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_call">'<tt>call</tt>' Instruction</a></li>
104 <li><a href="#i_vaarg">'<tt>vaarg</tt>' Instruction</a></li>
109 <li><a href="#intrinsics">Intrinsic Functions</a>
111 <li><a href="#int_varargs">Variable Argument Handling Intrinsics</a>
113 <li><a href="#i_va_start">'<tt>llvm.va_start</tt>' Intrinsic</a></li>
114 <li><a href="#i_va_end">'<tt>llvm.va_end</tt>' Intrinsic</a></li>
115 <li><a href="#i_va_copy">'<tt>llvm.va_copy</tt>' Intrinsic</a></li>
118 <li><a href="#int_gc">Accurate Garbage Collection Intrinsics</a>
120 <li><a href="#i_gcroot">'<tt>llvm.gcroot</tt>' Intrinsic</a></li>
121 <li><a href="#i_gcread">'<tt>llvm.gcread</tt>' Intrinsic</a></li>
122 <li><a href="#i_gcwrite">'<tt>llvm.gcwrite</tt>' Intrinsic</a></li>
125 <li><a href="#int_codegen">Code Generator Intrinsics</a>
127 <li><a href="#i_returnaddress">'<tt>llvm.returnaddress</tt>' Intrinsic</a></li>
128 <li><a href="#i_frameaddress">'<tt>llvm.frameaddress</tt>' Intrinsic</a></li>
129 <li><a href="#i_prefetch">'<tt>llvm.prefetch</tt>' Intrinsic</a></li>
130 <li><a href="#i_pcmarker">'<tt>llvm.pcmarker</tt>' Intrinsic</a></li>
131 <li><a href="#i_readcyclecounter"><tt>llvm.readcyclecounter</tt>' Intrinsic</a></li>
134 <li><a href="#int_os">Operating System Intrinsics</a>
136 <li><a href="#i_readport">'<tt>llvm.readport</tt>' Intrinsic</a></li>
137 <li><a href="#i_writeport">'<tt>llvm.writeport</tt>' Intrinsic</a></li>
138 <li><a href="#i_readio">'<tt>llvm.readio</tt>' Intrinsic</a></li>
139 <li><a href="#i_writeio">'<tt>llvm.writeio</tt>' Intrinsic</a></li>
141 <li><a href="#int_libc">Standard C Library Intrinsics</a>
143 <li><a href="#i_memcpy">'<tt>llvm.memcpy</tt>' Intrinsic</a></li>
144 <li><a href="#i_memmove">'<tt>llvm.memmove</tt>' Intrinsic</a></li>
145 <li><a href="#i_memset">'<tt>llvm.memset</tt>' Intrinsic</a></li>
146 <li><a href="#i_isunordered">'<tt>llvm.isunordered</tt>' Intrinsic</a></li>
147 <li><a href="#i_sqrt">'<tt>llvm.sqrt</tt>' Intrinsic</a></li>
151 <li><a href="#int_count">Bit counting Intrinsics</a>
153 <li><a href="#int_ctpop">'<tt>llvm.ctpop</tt>' Intrinsic </a></li>
154 <li><a href="#int_ctlz">'<tt>llvm.ctlz</tt>' Intrinsic </a></li>
155 <li><a href="#int_cttz">'<tt>llvm.cttz</tt>' Intrinsic </a></li>
158 <li><a href="#int_debugger">Debugger intrinsics</a></li>
163 <div class="doc_author">
164 <p>Written by <a href="mailto:sabre@nondot.org">Chris Lattner</a>
165 and <a href="mailto:vadve@cs.uiuc.edu">Vikram Adve</a></p>
168 <!-- *********************************************************************** -->
169 <div class="doc_section"> <a name="abstract">Abstract </a></div>
170 <!-- *********************************************************************** -->
172 <div class="doc_text">
173 <p>This document is a reference manual for the LLVM assembly language.
174 LLVM is an SSA based representation that provides type safety,
175 low-level operations, flexibility, and the capability of representing
176 'all' high-level languages cleanly. It is the common code
177 representation used throughout all phases of the LLVM compilation
181 <!-- *********************************************************************** -->
182 <div class="doc_section"> <a name="introduction">Introduction</a> </div>
183 <!-- *********************************************************************** -->
185 <div class="doc_text">
187 <p>The LLVM code representation is designed to be used in three
188 different forms: as an in-memory compiler IR, as an on-disk bytecode
189 representation (suitable for fast loading by a Just-In-Time compiler),
190 and as a human readable assembly language representation. This allows
191 LLVM to provide a powerful intermediate representation for efficient
192 compiler transformations and analysis, while providing a natural means
193 to debug and visualize the transformations. The three different forms
194 of LLVM are all equivalent. This document describes the human readable
195 representation and notation.</p>
197 <p>The LLVM representation aims to be light-weight and low-level
198 while being expressive, typed, and extensible at the same time. It
199 aims to be a "universal IR" of sorts, by being at a low enough level
200 that high-level ideas may be cleanly mapped to it (similar to how
201 microprocessors are "universal IR's", allowing many source languages to
202 be mapped to them). By providing type information, LLVM can be used as
203 the target of optimizations: for example, through pointer analysis, it
204 can be proven that a C automatic variable is never accessed outside of
205 the current function... allowing it to be promoted to a simple SSA
206 value instead of a memory location.</p>
210 <!-- _______________________________________________________________________ -->
211 <div class="doc_subsubsection"> <a name="wellformed">Well-Formedness</a> </div>
213 <div class="doc_text">
215 <p>It is important to note that this document describes 'well formed'
216 LLVM assembly language. There is a difference between what the parser
217 accepts and what is considered 'well formed'. For example, the
218 following instruction is syntactically okay, but not well formed:</p>
221 %x = <a href="#i_add">add</a> int 1, %x
224 <p>...because the definition of <tt>%x</tt> does not dominate all of
225 its uses. The LLVM infrastructure provides a verification pass that may
226 be used to verify that an LLVM module is well formed. This pass is
227 automatically run by the parser after parsing input assembly and by
228 the optimizer before it outputs bytecode. The violations pointed out
229 by the verifier pass indicate bugs in transformation passes or input to
232 <!-- Describe the typesetting conventions here. --> </div>
234 <!-- *********************************************************************** -->
235 <div class="doc_section"> <a name="identifiers">Identifiers</a> </div>
236 <!-- *********************************************************************** -->
238 <div class="doc_text">
240 <p>LLVM uses three different forms of identifiers, for different
244 <li>Named values are represented as a string of characters with a '%' prefix.
245 For example, %foo, %DivisionByZero, %a.really.long.identifier. The actual
246 regular expression used is '<tt>%[a-zA-Z$._][a-zA-Z$._0-9]*</tt>'.
247 Identifiers which require other characters in their names can be surrounded
248 with quotes. In this way, anything except a <tt>"</tt> character can be used
251 <li>Unnamed values are represented as an unsigned numeric value with a '%'
252 prefix. For example, %12, %2, %44.</li>
254 <li>Constants, which are described in a <a href="#constants">section about
255 constants</a>, below.</li>
258 <p>LLVM requires that values start with a '%' sign for two reasons: Compilers
259 don't need to worry about name clashes with reserved words, and the set of
260 reserved words may be expanded in the future without penalty. Additionally,
261 unnamed identifiers allow a compiler to quickly come up with a temporary
262 variable without having to avoid symbol table conflicts.</p>
264 <p>Reserved words in LLVM are very similar to reserved words in other
265 languages. There are keywords for different opcodes ('<tt><a
266 href="#i_add">add</a></tt>', '<tt><a href="#i_cast">cast</a></tt>', '<tt><a
267 href="#i_ret">ret</a></tt>', etc...), for primitive type names ('<tt><a
268 href="#t_void">void</a></tt>', '<tt><a href="#t_uint">uint</a></tt>', etc...),
269 and others. These reserved words cannot conflict with variable names, because
270 none of them start with a '%' character.</p>
272 <p>Here is an example of LLVM code to multiply the integer variable
273 '<tt>%X</tt>' by 8:</p>
278 %result = <a href="#i_mul">mul</a> uint %X, 8
281 <p>After strength reduction:</p>
284 %result = <a href="#i_shl">shl</a> uint %X, ubyte 3
287 <p>And the hard way:</p>
290 <a href="#i_add">add</a> uint %X, %X <i>; yields {uint}:%0</i>
291 <a href="#i_add">add</a> uint %0, %0 <i>; yields {uint}:%1</i>
292 %result = <a href="#i_add">add</a> uint %1, %1
295 <p>This last way of multiplying <tt>%X</tt> by 8 illustrates several
296 important lexical features of LLVM:</p>
300 <li>Comments are delimited with a '<tt>;</tt>' and go until the end of
303 <li>Unnamed temporaries are created when the result of a computation is not
304 assigned to a named value.</li>
306 <li>Unnamed temporaries are numbered sequentially</li>
310 <p>...and it also shows a convention that we follow in this document. When
311 demonstrating instructions, we will follow an instruction with a comment that
312 defines the type and name of value produced. Comments are shown in italic
317 <!-- *********************************************************************** -->
318 <div class="doc_section"> <a name="highlevel">High Level Structure</a> </div>
319 <!-- *********************************************************************** -->
321 <!-- ======================================================================= -->
322 <div class="doc_subsection"> <a name="modulestructure">Module Structure</a>
325 <div class="doc_text">
327 <p>LLVM programs are composed of "Module"s, each of which is a
328 translation unit of the input programs. Each module consists of
329 functions, global variables, and symbol table entries. Modules may be
330 combined together with the LLVM linker, which merges function (and
331 global variable) definitions, resolves forward declarations, and merges
332 symbol table entries. Here is an example of the "hello world" module:</p>
334 <pre><i>; Declare the string constant as a global constant...</i>
335 <a href="#identifiers">%.LC0</a> = <a href="#linkage_internal">internal</a> <a
336 href="#globalvars">constant</a> <a href="#t_array">[13 x sbyte]</a> c"hello world\0A\00" <i>; [13 x sbyte]*</i>
338 <i>; External declaration of the puts function</i>
339 <a href="#functionstructure">declare</a> int %puts(sbyte*) <i>; int(sbyte*)* </i>
341 <i>; Definition of main function</i>
342 int %main() { <i>; int()* </i>
343 <i>; Convert [13x sbyte]* to sbyte *...</i>
345 href="#i_getelementptr">getelementptr</a> [13 x sbyte]* %.LC0, long 0, long 0 <i>; sbyte*</i>
347 <i>; Call puts function to write out the string to stdout...</i>
349 href="#i_call">call</a> int %puts(sbyte* %cast210) <i>; int</i>
351 href="#i_ret">ret</a> int 0<br>}<br></pre>
353 <p>This example is made up of a <a href="#globalvars">global variable</a>
354 named "<tt>.LC0</tt>", an external declaration of the "<tt>puts</tt>"
355 function, and a <a href="#functionstructure">function definition</a>
356 for "<tt>main</tt>".</p>
358 <p>In general, a module is made up of a list of global values,
359 where both functions and global variables are global values. Global values are
360 represented by a pointer to a memory location (in this case, a pointer to an
361 array of char, and a pointer to a function), and have one of the following <a
362 href="#linkage">linkage types</a>.</p>
366 <!-- ======================================================================= -->
367 <div class="doc_subsection">
368 <a name="linkage">Linkage Types</a>
371 <div class="doc_text">
374 All Global Variables and Functions have one of the following types of linkage:
379 <dt><tt><b><a name="linkage_internal">internal</a></b></tt> </dt>
381 <dd>Global values with internal linkage are only directly accessible by
382 objects in the current module. In particular, linking code into a module with
383 an internal global value may cause the internal to be renamed as necessary to
384 avoid collisions. Because the symbol is internal to the module, all
385 references can be updated. This corresponds to the notion of the
386 '<tt>static</tt>' keyword in C, or the idea of "anonymous namespaces" in C++.
389 <dt><tt><b><a name="linkage_linkonce">linkonce</a></b></tt>: </dt>
391 <dd>"<tt>linkonce</tt>" linkage is similar to <tt>internal</tt> linkage, with
392 the twist that linking together two modules defining the same
393 <tt>linkonce</tt> globals will cause one of the globals to be discarded. This
394 is typically used to implement inline functions. Unreferenced
395 <tt>linkonce</tt> globals are allowed to be discarded.
398 <dt><tt><b><a name="linkage_weak">weak</a></b></tt>: </dt>
400 <dd>"<tt>weak</tt>" linkage is exactly the same as <tt>linkonce</tt> linkage,
401 except that unreferenced <tt>weak</tt> globals may not be discarded. This is
402 used to implement constructs in C such as "<tt>int X;</tt>" at global scope.
405 <dt><tt><b><a name="linkage_appending">appending</a></b></tt>: </dt>
407 <dd>"<tt>appending</tt>" linkage may only be applied to global variables of
408 pointer to array type. When two global variables with appending linkage are
409 linked together, the two global arrays are appended together. This is the
410 LLVM, typesafe, equivalent of having the system linker append together
411 "sections" with identical names when .o files are linked.
414 <dt><tt><b><a name="linkage_external">externally visible</a></b></tt>:</dt>
416 <dd>If none of the above identifiers are used, the global is externally
417 visible, meaning that it participates in linkage and can be used to resolve
418 external symbol references.
422 <p><a name="linkage_external">For example, since the "<tt>.LC0</tt>"
423 variable is defined to be internal, if another module defined a "<tt>.LC0</tt>"
424 variable and was linked with this one, one of the two would be renamed,
425 preventing a collision. Since "<tt>main</tt>" and "<tt>puts</tt>" are
426 external (i.e., lacking any linkage declarations), they are accessible
427 outside of the current module. It is illegal for a function <i>declaration</i>
428 to have any linkage type other than "externally visible".</a></p>
432 <!-- ======================================================================= -->
433 <div class="doc_subsection">
434 <a name="callingconv">Calling Conventions</a>
437 <div class="doc_text">
439 <p>LLVM <a href="#functionstructure">functions</a>, <a href="#i_call">calls</a>
440 and <a href="#i_invoke">invokes</a> can all have an optional calling convention
441 specified for the call. The calling convention of any pair of dynamic
442 caller/callee must match, or the behavior of the program is undefined. The
443 following calling conventions are supported by LLVM, and more may be added in
447 <dt><b>"<tt>ccc</tt>" - The C calling convention</b>:</dt>
449 <dd>This calling convention (the default if no other calling convention is
450 specified) matches the target C calling conventions. This calling convention
451 supports varargs function calls and tolerates some mismatch in the declared
452 prototype and implemented declaration of the function (as does normal C).
455 <dt><b>"<tt>fastcc</tt>" - The fast calling convention</b>:</dt>
457 <dd>This calling convention attempts to make calls as fast as possible
458 (e.g. by passing things in registers). This calling convention allows the
459 target to use whatever tricks it wants to produce fast code for the target,
460 without having to conform to an externally specified ABI. Implementations of
461 this convention should allow arbitrary tail call optimization to be supported.
462 This calling convention does not support varargs and requires the prototype of
463 all callees to exactly match the prototype of the function definition.
466 <dt><b>"<tt>coldcc</tt>" - The cold calling convention</b>:</dt>
468 <dd>This calling convention attempts to make code in the caller as efficient
469 as possible under the assumption that the call is not commonly executed. As
470 such, these calls often preserve all registers so that the call does not break
471 any live ranges in the caller side. This calling convention does not support
472 varargs and requires the prototype of all callees to exactly match the
473 prototype of the function definition.
476 <dt><b>"<tt>cc <<em>n</em>></tt>" - Numbered convention</b>:</dt>
478 <dd>Any calling convention may be specified by number, allowing
479 target-specific calling conventions to be used. Target specific calling
480 conventions start at 64.
484 <p>More calling conventions can be added/defined on an as-needed basis, to
485 support pascal conventions or any other well-known target-independent
490 <!-- ======================================================================= -->
491 <div class="doc_subsection">
492 <a name="globalvars">Global Variables</a>
495 <div class="doc_text">
497 <p>Global variables define regions of memory allocated at compilation time
498 instead of run-time. Global variables may optionally be initialized, may have
499 an explicit section to be placed in, and may
500 have an optional explicit alignment specified. A
501 variable may be defined as a global "constant," which indicates that the
502 contents of the variable will <b>never</b> be modified (enabling better
503 optimization, allowing the global data to be placed in the read-only section of
504 an executable, etc). Note that variables that need runtime initialization
505 cannot be marked "constant" as there is a store to the variable.</p>
508 LLVM explicitly allows <em>declarations</em> of global variables to be marked
509 constant, even if the final definition of the global is not. This capability
510 can be used to enable slightly better optimization of the program, but requires
511 the language definition to guarantee that optimizations based on the
512 'constantness' are valid for the translation units that do not include the
516 <p>As SSA values, global variables define pointer values that are in
517 scope (i.e. they dominate) all basic blocks in the program. Global
518 variables always define a pointer to their "content" type because they
519 describe a region of memory, and all memory objects in LLVM are
520 accessed through pointers.</p>
522 <p>LLVM allows an explicit section to be specified for globals. If the target
523 supports it, it will emit globals to the section specified.</p>
525 <p>An explicit alignment may be specified for a global. If not present, or if
526 the alignment is set to zero, the alignment of the global is set by the target
527 to whatever it feels convenient. If an explicit alignment is specified, the
528 global is forced to have at least that much alignment. All alignments must be
534 <!-- ======================================================================= -->
535 <div class="doc_subsection">
536 <a name="functionstructure">Functions</a>
539 <div class="doc_text">
541 <p>LLVM function definitions consist of an optional <a href="#linkage">linkage
542 type</a>, an optional <a href="#callingconv">calling convention</a>, a return
543 type, a function name, a (possibly empty) argument list, an optional section,
544 an optional alignment, an opening curly brace,
545 a list of basic blocks, and a closing curly brace. LLVM function declarations
546 are defined with the "<tt>declare</tt>" keyword, an optional <a
547 href="#callingconv">calling convention</a>, a return type, a function name,
548 a possibly empty list of arguments, and an optional alignment.</p>
550 <p>A function definition contains a list of basic blocks, forming the CFG for
551 the function. Each basic block may optionally start with a label (giving the
552 basic block a symbol table entry), contains a list of instructions, and ends
553 with a <a href="#terminators">terminator</a> instruction (such as a branch or
554 function return).</p>
556 <p>The first basic block in a program is special in two ways: it is immediately
557 executed on entrance to the function, and it is not allowed to have predecessor
558 basic blocks (i.e. there can not be any branches to the entry block of a
559 function). Because the block can have no predecessors, it also cannot have any
560 <a href="#i_phi">PHI nodes</a>.</p>
562 <p>LLVM functions are identified by their name and type signature. Hence, two
563 functions with the same name but different parameter lists or return values are
564 considered different functions, and LLVM will resolve references to each
567 <p>LLVM allows an explicit section to be specified for functions. If the target
568 supports it, it will emit functions to the section specified.</p>
570 <p>An explicit alignment may be specified for a function. If not present, or if
571 the alignment is set to zero, the alignment of the function is set by the target
572 to whatever it feels convenient. If an explicit alignment is specified, the
573 function is forced to have at least that much alignment. All alignments must be
580 <!-- *********************************************************************** -->
581 <div class="doc_section"> <a name="typesystem">Type System</a> </div>
582 <!-- *********************************************************************** -->
584 <div class="doc_text">
586 <p>The LLVM type system is one of the most important features of the
587 intermediate representation. Being typed enables a number of
588 optimizations to be performed on the IR directly, without having to do
589 extra analyses on the side before the transformation. A strong type
590 system makes it easier to read the generated code and enables novel
591 analyses and transformations that are not feasible to perform on normal
592 three address code representations.</p>
596 <!-- ======================================================================= -->
597 <div class="doc_subsection"> <a name="t_primitive">Primitive Types</a> </div>
598 <div class="doc_text">
599 <p>The primitive types are the fundamental building blocks of the LLVM
600 system. The current set of primitive types is as follows:</p>
602 <table class="layout">
607 <tr><th>Type</th><th>Description</th></tr>
608 <tr><td><tt>void</tt></td><td>No value</td></tr>
609 <tr><td><tt>ubyte</tt></td><td>Unsigned 8-bit value</td></tr>
610 <tr><td><tt>ushort</tt></td><td>Unsigned 16-bit value</td></tr>
611 <tr><td><tt>uint</tt></td><td>Unsigned 32-bit value</td></tr>
612 <tr><td><tt>ulong</tt></td><td>Unsigned 64-bit value</td></tr>
613 <tr><td><tt>float</tt></td><td>32-bit floating point value</td></tr>
614 <tr><td><tt>label</tt></td><td>Branch destination</td></tr>
621 <tr><th>Type</th><th>Description</th></tr>
622 <tr><td><tt>bool</tt></td><td>True or False value</td></tr>
623 <tr><td><tt>sbyte</tt></td><td>Signed 8-bit value</td></tr>
624 <tr><td><tt>short</tt></td><td>Signed 16-bit value</td></tr>
625 <tr><td><tt>int</tt></td><td>Signed 32-bit value</td></tr>
626 <tr><td><tt>long</tt></td><td>Signed 64-bit value</td></tr>
627 <tr><td><tt>double</tt></td><td>64-bit floating point value</td></tr>
635 <!-- _______________________________________________________________________ -->
636 <div class="doc_subsubsection"> <a name="t_classifications">Type
637 Classifications</a> </div>
638 <div class="doc_text">
639 <p>These different primitive types fall into a few useful
642 <table border="1" cellspacing="0" cellpadding="4">
644 <tr><th>Classification</th><th>Types</th></tr>
646 <td><a name="t_signed">signed</a></td>
647 <td><tt>sbyte, short, int, long, float, double</tt></td>
650 <td><a name="t_unsigned">unsigned</a></td>
651 <td><tt>ubyte, ushort, uint, ulong</tt></td>
654 <td><a name="t_integer">integer</a></td>
655 <td><tt>ubyte, sbyte, ushort, short, uint, int, ulong, long</tt></td>
658 <td><a name="t_integral">integral</a></td>
659 <td><tt>bool, ubyte, sbyte, ushort, short, uint, int, ulong, long</tt>
663 <td><a name="t_floating">floating point</a></td>
664 <td><tt>float, double</tt></td>
667 <td><a name="t_firstclass">first class</a></td>
668 <td><tt>bool, ubyte, sbyte, ushort, short, uint, int, ulong, long,<br>
669 float, double, <a href="#t_pointer">pointer</a>,
670 <a href="#t_packed">packed</a></tt></td>
675 <p>The <a href="#t_firstclass">first class</a> types are perhaps the
676 most important. Values of these types are the only ones which can be
677 produced by instructions, passed as arguments, or used as operands to
678 instructions. This means that all structures and arrays must be
679 manipulated either by pointer or by component.</p>
682 <!-- ======================================================================= -->
683 <div class="doc_subsection"> <a name="t_derived">Derived Types</a> </div>
685 <div class="doc_text">
687 <p>The real power in LLVM comes from the derived types in the system.
688 This is what allows a programmer to represent arrays, functions,
689 pointers, and other useful types. Note that these derived types may be
690 recursive: For example, it is possible to have a two dimensional array.</p>
694 <!-- _______________________________________________________________________ -->
695 <div class="doc_subsubsection"> <a name="t_array">Array Type</a> </div>
697 <div class="doc_text">
701 <p>The array type is a very simple derived type that arranges elements
702 sequentially in memory. The array type requires a size (number of
703 elements) and an underlying data type.</p>
708 [<# elements> x <elementtype>]
711 <p>The number of elements is a constant integer value; elementtype may
712 be any type with a size.</p>
715 <table class="layout">
718 <tt>[40 x int ]</tt><br/>
719 <tt>[41 x int ]</tt><br/>
720 <tt>[40 x uint]</tt><br/>
723 Array of 40 integer values.<br/>
724 Array of 41 integer values.<br/>
725 Array of 40 unsigned integer values.<br/>
729 <p>Here are some examples of multidimensional arrays:</p>
730 <table class="layout">
733 <tt>[3 x [4 x int]]</tt><br/>
734 <tt>[12 x [10 x float]]</tt><br/>
735 <tt>[2 x [3 x [4 x uint]]]</tt><br/>
738 3x4 array of integer values.<br/>
739 12x10 array of single precision floating point values.<br/>
740 2x3x4 array of unsigned integer values.<br/>
745 <p>Note that 'variable sized arrays' can be implemented in LLVM with a zero
746 length array. Normally, accesses past the end of an array are undefined in
747 LLVM (e.g. it is illegal to access the 5th element of a 3 element array).
748 As a special case, however, zero length arrays are recognized to be variable
749 length. This allows implementation of 'pascal style arrays' with the LLVM
750 type "{ int, [0 x float]}", for example.</p>
754 <!-- _______________________________________________________________________ -->
755 <div class="doc_subsubsection"> <a name="t_function">Function Type</a> </div>
756 <div class="doc_text">
758 <p>The function type can be thought of as a function signature. It
759 consists of a return type and a list of formal parameter types.
760 Function types are usually used to build virtual function tables
761 (which are structures of pointers to functions), for indirect function
762 calls, and when defining a function.</p>
764 The return type of a function type cannot be an aggregate type.
767 <pre> <returntype> (<parameter list>)<br></pre>
768 <p>...where '<tt><parameter list></tt>' is a comma-separated list of type
769 specifiers. Optionally, the parameter list may include a type <tt>...</tt>,
770 which indicates that the function takes a variable number of arguments.
771 Variable argument functions can access their arguments with the <a
772 href="#int_varargs">variable argument handling intrinsic</a> functions.</p>
774 <table class="layout">
777 <tt>int (int)</tt> <br/>
778 <tt>float (int, int *) *</tt><br/>
779 <tt>int (sbyte *, ...)</tt><br/>
782 function taking an <tt>int</tt>, returning an <tt>int</tt><br/>
783 <a href="#t_pointer">Pointer</a> to a function that takes an
784 <tt>int</tt> and a <a href="#t_pointer">pointer</a> to <tt>int</tt>,
785 returning <tt>float</tt>.<br/>
786 A vararg function that takes at least one <a href="#t_pointer">pointer</a>
787 to <tt>sbyte</tt> (signed char in C), which returns an integer. This is
788 the signature for <tt>printf</tt> in LLVM.<br/>
794 <!-- _______________________________________________________________________ -->
795 <div class="doc_subsubsection"> <a name="t_struct">Structure Type</a> </div>
796 <div class="doc_text">
798 <p>The structure type is used to represent a collection of data members
799 together in memory. The packing of the field types is defined to match
800 the ABI of the underlying processor. The elements of a structure may
801 be any type that has a size.</p>
802 <p>Structures are accessed using '<tt><a href="#i_load">load</a></tt>
803 and '<tt><a href="#i_store">store</a></tt>' by getting a pointer to a
804 field with the '<tt><a href="#i_getelementptr">getelementptr</a></tt>'
807 <pre> { <type list> }<br></pre>
809 <table class="layout">
812 <tt>{ int, int, int }</tt><br/>
813 <tt>{ float, int (int) * }</tt><br/>
816 a triple of three <tt>int</tt> values<br/>
817 A pair, where the first element is a <tt>float</tt> and the second element
818 is a <a href="#t_pointer">pointer</a> to a <a href="#t_function">function</a>
819 that takes an <tt>int</tt>, returning an <tt>int</tt>.<br/>
825 <!-- _______________________________________________________________________ -->
826 <div class="doc_subsubsection"> <a name="t_pointer">Pointer Type</a> </div>
827 <div class="doc_text">
829 <p>As in many languages, the pointer type represents a pointer or
830 reference to another object, which must live in memory.</p>
832 <pre> <type> *<br></pre>
834 <table class="layout">
837 <tt>[4x int]*</tt><br/>
838 <tt>int (int *) *</tt><br/>
841 A <a href="#t_pointer">pointer</a> to <a href="#t_array">array</a> of
842 four <tt>int</tt> values<br/>
843 A <a href="#t_pointer">pointer</a> to a <a
844 href="#t_function">function</a> that takes an <tt>int*</tt>, returning an
851 <!-- _______________________________________________________________________ -->
852 <div class="doc_subsubsection"> <a name="t_packed">Packed Type</a> </div>
853 <div class="doc_text">
857 <p>A packed type is a simple derived type that represents a vector
858 of elements. Packed types are used when multiple primitive data
859 are operated in parallel using a single instruction (SIMD).
860 A packed type requires a size (number of
861 elements) and an underlying primitive data type. Vectors must have a power
862 of two length (1, 2, 4, 8, 16 ...). Packed types are
863 considered <a href="#t_firstclass">first class</a>.</p>
868 < <# elements> x <elementtype> >
871 <p>The number of elements is a constant integer value; elementtype may
872 be any integral or floating point type.</p>
876 <table class="layout">
879 <tt><4 x int></tt><br/>
880 <tt><8 x float></tt><br/>
881 <tt><2 x uint></tt><br/>
884 Packed vector of 4 integer values.<br/>
885 Packed vector of 8 floating-point values.<br/>
886 Packed vector of 2 unsigned integer values.<br/>
892 <!-- _______________________________________________________________________ -->
893 <div class="doc_subsubsection"> <a name="t_opaque">Opaque Type</a> </div>
894 <div class="doc_text">
898 <p>Opaque types are used to represent unknown types in the system. This
899 corresponds (for example) to the C notion of a foward declared structure type.
900 In LLVM, opaque types can eventually be resolved to any type (not just a
911 <table class="layout">
924 <!-- *********************************************************************** -->
925 <div class="doc_section"> <a name="constants">Constants</a> </div>
926 <!-- *********************************************************************** -->
928 <div class="doc_text">
930 <p>LLVM has several different basic types of constants. This section describes
931 them all and their syntax.</p>
935 <!-- ======================================================================= -->
936 <div class="doc_subsection"><a name="simpleconstants">Simple Constants</a></div>
938 <div class="doc_text">
941 <dt><b>Boolean constants</b></dt>
943 <dd>The two strings '<tt>true</tt>' and '<tt>false</tt>' are both valid
944 constants of the <tt><a href="#t_primitive">bool</a></tt> type.
947 <dt><b>Integer constants</b></dt>
949 <dd>Standard integers (such as '4') are constants of the <a
950 href="#t_integer">integer</a> type. Negative numbers may be used with signed
954 <dt><b>Floating point constants</b></dt>
956 <dd>Floating point constants use standard decimal notation (e.g. 123.421),
957 exponential notation (e.g. 1.23421e+2), or a more precise hexadecimal
958 notation (see below). Floating point constants must have a <a
959 href="#t_floating">floating point</a> type. </dd>
961 <dt><b>Null pointer constants</b></dt>
963 <dd>The identifier '<tt>null</tt>' is recognized as a null pointer constant
964 and must be of <a href="#t_pointer">pointer type</a>.</dd>
968 <p>The one non-intuitive notation for constants is the optional hexadecimal form
969 of floating point constants. For example, the form '<tt>double
970 0x432ff973cafa8000</tt>' is equivalent to (but harder to read than) '<tt>double
971 4.5e+15</tt>'. The only time hexadecimal floating point constants are required
972 (and the only time that they are generated by the disassembler) is when a
973 floating point constant must be emitted but it cannot be represented as a
974 decimal floating point number. For example, NaN's, infinities, and other
975 special values are represented in their IEEE hexadecimal format so that
976 assembly and disassembly do not cause any bits to change in the constants.</p>
980 <!-- ======================================================================= -->
981 <div class="doc_subsection"><a name="aggregateconstants">Aggregate Constants</a>
984 <div class="doc_text">
985 <p>Aggregate constants arise from aggregation of simple constants
986 and smaller aggregate constants.</p>
989 <dt><b>Structure constants</b></dt>
991 <dd>Structure constants are represented with notation similar to structure
992 type definitions (a comma separated list of elements, surrounded by braces
993 (<tt>{}</tt>)). For example: "<tt>{ int 4, float 17.0, int* %G }</tt>",
994 where "<tt>%G</tt>" is declared as "<tt>%G = external global int</tt>". Structure constants
995 must have <a href="#t_struct">structure type</a>, and the number and
996 types of elements must match those specified by the type.
999 <dt><b>Array constants</b></dt>
1001 <dd>Array constants are represented with notation similar to array type
1002 definitions (a comma separated list of elements, surrounded by square brackets
1003 (<tt>[]</tt>)). For example: "<tt>[ int 42, int 11, int 74 ]</tt>". Array
1004 constants must have <a href="#t_array">array type</a>, and the number and
1005 types of elements must match those specified by the type.
1008 <dt><b>Packed constants</b></dt>
1010 <dd>Packed constants are represented with notation similar to packed type
1011 definitions (a comma separated list of elements, surrounded by
1012 less-than/greater-than's (<tt><></tt>)). For example: "<tt>< int 42,
1013 int 11, int 74, int 100 ></tt>". Packed constants must have <a
1014 href="#t_packed">packed type</a>, and the number and types of elements must
1015 match those specified by the type.
1018 <dt><b>Zero initialization</b></dt>
1020 <dd>The string '<tt>zeroinitializer</tt>' can be used to zero initialize a
1021 value to zero of <em>any</em> type, including scalar and aggregate types.
1022 This is often used to avoid having to print large zero initializers (e.g. for
1023 large arrays) and is always exactly equivalent to using explicit zero
1030 <!-- ======================================================================= -->
1031 <div class="doc_subsection">
1032 <a name="globalconstants">Global Variable and Function Addresses</a>
1035 <div class="doc_text">
1037 <p>The addresses of <a href="#globalvars">global variables</a> and <a
1038 href="#functionstructure">functions</a> are always implicitly valid (link-time)
1039 constants. These constants are explicitly referenced when the <a
1040 href="#identifiers">identifier for the global</a> is used and always have <a
1041 href="#t_pointer">pointer</a> type. For example, the following is a legal LLVM
1047 %Z = global [2 x int*] [ int* %X, int* %Y ]
1052 <!-- ======================================================================= -->
1053 <div class="doc_subsection"><a name="undefvalues">Undefined Values</a></div>
1054 <div class="doc_text">
1055 <p>The string '<tt>undef</tt>' is recognized as a type-less constant that has
1056 no specific value. Undefined values may be of any type and be used anywhere
1057 a constant is permitted.</p>
1059 <p>Undefined values indicate to the compiler that the program is well defined
1060 no matter what value is used, giving the compiler more freedom to optimize.
1064 <!-- ======================================================================= -->
1065 <div class="doc_subsection"><a name="constantexprs">Constant Expressions</a>
1068 <div class="doc_text">
1070 <p>Constant expressions are used to allow expressions involving other constants
1071 to be used as constants. Constant expressions may be of any <a
1072 href="#t_firstclass">first class</a> type and may involve any LLVM operation
1073 that does not have side effects (e.g. load and call are not supported). The
1074 following is the syntax for constant expressions:</p>
1077 <dt><b><tt>cast ( CST to TYPE )</tt></b></dt>
1079 <dd>Cast a constant to another type.</dd>
1081 <dt><b><tt>getelementptr ( CSTPTR, IDX0, IDX1, ... )</tt></b></dt>
1083 <dd>Perform the <a href="#i_getelementptr">getelementptr operation</a> on
1084 constants. As with the <a href="#i_getelementptr">getelementptr</a>
1085 instruction, the index list may have zero or more indexes, which are required
1086 to make sense for the type of "CSTPTR".</dd>
1088 <dt><b><tt>OPCODE ( LHS, RHS )</tt></b></dt>
1090 <dd>Perform the specified operation of the LHS and RHS constants. OPCODE may
1091 be any of the <a href="#binaryops">binary</a> or <a href="#bitwiseops">bitwise
1092 binary</a> operations. The constraints on operands are the same as those for
1093 the corresponding instruction (e.g. no bitwise operations on floating point
1094 values are allowed).</dd>
1098 <!-- *********************************************************************** -->
1099 <div class="doc_section"> <a name="instref">Instruction Reference</a> </div>
1100 <!-- *********************************************************************** -->
1102 <div class="doc_text">
1104 <p>The LLVM instruction set consists of several different
1105 classifications of instructions: <a href="#terminators">terminator
1106 instructions</a>, <a href="#binaryops">binary instructions</a>,
1107 <a href="#bitwiseops">bitwise binary instructions</a>, <a
1108 href="#memoryops">memory instructions</a>, and <a href="#otherops">other
1109 instructions</a>.</p>
1113 <!-- ======================================================================= -->
1114 <div class="doc_subsection"> <a name="terminators">Terminator
1115 Instructions</a> </div>
1117 <div class="doc_text">
1119 <p>As mentioned <a href="#functionstructure">previously</a>, every
1120 basic block in a program ends with a "Terminator" instruction, which
1121 indicates which block should be executed after the current block is
1122 finished. These terminator instructions typically yield a '<tt>void</tt>'
1123 value: they produce control flow, not values (the one exception being
1124 the '<a href="#i_invoke"><tt>invoke</tt></a>' instruction).</p>
1125 <p>There are six different terminator instructions: the '<a
1126 href="#i_ret"><tt>ret</tt></a>' instruction, the '<a href="#i_br"><tt>br</tt></a>'
1127 instruction, the '<a href="#i_switch"><tt>switch</tt></a>' instruction,
1128 the '<a href="#i_invoke"><tt>invoke</tt></a>' instruction, the '<a
1129 href="#i_unwind"><tt>unwind</tt></a>' instruction, and the '<a
1130 href="#i_unreachable"><tt>unreachable</tt></a>' instruction.</p>
1134 <!-- _______________________________________________________________________ -->
1135 <div class="doc_subsubsection"> <a name="i_ret">'<tt>ret</tt>'
1136 Instruction</a> </div>
1137 <div class="doc_text">
1139 <pre> ret <type> <value> <i>; Return a value from a non-void function</i>
1140 ret void <i>; Return from void function</i>
1143 <p>The '<tt>ret</tt>' instruction is used to return control flow (and a
1144 value) from a function back to the caller.</p>
1145 <p>There are two forms of the '<tt>ret</tt>' instruction: one that
1146 returns a value and then causes control flow, and one that just causes
1147 control flow to occur.</p>
1149 <p>The '<tt>ret</tt>' instruction may return any '<a
1150 href="#t_firstclass">first class</a>' type. Notice that a function is
1151 not <a href="#wellformed">well formed</a> if there exists a '<tt>ret</tt>'
1152 instruction inside of the function that returns a value that does not
1153 match the return type of the function.</p>
1155 <p>When the '<tt>ret</tt>' instruction is executed, control flow
1156 returns back to the calling function's context. If the caller is a "<a
1157 href="#i_call"><tt>call</tt></a>" instruction, execution continues at
1158 the instruction after the call. If the caller was an "<a
1159 href="#i_invoke"><tt>invoke</tt></a>" instruction, execution continues
1160 at the beginning of the "normal" destination block. If the instruction
1161 returns a value, that value shall set the call or invoke instruction's
1164 <pre> ret int 5 <i>; Return an integer value of 5</i>
1165 ret void <i>; Return from a void function</i>
1168 <!-- _______________________________________________________________________ -->
1169 <div class="doc_subsubsection"> <a name="i_br">'<tt>br</tt>' Instruction</a> </div>
1170 <div class="doc_text">
1172 <pre> br bool <cond>, label <iftrue>, label <iffalse><br> br label <dest> <i>; Unconditional branch</i>
1175 <p>The '<tt>br</tt>' instruction is used to cause control flow to
1176 transfer to a different basic block in the current function. There are
1177 two forms of this instruction, corresponding to a conditional branch
1178 and an unconditional branch.</p>
1180 <p>The conditional branch form of the '<tt>br</tt>' instruction takes a
1181 single '<tt>bool</tt>' value and two '<tt>label</tt>' values. The
1182 unconditional form of the '<tt>br</tt>' instruction takes a single '<tt>label</tt>'
1183 value as a target.</p>
1185 <p>Upon execution of a conditional '<tt>br</tt>' instruction, the '<tt>bool</tt>'
1186 argument is evaluated. If the value is <tt>true</tt>, control flows
1187 to the '<tt>iftrue</tt>' <tt>label</tt> argument. If "cond" is <tt>false</tt>,
1188 control flows to the '<tt>iffalse</tt>' <tt>label</tt> argument.</p>
1190 <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
1191 href="#i_ret">ret</a> int 1<br>IfUnequal:<br> <a href="#i_ret">ret</a> int 0<br></pre>
1193 <!-- _______________________________________________________________________ -->
1194 <div class="doc_subsubsection">
1195 <a name="i_switch">'<tt>switch</tt>' Instruction</a>
1198 <div class="doc_text">
1202 switch <intty> <value>, label <defaultdest> [ <intty> <val>, label <dest> ... ]
1207 <p>The '<tt>switch</tt>' instruction is used to transfer control flow to one of
1208 several different places. It is a generalization of the '<tt>br</tt>'
1209 instruction, allowing a branch to occur to one of many possible
1215 <p>The '<tt>switch</tt>' instruction uses three parameters: an integer
1216 comparison value '<tt>value</tt>', a default '<tt>label</tt>' destination, and
1217 an array of pairs of comparison value constants and '<tt>label</tt>'s. The
1218 table is not allowed to contain duplicate constant entries.</p>
1222 <p>The <tt>switch</tt> instruction specifies a table of values and
1223 destinations. When the '<tt>switch</tt>' instruction is executed, this
1224 table is searched for the given value. If the value is found, control flow is
1225 transfered to the corresponding destination; otherwise, control flow is
1226 transfered to the default destination.</p>
1228 <h5>Implementation:</h5>
1230 <p>Depending on properties of the target machine and the particular
1231 <tt>switch</tt> instruction, this instruction may be code generated in different
1232 ways. For example, it could be generated as a series of chained conditional
1233 branches or with a lookup table.</p>
1238 <i>; Emulate a conditional br instruction</i>
1239 %Val = <a href="#i_cast">cast</a> bool %value to int
1240 switch int %Val, label %truedest [int 0, label %falsedest ]
1242 <i>; Emulate an unconditional br instruction</i>
1243 switch uint 0, label %dest [ ]
1245 <i>; Implement a jump table:</i>
1246 switch uint %val, label %otherwise [ uint 0, label %onzero
1247 uint 1, label %onone
1248 uint 2, label %ontwo ]
1252 <!-- _______________________________________________________________________ -->
1253 <div class="doc_subsubsection">
1254 <a name="i_invoke">'<tt>invoke</tt>' Instruction</a>
1257 <div class="doc_text">
1262 <result> = invoke [<a href="#callingconv">cconv</a>] <ptr to function ty> %<function ptr val>(<function args>)
1263 to label <normal label> except label <exception label>
1268 <p>The '<tt>invoke</tt>' instruction causes control to transfer to a specified
1269 function, with the possibility of control flow transfer to either the
1270 '<tt>normal</tt>' label or the
1271 '<tt>exception</tt>' label. If the callee function returns with the
1272 "<tt><a href="#i_ret">ret</a></tt>" instruction, control flow will return to the
1273 "normal" label. If the callee (or any indirect callees) returns with the "<a
1274 href="#i_unwind"><tt>unwind</tt></a>" instruction, control is interrupted and
1275 continued at the dynamically nearest "exception" label.</p>
1279 <p>This instruction requires several arguments:</p>
1283 The optional "cconv" marker indicates which <a href="callingconv">calling
1284 convention</a> the call should use. If none is specified, the call defaults
1285 to using C calling conventions.
1287 <li>'<tt>ptr to function ty</tt>': shall be the signature of the pointer to
1288 function value being invoked. In most cases, this is a direct function
1289 invocation, but indirect <tt>invoke</tt>s are just as possible, branching off
1290 an arbitrary pointer to function value.
1293 <li>'<tt>function ptr val</tt>': An LLVM value containing a pointer to a
1294 function to be invoked. </li>
1296 <li>'<tt>function args</tt>': argument list whose types match the function
1297 signature argument types. If the function signature indicates the function
1298 accepts a variable number of arguments, the extra arguments can be
1301 <li>'<tt>normal label</tt>': the label reached when the called function
1302 executes a '<tt><a href="#i_ret">ret</a></tt>' instruction. </li>
1304 <li>'<tt>exception label</tt>': the label reached when a callee returns with
1305 the <a href="#i_unwind"><tt>unwind</tt></a> instruction. </li>
1311 <p>This instruction is designed to operate as a standard '<tt><a
1312 href="#i_call">call</a></tt>' instruction in most regards. The primary
1313 difference is that it establishes an association with a label, which is used by
1314 the runtime library to unwind the stack.</p>
1316 <p>This instruction is used in languages with destructors to ensure that proper
1317 cleanup is performed in the case of either a <tt>longjmp</tt> or a thrown
1318 exception. Additionally, this is important for implementation of
1319 '<tt>catch</tt>' clauses in high-level languages that support them.</p>
1323 %retval = invoke int %Test(int 15) to label %Continue
1324 except label %TestCleanup <i>; {int}:retval set</i>
1325 %retval = invoke <a href="#callingconv">coldcc</a> int %Test(int 15) to label %Continue
1326 except label %TestCleanup <i>; {int}:retval set</i>
1331 <!-- _______________________________________________________________________ -->
1333 <div class="doc_subsubsection"> <a name="i_unwind">'<tt>unwind</tt>'
1334 Instruction</a> </div>
1336 <div class="doc_text">
1345 <p>The '<tt>unwind</tt>' instruction unwinds the stack, continuing control flow
1346 at the first callee in the dynamic call stack which used an <a
1347 href="#i_invoke"><tt>invoke</tt></a> instruction to perform the call. This is
1348 primarily used to implement exception handling.</p>
1352 <p>The '<tt>unwind</tt>' intrinsic causes execution of the current function to
1353 immediately halt. The dynamic call stack is then searched for the first <a
1354 href="#i_invoke"><tt>invoke</tt></a> instruction on the call stack. Once found,
1355 execution continues at the "exceptional" destination block specified by the
1356 <tt>invoke</tt> instruction. If there is no <tt>invoke</tt> instruction in the
1357 dynamic call chain, undefined behavior results.</p>
1360 <!-- _______________________________________________________________________ -->
1362 <div class="doc_subsubsection"> <a name="i_unreachable">'<tt>unreachable</tt>'
1363 Instruction</a> </div>
1365 <div class="doc_text">
1374 <p>The '<tt>unreachable</tt>' instruction has no defined semantics. This
1375 instruction is used to inform the optimizer that a particular portion of the
1376 code is not reachable. This can be used to indicate that the code after a
1377 no-return function cannot be reached, and other facts.</p>
1381 <p>The '<tt>unreachable</tt>' instruction has no defined semantics.</p>
1386 <!-- ======================================================================= -->
1387 <div class="doc_subsection"> <a name="binaryops">Binary Operations</a> </div>
1388 <div class="doc_text">
1389 <p>Binary operators are used to do most of the computation in a
1390 program. They require two operands, execute an operation on them, and
1391 produce a single value. The operands might represent
1392 multiple data, as is the case with the <a href="#t_packed">packed</a> data type.
1393 The result value of a binary operator is not
1394 necessarily the same type as its operands.</p>
1395 <p>There are several different binary operators:</p>
1397 <!-- _______________________________________________________________________ -->
1398 <div class="doc_subsubsection"> <a name="i_add">'<tt>add</tt>'
1399 Instruction</a> </div>
1400 <div class="doc_text">
1402 <pre> <result> = add <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1405 <p>The '<tt>add</tt>' instruction returns the sum of its two operands.</p>
1407 <p>The two arguments to the '<tt>add</tt>' instruction must be either <a
1408 href="#t_integer">integer</a> or <a href="#t_floating">floating point</a> values.
1409 This instruction can also take <a href="#t_packed">packed</a> versions of the values.
1410 Both arguments must have identical types.</p>
1412 <p>The value produced is the integer or floating point sum of the two
1415 <pre> <result> = add int 4, %var <i>; yields {int}:result = 4 + %var</i>
1418 <!-- _______________________________________________________________________ -->
1419 <div class="doc_subsubsection"> <a name="i_sub">'<tt>sub</tt>'
1420 Instruction</a> </div>
1421 <div class="doc_text">
1423 <pre> <result> = sub <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1426 <p>The '<tt>sub</tt>' instruction returns the difference of its two
1428 <p>Note that the '<tt>sub</tt>' instruction is used to represent the '<tt>neg</tt>'
1429 instruction present in most other intermediate representations.</p>
1431 <p>The two arguments to the '<tt>sub</tt>' instruction must be either <a
1432 href="#t_integer">integer</a> or <a href="#t_floating">floating point</a>
1434 This instruction can also take <a href="#t_packed">packed</a> versions of the values.
1435 Both arguments must have identical types.</p>
1437 <p>The value produced is the integer or floating point difference of
1438 the two operands.</p>
1440 <pre> <result> = sub int 4, %var <i>; yields {int}:result = 4 - %var</i>
1441 <result> = sub int 0, %val <i>; yields {int}:result = -%var</i>
1444 <!-- _______________________________________________________________________ -->
1445 <div class="doc_subsubsection"> <a name="i_mul">'<tt>mul</tt>'
1446 Instruction</a> </div>
1447 <div class="doc_text">
1449 <pre> <result> = mul <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1452 <p>The '<tt>mul</tt>' instruction returns the product of its two
1455 <p>The two arguments to the '<tt>mul</tt>' instruction must be either <a
1456 href="#t_integer">integer</a> or <a href="#t_floating">floating point</a>
1458 This instruction can also take <a href="#t_packed">packed</a> versions of the values.
1459 Both arguments must have identical types.</p>
1461 <p>The value produced is the integer or floating point product of the
1463 <p>There is no signed vs unsigned multiplication. The appropriate
1464 action is taken based on the type of the operand.</p>
1466 <pre> <result> = mul int 4, %var <i>; yields {int}:result = 4 * %var</i>
1469 <!-- _______________________________________________________________________ -->
1470 <div class="doc_subsubsection"> <a name="i_div">'<tt>div</tt>'
1471 Instruction</a> </div>
1472 <div class="doc_text">
1474 <pre> <result> = div <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1477 <p>The '<tt>div</tt>' instruction returns the quotient of its two
1480 <p>The two arguments to the '<tt>div</tt>' instruction must be either <a
1481 href="#t_integer">integer</a> or <a href="#t_floating">floating point</a>
1483 This instruction can also take <a href="#t_packed">packed</a> versions of the values.
1484 Both arguments must have identical types.</p>
1486 <p>The value produced is the integer or floating point quotient of the
1489 <pre> <result> = div int 4, %var <i>; yields {int}:result = 4 / %var</i>
1492 <!-- _______________________________________________________________________ -->
1493 <div class="doc_subsubsection"> <a name="i_rem">'<tt>rem</tt>'
1494 Instruction</a> </div>
1495 <div class="doc_text">
1497 <pre> <result> = rem <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1500 <p>The '<tt>rem</tt>' instruction returns the remainder from the
1501 division of its two operands.</p>
1503 <p>The two arguments to the '<tt>rem</tt>' instruction must be either <a
1504 href="#t_integer">integer</a> or <a href="#t_floating">floating point</a>
1506 This instruction can also take <a href="#t_packed">packed</a> versions of the values.
1507 Both arguments must have identical types.</p>
1509 <p>This returns the <i>remainder</i> of a division (where the result
1510 has the same sign as the divisor), not the <i>modulus</i> (where the
1511 result has the same sign as the dividend) of a value. For more
1512 information about the difference, see <a
1513 href="http://mathforum.org/dr.math/problems/anne.4.28.99.html">The
1516 <pre> <result> = rem int 4, %var <i>; yields {int}:result = 4 % %var</i>
1519 <!-- _______________________________________________________________________ -->
1520 <div class="doc_subsubsection"> <a name="i_setcc">'<tt>set<i>cc</i></tt>'
1521 Instructions</a> </div>
1522 <div class="doc_text">
1524 <pre> <result> = seteq <ty> <var1>, <var2> <i>; yields {bool}:result</i>
1525 <result> = setne <ty> <var1>, <var2> <i>; yields {bool}:result</i>
1526 <result> = setlt <ty> <var1>, <var2> <i>; yields {bool}:result</i>
1527 <result> = setgt <ty> <var1>, <var2> <i>; yields {bool}:result</i>
1528 <result> = setle <ty> <var1>, <var2> <i>; yields {bool}:result</i>
1529 <result> = setge <ty> <var1>, <var2> <i>; yields {bool}:result</i>
1532 <p>The '<tt>set<i>cc</i></tt>' family of instructions returns a boolean
1533 value based on a comparison of their two operands.</p>
1535 <p>The two arguments to the '<tt>set<i>cc</i></tt>' instructions must
1536 be of <a href="#t_firstclass">first class</a> type (it is not possible
1537 to compare '<tt>label</tt>'s, '<tt>array</tt>'s, '<tt>structure</tt>'
1538 or '<tt>void</tt>' values, etc...). Both arguments must have identical
1541 <p>The '<tt>seteq</tt>' instruction yields a <tt>true</tt> '<tt>bool</tt>'
1542 value if both operands are equal.<br>
1543 The '<tt>setne</tt>' instruction yields a <tt>true</tt> '<tt>bool</tt>'
1544 value if both operands are unequal.<br>
1545 The '<tt>setlt</tt>' instruction yields a <tt>true</tt> '<tt>bool</tt>'
1546 value if the first operand is less than the second operand.<br>
1547 The '<tt>setgt</tt>' instruction yields a <tt>true</tt> '<tt>bool</tt>'
1548 value if the first operand is greater than the second operand.<br>
1549 The '<tt>setle</tt>' instruction yields a <tt>true</tt> '<tt>bool</tt>'
1550 value if the first operand is less than or equal to the second operand.<br>
1551 The '<tt>setge</tt>' instruction yields a <tt>true</tt> '<tt>bool</tt>'
1552 value if the first operand is greater than or equal to the second
1555 <pre> <result> = seteq int 4, 5 <i>; yields {bool}:result = false</i>
1556 <result> = setne float 4, 5 <i>; yields {bool}:result = true</i>
1557 <result> = setlt uint 4, 5 <i>; yields {bool}:result = true</i>
1558 <result> = setgt sbyte 4, 5 <i>; yields {bool}:result = false</i>
1559 <result> = setle sbyte 4, 5 <i>; yields {bool}:result = true</i>
1560 <result> = setge sbyte 4, 5 <i>; yields {bool}:result = false</i>
1563 <!-- ======================================================================= -->
1564 <div class="doc_subsection"> <a name="bitwiseops">Bitwise Binary
1565 Operations</a> </div>
1566 <div class="doc_text">
1567 <p>Bitwise binary operators are used to do various forms of
1568 bit-twiddling in a program. They are generally very efficient
1569 instructions and can commonly be strength reduced from other
1570 instructions. They require two operands, execute an operation on them,
1571 and produce a single value. The resulting value of the bitwise binary
1572 operators is always the same type as its first operand.</p>
1574 <!-- _______________________________________________________________________ -->
1575 <div class="doc_subsubsection"> <a name="i_and">'<tt>and</tt>'
1576 Instruction</a> </div>
1577 <div class="doc_text">
1579 <pre> <result> = and <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1582 <p>The '<tt>and</tt>' instruction returns the bitwise logical and of
1583 its two operands.</p>
1585 <p>The two arguments to the '<tt>and</tt>' instruction must be <a
1586 href="#t_integral">integral</a> values. Both arguments must have
1587 identical types.</p>
1589 <p>The truth table used for the '<tt>and</tt>' instruction is:</p>
1591 <div style="align: center">
1592 <table border="1" cellspacing="0" cellpadding="4">
1623 <pre> <result> = and int 4, %var <i>; yields {int}:result = 4 & %var</i>
1624 <result> = and int 15, 40 <i>; yields {int}:result = 8</i>
1625 <result> = and int 4, 8 <i>; yields {int}:result = 0</i>
1628 <!-- _______________________________________________________________________ -->
1629 <div class="doc_subsubsection"> <a name="i_or">'<tt>or</tt>' Instruction</a> </div>
1630 <div class="doc_text">
1632 <pre> <result> = or <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1635 <p>The '<tt>or</tt>' instruction returns the bitwise logical inclusive
1636 or of its two operands.</p>
1638 <p>The two arguments to the '<tt>or</tt>' instruction must be <a
1639 href="#t_integral">integral</a> values. Both arguments must have
1640 identical types.</p>
1642 <p>The truth table used for the '<tt>or</tt>' instruction is:</p>
1644 <div style="align: center">
1645 <table border="1" cellspacing="0" cellpadding="4">
1676 <pre> <result> = or int 4, %var <i>; yields {int}:result = 4 | %var</i>
1677 <result> = or int 15, 40 <i>; yields {int}:result = 47</i>
1678 <result> = or int 4, 8 <i>; yields {int}:result = 12</i>
1681 <!-- _______________________________________________________________________ -->
1682 <div class="doc_subsubsection"> <a name="i_xor">'<tt>xor</tt>'
1683 Instruction</a> </div>
1684 <div class="doc_text">
1686 <pre> <result> = xor <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1689 <p>The '<tt>xor</tt>' instruction returns the bitwise logical exclusive
1690 or of its two operands. The <tt>xor</tt> is used to implement the
1691 "one's complement" operation, which is the "~" operator in C.</p>
1693 <p>The two arguments to the '<tt>xor</tt>' instruction must be <a
1694 href="#t_integral">integral</a> values. Both arguments must have
1695 identical types.</p>
1697 <p>The truth table used for the '<tt>xor</tt>' instruction is:</p>
1699 <div style="align: center">
1700 <table border="1" cellspacing="0" cellpadding="4">
1732 <pre> <result> = xor int 4, %var <i>; yields {int}:result = 4 ^ %var</i>
1733 <result> = xor int 15, 40 <i>; yields {int}:result = 39</i>
1734 <result> = xor int 4, 8 <i>; yields {int}:result = 12</i>
1735 <result> = xor int %V, -1 <i>; yields {int}:result = ~%V</i>
1738 <!-- _______________________________________________________________________ -->
1739 <div class="doc_subsubsection"> <a name="i_shl">'<tt>shl</tt>'
1740 Instruction</a> </div>
1741 <div class="doc_text">
1743 <pre> <result> = shl <ty> <var1>, ubyte <var2> <i>; yields {ty}:result</i>
1746 <p>The '<tt>shl</tt>' instruction returns the first operand shifted to
1747 the left a specified number of bits.</p>
1749 <p>The first argument to the '<tt>shl</tt>' instruction must be an <a
1750 href="#t_integer">integer</a> type. The second argument must be an '<tt>ubyte</tt>'
1753 <p>The value produced is <tt>var1</tt> * 2<sup><tt>var2</tt></sup>.</p>
1755 <pre> <result> = shl int 4, ubyte %var <i>; yields {int}:result = 4 << %var</i>
1756 <result> = shl int 4, ubyte 2 <i>; yields {int}:result = 16</i>
1757 <result> = shl int 1, ubyte 10 <i>; yields {int}:result = 1024</i>
1760 <!-- _______________________________________________________________________ -->
1761 <div class="doc_subsubsection"> <a name="i_shr">'<tt>shr</tt>'
1762 Instruction</a> </div>
1763 <div class="doc_text">
1765 <pre> <result> = shr <ty> <var1>, ubyte <var2> <i>; yields {ty}:result</i>
1768 <p>The '<tt>shr</tt>' instruction returns the first operand shifted to
1769 the right a specified number of bits.</p>
1771 <p>The first argument to the '<tt>shr</tt>' instruction must be an <a
1772 href="#t_integer">integer</a> type. The second argument must be an '<tt>ubyte</tt>'
1775 <p>If the first argument is a <a href="#t_signed">signed</a> type, the
1776 most significant bit is duplicated in the newly free'd bit positions.
1777 If the first argument is unsigned, zero bits shall fill the empty
1780 <pre> <result> = shr int 4, ubyte %var <i>; yields {int}:result = 4 >> %var</i>
1781 <result> = shr uint 4, ubyte 1 <i>; yields {uint}:result = 2</i>
1782 <result> = shr int 4, ubyte 2 <i>; yields {int}:result = 1</i>
1783 <result> = shr sbyte 4, ubyte 3 <i>; yields {sbyte}:result = 0</i>
1784 <result> = shr sbyte -2, ubyte 1 <i>; yields {sbyte}:result = -1</i>
1788 <!-- ======================================================================= -->
1789 <div class="doc_subsection">
1790 <a name="memoryops">Memory Access Operations</a>
1793 <div class="doc_text">
1795 <p>A key design point of an SSA-based representation is how it
1796 represents memory. In LLVM, no memory locations are in SSA form, which
1797 makes things very simple. This section describes how to read, write,
1798 allocate, and free memory in LLVM.</p>
1802 <!-- _______________________________________________________________________ -->
1803 <div class="doc_subsubsection">
1804 <a name="i_malloc">'<tt>malloc</tt>' Instruction</a>
1807 <div class="doc_text">
1812 <result> = malloc <type>[, uint <NumElements>][, align <alignment>] <i>; yields {type*}:result</i>
1817 <p>The '<tt>malloc</tt>' instruction allocates memory from the system
1818 heap and returns a pointer to it.</p>
1822 <p>The '<tt>malloc</tt>' instruction allocates
1823 <tt>sizeof(<type>)*NumElements</tt>
1824 bytes of memory from the operating system and returns a pointer of the
1825 appropriate type to the program. If "NumElements" is specified, it is the
1826 number of elements allocated. If an alignment is specified, the value result
1827 of the allocation is guaranteed to be aligned to at least that boundary. If
1828 not specified, or if zero, the target can choose to align the allocation on any
1829 convenient boundary.</p>
1831 <p>'<tt>type</tt>' must be a sized type.</p>
1835 <p>Memory is allocated using the system "<tt>malloc</tt>" function, and
1836 a pointer is returned.</p>
1841 %array = malloc [4 x ubyte ] <i>; yields {[%4 x ubyte]*}:array</i>
1843 %size = <a href="#i_add">add</a> uint 2, 2 <i>; yields {uint}:size = uint 4</i>
1844 %array1 = malloc ubyte, uint 4 <i>; yields {ubyte*}:array1</i>
1845 %array2 = malloc [12 x ubyte], uint %size <i>; yields {[12 x ubyte]*}:array2</i>
1846 %array3 = malloc int, uint 4, align 1024 <i>; yields {int*}:array3</i>
1847 %array4 = malloc int, align 1024 <i>; yields {int*}:array4</i>
1851 <!-- _______________________________________________________________________ -->
1852 <div class="doc_subsubsection">
1853 <a name="i_free">'<tt>free</tt>' Instruction</a>
1856 <div class="doc_text">
1861 free <type> <value> <i>; yields {void}</i>
1866 <p>The '<tt>free</tt>' instruction returns memory back to the unused
1867 memory heap to be reallocated in the future.</p>
1871 <p>'<tt>value</tt>' shall be a pointer value that points to a value
1872 that was allocated with the '<tt><a href="#i_malloc">malloc</a></tt>'
1877 <p>Access to the memory pointed to by the pointer is no longer defined
1878 after this instruction executes.</p>
1883 %array = <a href="#i_malloc">malloc</a> [4 x ubyte] <i>; yields {[4 x ubyte]*}:array</i>
1884 free [4 x ubyte]* %array
1888 <!-- _______________________________________________________________________ -->
1889 <div class="doc_subsubsection">
1890 <a name="i_alloca">'<tt>alloca</tt>' Instruction</a>
1893 <div class="doc_text">
1898 <result> = alloca <type>[, uint <NumElements>][, align <alignment>] <i>; yields {type*}:result</i>
1903 <p>The '<tt>alloca</tt>' instruction allocates memory on the current
1904 stack frame of the procedure that is live until the current function
1905 returns to its caller.</p>
1909 <p>The '<tt>alloca</tt>' instruction allocates <tt>sizeof(<type>)*NumElements</tt>
1910 bytes of memory on the runtime stack, returning a pointer of the
1911 appropriate type to the program. If "NumElements" is specified, it is the
1912 number of elements allocated. If an alignment is specified, the value result
1913 of the allocation is guaranteed to be aligned to at least that boundary. If
1914 not specified, or if zero, the target can choose to align the allocation on any
1915 convenient boundary.</p>
1917 <p>'<tt>type</tt>' may be any sized type.</p>
1921 <p>Memory is allocated; a pointer is returned. '<tt>alloca</tt>'d
1922 memory is automatically released when the function returns. The '<tt>alloca</tt>'
1923 instruction is commonly used to represent automatic variables that must
1924 have an address available. When the function returns (either with the <tt><a
1925 href="#i_ret">ret</a></tt> or <tt><a href="#i_unwind">unwind</a></tt>
1926 instructions), the memory is reclaimed.</p>
1931 %ptr = alloca int <i>; yields {int*}:ptr</i>
1932 %ptr = alloca int, uint 4 <i>; yields {int*}:ptr</i>
1933 %ptr = alloca int, uint 4, align 1024 <i>; yields {int*}:ptr</i>
1934 %ptr = alloca int, align 1024 <i>; yields {int*}:ptr</i>
1938 <!-- _______________________________________________________________________ -->
1939 <div class="doc_subsubsection"> <a name="i_load">'<tt>load</tt>'
1940 Instruction</a> </div>
1941 <div class="doc_text">
1943 <pre> <result> = load <ty>* <pointer><br> <result> = volatile load <ty>* <pointer><br></pre>
1945 <p>The '<tt>load</tt>' instruction is used to read from memory.</p>
1947 <p>The argument to the '<tt>load</tt>' instruction specifies the memory
1948 address from which to load. The pointer must point to a <a
1949 href="#t_firstclass">first class</a> type. If the <tt>load</tt> is
1950 marked as <tt>volatile</tt>, then the optimizer is not allowed to modify
1951 the number or order of execution of this <tt>load</tt> with other
1952 volatile <tt>load</tt> and <tt><a href="#i_store">store</a></tt>
1955 <p>The location of memory pointed to is loaded.</p>
1957 <pre> %ptr = <a href="#i_alloca">alloca</a> int <i>; yields {int*}:ptr</i>
1959 href="#i_store">store</a> int 3, int* %ptr <i>; yields {void}</i>
1960 %val = load int* %ptr <i>; yields {int}:val = int 3</i>
1963 <!-- _______________________________________________________________________ -->
1964 <div class="doc_subsubsection"> <a name="i_store">'<tt>store</tt>'
1965 Instruction</a> </div>
1967 <pre> store <ty> <value>, <ty>* <pointer> <i>; yields {void}</i>
1968 volatile store <ty> <value>, <ty>* <pointer> <i>; yields {void}</i>
1971 <p>The '<tt>store</tt>' instruction is used to write to memory.</p>
1973 <p>There are two arguments to the '<tt>store</tt>' instruction: a value
1974 to store and an address in which to store it. The type of the '<tt><pointer></tt>'
1975 operand must be a pointer to the type of the '<tt><value></tt>'
1976 operand. If the <tt>store</tt> is marked as <tt>volatile</tt>, then the
1977 optimizer is not allowed to modify the number or order of execution of
1978 this <tt>store</tt> with other volatile <tt>load</tt> and <tt><a
1979 href="#i_store">store</a></tt> instructions.</p>
1981 <p>The contents of memory are updated to contain '<tt><value></tt>'
1982 at the location specified by the '<tt><pointer></tt>' operand.</p>
1984 <pre> %ptr = <a href="#i_alloca">alloca</a> int <i>; yields {int*}:ptr</i>
1986 href="#i_store">store</a> int 3, int* %ptr <i>; yields {void}</i>
1987 %val = load int* %ptr <i>; yields {int}:val = int 3</i>
1989 <!-- _______________________________________________________________________ -->
1990 <div class="doc_subsubsection">
1991 <a name="i_getelementptr">'<tt>getelementptr</tt>' Instruction</a>
1994 <div class="doc_text">
1997 <result> = getelementptr <ty>* <ptrval>{, <ty> <idx>}*
2003 The '<tt>getelementptr</tt>' instruction is used to get the address of a
2004 subelement of an aggregate data structure.</p>
2008 <p>This instruction takes a list of integer constants that indicate what
2009 elements of the aggregate object to index to. The actual types of the arguments
2010 provided depend on the type of the first pointer argument. The
2011 '<tt>getelementptr</tt>' instruction is used to index down through the type
2012 levels of a structure or to a specific index in an array. When indexing into a
2013 structure, only <tt>uint</tt>
2014 integer constants are allowed. When indexing into an array or pointer,
2015 <tt>int</tt> and <tt>long</tt> indexes are allowed of any sign.</p>
2017 <p>For example, let's consider a C code fragment and how it gets
2018 compiled to LLVM:</p>
2032 int *foo(struct ST *s) {
2033 return &s[1].Z.B[5][13];
2037 <p>The LLVM code generated by the GCC frontend is:</p>
2040 %RT = type { sbyte, [10 x [20 x int]], sbyte }
2041 %ST = type { int, double, %RT }
2045 int* %foo(%ST* %s) {
2047 %reg = getelementptr %ST* %s, int 1, uint 2, uint 1, int 5, int 13
2054 <p>The index types specified for the '<tt>getelementptr</tt>' instruction depend
2055 on the pointer type that is being indexed into. <a href="#t_pointer">Pointer</a>
2056 and <a href="#t_array">array</a> types require <tt>uint</tt>, <tt>int</tt>,
2057 <tt>ulong</tt>, or <tt>long</tt> values, and <a href="#t_struct">structure</a>
2058 types require <tt>uint</tt> <b>constants</b>.</p>
2060 <p>In the example above, the first index is indexing into the '<tt>%ST*</tt>'
2061 type, which is a pointer, yielding a '<tt>%ST</tt>' = '<tt>{ int, double, %RT
2062 }</tt>' type, a structure. The second index indexes into the third element of
2063 the structure, yielding a '<tt>%RT</tt>' = '<tt>{ sbyte, [10 x [20 x int]],
2064 sbyte }</tt>' type, another structure. The third index indexes into the second
2065 element of the structure, yielding a '<tt>[10 x [20 x int]]</tt>' type, an
2066 array. The two dimensions of the array are subscripted into, yielding an
2067 '<tt>int</tt>' type. The '<tt>getelementptr</tt>' instruction returns a pointer
2068 to this element, thus computing a value of '<tt>int*</tt>' type.</p>
2070 <p>Note that it is perfectly legal to index partially through a
2071 structure, returning a pointer to an inner element. Because of this,
2072 the LLVM code for the given testcase is equivalent to:</p>
2075 int* %foo(%ST* %s) {
2076 %t1 = getelementptr %ST* %s, int 1 <i>; yields %ST*:%t1</i>
2077 %t2 = getelementptr %ST* %t1, int 0, uint 2 <i>; yields %RT*:%t2</i>
2078 %t3 = getelementptr %RT* %t2, int 0, uint 1 <i>; yields [10 x [20 x int]]*:%t3</i>
2079 %t4 = getelementptr [10 x [20 x int]]* %t3, int 0, int 5 <i>; yields [20 x int]*:%t4</i>
2080 %t5 = getelementptr [20 x int]* %t4, int 0, int 13 <i>; yields int*:%t5</i>
2085 <p>Note that it is undefined to access an array out of bounds: array and
2086 pointer indexes must always be within the defined bounds of the array type.
2087 The one exception for this rules is zero length arrays. These arrays are
2088 defined to be accessible as variable length arrays, which requires access
2089 beyond the zero'th element.</p>
2094 <i>; yields [12 x ubyte]*:aptr</i>
2095 %aptr = getelementptr {int, [12 x ubyte]}* %sptr, long 0, uint 1
2099 <!-- ======================================================================= -->
2100 <div class="doc_subsection"> <a name="otherops">Other Operations</a> </div>
2101 <div class="doc_text">
2102 <p>The instructions in this category are the "miscellaneous"
2103 instructions, which defy better classification.</p>
2105 <!-- _______________________________________________________________________ -->
2106 <div class="doc_subsubsection"> <a name="i_phi">'<tt>phi</tt>'
2107 Instruction</a> </div>
2108 <div class="doc_text">
2110 <pre> <result> = phi <ty> [ <val0>, <label0>], ...<br></pre>
2112 <p>The '<tt>phi</tt>' instruction is used to implement the φ node in
2113 the SSA graph representing the function.</p>
2115 <p>The type of the incoming values are specified with the first type
2116 field. After this, the '<tt>phi</tt>' instruction takes a list of pairs
2117 as arguments, with one pair for each predecessor basic block of the
2118 current block. Only values of <a href="#t_firstclass">first class</a>
2119 type may be used as the value arguments to the PHI node. Only labels
2120 may be used as the label arguments.</p>
2121 <p>There must be no non-phi instructions between the start of a basic
2122 block and the PHI instructions: i.e. PHI instructions must be first in
2125 <p>At runtime, the '<tt>phi</tt>' instruction logically takes on the
2126 value specified by the parameter, depending on which basic block we
2127 came from in the last <a href="#terminators">terminator</a> instruction.</p>
2129 <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>
2132 <!-- _______________________________________________________________________ -->
2133 <div class="doc_subsubsection">
2134 <a name="i_cast">'<tt>cast .. to</tt>' Instruction</a>
2137 <div class="doc_text">
2142 <result> = cast <ty> <value> to <ty2> <i>; yields ty2</i>
2148 The '<tt>cast</tt>' instruction is used as the primitive means to convert
2149 integers to floating point, change data type sizes, and break type safety (by
2157 The '<tt>cast</tt>' instruction takes a value to cast, which must be a first
2158 class value, and a type to cast it to, which must also be a <a
2159 href="#t_firstclass">first class</a> type.
2165 This instruction follows the C rules for explicit casts when determining how the
2166 data being cast must change to fit in its new container.
2170 When casting to bool, any value that would be considered true in the context of
2171 a C '<tt>if</tt>' condition is converted to the boolean '<tt>true</tt>' values,
2172 all else are '<tt>false</tt>'.
2176 When extending an integral value from a type of one signness to another (for
2177 example '<tt>sbyte</tt>' to '<tt>ulong</tt>'), the value is sign-extended if the
2178 <b>source</b> value is signed, and zero-extended if the source value is
2179 unsigned. <tt>bool</tt> values are always zero extended into either zero or
2186 %X = cast int 257 to ubyte <i>; yields ubyte:1</i>
2187 %Y = cast int 123 to bool <i>; yields bool:true</i>
2191 <!-- _______________________________________________________________________ -->
2192 <div class="doc_subsubsection">
2193 <a name="i_select">'<tt>select</tt>' Instruction</a>
2196 <div class="doc_text">
2201 <result> = select bool <cond>, <ty> <val1>, <ty> <val2> <i>; yields ty</i>
2207 The '<tt>select</tt>' instruction is used to choose one value based on a
2208 condition, without branching.
2215 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.
2221 If the boolean condition evaluates to true, the instruction returns the first
2222 value argument; otherwise, it returns the second value argument.
2228 %X = select bool true, ubyte 17, ubyte 42 <i>; yields ubyte:17</i>
2236 <!-- _______________________________________________________________________ -->
2237 <div class="doc_subsubsection">
2238 <a name="i_call">'<tt>call</tt>' Instruction</a>
2241 <div class="doc_text">
2245 <result> = [tail] call [<a href="#callingconv">cconv</a>] <ty>* <fnptrval>(<param list>)
2250 <p>The '<tt>call</tt>' instruction represents a simple function call.</p>
2254 <p>This instruction requires several arguments:</p>
2258 <p>The optional "tail" marker indicates whether the callee function accesses
2259 any allocas or varargs in the caller. If the "tail" marker is present, the
2260 function call is eligible for tail call optimization. Note that calls may
2261 be marked "tail" even if they do not occur before a <a
2262 href="#i_ret"><tt>ret</tt></a> instruction.
2265 <p>The optional "cconv" marker indicates which <a href="callingconv">calling
2266 convention</a> the call should use. If none is specified, the call defaults
2267 to using C calling conventions.
2270 <p>'<tt>ty</tt>': shall be the signature of the pointer to function value
2271 being invoked. The argument types must match the types implied by this
2272 signature. This type can be omitted if the function is not varargs and
2273 if the function type does not return a pointer to a function.</p>
2276 <p>'<tt>fnptrval</tt>': An LLVM value containing a pointer to a function to
2277 be invoked. In most cases, this is a direct function invocation, but
2278 indirect <tt>call</tt>s are just as possible, calling an arbitrary pointer
2279 to function value.</p>
2282 <p>'<tt>function args</tt>': argument list whose types match the
2283 function signature argument types. All arguments must be of
2284 <a href="#t_firstclass">first class</a> type. If the function signature
2285 indicates the function accepts a variable number of arguments, the extra
2286 arguments can be specified.</p>
2292 <p>The '<tt>call</tt>' instruction is used to cause control flow to
2293 transfer to a specified function, with its incoming arguments bound to
2294 the specified values. Upon a '<tt><a href="#i_ret">ret</a></tt>'
2295 instruction in the called function, control flow continues with the
2296 instruction after the function call, and the return value of the
2297 function is bound to the result argument. This is a simpler case of
2298 the <a href="#i_invoke">invoke</a> instruction.</p>
2303 %retval = call int %test(int %argc)
2304 call int(sbyte*, ...) *%printf(sbyte* %msg, int 12, sbyte 42);
2305 %X = tail call int %foo()
2306 %Y = tail call <a href="#callingconv">fastcc</a> int %foo()
2311 <!-- _______________________________________________________________________ -->
2312 <div class="doc_subsubsection">
2313 <a name="i_vaarg">'<tt>vaarg</tt>' Instruction</a>
2316 <div class="doc_text">
2321 <resultval> = va_arg <va_list*> <arglist>, <argty>
2326 <p>The '<tt>va_arg</tt>' instruction is used to access arguments passed through
2327 the "variable argument" area of a function call. It is used to implement the
2328 <tt>va_arg</tt> macro in C.</p>
2332 <p>This instruction takes a <tt>va_list*</tt> value and the type of
2333 the argument. It returns a value of the specified argument type and
2334 increments the <tt>va_list</tt> to point to the next argument. Again, the
2335 actual type of <tt>va_list</tt> is target specific.</p>
2339 <p>The '<tt>va_arg</tt>' instruction loads an argument of the specified
2340 type from the specified <tt>va_list</tt> and causes the
2341 <tt>va_list</tt> to point to the next argument. For more information,
2342 see the variable argument handling <a href="#int_varargs">Intrinsic
2345 <p>It is legal for this instruction to be called in a function which does not
2346 take a variable number of arguments, for example, the <tt>vfprintf</tt>
2349 <p><tt>va_arg</tt> is an LLVM instruction instead of an <a
2350 href="#intrinsics">intrinsic function</a> because it takes a type as an
2355 <p>See the <a href="#int_varargs">variable argument processing</a> section.</p>
2359 <!-- *********************************************************************** -->
2360 <div class="doc_section"> <a name="intrinsics">Intrinsic Functions</a> </div>
2361 <!-- *********************************************************************** -->
2363 <div class="doc_text">
2365 <p>LLVM supports the notion of an "intrinsic function". These functions have
2366 well known names and semantics and are required to follow certain
2367 restrictions. Overall, these instructions represent an extension mechanism for
2368 the LLVM language that does not require changing all of the transformations in
2369 LLVM to add to the language (or the bytecode reader/writer, the parser,
2372 <p>Intrinsic function names must all start with an "<tt>llvm.</tt>" prefix. This
2373 prefix is reserved in LLVM for intrinsic names; thus, functions may not be named
2374 this. Intrinsic functions must always be external functions: you cannot define
2375 the body of intrinsic functions. Intrinsic functions may only be used in call
2376 or invoke instructions: it is illegal to take the address of an intrinsic
2377 function. Additionally, because intrinsic functions are part of the LLVM
2378 language, it is required that they all be documented here if any are added.</p>
2381 <p>To learn how to add an intrinsic function, please see the <a
2382 href="ExtendingLLVM.html">Extending LLVM Guide</a>.
2387 <!-- ======================================================================= -->
2388 <div class="doc_subsection">
2389 <a name="int_varargs">Variable Argument Handling Intrinsics</a>
2392 <div class="doc_text">
2394 <p>Variable argument support is defined in LLVM with the <a
2395 href="#i_vanext"><tt>vanext</tt></a> instruction and these three
2396 intrinsic functions. These functions are related to the similarly
2397 named macros defined in the <tt><stdarg.h></tt> header file.</p>
2399 <p>All of these functions operate on arguments that use a
2400 target-specific value type "<tt>va_list</tt>". The LLVM assembly
2401 language reference manual does not define what this type is, so all
2402 transformations should be prepared to handle intrinsics with any type
2405 <p>This example shows how the <a href="#i_vanext"><tt>vanext</tt></a>
2406 instruction and the variable argument handling intrinsic functions are
2410 int %test(int %X, ...) {
2411 ; Initialize variable argument processing
2413 call void %<a href="#i_va_start">llvm.va_start</a>(sbyte** %ap)
2415 ; Read a single integer argument
2416 %tmp = va_arg sbyte** %ap, int
2418 ; Demonstrate usage of llvm.va_copy and llvm.va_end
2420 call void %<a href="#i_va_copy">llvm.va_copy</a>(sbyte** %aq, sbyte** %ap)
2421 call void %<a href="#i_va_end">llvm.va_end</a>(sbyte** %aq)
2423 ; Stop processing of arguments.
2424 call void %<a href="#i_va_end">llvm.va_end</a>(sbyte** %ap)
2430 <!-- _______________________________________________________________________ -->
2431 <div class="doc_subsubsection">
2432 <a name="i_va_start">'<tt>llvm.va_start</tt>' Intrinsic</a>
2436 <div class="doc_text">
2438 <pre> declare void %llvm.va_start(<va_list>* <arglist>)<br></pre>
2440 <P>The '<tt>llvm.va_start</tt>' intrinsic initializes
2441 <tt>*<arglist></tt> for subsequent use by <tt><a
2442 href="#i_va_arg">va_arg</a></tt>.</p>
2446 <P>The argument is a pointer to a <tt>va_list</tt> element to initialize.</p>
2450 <P>The '<tt>llvm.va_start</tt>' intrinsic works just like the <tt>va_start</tt>
2451 macro available in C. In a target-dependent way, it initializes the
2452 <tt>va_list</tt> element the argument points to, so that the next call to
2453 <tt>va_arg</tt> will produce the first variable argument passed to the function.
2454 Unlike the C <tt>va_start</tt> macro, this intrinsic does not need to know the
2455 last argument of the function, the compiler can figure that out.</p>
2459 <!-- _______________________________________________________________________ -->
2460 <div class="doc_subsubsection">
2461 <a name="i_va_end">'<tt>llvm.va_end</tt>' Intrinsic</a>
2464 <div class="doc_text">
2466 <pre> declare void %llvm.va_end(<va_list*> <arglist>)<br></pre>
2468 <p>The '<tt>llvm.va_end</tt>' intrinsic destroys <tt><arglist></tt>
2469 which has been initialized previously with <tt><a href="#i_va_start">llvm.va_start</a></tt>
2470 or <tt><a href="#i_va_copy">llvm.va_copy</a></tt>.</p>
2472 <p>The argument is a <tt>va_list</tt> to destroy.</p>
2474 <p>The '<tt>llvm.va_end</tt>' intrinsic works just like the <tt>va_end</tt>
2475 macro available in C. In a target-dependent way, it destroys the <tt>va_list</tt>.
2476 Calls to <a href="#i_va_start"><tt>llvm.va_start</tt></a> and <a
2477 href="#i_va_copy"><tt>llvm.va_copy</tt></a> must be matched exactly
2478 with calls to <tt>llvm.va_end</tt>.</p>
2481 <!-- _______________________________________________________________________ -->
2482 <div class="doc_subsubsection">
2483 <a name="i_va_copy">'<tt>llvm.va_copy</tt>' Intrinsic</a>
2486 <div class="doc_text">
2491 declare void %llvm.va_copy(<va_list>* <destarglist>,
2492 <va_list>* <srcarglist>)
2497 <p>The '<tt>llvm.va_copy</tt>' intrinsic copies the current argument position from
2498 the source argument list to the destination argument list.</p>
2502 <p>The first argument is a pointer to a <tt>va_list</tt> element to initialize.
2503 The second argument is a pointer to a <tt>va_list</tt> element to copy from.</p>
2508 <p>The '<tt>llvm.va_copy</tt>' intrinsic works just like the <tt>va_copy</tt> macro
2509 available in C. In a target-dependent way, it copies the source
2510 <tt>va_list</tt> element into the destination list. This intrinsic is necessary
2511 because the <tt><a href="i_va_begin">llvm.va_begin</a></tt> intrinsic may be
2512 arbitrarily complex and require memory allocation, for example.</p>
2516 <!-- ======================================================================= -->
2517 <div class="doc_subsection">
2518 <a name="int_gc">Accurate Garbage Collection Intrinsics</a>
2521 <div class="doc_text">
2524 LLVM support for <a href="GarbageCollection.html">Accurate Garbage
2525 Collection</a> requires the implementation and generation of these intrinsics.
2526 These intrinsics allow identification of <a href="#i_gcroot">GC roots on the
2527 stack</a>, as well as garbage collector implementations that require <a
2528 href="#i_gcread">read</a> and <a href="#i_gcwrite">write</a> barriers.
2529 Front-ends for type-safe garbage collected languages should generate these
2530 intrinsics to make use of the LLVM garbage collectors. For more details, see <a
2531 href="GarbageCollection.html">Accurate Garbage Collection with LLVM</a>.
2535 <!-- _______________________________________________________________________ -->
2536 <div class="doc_subsubsection">
2537 <a name="i_gcroot">'<tt>llvm.gcroot</tt>' Intrinsic</a>
2540 <div class="doc_text">
2545 declare void %llvm.gcroot(<ty>** %ptrloc, <ty2>* %metadata)
2550 <p>The '<tt>llvm.gcroot</tt>' intrinsic declares the existence of a GC root to
2551 the code generator, and allows some metadata to be associated with it.</p>
2555 <p>The first argument specifies the address of a stack object that contains the
2556 root pointer. The second pointer (which must be either a constant or a global
2557 value address) contains the meta-data to be associated with the root.</p>
2561 <p>At runtime, a call to this intrinsics stores a null pointer into the "ptrloc"
2562 location. At compile-time, the code generator generates information to allow
2563 the runtime to find the pointer at GC safe points.
2569 <!-- _______________________________________________________________________ -->
2570 <div class="doc_subsubsection">
2571 <a name="i_gcread">'<tt>llvm.gcread</tt>' Intrinsic</a>
2574 <div class="doc_text">
2579 declare sbyte* %llvm.gcread(sbyte** %Ptr)
2584 <p>The '<tt>llvm.gcread</tt>' intrinsic identifies reads of references from heap
2585 locations, allowing garbage collector implementations that require read
2590 <p>The argument is the address to read from, which should be an address
2591 allocated from the garbage collector.</p>
2595 <p>The '<tt>llvm.gcread</tt>' intrinsic has the same semantics as a load
2596 instruction, but may be replaced with substantially more complex code by the
2597 garbage collector runtime, as needed.</p>
2602 <!-- _______________________________________________________________________ -->
2603 <div class="doc_subsubsection">
2604 <a name="i_gcwrite">'<tt>llvm.gcwrite</tt>' Intrinsic</a>
2607 <div class="doc_text">
2612 declare void %llvm.gcwrite(sbyte* %P1, sbyte** %P2)
2617 <p>The '<tt>llvm.gcwrite</tt>' intrinsic identifies writes of references to heap
2618 locations, allowing garbage collector implementations that require write
2619 barriers (such as generational or reference counting collectors).</p>
2623 <p>The first argument is the reference to store, and the second is the heap
2624 location to store to.</p>
2628 <p>The '<tt>llvm.gcwrite</tt>' intrinsic has the same semantics as a store
2629 instruction, but may be replaced with substantially more complex code by the
2630 garbage collector runtime, as needed.</p>
2636 <!-- ======================================================================= -->
2637 <div class="doc_subsection">
2638 <a name="int_codegen">Code Generator Intrinsics</a>
2641 <div class="doc_text">
2643 These intrinsics are provided by LLVM to expose special features that may only
2644 be implemented with code generator support.
2649 <!-- _______________________________________________________________________ -->
2650 <div class="doc_subsubsection">
2651 <a name="i_returnaddress">'<tt>llvm.returnaddress</tt>' Intrinsic</a>
2654 <div class="doc_text">
2658 declare void* %llvm.returnaddress(uint <level>)
2664 The '<tt>llvm.returnaddress</tt>' intrinsic returns a target-specific value
2665 indicating the return address of the current function or one of its callers.
2671 The argument to this intrinsic indicates which function to return the address
2672 for. Zero indicates the calling function, one indicates its caller, etc. The
2673 argument is <b>required</b> to be a constant integer value.
2679 The '<tt>llvm.returnaddress</tt>' intrinsic either returns a pointer indicating
2680 the return address of the specified call frame, or zero if it cannot be
2681 identified. The value returned by this intrinsic is likely to be incorrect or 0
2682 for arguments other than zero, so it should only be used for debugging purposes.
2686 Note that calling this intrinsic does not prevent function inlining or other
2687 aggressive transformations, so the value returned may not be that of the obvious
2688 source-language caller.
2693 <!-- _______________________________________________________________________ -->
2694 <div class="doc_subsubsection">
2695 <a name="i_frameaddress">'<tt>llvm.frameaddress</tt>' Intrinsic</a>
2698 <div class="doc_text">
2702 declare void* %llvm.frameaddress(uint <level>)
2708 The '<tt>llvm.frameaddress</tt>' intrinsic returns the target-specific frame
2709 pointer value for the specified stack frame.
2715 The argument to this intrinsic indicates which function to return the frame
2716 pointer for. Zero indicates the calling function, one indicates its caller,
2717 etc. The argument is <b>required</b> to be a constant integer value.
2723 The '<tt>llvm.frameaddress</tt>' intrinsic either returns a pointer indicating
2724 the frame address of the specified call frame, or zero if it cannot be
2725 identified. The value returned by this intrinsic is likely to be incorrect or 0
2726 for arguments other than zero, so it should only be used for debugging purposes.
2730 Note that calling this intrinsic does not prevent function inlining or other
2731 aggressive transformations, so the value returned may not be that of the obvious
2732 source-language caller.
2736 <!-- _______________________________________________________________________ -->
2737 <div class="doc_subsubsection">
2738 <a name="i_prefetch">'<tt>llvm.prefetch</tt>' Intrinsic</a>
2741 <div class="doc_text">
2745 declare void %llvm.prefetch(sbyte * <address>,
2746 uint <rw>, uint <locality>)
2753 The '<tt>llvm.prefetch</tt>' intrinsic is a hint to the code generator to insert
2754 a prefetch instruction if supported; otherwise, it is a noop. Prefetches have
2756 effect on the behavior of the program but can change its performance
2763 <tt>address</tt> is the address to be prefetched, <tt>rw</tt> is the specifier
2764 determining if the fetch should be for a read (0) or write (1), and
2765 <tt>locality</tt> is a temporal locality specifier ranging from (0) - no
2766 locality, to (3) - extremely local keep in cache. The <tt>rw</tt> and
2767 <tt>locality</tt> arguments must be constant integers.
2773 This intrinsic does not modify the behavior of the program. In particular,
2774 prefetches cannot trap and do not produce a value. On targets that support this
2775 intrinsic, the prefetch can provide hints to the processor cache for better
2781 <!-- _______________________________________________________________________ -->
2782 <div class="doc_subsubsection">
2783 <a name="i_pcmarker">'<tt>llvm.pcmarker</tt>' Intrinsic</a>
2786 <div class="doc_text">
2790 declare void %llvm.pcmarker( uint <id> )
2797 The '<tt>llvm.pcmarker</tt>' intrinsic is a method to export a Program Counter
2799 code to simulators and other tools. The method is target specific, but it is
2800 expected that the marker will use exported symbols to transmit the PC of the marker.
2801 The marker makes no guarantees that it will remain with any specific instruction
2802 after optimizations. It is possible that the presense of a marker will inhibit
2803 optimizations. The intended use is to be inserted after optmizations to allow
2804 correlations of simulation runs.
2810 <tt>id</tt> is a numerical id identifying the marker.
2816 This intrinsic does not modify the behavior of the program. Backends that do not
2817 support this intrinisic may ignore it.
2822 <!-- _______________________________________________________________________ -->
2823 <div class="doc_subsubsection">
2824 <a name="i_readcyclecounter">'<tt>llvm.readcyclecounter</tt>' Intrinsic</a>
2827 <div class="doc_text">
2831 declare ulong %llvm.readcyclecounter( )
2838 The '<tt>llvm.readcyclecounter</tt>' intrinsic provides access to the cycle
2839 counter register (or similar low latency, high accuracy clocks) on those targets
2840 that support it. On X86, it should map to RDTSC. On Alpha, it should map to RPCC.
2841 As the backing counters overflow quickly (on the order of 9 seconds on alpha), this
2842 should only be used for small timings.
2848 When directly supported, reading the cycle counter should not modify any memory.
2849 Implementations are allowed to either return a application specific value or a
2850 system wide value. On backends without support, this is lowered to a constant 0.
2856 <!-- ======================================================================= -->
2857 <div class="doc_subsection">
2858 <a name="int_os">Operating System Intrinsics</a>
2861 <div class="doc_text">
2863 These intrinsics are provided by LLVM to support the implementation of
2864 operating system level code.
2869 <!-- _______________________________________________________________________ -->
2870 <div class="doc_subsubsection">
2871 <a name="i_readport">'<tt>llvm.readport</tt>' Intrinsic</a>
2874 <div class="doc_text">
2878 declare <integer type> %llvm.readport (<integer type> <address>)
2884 The '<tt>llvm.readport</tt>' intrinsic reads data from the specified hardware
2891 The argument to this intrinsic indicates the hardware I/O address from which
2892 to read the data. The address is in the hardware I/O address namespace (as
2893 opposed to being a memory location for memory mapped I/O).
2899 The '<tt>llvm.readport</tt>' intrinsic reads data from the hardware I/O port
2900 specified by <i>address</i> and returns the value. The address and return
2901 value must be integers, but the size is dependent upon the platform upon which
2902 the program is code generated. For example, on x86, the address must be an
2903 unsigned 16-bit value, and the return value must be 8, 16, or 32 bits.
2908 <!-- _______________________________________________________________________ -->
2909 <div class="doc_subsubsection">
2910 <a name="i_writeport">'<tt>llvm.writeport</tt>' Intrinsic</a>
2913 <div class="doc_text">
2917 call void (<integer type>, <integer type>)*
2918 %llvm.writeport (<integer type> <value>,
2919 <integer type> <address>)
2925 The '<tt>llvm.writeport</tt>' intrinsic writes data to the specified hardware
2932 The first argument is the value to write to the I/O port.
2936 The second argument indicates the hardware I/O address to which data should be
2937 written. The address is in the hardware I/O address namespace (as opposed to
2938 being a memory location for memory mapped I/O).
2944 The '<tt>llvm.writeport</tt>' intrinsic writes <i>value</i> to the I/O port
2945 specified by <i>address</i>. The address and value must be integers, but the
2946 size is dependent upon the platform upon which the program is code generated.
2947 For example, on x86, the address must be an unsigned 16-bit value, and the
2948 value written must be 8, 16, or 32 bits in length.
2953 <!-- _______________________________________________________________________ -->
2954 <div class="doc_subsubsection">
2955 <a name="i_readio">'<tt>llvm.readio</tt>' Intrinsic</a>
2958 <div class="doc_text">
2962 declare <result> %llvm.readio (<ty> * <pointer>)
2968 The '<tt>llvm.readio</tt>' intrinsic reads data from a memory mapped I/O
2975 The argument to this intrinsic is a pointer indicating the memory address from
2976 which to read the data. The data must be a
2977 <a href="#t_firstclass">first class</a> type.
2983 The '<tt>llvm.readio</tt>' intrinsic reads data from a memory mapped I/O
2984 location specified by <i>pointer</i> and returns the value. The argument must
2985 be a pointer, and the return value must be a
2986 <a href="#t_firstclass">first class</a> type. However, certain architectures
2987 may not support I/O on all first class types. For example, 32-bit processors
2988 may only support I/O on data types that are 32 bits or less.
2992 This intrinsic enforces an in-order memory model for llvm.readio and
2993 llvm.writeio calls on machines that use dynamic scheduling. Dynamically
2994 scheduled processors may execute loads and stores out of order, re-ordering at
2995 run time accesses to memory mapped I/O registers. Using these intrinsics
2996 ensures that accesses to memory mapped I/O registers occur in program order.
3001 <!-- _______________________________________________________________________ -->
3002 <div class="doc_subsubsection">
3003 <a name="i_writeio">'<tt>llvm.writeio</tt>' Intrinsic</a>
3006 <div class="doc_text">
3010 declare void %llvm.writeio (<ty1> <value>, <ty2> * <pointer>)
3016 The '<tt>llvm.writeio</tt>' intrinsic writes data to the specified memory
3023 The first argument is the value to write to the memory mapped I/O location.
3024 The second argument is a pointer indicating the memory address to which the
3025 data should be written.
3031 The '<tt>llvm.writeio</tt>' intrinsic writes <i>value</i> to the memory mapped
3032 I/O address specified by <i>pointer</i>. The value must be a
3033 <a href="#t_firstclass">first class</a> type. However, certain architectures
3034 may not support I/O on all first class types. For example, 32-bit processors
3035 may only support I/O on data types that are 32 bits or less.
3039 This intrinsic enforces an in-order memory model for llvm.readio and
3040 llvm.writeio calls on machines that use dynamic scheduling. Dynamically
3041 scheduled processors may execute loads and stores out of order, re-ordering at
3042 run time accesses to memory mapped I/O registers. Using these intrinsics
3043 ensures that accesses to memory mapped I/O registers occur in program order.
3048 <!-- ======================================================================= -->
3049 <div class="doc_subsection">
3050 <a name="int_libc">Standard C Library Intrinsics</a>
3053 <div class="doc_text">
3055 LLVM provides intrinsics for a few important standard C library functions.
3056 These intrinsics allow source-language front-ends to pass information about the
3057 alignment of the pointer arguments to the code generator, providing opportunity
3058 for more efficient code generation.
3063 <!-- _______________________________________________________________________ -->
3064 <div class="doc_subsubsection">
3065 <a name="i_memcpy">'<tt>llvm.memcpy</tt>' Intrinsic</a>
3068 <div class="doc_text">
3072 declare void %llvm.memcpy(sbyte* <dest>, sbyte* <src>,
3073 uint <len>, uint <align>)
3079 The '<tt>llvm.memcpy</tt>' intrinsic copies a block of memory from the source
3080 location to the destination location.
3084 Note that, unlike the standard libc function, the <tt>llvm.memcpy</tt> intrinsic
3085 does not return a value, and takes an extra alignment argument.
3091 The first argument is a pointer to the destination, the second is a pointer to
3092 the source. The third argument is an (arbitrarily sized) integer argument
3093 specifying the number of bytes to copy, and the fourth argument is the alignment
3094 of the source and destination locations.
3098 If the call to this intrinisic has an alignment value that is not 0 or 1, then
3099 the caller guarantees that the size of the copy is a multiple of the alignment
3100 and that both the source and destination pointers are aligned to that boundary.
3106 The '<tt>llvm.memcpy</tt>' intrinsic copies a block of memory from the source
3107 location to the destination location, which are not allowed to overlap. It
3108 copies "len" bytes of memory over. If the argument is known to be aligned to
3109 some boundary, this can be specified as the fourth argument, otherwise it should
3115 <!-- _______________________________________________________________________ -->
3116 <div class="doc_subsubsection">
3117 <a name="i_memmove">'<tt>llvm.memmove</tt>' Intrinsic</a>
3120 <div class="doc_text">
3124 declare void %llvm.memmove(sbyte* <dest>, sbyte* <src>,
3125 uint <len>, uint <align>)
3131 The '<tt>llvm.memmove</tt>' intrinsic moves a block of memory from the source
3132 location to the destination location. It is similar to the '<tt>llvm.memcpy</tt>'
3133 intrinsic but allows the two memory locations to overlap.
3137 Note that, unlike the standard libc function, the <tt>llvm.memmove</tt> intrinsic
3138 does not return a value, and takes an extra alignment argument.
3144 The first argument is a pointer to the destination, the second is a pointer to
3145 the source. The third argument is an (arbitrarily sized) integer argument
3146 specifying the number of bytes to copy, and the fourth argument is the alignment
3147 of the source and destination locations.
3151 If the call to this intrinisic has an alignment value that is not 0 or 1, then
3152 the caller guarantees that the size of the copy is a multiple of the alignment
3153 and that both the source and destination pointers are aligned to that boundary.
3159 The '<tt>llvm.memmove</tt>' intrinsic copies a block of memory from the source
3160 location to the destination location, which may overlap. It
3161 copies "len" bytes of memory over. If the argument is known to be aligned to
3162 some boundary, this can be specified as the fourth argument, otherwise it should
3168 <!-- _______________________________________________________________________ -->
3169 <div class="doc_subsubsection">
3170 <a name="i_memset">'<tt>llvm.memset</tt>' Intrinsic</a>
3173 <div class="doc_text">
3177 declare void %llvm.memset(sbyte* <dest>, ubyte <val>,
3178 uint <len>, uint <align>)
3184 The '<tt>llvm.memset</tt>' intrinsic fills a block of memory with a particular
3189 Note that, unlike the standard libc function, the <tt>llvm.memset</tt> intrinsic
3190 does not return a value, and takes an extra alignment argument.
3196 The first argument is a pointer to the destination to fill, the second is the
3197 byte value to fill it with, the third argument is an (arbitrarily sized) integer
3198 argument specifying the number of bytes to fill, and the fourth argument is the
3199 known alignment of destination location.
3203 If the call to this intrinisic has an alignment value that is not 0 or 1, then
3204 the caller guarantees that the size of the copy is a multiple of the alignment
3205 and that the destination pointer is aligned to that boundary.
3211 The '<tt>llvm.memset</tt>' intrinsic fills "len" bytes of memory starting at the
3212 destination location. If the argument is known to be aligned to some boundary,
3213 this can be specified as the fourth argument, otherwise it should be set to 0 or
3219 <!-- _______________________________________________________________________ -->
3220 <div class="doc_subsubsection">
3221 <a name="i_isunordered">'<tt>llvm.isunordered</tt>' Intrinsic</a>
3224 <div class="doc_text">
3228 declare bool %llvm.isunordered(<float or double> Val1, <float or double> Val2)
3234 The '<tt>llvm.isunordered</tt>' intrinsic returns true if either or both of the
3235 specified floating point values is a NAN.
3241 The arguments are floating point numbers of the same type.
3247 If either or both of the arguments is a SNAN or QNAN, it returns true, otherwise
3253 <!-- _______________________________________________________________________ -->
3254 <div class="doc_subsubsection">
3255 <a name="i_sqrt">'<tt>llvm.sqrt</tt>' Intrinsic</a>
3258 <div class="doc_text">
3262 declare <float or double> %llvm.sqrt(<float or double> Val)
3268 The '<tt>llvm.sqrt</tt>' intrinsic returns the sqrt of the specified operand,
3269 returning the same value as the libm '<tt>sqrt</tt>' function would. Unlike
3270 <tt>sqrt</tt> in libm, however, <tt>llvm.sqrt</tt> has undefined behavior for
3271 negative numbers (which allows for better optimization).
3277 The argument and return value are floating point numbers of the same type.
3283 This function returns the sqrt of the specified operand if it is a positive
3284 floating point number.
3288 <!-- ======================================================================= -->
3289 <div class="doc_subsection">
3290 <a name="int_count">Bit Counting Intrinsics</a>
3293 <div class="doc_text">
3295 LLVM provides intrinsics for a few important bit counting operations.
3296 These allow efficient code generation for some algorithms.
3301 <!-- _______________________________________________________________________ -->
3302 <div class="doc_subsubsection">
3303 <a name="int_ctpop">'<tt>llvm.ctpop</tt>' Intrinsic</a>
3306 <div class="doc_text">
3310 declare int %llvm.ctpop(int <src>)
3317 The '<tt>llvm.ctpop</tt>' intrinsic counts the number of ones in a variable.
3323 The only argument is the value to be counted. The argument may be of any
3324 integer type. The return type must match the argument type.
3330 The '<tt>llvm.ctpop</tt>' intrinsic counts the 1's in a variable.
3334 <!-- _______________________________________________________________________ -->
3335 <div class="doc_subsubsection">
3336 <a name="int_ctlz">'<tt>llvm.ctlz</tt>' Intrinsic</a>
3339 <div class="doc_text">
3343 declare int %llvm.ctlz(int <src>)
3350 The '<tt>llvm.ctlz</tt>' intrinsic counts the number of leading zeros in a
3357 The only argument is the value to be counted. The argument may be of any
3358 integer type. The return type must match the argument type.
3364 The '<tt>llvm.ctlz</tt>' intrinsic counts the leading (most significant) zeros
3365 in a variable. If the src == 0 then the result is the size in bits of the type
3366 of src. For example, <tt>llvm.cttz(int 2) = 30</tt>.
3372 <!-- _______________________________________________________________________ -->
3373 <div class="doc_subsubsection">
3374 <a name="int_cttz">'<tt>llvm.cttz</tt>' Intrinsic</a>
3377 <div class="doc_text">
3381 declare int %llvm.cttz(int <src>)
3388 The '<tt>llvm.cttz</tt>' intrinsic counts the number of trailing zeros.
3394 The only argument is the value to be counted. The argument may be of any
3395 integer type. The return type must match the argument type.
3401 The '<tt>llvm.cttz</tt>' intrinsic counts the trailing (least significant) zeros
3402 in a variable. If the src == 0 then the result is the size in bits of the type
3403 of src. For example, <tt>llvm.cttz(2) = 1</tt>.
3407 <!-- ======================================================================= -->
3408 <div class="doc_subsection">
3409 <a name="int_debugger">Debugger Intrinsics</a>
3412 <div class="doc_text">
3414 The LLVM debugger intrinsics (which all start with <tt>llvm.dbg.</tt> prefix),
3415 are described in the <a
3416 href="SourceLevelDebugging.html#format_common_intrinsics">LLVM Source Level
3417 Debugging</a> document.
3422 <!-- *********************************************************************** -->
3425 <a href="http://jigsaw.w3.org/css-validator/check/referer"><img
3426 src="http://jigsaw.w3.org/css-validator/images/vcss" alt="Valid CSS!"></a>
3427 <a href="http://validator.w3.org/check/referer"><img
3428 src="http://www.w3.org/Icons/valid-html401" alt="Valid HTML 4.01!" /></a>
3430 <a href="mailto:sabre@nondot.org">Chris Lattner</a><br>
3431 <a href="http://llvm.cs.uiuc.edu">The LLVM Compiler Infrastructure</a><br>
3432 Last modified: $Date$