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_div">'<tt>div</tt>' Instruction</a></li>
81 <li><a href="#i_rem">'<tt>rem</tt>' Instruction</a></li>
82 <li><a href="#i_setcc">'<tt>set<i>cc</i></tt>' Instructions</a></li>
85 <li><a href="#bitwiseops">Bitwise Binary Operations</a>
87 <li><a href="#i_and">'<tt>and</tt>' Instruction</a></li>
88 <li><a href="#i_or">'<tt>or</tt>' Instruction</a></li>
89 <li><a href="#i_xor">'<tt>xor</tt>' Instruction</a></li>
90 <li><a href="#i_shl">'<tt>shl</tt>' Instruction</a></li>
91 <li><a href="#i_shr">'<tt>shr</tt>' Instruction</a></li>
94 <li><a href="#vectorops">Vector Operations</a>
96 <li><a href="#i_extractelement">'<tt>extractelement</tt>' Instruction</a></li>
97 <li><a href="#i_insertelement">'<tt>insertelement</tt>' Instruction</a></li>
98 <li><a href="#i_shufflevector">'<tt>shufflevector</tt>' Instruction</a></li>
99 <li><a href="#i_vsetint">'<tt>vsetint</tt>' Instruction</a></li>
100 <li><a href="#i_vsetfp">'<tt>vsetfp</tt>' Instruction</a></li>
101 <li><a href="#i_vselect">'<tt>vselect</tt>' Instruction</a></li>
104 <li><a href="#memoryops">Memory Access and Addressing Operations</a>
106 <li><a href="#i_malloc">'<tt>malloc</tt>' Instruction</a></li>
107 <li><a href="#i_free">'<tt>free</tt>' Instruction</a></li>
108 <li><a href="#i_alloca">'<tt>alloca</tt>' Instruction</a></li>
109 <li><a href="#i_load">'<tt>load</tt>' Instruction</a></li>
110 <li><a href="#i_store">'<tt>store</tt>' Instruction</a></li>
111 <li><a href="#i_getelementptr">'<tt>getelementptr</tt>' Instruction</a></li>
114 <li><a href="#otherops">Other Operations</a>
116 <li><a href="#i_phi">'<tt>phi</tt>' Instruction</a></li>
117 <li><a href="#i_cast">'<tt>cast .. to</tt>' Instruction</a></li>
118 <li><a href="#i_select">'<tt>select</tt>' Instruction</a></li>
119 <li><a href="#i_call">'<tt>call</tt>' Instruction</a></li>
120 <li><a href="#i_va_arg">'<tt>va_arg</tt>' Instruction</a></li>
125 <li><a href="#intrinsics">Intrinsic Functions</a>
127 <li><a href="#int_varargs">Variable Argument Handling Intrinsics</a>
129 <li><a href="#i_va_start">'<tt>llvm.va_start</tt>' Intrinsic</a></li>
130 <li><a href="#i_va_end">'<tt>llvm.va_end</tt>' Intrinsic</a></li>
131 <li><a href="#i_va_copy">'<tt>llvm.va_copy</tt>' Intrinsic</a></li>
134 <li><a href="#int_gc">Accurate Garbage Collection Intrinsics</a>
136 <li><a href="#i_gcroot">'<tt>llvm.gcroot</tt>' Intrinsic</a></li>
137 <li><a href="#i_gcread">'<tt>llvm.gcread</tt>' Intrinsic</a></li>
138 <li><a href="#i_gcwrite">'<tt>llvm.gcwrite</tt>' Intrinsic</a></li>
141 <li><a href="#int_codegen">Code Generator Intrinsics</a>
143 <li><a href="#i_returnaddress">'<tt>llvm.returnaddress</tt>' Intrinsic</a></li>
144 <li><a href="#i_frameaddress">'<tt>llvm.frameaddress</tt>' Intrinsic</a></li>
145 <li><a href="#i_stacksave">'<tt>llvm.stacksave</tt>' Intrinsic</a></li>
146 <li><a href="#i_stackrestore">'<tt>llvm.stackrestore</tt>' Intrinsic</a></li>
147 <li><a href="#i_prefetch">'<tt>llvm.prefetch</tt>' Intrinsic</a></li>
148 <li><a href="#i_pcmarker">'<tt>llvm.pcmarker</tt>' Intrinsic</a></li>
149 <li><a href="#i_readcyclecounter"><tt>llvm.readcyclecounter</tt>' Intrinsic</a></li>
152 <li><a href="#int_libc">Standard C Library Intrinsics</a>
154 <li><a href="#i_memcpy">'<tt>llvm.memcpy.*</tt>' Intrinsic</a></li>
155 <li><a href="#i_memmove">'<tt>llvm.memmove.*</tt>' Intrinsic</a></li>
156 <li><a href="#i_memset">'<tt>llvm.memset.*</tt>' Intrinsic</a></li>
157 <li><a href="#i_isunordered">'<tt>llvm.isunordered.*</tt>' Intrinsic</a></li>
158 <li><a href="#i_sqrt">'<tt>llvm.sqrt.*</tt>' Intrinsic</a></li>
159 <li><a href="#i_powi">'<tt>llvm.powi.*</tt>' Intrinsic</a></li>
162 <li><a href="#int_manip">Bit Manipulation Intrinsics</a>
164 <li><a href="#i_bswap">'<tt>llvm.bswap.*</tt>' Intrinsics</a></li>
165 <li><a href="#int_ctpop">'<tt>llvm.ctpop.*</tt>' Intrinsic </a></li>
166 <li><a href="#int_ctlz">'<tt>llvm.ctlz.*</tt>' Intrinsic </a></li>
167 <li><a href="#int_cttz">'<tt>llvm.cttz.*</tt>' Intrinsic </a></li>
170 <li><a href="#int_debugger">Debugger intrinsics</a></li>
175 <div class="doc_author">
176 <p>Written by <a href="mailto:sabre@nondot.org">Chris Lattner</a>
177 and <a href="mailto:vadve@cs.uiuc.edu">Vikram Adve</a></p>
180 <!-- *********************************************************************** -->
181 <div class="doc_section"> <a name="abstract">Abstract </a></div>
182 <!-- *********************************************************************** -->
184 <div class="doc_text">
185 <p>This document is a reference manual for the LLVM assembly language.
186 LLVM is an SSA based representation that provides type safety,
187 low-level operations, flexibility, and the capability of representing
188 'all' high-level languages cleanly. It is the common code
189 representation used throughout all phases of the LLVM compilation
193 <!-- *********************************************************************** -->
194 <div class="doc_section"> <a name="introduction">Introduction</a> </div>
195 <!-- *********************************************************************** -->
197 <div class="doc_text">
199 <p>The LLVM code representation is designed to be used in three
200 different forms: as an in-memory compiler IR, as an on-disk bytecode
201 representation (suitable for fast loading by a Just-In-Time compiler),
202 and as a human readable assembly language representation. This allows
203 LLVM to provide a powerful intermediate representation for efficient
204 compiler transformations and analysis, while providing a natural means
205 to debug and visualize the transformations. The three different forms
206 of LLVM are all equivalent. This document describes the human readable
207 representation and notation.</p>
209 <p>The LLVM representation aims to be light-weight and low-level
210 while being expressive, typed, and extensible at the same time. It
211 aims to be a "universal IR" of sorts, by being at a low enough level
212 that high-level ideas may be cleanly mapped to it (similar to how
213 microprocessors are "universal IR's", allowing many source languages to
214 be mapped to them). By providing type information, LLVM can be used as
215 the target of optimizations: for example, through pointer analysis, it
216 can be proven that a C automatic variable is never accessed outside of
217 the current function... allowing it to be promoted to a simple SSA
218 value instead of a memory location.</p>
222 <!-- _______________________________________________________________________ -->
223 <div class="doc_subsubsection"> <a name="wellformed">Well-Formedness</a> </div>
225 <div class="doc_text">
227 <p>It is important to note that this document describes 'well formed'
228 LLVM assembly language. There is a difference between what the parser
229 accepts and what is considered 'well formed'. For example, the
230 following instruction is syntactically okay, but not well formed:</p>
233 %x = <a href="#i_add">add</a> int 1, %x
236 <p>...because the definition of <tt>%x</tt> does not dominate all of
237 its uses. The LLVM infrastructure provides a verification pass that may
238 be used to verify that an LLVM module is well formed. This pass is
239 automatically run by the parser after parsing input assembly and by
240 the optimizer before it outputs bytecode. The violations pointed out
241 by the verifier pass indicate bugs in transformation passes or input to
244 <!-- Describe the typesetting conventions here. --> </div>
246 <!-- *********************************************************************** -->
247 <div class="doc_section"> <a name="identifiers">Identifiers</a> </div>
248 <!-- *********************************************************************** -->
250 <div class="doc_text">
252 <p>LLVM uses three different forms of identifiers, for different
256 <li>Named values are represented as a string of characters with a '%' prefix.
257 For example, %foo, %DivisionByZero, %a.really.long.identifier. The actual
258 regular expression used is '<tt>%[a-zA-Z$._][a-zA-Z$._0-9]*</tt>'.
259 Identifiers which require other characters in their names can be surrounded
260 with quotes. In this way, anything except a <tt>"</tt> character can be used
263 <li>Unnamed values are represented as an unsigned numeric value with a '%'
264 prefix. For example, %12, %2, %44.</li>
266 <li>Constants, which are described in a <a href="#constants">section about
267 constants</a>, below.</li>
270 <p>LLVM requires that values start with a '%' sign for two reasons: Compilers
271 don't need to worry about name clashes with reserved words, and the set of
272 reserved words may be expanded in the future without penalty. Additionally,
273 unnamed identifiers allow a compiler to quickly come up with a temporary
274 variable without having to avoid symbol table conflicts.</p>
276 <p>Reserved words in LLVM are very similar to reserved words in other
277 languages. There are keywords for different opcodes ('<tt><a
278 href="#i_add">add</a></tt>', '<tt><a href="#i_cast">cast</a></tt>', '<tt><a
279 href="#i_ret">ret</a></tt>', etc...), for primitive type names ('<tt><a
280 href="#t_void">void</a></tt>', '<tt><a href="#t_uint">uint</a></tt>', etc...),
281 and others. These reserved words cannot conflict with variable names, because
282 none of them start with a '%' character.</p>
284 <p>Here is an example of LLVM code to multiply the integer variable
285 '<tt>%X</tt>' by 8:</p>
290 %result = <a href="#i_mul">mul</a> uint %X, 8
293 <p>After strength reduction:</p>
296 %result = <a href="#i_shl">shl</a> uint %X, ubyte 3
299 <p>And the hard way:</p>
302 <a href="#i_add">add</a> uint %X, %X <i>; yields {uint}:%0</i>
303 <a href="#i_add">add</a> uint %0, %0 <i>; yields {uint}:%1</i>
304 %result = <a href="#i_add">add</a> uint %1, %1
307 <p>This last way of multiplying <tt>%X</tt> by 8 illustrates several
308 important lexical features of LLVM:</p>
312 <li>Comments are delimited with a '<tt>;</tt>' and go until the end of
315 <li>Unnamed temporaries are created when the result of a computation is not
316 assigned to a named value.</li>
318 <li>Unnamed temporaries are numbered sequentially</li>
322 <p>...and it also shows a convention that we follow in this document. When
323 demonstrating instructions, we will follow an instruction with a comment that
324 defines the type and name of value produced. Comments are shown in italic
329 <!-- *********************************************************************** -->
330 <div class="doc_section"> <a name="highlevel">High Level Structure</a> </div>
331 <!-- *********************************************************************** -->
333 <!-- ======================================================================= -->
334 <div class="doc_subsection"> <a name="modulestructure">Module Structure</a>
337 <div class="doc_text">
339 <p>LLVM programs are composed of "Module"s, each of which is a
340 translation unit of the input programs. Each module consists of
341 functions, global variables, and symbol table entries. Modules may be
342 combined together with the LLVM linker, which merges function (and
343 global variable) definitions, resolves forward declarations, and merges
344 symbol table entries. Here is an example of the "hello world" module:</p>
346 <pre><i>; Declare the string constant as a global constant...</i>
347 <a href="#identifiers">%.LC0</a> = <a href="#linkage_internal">internal</a> <a
348 href="#globalvars">constant</a> <a href="#t_array">[13 x sbyte]</a> c"hello world\0A\00" <i>; [13 x sbyte]*</i>
350 <i>; External declaration of the puts function</i>
351 <a href="#functionstructure">declare</a> int %puts(sbyte*) <i>; int(sbyte*)* </i>
353 <i>; Global variable / Function body section separator</i>
356 <i>; Definition of main function</i>
357 int %main() { <i>; int()* </i>
358 <i>; Convert [13x sbyte]* to sbyte *...</i>
360 href="#i_getelementptr">getelementptr</a> [13 x sbyte]* %.LC0, long 0, long 0 <i>; sbyte*</i>
362 <i>; Call puts function to write out the string to stdout...</i>
364 href="#i_call">call</a> int %puts(sbyte* %cast210) <i>; int</i>
366 href="#i_ret">ret</a> int 0<br>}<br></pre>
368 <p>This example is made up of a <a href="#globalvars">global variable</a>
369 named "<tt>.LC0</tt>", an external declaration of the "<tt>puts</tt>"
370 function, and a <a href="#functionstructure">function definition</a>
371 for "<tt>main</tt>".</p>
373 <p>In general, a module is made up of a list of global values,
374 where both functions and global variables are global values. Global values are
375 represented by a pointer to a memory location (in this case, a pointer to an
376 array of char, and a pointer to a function), and have one of the following <a
377 href="#linkage">linkage types</a>.</p>
379 <p>Due to a limitation in the current LLVM assembly parser (it is limited by
380 one-token lookahead), modules are split into two pieces by the "implementation"
381 keyword. Global variable prototypes and definitions must occur before the
382 keyword, and function definitions must occur after it. Function prototypes may
383 occur either before or after it. In the future, the implementation keyword may
384 become a noop, if the parser gets smarter.</p>
388 <!-- ======================================================================= -->
389 <div class="doc_subsection">
390 <a name="linkage">Linkage Types</a>
393 <div class="doc_text">
396 All Global Variables and Functions have one of the following types of linkage:
401 <dt><tt><b><a name="linkage_internal">internal</a></b></tt> </dt>
403 <dd>Global values with internal linkage are only directly accessible by
404 objects in the current module. In particular, linking code into a module with
405 an internal global value may cause the internal to be renamed as necessary to
406 avoid collisions. Because the symbol is internal to the module, all
407 references can be updated. This corresponds to the notion of the
408 '<tt>static</tt>' keyword in C, or the idea of "anonymous namespaces" in C++.
411 <dt><tt><b><a name="linkage_linkonce">linkonce</a></b></tt>: </dt>
413 <dd>"<tt>linkonce</tt>" linkage is similar to <tt>internal</tt> linkage, with
414 the twist that linking together two modules defining the same
415 <tt>linkonce</tt> globals will cause one of the globals to be discarded. This
416 is typically used to implement inline functions. Unreferenced
417 <tt>linkonce</tt> globals are allowed to be discarded.
420 <dt><tt><b><a name="linkage_weak">weak</a></b></tt>: </dt>
422 <dd>"<tt>weak</tt>" linkage is exactly the same as <tt>linkonce</tt> linkage,
423 except that unreferenced <tt>weak</tt> globals may not be discarded. This is
424 used to implement constructs in C such as "<tt>int X;</tt>" at global scope.
427 <dt><tt><b><a name="linkage_appending">appending</a></b></tt>: </dt>
429 <dd>"<tt>appending</tt>" linkage may only be applied to global variables of
430 pointer to array type. When two global variables with appending linkage are
431 linked together, the two global arrays are appended together. This is the
432 LLVM, typesafe, equivalent of having the system linker append together
433 "sections" with identical names when .o files are linked.
436 <dt><tt><b><a name="linkage_external">externally visible</a></b></tt>:</dt>
438 <dd>If none of the above identifiers are used, the global is externally
439 visible, meaning that it participates in linkage and can be used to resolve
440 external symbol references.
443 <dt><tt><b><a name="linkage_externweak">extern_weak</a></b></tt>: </dt>
445 <dd>"<tt>extern_weak</tt>" TBD
449 The next two types of linkage are targeted for Microsoft Windows platform
450 only. They are designed to support importing (exporting) symbols from (to)
454 <dt><tt><b><a name="linkage_dllimport">dllimport</a></b></tt>: </dt>
456 <dd>"<tt>dllimport</tt>" linkage causes the compiler to reference a function
457 or variable via a global pointer to a pointer that is set up by the DLL
458 exporting the symbol. On Microsoft Windows targets, the pointer name is
459 formed by combining <code>_imp__</code> and the function or variable name.
462 <dt><tt><b><a name="linkage_dllexport">dllexport</a></b></tt>: </dt>
464 <dd>"<tt>dllexport</tt>" linkage causes the compiler to provide a global
465 pointer to a pointer in a DLL, so that it can be referenced with the
466 <tt>dllimport</tt> attribute. On Microsoft Windows targets, the pointer
467 name is formed by combining <code>_imp__</code> and the function or variable
473 <p><a name="linkage_external">For example, since the "<tt>.LC0</tt>"
474 variable is defined to be internal, if another module defined a "<tt>.LC0</tt>"
475 variable and was linked with this one, one of the two would be renamed,
476 preventing a collision. Since "<tt>main</tt>" and "<tt>puts</tt>" are
477 external (i.e., lacking any linkage declarations), they are accessible
478 outside of the current module. It is illegal for a function <i>declaration</i>
479 to have any linkage type other than "externally visible".</a></p>
483 <!-- ======================================================================= -->
484 <div class="doc_subsection">
485 <a name="callingconv">Calling Conventions</a>
488 <div class="doc_text">
490 <p>LLVM <a href="#functionstructure">functions</a>, <a href="#i_call">calls</a>
491 and <a href="#i_invoke">invokes</a> can all have an optional calling convention
492 specified for the call. The calling convention of any pair of dynamic
493 caller/callee must match, or the behavior of the program is undefined. The
494 following calling conventions are supported by LLVM, and more may be added in
498 <dt><b>"<tt>ccc</tt>" - The C calling convention</b>:</dt>
500 <dd>This calling convention (the default if no other calling convention is
501 specified) matches the target C calling conventions. This calling convention
502 supports varargs function calls and tolerates some mismatch in the declared
503 prototype and implemented declaration of the function (as does normal C).
506 <dt><b>"<tt>csretcc</tt>" - The C struct return calling convention</b>:</dt>
508 <dd>This calling convention matches the target C calling conventions, except
509 that functions with this convention are required to take a pointer as their
510 first argument, and the return type of the function must be void. This is
511 used for C functions that return aggregates by-value. In this case, the
512 function has been transformed to take a pointer to the struct as the first
513 argument to the function. For targets where the ABI specifies specific
514 behavior for structure-return calls, the calling convention can be used to
515 distinguish between struct return functions and other functions that take a
516 pointer to a struct as the first argument.
519 <dt><b>"<tt>fastcc</tt>" - The fast calling convention</b>:</dt>
521 <dd>This calling convention attempts to make calls as fast as possible
522 (e.g. by passing things in registers). This calling convention allows the
523 target to use whatever tricks it wants to produce fast code for the target,
524 without having to conform to an externally specified ABI. Implementations of
525 this convention should allow arbitrary tail call optimization to be supported.
526 This calling convention does not support varargs and requires the prototype of
527 all callees to exactly match the prototype of the function definition.
530 <dt><b>"<tt>coldcc</tt>" - The cold calling convention</b>:</dt>
532 <dd>This calling convention attempts to make code in the caller as efficient
533 as possible under the assumption that the call is not commonly executed. As
534 such, these calls often preserve all registers so that the call does not break
535 any live ranges in the caller side. This calling convention does not support
536 varargs and requires the prototype of all callees to exactly match the
537 prototype of the function definition.
540 <dt><b>"<tt>cc <<em>n</em>></tt>" - Numbered convention</b>:</dt>
542 <dd>Any calling convention may be specified by number, allowing
543 target-specific calling conventions to be used. Target specific calling
544 conventions start at 64.
548 <p>More calling conventions can be added/defined on an as-needed basis, to
549 support pascal conventions or any other well-known target-independent
554 <!-- ======================================================================= -->
555 <div class="doc_subsection">
556 <a name="globalvars">Global Variables</a>
559 <div class="doc_text">
561 <p>Global variables define regions of memory allocated at compilation time
562 instead of run-time. Global variables may optionally be initialized, may have
563 an explicit section to be placed in, and may
564 have an optional explicit alignment specified. A
565 variable may be defined as a global "constant," which indicates that the
566 contents of the variable will <b>never</b> be modified (enabling better
567 optimization, allowing the global data to be placed in the read-only section of
568 an executable, etc). Note that variables that need runtime initialization
569 cannot be marked "constant" as there is a store to the variable.</p>
572 LLVM explicitly allows <em>declarations</em> of global variables to be marked
573 constant, even if the final definition of the global is not. This capability
574 can be used to enable slightly better optimization of the program, but requires
575 the language definition to guarantee that optimizations based on the
576 'constantness' are valid for the translation units that do not include the
580 <p>As SSA values, global variables define pointer values that are in
581 scope (i.e. they dominate) all basic blocks in the program. Global
582 variables always define a pointer to their "content" type because they
583 describe a region of memory, and all memory objects in LLVM are
584 accessed through pointers.</p>
586 <p>LLVM allows an explicit section to be specified for globals. If the target
587 supports it, it will emit globals to the section specified.</p>
589 <p>An explicit alignment may be specified for a global. If not present, or if
590 the alignment is set to zero, the alignment of the global is set by the target
591 to whatever it feels convenient. If an explicit alignment is specified, the
592 global is forced to have at least that much alignment. All alignments must be
598 <!-- ======================================================================= -->
599 <div class="doc_subsection">
600 <a name="functionstructure">Functions</a>
603 <div class="doc_text">
605 <p>LLVM function definitions consist of an optional <a href="#linkage">linkage
606 type</a>, an optional <a href="#callingconv">calling convention</a>, a return
607 type, a function name, a (possibly empty) argument list, an optional section,
608 an optional alignment, an opening curly brace,
609 a list of basic blocks, and a closing curly brace. LLVM function declarations
610 are defined with the "<tt>declare</tt>" keyword, an optional <a
611 href="#callingconv">calling convention</a>, a return type, a function name,
612 a possibly empty list of arguments, and an optional alignment.</p>
614 <p>A function definition contains a list of basic blocks, forming the CFG for
615 the function. Each basic block may optionally start with a label (giving the
616 basic block a symbol table entry), contains a list of instructions, and ends
617 with a <a href="#terminators">terminator</a> instruction (such as a branch or
618 function return).</p>
620 <p>The first basic block in a program is special in two ways: it is immediately
621 executed on entrance to the function, and it is not allowed to have predecessor
622 basic blocks (i.e. there can not be any branches to the entry block of a
623 function). Because the block can have no predecessors, it also cannot have any
624 <a href="#i_phi">PHI nodes</a>.</p>
626 <p>LLVM functions are identified by their name and type signature. Hence, two
627 functions with the same name but different parameter lists or return values are
628 considered different functions, and LLVM will resolve references to each
631 <p>LLVM allows an explicit section to be specified for functions. If the target
632 supports it, it will emit functions to the section specified.</p>
634 <p>An explicit alignment may be specified for a function. If not present, or if
635 the alignment is set to zero, the alignment of the function is set by the target
636 to whatever it feels convenient. If an explicit alignment is specified, the
637 function is forced to have at least that much alignment. All alignments must be
642 <!-- ======================================================================= -->
643 <div class="doc_subsection">
644 <a name="moduleasm">Module-Level Inline Assembly</a>
647 <div class="doc_text">
649 Modules may contain "module-level inline asm" blocks, which corresponds to the
650 GCC "file scope inline asm" blocks. These blocks are internally concatenated by
651 LLVM and treated as a single unit, but may be separated in the .ll file if
652 desired. The syntax is very simple:
655 <div class="doc_code"><pre>
656 module asm "inline asm code goes here"
657 module asm "more can go here"
660 <p>The strings can contain any character by escaping non-printable characters.
661 The escape sequence used is simply "\xx" where "xx" is the two digit hex code
666 The inline asm code is simply printed to the machine code .s file when
667 assembly code is generated.
672 <!-- *********************************************************************** -->
673 <div class="doc_section"> <a name="typesystem">Type System</a> </div>
674 <!-- *********************************************************************** -->
676 <div class="doc_text">
678 <p>The LLVM type system is one of the most important features of the
679 intermediate representation. Being typed enables a number of
680 optimizations to be performed on the IR directly, without having to do
681 extra analyses on the side before the transformation. A strong type
682 system makes it easier to read the generated code and enables novel
683 analyses and transformations that are not feasible to perform on normal
684 three address code representations.</p>
688 <!-- ======================================================================= -->
689 <div class="doc_subsection"> <a name="t_primitive">Primitive Types</a> </div>
690 <div class="doc_text">
691 <p>The primitive types are the fundamental building blocks of the LLVM
692 system. The current set of primitive types is as follows:</p>
694 <table class="layout">
699 <tr><th>Type</th><th>Description</th></tr>
700 <tr><td><tt>void</tt></td><td>No value</td></tr>
701 <tr><td><tt>ubyte</tt></td><td>Unsigned 8-bit value</td></tr>
702 <tr><td><tt>ushort</tt></td><td>Unsigned 16-bit value</td></tr>
703 <tr><td><tt>uint</tt></td><td>Unsigned 32-bit value</td></tr>
704 <tr><td><tt>ulong</tt></td><td>Unsigned 64-bit value</td></tr>
705 <tr><td><tt>float</tt></td><td>32-bit floating point value</td></tr>
706 <tr><td><tt>label</tt></td><td>Branch destination</td></tr>
713 <tr><th>Type</th><th>Description</th></tr>
714 <tr><td><tt>bool</tt></td><td>True or False value</td></tr>
715 <tr><td><tt>sbyte</tt></td><td>Signed 8-bit value</td></tr>
716 <tr><td><tt>short</tt></td><td>Signed 16-bit value</td></tr>
717 <tr><td><tt>int</tt></td><td>Signed 32-bit value</td></tr>
718 <tr><td><tt>long</tt></td><td>Signed 64-bit value</td></tr>
719 <tr><td><tt>double</tt></td><td>64-bit floating point value</td></tr>
727 <!-- _______________________________________________________________________ -->
728 <div class="doc_subsubsection"> <a name="t_classifications">Type
729 Classifications</a> </div>
730 <div class="doc_text">
731 <p>These different primitive types fall into a few useful
734 <table border="1" cellspacing="0" cellpadding="4">
736 <tr><th>Classification</th><th>Types</th></tr>
738 <td><a name="t_signed">signed</a></td>
739 <td><tt>sbyte, short, int, long, float, double</tt></td>
742 <td><a name="t_unsigned">unsigned</a></td>
743 <td><tt>ubyte, ushort, uint, ulong</tt></td>
746 <td><a name="t_integer">integer</a></td>
747 <td><tt>ubyte, sbyte, ushort, short, uint, int, ulong, long</tt></td>
750 <td><a name="t_integral">integral</a></td>
751 <td><tt>bool, ubyte, sbyte, ushort, short, uint, int, ulong, long</tt>
755 <td><a name="t_floating">floating point</a></td>
756 <td><tt>float, double</tt></td>
759 <td><a name="t_firstclass">first class</a></td>
760 <td><tt>bool, ubyte, sbyte, ushort, short, uint, int, ulong, long,<br>
761 float, double, <a href="#t_pointer">pointer</a>,
762 <a href="#t_packed">packed</a></tt></td>
767 <p>The <a href="#t_firstclass">first class</a> types are perhaps the
768 most important. Values of these types are the only ones which can be
769 produced by instructions, passed as arguments, or used as operands to
770 instructions. This means that all structures and arrays must be
771 manipulated either by pointer or by component.</p>
774 <!-- ======================================================================= -->
775 <div class="doc_subsection"> <a name="t_derived">Derived Types</a> </div>
777 <div class="doc_text">
779 <p>The real power in LLVM comes from the derived types in the system.
780 This is what allows a programmer to represent arrays, functions,
781 pointers, and other useful types. Note that these derived types may be
782 recursive: For example, it is possible to have a two dimensional array.</p>
786 <!-- _______________________________________________________________________ -->
787 <div class="doc_subsubsection"> <a name="t_array">Array Type</a> </div>
789 <div class="doc_text">
793 <p>The array type is a very simple derived type that arranges elements
794 sequentially in memory. The array type requires a size (number of
795 elements) and an underlying data type.</p>
800 [<# elements> x <elementtype>]
803 <p>The number of elements is a constant integer value; elementtype may
804 be any type with a size.</p>
807 <table class="layout">
810 <tt>[40 x int ]</tt><br/>
811 <tt>[41 x int ]</tt><br/>
812 <tt>[40 x uint]</tt><br/>
815 Array of 40 integer values.<br/>
816 Array of 41 integer values.<br/>
817 Array of 40 unsigned integer values.<br/>
821 <p>Here are some examples of multidimensional arrays:</p>
822 <table class="layout">
825 <tt>[3 x [4 x int]]</tt><br/>
826 <tt>[12 x [10 x float]]</tt><br/>
827 <tt>[2 x [3 x [4 x uint]]]</tt><br/>
830 3x4 array of integer values.<br/>
831 12x10 array of single precision floating point values.<br/>
832 2x3x4 array of unsigned integer values.<br/>
837 <p>Note that 'variable sized arrays' can be implemented in LLVM with a zero
838 length array. Normally, accesses past the end of an array are undefined in
839 LLVM (e.g. it is illegal to access the 5th element of a 3 element array).
840 As a special case, however, zero length arrays are recognized to be variable
841 length. This allows implementation of 'pascal style arrays' with the LLVM
842 type "{ int, [0 x float]}", for example.</p>
846 <!-- _______________________________________________________________________ -->
847 <div class="doc_subsubsection"> <a name="t_function">Function Type</a> </div>
848 <div class="doc_text">
850 <p>The function type can be thought of as a function signature. It
851 consists of a return type and a list of formal parameter types.
852 Function types are usually used to build virtual function tables
853 (which are structures of pointers to functions), for indirect function
854 calls, and when defining a function.</p>
856 The return type of a function type cannot be an aggregate type.
859 <pre> <returntype> (<parameter list>)<br></pre>
860 <p>...where '<tt><parameter list></tt>' is a comma-separated list of type
861 specifiers. Optionally, the parameter list may include a type <tt>...</tt>,
862 which indicates that the function takes a variable number of arguments.
863 Variable argument functions can access their arguments with the <a
864 href="#int_varargs">variable argument handling intrinsic</a> functions.</p>
866 <table class="layout">
869 <tt>int (int)</tt> <br/>
870 <tt>float (int, int *) *</tt><br/>
871 <tt>int (sbyte *, ...)</tt><br/>
874 function taking an <tt>int</tt>, returning an <tt>int</tt><br/>
875 <a href="#t_pointer">Pointer</a> to a function that takes an
876 <tt>int</tt> and a <a href="#t_pointer">pointer</a> to <tt>int</tt>,
877 returning <tt>float</tt>.<br/>
878 A vararg function that takes at least one <a href="#t_pointer">pointer</a>
879 to <tt>sbyte</tt> (signed char in C), which returns an integer. This is
880 the signature for <tt>printf</tt> in LLVM.<br/>
886 <!-- _______________________________________________________________________ -->
887 <div class="doc_subsubsection"> <a name="t_struct">Structure Type</a> </div>
888 <div class="doc_text">
890 <p>The structure type is used to represent a collection of data members
891 together in memory. The packing of the field types is defined to match
892 the ABI of the underlying processor. The elements of a structure may
893 be any type that has a size.</p>
894 <p>Structures are accessed using '<tt><a href="#i_load">load</a></tt>
895 and '<tt><a href="#i_store">store</a></tt>' by getting a pointer to a
896 field with the '<tt><a href="#i_getelementptr">getelementptr</a></tt>'
899 <pre> { <type list> }<br></pre>
901 <table class="layout">
904 <tt>{ int, int, int }</tt><br/>
905 <tt>{ float, int (int) * }</tt><br/>
908 a triple of three <tt>int</tt> values<br/>
909 A pair, where the first element is a <tt>float</tt> and the second element
910 is a <a href="#t_pointer">pointer</a> to a <a href="#t_function">function</a>
911 that takes an <tt>int</tt>, returning an <tt>int</tt>.<br/>
917 <!-- _______________________________________________________________________ -->
918 <div class="doc_subsubsection"> <a name="t_pointer">Pointer Type</a> </div>
919 <div class="doc_text">
921 <p>As in many languages, the pointer type represents a pointer or
922 reference to another object, which must live in memory.</p>
924 <pre> <type> *<br></pre>
926 <table class="layout">
929 <tt>[4x int]*</tt><br/>
930 <tt>int (int *) *</tt><br/>
933 A <a href="#t_pointer">pointer</a> to <a href="#t_array">array</a> of
934 four <tt>int</tt> values<br/>
935 A <a href="#t_pointer">pointer</a> to a <a
936 href="#t_function">function</a> that takes an <tt>int*</tt>, returning an
943 <!-- _______________________________________________________________________ -->
944 <div class="doc_subsubsection"> <a name="t_packed">Packed Type</a> </div>
945 <div class="doc_text">
949 <p>A packed type is a simple derived type that represents a vector
950 of elements. Packed types are used when multiple primitive data
951 are operated in parallel using a single instruction (SIMD).
952 A packed type requires a size (number of
953 elements) and an underlying primitive data type. Vectors must have a power
954 of two length (1, 2, 4, 8, 16 ...). Packed types are
955 considered <a href="#t_firstclass">first class</a>.</p>
960 < <# elements> x <elementtype> >
963 <p>The number of elements is a constant integer value; elementtype may
964 be any integral or floating point type.</p>
968 <table class="layout">
971 <tt><4 x int></tt><br/>
972 <tt><8 x float></tt><br/>
973 <tt><2 x uint></tt><br/>
976 Packed vector of 4 integer values.<br/>
977 Packed vector of 8 floating-point values.<br/>
978 Packed vector of 2 unsigned integer values.<br/>
984 <!-- _______________________________________________________________________ -->
985 <div class="doc_subsubsection"> <a name="t_opaque">Opaque Type</a> </div>
986 <div class="doc_text">
990 <p>Opaque types are used to represent unknown types in the system. This
991 corresponds (for example) to the C notion of a foward declared structure type.
992 In LLVM, opaque types can eventually be resolved to any type (not just a
1003 <table class="layout">
1009 An opaque type.<br/>
1016 <!-- *********************************************************************** -->
1017 <div class="doc_section"> <a name="constants">Constants</a> </div>
1018 <!-- *********************************************************************** -->
1020 <div class="doc_text">
1022 <p>LLVM has several different basic types of constants. This section describes
1023 them all and their syntax.</p>
1027 <!-- ======================================================================= -->
1028 <div class="doc_subsection"><a name="simpleconstants">Simple Constants</a></div>
1030 <div class="doc_text">
1033 <dt><b>Boolean constants</b></dt>
1035 <dd>The two strings '<tt>true</tt>' and '<tt>false</tt>' are both valid
1036 constants of the <tt><a href="#t_primitive">bool</a></tt> type.
1039 <dt><b>Integer constants</b></dt>
1041 <dd>Standard integers (such as '4') are constants of the <a
1042 href="#t_integer">integer</a> type. Negative numbers may be used with signed
1046 <dt><b>Floating point constants</b></dt>
1048 <dd>Floating point constants use standard decimal notation (e.g. 123.421),
1049 exponential notation (e.g. 1.23421e+2), or a more precise hexadecimal
1050 notation (see below). Floating point constants must have a <a
1051 href="#t_floating">floating point</a> type. </dd>
1053 <dt><b>Null pointer constants</b></dt>
1055 <dd>The identifier '<tt>null</tt>' is recognized as a null pointer constant
1056 and must be of <a href="#t_pointer">pointer type</a>.</dd>
1060 <p>The one non-intuitive notation for constants is the optional hexadecimal form
1061 of floating point constants. For example, the form '<tt>double
1062 0x432ff973cafa8000</tt>' is equivalent to (but harder to read than) '<tt>double
1063 4.5e+15</tt>'. The only time hexadecimal floating point constants are required
1064 (and the only time that they are generated by the disassembler) is when a
1065 floating point constant must be emitted but it cannot be represented as a
1066 decimal floating point number. For example, NaN's, infinities, and other
1067 special values are represented in their IEEE hexadecimal format so that
1068 assembly and disassembly do not cause any bits to change in the constants.</p>
1072 <!-- ======================================================================= -->
1073 <div class="doc_subsection"><a name="aggregateconstants">Aggregate Constants</a>
1076 <div class="doc_text">
1077 <p>Aggregate constants arise from aggregation of simple constants
1078 and smaller aggregate constants.</p>
1081 <dt><b>Structure constants</b></dt>
1083 <dd>Structure constants are represented with notation similar to structure
1084 type definitions (a comma separated list of elements, surrounded by braces
1085 (<tt>{}</tt>)). For example: "<tt>{ int 4, float 17.0, int* %G }</tt>",
1086 where "<tt>%G</tt>" is declared as "<tt>%G = external global int</tt>". Structure constants
1087 must have <a href="#t_struct">structure type</a>, and the number and
1088 types of elements must match those specified by the type.
1091 <dt><b>Array constants</b></dt>
1093 <dd>Array constants are represented with notation similar to array type
1094 definitions (a comma separated list of elements, surrounded by square brackets
1095 (<tt>[]</tt>)). For example: "<tt>[ int 42, int 11, int 74 ]</tt>". Array
1096 constants must have <a href="#t_array">array type</a>, and the number and
1097 types of elements must match those specified by the type.
1100 <dt><b>Packed constants</b></dt>
1102 <dd>Packed constants are represented with notation similar to packed type
1103 definitions (a comma separated list of elements, surrounded by
1104 less-than/greater-than's (<tt><></tt>)). For example: "<tt>< int 42,
1105 int 11, int 74, int 100 ></tt>". Packed constants must have <a
1106 href="#t_packed">packed type</a>, and the number and types of elements must
1107 match those specified by the type.
1110 <dt><b>Zero initialization</b></dt>
1112 <dd>The string '<tt>zeroinitializer</tt>' can be used to zero initialize a
1113 value to zero of <em>any</em> type, including scalar and aggregate types.
1114 This is often used to avoid having to print large zero initializers (e.g. for
1115 large arrays) and is always exactly equivalent to using explicit zero
1122 <!-- ======================================================================= -->
1123 <div class="doc_subsection">
1124 <a name="globalconstants">Global Variable and Function Addresses</a>
1127 <div class="doc_text">
1129 <p>The addresses of <a href="#globalvars">global variables</a> and <a
1130 href="#functionstructure">functions</a> are always implicitly valid (link-time)
1131 constants. These constants are explicitly referenced when the <a
1132 href="#identifiers">identifier for the global</a> is used and always have <a
1133 href="#t_pointer">pointer</a> type. For example, the following is a legal LLVM
1139 %Z = global [2 x int*] [ int* %X, int* %Y ]
1144 <!-- ======================================================================= -->
1145 <div class="doc_subsection"><a name="undefvalues">Undefined Values</a></div>
1146 <div class="doc_text">
1147 <p>The string '<tt>undef</tt>' is recognized as a type-less constant that has
1148 no specific value. Undefined values may be of any type and be used anywhere
1149 a constant is permitted.</p>
1151 <p>Undefined values indicate to the compiler that the program is well defined
1152 no matter what value is used, giving the compiler more freedom to optimize.
1156 <!-- ======================================================================= -->
1157 <div class="doc_subsection"><a name="constantexprs">Constant Expressions</a>
1160 <div class="doc_text">
1162 <p>Constant expressions are used to allow expressions involving other constants
1163 to be used as constants. Constant expressions may be of any <a
1164 href="#t_firstclass">first class</a> type and may involve any LLVM operation
1165 that does not have side effects (e.g. load and call are not supported). The
1166 following is the syntax for constant expressions:</p>
1169 <dt><b><tt>cast ( CST to TYPE )</tt></b></dt>
1171 <dd>Cast a constant to another type.</dd>
1173 <dt><b><tt>getelementptr ( CSTPTR, IDX0, IDX1, ... )</tt></b></dt>
1175 <dd>Perform the <a href="#i_getelementptr">getelementptr operation</a> on
1176 constants. As with the <a href="#i_getelementptr">getelementptr</a>
1177 instruction, the index list may have zero or more indexes, which are required
1178 to make sense for the type of "CSTPTR".</dd>
1180 <dt><b><tt>select ( COND, VAL1, VAL2 )</tt></b></dt>
1182 <dd>Perform the <a href="#i_select">select operation</a> on
1185 <dt><b><tt>extractelement ( VAL, IDX )</tt></b></dt>
1187 <dd>Perform the <a href="#i_extractelement">extractelement
1188 operation</a> on constants.
1190 <dt><b><tt>insertelement ( VAL, ELT, IDX )</tt></b></dt>
1192 <dd>Perform the <a href="#i_insertelement">insertelement
1193 operation</a> on constants.
1196 <dt><b><tt>shufflevector ( VEC1, VEC2, IDXMASK )</tt></b></dt>
1198 <dd>Perform the <a href="#i_shufflevector">shufflevector
1199 operation</a> on constants.
1201 <dt><b><tt>OPCODE ( LHS, RHS )</tt></b></dt>
1203 <dd>Perform the specified operation of the LHS and RHS constants. OPCODE may
1204 be any of the <a href="#binaryops">binary</a> or <a href="#bitwiseops">bitwise
1205 binary</a> operations. The constraints on operands are the same as those for
1206 the corresponding instruction (e.g. no bitwise operations on floating point
1207 values are allowed).</dd>
1211 <!-- *********************************************************************** -->
1212 <div class="doc_section"> <a name="othervalues">Other Values</a> </div>
1213 <!-- *********************************************************************** -->
1215 <!-- ======================================================================= -->
1216 <div class="doc_subsection">
1217 <a name="inlineasm">Inline Assembler Expressions</a>
1220 <div class="doc_text">
1223 LLVM supports inline assembler expressions (as opposed to <a href="#moduleasm">
1224 Module-Level Inline Assembly</a>) through the use of a special value. This
1225 value represents the inline assembler as a string (containing the instructions
1226 to emit), a list of operand constraints (stored as a string), and a flag that
1227 indicates whether or not the inline asm expression has side effects. An example
1228 inline assembler expression is:
1232 int(int) asm "bswap $0", "=r,r"
1236 Inline assembler expressions may <b>only</b> be used as the callee operand of
1237 a <a href="#i_call"><tt>call</tt> instruction</a>. Thus, typically we have:
1241 %X = call int asm "<a href="#i_bswap">bswap</a> $0", "=r,r"(int %Y)
1245 Inline asms with side effects not visible in the constraint list must be marked
1246 as having side effects. This is done through the use of the
1247 '<tt>sideeffect</tt>' keyword, like so:
1251 call void asm sideeffect "eieio", ""()
1254 <p>TODO: The format of the asm and constraints string still need to be
1255 documented here. Constraints on what can be done (e.g. duplication, moving, etc
1256 need to be documented).
1261 <!-- *********************************************************************** -->
1262 <div class="doc_section"> <a name="instref">Instruction Reference</a> </div>
1263 <!-- *********************************************************************** -->
1265 <div class="doc_text">
1267 <p>The LLVM instruction set consists of several different
1268 classifications of instructions: <a href="#terminators">terminator
1269 instructions</a>, <a href="#binaryops">binary instructions</a>,
1270 <a href="#bitwiseops">bitwise binary instructions</a>, <a
1271 href="#memoryops">memory instructions</a>, and <a href="#otherops">other
1272 instructions</a>.</p>
1276 <!-- ======================================================================= -->
1277 <div class="doc_subsection"> <a name="terminators">Terminator
1278 Instructions</a> </div>
1280 <div class="doc_text">
1282 <p>As mentioned <a href="#functionstructure">previously</a>, every
1283 basic block in a program ends with a "Terminator" instruction, which
1284 indicates which block should be executed after the current block is
1285 finished. These terminator instructions typically yield a '<tt>void</tt>'
1286 value: they produce control flow, not values (the one exception being
1287 the '<a href="#i_invoke"><tt>invoke</tt></a>' instruction).</p>
1288 <p>There are six different terminator instructions: the '<a
1289 href="#i_ret"><tt>ret</tt></a>' instruction, the '<a href="#i_br"><tt>br</tt></a>'
1290 instruction, the '<a href="#i_switch"><tt>switch</tt></a>' instruction,
1291 the '<a href="#i_invoke"><tt>invoke</tt></a>' instruction, the '<a
1292 href="#i_unwind"><tt>unwind</tt></a>' instruction, and the '<a
1293 href="#i_unreachable"><tt>unreachable</tt></a>' instruction.</p>
1297 <!-- _______________________________________________________________________ -->
1298 <div class="doc_subsubsection"> <a name="i_ret">'<tt>ret</tt>'
1299 Instruction</a> </div>
1300 <div class="doc_text">
1302 <pre> ret <type> <value> <i>; Return a value from a non-void function</i>
1303 ret void <i>; Return from void function</i>
1306 <p>The '<tt>ret</tt>' instruction is used to return control flow (and a
1307 value) from a function back to the caller.</p>
1308 <p>There are two forms of the '<tt>ret</tt>' instruction: one that
1309 returns a value and then causes control flow, and one that just causes
1310 control flow to occur.</p>
1312 <p>The '<tt>ret</tt>' instruction may return any '<a
1313 href="#t_firstclass">first class</a>' type. Notice that a function is
1314 not <a href="#wellformed">well formed</a> if there exists a '<tt>ret</tt>'
1315 instruction inside of the function that returns a value that does not
1316 match the return type of the function.</p>
1318 <p>When the '<tt>ret</tt>' instruction is executed, control flow
1319 returns back to the calling function's context. If the caller is a "<a
1320 href="#i_call"><tt>call</tt></a>" instruction, execution continues at
1321 the instruction after the call. If the caller was an "<a
1322 href="#i_invoke"><tt>invoke</tt></a>" instruction, execution continues
1323 at the beginning of the "normal" destination block. If the instruction
1324 returns a value, that value shall set the call or invoke instruction's
1327 <pre> ret int 5 <i>; Return an integer value of 5</i>
1328 ret void <i>; Return from a void function</i>
1331 <!-- _______________________________________________________________________ -->
1332 <div class="doc_subsubsection"> <a name="i_br">'<tt>br</tt>' Instruction</a> </div>
1333 <div class="doc_text">
1335 <pre> br bool <cond>, label <iftrue>, label <iffalse><br> br label <dest> <i>; Unconditional branch</i>
1338 <p>The '<tt>br</tt>' instruction is used to cause control flow to
1339 transfer to a different basic block in the current function. There are
1340 two forms of this instruction, corresponding to a conditional branch
1341 and an unconditional branch.</p>
1343 <p>The conditional branch form of the '<tt>br</tt>' instruction takes a
1344 single '<tt>bool</tt>' value and two '<tt>label</tt>' values. The
1345 unconditional form of the '<tt>br</tt>' instruction takes a single '<tt>label</tt>'
1346 value as a target.</p>
1348 <p>Upon execution of a conditional '<tt>br</tt>' instruction, the '<tt>bool</tt>'
1349 argument is evaluated. If the value is <tt>true</tt>, control flows
1350 to the '<tt>iftrue</tt>' <tt>label</tt> argument. If "cond" is <tt>false</tt>,
1351 control flows to the '<tt>iffalse</tt>' <tt>label</tt> argument.</p>
1353 <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
1354 href="#i_ret">ret</a> int 1<br>IfUnequal:<br> <a href="#i_ret">ret</a> int 0<br></pre>
1356 <!-- _______________________________________________________________________ -->
1357 <div class="doc_subsubsection">
1358 <a name="i_switch">'<tt>switch</tt>' Instruction</a>
1361 <div class="doc_text">
1365 switch <intty> <value>, label <defaultdest> [ <intty> <val>, label <dest> ... ]
1370 <p>The '<tt>switch</tt>' instruction is used to transfer control flow to one of
1371 several different places. It is a generalization of the '<tt>br</tt>'
1372 instruction, allowing a branch to occur to one of many possible
1378 <p>The '<tt>switch</tt>' instruction uses three parameters: an integer
1379 comparison value '<tt>value</tt>', a default '<tt>label</tt>' destination, and
1380 an array of pairs of comparison value constants and '<tt>label</tt>'s. The
1381 table is not allowed to contain duplicate constant entries.</p>
1385 <p>The <tt>switch</tt> instruction specifies a table of values and
1386 destinations. When the '<tt>switch</tt>' instruction is executed, this
1387 table is searched for the given value. If the value is found, control flow is
1388 transfered to the corresponding destination; otherwise, control flow is
1389 transfered to the default destination.</p>
1391 <h5>Implementation:</h5>
1393 <p>Depending on properties of the target machine and the particular
1394 <tt>switch</tt> instruction, this instruction may be code generated in different
1395 ways. For example, it could be generated as a series of chained conditional
1396 branches or with a lookup table.</p>
1401 <i>; Emulate a conditional br instruction</i>
1402 %Val = <a href="#i_cast">cast</a> bool %value to int
1403 switch int %Val, label %truedest [int 0, label %falsedest ]
1405 <i>; Emulate an unconditional br instruction</i>
1406 switch uint 0, label %dest [ ]
1408 <i>; Implement a jump table:</i>
1409 switch uint %val, label %otherwise [ uint 0, label %onzero
1410 uint 1, label %onone
1411 uint 2, label %ontwo ]
1415 <!-- _______________________________________________________________________ -->
1416 <div class="doc_subsubsection">
1417 <a name="i_invoke">'<tt>invoke</tt>' Instruction</a>
1420 <div class="doc_text">
1425 <result> = invoke [<a href="#callingconv">cconv</a>] <ptr to function ty> %<function ptr val>(<function args>)
1426 to label <normal label> unwind label <exception label>
1431 <p>The '<tt>invoke</tt>' instruction causes control to transfer to a specified
1432 function, with the possibility of control flow transfer to either the
1433 '<tt>normal</tt>' label or the
1434 '<tt>exception</tt>' label. If the callee function returns with the
1435 "<tt><a href="#i_ret">ret</a></tt>" instruction, control flow will return to the
1436 "normal" label. If the callee (or any indirect callees) returns with the "<a
1437 href="#i_unwind"><tt>unwind</tt></a>" instruction, control is interrupted and
1438 continued at the dynamically nearest "exception" label.</p>
1442 <p>This instruction requires several arguments:</p>
1446 The optional "cconv" marker indicates which <a href="callingconv">calling
1447 convention</a> the call should use. If none is specified, the call defaults
1448 to using C calling conventions.
1450 <li>'<tt>ptr to function ty</tt>': shall be the signature of the pointer to
1451 function value being invoked. In most cases, this is a direct function
1452 invocation, but indirect <tt>invoke</tt>s are just as possible, branching off
1453 an arbitrary pointer to function value.
1456 <li>'<tt>function ptr val</tt>': An LLVM value containing a pointer to a
1457 function to be invoked. </li>
1459 <li>'<tt>function args</tt>': argument list whose types match the function
1460 signature argument types. If the function signature indicates the function
1461 accepts a variable number of arguments, the extra arguments can be
1464 <li>'<tt>normal label</tt>': the label reached when the called function
1465 executes a '<tt><a href="#i_ret">ret</a></tt>' instruction. </li>
1467 <li>'<tt>exception label</tt>': the label reached when a callee returns with
1468 the <a href="#i_unwind"><tt>unwind</tt></a> instruction. </li>
1474 <p>This instruction is designed to operate as a standard '<tt><a
1475 href="#i_call">call</a></tt>' instruction in most regards. The primary
1476 difference is that it establishes an association with a label, which is used by
1477 the runtime library to unwind the stack.</p>
1479 <p>This instruction is used in languages with destructors to ensure that proper
1480 cleanup is performed in the case of either a <tt>longjmp</tt> or a thrown
1481 exception. Additionally, this is important for implementation of
1482 '<tt>catch</tt>' clauses in high-level languages that support them.</p>
1486 %retval = invoke int %Test(int 15) to label %Continue
1487 unwind label %TestCleanup <i>; {int}:retval set</i>
1488 %retval = invoke <a href="#callingconv">coldcc</a> int %Test(int 15) to label %Continue
1489 unwind label %TestCleanup <i>; {int}:retval set</i>
1494 <!-- _______________________________________________________________________ -->
1496 <div class="doc_subsubsection"> <a name="i_unwind">'<tt>unwind</tt>'
1497 Instruction</a> </div>
1499 <div class="doc_text">
1508 <p>The '<tt>unwind</tt>' instruction unwinds the stack, continuing control flow
1509 at the first callee in the dynamic call stack which used an <a
1510 href="#i_invoke"><tt>invoke</tt></a> instruction to perform the call. This is
1511 primarily used to implement exception handling.</p>
1515 <p>The '<tt>unwind</tt>' intrinsic causes execution of the current function to
1516 immediately halt. The dynamic call stack is then searched for the first <a
1517 href="#i_invoke"><tt>invoke</tt></a> instruction on the call stack. Once found,
1518 execution continues at the "exceptional" destination block specified by the
1519 <tt>invoke</tt> instruction. If there is no <tt>invoke</tt> instruction in the
1520 dynamic call chain, undefined behavior results.</p>
1523 <!-- _______________________________________________________________________ -->
1525 <div class="doc_subsubsection"> <a name="i_unreachable">'<tt>unreachable</tt>'
1526 Instruction</a> </div>
1528 <div class="doc_text">
1537 <p>The '<tt>unreachable</tt>' instruction has no defined semantics. This
1538 instruction is used to inform the optimizer that a particular portion of the
1539 code is not reachable. This can be used to indicate that the code after a
1540 no-return function cannot be reached, and other facts.</p>
1544 <p>The '<tt>unreachable</tt>' instruction has no defined semantics.</p>
1549 <!-- ======================================================================= -->
1550 <div class="doc_subsection"> <a name="binaryops">Binary Operations</a> </div>
1551 <div class="doc_text">
1552 <p>Binary operators are used to do most of the computation in a
1553 program. They require two operands, execute an operation on them, and
1554 produce a single value. The operands might represent
1555 multiple data, as is the case with the <a href="#t_packed">packed</a> data type.
1556 The result value of a binary operator is not
1557 necessarily the same type as its operands.</p>
1558 <p>There are several different binary operators:</p>
1560 <!-- _______________________________________________________________________ -->
1561 <div class="doc_subsubsection"> <a name="i_add">'<tt>add</tt>'
1562 Instruction</a> </div>
1563 <div class="doc_text">
1565 <pre> <result> = add <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1568 <p>The '<tt>add</tt>' instruction returns the sum of its two operands.</p>
1570 <p>The two arguments to the '<tt>add</tt>' instruction must be either <a
1571 href="#t_integer">integer</a> or <a href="#t_floating">floating point</a> values.
1572 This instruction can also take <a href="#t_packed">packed</a> versions of the values.
1573 Both arguments must have identical types.</p>
1575 <p>The value produced is the integer or floating point sum of the two
1578 <pre> <result> = add int 4, %var <i>; yields {int}:result = 4 + %var</i>
1581 <!-- _______________________________________________________________________ -->
1582 <div class="doc_subsubsection"> <a name="i_sub">'<tt>sub</tt>'
1583 Instruction</a> </div>
1584 <div class="doc_text">
1586 <pre> <result> = sub <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1589 <p>The '<tt>sub</tt>' instruction returns the difference of its two
1591 <p>Note that the '<tt>sub</tt>' instruction is used to represent the '<tt>neg</tt>'
1592 instruction present in most other intermediate representations.</p>
1594 <p>The two arguments to the '<tt>sub</tt>' instruction must be either <a
1595 href="#t_integer">integer</a> or <a href="#t_floating">floating point</a>
1597 This instruction can also take <a href="#t_packed">packed</a> versions of the values.
1598 Both arguments must have identical types.</p>
1600 <p>The value produced is the integer or floating point difference of
1601 the two operands.</p>
1603 <pre> <result> = sub int 4, %var <i>; yields {int}:result = 4 - %var</i>
1604 <result> = sub int 0, %val <i>; yields {int}:result = -%var</i>
1607 <!-- _______________________________________________________________________ -->
1608 <div class="doc_subsubsection"> <a name="i_mul">'<tt>mul</tt>'
1609 Instruction</a> </div>
1610 <div class="doc_text">
1612 <pre> <result> = mul <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1615 <p>The '<tt>mul</tt>' instruction returns the product of its two
1618 <p>The two arguments to the '<tt>mul</tt>' instruction must be either <a
1619 href="#t_integer">integer</a> or <a href="#t_floating">floating point</a>
1621 This instruction can also take <a href="#t_packed">packed</a> versions of the values.
1622 Both arguments must have identical types.</p>
1624 <p>The value produced is the integer or floating point product of the
1626 <p>There is no signed vs unsigned multiplication. The appropriate
1627 action is taken based on the type of the operand.</p>
1629 <pre> <result> = mul int 4, %var <i>; yields {int}:result = 4 * %var</i>
1632 <!-- _______________________________________________________________________ -->
1633 <div class="doc_subsubsection"> <a name="i_div">'<tt>div</tt>'
1634 Instruction</a> </div>
1635 <div class="doc_text">
1637 <pre> <result> = div <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1640 <p>The '<tt>div</tt>' instruction returns the quotient of its two
1643 <p>The two arguments to the '<tt>div</tt>' instruction must be either <a
1644 href="#t_integer">integer</a> or <a href="#t_floating">floating point</a>
1646 This instruction can also take <a href="#t_packed">packed</a> versions of the values.
1647 Both arguments must have identical types.</p>
1649 <p>The value produced is the integer or floating point quotient of the
1652 <pre> <result> = div int 4, %var <i>; yields {int}:result = 4 / %var</i>
1655 <!-- _______________________________________________________________________ -->
1656 <div class="doc_subsubsection"> <a name="i_rem">'<tt>rem</tt>'
1657 Instruction</a> </div>
1658 <div class="doc_text">
1660 <pre> <result> = rem <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1663 <p>The '<tt>rem</tt>' instruction returns the remainder from the
1664 division of its two operands.</p>
1666 <p>The two arguments to the '<tt>rem</tt>' instruction must be either <a
1667 href="#t_integer">integer</a> or <a href="#t_floating">floating point</a>
1669 This instruction can also take <a href="#t_packed">packed</a> versions of the values.
1670 Both arguments must have identical types.</p>
1672 <p>This returns the <i>remainder</i> of a division (where the result
1673 has the same sign as the divisor), not the <i>modulus</i> (where the
1674 result has the same sign as the dividend) of a value. For more
1675 information about the difference, see <a
1676 href="http://mathforum.org/dr.math/problems/anne.4.28.99.html">The
1679 <pre> <result> = rem int 4, %var <i>; yields {int}:result = 4 % %var</i>
1683 <!-- _______________________________________________________________________ -->
1684 <div class="doc_subsubsection"> <a name="i_setcc">'<tt>set<i>cc</i></tt>'
1685 Instructions</a> </div>
1686 <div class="doc_text">
1688 <pre> <result> = seteq <ty> <var1>, <var2> <i>; yields {bool}:result</i>
1689 <result> = setne <ty> <var1>, <var2> <i>; yields {bool}:result</i>
1690 <result> = setlt <ty> <var1>, <var2> <i>; yields {bool}:result</i>
1691 <result> = setgt <ty> <var1>, <var2> <i>; yields {bool}:result</i>
1692 <result> = setle <ty> <var1>, <var2> <i>; yields {bool}:result</i>
1693 <result> = setge <ty> <var1>, <var2> <i>; yields {bool}:result</i>
1696 <p>The '<tt>set<i>cc</i></tt>' family of instructions returns a boolean
1697 value based on a comparison of their two operands.</p>
1699 <p>The two arguments to the '<tt>set<i>cc</i></tt>' instructions must
1700 be of <a href="#t_firstclass">first class</a> type (it is not possible
1701 to compare '<tt>label</tt>'s, '<tt>array</tt>'s, '<tt>structure</tt>'
1702 or '<tt>void</tt>' values, etc...). Both arguments must have identical
1705 <p>The '<tt>seteq</tt>' instruction yields a <tt>true</tt> '<tt>bool</tt>'
1706 value if both operands are equal.<br>
1707 The '<tt>setne</tt>' instruction yields a <tt>true</tt> '<tt>bool</tt>'
1708 value if both operands are unequal.<br>
1709 The '<tt>setlt</tt>' instruction yields a <tt>true</tt> '<tt>bool</tt>'
1710 value if the first operand is less than the second operand.<br>
1711 The '<tt>setgt</tt>' instruction yields a <tt>true</tt> '<tt>bool</tt>'
1712 value if the first operand is greater than the second operand.<br>
1713 The '<tt>setle</tt>' instruction yields a <tt>true</tt> '<tt>bool</tt>'
1714 value if the first operand is less than or equal to the second operand.<br>
1715 The '<tt>setge</tt>' instruction yields a <tt>true</tt> '<tt>bool</tt>'
1716 value if the first operand is greater than or equal to the second
1719 <pre> <result> = seteq int 4, 5 <i>; yields {bool}:result = false</i>
1720 <result> = setne float 4, 5 <i>; yields {bool}:result = true</i>
1721 <result> = setlt uint 4, 5 <i>; yields {bool}:result = true</i>
1722 <result> = setgt sbyte 4, 5 <i>; yields {bool}:result = false</i>
1723 <result> = setle sbyte 4, 5 <i>; yields {bool}:result = true</i>
1724 <result> = setge sbyte 4, 5 <i>; yields {bool}:result = false</i>
1728 <!-- ======================================================================= -->
1729 <div class="doc_subsection"> <a name="bitwiseops">Bitwise Binary
1730 Operations</a> </div>
1731 <div class="doc_text">
1732 <p>Bitwise binary operators are used to do various forms of
1733 bit-twiddling in a program. They are generally very efficient
1734 instructions and can commonly be strength reduced from other
1735 instructions. They require two operands, execute an operation on them,
1736 and produce a single value. The resulting value of the bitwise binary
1737 operators is always the same type as its first operand.</p>
1739 <!-- _______________________________________________________________________ -->
1740 <div class="doc_subsubsection"> <a name="i_and">'<tt>and</tt>'
1741 Instruction</a> </div>
1742 <div class="doc_text">
1744 <pre> <result> = and <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1747 <p>The '<tt>and</tt>' instruction returns the bitwise logical and of
1748 its two operands.</p>
1750 <p>The two arguments to the '<tt>and</tt>' instruction must be <a
1751 href="#t_integral">integral</a> values. Both arguments must have
1752 identical types.</p>
1754 <p>The truth table used for the '<tt>and</tt>' instruction is:</p>
1756 <div style="align: center">
1757 <table border="1" cellspacing="0" cellpadding="4">
1788 <pre> <result> = and int 4, %var <i>; yields {int}:result = 4 & %var</i>
1789 <result> = and int 15, 40 <i>; yields {int}:result = 8</i>
1790 <result> = and int 4, 8 <i>; yields {int}:result = 0</i>
1793 <!-- _______________________________________________________________________ -->
1794 <div class="doc_subsubsection"> <a name="i_or">'<tt>or</tt>' Instruction</a> </div>
1795 <div class="doc_text">
1797 <pre> <result> = or <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1800 <p>The '<tt>or</tt>' instruction returns the bitwise logical inclusive
1801 or of its two operands.</p>
1803 <p>The two arguments to the '<tt>or</tt>' instruction must be <a
1804 href="#t_integral">integral</a> values. Both arguments must have
1805 identical types.</p>
1807 <p>The truth table used for the '<tt>or</tt>' instruction is:</p>
1809 <div style="align: center">
1810 <table border="1" cellspacing="0" cellpadding="4">
1841 <pre> <result> = or int 4, %var <i>; yields {int}:result = 4 | %var</i>
1842 <result> = or int 15, 40 <i>; yields {int}:result = 47</i>
1843 <result> = or int 4, 8 <i>; yields {int}:result = 12</i>
1846 <!-- _______________________________________________________________________ -->
1847 <div class="doc_subsubsection"> <a name="i_xor">'<tt>xor</tt>'
1848 Instruction</a> </div>
1849 <div class="doc_text">
1851 <pre> <result> = xor <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1854 <p>The '<tt>xor</tt>' instruction returns the bitwise logical exclusive
1855 or of its two operands. The <tt>xor</tt> is used to implement the
1856 "one's complement" operation, which is the "~" operator in C.</p>
1858 <p>The two arguments to the '<tt>xor</tt>' instruction must be <a
1859 href="#t_integral">integral</a> values. Both arguments must have
1860 identical types.</p>
1862 <p>The truth table used for the '<tt>xor</tt>' instruction is:</p>
1864 <div style="align: center">
1865 <table border="1" cellspacing="0" cellpadding="4">
1897 <pre> <result> = xor int 4, %var <i>; yields {int}:result = 4 ^ %var</i>
1898 <result> = xor int 15, 40 <i>; yields {int}:result = 39</i>
1899 <result> = xor int 4, 8 <i>; yields {int}:result = 12</i>
1900 <result> = xor int %V, -1 <i>; yields {int}:result = ~%V</i>
1903 <!-- _______________________________________________________________________ -->
1904 <div class="doc_subsubsection"> <a name="i_shl">'<tt>shl</tt>'
1905 Instruction</a> </div>
1906 <div class="doc_text">
1908 <pre> <result> = shl <ty> <var1>, ubyte <var2> <i>; yields {ty}:result</i>
1911 <p>The '<tt>shl</tt>' instruction returns the first operand shifted to
1912 the left a specified number of bits.</p>
1914 <p>The first argument to the '<tt>shl</tt>' instruction must be an <a
1915 href="#t_integer">integer</a> type. The second argument must be an '<tt>ubyte</tt>'
1918 <p>The value produced is <tt>var1</tt> * 2<sup><tt>var2</tt></sup>.</p>
1920 <pre> <result> = shl int 4, ubyte %var <i>; yields {int}:result = 4 << %var</i>
1921 <result> = shl int 4, ubyte 2 <i>; yields {int}:result = 16</i>
1922 <result> = shl int 1, ubyte 10 <i>; yields {int}:result = 1024</i>
1925 <!-- _______________________________________________________________________ -->
1926 <div class="doc_subsubsection"> <a name="i_shr">'<tt>shr</tt>'
1927 Instruction</a> </div>
1928 <div class="doc_text">
1930 <pre> <result> = shr <ty> <var1>, ubyte <var2> <i>; yields {ty}:result</i>
1933 <p>The '<tt>shr</tt>' instruction returns the first operand shifted to
1934 the right a specified number of bits.</p>
1936 <p>The first argument to the '<tt>shr</tt>' instruction must be an <a
1937 href="#t_integer">integer</a> type. The second argument must be an '<tt>ubyte</tt>'
1940 <p>If the first argument is a <a href="#t_signed">signed</a> type, the
1941 most significant bit is duplicated in the newly free'd bit positions.
1942 If the first argument is unsigned, zero bits shall fill the empty
1945 <pre> <result> = shr int 4, ubyte %var <i>; yields {int}:result = 4 >> %var</i>
1946 <result> = shr uint 4, ubyte 1 <i>; yields {uint}:result = 2</i>
1947 <result> = shr int 4, ubyte 2 <i>; yields {int}:result = 1</i>
1948 <result> = shr sbyte 4, ubyte 3 <i>; yields {sbyte}:result = 0</i>
1949 <result> = shr sbyte -2, ubyte 1 <i>; yields {sbyte}:result = -1</i>
1953 <!-- ======================================================================= -->
1954 <div class="doc_subsection">
1955 <a name="vectorops">Vector Operations</a>
1958 <div class="doc_text">
1960 <p>LLVM supports several instructions to represent vector operations in a
1961 target-independent manner. This instructions cover the element-access and
1962 vector-specific operations needed to process vectors effectively. While LLVM
1963 does directly support these vector operations, many sophisticated algorithms
1964 will want to use target-specific intrinsics to take full advantage of a specific
1969 <!-- _______________________________________________________________________ -->
1970 <div class="doc_subsubsection">
1971 <a name="i_extractelement">'<tt>extractelement</tt>' Instruction</a>
1974 <div class="doc_text">
1979 <result> = extractelement <n x <ty>> <val>, uint <idx> <i>; yields <ty></i>
1985 The '<tt>extractelement</tt>' instruction extracts a single scalar
1986 element from a packed vector at a specified index.
1993 The first operand of an '<tt>extractelement</tt>' instruction is a
1994 value of <a href="#t_packed">packed</a> type. The second operand is
1995 an index indicating the position from which to extract the element.
1996 The index may be a variable.</p>
2001 The result is a scalar of the same type as the element type of
2002 <tt>val</tt>. Its value is the value at position <tt>idx</tt> of
2003 <tt>val</tt>. If <tt>idx</tt> exceeds the length of <tt>val</tt>, the
2004 results are undefined.
2010 %result = extractelement <4 x int> %vec, uint 0 <i>; yields int</i>
2015 <!-- _______________________________________________________________________ -->
2016 <div class="doc_subsubsection">
2017 <a name="i_insertelement">'<tt>insertelement</tt>' Instruction</a>
2020 <div class="doc_text">
2025 <result> = insertelement <n x <ty>> <val>, <ty> <elt>, uint <idx> <i>; yields <n x <ty>></i>
2031 The '<tt>insertelement</tt>' instruction inserts a scalar
2032 element into a packed vector at a specified index.
2039 The first operand of an '<tt>insertelement</tt>' instruction is a
2040 value of <a href="#t_packed">packed</a> type. The second operand is a
2041 scalar value whose type must equal the element type of the first
2042 operand. The third operand is an index indicating the position at
2043 which to insert the value. The index may be a variable.</p>
2048 The result is a packed vector of the same type as <tt>val</tt>. Its
2049 element values are those of <tt>val</tt> except at position
2050 <tt>idx</tt>, where it gets the value <tt>elt</tt>. If <tt>idx</tt>
2051 exceeds the length of <tt>val</tt>, the results are undefined.
2057 %result = insertelement <4 x int> %vec, int 1, uint 0 <i>; yields <4 x int></i>
2061 <!-- _______________________________________________________________________ -->
2062 <div class="doc_subsubsection">
2063 <a name="i_shufflevector">'<tt>shufflevector</tt>' Instruction</a>
2066 <div class="doc_text">
2071 <result> = shufflevector <n x <ty>> <v1>, <n x <ty>> <v2>, <n x uint> <mask> <i>; yields <n x <ty>></i>
2077 The '<tt>shufflevector</tt>' instruction constructs a permutation of elements
2078 from two input vectors, returning a vector of the same type.
2084 The first two operands of a '<tt>shufflevector</tt>' instruction are vectors
2085 with types that match each other and types that match the result of the
2086 instruction. The third argument is a shuffle mask, which has the same number
2087 of elements as the other vector type, but whose element type is always 'uint'.
2091 The shuffle mask operand is required to be a constant vector with either
2092 constant integer or undef values.
2098 The elements of the two input vectors are numbered from left to right across
2099 both of the vectors. The shuffle mask operand specifies, for each element of
2100 the result vector, which element of the two input registers the result element
2101 gets. The element selector may be undef (meaning "don't care") and the second
2102 operand may be undef if performing a shuffle from only one vector.
2108 %result = shufflevector <4 x int> %v1, <4 x int> %v2,
2109 <4 x uint> <uint 0, uint 4, uint 1, uint 5> <i>; yields <4 x int></i>
2110 %result = shufflevector <4 x int> %v1, <4 x int> undef,
2111 <4 x uint> <uint 0, uint 1, uint 2, uint 3> <i>; yields <4 x int></i> - Identity shuffle.
2116 <!-- _______________________________________________________________________ -->
2117 <div class="doc_subsubsection"> <a name="i_vsetint">'<tt>vsetint</tt>'
2118 Instruction</a> </div>
2119 <div class="doc_text">
2121 <pre><result> = vsetint <op>, <n x <ty>> <var1>, <var2> <i>; yields <n x bool></i>
2126 <p>The '<tt>vsetint</tt>' instruction takes two integer vectors and
2127 returns a vector of boolean values representing, at each position, the
2128 result of the comparison between the values at that position in the
2133 <p>The arguments to a '<tt>vsetint</tt>' instruction are a comparison
2134 operation and two value arguments. The value arguments must be of <a
2135 href="#t_integral">integral</a> <a href="#t_packed">packed</a> type,
2136 and they must have identical types. The operation argument must be
2137 one of <tt>eq</tt>, <tt>ne</tt>, <tt>slt</tt>, <tt>sgt</tt>,
2138 <tt>sle</tt>, <tt>sge</tt>, <tt>ult</tt>, <tt>ugt</tt>, <tt>ule</tt>,
2139 <tt>uge</tt>, <tt>true</tt>, and <tt>false</tt>. The result is a
2140 packed <tt>bool</tt> value with the same length as each operand.</p>
2144 <p>The following table shows the semantics of '<tt>vsetint</tt>'. For
2145 each position of the result, the comparison is done on the
2146 corresponding positions of the two value arguments. Note that the
2147 signedness of the comparison depends on the comparison opcode and
2148 <i>not</i> on the signedness of the value operands. E.g., <tt>vsetint
2149 slt <4 x unsigned> %x, %y</tt> does an elementwise <i>signed</i>
2150 comparison of <tt>%x</tt> and <tt>%y</tt>.</p>
2152 <table border="1" cellspacing="0" cellpadding="4">
2154 <tr><th>Operation</th><th>Result is true iff</th><th>Comparison is</th></tr>
2155 <tr><td><tt>eq</tt></td><td>var1 == var2</td><td>--</td></tr>
2156 <tr><td><tt>ne</tt></td><td>var1 != var2</td><td>--</td></tr>
2157 <tr><td><tt>slt</tt></td><td>var1 < var2</td><td>signed</td></tr>
2158 <tr><td><tt>sgt</tt></td><td>var1 > var2</td><td>signed</td></tr>
2159 <tr><td><tt>sle</tt></td><td>var1 <= var2</td><td>signed</td></tr>
2160 <tr><td><tt>sge</tt></td><td>var1 >= var2</td><td>signed</td></tr>
2161 <tr><td><tt>ult</tt></td><td>var1 < var2</td><td>unsigned</td></tr>
2162 <tr><td><tt>ugt</tt></td><td>var1 > var2</td><td>unsigned</td></tr>
2163 <tr><td><tt>ule</tt></td><td>var1 <= var2</td><td>unsigned</td></tr>
2164 <tr><td><tt>uge</tt></td><td>var1 >= var2</td><td>unsigned</td></tr>
2165 <tr><td><tt>true</tt></td><td>always</td><td>--</td></tr>
2166 <tr><td><tt>false</tt></td><td>never</td><td>--</td></tr>
2171 <pre> <result> = vsetint eq <2 x int> <int 0, int 1>, <int 1, int 0> <i>; yields {<2 x bool>}:result = false, false</i>
2172 <result> = vsetint ne <2 x int> <int 0, int 1>, <int 1, int 0> <i>; yields {<2 x bool>}:result = true, true</i>
2173 <result> = vsetint slt <2 x int> <int 0, int 1>, <int 1, int 0> <i>; yields {<2 x bool>}:result = true, false</i>
2174 <result> = vsetint sgt <2 x int> <int 0, int 1>, <int 1, int 0> <i>; yields {<2 x bool>}:result = false, true</i>
2175 <result> = vsetint sle <2 x int> <int 0, int 1>, <int 1, int 0> <i>; yields {<2 x bool>}:result = true, false</i>
2176 <result> = vsetint sge <2 x int> <int 0, int 1>, <int 1, int 0> <i>; yields {<2 x bool>}:result = false, true</i>
2180 <!-- _______________________________________________________________________ -->
2181 <div class="doc_subsubsection"> <a name="i_vsetfp">'<tt>vsetfp</tt>'
2182 Instruction</a> </div>
2183 <div class="doc_text">
2185 <pre><result> = vsetfp <op>, <n x <ty>> <var1>, <var2> <i>; yields <n x bool></i>
2190 <p>The '<tt>vsetfp</tt>' instruction takes two floating point vector
2191 arguments and returns a vector of boolean values representing, at each
2192 position, the result of the comparison between the values at that
2193 position in the two operands.</p>
2197 <p>The arguments to a '<tt>vsetfp</tt>' instruction are a comparison
2198 operation and two value arguments. The value arguments must be of <a
2199 href="t_floating">floating point</a> <a href="#t_packed">packed</a>
2200 type, and they must have identical types. The operation argument must
2201 be one of <tt>eq</tt>, <tt>ne</tt>, <tt>lt</tt>, <tt>gt</tt>,
2202 <tt>le</tt>, <tt>ge</tt>, <tt>oeq</tt>, <tt>one</tt>, <tt>olt</tt>,
2203 <tt>ogt</tt>, <tt>ole</tt>, <tt>oge</tt>, <tt>ueq</tt>, <tt>une</tt>,
2204 <tt>ult</tt>, <tt>ugt</tt>, <tt>ule</tt>, <tt>uge</tt>, <tt>o</tt>,
2205 <tt>u</tt>, <tt>true</tt>, and <tt>false</tt>. The result is a packed
2206 <tt>bool</tt> value with the same length as each operand.</p>
2210 <p>The following table shows the semantics of '<tt>vsetfp</tt>' for
2211 floating point types. If either operand is a floating point Not a
2212 Number (NaN) value, the operation is unordered, and the value in the
2213 first column below is produced at that position. Otherwise, the
2214 operation is ordered, and the value in the second column is
2217 <table border="1" cellspacing="0" cellpadding="4">
2219 <tr><th>Operation</th><th>If unordered<th>Otherwise true iff</th></tr>
2220 <tr><td><tt>eq</tt></td><td>undefined</td><td>var1 == var2</td></tr>
2221 <tr><td><tt>ne</tt></td><td>undefined</td><td>var1 != var2</td></tr>
2222 <tr><td><tt>lt</tt></td><td>undefined</td><td>var1 < var2</td></tr>
2223 <tr><td><tt>gt</tt></td><td>undefined</td><td>var1 > var2</td></tr>
2224 <tr><td><tt>le</tt></td><td>undefined</td><td>var1 <= var2</td></tr>
2225 <tr><td><tt>ge</tt></td><td>undefined</td><td>var1 >= var2</td></tr>
2226 <tr><td><tt>oeq</tt></td><td>false</td><td>var1 == var2</td></tr>
2227 <tr><td><tt>one</tt></td><td>false</td><td>var1 != var2</td></tr>
2228 <tr><td><tt>olt</tt></td><td>false</td><td>var1 < var2</td></tr>
2229 <tr><td><tt>ogt</tt></td><td>false</td><td>var1 > var2</td></tr>
2230 <tr><td><tt>ole</tt></td><td>false</td><td>var1 <= var2</td></tr>
2231 <tr><td><tt>oge</tt></td><td>false</td><td>var1 >= var2</td></tr>
2232 <tr><td><tt>ueq</tt></td><td>true</td><td>var1 == var2</td></tr>
2233 <tr><td><tt>une</tt></td><td>true</td><td>var1 != var2</td></tr>
2234 <tr><td><tt>ult</tt></td><td>true</td><td>var1 < var2</td></tr>
2235 <tr><td><tt>ugt</tt></td><td>true</td><td>var1 > var2</td></tr>
2236 <tr><td><tt>ule</tt></td><td>true</td><td>var1 <= var2</td></tr>
2237 <tr><td><tt>uge</tt></td><td>true</td><td>var1 >= var2</td></tr>
2238 <tr><td><tt>o</tt></td><td>false</td><td>always</td></tr>
2239 <tr><td><tt>u</tt></td><td>true</td><td>never</td></tr>
2240 <tr><td><tt>true</tt></td><td>true</td><td>always</td></tr>
2241 <tr><td><tt>false</tt></td><td>false</td><td>never</td></tr>
2246 <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>
2247 <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>
2248 <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>
2249 <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>
2250 <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>
2251 <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>
2255 <!-- _______________________________________________________________________ -->
2256 <div class="doc_subsubsection">
2257 <a name="i_vselect">'<tt>vselect</tt>' Instruction</a>
2260 <div class="doc_text">
2265 <result> = vselect <n x bool> <cond>, <n x <ty>> <val1>, <n x <ty>> <val2> <i>; yields <n x <ty>></i>
2271 The '<tt>vselect</tt>' instruction chooses one value at each position
2272 of a vector based on a condition.
2279 The '<tt>vselect</tt>' instruction requires a <a
2280 href="#t_packed">packed</a> <tt>bool</tt> value indicating the
2281 condition at each vector position, and two values of the same packed
2282 type. All three operands must have the same length. The type of the
2283 result is the same as the type of the two value operands.</p>
2288 At each position where the <tt>bool</tt> vector is true, that position
2289 of the result gets its value from the first value argument; otherwise,
2290 it gets its value from the second value argument.
2296 %X = vselect bool <2 x bool> <bool true, bool false>, <2 x ubyte> <ubyte 17, ubyte 17>,
2297 <2 x ubyte> <ubyte 42, ubyte 42> <i>; yields <2 x ubyte>:17, 42</i>
2303 <!-- ======================================================================= -->
2304 <div class="doc_subsection">
2305 <a name="memoryops">Memory Access and Addressing Operations</a>
2308 <div class="doc_text">
2310 <p>A key design point of an SSA-based representation is how it
2311 represents memory. In LLVM, no memory locations are in SSA form, which
2312 makes things very simple. This section describes how to read, write,
2313 allocate, and free memory in LLVM.</p>
2317 <!-- _______________________________________________________________________ -->
2318 <div class="doc_subsubsection">
2319 <a name="i_malloc">'<tt>malloc</tt>' Instruction</a>
2322 <div class="doc_text">
2327 <result> = malloc <type>[, uint <NumElements>][, align <alignment>] <i>; yields {type*}:result</i>
2332 <p>The '<tt>malloc</tt>' instruction allocates memory from the system
2333 heap and returns a pointer to it.</p>
2337 <p>The '<tt>malloc</tt>' instruction allocates
2338 <tt>sizeof(<type>)*NumElements</tt>
2339 bytes of memory from the operating system and returns a pointer of the
2340 appropriate type to the program. If "NumElements" is specified, it is the
2341 number of elements allocated. If an alignment is specified, the value result
2342 of the allocation is guaranteed to be aligned to at least that boundary. If
2343 not specified, or if zero, the target can choose to align the allocation on any
2344 convenient boundary.</p>
2346 <p>'<tt>type</tt>' must be a sized type.</p>
2350 <p>Memory is allocated using the system "<tt>malloc</tt>" function, and
2351 a pointer is returned.</p>
2356 %array = malloc [4 x ubyte ] <i>; yields {[%4 x ubyte]*}:array</i>
2358 %size = <a href="#i_add">add</a> uint 2, 2 <i>; yields {uint}:size = uint 4</i>
2359 %array1 = malloc ubyte, uint 4 <i>; yields {ubyte*}:array1</i>
2360 %array2 = malloc [12 x ubyte], uint %size <i>; yields {[12 x ubyte]*}:array2</i>
2361 %array3 = malloc int, uint 4, align 1024 <i>; yields {int*}:array3</i>
2362 %array4 = malloc int, align 1024 <i>; yields {int*}:array4</i>
2366 <!-- _______________________________________________________________________ -->
2367 <div class="doc_subsubsection">
2368 <a name="i_free">'<tt>free</tt>' Instruction</a>
2371 <div class="doc_text">
2376 free <type> <value> <i>; yields {void}</i>
2381 <p>The '<tt>free</tt>' instruction returns memory back to the unused
2382 memory heap to be reallocated in the future.</p>
2386 <p>'<tt>value</tt>' shall be a pointer value that points to a value
2387 that was allocated with the '<tt><a href="#i_malloc">malloc</a></tt>'
2392 <p>Access to the memory pointed to by the pointer is no longer defined
2393 after this instruction executes.</p>
2398 %array = <a href="#i_malloc">malloc</a> [4 x ubyte] <i>; yields {[4 x ubyte]*}:array</i>
2399 free [4 x ubyte]* %array
2403 <!-- _______________________________________________________________________ -->
2404 <div class="doc_subsubsection">
2405 <a name="i_alloca">'<tt>alloca</tt>' Instruction</a>
2408 <div class="doc_text">
2413 <result> = alloca <type>[, uint <NumElements>][, align <alignment>] <i>; yields {type*}:result</i>
2418 <p>The '<tt>alloca</tt>' instruction allocates memory on the current
2419 stack frame of the procedure that is live until the current function
2420 returns to its caller.</p>
2424 <p>The '<tt>alloca</tt>' instruction allocates <tt>sizeof(<type>)*NumElements</tt>
2425 bytes of memory on the runtime stack, returning a pointer of the
2426 appropriate type to the program. If "NumElements" is specified, it is the
2427 number of elements allocated. If an alignment is specified, the value result
2428 of the allocation is guaranteed to be aligned to at least that boundary. If
2429 not specified, or if zero, the target can choose to align the allocation on any
2430 convenient boundary.</p>
2432 <p>'<tt>type</tt>' may be any sized type.</p>
2436 <p>Memory is allocated; a pointer is returned. '<tt>alloca</tt>'d
2437 memory is automatically released when the function returns. The '<tt>alloca</tt>'
2438 instruction is commonly used to represent automatic variables that must
2439 have an address available. When the function returns (either with the <tt><a
2440 href="#i_ret">ret</a></tt> or <tt><a href="#i_unwind">unwind</a></tt>
2441 instructions), the memory is reclaimed.</p>
2446 %ptr = alloca int <i>; yields {int*}:ptr</i>
2447 %ptr = alloca int, uint 4 <i>; yields {int*}:ptr</i>
2448 %ptr = alloca int, uint 4, align 1024 <i>; yields {int*}:ptr</i>
2449 %ptr = alloca int, align 1024 <i>; yields {int*}:ptr</i>
2453 <!-- _______________________________________________________________________ -->
2454 <div class="doc_subsubsection"> <a name="i_load">'<tt>load</tt>'
2455 Instruction</a> </div>
2456 <div class="doc_text">
2458 <pre> <result> = load <ty>* <pointer><br> <result> = volatile load <ty>* <pointer><br></pre>
2460 <p>The '<tt>load</tt>' instruction is used to read from memory.</p>
2462 <p>The argument to the '<tt>load</tt>' instruction specifies the memory
2463 address from which to load. The pointer must point to a <a
2464 href="#t_firstclass">first class</a> type. If the <tt>load</tt> is
2465 marked as <tt>volatile</tt>, then the optimizer is not allowed to modify
2466 the number or order of execution of this <tt>load</tt> with other
2467 volatile <tt>load</tt> and <tt><a href="#i_store">store</a></tt>
2470 <p>The location of memory pointed to is loaded.</p>
2472 <pre> %ptr = <a href="#i_alloca">alloca</a> int <i>; yields {int*}:ptr</i>
2474 href="#i_store">store</a> int 3, int* %ptr <i>; yields {void}</i>
2475 %val = load int* %ptr <i>; yields {int}:val = int 3</i>
2478 <!-- _______________________________________________________________________ -->
2479 <div class="doc_subsubsection"> <a name="i_store">'<tt>store</tt>'
2480 Instruction</a> </div>
2482 <pre> store <ty> <value>, <ty>* <pointer> <i>; yields {void}</i>
2483 volatile store <ty> <value>, <ty>* <pointer> <i>; yields {void}</i>
2486 <p>The '<tt>store</tt>' instruction is used to write to memory.</p>
2488 <p>There are two arguments to the '<tt>store</tt>' instruction: a value
2489 to store and an address in which to store it. The type of the '<tt><pointer></tt>'
2490 operand must be a pointer to the type of the '<tt><value></tt>'
2491 operand. If the <tt>store</tt> is marked as <tt>volatile</tt>, then the
2492 optimizer is not allowed to modify the number or order of execution of
2493 this <tt>store</tt> with other volatile <tt>load</tt> and <tt><a
2494 href="#i_store">store</a></tt> instructions.</p>
2496 <p>The contents of memory are updated to contain '<tt><value></tt>'
2497 at the location specified by the '<tt><pointer></tt>' operand.</p>
2499 <pre> %ptr = <a href="#i_alloca">alloca</a> int <i>; yields {int*}:ptr</i>
2501 href="#i_store">store</a> int 3, int* %ptr <i>; yields {void}</i>
2502 %val = load int* %ptr <i>; yields {int}:val = int 3</i>
2504 <!-- _______________________________________________________________________ -->
2505 <div class="doc_subsubsection">
2506 <a name="i_getelementptr">'<tt>getelementptr</tt>' Instruction</a>
2509 <div class="doc_text">
2512 <result> = getelementptr <ty>* <ptrval>{, <ty> <idx>}*
2518 The '<tt>getelementptr</tt>' instruction is used to get the address of a
2519 subelement of an aggregate data structure.</p>
2523 <p>This instruction takes a list of integer constants that indicate what
2524 elements of the aggregate object to index to. The actual types of the arguments
2525 provided depend on the type of the first pointer argument. The
2526 '<tt>getelementptr</tt>' instruction is used to index down through the type
2527 levels of a structure or to a specific index in an array. When indexing into a
2528 structure, only <tt>uint</tt>
2529 integer constants are allowed. When indexing into an array or pointer,
2530 <tt>int</tt> and <tt>long</tt> indexes are allowed of any sign.</p>
2532 <p>For example, let's consider a C code fragment and how it gets
2533 compiled to LLVM:</p>
2547 int *foo(struct ST *s) {
2548 return &s[1].Z.B[5][13];
2552 <p>The LLVM code generated by the GCC frontend is:</p>
2555 %RT = type { sbyte, [10 x [20 x int]], sbyte }
2556 %ST = type { int, double, %RT }
2560 int* %foo(%ST* %s) {
2562 %reg = getelementptr %ST* %s, int 1, uint 2, uint 1, int 5, int 13
2569 <p>The index types specified for the '<tt>getelementptr</tt>' instruction depend
2570 on the pointer type that is being indexed into. <a href="#t_pointer">Pointer</a>
2571 and <a href="#t_array">array</a> types require <tt>uint</tt>, <tt>int</tt>,
2572 <tt>ulong</tt>, or <tt>long</tt> values, and <a href="#t_struct">structure</a>
2573 types require <tt>uint</tt> <b>constants</b>.</p>
2575 <p>In the example above, the first index is indexing into the '<tt>%ST*</tt>'
2576 type, which is a pointer, yielding a '<tt>%ST</tt>' = '<tt>{ int, double, %RT
2577 }</tt>' type, a structure. The second index indexes into the third element of
2578 the structure, yielding a '<tt>%RT</tt>' = '<tt>{ sbyte, [10 x [20 x int]],
2579 sbyte }</tt>' type, another structure. The third index indexes into the second
2580 element of the structure, yielding a '<tt>[10 x [20 x int]]</tt>' type, an
2581 array. The two dimensions of the array are subscripted into, yielding an
2582 '<tt>int</tt>' type. The '<tt>getelementptr</tt>' instruction returns a pointer
2583 to this element, thus computing a value of '<tt>int*</tt>' type.</p>
2585 <p>Note that it is perfectly legal to index partially through a
2586 structure, returning a pointer to an inner element. Because of this,
2587 the LLVM code for the given testcase is equivalent to:</p>
2590 int* %foo(%ST* %s) {
2591 %t1 = getelementptr %ST* %s, int 1 <i>; yields %ST*:%t1</i>
2592 %t2 = getelementptr %ST* %t1, int 0, uint 2 <i>; yields %RT*:%t2</i>
2593 %t3 = getelementptr %RT* %t2, int 0, uint 1 <i>; yields [10 x [20 x int]]*:%t3</i>
2594 %t4 = getelementptr [10 x [20 x int]]* %t3, int 0, int 5 <i>; yields [20 x int]*:%t4</i>
2595 %t5 = getelementptr [20 x int]* %t4, int 0, int 13 <i>; yields int*:%t5</i>
2600 <p>Note that it is undefined to access an array out of bounds: array and
2601 pointer indexes must always be within the defined bounds of the array type.
2602 The one exception for this rules is zero length arrays. These arrays are
2603 defined to be accessible as variable length arrays, which requires access
2604 beyond the zero'th element.</p>
2606 <p>The getelementptr instruction is often confusing. For some more insight
2607 into how it works, see <a href="GetElementPtr.html">the getelementptr
2613 <i>; yields [12 x ubyte]*:aptr</i>
2614 %aptr = getelementptr {int, [12 x ubyte]}* %sptr, long 0, uint 1
2618 <!-- ======================================================================= -->
2619 <div class="doc_subsection"> <a name="otherops">Other Operations</a> </div>
2620 <div class="doc_text">
2621 <p>The instructions in this category are the "miscellaneous"
2622 instructions, which defy better classification.</p>
2624 <!-- _______________________________________________________________________ -->
2625 <div class="doc_subsubsection"> <a name="i_phi">'<tt>phi</tt>'
2626 Instruction</a> </div>
2627 <div class="doc_text">
2629 <pre> <result> = phi <ty> [ <val0>, <label0>], ...<br></pre>
2631 <p>The '<tt>phi</tt>' instruction is used to implement the φ node in
2632 the SSA graph representing the function.</p>
2634 <p>The type of the incoming values are specified with the first type
2635 field. After this, the '<tt>phi</tt>' instruction takes a list of pairs
2636 as arguments, with one pair for each predecessor basic block of the
2637 current block. Only values of <a href="#t_firstclass">first class</a>
2638 type may be used as the value arguments to the PHI node. Only labels
2639 may be used as the label arguments.</p>
2640 <p>There must be no non-phi instructions between the start of a basic
2641 block and the PHI instructions: i.e. PHI instructions must be first in
2644 <p>At runtime, the '<tt>phi</tt>' instruction logically takes on the
2645 value specified by the parameter, depending on which basic block we
2646 came from in the last <a href="#terminators">terminator</a> instruction.</p>
2648 <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>
2651 <!-- _______________________________________________________________________ -->
2652 <div class="doc_subsubsection">
2653 <a name="i_cast">'<tt>cast .. to</tt>' Instruction</a>
2656 <div class="doc_text">
2661 <result> = cast <ty> <value> to <ty2> <i>; yields ty2</i>
2667 The '<tt>cast</tt>' instruction is used as the primitive means to convert
2668 integers to floating point, change data type sizes, and break type safety (by
2676 The '<tt>cast</tt>' instruction takes a value to cast, which must be a first
2677 class value, and a type to cast it to, which must also be a <a
2678 href="#t_firstclass">first class</a> type.
2684 This instruction follows the C rules for explicit casts when determining how the
2685 data being cast must change to fit in its new container.
2689 When casting to bool, any value that would be considered true in the context of
2690 a C '<tt>if</tt>' condition is converted to the boolean '<tt>true</tt>' values,
2691 all else are '<tt>false</tt>'.
2695 When extending an integral value from a type of one signness to another (for
2696 example '<tt>sbyte</tt>' to '<tt>ulong</tt>'), the value is sign-extended if the
2697 <b>source</b> value is signed, and zero-extended if the source value is
2698 unsigned. <tt>bool</tt> values are always zero extended into either zero or
2705 %X = cast int 257 to ubyte <i>; yields ubyte:1</i>
2706 %Y = cast int 123 to bool <i>; yields bool:true</i>
2710 <!-- _______________________________________________________________________ -->
2711 <div class="doc_subsubsection">
2712 <a name="i_select">'<tt>select</tt>' Instruction</a>
2715 <div class="doc_text">
2720 <result> = select bool <cond>, <ty> <val1>, <ty> <val2> <i>; yields ty</i>
2726 The '<tt>select</tt>' instruction is used to choose one value based on a
2727 condition, without branching.
2734 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.
2740 If the boolean condition evaluates to true, the instruction returns the first
2741 value argument; otherwise, it returns the second value argument.
2747 %X = select bool true, ubyte 17, ubyte 42 <i>; yields ubyte:17</i>
2752 <!-- _______________________________________________________________________ -->
2753 <div class="doc_subsubsection">
2754 <a name="i_call">'<tt>call</tt>' Instruction</a>
2757 <div class="doc_text">
2761 <result> = [tail] call [<a href="#callingconv">cconv</a>] <ty>* <fnptrval>(<param list>)
2766 <p>The '<tt>call</tt>' instruction represents a simple function call.</p>
2770 <p>This instruction requires several arguments:</p>
2774 <p>The optional "tail" marker indicates whether the callee function accesses
2775 any allocas or varargs in the caller. If the "tail" marker is present, the
2776 function call is eligible for tail call optimization. Note that calls may
2777 be marked "tail" even if they do not occur before a <a
2778 href="#i_ret"><tt>ret</tt></a> instruction.
2781 <p>The optional "cconv" marker indicates which <a href="callingconv">calling
2782 convention</a> the call should use. If none is specified, the call defaults
2783 to using C calling conventions.
2786 <p>'<tt>ty</tt>': shall be the signature of the pointer to function value
2787 being invoked. The argument types must match the types implied by this
2788 signature. This type can be omitted if the function is not varargs and
2789 if the function type does not return a pointer to a function.</p>
2792 <p>'<tt>fnptrval</tt>': An LLVM value containing a pointer to a function to
2793 be invoked. In most cases, this is a direct function invocation, but
2794 indirect <tt>call</tt>s are just as possible, calling an arbitrary pointer
2795 to function value.</p>
2798 <p>'<tt>function args</tt>': argument list whose types match the
2799 function signature argument types. All arguments must be of
2800 <a href="#t_firstclass">first class</a> type. If the function signature
2801 indicates the function accepts a variable number of arguments, the extra
2802 arguments can be specified.</p>
2808 <p>The '<tt>call</tt>' instruction is used to cause control flow to
2809 transfer to a specified function, with its incoming arguments bound to
2810 the specified values. Upon a '<tt><a href="#i_ret">ret</a></tt>'
2811 instruction in the called function, control flow continues with the
2812 instruction after the function call, and the return value of the
2813 function is bound to the result argument. This is a simpler case of
2814 the <a href="#i_invoke">invoke</a> instruction.</p>
2819 %retval = call int %test(int %argc)
2820 call int(sbyte*, ...) *%printf(sbyte* %msg, int 12, sbyte 42);
2821 %X = tail call int %foo()
2822 %Y = tail call <a href="#callingconv">fastcc</a> int %foo()
2827 <!-- _______________________________________________________________________ -->
2828 <div class="doc_subsubsection">
2829 <a name="i_va_arg">'<tt>va_arg</tt>' Instruction</a>
2832 <div class="doc_text">
2837 <resultval> = va_arg <va_list*> <arglist>, <argty>
2842 <p>The '<tt>va_arg</tt>' instruction is used to access arguments passed through
2843 the "variable argument" area of a function call. It is used to implement the
2844 <tt>va_arg</tt> macro in C.</p>
2848 <p>This instruction takes a <tt>va_list*</tt> value and the type of
2849 the argument. It returns a value of the specified argument type and
2850 increments the <tt>va_list</tt> to point to the next argument. Again, the
2851 actual type of <tt>va_list</tt> is target specific.</p>
2855 <p>The '<tt>va_arg</tt>' instruction loads an argument of the specified
2856 type from the specified <tt>va_list</tt> and causes the
2857 <tt>va_list</tt> to point to the next argument. For more information,
2858 see the variable argument handling <a href="#int_varargs">Intrinsic
2861 <p>It is legal for this instruction to be called in a function which does not
2862 take a variable number of arguments, for example, the <tt>vfprintf</tt>
2865 <p><tt>va_arg</tt> is an LLVM instruction instead of an <a
2866 href="#intrinsics">intrinsic function</a> because it takes a type as an
2871 <p>See the <a href="#int_varargs">variable argument processing</a> section.</p>
2875 <!-- *********************************************************************** -->
2876 <div class="doc_section"> <a name="intrinsics">Intrinsic Functions</a> </div>
2877 <!-- *********************************************************************** -->
2879 <div class="doc_text">
2881 <p>LLVM supports the notion of an "intrinsic function". These functions have
2882 well known names and semantics and are required to follow certain
2883 restrictions. Overall, these instructions represent an extension mechanism for
2884 the LLVM language that does not require changing all of the transformations in
2885 LLVM to add to the language (or the bytecode reader/writer, the parser,
2888 <p>Intrinsic function names must all start with an "<tt>llvm.</tt>" prefix. This
2889 prefix is reserved in LLVM for intrinsic names; thus, functions may not be named
2890 this. Intrinsic functions must always be external functions: you cannot define
2891 the body of intrinsic functions. Intrinsic functions may only be used in call
2892 or invoke instructions: it is illegal to take the address of an intrinsic
2893 function. Additionally, because intrinsic functions are part of the LLVM
2894 language, it is required that they all be documented here if any are added.</p>
2897 <p>To learn how to add an intrinsic function, please see the <a
2898 href="ExtendingLLVM.html">Extending LLVM Guide</a>.
2903 <!-- ======================================================================= -->
2904 <div class="doc_subsection">
2905 <a name="int_varargs">Variable Argument Handling Intrinsics</a>
2908 <div class="doc_text">
2910 <p>Variable argument support is defined in LLVM with the <a
2911 href="#i_va_arg"><tt>va_arg</tt></a> instruction and these three
2912 intrinsic functions. These functions are related to the similarly
2913 named macros defined in the <tt><stdarg.h></tt> header file.</p>
2915 <p>All of these functions operate on arguments that use a
2916 target-specific value type "<tt>va_list</tt>". The LLVM assembly
2917 language reference manual does not define what this type is, so all
2918 transformations should be prepared to handle intrinsics with any type
2921 <p>This example shows how the <a href="#i_va_arg"><tt>va_arg</tt></a>
2922 instruction and the variable argument handling intrinsic functions are
2926 int %test(int %X, ...) {
2927 ; Initialize variable argument processing
2929 call void %<a href="#i_va_start">llvm.va_start</a>(sbyte** %ap)
2931 ; Read a single integer argument
2932 %tmp = va_arg sbyte** %ap, int
2934 ; Demonstrate usage of llvm.va_copy and llvm.va_end
2936 call void %<a href="#i_va_copy">llvm.va_copy</a>(sbyte** %aq, sbyte** %ap)
2937 call void %<a href="#i_va_end">llvm.va_end</a>(sbyte** %aq)
2939 ; Stop processing of arguments.
2940 call void %<a href="#i_va_end">llvm.va_end</a>(sbyte** %ap)
2946 <!-- _______________________________________________________________________ -->
2947 <div class="doc_subsubsection">
2948 <a name="i_va_start">'<tt>llvm.va_start</tt>' Intrinsic</a>
2952 <div class="doc_text">
2954 <pre> declare void %llvm.va_start(<va_list>* <arglist>)<br></pre>
2956 <P>The '<tt>llvm.va_start</tt>' intrinsic initializes
2957 <tt>*<arglist></tt> for subsequent use by <tt><a
2958 href="#i_va_arg">va_arg</a></tt>.</p>
2962 <P>The argument is a pointer to a <tt>va_list</tt> element to initialize.</p>
2966 <P>The '<tt>llvm.va_start</tt>' intrinsic works just like the <tt>va_start</tt>
2967 macro available in C. In a target-dependent way, it initializes the
2968 <tt>va_list</tt> element the argument points to, so that the next call to
2969 <tt>va_arg</tt> will produce the first variable argument passed to the function.
2970 Unlike the C <tt>va_start</tt> macro, this intrinsic does not need to know the
2971 last argument of the function, the compiler can figure that out.</p>
2975 <!-- _______________________________________________________________________ -->
2976 <div class="doc_subsubsection">
2977 <a name="i_va_end">'<tt>llvm.va_end</tt>' Intrinsic</a>
2980 <div class="doc_text">
2982 <pre> declare void %llvm.va_end(<va_list*> <arglist>)<br></pre>
2984 <p>The '<tt>llvm.va_end</tt>' intrinsic destroys <tt><arglist></tt>
2985 which has been initialized previously with <tt><a href="#i_va_start">llvm.va_start</a></tt>
2986 or <tt><a href="#i_va_copy">llvm.va_copy</a></tt>.</p>
2988 <p>The argument is a <tt>va_list</tt> to destroy.</p>
2990 <p>The '<tt>llvm.va_end</tt>' intrinsic works just like the <tt>va_end</tt>
2991 macro available in C. In a target-dependent way, it destroys the <tt>va_list</tt>.
2992 Calls to <a href="#i_va_start"><tt>llvm.va_start</tt></a> and <a
2993 href="#i_va_copy"><tt>llvm.va_copy</tt></a> must be matched exactly
2994 with calls to <tt>llvm.va_end</tt>.</p>
2997 <!-- _______________________________________________________________________ -->
2998 <div class="doc_subsubsection">
2999 <a name="i_va_copy">'<tt>llvm.va_copy</tt>' Intrinsic</a>
3002 <div class="doc_text">
3007 declare void %llvm.va_copy(<va_list>* <destarglist>,
3008 <va_list>* <srcarglist>)
3013 <p>The '<tt>llvm.va_copy</tt>' intrinsic copies the current argument position from
3014 the source argument list to the destination argument list.</p>
3018 <p>The first argument is a pointer to a <tt>va_list</tt> element to initialize.
3019 The second argument is a pointer to a <tt>va_list</tt> element to copy from.</p>
3024 <p>The '<tt>llvm.va_copy</tt>' intrinsic works just like the <tt>va_copy</tt> macro
3025 available in C. In a target-dependent way, it copies the source
3026 <tt>va_list</tt> element into the destination list. This intrinsic is necessary
3027 because the <tt><a href="i_va_begin">llvm.va_begin</a></tt> intrinsic may be
3028 arbitrarily complex and require memory allocation, for example.</p>
3032 <!-- ======================================================================= -->
3033 <div class="doc_subsection">
3034 <a name="int_gc">Accurate Garbage Collection Intrinsics</a>
3037 <div class="doc_text">
3040 LLVM support for <a href="GarbageCollection.html">Accurate Garbage
3041 Collection</a> requires the implementation and generation of these intrinsics.
3042 These intrinsics allow identification of <a href="#i_gcroot">GC roots on the
3043 stack</a>, as well as garbage collector implementations that require <a
3044 href="#i_gcread">read</a> and <a href="#i_gcwrite">write</a> barriers.
3045 Front-ends for type-safe garbage collected languages should generate these
3046 intrinsics to make use of the LLVM garbage collectors. For more details, see <a
3047 href="GarbageCollection.html">Accurate Garbage Collection with LLVM</a>.
3051 <!-- _______________________________________________________________________ -->
3052 <div class="doc_subsubsection">
3053 <a name="i_gcroot">'<tt>llvm.gcroot</tt>' Intrinsic</a>
3056 <div class="doc_text">
3061 declare void %llvm.gcroot(<ty>** %ptrloc, <ty2>* %metadata)
3066 <p>The '<tt>llvm.gcroot</tt>' intrinsic declares the existence of a GC root to
3067 the code generator, and allows some metadata to be associated with it.</p>
3071 <p>The first argument specifies the address of a stack object that contains the
3072 root pointer. The second pointer (which must be either a constant or a global
3073 value address) contains the meta-data to be associated with the root.</p>
3077 <p>At runtime, a call to this intrinsics stores a null pointer into the "ptrloc"
3078 location. At compile-time, the code generator generates information to allow
3079 the runtime to find the pointer at GC safe points.
3085 <!-- _______________________________________________________________________ -->
3086 <div class="doc_subsubsection">
3087 <a name="i_gcread">'<tt>llvm.gcread</tt>' Intrinsic</a>
3090 <div class="doc_text">
3095 declare sbyte* %llvm.gcread(sbyte* %ObjPtr, sbyte** %Ptr)
3100 <p>The '<tt>llvm.gcread</tt>' intrinsic identifies reads of references from heap
3101 locations, allowing garbage collector implementations that require read
3106 <p>The second argument is the address to read from, which should be an address
3107 allocated from the garbage collector. The first object is a pointer to the
3108 start of the referenced object, if needed by the language runtime (otherwise
3113 <p>The '<tt>llvm.gcread</tt>' intrinsic has the same semantics as a load
3114 instruction, but may be replaced with substantially more complex code by the
3115 garbage collector runtime, as needed.</p>
3120 <!-- _______________________________________________________________________ -->
3121 <div class="doc_subsubsection">
3122 <a name="i_gcwrite">'<tt>llvm.gcwrite</tt>' Intrinsic</a>
3125 <div class="doc_text">
3130 declare void %llvm.gcwrite(sbyte* %P1, sbyte* %Obj, sbyte** %P2)
3135 <p>The '<tt>llvm.gcwrite</tt>' intrinsic identifies writes of references to heap
3136 locations, allowing garbage collector implementations that require write
3137 barriers (such as generational or reference counting collectors).</p>
3141 <p>The first argument is the reference to store, the second is the start of the
3142 object to store it to, and the third is the address of the field of Obj to
3143 store to. If the runtime does not require a pointer to the object, Obj may be
3148 <p>The '<tt>llvm.gcwrite</tt>' intrinsic has the same semantics as a store
3149 instruction, but may be replaced with substantially more complex code by the
3150 garbage collector runtime, as needed.</p>
3156 <!-- ======================================================================= -->
3157 <div class="doc_subsection">
3158 <a name="int_codegen">Code Generator Intrinsics</a>
3161 <div class="doc_text">
3163 These intrinsics are provided by LLVM to expose special features that may only
3164 be implemented with code generator support.
3169 <!-- _______________________________________________________________________ -->
3170 <div class="doc_subsubsection">
3171 <a name="i_returnaddress">'<tt>llvm.returnaddress</tt>' Intrinsic</a>
3174 <div class="doc_text">
3178 declare sbyte *%llvm.returnaddress(uint <level>)
3184 The '<tt>llvm.returnaddress</tt>' intrinsic attempts to compute a
3185 target-specific value indicating the return address of the current function
3186 or one of its callers.
3192 The argument to this intrinsic indicates which function to return the address
3193 for. Zero indicates the calling function, one indicates its caller, etc. The
3194 argument is <b>required</b> to be a constant integer value.
3200 The '<tt>llvm.returnaddress</tt>' intrinsic either returns a pointer indicating
3201 the return address of the specified call frame, or zero if it cannot be
3202 identified. The value returned by this intrinsic is likely to be incorrect or 0
3203 for arguments other than zero, so it should only be used for debugging purposes.
3207 Note that calling this intrinsic does not prevent function inlining or other
3208 aggressive transformations, so the value returned may not be that of the obvious
3209 source-language caller.
3214 <!-- _______________________________________________________________________ -->
3215 <div class="doc_subsubsection">
3216 <a name="i_frameaddress">'<tt>llvm.frameaddress</tt>' Intrinsic</a>
3219 <div class="doc_text">
3223 declare sbyte *%llvm.frameaddress(uint <level>)
3229 The '<tt>llvm.frameaddress</tt>' intrinsic attempts to return the
3230 target-specific frame pointer value for the specified stack frame.
3236 The argument to this intrinsic indicates which function to return the frame
3237 pointer for. Zero indicates the calling function, one indicates its caller,
3238 etc. The argument is <b>required</b> to be a constant integer value.
3244 The '<tt>llvm.frameaddress</tt>' intrinsic either returns a pointer indicating
3245 the frame address of the specified call frame, or zero if it cannot be
3246 identified. The value returned by this intrinsic is likely to be incorrect or 0
3247 for arguments other than zero, so it should only be used for debugging purposes.
3251 Note that calling this intrinsic does not prevent function inlining or other
3252 aggressive transformations, so the value returned may not be that of the obvious
3253 source-language caller.
3257 <!-- _______________________________________________________________________ -->
3258 <div class="doc_subsubsection">
3259 <a name="i_stacksave">'<tt>llvm.stacksave</tt>' Intrinsic</a>
3262 <div class="doc_text">
3266 declare sbyte *%llvm.stacksave()
3272 The '<tt>llvm.stacksave</tt>' intrinsic is used to remember the current state of
3273 the function stack, for use with <a href="#i_stackrestore">
3274 <tt>llvm.stackrestore</tt></a>. This is useful for implementing language
3275 features like scoped automatic variable sized arrays in C99.
3281 This intrinsic returns a opaque pointer value that can be passed to <a
3282 href="#i_stackrestore"><tt>llvm.stackrestore</tt></a>. When an
3283 <tt>llvm.stackrestore</tt> intrinsic is executed with a value saved from
3284 <tt>llvm.stacksave</tt>, it effectively restores the state of the stack to the
3285 state it was in when the <tt>llvm.stacksave</tt> intrinsic executed. In
3286 practice, this pops any <a href="#i_alloca">alloca</a> blocks from the stack
3287 that were allocated after the <tt>llvm.stacksave</tt> was executed.
3292 <!-- _______________________________________________________________________ -->
3293 <div class="doc_subsubsection">
3294 <a name="i_stackrestore">'<tt>llvm.stackrestore</tt>' Intrinsic</a>
3297 <div class="doc_text">
3301 declare void %llvm.stackrestore(sbyte* %ptr)
3307 The '<tt>llvm.stackrestore</tt>' intrinsic is used to restore the state of
3308 the function stack to the state it was in when the corresponding <a
3309 href="#llvm.stacksave"><tt>llvm.stacksave</tt></a> intrinsic executed. This is
3310 useful for implementing language features like scoped automatic variable sized
3317 See the description for <a href="#i_stacksave"><tt>llvm.stacksave</tt></a>.
3323 <!-- _______________________________________________________________________ -->
3324 <div class="doc_subsubsection">
3325 <a name="i_prefetch">'<tt>llvm.prefetch</tt>' Intrinsic</a>
3328 <div class="doc_text">
3332 declare void %llvm.prefetch(sbyte * <address>,
3333 uint <rw>, uint <locality>)
3340 The '<tt>llvm.prefetch</tt>' intrinsic is a hint to the code generator to insert
3341 a prefetch instruction if supported; otherwise, it is a noop. Prefetches have
3343 effect on the behavior of the program but can change its performance
3350 <tt>address</tt> is the address to be prefetched, <tt>rw</tt> is the specifier
3351 determining if the fetch should be for a read (0) or write (1), and
3352 <tt>locality</tt> is a temporal locality specifier ranging from (0) - no
3353 locality, to (3) - extremely local keep in cache. The <tt>rw</tt> and
3354 <tt>locality</tt> arguments must be constant integers.
3360 This intrinsic does not modify the behavior of the program. In particular,
3361 prefetches cannot trap and do not produce a value. On targets that support this
3362 intrinsic, the prefetch can provide hints to the processor cache for better
3368 <!-- _______________________________________________________________________ -->
3369 <div class="doc_subsubsection">
3370 <a name="i_pcmarker">'<tt>llvm.pcmarker</tt>' Intrinsic</a>
3373 <div class="doc_text">
3377 declare void %llvm.pcmarker( uint <id> )
3384 The '<tt>llvm.pcmarker</tt>' intrinsic is a method to export a Program Counter
3386 code to simulators and other tools. The method is target specific, but it is
3387 expected that the marker will use exported symbols to transmit the PC of the marker.
3388 The marker makes no guarantees that it will remain with any specific instruction
3389 after optimizations. It is possible that the presence of a marker will inhibit
3390 optimizations. The intended use is to be inserted after optimizations to allow
3391 correlations of simulation runs.
3397 <tt>id</tt> is a numerical id identifying the marker.
3403 This intrinsic does not modify the behavior of the program. Backends that do not
3404 support this intrinisic may ignore it.
3409 <!-- _______________________________________________________________________ -->
3410 <div class="doc_subsubsection">
3411 <a name="i_readcyclecounter">'<tt>llvm.readcyclecounter</tt>' Intrinsic</a>
3414 <div class="doc_text">
3418 declare ulong %llvm.readcyclecounter( )
3425 The '<tt>llvm.readcyclecounter</tt>' intrinsic provides access to the cycle
3426 counter register (or similar low latency, high accuracy clocks) on those targets
3427 that support it. On X86, it should map to RDTSC. On Alpha, it should map to RPCC.
3428 As the backing counters overflow quickly (on the order of 9 seconds on alpha), this
3429 should only be used for small timings.
3435 When directly supported, reading the cycle counter should not modify any memory.
3436 Implementations are allowed to either return a application specific value or a
3437 system wide value. On backends without support, this is lowered to a constant 0.
3442 <!-- ======================================================================= -->
3443 <div class="doc_subsection">
3444 <a name="int_libc">Standard C Library Intrinsics</a>
3447 <div class="doc_text">
3449 LLVM provides intrinsics for a few important standard C library functions.
3450 These intrinsics allow source-language front-ends to pass information about the
3451 alignment of the pointer arguments to the code generator, providing opportunity
3452 for more efficient code generation.
3457 <!-- _______________________________________________________________________ -->
3458 <div class="doc_subsubsection">
3459 <a name="i_memcpy">'<tt>llvm.memcpy</tt>' Intrinsic</a>
3462 <div class="doc_text">
3466 declare void %llvm.memcpy.i32(sbyte* <dest>, sbyte* <src>,
3467 uint <len>, uint <align>)
3468 declare void %llvm.memcpy.i64(sbyte* <dest>, sbyte* <src>,
3469 ulong <len>, uint <align>)
3475 The '<tt>llvm.memcpy.*</tt>' intrinsics copy a block of memory from the source
3476 location to the destination location.
3480 Note that, unlike the standard libc function, the <tt>llvm.memcpy.*</tt>
3481 intrinsics do not return a value, and takes an extra alignment argument.
3487 The first argument is a pointer to the destination, the second is a pointer to
3488 the source. The third argument is an integer argument
3489 specifying the number of bytes to copy, and the fourth argument is the alignment
3490 of the source and destination locations.
3494 If the call to this intrinisic has an alignment value that is not 0 or 1, then
3495 the caller guarantees that both the source and destination pointers are aligned
3502 The '<tt>llvm.memcpy.*</tt>' intrinsics copy a block of memory from the source
3503 location to the destination location, which are not allowed to overlap. It
3504 copies "len" bytes of memory over. If the argument is known to be aligned to
3505 some boundary, this can be specified as the fourth argument, otherwise it should
3511 <!-- _______________________________________________________________________ -->
3512 <div class="doc_subsubsection">
3513 <a name="i_memmove">'<tt>llvm.memmove</tt>' Intrinsic</a>
3516 <div class="doc_text">
3520 declare void %llvm.memmove.i32(sbyte* <dest>, sbyte* <src>,
3521 uint <len>, uint <align>)
3522 declare void %llvm.memmove.i64(sbyte* <dest>, sbyte* <src>,
3523 ulong <len>, uint <align>)
3529 The '<tt>llvm.memmove.*</tt>' intrinsics move a block of memory from the source
3530 location to the destination location. It is similar to the
3531 '<tt>llvm.memcmp</tt>' intrinsic but allows the two memory locations to overlap.
3535 Note that, unlike the standard libc function, the <tt>llvm.memmove.*</tt>
3536 intrinsics do not return a value, and takes an extra alignment argument.
3542 The first argument is a pointer to the destination, the second is a pointer to
3543 the source. The third argument is an integer argument
3544 specifying the number of bytes to copy, and the fourth argument is the alignment
3545 of the source and destination locations.
3549 If the call to this intrinisic has an alignment value that is not 0 or 1, then
3550 the caller guarantees that the source and destination pointers are aligned to
3557 The '<tt>llvm.memmove.*</tt>' intrinsics copy a block of memory from the source
3558 location to the destination location, which may overlap. It
3559 copies "len" bytes of memory over. If the argument is known to be aligned to
3560 some boundary, this can be specified as the fourth argument, otherwise it should
3566 <!-- _______________________________________________________________________ -->
3567 <div class="doc_subsubsection">
3568 <a name="i_memset">'<tt>llvm.memset.*</tt>' Intrinsics</a>
3571 <div class="doc_text">
3575 declare void %llvm.memset.i32(sbyte* <dest>, ubyte <val>,
3576 uint <len>, uint <align>)
3577 declare void %llvm.memset.i64(sbyte* <dest>, ubyte <val>,
3578 ulong <len>, uint <align>)
3584 The '<tt>llvm.memset.*</tt>' intrinsics fill a block of memory with a particular
3589 Note that, unlike the standard libc function, the <tt>llvm.memset</tt> intrinsic
3590 does not return a value, and takes an extra alignment argument.
3596 The first argument is a pointer to the destination to fill, the second is the
3597 byte value to fill it with, the third argument is an integer
3598 argument specifying the number of bytes to fill, and the fourth argument is the
3599 known alignment of destination location.
3603 If the call to this intrinisic has an alignment value that is not 0 or 1, then
3604 the caller guarantees that the destination pointer is aligned to that boundary.
3610 The '<tt>llvm.memset.*</tt>' intrinsics fill "len" bytes of memory starting at
3612 destination location. If the argument is known to be aligned to some boundary,
3613 this can be specified as the fourth argument, otherwise it should be set to 0 or
3619 <!-- _______________________________________________________________________ -->
3620 <div class="doc_subsubsection">
3621 <a name="i_isunordered">'<tt>llvm.isunordered.*</tt>' Intrinsic</a>
3624 <div class="doc_text">
3628 declare bool %llvm.isunordered.f32(float Val1, float Val2)
3629 declare bool %llvm.isunordered.f64(double Val1, double Val2)
3635 The '<tt>llvm.isunordered</tt>' intrinsics return true if either or both of the
3636 specified floating point values is a NAN.
3642 The arguments are floating point numbers of the same type.
3648 If either or both of the arguments is a SNAN or QNAN, it returns true, otherwise
3654 <!-- _______________________________________________________________________ -->
3655 <div class="doc_subsubsection">
3656 <a name="i_sqrt">'<tt>llvm.sqrt.*</tt>' Intrinsic</a>
3659 <div class="doc_text">
3663 declare float %llvm.sqrt.f32(float %Val)
3664 declare double %llvm.sqrt.f64(double %Val)
3670 The '<tt>llvm.sqrt</tt>' intrinsics return the sqrt of the specified operand,
3671 returning the same value as the libm '<tt>sqrt</tt>' function would. Unlike
3672 <tt>sqrt</tt> in libm, however, <tt>llvm.sqrt</tt> has undefined behavior for
3673 negative numbers (which allows for better optimization).
3679 The argument and return value are floating point numbers of the same type.
3685 This function returns the sqrt of the specified operand if it is a positive
3686 floating point number.
3690 <!-- _______________________________________________________________________ -->
3691 <div class="doc_subsubsection">
3692 <a name="i_powi">'<tt>llvm.powi.*</tt>' Intrinsic</a>
3695 <div class="doc_text">
3699 declare float %llvm.powi.f32(float %Val, int %power)
3700 declare double %llvm.powi.f64(double %Val, int %power)
3706 The '<tt>llvm.powi.*</tt>' intrinsics return the first operand raised to the
3707 specified (positive or negative) power. The order of evaluation of
3708 multiplications is not defined.
3714 The second argument is an integer power, and the first is a value to raise to
3721 This function returns the first value raised to the second power with an
3722 unspecified sequence of rounding operations.</p>
3726 <!-- ======================================================================= -->
3727 <div class="doc_subsection">
3728 <a name="int_manip">Bit Manipulation Intrinsics</a>
3731 <div class="doc_text">
3733 LLVM provides intrinsics for a few important bit manipulation operations.
3734 These allow efficient code generation for some algorithms.
3739 <!-- _______________________________________________________________________ -->
3740 <div class="doc_subsubsection">
3741 <a name="i_bswap">'<tt>llvm.bswap.*</tt>' Intrinsics</a>
3744 <div class="doc_text">
3748 declare ushort %llvm.bswap.i16(ushort <id>)
3749 declare uint %llvm.bswap.i32(uint <id>)
3750 declare ulong %llvm.bswap.i64(ulong <id>)
3756 The '<tt>llvm.bwsap</tt>' family of intrinsics is used to byteswap a 16, 32 or
3757 64 bit quantity. These are useful for performing operations on data that is not
3758 in the target's native byte order.
3764 The <tt>llvm.bswap.16</tt> intrinsic returns a ushort value that has the high and low
3765 byte of the input ushort swapped. Similarly, the <tt>llvm.bswap.i32</tt> intrinsic
3766 returns a uint value that has the four bytes of the input uint swapped, so that
3767 if the input bytes are numbered 0, 1, 2, 3 then the returned uint will have its
3768 bytes in 3, 2, 1, 0 order. The <tt>llvm.bswap.i64</tt> intrinsic extends this concept
3774 <!-- _______________________________________________________________________ -->
3775 <div class="doc_subsubsection">
3776 <a name="int_ctpop">'<tt>llvm.ctpop.*</tt>' Intrinsic</a>
3779 <div class="doc_text">
3783 declare ubyte %llvm.ctpop.i8 (ubyte <src>)
3784 declare ushort %llvm.ctpop.i16(ushort <src>)
3785 declare uint %llvm.ctpop.i32(uint <src>)
3786 declare ulong %llvm.ctpop.i64(ulong <src>)
3792 The '<tt>llvm.ctpop</tt>' family of intrinsics counts the number of bits set in a
3799 The only argument is the value to be counted. The argument may be of any
3800 unsigned integer type. The return type must match the argument type.
3806 The '<tt>llvm.ctpop</tt>' intrinsic counts the 1's in a variable.
3810 <!-- _______________________________________________________________________ -->
3811 <div class="doc_subsubsection">
3812 <a name="int_ctlz">'<tt>llvm.ctlz.*</tt>' Intrinsic</a>
3815 <div class="doc_text">
3819 declare ubyte %llvm.ctlz.i8 (ubyte <src>)
3820 declare ushort %llvm.ctlz.i16(ushort <src>)
3821 declare uint %llvm.ctlz.i32(uint <src>)
3822 declare ulong %llvm.ctlz.i64(ulong <src>)
3828 The '<tt>llvm.ctlz</tt>' family of intrinsic functions counts the number of
3829 leading zeros in a variable.
3835 The only argument is the value to be counted. The argument may be of any
3836 unsigned integer type. The return type must match the argument type.
3842 The '<tt>llvm.ctlz</tt>' intrinsic counts the leading (most significant) zeros
3843 in a variable. If the src == 0 then the result is the size in bits of the type
3844 of src. For example, <tt>llvm.ctlz(int 2) = 30</tt>.
3850 <!-- _______________________________________________________________________ -->
3851 <div class="doc_subsubsection">
3852 <a name="int_cttz">'<tt>llvm.cttz.*</tt>' Intrinsic</a>
3855 <div class="doc_text">
3859 declare ubyte %llvm.cttz.i8 (ubyte <src>)
3860 declare ushort %llvm.cttz.i16(ushort <src>)
3861 declare uint %llvm.cttz.i32(uint <src>)
3862 declare ulong %llvm.cttz.i64(ulong <src>)
3868 The '<tt>llvm.cttz</tt>' family of intrinsic functions counts the number of
3875 The only argument is the value to be counted. The argument may be of any
3876 unsigned integer type. The return type must match the argument type.
3882 The '<tt>llvm.cttz</tt>' intrinsic counts the trailing (least significant) zeros
3883 in a variable. If the src == 0 then the result is the size in bits of the type
3884 of src. For example, <tt>llvm.cttz(2) = 1</tt>.
3888 <!-- ======================================================================= -->
3889 <div class="doc_subsection">
3890 <a name="int_debugger">Debugger Intrinsics</a>
3893 <div class="doc_text">
3895 The LLVM debugger intrinsics (which all start with <tt>llvm.dbg.</tt> prefix),
3896 are described in the <a
3897 href="SourceLevelDebugging.html#format_common_intrinsics">LLVM Source Level
3898 Debugging</a> document.
3903 <!-- *********************************************************************** -->
3906 <a href="http://jigsaw.w3.org/css-validator/check/referer"><img
3907 src="http://jigsaw.w3.org/css-validator/images/vcss" alt="Valid CSS!"></a>
3908 <a href="http://validator.w3.org/check/referer"><img
3909 src="http://www.w3.org/Icons/valid-html401" alt="Valid HTML 4.01!" /></a>
3911 <a href="mailto:sabre@nondot.org">Chris Lattner</a><br>
3912 <a href="http://llvm.org">The LLVM Compiler Infrastructure</a><br>
3913 Last modified: $Date$