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">Functions</a></li>
27 <li><a href="#moduleasm">Module-Level Inline Assembly</a></li>
30 <li><a href="#typesystem">Type System</a>
32 <li><a href="#t_primitive">Primitive Types</a>
34 <li><a href="#t_classifications">Type Classifications</a></li>
37 <li><a href="#t_derived">Derived Types</a>
39 <li><a href="#t_array">Array Type</a></li>
40 <li><a href="#t_function">Function Type</a></li>
41 <li><a href="#t_pointer">Pointer Type</a></li>
42 <li><a href="#t_struct">Structure Type</a></li>
43 <li><a href="#t_packed">Packed Type</a></li>
44 <li><a href="#t_opaque">Opaque Type</a></li>
49 <li><a href="#constants">Constants</a>
51 <li><a href="#simpleconstants">Simple Constants</a>
52 <li><a href="#aggregateconstants">Aggregate Constants</a>
53 <li><a href="#globalconstants">Global Variable and Function Addresses</a>
54 <li><a href="#undefvalues">Undefined Values</a>
55 <li><a href="#constantexprs">Constant Expressions</a>
58 <li><a href="#othervalues">Other Values</a>
60 <li><a href="#inlineasm">Inline Assembler Expressions</a>
63 <li><a href="#instref">Instruction Reference</a>
65 <li><a href="#terminators">Terminator Instructions</a>
67 <li><a href="#i_ret">'<tt>ret</tt>' Instruction</a></li>
68 <li><a href="#i_br">'<tt>br</tt>' Instruction</a></li>
69 <li><a href="#i_switch">'<tt>switch</tt>' Instruction</a></li>
70 <li><a href="#i_invoke">'<tt>invoke</tt>' Instruction</a></li>
71 <li><a href="#i_unwind">'<tt>unwind</tt>' Instruction</a></li>
72 <li><a href="#i_unreachable">'<tt>unreachable</tt>' Instruction</a></li>
75 <li><a href="#binaryops">Binary Operations</a>
77 <li><a href="#i_add">'<tt>add</tt>' Instruction</a></li>
78 <li><a href="#i_sub">'<tt>sub</tt>' Instruction</a></li>
79 <li><a href="#i_mul">'<tt>mul</tt>' Instruction</a></li>
80 <li><a href="#i_udiv">'<tt>udiv</tt>' Instruction</a></li>
81 <li><a href="#i_sdiv">'<tt>sdiv</tt>' Instruction</a></li>
82 <li><a href="#i_fdiv">'<tt>fdiv</tt>' Instruction</a></li>
83 <li><a href="#i_urem">'<tt>urem</tt>' Instruction</a></li>
84 <li><a href="#i_srem">'<tt>srem</tt>' Instruction</a></li>
85 <li><a href="#i_frem">'<tt>frem</tt>' Instruction</a></li>
86 <li><a href="#i_setcc">'<tt>set<i>cc</i></tt>' Instructions</a></li>
89 <li><a href="#bitwiseops">Bitwise Binary Operations</a>
91 <li><a href="#i_and">'<tt>and</tt>' Instruction</a></li>
92 <li><a href="#i_or">'<tt>or</tt>' Instruction</a></li>
93 <li><a href="#i_xor">'<tt>xor</tt>' Instruction</a></li>
94 <li><a href="#i_shl">'<tt>shl</tt>' Instruction</a></li>
95 <li><a href="#i_lshr">'<tt>lshr</tt>' Instruction</a></li>
96 <li><a href="#i_ashr">'<tt>ashr</tt>' Instruction</a></li>
99 <li><a href="#vectorops">Vector Operations</a>
101 <li><a href="#i_extractelement">'<tt>extractelement</tt>' Instruction</a></li>
102 <li><a href="#i_insertelement">'<tt>insertelement</tt>' Instruction</a></li>
103 <li><a href="#i_shufflevector">'<tt>shufflevector</tt>' Instruction</a></li>
104 <li><a href="#i_vsetint">'<tt>vsetint</tt>' Instruction</a></li>
105 <li><a href="#i_vsetfp">'<tt>vsetfp</tt>' Instruction</a></li>
106 <li><a href="#i_vselect">'<tt>vselect</tt>' Instruction</a></li>
109 <li><a href="#memoryops">Memory Access and Addressing Operations</a>
111 <li><a href="#i_malloc">'<tt>malloc</tt>' Instruction</a></li>
112 <li><a href="#i_free">'<tt>free</tt>' Instruction</a></li>
113 <li><a href="#i_alloca">'<tt>alloca</tt>' Instruction</a></li>
114 <li><a href="#i_load">'<tt>load</tt>' Instruction</a></li>
115 <li><a href="#i_store">'<tt>store</tt>' Instruction</a></li>
116 <li><a href="#i_getelementptr">'<tt>getelementptr</tt>' Instruction</a></li>
119 <li><a href="#convertops">Conversion Operations</a>
121 <li><a href="#i_trunc">'<tt>trunc .. to</tt>' Instruction</a></li>
122 <li><a href="#i_zext">'<tt>zext .. to</tt>' Instruction</a></li>
123 <li><a href="#i_sext">'<tt>sext .. to</tt>' Instruction</a></li>
124 <li><a href="#i_fptrunc">'<tt>fptrunc .. to</tt>' Instruction</a></li>
125 <li><a href="#i_fpext">'<tt>fpext .. to</tt>' Instruction</a></li>
126 <li><a href="#i_fp2uint">'<tt>fp2uint .. to</tt>' Instruction</a></li>
127 <li><a href="#i_fp2sint">'<tt>fp2sint .. to</tt>' Instruction</a></li>
128 <li><a href="#i_uint2fp">'<tt>uint2fp .. to</tt>' Instruction</a></li>
129 <li><a href="#i_sint2fp">'<tt>sint2fp .. to</tt>' Instruction</a></li>
130 <li><a href="#i_bitconvert">'<tt>bitconvert .. to</tt>' Instruction</a></li>
132 <li><a href="#otherops">Other Operations</a>
134 <li><a href="#i_phi">'<tt>phi</tt>' Instruction</a></li>
135 <li><a href="#i_select">'<tt>select</tt>' Instruction</a></li>
136 <li><a href="#i_call">'<tt>call</tt>' Instruction</a></li>
137 <li><a href="#i_va_arg">'<tt>va_arg</tt>' Instruction</a></li>
142 <li><a href="#intrinsics">Intrinsic Functions</a>
144 <li><a href="#int_varargs">Variable Argument Handling Intrinsics</a>
146 <li><a href="#i_va_start">'<tt>llvm.va_start</tt>' Intrinsic</a></li>
147 <li><a href="#i_va_end">'<tt>llvm.va_end</tt>' Intrinsic</a></li>
148 <li><a href="#i_va_copy">'<tt>llvm.va_copy</tt>' Intrinsic</a></li>
151 <li><a href="#int_gc">Accurate Garbage Collection Intrinsics</a>
153 <li><a href="#i_gcroot">'<tt>llvm.gcroot</tt>' Intrinsic</a></li>
154 <li><a href="#i_gcread">'<tt>llvm.gcread</tt>' Intrinsic</a></li>
155 <li><a href="#i_gcwrite">'<tt>llvm.gcwrite</tt>' Intrinsic</a></li>
158 <li><a href="#int_codegen">Code Generator Intrinsics</a>
160 <li><a href="#i_returnaddress">'<tt>llvm.returnaddress</tt>' Intrinsic</a></li>
161 <li><a href="#i_frameaddress">'<tt>llvm.frameaddress</tt>' Intrinsic</a></li>
162 <li><a href="#i_stacksave">'<tt>llvm.stacksave</tt>' Intrinsic</a></li>
163 <li><a href="#i_stackrestore">'<tt>llvm.stackrestore</tt>' Intrinsic</a></li>
164 <li><a href="#i_prefetch">'<tt>llvm.prefetch</tt>' Intrinsic</a></li>
165 <li><a href="#i_pcmarker">'<tt>llvm.pcmarker</tt>' Intrinsic</a></li>
166 <li><a href="#i_readcyclecounter"><tt>llvm.readcyclecounter</tt>' Intrinsic</a></li>
169 <li><a href="#int_libc">Standard C Library Intrinsics</a>
171 <li><a href="#i_memcpy">'<tt>llvm.memcpy.*</tt>' Intrinsic</a></li>
172 <li><a href="#i_memmove">'<tt>llvm.memmove.*</tt>' Intrinsic</a></li>
173 <li><a href="#i_memset">'<tt>llvm.memset.*</tt>' Intrinsic</a></li>
174 <li><a href="#i_isunordered">'<tt>llvm.isunordered.*</tt>' Intrinsic</a></li>
175 <li><a href="#i_sqrt">'<tt>llvm.sqrt.*</tt>' Intrinsic</a></li>
176 <li><a href="#i_powi">'<tt>llvm.powi.*</tt>' Intrinsic</a></li>
179 <li><a href="#int_manip">Bit Manipulation Intrinsics</a>
181 <li><a href="#i_bswap">'<tt>llvm.bswap.*</tt>' Intrinsics</a></li>
182 <li><a href="#int_ctpop">'<tt>llvm.ctpop.*</tt>' Intrinsic </a></li>
183 <li><a href="#int_ctlz">'<tt>llvm.ctlz.*</tt>' Intrinsic </a></li>
184 <li><a href="#int_cttz">'<tt>llvm.cttz.*</tt>' Intrinsic </a></li>
187 <li><a href="#int_debugger">Debugger intrinsics</a></li>
192 <div class="doc_author">
193 <p>Written by <a href="mailto:sabre@nondot.org">Chris Lattner</a>
194 and <a href="mailto:vadve@cs.uiuc.edu">Vikram Adve</a></p>
197 <!-- *********************************************************************** -->
198 <div class="doc_section"> <a name="abstract">Abstract </a></div>
199 <!-- *********************************************************************** -->
201 <div class="doc_text">
202 <p>This document is a reference manual for the LLVM assembly language.
203 LLVM is an SSA based representation that provides type safety,
204 low-level operations, flexibility, and the capability of representing
205 'all' high-level languages cleanly. It is the common code
206 representation used throughout all phases of the LLVM compilation
210 <!-- *********************************************************************** -->
211 <div class="doc_section"> <a name="introduction">Introduction</a> </div>
212 <!-- *********************************************************************** -->
214 <div class="doc_text">
216 <p>The LLVM code representation is designed to be used in three
217 different forms: as an in-memory compiler IR, as an on-disk bytecode
218 representation (suitable for fast loading by a Just-In-Time compiler),
219 and as a human readable assembly language representation. This allows
220 LLVM to provide a powerful intermediate representation for efficient
221 compiler transformations and analysis, while providing a natural means
222 to debug and visualize the transformations. The three different forms
223 of LLVM are all equivalent. This document describes the human readable
224 representation and notation.</p>
226 <p>The LLVM representation aims to be light-weight and low-level
227 while being expressive, typed, and extensible at the same time. It
228 aims to be a "universal IR" of sorts, by being at a low enough level
229 that high-level ideas may be cleanly mapped to it (similar to how
230 microprocessors are "universal IR's", allowing many source languages to
231 be mapped to them). By providing type information, LLVM can be used as
232 the target of optimizations: for example, through pointer analysis, it
233 can be proven that a C automatic variable is never accessed outside of
234 the current function... allowing it to be promoted to a simple SSA
235 value instead of a memory location.</p>
239 <!-- _______________________________________________________________________ -->
240 <div class="doc_subsubsection"> <a name="wellformed">Well-Formedness</a> </div>
242 <div class="doc_text">
244 <p>It is important to note that this document describes 'well formed'
245 LLVM assembly language. There is a difference between what the parser
246 accepts and what is considered 'well formed'. For example, the
247 following instruction is syntactically okay, but not well formed:</p>
250 %x = <a href="#i_add">add</a> int 1, %x
253 <p>...because the definition of <tt>%x</tt> does not dominate all of
254 its uses. The LLVM infrastructure provides a verification pass that may
255 be used to verify that an LLVM module is well formed. This pass is
256 automatically run by the parser after parsing input assembly and by
257 the optimizer before it outputs bytecode. The violations pointed out
258 by the verifier pass indicate bugs in transformation passes or input to
261 <!-- Describe the typesetting conventions here. --> </div>
263 <!-- *********************************************************************** -->
264 <div class="doc_section"> <a name="identifiers">Identifiers</a> </div>
265 <!-- *********************************************************************** -->
267 <div class="doc_text">
269 <p>LLVM uses three different forms of identifiers, for different
273 <li>Named values are represented as a string of characters with a '%' prefix.
274 For example, %foo, %DivisionByZero, %a.really.long.identifier. The actual
275 regular expression used is '<tt>%[a-zA-Z$._][a-zA-Z$._0-9]*</tt>'.
276 Identifiers which require other characters in their names can be surrounded
277 with quotes. In this way, anything except a <tt>"</tt> character can be used
280 <li>Unnamed values are represented as an unsigned numeric value with a '%'
281 prefix. For example, %12, %2, %44.</li>
283 <li>Constants, which are described in a <a href="#constants">section about
284 constants</a>, below.</li>
287 <p>LLVM requires that values start with a '%' sign for two reasons: Compilers
288 don't need to worry about name clashes with reserved words, and the set of
289 reserved words may be expanded in the future without penalty. Additionally,
290 unnamed identifiers allow a compiler to quickly come up with a temporary
291 variable without having to avoid symbol table conflicts.</p>
293 <p>Reserved words in LLVM are very similar to reserved words in other
294 languages. There are keywords for different opcodes ('<tt><a
295 href="#i_add">add</a></tt>', '<tt><a href="#i_cast">cast</a></tt>', '<tt><a
296 href="#i_ret">ret</a></tt>', etc...), for primitive type names ('<tt><a
297 href="#t_void">void</a></tt>', '<tt><a href="#t_uint">uint</a></tt>', etc...),
298 and others. These reserved words cannot conflict with variable names, because
299 none of them start with a '%' character.</p>
301 <p>Here is an example of LLVM code to multiply the integer variable
302 '<tt>%X</tt>' by 8:</p>
307 %result = <a href="#i_mul">mul</a> uint %X, 8
310 <p>After strength reduction:</p>
313 %result = <a href="#i_shl">shl</a> uint %X, ubyte 3
316 <p>And the hard way:</p>
319 <a href="#i_add">add</a> uint %X, %X <i>; yields {uint}:%0</i>
320 <a href="#i_add">add</a> uint %0, %0 <i>; yields {uint}:%1</i>
321 %result = <a href="#i_add">add</a> uint %1, %1
324 <p>This last way of multiplying <tt>%X</tt> by 8 illustrates several
325 important lexical features of LLVM:</p>
329 <li>Comments are delimited with a '<tt>;</tt>' and go until the end of
332 <li>Unnamed temporaries are created when the result of a computation is not
333 assigned to a named value.</li>
335 <li>Unnamed temporaries are numbered sequentially</li>
339 <p>...and it also shows a convention that we follow in this document. When
340 demonstrating instructions, we will follow an instruction with a comment that
341 defines the type and name of value produced. Comments are shown in italic
346 <!-- *********************************************************************** -->
347 <div class="doc_section"> <a name="highlevel">High Level Structure</a> </div>
348 <!-- *********************************************************************** -->
350 <!-- ======================================================================= -->
351 <div class="doc_subsection"> <a name="modulestructure">Module Structure</a>
354 <div class="doc_text">
356 <p>LLVM programs are composed of "Module"s, each of which is a
357 translation unit of the input programs. Each module consists of
358 functions, global variables, and symbol table entries. Modules may be
359 combined together with the LLVM linker, which merges function (and
360 global variable) definitions, resolves forward declarations, and merges
361 symbol table entries. Here is an example of the "hello world" module:</p>
363 <pre><i>; Declare the string constant as a global constant...</i>
364 <a href="#identifiers">%.LC0</a> = <a href="#linkage_internal">internal</a> <a
365 href="#globalvars">constant</a> <a href="#t_array">[13 x sbyte]</a> c"hello world\0A\00" <i>; [13 x sbyte]*</i>
367 <i>; External declaration of the puts function</i>
368 <a href="#functionstructure">declare</a> int %puts(sbyte*) <i>; int(sbyte*)* </i>
370 <i>; Global variable / Function body section separator</i>
373 <i>; Definition of main function</i>
374 int %main() { <i>; int()* </i>
375 <i>; Convert [13x sbyte]* to sbyte *...</i>
377 href="#i_getelementptr">getelementptr</a> [13 x sbyte]* %.LC0, long 0, long 0 <i>; sbyte*</i>
379 <i>; Call puts function to write out the string to stdout...</i>
381 href="#i_call">call</a> int %puts(sbyte* %cast210) <i>; int</i>
383 href="#i_ret">ret</a> int 0<br>}<br></pre>
385 <p>This example is made up of a <a href="#globalvars">global variable</a>
386 named "<tt>.LC0</tt>", an external declaration of the "<tt>puts</tt>"
387 function, and a <a href="#functionstructure">function definition</a>
388 for "<tt>main</tt>".</p>
390 <p>In general, a module is made up of a list of global values,
391 where both functions and global variables are global values. Global values are
392 represented by a pointer to a memory location (in this case, a pointer to an
393 array of char, and a pointer to a function), and have one of the following <a
394 href="#linkage">linkage types</a>.</p>
396 <p>Due to a limitation in the current LLVM assembly parser (it is limited by
397 one-token lookahead), modules are split into two pieces by the "implementation"
398 keyword. Global variable prototypes and definitions must occur before the
399 keyword, and function definitions must occur after it. Function prototypes may
400 occur either before or after it. In the future, the implementation keyword may
401 become a noop, if the parser gets smarter.</p>
405 <!-- ======================================================================= -->
406 <div class="doc_subsection">
407 <a name="linkage">Linkage Types</a>
410 <div class="doc_text">
413 All Global Variables and Functions have one of the following types of linkage:
418 <dt><tt><b><a name="linkage_internal">internal</a></b></tt> </dt>
420 <dd>Global values with internal linkage are only directly accessible by
421 objects in the current module. In particular, linking code into a module with
422 an internal global value may cause the internal to be renamed as necessary to
423 avoid collisions. Because the symbol is internal to the module, all
424 references can be updated. This corresponds to the notion of the
425 '<tt>static</tt>' keyword in C, or the idea of "anonymous namespaces" in C++.
428 <dt><tt><b><a name="linkage_linkonce">linkonce</a></b></tt>: </dt>
430 <dd>"<tt>linkonce</tt>" linkage is similar to <tt>internal</tt> linkage, with
431 the twist that linking together two modules defining the same
432 <tt>linkonce</tt> globals will cause one of the globals to be discarded. This
433 is typically used to implement inline functions. Unreferenced
434 <tt>linkonce</tt> globals are allowed to be discarded.
437 <dt><tt><b><a name="linkage_weak">weak</a></b></tt>: </dt>
439 <dd>"<tt>weak</tt>" linkage is exactly the same as <tt>linkonce</tt> linkage,
440 except that unreferenced <tt>weak</tt> globals may not be discarded. This is
441 used to implement constructs in C such as "<tt>int X;</tt>" at global scope.
444 <dt><tt><b><a name="linkage_appending">appending</a></b></tt>: </dt>
446 <dd>"<tt>appending</tt>" linkage may only be applied to global variables of
447 pointer to array type. When two global variables with appending linkage are
448 linked together, the two global arrays are appended together. This is the
449 LLVM, typesafe, equivalent of having the system linker append together
450 "sections" with identical names when .o files are linked.
453 <dt><tt><b><a name="linkage_external">externally visible</a></b></tt>:</dt>
455 <dd>If none of the above identifiers are used, the global is externally
456 visible, meaning that it participates in linkage and can be used to resolve
457 external symbol references.
460 <dt><tt><b><a name="linkage_externweak">extern_weak</a></b></tt>: </dt>
462 <dd>"<tt>extern_weak</tt>" TBD
466 The next two types of linkage are targeted for Microsoft Windows platform
467 only. They are designed to support importing (exporting) symbols from (to)
471 <dt><tt><b><a name="linkage_dllimport">dllimport</a></b></tt>: </dt>
473 <dd>"<tt>dllimport</tt>" linkage causes the compiler to reference a function
474 or variable via a global pointer to a pointer that is set up by the DLL
475 exporting the symbol. On Microsoft Windows targets, the pointer name is
476 formed by combining <code>_imp__</code> and the function or variable name.
479 <dt><tt><b><a name="linkage_dllexport">dllexport</a></b></tt>: </dt>
481 <dd>"<tt>dllexport</tt>" linkage causes the compiler to provide a global
482 pointer to a pointer in a DLL, so that it can be referenced with the
483 <tt>dllimport</tt> attribute. On Microsoft Windows targets, the pointer
484 name is formed by combining <code>_imp__</code> and the function or variable
490 <p><a name="linkage_external">For example, since the "<tt>.LC0</tt>"
491 variable is defined to be internal, if another module defined a "<tt>.LC0</tt>"
492 variable and was linked with this one, one of the two would be renamed,
493 preventing a collision. Since "<tt>main</tt>" and "<tt>puts</tt>" are
494 external (i.e., lacking any linkage declarations), they are accessible
495 outside of the current module. It is illegal for a function <i>declaration</i>
496 to have any linkage type other than "externally visible".</a></p>
500 <!-- ======================================================================= -->
501 <div class="doc_subsection">
502 <a name="callingconv">Calling Conventions</a>
505 <div class="doc_text">
507 <p>LLVM <a href="#functionstructure">functions</a>, <a href="#i_call">calls</a>
508 and <a href="#i_invoke">invokes</a> can all have an optional calling convention
509 specified for the call. The calling convention of any pair of dynamic
510 caller/callee must match, or the behavior of the program is undefined. The
511 following calling conventions are supported by LLVM, and more may be added in
515 <dt><b>"<tt>ccc</tt>" - The C calling convention</b>:</dt>
517 <dd>This calling convention (the default if no other calling convention is
518 specified) matches the target C calling conventions. This calling convention
519 supports varargs function calls and tolerates some mismatch in the declared
520 prototype and implemented declaration of the function (as does normal C).
523 <dt><b>"<tt>csretcc</tt>" - The C struct return calling convention</b>:</dt>
525 <dd>This calling convention matches the target C calling conventions, except
526 that functions with this convention are required to take a pointer as their
527 first argument, and the return type of the function must be void. This is
528 used for C functions that return aggregates by-value. In this case, the
529 function has been transformed to take a pointer to the struct as the first
530 argument to the function. For targets where the ABI specifies specific
531 behavior for structure-return calls, the calling convention can be used to
532 distinguish between struct return functions and other functions that take a
533 pointer to a struct as the first argument.
536 <dt><b>"<tt>fastcc</tt>" - The fast calling convention</b>:</dt>
538 <dd>This calling convention attempts to make calls as fast as possible
539 (e.g. by passing things in registers). This calling convention allows the
540 target to use whatever tricks it wants to produce fast code for the target,
541 without having to conform to an externally specified ABI. Implementations of
542 this convention should allow arbitrary tail call optimization to be supported.
543 This calling convention does not support varargs and requires the prototype of
544 all callees to exactly match the prototype of the function definition.
547 <dt><b>"<tt>coldcc</tt>" - The cold calling convention</b>:</dt>
549 <dd>This calling convention attempts to make code in the caller as efficient
550 as possible under the assumption that the call is not commonly executed. As
551 such, these calls often preserve all registers so that the call does not break
552 any live ranges in the caller side. This calling convention does not support
553 varargs and requires the prototype of all callees to exactly match the
554 prototype of the function definition.
557 <dt><b>"<tt>cc <<em>n</em>></tt>" - Numbered convention</b>:</dt>
559 <dd>Any calling convention may be specified by number, allowing
560 target-specific calling conventions to be used. Target specific calling
561 conventions start at 64.
565 <p>More calling conventions can be added/defined on an as-needed basis, to
566 support pascal conventions or any other well-known target-independent
571 <!-- ======================================================================= -->
572 <div class="doc_subsection">
573 <a name="globalvars">Global Variables</a>
576 <div class="doc_text">
578 <p>Global variables define regions of memory allocated at compilation time
579 instead of run-time. Global variables may optionally be initialized, may have
580 an explicit section to be placed in, and may
581 have an optional explicit alignment specified. A
582 variable may be defined as a global "constant," which indicates that the
583 contents of the variable will <b>never</b> be modified (enabling better
584 optimization, allowing the global data to be placed in the read-only section of
585 an executable, etc). Note that variables that need runtime initialization
586 cannot be marked "constant" as there is a store to the variable.</p>
589 LLVM explicitly allows <em>declarations</em> of global variables to be marked
590 constant, even if the final definition of the global is not. This capability
591 can be used to enable slightly better optimization of the program, but requires
592 the language definition to guarantee that optimizations based on the
593 'constantness' are valid for the translation units that do not include the
597 <p>As SSA values, global variables define pointer values that are in
598 scope (i.e. they dominate) all basic blocks in the program. Global
599 variables always define a pointer to their "content" type because they
600 describe a region of memory, and all memory objects in LLVM are
601 accessed through pointers.</p>
603 <p>LLVM allows an explicit section to be specified for globals. If the target
604 supports it, it will emit globals to the section specified.</p>
606 <p>An explicit alignment may be specified for a global. If not present, or if
607 the alignment is set to zero, the alignment of the global is set by the target
608 to whatever it feels convenient. If an explicit alignment is specified, the
609 global is forced to have at least that much alignment. All alignments must be
615 <!-- ======================================================================= -->
616 <div class="doc_subsection">
617 <a name="functionstructure">Functions</a>
620 <div class="doc_text">
622 <p>LLVM function definitions consist of an optional <a href="#linkage">linkage
623 type</a>, an optional <a href="#callingconv">calling convention</a>, a return
624 type, a function name, a (possibly empty) argument list, an optional section,
625 an optional alignment, an opening curly brace,
626 a list of basic blocks, and a closing curly brace. LLVM function declarations
627 are defined with the "<tt>declare</tt>" keyword, an optional <a
628 href="#callingconv">calling convention</a>, a return type, a function name,
629 a possibly empty list of arguments, and an optional alignment.</p>
631 <p>A function definition contains a list of basic blocks, forming the CFG for
632 the function. Each basic block may optionally start with a label (giving the
633 basic block a symbol table entry), contains a list of instructions, and ends
634 with a <a href="#terminators">terminator</a> instruction (such as a branch or
635 function return).</p>
637 <p>The first basic block in a program is special in two ways: it is immediately
638 executed on entrance to the function, and it is not allowed to have predecessor
639 basic blocks (i.e. there can not be any branches to the entry block of a
640 function). Because the block can have no predecessors, it also cannot have any
641 <a href="#i_phi">PHI nodes</a>.</p>
643 <p>LLVM functions are identified by their name and type signature. Hence, two
644 functions with the same name but different parameter lists or return values are
645 considered different functions, and LLVM will resolve references to each
648 <p>LLVM allows an explicit section to be specified for functions. If the target
649 supports it, it will emit functions to the section specified.</p>
651 <p>An explicit alignment may be specified for a function. If not present, or if
652 the alignment is set to zero, the alignment of the function is set by the target
653 to whatever it feels convenient. If an explicit alignment is specified, the
654 function is forced to have at least that much alignment. All alignments must be
659 <!-- ======================================================================= -->
660 <div class="doc_subsection">
661 <a name="moduleasm">Module-Level Inline Assembly</a>
664 <div class="doc_text">
666 Modules may contain "module-level inline asm" blocks, which corresponds to the
667 GCC "file scope inline asm" blocks. These blocks are internally concatenated by
668 LLVM and treated as a single unit, but may be separated in the .ll file if
669 desired. The syntax is very simple:
672 <div class="doc_code"><pre>
673 module asm "inline asm code goes here"
674 module asm "more can go here"
677 <p>The strings can contain any character by escaping non-printable characters.
678 The escape sequence used is simply "\xx" where "xx" is the two digit hex code
683 The inline asm code is simply printed to the machine code .s file when
684 assembly code is generated.
689 <!-- *********************************************************************** -->
690 <div class="doc_section"> <a name="typesystem">Type System</a> </div>
691 <!-- *********************************************************************** -->
693 <div class="doc_text">
695 <p>The LLVM type system is one of the most important features of the
696 intermediate representation. Being typed enables a number of
697 optimizations to be performed on the IR directly, without having to do
698 extra analyses on the side before the transformation. A strong type
699 system makes it easier to read the generated code and enables novel
700 analyses and transformations that are not feasible to perform on normal
701 three address code representations.</p>
705 <!-- ======================================================================= -->
706 <div class="doc_subsection"> <a name="t_primitive">Primitive Types</a> </div>
707 <div class="doc_text">
708 <p>The primitive types are the fundamental building blocks of the LLVM
709 system. The current set of primitive types is as follows:</p>
711 <table class="layout">
716 <tr><th>Type</th><th>Description</th></tr>
717 <tr><td><tt>void</tt></td><td>No value</td></tr>
718 <tr><td><tt>ubyte</tt></td><td>Unsigned 8-bit value</td></tr>
719 <tr><td><tt>ushort</tt></td><td>Unsigned 16-bit value</td></tr>
720 <tr><td><tt>uint</tt></td><td>Unsigned 32-bit value</td></tr>
721 <tr><td><tt>ulong</tt></td><td>Unsigned 64-bit value</td></tr>
722 <tr><td><tt>float</tt></td><td>32-bit floating point value</td></tr>
723 <tr><td><tt>label</tt></td><td>Branch destination</td></tr>
730 <tr><th>Type</th><th>Description</th></tr>
731 <tr><td><tt>bool</tt></td><td>True or False value</td></tr>
732 <tr><td><tt>sbyte</tt></td><td>Signed 8-bit value</td></tr>
733 <tr><td><tt>short</tt></td><td>Signed 16-bit value</td></tr>
734 <tr><td><tt>int</tt></td><td>Signed 32-bit value</td></tr>
735 <tr><td><tt>long</tt></td><td>Signed 64-bit value</td></tr>
736 <tr><td><tt>double</tt></td><td>64-bit floating point value</td></tr>
744 <!-- _______________________________________________________________________ -->
745 <div class="doc_subsubsection"> <a name="t_classifications">Type
746 Classifications</a> </div>
747 <div class="doc_text">
748 <p>These different primitive types fall into a few useful
751 <table border="1" cellspacing="0" cellpadding="4">
753 <tr><th>Classification</th><th>Types</th></tr>
755 <td><a name="t_signed">signed</a></td>
756 <td><tt>sbyte, short, int, long, float, double</tt></td>
759 <td><a name="t_unsigned">unsigned</a></td>
760 <td><tt>ubyte, ushort, uint, ulong</tt></td>
763 <td><a name="t_integer">integer</a></td>
764 <td><tt>ubyte, sbyte, ushort, short, uint, int, ulong, long</tt></td>
767 <td><a name="t_integral">integral</a></td>
768 <td><tt>bool, ubyte, sbyte, ushort, short, uint, int, ulong, long</tt>
772 <td><a name="t_floating">floating point</a></td>
773 <td><tt>float, double</tt></td>
776 <td><a name="t_firstclass">first class</a></td>
777 <td><tt>bool, ubyte, sbyte, ushort, short, uint, int, ulong, long,<br>
778 float, double, <a href="#t_pointer">pointer</a>,
779 <a href="#t_packed">packed</a></tt></td>
784 <p>The <a href="#t_firstclass">first class</a> types are perhaps the
785 most important. Values of these types are the only ones which can be
786 produced by instructions, passed as arguments, or used as operands to
787 instructions. This means that all structures and arrays must be
788 manipulated either by pointer or by component.</p>
791 <!-- ======================================================================= -->
792 <div class="doc_subsection"> <a name="t_derived">Derived Types</a> </div>
794 <div class="doc_text">
796 <p>The real power in LLVM comes from the derived types in the system.
797 This is what allows a programmer to represent arrays, functions,
798 pointers, and other useful types. Note that these derived types may be
799 recursive: For example, it is possible to have a two dimensional array.</p>
803 <!-- _______________________________________________________________________ -->
804 <div class="doc_subsubsection"> <a name="t_array">Array Type</a> </div>
806 <div class="doc_text">
810 <p>The array type is a very simple derived type that arranges elements
811 sequentially in memory. The array type requires a size (number of
812 elements) and an underlying data type.</p>
817 [<# elements> x <elementtype>]
820 <p>The number of elements is a constant integer value; elementtype may
821 be any type with a size.</p>
824 <table class="layout">
827 <tt>[40 x int ]</tt><br/>
828 <tt>[41 x int ]</tt><br/>
829 <tt>[40 x uint]</tt><br/>
832 Array of 40 integer values.<br/>
833 Array of 41 integer values.<br/>
834 Array of 40 unsigned integer values.<br/>
838 <p>Here are some examples of multidimensional arrays:</p>
839 <table class="layout">
842 <tt>[3 x [4 x int]]</tt><br/>
843 <tt>[12 x [10 x float]]</tt><br/>
844 <tt>[2 x [3 x [4 x uint]]]</tt><br/>
847 3x4 array of integer values.<br/>
848 12x10 array of single precision floating point values.<br/>
849 2x3x4 array of unsigned integer values.<br/>
854 <p>Note that 'variable sized arrays' can be implemented in LLVM with a zero
855 length array. Normally, accesses past the end of an array are undefined in
856 LLVM (e.g. it is illegal to access the 5th element of a 3 element array).
857 As a special case, however, zero length arrays are recognized to be variable
858 length. This allows implementation of 'pascal style arrays' with the LLVM
859 type "{ int, [0 x float]}", for example.</p>
863 <!-- _______________________________________________________________________ -->
864 <div class="doc_subsubsection"> <a name="t_function">Function Type</a> </div>
865 <div class="doc_text">
867 <p>The function type can be thought of as a function signature. It
868 consists of a return type and a list of formal parameter types.
869 Function types are usually used to build virtual function tables
870 (which are structures of pointers to functions), for indirect function
871 calls, and when defining a function.</p>
873 The return type of a function type cannot be an aggregate type.
876 <pre> <returntype> (<parameter list>)<br></pre>
877 <p>...where '<tt><parameter list></tt>' is a comma-separated list of type
878 specifiers. Optionally, the parameter list may include a type <tt>...</tt>,
879 which indicates that the function takes a variable number of arguments.
880 Variable argument functions can access their arguments with the <a
881 href="#int_varargs">variable argument handling intrinsic</a> functions.</p>
883 <table class="layout">
886 <tt>int (int)</tt> <br/>
887 <tt>float (int, int *) *</tt><br/>
888 <tt>int (sbyte *, ...)</tt><br/>
891 function taking an <tt>int</tt>, returning an <tt>int</tt><br/>
892 <a href="#t_pointer">Pointer</a> to a function that takes an
893 <tt>int</tt> and a <a href="#t_pointer">pointer</a> to <tt>int</tt>,
894 returning <tt>float</tt>.<br/>
895 A vararg function that takes at least one <a href="#t_pointer">pointer</a>
896 to <tt>sbyte</tt> (signed char in C), which returns an integer. This is
897 the signature for <tt>printf</tt> in LLVM.<br/>
903 <!-- _______________________________________________________________________ -->
904 <div class="doc_subsubsection"> <a name="t_struct">Structure Type</a> </div>
905 <div class="doc_text">
907 <p>The structure type is used to represent a collection of data members
908 together in memory. The packing of the field types is defined to match
909 the ABI of the underlying processor. The elements of a structure may
910 be any type that has a size.</p>
911 <p>Structures are accessed using '<tt><a href="#i_load">load</a></tt>
912 and '<tt><a href="#i_store">store</a></tt>' by getting a pointer to a
913 field with the '<tt><a href="#i_getelementptr">getelementptr</a></tt>'
916 <pre> { <type list> }<br></pre>
918 <table class="layout">
921 <tt>{ int, int, int }</tt><br/>
922 <tt>{ float, int (int) * }</tt><br/>
925 a triple of three <tt>int</tt> values<br/>
926 A pair, where the first element is a <tt>float</tt> and the second element
927 is a <a href="#t_pointer">pointer</a> to a <a href="#t_function">function</a>
928 that takes an <tt>int</tt>, returning an <tt>int</tt>.<br/>
934 <!-- _______________________________________________________________________ -->
935 <div class="doc_subsubsection"> <a name="t_pointer">Pointer Type</a> </div>
936 <div class="doc_text">
938 <p>As in many languages, the pointer type represents a pointer or
939 reference to another object, which must live in memory.</p>
941 <pre> <type> *<br></pre>
943 <table class="layout">
946 <tt>[4x int]*</tt><br/>
947 <tt>int (int *) *</tt><br/>
950 A <a href="#t_pointer">pointer</a> to <a href="#t_array">array</a> of
951 four <tt>int</tt> values<br/>
952 A <a href="#t_pointer">pointer</a> to a <a
953 href="#t_function">function</a> that takes an <tt>int*</tt>, returning an
960 <!-- _______________________________________________________________________ -->
961 <div class="doc_subsubsection"> <a name="t_packed">Packed Type</a> </div>
962 <div class="doc_text">
966 <p>A packed type is a simple derived type that represents a vector
967 of elements. Packed types are used when multiple primitive data
968 are operated in parallel using a single instruction (SIMD).
969 A packed type requires a size (number of
970 elements) and an underlying primitive data type. Vectors must have a power
971 of two length (1, 2, 4, 8, 16 ...). Packed types are
972 considered <a href="#t_firstclass">first class</a>.</p>
977 < <# elements> x <elementtype> >
980 <p>The number of elements is a constant integer value; elementtype may
981 be any integral or floating point type.</p>
985 <table class="layout">
988 <tt><4 x int></tt><br/>
989 <tt><8 x float></tt><br/>
990 <tt><2 x uint></tt><br/>
993 Packed vector of 4 integer values.<br/>
994 Packed vector of 8 floating-point values.<br/>
995 Packed vector of 2 unsigned integer values.<br/>
1001 <!-- _______________________________________________________________________ -->
1002 <div class="doc_subsubsection"> <a name="t_opaque">Opaque Type</a> </div>
1003 <div class="doc_text">
1007 <p>Opaque types are used to represent unknown types in the system. This
1008 corresponds (for example) to the C notion of a foward declared structure type.
1009 In LLVM, opaque types can eventually be resolved to any type (not just a
1010 structure type).</p>
1020 <table class="layout">
1026 An opaque type.<br/>
1033 <!-- *********************************************************************** -->
1034 <div class="doc_section"> <a name="constants">Constants</a> </div>
1035 <!-- *********************************************************************** -->
1037 <div class="doc_text">
1039 <p>LLVM has several different basic types of constants. This section describes
1040 them all and their syntax.</p>
1044 <!-- ======================================================================= -->
1045 <div class="doc_subsection"><a name="simpleconstants">Simple Constants</a></div>
1047 <div class="doc_text">
1050 <dt><b>Boolean constants</b></dt>
1052 <dd>The two strings '<tt>true</tt>' and '<tt>false</tt>' are both valid
1053 constants of the <tt><a href="#t_primitive">bool</a></tt> type.
1056 <dt><b>Integer constants</b></dt>
1058 <dd>Standard integers (such as '4') are constants of the <a
1059 href="#t_integer">integer</a> type. Negative numbers may be used with signed
1063 <dt><b>Floating point constants</b></dt>
1065 <dd>Floating point constants use standard decimal notation (e.g. 123.421),
1066 exponential notation (e.g. 1.23421e+2), or a more precise hexadecimal
1067 notation (see below). Floating point constants must have a <a
1068 href="#t_floating">floating point</a> type. </dd>
1070 <dt><b>Null pointer constants</b></dt>
1072 <dd>The identifier '<tt>null</tt>' is recognized as a null pointer constant
1073 and must be of <a href="#t_pointer">pointer type</a>.</dd>
1077 <p>The one non-intuitive notation for constants is the optional hexadecimal form
1078 of floating point constants. For example, the form '<tt>double
1079 0x432ff973cafa8000</tt>' is equivalent to (but harder to read than) '<tt>double
1080 4.5e+15</tt>'. The only time hexadecimal floating point constants are required
1081 (and the only time that they are generated by the disassembler) is when a
1082 floating point constant must be emitted but it cannot be represented as a
1083 decimal floating point number. For example, NaN's, infinities, and other
1084 special values are represented in their IEEE hexadecimal format so that
1085 assembly and disassembly do not cause any bits to change in the constants.</p>
1089 <!-- ======================================================================= -->
1090 <div class="doc_subsection"><a name="aggregateconstants">Aggregate Constants</a>
1093 <div class="doc_text">
1094 <p>Aggregate constants arise from aggregation of simple constants
1095 and smaller aggregate constants.</p>
1098 <dt><b>Structure constants</b></dt>
1100 <dd>Structure constants are represented with notation similar to structure
1101 type definitions (a comma separated list of elements, surrounded by braces
1102 (<tt>{}</tt>)). For example: "<tt>{ int 4, float 17.0, int* %G }</tt>",
1103 where "<tt>%G</tt>" is declared as "<tt>%G = external global int</tt>". Structure constants
1104 must have <a href="#t_struct">structure type</a>, and the number and
1105 types of elements must match those specified by the type.
1108 <dt><b>Array constants</b></dt>
1110 <dd>Array constants are represented with notation similar to array type
1111 definitions (a comma separated list of elements, surrounded by square brackets
1112 (<tt>[]</tt>)). For example: "<tt>[ int 42, int 11, int 74 ]</tt>". Array
1113 constants must have <a href="#t_array">array type</a>, and the number and
1114 types of elements must match those specified by the type.
1117 <dt><b>Packed constants</b></dt>
1119 <dd>Packed constants are represented with notation similar to packed type
1120 definitions (a comma separated list of elements, surrounded by
1121 less-than/greater-than's (<tt><></tt>)). For example: "<tt>< int 42,
1122 int 11, int 74, int 100 ></tt>". Packed constants must have <a
1123 href="#t_packed">packed type</a>, and the number and types of elements must
1124 match those specified by the type.
1127 <dt><b>Zero initialization</b></dt>
1129 <dd>The string '<tt>zeroinitializer</tt>' can be used to zero initialize a
1130 value to zero of <em>any</em> type, including scalar and aggregate types.
1131 This is often used to avoid having to print large zero initializers (e.g. for
1132 large arrays) and is always exactly equivalent to using explicit zero
1139 <!-- ======================================================================= -->
1140 <div class="doc_subsection">
1141 <a name="globalconstants">Global Variable and Function Addresses</a>
1144 <div class="doc_text">
1146 <p>The addresses of <a href="#globalvars">global variables</a> and <a
1147 href="#functionstructure">functions</a> are always implicitly valid (link-time)
1148 constants. These constants are explicitly referenced when the <a
1149 href="#identifiers">identifier for the global</a> is used and always have <a
1150 href="#t_pointer">pointer</a> type. For example, the following is a legal LLVM
1156 %Z = global [2 x int*] [ int* %X, int* %Y ]
1161 <!-- ======================================================================= -->
1162 <div class="doc_subsection"><a name="undefvalues">Undefined Values</a></div>
1163 <div class="doc_text">
1164 <p>The string '<tt>undef</tt>' is recognized as a type-less constant that has
1165 no specific value. Undefined values may be of any type and be used anywhere
1166 a constant is permitted.</p>
1168 <p>Undefined values indicate to the compiler that the program is well defined
1169 no matter what value is used, giving the compiler more freedom to optimize.
1173 <!-- ======================================================================= -->
1174 <div class="doc_subsection"><a name="constantexprs">Constant Expressions</a>
1177 <div class="doc_text">
1179 <p>Constant expressions are used to allow expressions involving other constants
1180 to be used as constants. Constant expressions may be of any <a
1181 href="#t_firstclass">first class</a> type and may involve any LLVM operation
1182 that does not have side effects (e.g. load and call are not supported). The
1183 following is the syntax for constant expressions:</p>
1186 <dt><b><tt>trunc ( CST to TYPE )</tt></b></dt>
1187 <dd>Truncate a constant to another type. The bit size of CST must be larger
1188 than the bit size of TYPE. Both types must be integral.</dd>
1190 <dt><b><tt>zext ( CST to TYPE )</tt></b></dt>
1191 <dd>Zero extend a constant to another type. The bit size of CST must be
1192 smaller or equal to the bit size of TYPE. Both types must be integral.</dd>
1194 <dt><b><tt>sext ( CST to TYPE )</tt></b></dt>
1195 <dd>Sign extend a constant to another type. The bit size of CST must be
1196 smaller or equal to the bit size of TYPE. Both types must be integral.</dd>
1198 <dt><b><tt>fptrunc ( CST to TYPE )</tt></b></dt>
1199 <dd>Truncate a floating point constant to another floating point type. The
1200 size of CST must be larger than the size of TYPE. Both types must be
1201 floating point.</dd>
1203 <dt><b><tt>fpext ( CST to TYPE )</tt></b></dt>
1204 <dd>Floating point extend a constant to another type. The size of CST must be
1205 smaller or equal to the size of TYPE. Both types must be floating point.</dd>
1207 <dt><b><tt>fp2uint ( CST to TYPE )</tt></b></dt>
1208 <dd>Convert a floating point constant to the corresponding unsigned integer
1209 constant. TYPE must be an integer type. CST must be floating point. If the
1210 value won't fit in the integer type, the results are undefined.</dd>
1212 <dt><b><tt>fp2sint ( CST to TYPE )</tt></b></dt>
1213 <dd>Convert a floating point constant to the corresponding signed integer
1214 constant. TYPE must be an integer type. CST must be floating point. If the
1215 value won't fit in the integer type, the results are undefined.</dd>
1217 <dt><b><tt>uint2fp ( CST to TYPE )</tt></b></dt>
1218 <dd>Convert an unsigned integer constant to the corresponding floating point
1219 constant. TYPE must be floating point. CST must be of integer type. If the
1220 value won't fit in the floating point type, the results are undefined.</dd>
1222 <dt><b><tt>sint2fp ( CST to TYPE )</tt></b></dt>
1223 <dd>Convert a signed integer constant to the corresponding floating point
1224 constant. TYPE must be floating point. CST must be of integer type. If the
1225 value won't fit in the floating point type, the results are undefined.</dd>
1227 <dt><b><tt>bitconvert ( CST to TYPE )</tt></b></dt>
1228 <dd>Convert a constant, CST, to another TYPE. The size of CST and TYPE must be
1229 identical (same number of bits). The conversion is done as if the CST value
1230 was stored to memory and read back as TYPE. In other words, no bits change
1231 with this operator, just the type. This can be used for conversion of pointer
1232 and packed types to any other type, as long as they have the same bit width.
1235 <dt><b><tt>getelementptr ( CSTPTR, IDX0, IDX1, ... )</tt></b></dt>
1237 <dd>Perform the <a href="#i_getelementptr">getelementptr operation</a> on
1238 constants. As with the <a href="#i_getelementptr">getelementptr</a>
1239 instruction, the index list may have zero or more indexes, which are required
1240 to make sense for the type of "CSTPTR".</dd>
1242 <dt><b><tt>select ( COND, VAL1, VAL2 )</tt></b></dt>
1244 <dd>Perform the <a href="#i_select">select operation</a> on
1247 <dt><b><tt>extractelement ( VAL, IDX )</tt></b></dt>
1249 <dd>Perform the <a href="#i_extractelement">extractelement
1250 operation</a> on constants.
1252 <dt><b><tt>insertelement ( VAL, ELT, IDX )</tt></b></dt>
1254 <dd>Perform the <a href="#i_insertelement">insertelement
1255 operation</a> on constants.
1258 <dt><b><tt>shufflevector ( VEC1, VEC2, IDXMASK )</tt></b></dt>
1260 <dd>Perform the <a href="#i_shufflevector">shufflevector
1261 operation</a> on constants.
1263 <dt><b><tt>OPCODE ( LHS, RHS )</tt></b></dt>
1265 <dd>Perform the specified operation of the LHS and RHS constants. OPCODE may
1266 be any of the <a href="#binaryops">binary</a> or <a href="#bitwiseops">bitwise
1267 binary</a> operations. The constraints on operands are the same as those for
1268 the corresponding instruction (e.g. no bitwise operations on floating point
1269 values are allowed).</dd>
1273 <!-- *********************************************************************** -->
1274 <div class="doc_section"> <a name="othervalues">Other Values</a> </div>
1275 <!-- *********************************************************************** -->
1277 <!-- ======================================================================= -->
1278 <div class="doc_subsection">
1279 <a name="inlineasm">Inline Assembler Expressions</a>
1282 <div class="doc_text">
1285 LLVM supports inline assembler expressions (as opposed to <a href="#moduleasm">
1286 Module-Level Inline Assembly</a>) through the use of a special value. This
1287 value represents the inline assembler as a string (containing the instructions
1288 to emit), a list of operand constraints (stored as a string), and a flag that
1289 indicates whether or not the inline asm expression has side effects. An example
1290 inline assembler expression is:
1294 int(int) asm "bswap $0", "=r,r"
1298 Inline assembler expressions may <b>only</b> be used as the callee operand of
1299 a <a href="#i_call"><tt>call</tt> instruction</a>. Thus, typically we have:
1303 %X = call int asm "<a href="#i_bswap">bswap</a> $0", "=r,r"(int %Y)
1307 Inline asms with side effects not visible in the constraint list must be marked
1308 as having side effects. This is done through the use of the
1309 '<tt>sideeffect</tt>' keyword, like so:
1313 call void asm sideeffect "eieio", ""()
1316 <p>TODO: The format of the asm and constraints string still need to be
1317 documented here. Constraints on what can be done (e.g. duplication, moving, etc
1318 need to be documented).
1323 <!-- *********************************************************************** -->
1324 <div class="doc_section"> <a name="instref">Instruction Reference</a> </div>
1325 <!-- *********************************************************************** -->
1327 <div class="doc_text">
1329 <p>The LLVM instruction set consists of several different
1330 classifications of instructions: <a href="#terminators">terminator
1331 instructions</a>, <a href="#binaryops">binary instructions</a>,
1332 <a href="#bitwiseops">bitwise binary instructions</a>, <a
1333 href="#memoryops">memory instructions</a>, and <a href="#otherops">other
1334 instructions</a>.</p>
1338 <!-- ======================================================================= -->
1339 <div class="doc_subsection"> <a name="terminators">Terminator
1340 Instructions</a> </div>
1342 <div class="doc_text">
1344 <p>As mentioned <a href="#functionstructure">previously</a>, every
1345 basic block in a program ends with a "Terminator" instruction, which
1346 indicates which block should be executed after the current block is
1347 finished. These terminator instructions typically yield a '<tt>void</tt>'
1348 value: they produce control flow, not values (the one exception being
1349 the '<a href="#i_invoke"><tt>invoke</tt></a>' instruction).</p>
1350 <p>There are six different terminator instructions: the '<a
1351 href="#i_ret"><tt>ret</tt></a>' instruction, the '<a href="#i_br"><tt>br</tt></a>'
1352 instruction, the '<a href="#i_switch"><tt>switch</tt></a>' instruction,
1353 the '<a href="#i_invoke"><tt>invoke</tt></a>' instruction, the '<a
1354 href="#i_unwind"><tt>unwind</tt></a>' instruction, and the '<a
1355 href="#i_unreachable"><tt>unreachable</tt></a>' instruction.</p>
1359 <!-- _______________________________________________________________________ -->
1360 <div class="doc_subsubsection"> <a name="i_ret">'<tt>ret</tt>'
1361 Instruction</a> </div>
1362 <div class="doc_text">
1364 <pre> ret <type> <value> <i>; Return a value from a non-void function</i>
1365 ret void <i>; Return from void function</i>
1368 <p>The '<tt>ret</tt>' instruction is used to return control flow (and a
1369 value) from a function back to the caller.</p>
1370 <p>There are two forms of the '<tt>ret</tt>' instruction: one that
1371 returns a value and then causes control flow, and one that just causes
1372 control flow to occur.</p>
1374 <p>The '<tt>ret</tt>' instruction may return any '<a
1375 href="#t_firstclass">first class</a>' type. Notice that a function is
1376 not <a href="#wellformed">well formed</a> if there exists a '<tt>ret</tt>'
1377 instruction inside of the function that returns a value that does not
1378 match the return type of the function.</p>
1380 <p>When the '<tt>ret</tt>' instruction is executed, control flow
1381 returns back to the calling function's context. If the caller is a "<a
1382 href="#i_call"><tt>call</tt></a>" instruction, execution continues at
1383 the instruction after the call. If the caller was an "<a
1384 href="#i_invoke"><tt>invoke</tt></a>" instruction, execution continues
1385 at the beginning of the "normal" destination block. If the instruction
1386 returns a value, that value shall set the call or invoke instruction's
1389 <pre> ret int 5 <i>; Return an integer value of 5</i>
1390 ret void <i>; Return from a void function</i>
1393 <!-- _______________________________________________________________________ -->
1394 <div class="doc_subsubsection"> <a name="i_br">'<tt>br</tt>' Instruction</a> </div>
1395 <div class="doc_text">
1397 <pre> br bool <cond>, label <iftrue>, label <iffalse><br> br label <dest> <i>; Unconditional branch</i>
1400 <p>The '<tt>br</tt>' instruction is used to cause control flow to
1401 transfer to a different basic block in the current function. There are
1402 two forms of this instruction, corresponding to a conditional branch
1403 and an unconditional branch.</p>
1405 <p>The conditional branch form of the '<tt>br</tt>' instruction takes a
1406 single '<tt>bool</tt>' value and two '<tt>label</tt>' values. The
1407 unconditional form of the '<tt>br</tt>' instruction takes a single '<tt>label</tt>'
1408 value as a target.</p>
1410 <p>Upon execution of a conditional '<tt>br</tt>' instruction, the '<tt>bool</tt>'
1411 argument is evaluated. If the value is <tt>true</tt>, control flows
1412 to the '<tt>iftrue</tt>' <tt>label</tt> argument. If "cond" is <tt>false</tt>,
1413 control flows to the '<tt>iffalse</tt>' <tt>label</tt> argument.</p>
1415 <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
1416 href="#i_ret">ret</a> int 1<br>IfUnequal:<br> <a href="#i_ret">ret</a> int 0<br></pre>
1418 <!-- _______________________________________________________________________ -->
1419 <div class="doc_subsubsection">
1420 <a name="i_switch">'<tt>switch</tt>' Instruction</a>
1423 <div class="doc_text">
1427 switch <intty> <value>, label <defaultdest> [ <intty> <val>, label <dest> ... ]
1432 <p>The '<tt>switch</tt>' instruction is used to transfer control flow to one of
1433 several different places. It is a generalization of the '<tt>br</tt>'
1434 instruction, allowing a branch to occur to one of many possible
1440 <p>The '<tt>switch</tt>' instruction uses three parameters: an integer
1441 comparison value '<tt>value</tt>', a default '<tt>label</tt>' destination, and
1442 an array of pairs of comparison value constants and '<tt>label</tt>'s. The
1443 table is not allowed to contain duplicate constant entries.</p>
1447 <p>The <tt>switch</tt> instruction specifies a table of values and
1448 destinations. When the '<tt>switch</tt>' instruction is executed, this
1449 table is searched for the given value. If the value is found, control flow is
1450 transfered to the corresponding destination; otherwise, control flow is
1451 transfered to the default destination.</p>
1453 <h5>Implementation:</h5>
1455 <p>Depending on properties of the target machine and the particular
1456 <tt>switch</tt> instruction, this instruction may be code generated in different
1457 ways. For example, it could be generated as a series of chained conditional
1458 branches or with a lookup table.</p>
1463 <i>; Emulate a conditional br instruction</i>
1464 %Val = <a href="#i_zext">zext</a> bool %value to int
1465 switch int %Val, label %truedest [int 0, label %falsedest ]
1467 <i>; Emulate an unconditional br instruction</i>
1468 switch uint 0, label %dest [ ]
1470 <i>; Implement a jump table:</i>
1471 switch uint %val, label %otherwise [ uint 0, label %onzero
1472 uint 1, label %onone
1473 uint 2, label %ontwo ]
1477 <!-- _______________________________________________________________________ -->
1478 <div class="doc_subsubsection">
1479 <a name="i_invoke">'<tt>invoke</tt>' Instruction</a>
1482 <div class="doc_text">
1487 <result> = invoke [<a href="#callingconv">cconv</a>] <ptr to function ty> %<function ptr val>(<function args>)
1488 to label <normal label> unwind label <exception label>
1493 <p>The '<tt>invoke</tt>' instruction causes control to transfer to a specified
1494 function, with the possibility of control flow transfer to either the
1495 '<tt>normal</tt>' label or the
1496 '<tt>exception</tt>' label. If the callee function returns with the
1497 "<tt><a href="#i_ret">ret</a></tt>" instruction, control flow will return to the
1498 "normal" label. If the callee (or any indirect callees) returns with the "<a
1499 href="#i_unwind"><tt>unwind</tt></a>" instruction, control is interrupted and
1500 continued at the dynamically nearest "exception" label.</p>
1504 <p>This instruction requires several arguments:</p>
1508 The optional "cconv" marker indicates which <a href="callingconv">calling
1509 convention</a> the call should use. If none is specified, the call defaults
1510 to using C calling conventions.
1512 <li>'<tt>ptr to function ty</tt>': shall be the signature of the pointer to
1513 function value being invoked. In most cases, this is a direct function
1514 invocation, but indirect <tt>invoke</tt>s are just as possible, branching off
1515 an arbitrary pointer to function value.
1518 <li>'<tt>function ptr val</tt>': An LLVM value containing a pointer to a
1519 function to be invoked. </li>
1521 <li>'<tt>function args</tt>': argument list whose types match the function
1522 signature argument types. If the function signature indicates the function
1523 accepts a variable number of arguments, the extra arguments can be
1526 <li>'<tt>normal label</tt>': the label reached when the called function
1527 executes a '<tt><a href="#i_ret">ret</a></tt>' instruction. </li>
1529 <li>'<tt>exception label</tt>': the label reached when a callee returns with
1530 the <a href="#i_unwind"><tt>unwind</tt></a> instruction. </li>
1536 <p>This instruction is designed to operate as a standard '<tt><a
1537 href="#i_call">call</a></tt>' instruction in most regards. The primary
1538 difference is that it establishes an association with a label, which is used by
1539 the runtime library to unwind the stack.</p>
1541 <p>This instruction is used in languages with destructors to ensure that proper
1542 cleanup is performed in the case of either a <tt>longjmp</tt> or a thrown
1543 exception. Additionally, this is important for implementation of
1544 '<tt>catch</tt>' clauses in high-level languages that support them.</p>
1548 %retval = invoke int %Test(int 15) to label %Continue
1549 unwind label %TestCleanup <i>; {int}:retval set</i>
1550 %retval = invoke <a href="#callingconv">coldcc</a> int %Test(int 15) to label %Continue
1551 unwind label %TestCleanup <i>; {int}:retval set</i>
1556 <!-- _______________________________________________________________________ -->
1558 <div class="doc_subsubsection"> <a name="i_unwind">'<tt>unwind</tt>'
1559 Instruction</a> </div>
1561 <div class="doc_text">
1570 <p>The '<tt>unwind</tt>' instruction unwinds the stack, continuing control flow
1571 at the first callee in the dynamic call stack which used an <a
1572 href="#i_invoke"><tt>invoke</tt></a> instruction to perform the call. This is
1573 primarily used to implement exception handling.</p>
1577 <p>The '<tt>unwind</tt>' intrinsic causes execution of the current function to
1578 immediately halt. The dynamic call stack is then searched for the first <a
1579 href="#i_invoke"><tt>invoke</tt></a> instruction on the call stack. Once found,
1580 execution continues at the "exceptional" destination block specified by the
1581 <tt>invoke</tt> instruction. If there is no <tt>invoke</tt> instruction in the
1582 dynamic call chain, undefined behavior results.</p>
1585 <!-- _______________________________________________________________________ -->
1587 <div class="doc_subsubsection"> <a name="i_unreachable">'<tt>unreachable</tt>'
1588 Instruction</a> </div>
1590 <div class="doc_text">
1599 <p>The '<tt>unreachable</tt>' instruction has no defined semantics. This
1600 instruction is used to inform the optimizer that a particular portion of the
1601 code is not reachable. This can be used to indicate that the code after a
1602 no-return function cannot be reached, and other facts.</p>
1606 <p>The '<tt>unreachable</tt>' instruction has no defined semantics.</p>
1611 <!-- ======================================================================= -->
1612 <div class="doc_subsection"> <a name="binaryops">Binary Operations</a> </div>
1613 <div class="doc_text">
1614 <p>Binary operators are used to do most of the computation in a
1615 program. They require two operands, execute an operation on them, and
1616 produce a single value. The operands might represent
1617 multiple data, as is the case with the <a href="#t_packed">packed</a> data type.
1618 The result value of a binary operator is not
1619 necessarily the same type as its operands.</p>
1620 <p>There are several different binary operators:</p>
1622 <!-- _______________________________________________________________________ -->
1623 <div class="doc_subsubsection"> <a name="i_add">'<tt>add</tt>'
1624 Instruction</a> </div>
1625 <div class="doc_text">
1627 <pre> <result> = add <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1630 <p>The '<tt>add</tt>' instruction returns the sum of its two operands.</p>
1632 <p>The two arguments to the '<tt>add</tt>' instruction must be either <a
1633 href="#t_integer">integer</a> or <a href="#t_floating">floating point</a> values.
1634 This instruction can also take <a href="#t_packed">packed</a> versions of the values.
1635 Both arguments must have identical types.</p>
1637 <p>The value produced is the integer or floating point sum of the two
1640 <pre> <result> = add int 4, %var <i>; yields {int}:result = 4 + %var</i>
1643 <!-- _______________________________________________________________________ -->
1644 <div class="doc_subsubsection"> <a name="i_sub">'<tt>sub</tt>'
1645 Instruction</a> </div>
1646 <div class="doc_text">
1648 <pre> <result> = sub <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1651 <p>The '<tt>sub</tt>' instruction returns the difference of its two
1653 <p>Note that the '<tt>sub</tt>' instruction is used to represent the '<tt>neg</tt>'
1654 instruction present in most other intermediate representations.</p>
1656 <p>The two arguments to the '<tt>sub</tt>' instruction must be either <a
1657 href="#t_integer">integer</a> or <a href="#t_floating">floating point</a>
1659 This instruction can also take <a href="#t_packed">packed</a> versions of the values.
1660 Both arguments must have identical types.</p>
1662 <p>The value produced is the integer or floating point difference of
1663 the two operands.</p>
1665 <pre> <result> = sub int 4, %var <i>; yields {int}:result = 4 - %var</i>
1666 <result> = sub int 0, %val <i>; yields {int}:result = -%var</i>
1669 <!-- _______________________________________________________________________ -->
1670 <div class="doc_subsubsection"> <a name="i_mul">'<tt>mul</tt>'
1671 Instruction</a> </div>
1672 <div class="doc_text">
1674 <pre> <result> = mul <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1677 <p>The '<tt>mul</tt>' instruction returns the product of its two
1680 <p>The two arguments to the '<tt>mul</tt>' instruction must be either <a
1681 href="#t_integer">integer</a> or <a href="#t_floating">floating point</a>
1683 This instruction can also take <a href="#t_packed">packed</a> versions of the values.
1684 Both arguments must have identical types.</p>
1686 <p>The value produced is the integer or floating point product of the
1688 <p>There is no signed vs unsigned multiplication. The appropriate
1689 action is taken based on the type of the operand.</p>
1691 <pre> <result> = mul int 4, %var <i>; yields {int}:result = 4 * %var</i>
1694 <!-- _______________________________________________________________________ -->
1695 <div class="doc_subsubsection"> <a name="i_udiv">'<tt>udiv</tt>' Instruction
1697 <div class="doc_text">
1699 <pre> <result> = udiv <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1702 <p>The '<tt>udiv</tt>' instruction returns the quotient of its two
1705 <p>The two arguments to the '<tt>udiv</tt>' instruction must be
1706 <a href="#t_integer">integer</a> values. Both arguments must have identical
1707 types. This instruction can also take <a href="#t_packed">packed</a> versions
1708 of the values in which case the elements must be integers.</p>
1710 <p>The value produced is the unsigned integer quotient of the two operands. This
1711 instruction always performs an unsigned division operation, regardless of
1712 whether the arguments are unsigned or not.</p>
1714 <pre> <result> = udiv uint 4, %var <i>; yields {uint}:result = 4 / %var</i>
1717 <!-- _______________________________________________________________________ -->
1718 <div class="doc_subsubsection"> <a name="i_sdiv">'<tt>sdiv</tt>' Instruction
1720 <div class="doc_text">
1722 <pre> <result> = sdiv <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1725 <p>The '<tt>sdiv</tt>' instruction returns the quotient of its two
1728 <p>The two arguments to the '<tt>sdiv</tt>' instruction must be
1729 <a href="#t_integer">integer</a> values. Both arguments must have identical
1730 types. This instruction can also take <a href="#t_packed">packed</a> versions
1731 of the values in which case the elements must be integers.</p>
1733 <p>The value produced is the signed integer quotient of the two operands. This
1734 instruction always performs a signed division operation, regardless of whether
1735 the arguments are signed or not.</p>
1737 <pre> <result> = sdiv int 4, %var <i>; yields {int}:result = 4 / %var</i>
1740 <!-- _______________________________________________________________________ -->
1741 <div class="doc_subsubsection"> <a name="i_fdiv">'<tt>fdiv</tt>'
1742 Instruction</a> </div>
1743 <div class="doc_text">
1745 <pre> <result> = fdiv <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1748 <p>The '<tt>fdiv</tt>' instruction returns the quotient of its two
1751 <p>The two arguments to the '<tt>div</tt>' instruction must be
1752 <a href="#t_floating">floating point</a> values. Both arguments must have
1753 identical types. This instruction can also take <a href="#t_packed">packed</a>
1754 versions of the values in which case the elements must be floating point.</p>
1756 <p>The value produced is the floating point quotient of the two operands.</p>
1758 <pre> <result> = fdiv float 4.0, %var <i>; yields {float}:result = 4.0 / %var</i>
1761 <!-- _______________________________________________________________________ -->
1762 <div class="doc_subsubsection"> <a name="i_urem">'<tt>urem</tt>' Instruction</a>
1764 <div class="doc_text">
1766 <pre> <result> = urem <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1769 <p>The '<tt>urem</tt>' instruction returns the remainder from the
1770 unsigned division of its two arguments.</p>
1772 <p>The two arguments to the '<tt>urem</tt>' instruction must be
1773 <a href="#t_integer">integer</a> values. Both arguments must have identical
1776 <p>This instruction returns the unsigned integer <i>remainder</i> of a division.
1777 This instruction always performs an unsigned division to get the remainder,
1778 regardless of whether the arguments are unsigned or not.</p>
1780 <pre> <result> = urem uint 4, %var <i>; yields {uint}:result = 4 % %var</i>
1784 <!-- _______________________________________________________________________ -->
1785 <div class="doc_subsubsection"> <a name="i_srem">'<tt>srem</tt>'
1786 Instruction</a> </div>
1787 <div class="doc_text">
1789 <pre> <result> = srem <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1792 <p>The '<tt>srem</tt>' instruction returns the remainder from the
1793 signed division of its two operands.</p>
1795 <p>The two arguments to the '<tt>srem</tt>' instruction must be
1796 <a href="#t_integer">integer</a> values. Both arguments must have identical
1799 <p>This instruction returns the <i>remainder</i> of a division (where the result
1800 has the same sign as the divisor), not the <i>modulus</i> (where the
1801 result has the same sign as the dividend) of a value. For more
1802 information about the difference, see <a
1803 href="http://mathforum.org/dr.math/problems/anne.4.28.99.html">The
1806 <pre> <result> = srem int 4, %var <i>; yields {int}:result = 4 % %var</i>
1810 <!-- _______________________________________________________________________ -->
1811 <div class="doc_subsubsection"> <a name="i_frem">'<tt>frem</tt>'
1812 Instruction</a> </div>
1813 <div class="doc_text">
1815 <pre> <result> = frem <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1818 <p>The '<tt>frem</tt>' instruction returns the remainder from the
1819 division of its two operands.</p>
1821 <p>The two arguments to the '<tt>frem</tt>' instruction must be
1822 <a href="#t_floating">floating point</a> values. Both arguments must have
1823 identical types.</p>
1825 <p>This instruction returns the <i>remainder</i> of a division.</p>
1827 <pre> <result> = frem float 4.0, %var <i>; yields {float}:result = 4.0 % %var</i>
1831 <!-- _______________________________________________________________________ -->
1832 <div class="doc_subsubsection"> <a name="i_setcc">'<tt>set<i>cc</i></tt>'
1833 Instructions</a> </div>
1834 <div class="doc_text">
1836 <pre> <result> = seteq <ty> <var1>, <var2> <i>; yields {bool}:result</i>
1837 <result> = setne <ty> <var1>, <var2> <i>; yields {bool}:result</i>
1838 <result> = setlt <ty> <var1>, <var2> <i>; yields {bool}:result</i>
1839 <result> = setgt <ty> <var1>, <var2> <i>; yields {bool}:result</i>
1840 <result> = setle <ty> <var1>, <var2> <i>; yields {bool}:result</i>
1841 <result> = setge <ty> <var1>, <var2> <i>; yields {bool}:result</i>
1844 <p>The '<tt>set<i>cc</i></tt>' family of instructions returns a boolean
1845 value based on a comparison of their two operands.</p>
1847 <p>The two arguments to the '<tt>set<i>cc</i></tt>' instructions must
1848 be of <a href="#t_firstclass">first class</a> type (it is not possible
1849 to compare '<tt>label</tt>'s, '<tt>array</tt>'s, '<tt>structure</tt>'
1850 or '<tt>void</tt>' values, etc...). Both arguments must have identical
1853 <p>The '<tt>seteq</tt>' instruction yields a <tt>true</tt> '<tt>bool</tt>'
1854 value if both operands are equal.<br>
1855 The '<tt>setne</tt>' instruction yields a <tt>true</tt> '<tt>bool</tt>'
1856 value if both operands are unequal.<br>
1857 The '<tt>setlt</tt>' instruction yields a <tt>true</tt> '<tt>bool</tt>'
1858 value if the first operand is less than the second operand.<br>
1859 The '<tt>setgt</tt>' instruction yields a <tt>true</tt> '<tt>bool</tt>'
1860 value if the first operand is greater than the second operand.<br>
1861 The '<tt>setle</tt>' instruction yields a <tt>true</tt> '<tt>bool</tt>'
1862 value if the first operand is less than or equal to the second operand.<br>
1863 The '<tt>setge</tt>' instruction yields a <tt>true</tt> '<tt>bool</tt>'
1864 value if the first operand is greater than or equal to the second
1867 <pre> <result> = seteq int 4, 5 <i>; yields {bool}:result = false</i>
1868 <result> = setne float 4, 5 <i>; yields {bool}:result = true</i>
1869 <result> = setlt uint 4, 5 <i>; yields {bool}:result = true</i>
1870 <result> = setgt sbyte 4, 5 <i>; yields {bool}:result = false</i>
1871 <result> = setle sbyte 4, 5 <i>; yields {bool}:result = true</i>
1872 <result> = setge sbyte 4, 5 <i>; yields {bool}:result = false</i>
1876 <!-- ======================================================================= -->
1877 <div class="doc_subsection"> <a name="bitwiseops">Bitwise Binary
1878 Operations</a> </div>
1879 <div class="doc_text">
1880 <p>Bitwise binary operators are used to do various forms of
1881 bit-twiddling in a program. They are generally very efficient
1882 instructions and can commonly be strength reduced from other
1883 instructions. They require two operands, execute an operation on them,
1884 and produce a single value. The resulting value of the bitwise binary
1885 operators is always the same type as its first operand.</p>
1887 <!-- _______________________________________________________________________ -->
1888 <div class="doc_subsubsection"> <a name="i_and">'<tt>and</tt>'
1889 Instruction</a> </div>
1890 <div class="doc_text">
1892 <pre> <result> = and <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1895 <p>The '<tt>and</tt>' instruction returns the bitwise logical and of
1896 its two operands.</p>
1898 <p>The two arguments to the '<tt>and</tt>' instruction must be <a
1899 href="#t_integral">integral</a> values. Both arguments must have
1900 identical types.</p>
1902 <p>The truth table used for the '<tt>and</tt>' instruction is:</p>
1904 <div style="align: center">
1905 <table border="1" cellspacing="0" cellpadding="4">
1936 <pre> <result> = and int 4, %var <i>; yields {int}:result = 4 & %var</i>
1937 <result> = and int 15, 40 <i>; yields {int}:result = 8</i>
1938 <result> = and int 4, 8 <i>; yields {int}:result = 0</i>
1941 <!-- _______________________________________________________________________ -->
1942 <div class="doc_subsubsection"> <a name="i_or">'<tt>or</tt>' Instruction</a> </div>
1943 <div class="doc_text">
1945 <pre> <result> = or <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1948 <p>The '<tt>or</tt>' instruction returns the bitwise logical inclusive
1949 or of its two operands.</p>
1951 <p>The two arguments to the '<tt>or</tt>' instruction must be <a
1952 href="#t_integral">integral</a> values. Both arguments must have
1953 identical types.</p>
1955 <p>The truth table used for the '<tt>or</tt>' instruction is:</p>
1957 <div style="align: center">
1958 <table border="1" cellspacing="0" cellpadding="4">
1989 <pre> <result> = or int 4, %var <i>; yields {int}:result = 4 | %var</i>
1990 <result> = or int 15, 40 <i>; yields {int}:result = 47</i>
1991 <result> = or int 4, 8 <i>; yields {int}:result = 12</i>
1994 <!-- _______________________________________________________________________ -->
1995 <div class="doc_subsubsection"> <a name="i_xor">'<tt>xor</tt>'
1996 Instruction</a> </div>
1997 <div class="doc_text">
1999 <pre> <result> = xor <ty> <var1>, <var2> <i>; yields {ty}:result</i>
2002 <p>The '<tt>xor</tt>' instruction returns the bitwise logical exclusive
2003 or of its two operands. The <tt>xor</tt> is used to implement the
2004 "one's complement" operation, which is the "~" operator in C.</p>
2006 <p>The two arguments to the '<tt>xor</tt>' instruction must be <a
2007 href="#t_integral">integral</a> values. Both arguments must have
2008 identical types.</p>
2010 <p>The truth table used for the '<tt>xor</tt>' instruction is:</p>
2012 <div style="align: center">
2013 <table border="1" cellspacing="0" cellpadding="4">
2045 <pre> <result> = xor int 4, %var <i>; yields {int}:result = 4 ^ %var</i>
2046 <result> = xor int 15, 40 <i>; yields {int}:result = 39</i>
2047 <result> = xor int 4, 8 <i>; yields {int}:result = 12</i>
2048 <result> = xor int %V, -1 <i>; yields {int}:result = ~%V</i>
2051 <!-- _______________________________________________________________________ -->
2052 <div class="doc_subsubsection"> <a name="i_shl">'<tt>shl</tt>'
2053 Instruction</a> </div>
2054 <div class="doc_text">
2056 <pre> <result> = shl <ty> <var1>, ubyte <var2> <i>; yields {ty}:result</i>
2059 <p>The '<tt>shl</tt>' instruction returns the first operand shifted to
2060 the left a specified number of bits.</p>
2062 <p>The first argument to the '<tt>shl</tt>' instruction must be an <a
2063 href="#t_integer">integer</a> type. The second argument must be an '<tt>ubyte</tt>'
2066 <p>The value produced is <tt>var1</tt> * 2<sup><tt>var2</tt></sup>.</p>
2068 <pre> <result> = shl int 4, ubyte %var <i>; yields {int}:result = 4 << %var</i>
2069 <result> = shl int 4, ubyte 2 <i>; yields {int}:result = 16</i>
2070 <result> = shl int 1, ubyte 10 <i>; yields {int}:result = 1024</i>
2073 <!-- _______________________________________________________________________ -->
2074 <div class="doc_subsubsection"> <a name="i_lshr">'<tt>lshr</tt>'
2075 Instruction</a> </div>
2076 <div class="doc_text">
2078 <pre> <result> = lshr <ty> <var1>, ubyte <var2> <i>; yields {ty}:result</i>
2082 <p>The '<tt>lshr</tt>' instruction (logical shift right) returns the first
2083 operand shifted to the right a specified number of bits.</p>
2086 <p>The first argument to the '<tt>lshr</tt>' instruction must be an <a
2087 href="#t_integer">integer</a> type. The second argument must be an '<tt>ubyte</tt>' type.</p>
2090 <p>This instruction always performs a logical shift right operation, regardless
2091 of whether the arguments are unsigned or not. The <tt>var2</tt> most significant
2092 bits will be filled with zero bits after the shift.</p>
2096 <result> = lshr uint 4, ubyte 1 <i>; yields {uint}:result = 2</i>
2097 <result> = lshr int 4, ubyte 2 <i>; yields {uint}:result = 1</i>
2098 <result> = lshr sbyte 4, ubyte 3 <i>; yields {sbyte}:result = 0</i>
2099 <result> = lshr sbyte -2, ubyte 1 <i>; yields {sbyte}:result = 0x7FFFFFFF </i>
2103 <!-- ======================================================================= -->
2104 <div class="doc_subsubsection"> <a name="i_ashr">'<tt>ashr</tt>'
2105 Instruction</a> </div>
2106 <div class="doc_text">
2109 <pre> <result> = ashr <ty> <var1>, ubyte <var2> <i>; yields {ty}:result</i>
2113 <p>The '<tt>ashr</tt>' instruction (arithmetic shift right) returns the first
2114 operand shifted to the right a specified number of bits.</p>
2117 <p>The first argument to the '<tt>ashr</tt>' instruction must be an
2118 <a href="#t_integer">integer</a> type. The second argument must be an
2119 '<tt>ubyte</tt>' type.</p>
2122 <p>This instruction always performs an arithmetic shift right operation,
2123 regardless of whether the arguments are signed or not. The <tt>var2</tt> most
2124 significant bits will be filled with the sign bit of <tt>var1</tt>.</p>
2128 <result> = ashr uint 4, ubyte 1 <i>; yields {uint}:result = 2</i>
2129 <result> = ashr int 4, ubyte 2 <i>; yields {int}:result = 1</i>
2130 <result> = ashr ubyte 4, ubyte 3 <i>; yields {ubyte}:result = 0</i>
2131 <result> = ashr sbyte -2, ubyte 1 <i>; yields {sbyte}:result = -1</i>
2135 <!-- ======================================================================= -->
2136 <div class="doc_subsection">
2137 <a name="vectorops">Vector Operations</a>
2140 <div class="doc_text">
2142 <p>LLVM supports several instructions to represent vector operations in a
2143 target-independent manner. This instructions cover the element-access and
2144 vector-specific operations needed to process vectors effectively. While LLVM
2145 does directly support these vector operations, many sophisticated algorithms
2146 will want to use target-specific intrinsics to take full advantage of a specific
2151 <!-- _______________________________________________________________________ -->
2152 <div class="doc_subsubsection">
2153 <a name="i_extractelement">'<tt>extractelement</tt>' Instruction</a>
2156 <div class="doc_text">
2161 <result> = extractelement <n x <ty>> <val>, uint <idx> <i>; yields <ty></i>
2167 The '<tt>extractelement</tt>' instruction extracts a single scalar
2168 element from a packed vector at a specified index.
2175 The first operand of an '<tt>extractelement</tt>' instruction is a
2176 value of <a href="#t_packed">packed</a> type. The second operand is
2177 an index indicating the position from which to extract the element.
2178 The index may be a variable.</p>
2183 The result is a scalar of the same type as the element type of
2184 <tt>val</tt>. Its value is the value at position <tt>idx</tt> of
2185 <tt>val</tt>. If <tt>idx</tt> exceeds the length of <tt>val</tt>, the
2186 results are undefined.
2192 %result = extractelement <4 x int> %vec, uint 0 <i>; yields int</i>
2197 <!-- _______________________________________________________________________ -->
2198 <div class="doc_subsubsection">
2199 <a name="i_insertelement">'<tt>insertelement</tt>' Instruction</a>
2202 <div class="doc_text">
2207 <result> = insertelement <n x <ty>> <val>, <ty> <elt>, uint <idx> <i>; yields <n x <ty>></i>
2213 The '<tt>insertelement</tt>' instruction inserts a scalar
2214 element into a packed vector at a specified index.
2221 The first operand of an '<tt>insertelement</tt>' instruction is a
2222 value of <a href="#t_packed">packed</a> type. The second operand is a
2223 scalar value whose type must equal the element type of the first
2224 operand. The third operand is an index indicating the position at
2225 which to insert the value. The index may be a variable.</p>
2230 The result is a packed vector of the same type as <tt>val</tt>. Its
2231 element values are those of <tt>val</tt> except at position
2232 <tt>idx</tt>, where it gets the value <tt>elt</tt>. If <tt>idx</tt>
2233 exceeds the length of <tt>val</tt>, the results are undefined.
2239 %result = insertelement <4 x int> %vec, int 1, uint 0 <i>; yields <4 x int></i>
2243 <!-- _______________________________________________________________________ -->
2244 <div class="doc_subsubsection">
2245 <a name="i_shufflevector">'<tt>shufflevector</tt>' Instruction</a>
2248 <div class="doc_text">
2253 <result> = shufflevector <n x <ty>> <v1>, <n x <ty>> <v2>, <n x uint> <mask> <i>; yields <n x <ty>></i>
2259 The '<tt>shufflevector</tt>' instruction constructs a permutation of elements
2260 from two input vectors, returning a vector of the same type.
2266 The first two operands of a '<tt>shufflevector</tt>' instruction are vectors
2267 with types that match each other and types that match the result of the
2268 instruction. The third argument is a shuffle mask, which has the same number
2269 of elements as the other vector type, but whose element type is always 'uint'.
2273 The shuffle mask operand is required to be a constant vector with either
2274 constant integer or undef values.
2280 The elements of the two input vectors are numbered from left to right across
2281 both of the vectors. The shuffle mask operand specifies, for each element of
2282 the result vector, which element of the two input registers the result element
2283 gets. The element selector may be undef (meaning "don't care") and the second
2284 operand may be undef if performing a shuffle from only one vector.
2290 %result = shufflevector <4 x int> %v1, <4 x int> %v2,
2291 <4 x uint> <uint 0, uint 4, uint 1, uint 5> <i>; yields <4 x int></i>
2292 %result = shufflevector <4 x int> %v1, <4 x int> undef,
2293 <4 x uint> <uint 0, uint 1, uint 2, uint 3> <i>; yields <4 x int></i> - Identity shuffle.
2298 <!-- _______________________________________________________________________ -->
2299 <div class="doc_subsubsection"> <a name="i_vsetint">'<tt>vsetint</tt>'
2300 Instruction</a> </div>
2301 <div class="doc_text">
2303 <pre><result> = vsetint <op>, <n x <ty>> <var1>, <var2> <i>; yields <n x bool></i>
2308 <p>The '<tt>vsetint</tt>' instruction takes two integer vectors and
2309 returns a vector of boolean values representing, at each position, the
2310 result of the comparison between the values at that position in the
2315 <p>The arguments to a '<tt>vsetint</tt>' instruction are a comparison
2316 operation and two value arguments. The value arguments must be of <a
2317 href="#t_integral">integral</a> <a href="#t_packed">packed</a> type,
2318 and they must have identical types. The operation argument must be
2319 one of <tt>eq</tt>, <tt>ne</tt>, <tt>slt</tt>, <tt>sgt</tt>,
2320 <tt>sle</tt>, <tt>sge</tt>, <tt>ult</tt>, <tt>ugt</tt>, <tt>ule</tt>,
2321 <tt>uge</tt>, <tt>true</tt>, and <tt>false</tt>. The result is a
2322 packed <tt>bool</tt> value with the same length as each operand.</p>
2326 <p>The following table shows the semantics of '<tt>vsetint</tt>'. For
2327 each position of the result, the comparison is done on the
2328 corresponding positions of the two value arguments. Note that the
2329 signedness of the comparison depends on the comparison opcode and
2330 <i>not</i> on the signedness of the value operands. E.g., <tt>vsetint
2331 slt <4 x unsigned> %x, %y</tt> does an elementwise <i>signed</i>
2332 comparison of <tt>%x</tt> and <tt>%y</tt>.</p>
2334 <table border="1" cellspacing="0" cellpadding="4">
2336 <tr><th>Operation</th><th>Result is true iff</th><th>Comparison is</th></tr>
2337 <tr><td><tt>eq</tt></td><td>var1 == var2</td><td>--</td></tr>
2338 <tr><td><tt>ne</tt></td><td>var1 != var2</td><td>--</td></tr>
2339 <tr><td><tt>slt</tt></td><td>var1 < var2</td><td>signed</td></tr>
2340 <tr><td><tt>sgt</tt></td><td>var1 > var2</td><td>signed</td></tr>
2341 <tr><td><tt>sle</tt></td><td>var1 <= var2</td><td>signed</td></tr>
2342 <tr><td><tt>sge</tt></td><td>var1 >= var2</td><td>signed</td></tr>
2343 <tr><td><tt>ult</tt></td><td>var1 < var2</td><td>unsigned</td></tr>
2344 <tr><td><tt>ugt</tt></td><td>var1 > var2</td><td>unsigned</td></tr>
2345 <tr><td><tt>ule</tt></td><td>var1 <= var2</td><td>unsigned</td></tr>
2346 <tr><td><tt>uge</tt></td><td>var1 >= var2</td><td>unsigned</td></tr>
2347 <tr><td><tt>true</tt></td><td>always</td><td>--</td></tr>
2348 <tr><td><tt>false</tt></td><td>never</td><td>--</td></tr>
2353 <pre> <result> = vsetint eq <2 x int> <int 0, int 1>, <int 1, int 0> <i>; yields {<2 x bool>}:result = false, false</i>
2354 <result> = vsetint ne <2 x int> <int 0, int 1>, <int 1, int 0> <i>; yields {<2 x bool>}:result = true, true</i>
2355 <result> = vsetint slt <2 x int> <int 0, int 1>, <int 1, int 0> <i>; yields {<2 x bool>}:result = true, false</i>
2356 <result> = vsetint sgt <2 x int> <int 0, int 1>, <int 1, int 0> <i>; yields {<2 x bool>}:result = false, true</i>
2357 <result> = vsetint sle <2 x int> <int 0, int 1>, <int 1, int 0> <i>; yields {<2 x bool>}:result = true, false</i>
2358 <result> = vsetint sge <2 x int> <int 0, int 1>, <int 1, int 0> <i>; yields {<2 x bool>}:result = false, true</i>
2362 <!-- _______________________________________________________________________ -->
2363 <div class="doc_subsubsection"> <a name="i_vsetfp">'<tt>vsetfp</tt>'
2364 Instruction</a> </div>
2365 <div class="doc_text">
2367 <pre><result> = vsetfp <op>, <n x <ty>> <var1>, <var2> <i>; yields <n x bool></i>
2372 <p>The '<tt>vsetfp</tt>' instruction takes two floating point vector
2373 arguments and returns a vector of boolean values representing, at each
2374 position, the result of the comparison between the values at that
2375 position in the two operands.</p>
2379 <p>The arguments to a '<tt>vsetfp</tt>' instruction are a comparison
2380 operation and two value arguments. The value arguments must be of <a
2381 href="t_floating">floating point</a> <a href="#t_packed">packed</a>
2382 type, and they must have identical types. The operation argument must
2383 be one of <tt>eq</tt>, <tt>ne</tt>, <tt>lt</tt>, <tt>gt</tt>,
2384 <tt>le</tt>, <tt>ge</tt>, <tt>oeq</tt>, <tt>one</tt>, <tt>olt</tt>,
2385 <tt>ogt</tt>, <tt>ole</tt>, <tt>oge</tt>, <tt>ueq</tt>, <tt>une</tt>,
2386 <tt>ult</tt>, <tt>ugt</tt>, <tt>ule</tt>, <tt>uge</tt>, <tt>o</tt>,
2387 <tt>u</tt>, <tt>true</tt>, and <tt>false</tt>. The result is a packed
2388 <tt>bool</tt> value with the same length as each operand.</p>
2392 <p>The following table shows the semantics of '<tt>vsetfp</tt>' for
2393 floating point types. If either operand is a floating point Not a
2394 Number (NaN) value, the operation is unordered, and the value in the
2395 first column below is produced at that position. Otherwise, the
2396 operation is ordered, and the value in the second column is
2399 <table border="1" cellspacing="0" cellpadding="4">
2401 <tr><th>Operation</th><th>If unordered<th>Otherwise true iff</th></tr>
2402 <tr><td><tt>eq</tt></td><td>undefined</td><td>var1 == var2</td></tr>
2403 <tr><td><tt>ne</tt></td><td>undefined</td><td>var1 != var2</td></tr>
2404 <tr><td><tt>lt</tt></td><td>undefined</td><td>var1 < var2</td></tr>
2405 <tr><td><tt>gt</tt></td><td>undefined</td><td>var1 > var2</td></tr>
2406 <tr><td><tt>le</tt></td><td>undefined</td><td>var1 <= var2</td></tr>
2407 <tr><td><tt>ge</tt></td><td>undefined</td><td>var1 >= var2</td></tr>
2408 <tr><td><tt>oeq</tt></td><td>false</td><td>var1 == var2</td></tr>
2409 <tr><td><tt>one</tt></td><td>false</td><td>var1 != var2</td></tr>
2410 <tr><td><tt>olt</tt></td><td>false</td><td>var1 < var2</td></tr>
2411 <tr><td><tt>ogt</tt></td><td>false</td><td>var1 > var2</td></tr>
2412 <tr><td><tt>ole</tt></td><td>false</td><td>var1 <= var2</td></tr>
2413 <tr><td><tt>oge</tt></td><td>false</td><td>var1 >= var2</td></tr>
2414 <tr><td><tt>ueq</tt></td><td>true</td><td>var1 == var2</td></tr>
2415 <tr><td><tt>une</tt></td><td>true</td><td>var1 != var2</td></tr>
2416 <tr><td><tt>ult</tt></td><td>true</td><td>var1 < var2</td></tr>
2417 <tr><td><tt>ugt</tt></td><td>true</td><td>var1 > var2</td></tr>
2418 <tr><td><tt>ule</tt></td><td>true</td><td>var1 <= var2</td></tr>
2419 <tr><td><tt>uge</tt></td><td>true</td><td>var1 >= var2</td></tr>
2420 <tr><td><tt>o</tt></td><td>false</td><td>always</td></tr>
2421 <tr><td><tt>u</tt></td><td>true</td><td>never</td></tr>
2422 <tr><td><tt>true</tt></td><td>true</td><td>always</td></tr>
2423 <tr><td><tt>false</tt></td><td>false</td><td>never</td></tr>
2428 <pre> <result> = vsetfp eq <2 x float> <float 0.0, float 1.0>, <float 1.0, float 0.0> <i>; yields {<2 x bool>}:result = false, false</i>
2429 <result> = vsetfp ne <2 x float> <float 0.0, float 1.0>, <float 1.0, float 0.0> <i>; yields {<2 x bool>}:result = true, true</i>
2430 <result> = vsetfp lt <2 x float> <float 0.0, float 1.0>, <float 1.0, float 0.0> <i>; yields {<2 x bool>}:result = true, false</i>
2431 <result> = vsetfp gt <2 x float> <float 0.0, float 1.0>, <float 1.0, float 0.0> <i>; yields {<2 x bool>}:result = false, true</i>
2432 <result> = vsetfp le <2 x float> <float 0.0, float 1.0>, <float 1.0, float 0.0> <i>; yields {<2 x bool>}:result = true, false</i>
2433 <result> = vsetfp ge <2 x float> <float 0.0, float 1.0>, <float 1.0, float 0.0> <i>; yields {<2 x bool>}:result = false, true</i>
2437 <!-- _______________________________________________________________________ -->
2438 <div class="doc_subsubsection">
2439 <a name="i_vselect">'<tt>vselect</tt>' Instruction</a>
2442 <div class="doc_text">
2447 <result> = vselect <n x bool> <cond>, <n x <ty>> <val1>, <n x <ty>> <val2> <i>; yields <n x <ty>></i>
2453 The '<tt>vselect</tt>' instruction chooses one value at each position
2454 of a vector based on a condition.
2461 The '<tt>vselect</tt>' instruction requires a <a
2462 href="#t_packed">packed</a> <tt>bool</tt> value indicating the
2463 condition at each vector position, and two values of the same packed
2464 type. All three operands must have the same length. The type of the
2465 result is the same as the type of the two value operands.</p>
2470 At each position where the <tt>bool</tt> vector is true, that position
2471 of the result gets its value from the first value argument; otherwise,
2472 it gets its value from the second value argument.
2478 %X = vselect bool <2 x bool> <bool true, bool false>, <2 x ubyte> <ubyte 17, ubyte 17>,
2479 <2 x ubyte> <ubyte 42, ubyte 42> <i>; yields <2 x ubyte>:17, 42</i>
2485 <!-- ======================================================================= -->
2486 <div class="doc_subsection">
2487 <a name="memoryops">Memory Access and Addressing Operations</a>
2490 <div class="doc_text">
2492 <p>A key design point of an SSA-based representation is how it
2493 represents memory. In LLVM, no memory locations are in SSA form, which
2494 makes things very simple. This section describes how to read, write,
2495 allocate, and free memory in LLVM.</p>
2499 <!-- _______________________________________________________________________ -->
2500 <div class="doc_subsubsection">
2501 <a name="i_malloc">'<tt>malloc</tt>' Instruction</a>
2504 <div class="doc_text">
2509 <result> = malloc <type>[, uint <NumElements>][, align <alignment>] <i>; yields {type*}:result</i>
2514 <p>The '<tt>malloc</tt>' instruction allocates memory from the system
2515 heap and returns a pointer to it.</p>
2519 <p>The '<tt>malloc</tt>' instruction allocates
2520 <tt>sizeof(<type>)*NumElements</tt>
2521 bytes of memory from the operating system and returns a pointer of the
2522 appropriate type to the program. If "NumElements" is specified, it is the
2523 number of elements allocated. If an alignment is specified, the value result
2524 of the allocation is guaranteed to be aligned to at least that boundary. If
2525 not specified, or if zero, the target can choose to align the allocation on any
2526 convenient boundary.</p>
2528 <p>'<tt>type</tt>' must be a sized type.</p>
2532 <p>Memory is allocated using the system "<tt>malloc</tt>" function, and
2533 a pointer is returned.</p>
2538 %array = malloc [4 x ubyte ] <i>; yields {[%4 x ubyte]*}:array</i>
2540 %size = <a href="#i_add">add</a> uint 2, 2 <i>; yields {uint}:size = uint 4</i>
2541 %array1 = malloc ubyte, uint 4 <i>; yields {ubyte*}:array1</i>
2542 %array2 = malloc [12 x ubyte], uint %size <i>; yields {[12 x ubyte]*}:array2</i>
2543 %array3 = malloc int, uint 4, align 1024 <i>; yields {int*}:array3</i>
2544 %array4 = malloc int, align 1024 <i>; yields {int*}:array4</i>
2548 <!-- _______________________________________________________________________ -->
2549 <div class="doc_subsubsection">
2550 <a name="i_free">'<tt>free</tt>' Instruction</a>
2553 <div class="doc_text">
2558 free <type> <value> <i>; yields {void}</i>
2563 <p>The '<tt>free</tt>' instruction returns memory back to the unused
2564 memory heap to be reallocated in the future.</p>
2568 <p>'<tt>value</tt>' shall be a pointer value that points to a value
2569 that was allocated with the '<tt><a href="#i_malloc">malloc</a></tt>'
2574 <p>Access to the memory pointed to by the pointer is no longer defined
2575 after this instruction executes.</p>
2580 %array = <a href="#i_malloc">malloc</a> [4 x ubyte] <i>; yields {[4 x ubyte]*}:array</i>
2581 free [4 x ubyte]* %array
2585 <!-- _______________________________________________________________________ -->
2586 <div class="doc_subsubsection">
2587 <a name="i_alloca">'<tt>alloca</tt>' Instruction</a>
2590 <div class="doc_text">
2595 <result> = alloca <type>[, uint <NumElements>][, align <alignment>] <i>; yields {type*}:result</i>
2600 <p>The '<tt>alloca</tt>' instruction allocates memory on the current
2601 stack frame of the procedure that is live until the current function
2602 returns to its caller.</p>
2606 <p>The '<tt>alloca</tt>' instruction allocates <tt>sizeof(<type>)*NumElements</tt>
2607 bytes of memory on the runtime stack, returning a pointer of the
2608 appropriate type to the program. If "NumElements" is specified, it is the
2609 number of elements allocated. If an alignment is specified, the value result
2610 of the allocation is guaranteed to be aligned to at least that boundary. If
2611 not specified, or if zero, the target can choose to align the allocation on any
2612 convenient boundary.</p>
2614 <p>'<tt>type</tt>' may be any sized type.</p>
2618 <p>Memory is allocated; a pointer is returned. '<tt>alloca</tt>'d
2619 memory is automatically released when the function returns. The '<tt>alloca</tt>'
2620 instruction is commonly used to represent automatic variables that must
2621 have an address available. When the function returns (either with the <tt><a
2622 href="#i_ret">ret</a></tt> or <tt><a href="#i_unwind">unwind</a></tt>
2623 instructions), the memory is reclaimed.</p>
2628 %ptr = alloca int <i>; yields {int*}:ptr</i>
2629 %ptr = alloca int, uint 4 <i>; yields {int*}:ptr</i>
2630 %ptr = alloca int, uint 4, align 1024 <i>; yields {int*}:ptr</i>
2631 %ptr = alloca int, align 1024 <i>; yields {int*}:ptr</i>
2635 <!-- _______________________________________________________________________ -->
2636 <div class="doc_subsubsection"> <a name="i_load">'<tt>load</tt>'
2637 Instruction</a> </div>
2638 <div class="doc_text">
2640 <pre> <result> = load <ty>* <pointer><br> <result> = volatile load <ty>* <pointer><br></pre>
2642 <p>The '<tt>load</tt>' instruction is used to read from memory.</p>
2644 <p>The argument to the '<tt>load</tt>' instruction specifies the memory
2645 address from which to load. The pointer must point to a <a
2646 href="#t_firstclass">first class</a> type. If the <tt>load</tt> is
2647 marked as <tt>volatile</tt>, then the optimizer is not allowed to modify
2648 the number or order of execution of this <tt>load</tt> with other
2649 volatile <tt>load</tt> and <tt><a href="#i_store">store</a></tt>
2652 <p>The location of memory pointed to is loaded.</p>
2654 <pre> %ptr = <a href="#i_alloca">alloca</a> int <i>; yields {int*}:ptr</i>
2656 href="#i_store">store</a> int 3, int* %ptr <i>; yields {void}</i>
2657 %val = load int* %ptr <i>; yields {int}:val = int 3</i>
2660 <!-- _______________________________________________________________________ -->
2661 <div class="doc_subsubsection"> <a name="i_store">'<tt>store</tt>'
2662 Instruction</a> </div>
2664 <pre> store <ty> <value>, <ty>* <pointer> <i>; yields {void}</i>
2665 volatile store <ty> <value>, <ty>* <pointer> <i>; yields {void}</i>
2668 <p>The '<tt>store</tt>' instruction is used to write to memory.</p>
2670 <p>There are two arguments to the '<tt>store</tt>' instruction: a value
2671 to store and an address in which to store it. The type of the '<tt><pointer></tt>'
2672 operand must be a pointer to the type of the '<tt><value></tt>'
2673 operand. If the <tt>store</tt> is marked as <tt>volatile</tt>, then the
2674 optimizer is not allowed to modify the number or order of execution of
2675 this <tt>store</tt> with other volatile <tt>load</tt> and <tt><a
2676 href="#i_store">store</a></tt> instructions.</p>
2678 <p>The contents of memory are updated to contain '<tt><value></tt>'
2679 at the location specified by the '<tt><pointer></tt>' operand.</p>
2681 <pre> %ptr = <a href="#i_alloca">alloca</a> int <i>; yields {int*}:ptr</i>
2683 href="#i_store">store</a> int 3, int* %ptr <i>; yields {void}</i>
2684 %val = load int* %ptr <i>; yields {int}:val = int 3</i>
2686 <!-- _______________________________________________________________________ -->
2687 <div class="doc_subsubsection">
2688 <a name="i_getelementptr">'<tt>getelementptr</tt>' Instruction</a>
2691 <div class="doc_text">
2694 <result> = getelementptr <ty>* <ptrval>{, <ty> <idx>}*
2700 The '<tt>getelementptr</tt>' instruction is used to get the address of a
2701 subelement of an aggregate data structure.</p>
2705 <p>This instruction takes a list of integer constants that indicate what
2706 elements of the aggregate object to index to. The actual types of the arguments
2707 provided depend on the type of the first pointer argument. The
2708 '<tt>getelementptr</tt>' instruction is used to index down through the type
2709 levels of a structure or to a specific index in an array. When indexing into a
2710 structure, only <tt>uint</tt>
2711 integer constants are allowed. When indexing into an array or pointer,
2712 <tt>int</tt> and <tt>long</tt> indexes are allowed of any sign.</p>
2714 <p>For example, let's consider a C code fragment and how it gets
2715 compiled to LLVM:</p>
2729 int *foo(struct ST *s) {
2730 return &s[1].Z.B[5][13];
2734 <p>The LLVM code generated by the GCC frontend is:</p>
2737 %RT = type { sbyte, [10 x [20 x int]], sbyte }
2738 %ST = type { int, double, %RT }
2742 int* %foo(%ST* %s) {
2744 %reg = getelementptr %ST* %s, int 1, uint 2, uint 1, int 5, int 13
2751 <p>The index types specified for the '<tt>getelementptr</tt>' instruction depend
2752 on the pointer type that is being indexed into. <a href="#t_pointer">Pointer</a>
2753 and <a href="#t_array">array</a> types require <tt>uint</tt>, <tt>int</tt>,
2754 <tt>ulong</tt>, or <tt>long</tt> values, and <a href="#t_struct">structure</a>
2755 types require <tt>uint</tt> <b>constants</b>.</p>
2757 <p>In the example above, the first index is indexing into the '<tt>%ST*</tt>'
2758 type, which is a pointer, yielding a '<tt>%ST</tt>' = '<tt>{ int, double, %RT
2759 }</tt>' type, a structure. The second index indexes into the third element of
2760 the structure, yielding a '<tt>%RT</tt>' = '<tt>{ sbyte, [10 x [20 x int]],
2761 sbyte }</tt>' type, another structure. The third index indexes into the second
2762 element of the structure, yielding a '<tt>[10 x [20 x int]]</tt>' type, an
2763 array. The two dimensions of the array are subscripted into, yielding an
2764 '<tt>int</tt>' type. The '<tt>getelementptr</tt>' instruction returns a pointer
2765 to this element, thus computing a value of '<tt>int*</tt>' type.</p>
2767 <p>Note that it is perfectly legal to index partially through a
2768 structure, returning a pointer to an inner element. Because of this,
2769 the LLVM code for the given testcase is equivalent to:</p>
2772 int* %foo(%ST* %s) {
2773 %t1 = getelementptr %ST* %s, int 1 <i>; yields %ST*:%t1</i>
2774 %t2 = getelementptr %ST* %t1, int 0, uint 2 <i>; yields %RT*:%t2</i>
2775 %t3 = getelementptr %RT* %t2, int 0, uint 1 <i>; yields [10 x [20 x int]]*:%t3</i>
2776 %t4 = getelementptr [10 x [20 x int]]* %t3, int 0, int 5 <i>; yields [20 x int]*:%t4</i>
2777 %t5 = getelementptr [20 x int]* %t4, int 0, int 13 <i>; yields int*:%t5</i>
2782 <p>Note that it is undefined to access an array out of bounds: array and
2783 pointer indexes must always be within the defined bounds of the array type.
2784 The one exception for this rules is zero length arrays. These arrays are
2785 defined to be accessible as variable length arrays, which requires access
2786 beyond the zero'th element.</p>
2788 <p>The getelementptr instruction is often confusing. For some more insight
2789 into how it works, see <a href="GetElementPtr.html">the getelementptr
2795 <i>; yields [12 x ubyte]*:aptr</i>
2796 %aptr = getelementptr {int, [12 x ubyte]}* %sptr, long 0, uint 1
2800 <!-- ======================================================================= -->
2801 <div class="doc_subsection"> <a name="convertops">Conversion Operations</a>
2803 <div class="doc_text">
2804 <p>The instructions in this category are the conversion instructions (casting)
2805 which all take a single operand and a type. They perform various bit conversions
2809 <!-- _______________________________________________________________________ -->
2810 <div class="doc_subsubsection">
2811 <a name="i_trunc">'<tt>trunc .. to</tt>' Instruction</a>
2813 <div class="doc_text">
2817 <result> = trunc <ty> <value> to <ty2> <i>; yields ty2</i>
2822 The '<tt>trunc</tt>' instruction truncates its operand to the type <tt>ty2</tt>.
2827 The '<tt>trunc</tt>' instruction takes a <tt>value</tt> to trunc, which must
2828 be an <a href="#t_integer">integer</a> type, and a type that specifies the size
2829 and type of the result, which must be an <a href="#t_integral">integral</a>
2834 The '<tt>trunc</tt>' instruction truncates the high order bits in <tt>value</tt>
2835 and converts the reamining bits to <tt>ty2</tt>. The bit size of <tt>value</tt>
2836 must be larger than the bit size of <tt>ty2</tt>. Equal sized types are not
2837 allowed. This implies that a <tt>trunc</tt> cannot be a <i>no-op cast</i>. It
2838 will always truncate bits.</p>
2840 <p>When truncating to bool, the truncation is done as a comparison against
2841 zero. If the <tt>value</tt> was zero, the bool result will be <tt>false</tt>.
2842 If the <tt>value</tt> was non-zero, the bool result will be <tt>true</tt>.</p>
2846 %X = trunc int 257 to ubyte <i>; yields ubyte:1</i>
2847 %Y = trunc int 123 to bool <i>; yields bool:true</i>
2851 <!-- _______________________________________________________________________ -->
2852 <div class="doc_subsubsection">
2853 <a name="i_zext">'<tt>zext .. to</tt>' Instruction</a>
2855 <div class="doc_text">
2859 <result> = zext <ty> <value> to <ty2> <i>; yields ty2</i>
2863 <p>The '<tt>zext</tt>' instruction zero extends its operand to type
2868 <p>The '<tt>zext</tt>' instruction takes a value to cast, which must be of
2869 <a href="#t_integral">integral</a> type, and a type to cast it to, which must
2870 also be of <a href="#t_integral">integral</a> type. The bit size of the
2871 <tt>value</tt> must be smaller than or equal to the bit size of the
2872 destination type, <tt>ty2</tt>.</p>
2875 <p>The <tt>zext</tt> fills the high order bits of the <tt>value</tt> with zero
2876 bits until it reaches the size of the destination type, <tt>ty2</tt>. When the
2877 the operand and the type are the same size, no bit filling is done and the
2878 cast is considered a <i>no-op cast</i> because no bits change (only the type
2881 <p>When zero extending to bool, the extension is done as a comparison against
2882 zero. If the <tt>value</tt> was zero, the bool result will be <tt>false</tt>.
2883 If the <tt>value</tt> was non-zero, the bool result will be <tt>true</tt>.</p>
2887 %X = zext int 257 to ulong <i>; yields ulong:257</i>
2888 %Y = zext bool true to int <i>; yields int:1</i>
2892 <!-- _______________________________________________________________________ -->
2893 <div class="doc_subsubsection">
2894 <a name="i_sext">'<tt>sext .. to</tt>' Instruction</a>
2896 <div class="doc_text">
2900 <result> = sext <ty> <value> to <ty2> <i>; yields ty2</i>
2904 <p>The '<tt>sext</tt>' sign extends <tt>value</tt> to the type <tt>ty2</tt>.</p>
2908 The '<tt>sext</tt>' instruction takes a value to cast, which must be of
2909 <a href="#t_integral">integral</a> type, and a type to cast it to, which must
2910 also be of <a href="#t_integral">integral</a> type.</p>
2914 The '<tt>sext</tt>' instruction performs a sign extension by copying the sign
2915 bit (highest order bit) of the <tt>value</tt> until it reaches the bit size of
2916 the type <tt>ty2</tt>. When the the operand and the type are the same size,
2917 no bit filling is done and the cast is considered a <i>no-op cast</i> because
2918 no bits change (only the type changes).</p>
2920 <p>When sign extending to bool, the extension is done as a comparison against
2921 zero. If the <tt>value</tt> was zero, the bool result will be <tt>false</tt>.
2922 If the <tt>value</tt> was non-zero, the bool result will be <tt>true</tt>.</p>
2927 %X = sext sbyte -1 to ushort <i>; yields ushort:65535</i>
2928 %Y = sext bool true to int <i>; yields int:-1</i>
2932 <!-- _______________________________________________________________________ -->
2933 <div class="doc_subsubsection">
2934 <a name="i_fpext">'<tt>fpext .. to</tt>' Instruction</a>
2936 <div class="doc_text">
2940 <result> = fpext <ty> <value> to <ty2> <i>; yields ty2</i>
2944 <p>The '<tt>fpext</tt>' extends a floating point <tt>value</tt> to a larger
2945 floating point value.</p>
2948 <p>The '<tt>fpext</tt>' instruction takes a
2949 <a href="#t_floating">floating point</a> <tt>value</tt> to cast,
2950 and a <a href="#t_floating">floating point</a> type to cast it to.</p>
2953 <p>The '<tt>fpext</tt>' instruction extends the <tt>value</tt> from one floating
2954 point type to another. If the type of the <tt>value</tt> and <tt>ty2</tt> are
2955 the same, the instruction is considered a <i>no-op cast</i> because no bits
2960 %X = fpext float 3.1415 to double <i>; yields double:3.1415</i>
2961 %Y = fpext float 1.0 to float <i>; yields float:1.0 (no-op)</i>
2965 <!-- _______________________________________________________________________ -->
2966 <div class="doc_subsubsection">
2967 <a name="i_fptrunc">'<tt>fptrunc .. to</tt>' Instruction</a>
2970 <div class="doc_text">
2975 <result> = fptrunc <ty> <value> to <ty2> <i>; yields ty2</i>
2979 <p>The '<tt>fptrunc</tt>' instruction truncates <tt>value</tt> to type
2984 <p>The '<tt>fptrunc</tt>' instruction takes a <a href="#t_floating">floating
2985 point</a> value to cast and a <a href="#t_floating">floating point</a> type to
2986 cast it to. The size of <tt>value</tt> must be larger than the size of
2987 <tt>ty2</a>. This implies that <tt>fptrunc</tt> cannot be used to make a
2988 <i>no-op cast</i>.</p>
2991 <p> The '<tt>fptrunc</tt>' instruction converts a
2992 <a href="#t_floating">floating point</a> value from a larger type to a smaller
2993 type. If the value cannot fit within the destination type, <tt>ty2</tt>, then
2994 the results are undefined.</p>
2998 %X = fptrunc double 123.0 to float <i>; yields float:123.0</i>
2999 %Y = fptrunc double 1.0E+300 to float <i>; yields undefined</i>
3003 <!-- _______________________________________________________________________ -->
3004 <div class="doc_subsubsection">
3005 <a name="i_fp2uint">'<tt>fp2uint .. to</tt>' Instruction</a>
3007 <div class="doc_text">
3011 <result> = fp2uint <ty> <value> to <ty2> <i>; yields ty2</i>
3015 <p>The '<tt>fp2uint</tt>' converts a floating point <tt>value</tt> to its
3016 unsigned integer equivalent of type <tt>ty2</tt>.
3020 <p>The '<tt>fp2uint</tt>' instruction takes a value to cast, which must be a
3021 <a href="#t_floating">floating point</a> value, and a type to cast it to, which
3022 must be an <a href="#t_integral">integral</a> type.</p>
3025 <p> The '<tt>fp2uint</tt>' instruction converts its
3026 <a href="#t_floating">floating point</a> operand into the nearest (rounding
3027 towards zero) unsigned integer value. If the value cannot fit in <tt>ty2</tt>,
3028 the results are undefined.</p>
3030 <p>When converting to bool, the conversion is done as a comparison against
3031 zero. If the <tt>value</tt> was zero, the bool result will be <tt>false</tt>.
3032 If the <tt>value</tt> was non-zero, the bool result will be <tt>true</tt>.</p>
3036 %X = fp2uint double 123.0 to int <i>; yields int:123</i>
3037 %Y = fp2uint float 1.0E+300 to bool <i>; yields bool:true</i>
3038 %X = fp2uint float 1.04E+17 to ubyte <i>; yields undefined:1</i>
3042 <!-- _______________________________________________________________________ -->
3043 <div class="doc_subsubsection">
3044 <a name="i_fp2sint">'<tt>fp2sint .. to</tt>' Instruction</a>
3046 <div class="doc_text">
3050 <result> = fp2sint <ty> <value> to <ty2> <i>; yields ty2</i>
3054 <p>The '<tt>fp2sint</tt>' instruction converts
3055 <a href="#t_floating">floating point</a> <tt>value</tt> to type <tt>ty2</tt>.
3060 <p> The '<tt>fp2sint</tt>' instruction takes a value to cast, which must be a
3061 <a href="#t_floating">floating point</a> value, and a type to cast it to, which
3062 must also be an <a href="#t_integral">integral</a> type.</p>
3065 <p>The '<tt>fp2sint</tt>' instruction converts its
3066 <a href="#t_floating">floating point</a> operand into the nearest (rounding
3067 towards zero) signed integer value. If the value cannot fit in <tt>ty2</tt>,
3068 the results are undefined.</p>
3070 <p>When converting to bool, the conversion is done as a comparison against
3071 zero. If the <tt>value</tt> was zero, the bool result will be <tt>false</tt>.
3072 If the <tt>value</tt> was non-zero, the bool result will be <tt>true</tt>.</p>
3076 %X = fp2sint double -123.0 to int <i>; yields int:-123</i>
3077 %Y = fp2sint float 1.0E-247 to bool <i>; yields bool:true</i>
3078 %X = fp2sint float 1.04E+17 to sbyte <i>; yields undefined:1</i>
3082 <!-- _______________________________________________________________________ -->
3083 <div class="doc_subsubsection">
3084 <a name="i_uint2fp">'<tt>uint2fp .. to</tt>' Instruction</a>
3086 <div class="doc_text">
3090 <result> = uint2fp <ty> <value> to <ty2> <i>; yields ty2</i>
3094 <p>The '<tt>uint2fp</tt>' instruction regards <tt>value</tt> as an unsigned
3095 integer and converts that value to the <tt>ty2</tt> type.</p>
3099 <p>The '<tt>uint2fp</tt>' instruction takes a value to cast, which must be an
3100 <a href="#t_integral">integral</a> value, and a type to cast it to, which must
3101 be a <a href="#t_floating">floating point</a> type.</p>
3104 <p>The '<tt>uint2fp</tt>' instruction interprets its operand as an unsigned
3105 integer quantity and converts it to the corresponding floating point value. If
3106 the value cannot fit in the floating point value, the results are undefined.</p>
3111 %X = uint2fp int 257 to float <i>; yields float:257.0</i>
3112 %Y = uint2fp sbyte -1 to double <i>; yields double:255.0</i>
3116 <!-- _______________________________________________________________________ -->
3117 <div class="doc_subsubsection">
3118 <a name="i_sint2fp">'<tt>sint2fp .. to</tt>' Instruction</a>
3120 <div class="doc_text">
3124 <result> = sint2fp <ty> <value> to <ty2> <i>; yields ty2</i>
3128 <p>The '<tt>sint2fp</tt>' instruction regards <tt>value</tt> as a signed
3129 integer and converts that value to the <tt>ty2</tt> type.</p>
3132 <p>The '<tt>sint2fp</tt>' instruction takes a value to cast, which must be an
3133 <a href="#t_integral">integral</a> value, and a type to cast it to, which must be
3134 a <a href="#t_floating">floating point</a> type.</p>
3137 <p>The '<tt>sint2fp</tt>' instruction interprets its operand as a signed
3138 integer quantity and converts it to the corresponding floating point value. If
3139 the value cannot fit in the floating point value, the results are undefined.</p>
3143 %X = sint2fp int 257 to float <i>; yields float:257.0</i>
3144 %Y = sint2fp sbyte -1 to double <i>; yields double:-1.0</i>
3148 <!-- _______________________________________________________________________ -->
3149 <div class="doc_subsubsection">
3150 <a name="i_bitconvert">'<tt>bitconvert .. to</tt>' Instruction</a>
3152 <div class="doc_text">
3156 <result> = bitconvert <ty> <value> to <ty2> <i>; yields ty2</i>
3160 <p>The '<tt>bitconvert</tt>' instruction converts <tt>value</tt> to type
3161 <tt>ty2</tt> without changing any bits.</p>
3164 <p>The '<tt>bitconvert</tt>' instruction takes a value to cast, which must be
3165 a first class value, and a type to cast it to, which must also be a <a
3166 href="#t_firstclass">first class</a> type. The bit sizes of <tt>value</tt>
3167 and the destination type, <tt>ty2</tt>, must be identical.</p>
3170 <p>The '<tt>bitconvert</tt>' instruction converts <tt>value</tt> to type
3171 <tt>ty2</tt> as if the value had been stored to memory and read back as type
3172 <tt>ty2</tt>. That is, no bits are changed during the conversion. The
3173 <tt>bitconvert</tt> instruction may be used to construct <i>no-op casts</i> that
3174 the <tt>zext, sext, and fpext</tt> instructions do not permit.</p>
3178 %X = bitconvert ubyte 255 to sbyte <i>; yields sbyte:-1</i>
3179 %Y = bitconvert uint* %x to uint <i>; yields uint:%x</i>
3183 <!-- ======================================================================= -->
3184 <div class="doc_subsection"> <a name="otherops">Other Operations</a> </div>
3185 <div class="doc_text">
3186 <p>The instructions in this category are the "miscellaneous"
3187 instructions, which defy better classification.</p>
3189 <!-- _______________________________________________________________________ -->
3190 <div class="doc_subsubsection"> <a name="i_phi">'<tt>phi</tt>'
3191 Instruction</a> </div>
3192 <div class="doc_text">
3194 <pre> <result> = phi <ty> [ <val0>, <label0>], ...<br></pre>
3196 <p>The '<tt>phi</tt>' instruction is used to implement the φ node in
3197 the SSA graph representing the function.</p>
3199 <p>The type of the incoming values are specified with the first type
3200 field. After this, the '<tt>phi</tt>' instruction takes a list of pairs
3201 as arguments, with one pair for each predecessor basic block of the
3202 current block. Only values of <a href="#t_firstclass">first class</a>
3203 type may be used as the value arguments to the PHI node. Only labels
3204 may be used as the label arguments.</p>
3205 <p>There must be no non-phi instructions between the start of a basic
3206 block and the PHI instructions: i.e. PHI instructions must be first in
3209 <p>At runtime, the '<tt>phi</tt>' instruction logically takes on the
3210 value specified by the parameter, depending on which basic block we
3211 came from in the last <a href="#terminators">terminator</a> instruction.</p>
3213 <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>
3216 <!-- _______________________________________________________________________ -->
3217 <div class="doc_subsubsection">
3218 <a name="i_select">'<tt>select</tt>' Instruction</a>
3221 <div class="doc_text">
3226 <result> = select bool <cond>, <ty> <val1>, <ty> <val2> <i>; yields ty</i>
3232 The '<tt>select</tt>' instruction is used to choose one value based on a
3233 condition, without branching.
3240 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.
3246 If the boolean condition evaluates to true, the instruction returns the first
3247 value argument; otherwise, it returns the second value argument.
3253 %X = select bool true, ubyte 17, ubyte 42 <i>; yields ubyte:17</i>
3258 <!-- _______________________________________________________________________ -->
3259 <div class="doc_subsubsection">
3260 <a name="i_call">'<tt>call</tt>' Instruction</a>
3263 <div class="doc_text">
3267 <result> = [tail] call [<a href="#callingconv">cconv</a>] <ty>* <fnptrval>(<param list>)
3272 <p>The '<tt>call</tt>' instruction represents a simple function call.</p>
3276 <p>This instruction requires several arguments:</p>
3280 <p>The optional "tail" marker indicates whether the callee function accesses
3281 any allocas or varargs in the caller. If the "tail" marker is present, the
3282 function call is eligible for tail call optimization. Note that calls may
3283 be marked "tail" even if they do not occur before a <a
3284 href="#i_ret"><tt>ret</tt></a> instruction.
3287 <p>The optional "cconv" marker indicates which <a href="callingconv">calling
3288 convention</a> the call should use. If none is specified, the call defaults
3289 to using C calling conventions.
3292 <p>'<tt>ty</tt>': shall be the signature of the pointer to function value
3293 being invoked. The argument types must match the types implied by this
3294 signature. This type can be omitted if the function is not varargs and
3295 if the function type does not return a pointer to a function.</p>
3298 <p>'<tt>fnptrval</tt>': An LLVM value containing a pointer to a function to
3299 be invoked. In most cases, this is a direct function invocation, but
3300 indirect <tt>call</tt>s are just as possible, calling an arbitrary pointer
3301 to function value.</p>
3304 <p>'<tt>function args</tt>': argument list whose types match the
3305 function signature argument types. All arguments must be of
3306 <a href="#t_firstclass">first class</a> type. If the function signature
3307 indicates the function accepts a variable number of arguments, the extra
3308 arguments can be specified.</p>
3314 <p>The '<tt>call</tt>' instruction is used to cause control flow to
3315 transfer to a specified function, with its incoming arguments bound to
3316 the specified values. Upon a '<tt><a href="#i_ret">ret</a></tt>'
3317 instruction in the called function, control flow continues with the
3318 instruction after the function call, and the return value of the
3319 function is bound to the result argument. This is a simpler case of
3320 the <a href="#i_invoke">invoke</a> instruction.</p>
3325 %retval = call int %test(int %argc)
3326 call int(sbyte*, ...) *%printf(sbyte* %msg, int 12, sbyte 42);
3327 %X = tail call int %foo()
3328 %Y = tail call <a href="#callingconv">fastcc</a> int %foo()
3333 <!-- _______________________________________________________________________ -->
3334 <div class="doc_subsubsection">
3335 <a name="i_va_arg">'<tt>va_arg</tt>' Instruction</a>
3338 <div class="doc_text">
3343 <resultval> = va_arg <va_list*> <arglist>, <argty>
3348 <p>The '<tt>va_arg</tt>' instruction is used to access arguments passed through
3349 the "variable argument" area of a function call. It is used to implement the
3350 <tt>va_arg</tt> macro in C.</p>
3354 <p>This instruction takes a <tt>va_list*</tt> value and the type of
3355 the argument. It returns a value of the specified argument type and
3356 increments the <tt>va_list</tt> to point to the next argument. Again, the
3357 actual type of <tt>va_list</tt> is target specific.</p>
3361 <p>The '<tt>va_arg</tt>' instruction loads an argument of the specified
3362 type from the specified <tt>va_list</tt> and causes the
3363 <tt>va_list</tt> to point to the next argument. For more information,
3364 see the variable argument handling <a href="#int_varargs">Intrinsic
3367 <p>It is legal for this instruction to be called in a function which does not
3368 take a variable number of arguments, for example, the <tt>vfprintf</tt>
3371 <p><tt>va_arg</tt> is an LLVM instruction instead of an <a
3372 href="#intrinsics">intrinsic function</a> because it takes a type as an
3377 <p>See the <a href="#int_varargs">variable argument processing</a> section.</p>
3381 <!-- *********************************************************************** -->
3382 <div class="doc_section"> <a name="intrinsics">Intrinsic Functions</a> </div>
3383 <!-- *********************************************************************** -->
3385 <div class="doc_text">
3387 <p>LLVM supports the notion of an "intrinsic function". These functions have
3388 well known names and semantics and are required to follow certain
3389 restrictions. Overall, these instructions represent an extension mechanism for
3390 the LLVM language that does not require changing all of the transformations in
3391 LLVM to add to the language (or the bytecode reader/writer, the parser,
3394 <p>Intrinsic function names must all start with an "<tt>llvm.</tt>" prefix. This
3395 prefix is reserved in LLVM for intrinsic names; thus, functions may not be named
3396 this. Intrinsic functions must always be external functions: you cannot define
3397 the body of intrinsic functions. Intrinsic functions may only be used in call
3398 or invoke instructions: it is illegal to take the address of an intrinsic
3399 function. Additionally, because intrinsic functions are part of the LLVM
3400 language, it is required that they all be documented here if any are added.</p>
3403 <p>To learn how to add an intrinsic function, please see the <a
3404 href="ExtendingLLVM.html">Extending LLVM Guide</a>.
3409 <!-- ======================================================================= -->
3410 <div class="doc_subsection">
3411 <a name="int_varargs">Variable Argument Handling Intrinsics</a>
3414 <div class="doc_text">
3416 <p>Variable argument support is defined in LLVM with the <a
3417 href="#i_va_arg"><tt>va_arg</tt></a> instruction and these three
3418 intrinsic functions. These functions are related to the similarly
3419 named macros defined in the <tt><stdarg.h></tt> header file.</p>
3421 <p>All of these functions operate on arguments that use a
3422 target-specific value type "<tt>va_list</tt>". The LLVM assembly
3423 language reference manual does not define what this type is, so all
3424 transformations should be prepared to handle intrinsics with any type
3427 <p>This example shows how the <a href="#i_va_arg"><tt>va_arg</tt></a>
3428 instruction and the variable argument handling intrinsic functions are
3432 int %test(int %X, ...) {
3433 ; Initialize variable argument processing
3435 call void %<a href="#i_va_start">llvm.va_start</a>(sbyte** %ap)
3437 ; Read a single integer argument
3438 %tmp = va_arg sbyte** %ap, int
3440 ; Demonstrate usage of llvm.va_copy and llvm.va_end
3442 call void %<a href="#i_va_copy">llvm.va_copy</a>(sbyte** %aq, sbyte** %ap)
3443 call void %<a href="#i_va_end">llvm.va_end</a>(sbyte** %aq)
3445 ; Stop processing of arguments.
3446 call void %<a href="#i_va_end">llvm.va_end</a>(sbyte** %ap)
3452 <!-- _______________________________________________________________________ -->
3453 <div class="doc_subsubsection">
3454 <a name="i_va_start">'<tt>llvm.va_start</tt>' Intrinsic</a>
3458 <div class="doc_text">
3460 <pre> declare void %llvm.va_start(<va_list>* <arglist>)<br></pre>
3462 <P>The '<tt>llvm.va_start</tt>' intrinsic initializes
3463 <tt>*<arglist></tt> for subsequent use by <tt><a
3464 href="#i_va_arg">va_arg</a></tt>.</p>
3468 <P>The argument is a pointer to a <tt>va_list</tt> element to initialize.</p>
3472 <P>The '<tt>llvm.va_start</tt>' intrinsic works just like the <tt>va_start</tt>
3473 macro available in C. In a target-dependent way, it initializes the
3474 <tt>va_list</tt> element the argument points to, so that the next call to
3475 <tt>va_arg</tt> will produce the first variable argument passed to the function.
3476 Unlike the C <tt>va_start</tt> macro, this intrinsic does not need to know the
3477 last argument of the function, the compiler can figure that out.</p>
3481 <!-- _______________________________________________________________________ -->
3482 <div class="doc_subsubsection">
3483 <a name="i_va_end">'<tt>llvm.va_end</tt>' Intrinsic</a>
3486 <div class="doc_text">
3488 <pre> declare void %llvm.va_end(<va_list*> <arglist>)<br></pre>
3490 <p>The '<tt>llvm.va_end</tt>' intrinsic destroys <tt><arglist></tt>
3491 which has been initialized previously with <tt><a href="#i_va_start">llvm.va_start</a></tt>
3492 or <tt><a href="#i_va_copy">llvm.va_copy</a></tt>.</p>
3494 <p>The argument is a <tt>va_list</tt> to destroy.</p>
3496 <p>The '<tt>llvm.va_end</tt>' intrinsic works just like the <tt>va_end</tt>
3497 macro available in C. In a target-dependent way, it destroys the <tt>va_list</tt>.
3498 Calls to <a href="#i_va_start"><tt>llvm.va_start</tt></a> and <a
3499 href="#i_va_copy"><tt>llvm.va_copy</tt></a> must be matched exactly
3500 with calls to <tt>llvm.va_end</tt>.</p>
3503 <!-- _______________________________________________________________________ -->
3504 <div class="doc_subsubsection">
3505 <a name="i_va_copy">'<tt>llvm.va_copy</tt>' Intrinsic</a>
3508 <div class="doc_text">
3513 declare void %llvm.va_copy(<va_list>* <destarglist>,
3514 <va_list>* <srcarglist>)
3519 <p>The '<tt>llvm.va_copy</tt>' intrinsic copies the current argument position from
3520 the source argument list to the destination argument list.</p>
3524 <p>The first argument is a pointer to a <tt>va_list</tt> element to initialize.
3525 The second argument is a pointer to a <tt>va_list</tt> element to copy from.</p>
3530 <p>The '<tt>llvm.va_copy</tt>' intrinsic works just like the <tt>va_copy</tt> macro
3531 available in C. In a target-dependent way, it copies the source
3532 <tt>va_list</tt> element into the destination list. This intrinsic is necessary
3533 because the <tt><a href="i_va_begin">llvm.va_begin</a></tt> intrinsic may be
3534 arbitrarily complex and require memory allocation, for example.</p>
3538 <!-- ======================================================================= -->
3539 <div class="doc_subsection">
3540 <a name="int_gc">Accurate Garbage Collection Intrinsics</a>
3543 <div class="doc_text">
3546 LLVM support for <a href="GarbageCollection.html">Accurate Garbage
3547 Collection</a> requires the implementation and generation of these intrinsics.
3548 These intrinsics allow identification of <a href="#i_gcroot">GC roots on the
3549 stack</a>, as well as garbage collector implementations that require <a
3550 href="#i_gcread">read</a> and <a href="#i_gcwrite">write</a> barriers.
3551 Front-ends for type-safe garbage collected languages should generate these
3552 intrinsics to make use of the LLVM garbage collectors. For more details, see <a
3553 href="GarbageCollection.html">Accurate Garbage Collection with LLVM</a>.
3557 <!-- _______________________________________________________________________ -->
3558 <div class="doc_subsubsection">
3559 <a name="i_gcroot">'<tt>llvm.gcroot</tt>' Intrinsic</a>
3562 <div class="doc_text">
3567 declare void %llvm.gcroot(<ty>** %ptrloc, <ty2>* %metadata)
3572 <p>The '<tt>llvm.gcroot</tt>' intrinsic declares the existence of a GC root to
3573 the code generator, and allows some metadata to be associated with it.</p>
3577 <p>The first argument specifies the address of a stack object that contains the
3578 root pointer. The second pointer (which must be either a constant or a global
3579 value address) contains the meta-data to be associated with the root.</p>
3583 <p>At runtime, a call to this intrinsics stores a null pointer into the "ptrloc"
3584 location. At compile-time, the code generator generates information to allow
3585 the runtime to find the pointer at GC safe points.
3591 <!-- _______________________________________________________________________ -->
3592 <div class="doc_subsubsection">
3593 <a name="i_gcread">'<tt>llvm.gcread</tt>' Intrinsic</a>
3596 <div class="doc_text">
3601 declare sbyte* %llvm.gcread(sbyte* %ObjPtr, sbyte** %Ptr)
3606 <p>The '<tt>llvm.gcread</tt>' intrinsic identifies reads of references from heap
3607 locations, allowing garbage collector implementations that require read
3612 <p>The second argument is the address to read from, which should be an address
3613 allocated from the garbage collector. The first object is a pointer to the
3614 start of the referenced object, if needed by the language runtime (otherwise
3619 <p>The '<tt>llvm.gcread</tt>' intrinsic has the same semantics as a load
3620 instruction, but may be replaced with substantially more complex code by the
3621 garbage collector runtime, as needed.</p>
3626 <!-- _______________________________________________________________________ -->
3627 <div class="doc_subsubsection">
3628 <a name="i_gcwrite">'<tt>llvm.gcwrite</tt>' Intrinsic</a>
3631 <div class="doc_text">
3636 declare void %llvm.gcwrite(sbyte* %P1, sbyte* %Obj, sbyte** %P2)
3641 <p>The '<tt>llvm.gcwrite</tt>' intrinsic identifies writes of references to heap
3642 locations, allowing garbage collector implementations that require write
3643 barriers (such as generational or reference counting collectors).</p>
3647 <p>The first argument is the reference to store, the second is the start of the
3648 object to store it to, and the third is the address of the field of Obj to
3649 store to. If the runtime does not require a pointer to the object, Obj may be
3654 <p>The '<tt>llvm.gcwrite</tt>' intrinsic has the same semantics as a store
3655 instruction, but may be replaced with substantially more complex code by the
3656 garbage collector runtime, as needed.</p>
3662 <!-- ======================================================================= -->
3663 <div class="doc_subsection">
3664 <a name="int_codegen">Code Generator Intrinsics</a>
3667 <div class="doc_text">
3669 These intrinsics are provided by LLVM to expose special features that may only
3670 be implemented with code generator support.
3675 <!-- _______________________________________________________________________ -->
3676 <div class="doc_subsubsection">
3677 <a name="i_returnaddress">'<tt>llvm.returnaddress</tt>' Intrinsic</a>
3680 <div class="doc_text">
3684 declare sbyte *%llvm.returnaddress(uint <level>)
3690 The '<tt>llvm.returnaddress</tt>' intrinsic attempts to compute a
3691 target-specific value indicating the return address of the current function
3692 or one of its callers.
3698 The argument to this intrinsic indicates which function to return the address
3699 for. Zero indicates the calling function, one indicates its caller, etc. The
3700 argument is <b>required</b> to be a constant integer value.
3706 The '<tt>llvm.returnaddress</tt>' intrinsic either returns a pointer indicating
3707 the return address of the specified call frame, or zero if it cannot be
3708 identified. The value returned by this intrinsic is likely to be incorrect or 0
3709 for arguments other than zero, so it should only be used for debugging purposes.
3713 Note that calling this intrinsic does not prevent function inlining or other
3714 aggressive transformations, so the value returned may not be that of the obvious
3715 source-language caller.
3720 <!-- _______________________________________________________________________ -->
3721 <div class="doc_subsubsection">
3722 <a name="i_frameaddress">'<tt>llvm.frameaddress</tt>' Intrinsic</a>
3725 <div class="doc_text">
3729 declare sbyte *%llvm.frameaddress(uint <level>)
3735 The '<tt>llvm.frameaddress</tt>' intrinsic attempts to return the
3736 target-specific frame pointer value for the specified stack frame.
3742 The argument to this intrinsic indicates which function to return the frame
3743 pointer for. Zero indicates the calling function, one indicates its caller,
3744 etc. The argument is <b>required</b> to be a constant integer value.
3750 The '<tt>llvm.frameaddress</tt>' intrinsic either returns a pointer indicating
3751 the frame address of the specified call frame, or zero if it cannot be
3752 identified. The value returned by this intrinsic is likely to be incorrect or 0
3753 for arguments other than zero, so it should only be used for debugging purposes.
3757 Note that calling this intrinsic does not prevent function inlining or other
3758 aggressive transformations, so the value returned may not be that of the obvious
3759 source-language caller.
3763 <!-- _______________________________________________________________________ -->
3764 <div class="doc_subsubsection">
3765 <a name="i_stacksave">'<tt>llvm.stacksave</tt>' Intrinsic</a>
3768 <div class="doc_text">
3772 declare sbyte *%llvm.stacksave()
3778 The '<tt>llvm.stacksave</tt>' intrinsic is used to remember the current state of
3779 the function stack, for use with <a href="#i_stackrestore">
3780 <tt>llvm.stackrestore</tt></a>. This is useful for implementing language
3781 features like scoped automatic variable sized arrays in C99.
3787 This intrinsic returns a opaque pointer value that can be passed to <a
3788 href="#i_stackrestore"><tt>llvm.stackrestore</tt></a>. When an
3789 <tt>llvm.stackrestore</tt> intrinsic is executed with a value saved from
3790 <tt>llvm.stacksave</tt>, it effectively restores the state of the stack to the
3791 state it was in when the <tt>llvm.stacksave</tt> intrinsic executed. In
3792 practice, this pops any <a href="#i_alloca">alloca</a> blocks from the stack
3793 that were allocated after the <tt>llvm.stacksave</tt> was executed.
3798 <!-- _______________________________________________________________________ -->
3799 <div class="doc_subsubsection">
3800 <a name="i_stackrestore">'<tt>llvm.stackrestore</tt>' Intrinsic</a>
3803 <div class="doc_text">
3807 declare void %llvm.stackrestore(sbyte* %ptr)
3813 The '<tt>llvm.stackrestore</tt>' intrinsic is used to restore the state of
3814 the function stack to the state it was in when the corresponding <a
3815 href="#llvm.stacksave"><tt>llvm.stacksave</tt></a> intrinsic executed. This is
3816 useful for implementing language features like scoped automatic variable sized
3823 See the description for <a href="#i_stacksave"><tt>llvm.stacksave</tt></a>.
3829 <!-- _______________________________________________________________________ -->
3830 <div class="doc_subsubsection">
3831 <a name="i_prefetch">'<tt>llvm.prefetch</tt>' Intrinsic</a>
3834 <div class="doc_text">
3838 declare void %llvm.prefetch(sbyte * <address>,
3839 uint <rw>, uint <locality>)
3846 The '<tt>llvm.prefetch</tt>' intrinsic is a hint to the code generator to insert
3847 a prefetch instruction if supported; otherwise, it is a noop. Prefetches have
3849 effect on the behavior of the program but can change its performance
3856 <tt>address</tt> is the address to be prefetched, <tt>rw</tt> is the specifier
3857 determining if the fetch should be for a read (0) or write (1), and
3858 <tt>locality</tt> is a temporal locality specifier ranging from (0) - no
3859 locality, to (3) - extremely local keep in cache. The <tt>rw</tt> and
3860 <tt>locality</tt> arguments must be constant integers.
3866 This intrinsic does not modify the behavior of the program. In particular,
3867 prefetches cannot trap and do not produce a value. On targets that support this
3868 intrinsic, the prefetch can provide hints to the processor cache for better
3874 <!-- _______________________________________________________________________ -->
3875 <div class="doc_subsubsection">
3876 <a name="i_pcmarker">'<tt>llvm.pcmarker</tt>' Intrinsic</a>
3879 <div class="doc_text">
3883 declare void %llvm.pcmarker( uint <id> )
3890 The '<tt>llvm.pcmarker</tt>' intrinsic is a method to export a Program Counter
3892 code to simulators and other tools. The method is target specific, but it is
3893 expected that the marker will use exported symbols to transmit the PC of the marker.
3894 The marker makes no guarantees that it will remain with any specific instruction
3895 after optimizations. It is possible that the presence of a marker will inhibit
3896 optimizations. The intended use is to be inserted after optimizations to allow
3897 correlations of simulation runs.
3903 <tt>id</tt> is a numerical id identifying the marker.
3909 This intrinsic does not modify the behavior of the program. Backends that do not
3910 support this intrinisic may ignore it.
3915 <!-- _______________________________________________________________________ -->
3916 <div class="doc_subsubsection">
3917 <a name="i_readcyclecounter">'<tt>llvm.readcyclecounter</tt>' Intrinsic</a>
3920 <div class="doc_text">
3924 declare ulong %llvm.readcyclecounter( )
3931 The '<tt>llvm.readcyclecounter</tt>' intrinsic provides access to the cycle
3932 counter register (or similar low latency, high accuracy clocks) on those targets
3933 that support it. On X86, it should map to RDTSC. On Alpha, it should map to RPCC.
3934 As the backing counters overflow quickly (on the order of 9 seconds on alpha), this
3935 should only be used for small timings.
3941 When directly supported, reading the cycle counter should not modify any memory.
3942 Implementations are allowed to either return a application specific value or a
3943 system wide value. On backends without support, this is lowered to a constant 0.
3948 <!-- ======================================================================= -->
3949 <div class="doc_subsection">
3950 <a name="int_libc">Standard C Library Intrinsics</a>
3953 <div class="doc_text">
3955 LLVM provides intrinsics for a few important standard C library functions.
3956 These intrinsics allow source-language front-ends to pass information about the
3957 alignment of the pointer arguments to the code generator, providing opportunity
3958 for more efficient code generation.
3963 <!-- _______________________________________________________________________ -->
3964 <div class="doc_subsubsection">
3965 <a name="i_memcpy">'<tt>llvm.memcpy</tt>' Intrinsic</a>
3968 <div class="doc_text">
3972 declare void %llvm.memcpy.i32(sbyte* <dest>, sbyte* <src>,
3973 uint <len>, uint <align>)
3974 declare void %llvm.memcpy.i64(sbyte* <dest>, sbyte* <src>,
3975 ulong <len>, uint <align>)
3981 The '<tt>llvm.memcpy.*</tt>' intrinsics copy a block of memory from the source
3982 location to the destination location.
3986 Note that, unlike the standard libc function, the <tt>llvm.memcpy.*</tt>
3987 intrinsics do not return a value, and takes an extra alignment argument.
3993 The first argument is a pointer to the destination, the second is a pointer to
3994 the source. The third argument is an integer argument
3995 specifying the number of bytes to copy, and the fourth argument is the alignment
3996 of the source and destination locations.
4000 If the call to this intrinisic has an alignment value that is not 0 or 1, then
4001 the caller guarantees that both the source and destination pointers are aligned
4008 The '<tt>llvm.memcpy.*</tt>' intrinsics copy a block of memory from the source
4009 location to the destination location, which are not allowed to overlap. It
4010 copies "len" bytes of memory over. If the argument is known to be aligned to
4011 some boundary, this can be specified as the fourth argument, otherwise it should
4017 <!-- _______________________________________________________________________ -->
4018 <div class="doc_subsubsection">
4019 <a name="i_memmove">'<tt>llvm.memmove</tt>' Intrinsic</a>
4022 <div class="doc_text">
4026 declare void %llvm.memmove.i32(sbyte* <dest>, sbyte* <src>,
4027 uint <len>, uint <align>)
4028 declare void %llvm.memmove.i64(sbyte* <dest>, sbyte* <src>,
4029 ulong <len>, uint <align>)
4035 The '<tt>llvm.memmove.*</tt>' intrinsics move a block of memory from the source
4036 location to the destination location. It is similar to the
4037 '<tt>llvm.memcmp</tt>' intrinsic but allows the two memory locations to overlap.
4041 Note that, unlike the standard libc function, the <tt>llvm.memmove.*</tt>
4042 intrinsics do not return a value, and takes an extra alignment argument.
4048 The first argument is a pointer to the destination, the second is a pointer to
4049 the source. The third argument is an integer argument
4050 specifying the number of bytes to copy, and the fourth argument is the alignment
4051 of the source and destination locations.
4055 If the call to this intrinisic has an alignment value that is not 0 or 1, then
4056 the caller guarantees that the source and destination pointers are aligned to
4063 The '<tt>llvm.memmove.*</tt>' intrinsics copy a block of memory from the source
4064 location to the destination location, which may overlap. It
4065 copies "len" bytes of memory over. If the argument is known to be aligned to
4066 some boundary, this can be specified as the fourth argument, otherwise it should
4072 <!-- _______________________________________________________________________ -->
4073 <div class="doc_subsubsection">
4074 <a name="i_memset">'<tt>llvm.memset.*</tt>' Intrinsics</a>
4077 <div class="doc_text">
4081 declare void %llvm.memset.i32(sbyte* <dest>, ubyte <val>,
4082 uint <len>, uint <align>)
4083 declare void %llvm.memset.i64(sbyte* <dest>, ubyte <val>,
4084 ulong <len>, uint <align>)
4090 The '<tt>llvm.memset.*</tt>' intrinsics fill a block of memory with a particular
4095 Note that, unlike the standard libc function, the <tt>llvm.memset</tt> intrinsic
4096 does not return a value, and takes an extra alignment argument.
4102 The first argument is a pointer to the destination to fill, the second is the
4103 byte value to fill it with, the third argument is an integer
4104 argument specifying the number of bytes to fill, and the fourth argument is the
4105 known alignment of destination location.
4109 If the call to this intrinisic has an alignment value that is not 0 or 1, then
4110 the caller guarantees that the destination pointer is aligned to that boundary.
4116 The '<tt>llvm.memset.*</tt>' intrinsics fill "len" bytes of memory starting at
4118 destination location. If the argument is known to be aligned to some boundary,
4119 this can be specified as the fourth argument, otherwise it should be set to 0 or
4125 <!-- _______________________________________________________________________ -->
4126 <div class="doc_subsubsection">
4127 <a name="i_isunordered">'<tt>llvm.isunordered.*</tt>' Intrinsic</a>
4130 <div class="doc_text">
4134 declare bool %llvm.isunordered.f32(float Val1, float Val2)
4135 declare bool %llvm.isunordered.f64(double Val1, double Val2)
4141 The '<tt>llvm.isunordered</tt>' intrinsics return true if either or both of the
4142 specified floating point values is a NAN.
4148 The arguments are floating point numbers of the same type.
4154 If either or both of the arguments is a SNAN or QNAN, it returns true, otherwise
4160 <!-- _______________________________________________________________________ -->
4161 <div class="doc_subsubsection">
4162 <a name="i_sqrt">'<tt>llvm.sqrt.*</tt>' Intrinsic</a>
4165 <div class="doc_text">
4169 declare float %llvm.sqrt.f32(float %Val)
4170 declare double %llvm.sqrt.f64(double %Val)
4176 The '<tt>llvm.sqrt</tt>' intrinsics return the sqrt of the specified operand,
4177 returning the same value as the libm '<tt>sqrt</tt>' function would. Unlike
4178 <tt>sqrt</tt> in libm, however, <tt>llvm.sqrt</tt> has undefined behavior for
4179 negative numbers (which allows for better optimization).
4185 The argument and return value are floating point numbers of the same type.
4191 This function returns the sqrt of the specified operand if it is a positive
4192 floating point number.
4196 <!-- _______________________________________________________________________ -->
4197 <div class="doc_subsubsection">
4198 <a name="i_powi">'<tt>llvm.powi.*</tt>' Intrinsic</a>
4201 <div class="doc_text">
4205 declare float %llvm.powi.f32(float %Val, int %power)
4206 declare double %llvm.powi.f64(double %Val, int %power)
4212 The '<tt>llvm.powi.*</tt>' intrinsics return the first operand raised to the
4213 specified (positive or negative) power. The order of evaluation of
4214 multiplications is not defined.
4220 The second argument is an integer power, and the first is a value to raise to
4227 This function returns the first value raised to the second power with an
4228 unspecified sequence of rounding operations.</p>
4232 <!-- ======================================================================= -->
4233 <div class="doc_subsection">
4234 <a name="int_manip">Bit Manipulation Intrinsics</a>
4237 <div class="doc_text">
4239 LLVM provides intrinsics for a few important bit manipulation operations.
4240 These allow efficient code generation for some algorithms.
4245 <!-- _______________________________________________________________________ -->
4246 <div class="doc_subsubsection">
4247 <a name="i_bswap">'<tt>llvm.bswap.*</tt>' Intrinsics</a>
4250 <div class="doc_text">
4254 declare ushort %llvm.bswap.i16(ushort <id>)
4255 declare uint %llvm.bswap.i32(uint <id>)
4256 declare ulong %llvm.bswap.i64(ulong <id>)
4262 The '<tt>llvm.bwsap</tt>' family of intrinsics is used to byteswap a 16, 32 or
4263 64 bit quantity. These are useful for performing operations on data that is not
4264 in the target's native byte order.
4270 The <tt>llvm.bswap.16</tt> intrinsic returns a ushort value that has the high and low
4271 byte of the input ushort swapped. Similarly, the <tt>llvm.bswap.i32</tt> intrinsic
4272 returns a uint value that has the four bytes of the input uint swapped, so that
4273 if the input bytes are numbered 0, 1, 2, 3 then the returned uint will have its
4274 bytes in 3, 2, 1, 0 order. The <tt>llvm.bswap.i64</tt> intrinsic extends this concept
4280 <!-- _______________________________________________________________________ -->
4281 <div class="doc_subsubsection">
4282 <a name="int_ctpop">'<tt>llvm.ctpop.*</tt>' Intrinsic</a>
4285 <div class="doc_text">
4289 declare ubyte %llvm.ctpop.i8 (ubyte <src>)
4290 declare ushort %llvm.ctpop.i16(ushort <src>)
4291 declare uint %llvm.ctpop.i32(uint <src>)
4292 declare ulong %llvm.ctpop.i64(ulong <src>)
4298 The '<tt>llvm.ctpop</tt>' family of intrinsics counts the number of bits set in a
4305 The only argument is the value to be counted. The argument may be of any
4306 unsigned integer type. The return type must match the argument type.
4312 The '<tt>llvm.ctpop</tt>' intrinsic counts the 1's in a variable.
4316 <!-- _______________________________________________________________________ -->
4317 <div class="doc_subsubsection">
4318 <a name="int_ctlz">'<tt>llvm.ctlz.*</tt>' Intrinsic</a>
4321 <div class="doc_text">
4325 declare ubyte %llvm.ctlz.i8 (ubyte <src>)
4326 declare ushort %llvm.ctlz.i16(ushort <src>)
4327 declare uint %llvm.ctlz.i32(uint <src>)
4328 declare ulong %llvm.ctlz.i64(ulong <src>)
4334 The '<tt>llvm.ctlz</tt>' family of intrinsic functions counts the number of
4335 leading zeros in a variable.
4341 The only argument is the value to be counted. The argument may be of any
4342 unsigned integer type. The return type must match the argument type.
4348 The '<tt>llvm.ctlz</tt>' intrinsic counts the leading (most significant) zeros
4349 in a variable. If the src == 0 then the result is the size in bits of the type
4350 of src. For example, <tt>llvm.ctlz(int 2) = 30</tt>.
4356 <!-- _______________________________________________________________________ -->
4357 <div class="doc_subsubsection">
4358 <a name="int_cttz">'<tt>llvm.cttz.*</tt>' Intrinsic</a>
4361 <div class="doc_text">
4365 declare ubyte %llvm.cttz.i8 (ubyte <src>)
4366 declare ushort %llvm.cttz.i16(ushort <src>)
4367 declare uint %llvm.cttz.i32(uint <src>)
4368 declare ulong %llvm.cttz.i64(ulong <src>)
4374 The '<tt>llvm.cttz</tt>' family of intrinsic functions counts the number of
4381 The only argument is the value to be counted. The argument may be of any
4382 unsigned integer type. The return type must match the argument type.
4388 The '<tt>llvm.cttz</tt>' intrinsic counts the trailing (least significant) zeros
4389 in a variable. If the src == 0 then the result is the size in bits of the type
4390 of src. For example, <tt>llvm.cttz(2) = 1</tt>.
4394 <!-- ======================================================================= -->
4395 <div class="doc_subsection">
4396 <a name="int_debugger">Debugger Intrinsics</a>
4399 <div class="doc_text">
4401 The LLVM debugger intrinsics (which all start with <tt>llvm.dbg.</tt> prefix),
4402 are described in the <a
4403 href="SourceLevelDebugging.html#format_common_intrinsics">LLVM Source Level
4404 Debugging</a> document.
4409 <!-- *********************************************************************** -->
4412 <a href="http://jigsaw.w3.org/css-validator/check/referer"><img
4413 src="http://jigsaw.w3.org/css-validator/images/vcss" alt="Valid CSS!"></a>
4414 <a href="http://validator.w3.org/check/referer"><img
4415 src="http://www.w3.org/Icons/valid-html401" alt="Valid HTML 4.01!" /></a>
4417 <a href="mailto:sabre@nondot.org">Chris Lattner</a><br>
4418 <a href="http://llvm.org">The LLVM Compiler Infrastructure</a><br>
4419 Last modified: $Date$