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 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>
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>; Definition of main function</i>
354 int %main() { <i>; int()* </i>
355 <i>; Convert [13x sbyte]* to sbyte *...</i>
357 href="#i_getelementptr">getelementptr</a> [13 x sbyte]* %.LC0, long 0, long 0 <i>; sbyte*</i>
359 <i>; Call puts function to write out the string to stdout...</i>
361 href="#i_call">call</a> int %puts(sbyte* %cast210) <i>; int</i>
363 href="#i_ret">ret</a> int 0<br>}<br></pre>
365 <p>This example is made up of a <a href="#globalvars">global variable</a>
366 named "<tt>.LC0</tt>", an external declaration of the "<tt>puts</tt>"
367 function, and a <a href="#functionstructure">function definition</a>
368 for "<tt>main</tt>".</p>
370 <p>In general, a module is made up of a list of global values,
371 where both functions and global variables are global values. Global values are
372 represented by a pointer to a memory location (in this case, a pointer to an
373 array of char, and a pointer to a function), and have one of the following <a
374 href="#linkage">linkage types</a>.</p>
378 <!-- ======================================================================= -->
379 <div class="doc_subsection">
380 <a name="linkage">Linkage Types</a>
383 <div class="doc_text">
386 All Global Variables and Functions have one of the following types of linkage:
391 <dt><tt><b><a name="linkage_internal">internal</a></b></tt> </dt>
393 <dd>Global values with internal linkage are only directly accessible by
394 objects in the current module. In particular, linking code into a module with
395 an internal global value may cause the internal to be renamed as necessary to
396 avoid collisions. Because the symbol is internal to the module, all
397 references can be updated. This corresponds to the notion of the
398 '<tt>static</tt>' keyword in C, or the idea of "anonymous namespaces" in C++.
401 <dt><tt><b><a name="linkage_linkonce">linkonce</a></b></tt>: </dt>
403 <dd>"<tt>linkonce</tt>" linkage is similar to <tt>internal</tt> linkage, with
404 the twist that linking together two modules defining the same
405 <tt>linkonce</tt> globals will cause one of the globals to be discarded. This
406 is typically used to implement inline functions. Unreferenced
407 <tt>linkonce</tt> globals are allowed to be discarded.
410 <dt><tt><b><a name="linkage_weak">weak</a></b></tt>: </dt>
412 <dd>"<tt>weak</tt>" linkage is exactly the same as <tt>linkonce</tt> linkage,
413 except that unreferenced <tt>weak</tt> globals may not be discarded. This is
414 used to implement constructs in C such as "<tt>int X;</tt>" at global scope.
417 <dt><tt><b><a name="linkage_appending">appending</a></b></tt>: </dt>
419 <dd>"<tt>appending</tt>" linkage may only be applied to global variables of
420 pointer to array type. When two global variables with appending linkage are
421 linked together, the two global arrays are appended together. This is the
422 LLVM, typesafe, equivalent of having the system linker append together
423 "sections" with identical names when .o files are linked.
426 <dt><tt><b><a name="linkage_external">externally visible</a></b></tt>:</dt>
428 <dd>If none of the above identifiers are used, the global is externally
429 visible, meaning that it participates in linkage and can be used to resolve
430 external symbol references.
434 <p><a name="linkage_external">For example, since the "<tt>.LC0</tt>"
435 variable is defined to be internal, if another module defined a "<tt>.LC0</tt>"
436 variable and was linked with this one, one of the two would be renamed,
437 preventing a collision. Since "<tt>main</tt>" and "<tt>puts</tt>" are
438 external (i.e., lacking any linkage declarations), they are accessible
439 outside of the current module. It is illegal for a function <i>declaration</i>
440 to have any linkage type other than "externally visible".</a></p>
444 <!-- ======================================================================= -->
445 <div class="doc_subsection">
446 <a name="callingconv">Calling Conventions</a>
449 <div class="doc_text">
451 <p>LLVM <a href="#functionstructure">functions</a>, <a href="#i_call">calls</a>
452 and <a href="#i_invoke">invokes</a> can all have an optional calling convention
453 specified for the call. The calling convention of any pair of dynamic
454 caller/callee must match, or the behavior of the program is undefined. The
455 following calling conventions are supported by LLVM, and more may be added in
459 <dt><b>"<tt>ccc</tt>" - The C calling convention</b>:</dt>
461 <dd>This calling convention (the default if no other calling convention is
462 specified) matches the target C calling conventions. This calling convention
463 supports varargs function calls and tolerates some mismatch in the declared
464 prototype and implemented declaration of the function (as does normal C).
467 <dt><b>"<tt>csretcc</tt>" - The C struct return calling convention</b>:</dt>
469 <dd>This calling convention matches the target C calling conventions, except
470 that functions with this convention are required to take a pointer as their
471 first argument, and the return type of the function must be void. This is
472 used for C functions that return aggregates by-value. In this case, the
473 function has been transformed to take a pointer to the struct as the first
474 argument to the function. For targets where the ABI specifies specific
475 behavior for structure-return calls, the calling convention can be used to
476 distinguish between struct return functions and other functions that take a
477 pointer to a struct as the first argument.
480 <dt><b>"<tt>fastcc</tt>" - The fast calling convention</b>:</dt>
482 <dd>This calling convention attempts to make calls as fast as possible
483 (e.g. by passing things in registers). This calling convention allows the
484 target to use whatever tricks it wants to produce fast code for the target,
485 without having to conform to an externally specified ABI. Implementations of
486 this convention should allow arbitrary tail call optimization to be supported.
487 This calling convention does not support varargs and requires the prototype of
488 all callees to exactly match the prototype of the function definition.
491 <dt><b>"<tt>coldcc</tt>" - The cold calling convention</b>:</dt>
493 <dd>This calling convention attempts to make code in the caller as efficient
494 as possible under the assumption that the call is not commonly executed. As
495 such, these calls often preserve all registers so that the call does not break
496 any live ranges in the caller side. This calling convention does not support
497 varargs and requires the prototype of all callees to exactly match the
498 prototype of the function definition.
501 <dt><b>"<tt>cc <<em>n</em>></tt>" - Numbered convention</b>:</dt>
503 <dd>Any calling convention may be specified by number, allowing
504 target-specific calling conventions to be used. Target specific calling
505 conventions start at 64.
509 <p>More calling conventions can be added/defined on an as-needed basis, to
510 support pascal conventions or any other well-known target-independent
515 <!-- ======================================================================= -->
516 <div class="doc_subsection">
517 <a name="globalvars">Global Variables</a>
520 <div class="doc_text">
522 <p>Global variables define regions of memory allocated at compilation time
523 instead of run-time. Global variables may optionally be initialized, may have
524 an explicit section to be placed in, and may
525 have an optional explicit alignment specified. A
526 variable may be defined as a global "constant," which indicates that the
527 contents of the variable will <b>never</b> be modified (enabling better
528 optimization, allowing the global data to be placed in the read-only section of
529 an executable, etc). Note that variables that need runtime initialization
530 cannot be marked "constant" as there is a store to the variable.</p>
533 LLVM explicitly allows <em>declarations</em> of global variables to be marked
534 constant, even if the final definition of the global is not. This capability
535 can be used to enable slightly better optimization of the program, but requires
536 the language definition to guarantee that optimizations based on the
537 'constantness' are valid for the translation units that do not include the
541 <p>As SSA values, global variables define pointer values that are in
542 scope (i.e. they dominate) all basic blocks in the program. Global
543 variables always define a pointer to their "content" type because they
544 describe a region of memory, and all memory objects in LLVM are
545 accessed through pointers.</p>
547 <p>LLVM allows an explicit section to be specified for globals. If the target
548 supports it, it will emit globals to the section specified.</p>
550 <p>An explicit alignment may be specified for a global. If not present, or if
551 the alignment is set to zero, the alignment of the global is set by the target
552 to whatever it feels convenient. If an explicit alignment is specified, the
553 global is forced to have at least that much alignment. All alignments must be
559 <!-- ======================================================================= -->
560 <div class="doc_subsection">
561 <a name="functionstructure">Functions</a>
564 <div class="doc_text">
566 <p>LLVM function definitions consist of an optional <a href="#linkage">linkage
567 type</a>, an optional <a href="#callingconv">calling convention</a>, a return
568 type, a function name, a (possibly empty) argument list, an optional section,
569 an optional alignment, an opening curly brace,
570 a list of basic blocks, and a closing curly brace. LLVM function declarations
571 are defined with the "<tt>declare</tt>" keyword, an optional <a
572 href="#callingconv">calling convention</a>, a return type, a function name,
573 a possibly empty list of arguments, and an optional alignment.</p>
575 <p>A function definition contains a list of basic blocks, forming the CFG for
576 the function. Each basic block may optionally start with a label (giving the
577 basic block a symbol table entry), contains a list of instructions, and ends
578 with a <a href="#terminators">terminator</a> instruction (such as a branch or
579 function return).</p>
581 <p>The first basic block in a program is special in two ways: it is immediately
582 executed on entrance to the function, and it is not allowed to have predecessor
583 basic blocks (i.e. there can not be any branches to the entry block of a
584 function). Because the block can have no predecessors, it also cannot have any
585 <a href="#i_phi">PHI nodes</a>.</p>
587 <p>LLVM functions are identified by their name and type signature. Hence, two
588 functions with the same name but different parameter lists or return values are
589 considered different functions, and LLVM will resolve references to each
592 <p>LLVM allows an explicit section to be specified for functions. If the target
593 supports it, it will emit functions to the section specified.</p>
595 <p>An explicit alignment may be specified for a function. If not present, or if
596 the alignment is set to zero, the alignment of the function is set by the target
597 to whatever it feels convenient. If an explicit alignment is specified, the
598 function is forced to have at least that much alignment. All alignments must be
603 <!-- ======================================================================= -->
604 <div class="doc_subsection">
605 <a name="moduleasm">Module-Level Inline Assembly</a>
608 <div class="doc_text">
610 Modules may contain "module-level inline asm" blocks, which corresponds to the
611 GCC "file scope inline asm" blocks. These blocks are internally concatenated by
612 LLVM and treated as a single unit, but may be separated in the .ll file if
613 desired. The syntax is very simple:
616 <div class="doc_code"><pre>
617 module asm "inline asm code goes here"
618 module asm "more can go here"
621 <p>The strings can contain any character by escaping non-printable characters.
622 The escape sequence used is simply "\xx" where "xx" is the two digit hex code
627 The inline asm code is simply printed to the machine code .s file when
628 assembly code is generated.
633 <!-- *********************************************************************** -->
634 <div class="doc_section"> <a name="typesystem">Type System</a> </div>
635 <!-- *********************************************************************** -->
637 <div class="doc_text">
639 <p>The LLVM type system is one of the most important features of the
640 intermediate representation. Being typed enables a number of
641 optimizations to be performed on the IR directly, without having to do
642 extra analyses on the side before the transformation. A strong type
643 system makes it easier to read the generated code and enables novel
644 analyses and transformations that are not feasible to perform on normal
645 three address code representations.</p>
649 <!-- ======================================================================= -->
650 <div class="doc_subsection"> <a name="t_primitive">Primitive Types</a> </div>
651 <div class="doc_text">
652 <p>The primitive types are the fundamental building blocks of the LLVM
653 system. The current set of primitive types is as follows:</p>
655 <table class="layout">
660 <tr><th>Type</th><th>Description</th></tr>
661 <tr><td><tt>void</tt></td><td>No value</td></tr>
662 <tr><td><tt>ubyte</tt></td><td>Unsigned 8-bit value</td></tr>
663 <tr><td><tt>ushort</tt></td><td>Unsigned 16-bit value</td></tr>
664 <tr><td><tt>uint</tt></td><td>Unsigned 32-bit value</td></tr>
665 <tr><td><tt>ulong</tt></td><td>Unsigned 64-bit value</td></tr>
666 <tr><td><tt>float</tt></td><td>32-bit floating point value</td></tr>
667 <tr><td><tt>label</tt></td><td>Branch destination</td></tr>
674 <tr><th>Type</th><th>Description</th></tr>
675 <tr><td><tt>bool</tt></td><td>True or False value</td></tr>
676 <tr><td><tt>sbyte</tt></td><td>Signed 8-bit value</td></tr>
677 <tr><td><tt>short</tt></td><td>Signed 16-bit value</td></tr>
678 <tr><td><tt>int</tt></td><td>Signed 32-bit value</td></tr>
679 <tr><td><tt>long</tt></td><td>Signed 64-bit value</td></tr>
680 <tr><td><tt>double</tt></td><td>64-bit floating point value</td></tr>
688 <!-- _______________________________________________________________________ -->
689 <div class="doc_subsubsection"> <a name="t_classifications">Type
690 Classifications</a> </div>
691 <div class="doc_text">
692 <p>These different primitive types fall into a few useful
695 <table border="1" cellspacing="0" cellpadding="4">
697 <tr><th>Classification</th><th>Types</th></tr>
699 <td><a name="t_signed">signed</a></td>
700 <td><tt>sbyte, short, int, long, float, double</tt></td>
703 <td><a name="t_unsigned">unsigned</a></td>
704 <td><tt>ubyte, ushort, uint, ulong</tt></td>
707 <td><a name="t_integer">integer</a></td>
708 <td><tt>ubyte, sbyte, ushort, short, uint, int, ulong, long</tt></td>
711 <td><a name="t_integral">integral</a></td>
712 <td><tt>bool, ubyte, sbyte, ushort, short, uint, int, ulong, long</tt>
716 <td><a name="t_floating">floating point</a></td>
717 <td><tt>float, double</tt></td>
720 <td><a name="t_firstclass">first class</a></td>
721 <td><tt>bool, ubyte, sbyte, ushort, short, uint, int, ulong, long,<br>
722 float, double, <a href="#t_pointer">pointer</a>,
723 <a href="#t_packed">packed</a></tt></td>
728 <p>The <a href="#t_firstclass">first class</a> types are perhaps the
729 most important. Values of these types are the only ones which can be
730 produced by instructions, passed as arguments, or used as operands to
731 instructions. This means that all structures and arrays must be
732 manipulated either by pointer or by component.</p>
735 <!-- ======================================================================= -->
736 <div class="doc_subsection"> <a name="t_derived">Derived Types</a> </div>
738 <div class="doc_text">
740 <p>The real power in LLVM comes from the derived types in the system.
741 This is what allows a programmer to represent arrays, functions,
742 pointers, and other useful types. Note that these derived types may be
743 recursive: For example, it is possible to have a two dimensional array.</p>
747 <!-- _______________________________________________________________________ -->
748 <div class="doc_subsubsection"> <a name="t_array">Array Type</a> </div>
750 <div class="doc_text">
754 <p>The array type is a very simple derived type that arranges elements
755 sequentially in memory. The array type requires a size (number of
756 elements) and an underlying data type.</p>
761 [<# elements> x <elementtype>]
764 <p>The number of elements is a constant integer value; elementtype may
765 be any type with a size.</p>
768 <table class="layout">
771 <tt>[40 x int ]</tt><br/>
772 <tt>[41 x int ]</tt><br/>
773 <tt>[40 x uint]</tt><br/>
776 Array of 40 integer values.<br/>
777 Array of 41 integer values.<br/>
778 Array of 40 unsigned integer values.<br/>
782 <p>Here are some examples of multidimensional arrays:</p>
783 <table class="layout">
786 <tt>[3 x [4 x int]]</tt><br/>
787 <tt>[12 x [10 x float]]</tt><br/>
788 <tt>[2 x [3 x [4 x uint]]]</tt><br/>
791 3x4 array of integer values.<br/>
792 12x10 array of single precision floating point values.<br/>
793 2x3x4 array of unsigned integer values.<br/>
798 <p>Note that 'variable sized arrays' can be implemented in LLVM with a zero
799 length array. Normally, accesses past the end of an array are undefined in
800 LLVM (e.g. it is illegal to access the 5th element of a 3 element array).
801 As a special case, however, zero length arrays are recognized to be variable
802 length. This allows implementation of 'pascal style arrays' with the LLVM
803 type "{ int, [0 x float]}", for example.</p>
807 <!-- _______________________________________________________________________ -->
808 <div class="doc_subsubsection"> <a name="t_function">Function Type</a> </div>
809 <div class="doc_text">
811 <p>The function type can be thought of as a function signature. It
812 consists of a return type and a list of formal parameter types.
813 Function types are usually used to build virtual function tables
814 (which are structures of pointers to functions), for indirect function
815 calls, and when defining a function.</p>
817 The return type of a function type cannot be an aggregate type.
820 <pre> <returntype> (<parameter list>)<br></pre>
821 <p>...where '<tt><parameter list></tt>' is a comma-separated list of type
822 specifiers. Optionally, the parameter list may include a type <tt>...</tt>,
823 which indicates that the function takes a variable number of arguments.
824 Variable argument functions can access their arguments with the <a
825 href="#int_varargs">variable argument handling intrinsic</a> functions.</p>
827 <table class="layout">
830 <tt>int (int)</tt> <br/>
831 <tt>float (int, int *) *</tt><br/>
832 <tt>int (sbyte *, ...)</tt><br/>
835 function taking an <tt>int</tt>, returning an <tt>int</tt><br/>
836 <a href="#t_pointer">Pointer</a> to a function that takes an
837 <tt>int</tt> and a <a href="#t_pointer">pointer</a> to <tt>int</tt>,
838 returning <tt>float</tt>.<br/>
839 A vararg function that takes at least one <a href="#t_pointer">pointer</a>
840 to <tt>sbyte</tt> (signed char in C), which returns an integer. This is
841 the signature for <tt>printf</tt> in LLVM.<br/>
847 <!-- _______________________________________________________________________ -->
848 <div class="doc_subsubsection"> <a name="t_struct">Structure Type</a> </div>
849 <div class="doc_text">
851 <p>The structure type is used to represent a collection of data members
852 together in memory. The packing of the field types is defined to match
853 the ABI of the underlying processor. The elements of a structure may
854 be any type that has a size.</p>
855 <p>Structures are accessed using '<tt><a href="#i_load">load</a></tt>
856 and '<tt><a href="#i_store">store</a></tt>' by getting a pointer to a
857 field with the '<tt><a href="#i_getelementptr">getelementptr</a></tt>'
860 <pre> { <type list> }<br></pre>
862 <table class="layout">
865 <tt>{ int, int, int }</tt><br/>
866 <tt>{ float, int (int) * }</tt><br/>
869 a triple of three <tt>int</tt> values<br/>
870 A pair, where the first element is a <tt>float</tt> and the second element
871 is a <a href="#t_pointer">pointer</a> to a <a href="#t_function">function</a>
872 that takes an <tt>int</tt>, returning an <tt>int</tt>.<br/>
878 <!-- _______________________________________________________________________ -->
879 <div class="doc_subsubsection"> <a name="t_pointer">Pointer Type</a> </div>
880 <div class="doc_text">
882 <p>As in many languages, the pointer type represents a pointer or
883 reference to another object, which must live in memory.</p>
885 <pre> <type> *<br></pre>
887 <table class="layout">
890 <tt>[4x int]*</tt><br/>
891 <tt>int (int *) *</tt><br/>
894 A <a href="#t_pointer">pointer</a> to <a href="#t_array">array</a> of
895 four <tt>int</tt> values<br/>
896 A <a href="#t_pointer">pointer</a> to a <a
897 href="#t_function">function</a> that takes an <tt>int*</tt>, returning an
904 <!-- _______________________________________________________________________ -->
905 <div class="doc_subsubsection"> <a name="t_packed">Packed Type</a> </div>
906 <div class="doc_text">
910 <p>A packed type is a simple derived type that represents a vector
911 of elements. Packed types are used when multiple primitive data
912 are operated in parallel using a single instruction (SIMD).
913 A packed type requires a size (number of
914 elements) and an underlying primitive data type. Vectors must have a power
915 of two length (1, 2, 4, 8, 16 ...). Packed types are
916 considered <a href="#t_firstclass">first class</a>.</p>
921 < <# elements> x <elementtype> >
924 <p>The number of elements is a constant integer value; elementtype may
925 be any integral or floating point type.</p>
929 <table class="layout">
932 <tt><4 x int></tt><br/>
933 <tt><8 x float></tt><br/>
934 <tt><2 x uint></tt><br/>
937 Packed vector of 4 integer values.<br/>
938 Packed vector of 8 floating-point values.<br/>
939 Packed vector of 2 unsigned integer values.<br/>
945 <!-- _______________________________________________________________________ -->
946 <div class="doc_subsubsection"> <a name="t_opaque">Opaque Type</a> </div>
947 <div class="doc_text">
951 <p>Opaque types are used to represent unknown types in the system. This
952 corresponds (for example) to the C notion of a foward declared structure type.
953 In LLVM, opaque types can eventually be resolved to any type (not just a
964 <table class="layout">
977 <!-- *********************************************************************** -->
978 <div class="doc_section"> <a name="constants">Constants</a> </div>
979 <!-- *********************************************************************** -->
981 <div class="doc_text">
983 <p>LLVM has several different basic types of constants. This section describes
984 them all and their syntax.</p>
988 <!-- ======================================================================= -->
989 <div class="doc_subsection"><a name="simpleconstants">Simple Constants</a></div>
991 <div class="doc_text">
994 <dt><b>Boolean constants</b></dt>
996 <dd>The two strings '<tt>true</tt>' and '<tt>false</tt>' are both valid
997 constants of the <tt><a href="#t_primitive">bool</a></tt> type.
1000 <dt><b>Integer constants</b></dt>
1002 <dd>Standard integers (such as '4') are constants of the <a
1003 href="#t_integer">integer</a> type. Negative numbers may be used with signed
1007 <dt><b>Floating point constants</b></dt>
1009 <dd>Floating point constants use standard decimal notation (e.g. 123.421),
1010 exponential notation (e.g. 1.23421e+2), or a more precise hexadecimal
1011 notation (see below). Floating point constants must have a <a
1012 href="#t_floating">floating point</a> type. </dd>
1014 <dt><b>Null pointer constants</b></dt>
1016 <dd>The identifier '<tt>null</tt>' is recognized as a null pointer constant
1017 and must be of <a href="#t_pointer">pointer type</a>.</dd>
1021 <p>The one non-intuitive notation for constants is the optional hexadecimal form
1022 of floating point constants. For example, the form '<tt>double
1023 0x432ff973cafa8000</tt>' is equivalent to (but harder to read than) '<tt>double
1024 4.5e+15</tt>'. The only time hexadecimal floating point constants are required
1025 (and the only time that they are generated by the disassembler) is when a
1026 floating point constant must be emitted but it cannot be represented as a
1027 decimal floating point number. For example, NaN's, infinities, and other
1028 special values are represented in their IEEE hexadecimal format so that
1029 assembly and disassembly do not cause any bits to change in the constants.</p>
1033 <!-- ======================================================================= -->
1034 <div class="doc_subsection"><a name="aggregateconstants">Aggregate Constants</a>
1037 <div class="doc_text">
1038 <p>Aggregate constants arise from aggregation of simple constants
1039 and smaller aggregate constants.</p>
1042 <dt><b>Structure constants</b></dt>
1044 <dd>Structure constants are represented with notation similar to structure
1045 type definitions (a comma separated list of elements, surrounded by braces
1046 (<tt>{}</tt>)). For example: "<tt>{ int 4, float 17.0, int* %G }</tt>",
1047 where "<tt>%G</tt>" is declared as "<tt>%G = external global int</tt>". Structure constants
1048 must have <a href="#t_struct">structure type</a>, and the number and
1049 types of elements must match those specified by the type.
1052 <dt><b>Array constants</b></dt>
1054 <dd>Array constants are represented with notation similar to array type
1055 definitions (a comma separated list of elements, surrounded by square brackets
1056 (<tt>[]</tt>)). For example: "<tt>[ int 42, int 11, int 74 ]</tt>". Array
1057 constants must have <a href="#t_array">array type</a>, and the number and
1058 types of elements must match those specified by the type.
1061 <dt><b>Packed constants</b></dt>
1063 <dd>Packed constants are represented with notation similar to packed type
1064 definitions (a comma separated list of elements, surrounded by
1065 less-than/greater-than's (<tt><></tt>)). For example: "<tt>< int 42,
1066 int 11, int 74, int 100 ></tt>". Packed constants must have <a
1067 href="#t_packed">packed type</a>, and the number and types of elements must
1068 match those specified by the type.
1071 <dt><b>Zero initialization</b></dt>
1073 <dd>The string '<tt>zeroinitializer</tt>' can be used to zero initialize a
1074 value to zero of <em>any</em> type, including scalar and aggregate types.
1075 This is often used to avoid having to print large zero initializers (e.g. for
1076 large arrays) and is always exactly equivalent to using explicit zero
1083 <!-- ======================================================================= -->
1084 <div class="doc_subsection">
1085 <a name="globalconstants">Global Variable and Function Addresses</a>
1088 <div class="doc_text">
1090 <p>The addresses of <a href="#globalvars">global variables</a> and <a
1091 href="#functionstructure">functions</a> are always implicitly valid (link-time)
1092 constants. These constants are explicitly referenced when the <a
1093 href="#identifiers">identifier for the global</a> is used and always have <a
1094 href="#t_pointer">pointer</a> type. For example, the following is a legal LLVM
1100 %Z = global [2 x int*] [ int* %X, int* %Y ]
1105 <!-- ======================================================================= -->
1106 <div class="doc_subsection"><a name="undefvalues">Undefined Values</a></div>
1107 <div class="doc_text">
1108 <p>The string '<tt>undef</tt>' is recognized as a type-less constant that has
1109 no specific value. Undefined values may be of any type and be used anywhere
1110 a constant is permitted.</p>
1112 <p>Undefined values indicate to the compiler that the program is well defined
1113 no matter what value is used, giving the compiler more freedom to optimize.
1117 <!-- ======================================================================= -->
1118 <div class="doc_subsection"><a name="constantexprs">Constant Expressions</a>
1121 <div class="doc_text">
1123 <p>Constant expressions are used to allow expressions involving other constants
1124 to be used as constants. Constant expressions may be of any <a
1125 href="#t_firstclass">first class</a> type and may involve any LLVM operation
1126 that does not have side effects (e.g. load and call are not supported). The
1127 following is the syntax for constant expressions:</p>
1130 <dt><b><tt>cast ( CST to TYPE )</tt></b></dt>
1132 <dd>Cast a constant to another type.</dd>
1134 <dt><b><tt>getelementptr ( CSTPTR, IDX0, IDX1, ... )</tt></b></dt>
1136 <dd>Perform the <a href="#i_getelementptr">getelementptr operation</a> on
1137 constants. As with the <a href="#i_getelementptr">getelementptr</a>
1138 instruction, the index list may have zero or more indexes, which are required
1139 to make sense for the type of "CSTPTR".</dd>
1141 <dt><b><tt>select ( COND, VAL1, VAL2 )</tt></b></dt>
1143 <dd>Perform the <a href="#i_select">select operation</a> on
1146 <dt><b><tt>extractelement ( VAL, IDX )</tt></b></dt>
1148 <dd>Perform the <a href="#i_extractelement">extractelement
1149 operation</a> on constants.
1151 <dt><b><tt>insertelement ( VAL, ELT, IDX )</tt></b></dt>
1153 <dd>Perform the <a href="#i_insertelement">insertelement
1154 operation</a> on constants.
1157 <dt><b><tt>shufflevector ( VEC1, VEC2, IDXMASK )</tt></b></dt>
1159 <dd>Perform the <a href="#i_shufflevector">shufflevector
1160 operation</a> on constants.
1162 <dt><b><tt>OPCODE ( LHS, RHS )</tt></b></dt>
1164 <dd>Perform the specified operation of the LHS and RHS constants. OPCODE may
1165 be any of the <a href="#binaryops">binary</a> or <a href="#bitwiseops">bitwise
1166 binary</a> operations. The constraints on operands are the same as those for
1167 the corresponding instruction (e.g. no bitwise operations on floating point
1168 values are allowed).</dd>
1172 <!-- *********************************************************************** -->
1173 <div class="doc_section"> <a name="othervalues">Other Values</a> </div>
1174 <!-- *********************************************************************** -->
1176 <!-- ======================================================================= -->
1177 <div class="doc_subsection">
1178 <a name="inlineasm">Inline Assembler Expressions</a>
1181 <div class="doc_text">
1184 LLVM supports inline assembler expressions (as opposed to <a href="#moduleasm">
1185 Module-Level Inline Assembly</a>) through the use of a special value. This
1186 value represents the inline assembler as a string (containing the instructions
1187 to emit), a list of operand constraints (stored as a string), and a flag that
1188 indicates whether or not the inline asm expression has side effects. An example
1189 inline assembler expression is:
1193 int(int) asm "bswap $0", "=r,r"
1197 Inline assembler expressions may <b>only</b> be used as the callee operand of
1198 a <a href="#i_call"><tt>call</tt> instruction</a>. Thus, typically we have:
1202 %X = call int asm "<a href="#i_bswap">bswap</a> $0", "=r,r"(int %Y)
1206 Inline asms with side effects not visible in the constraint list must be marked
1207 as having side effects. This is done through the use of the
1208 '<tt>sideeffect</tt>' keyword, like so:
1212 call void asm sideeffect "eieio", ""()
1215 <p>TODO: The format of the asm and constraints string still need to be
1216 documented here. Constraints on what can be done (e.g. duplication, moving, etc
1217 need to be documented).
1222 <!-- *********************************************************************** -->
1223 <div class="doc_section"> <a name="instref">Instruction Reference</a> </div>
1224 <!-- *********************************************************************** -->
1226 <div class="doc_text">
1228 <p>The LLVM instruction set consists of several different
1229 classifications of instructions: <a href="#terminators">terminator
1230 instructions</a>, <a href="#binaryops">binary instructions</a>,
1231 <a href="#bitwiseops">bitwise binary instructions</a>, <a
1232 href="#memoryops">memory instructions</a>, and <a href="#otherops">other
1233 instructions</a>.</p>
1237 <!-- ======================================================================= -->
1238 <div class="doc_subsection"> <a name="terminators">Terminator
1239 Instructions</a> </div>
1241 <div class="doc_text">
1243 <p>As mentioned <a href="#functionstructure">previously</a>, every
1244 basic block in a program ends with a "Terminator" instruction, which
1245 indicates which block should be executed after the current block is
1246 finished. These terminator instructions typically yield a '<tt>void</tt>'
1247 value: they produce control flow, not values (the one exception being
1248 the '<a href="#i_invoke"><tt>invoke</tt></a>' instruction).</p>
1249 <p>There are six different terminator instructions: the '<a
1250 href="#i_ret"><tt>ret</tt></a>' instruction, the '<a href="#i_br"><tt>br</tt></a>'
1251 instruction, the '<a href="#i_switch"><tt>switch</tt></a>' instruction,
1252 the '<a href="#i_invoke"><tt>invoke</tt></a>' instruction, the '<a
1253 href="#i_unwind"><tt>unwind</tt></a>' instruction, and the '<a
1254 href="#i_unreachable"><tt>unreachable</tt></a>' instruction.</p>
1258 <!-- _______________________________________________________________________ -->
1259 <div class="doc_subsubsection"> <a name="i_ret">'<tt>ret</tt>'
1260 Instruction</a> </div>
1261 <div class="doc_text">
1263 <pre> ret <type> <value> <i>; Return a value from a non-void function</i>
1264 ret void <i>; Return from void function</i>
1267 <p>The '<tt>ret</tt>' instruction is used to return control flow (and a
1268 value) from a function back to the caller.</p>
1269 <p>There are two forms of the '<tt>ret</tt>' instruction: one that
1270 returns a value and then causes control flow, and one that just causes
1271 control flow to occur.</p>
1273 <p>The '<tt>ret</tt>' instruction may return any '<a
1274 href="#t_firstclass">first class</a>' type. Notice that a function is
1275 not <a href="#wellformed">well formed</a> if there exists a '<tt>ret</tt>'
1276 instruction inside of the function that returns a value that does not
1277 match the return type of the function.</p>
1279 <p>When the '<tt>ret</tt>' instruction is executed, control flow
1280 returns back to the calling function's context. If the caller is a "<a
1281 href="#i_call"><tt>call</tt></a>" instruction, execution continues at
1282 the instruction after the call. If the caller was an "<a
1283 href="#i_invoke"><tt>invoke</tt></a>" instruction, execution continues
1284 at the beginning of the "normal" destination block. If the instruction
1285 returns a value, that value shall set the call or invoke instruction's
1288 <pre> ret int 5 <i>; Return an integer value of 5</i>
1289 ret void <i>; Return from a void function</i>
1292 <!-- _______________________________________________________________________ -->
1293 <div class="doc_subsubsection"> <a name="i_br">'<tt>br</tt>' Instruction</a> </div>
1294 <div class="doc_text">
1296 <pre> br bool <cond>, label <iftrue>, label <iffalse><br> br label <dest> <i>; Unconditional branch</i>
1299 <p>The '<tt>br</tt>' instruction is used to cause control flow to
1300 transfer to a different basic block in the current function. There are
1301 two forms of this instruction, corresponding to a conditional branch
1302 and an unconditional branch.</p>
1304 <p>The conditional branch form of the '<tt>br</tt>' instruction takes a
1305 single '<tt>bool</tt>' value and two '<tt>label</tt>' values. The
1306 unconditional form of the '<tt>br</tt>' instruction takes a single '<tt>label</tt>'
1307 value as a target.</p>
1309 <p>Upon execution of a conditional '<tt>br</tt>' instruction, the '<tt>bool</tt>'
1310 argument is evaluated. If the value is <tt>true</tt>, control flows
1311 to the '<tt>iftrue</tt>' <tt>label</tt> argument. If "cond" is <tt>false</tt>,
1312 control flows to the '<tt>iffalse</tt>' <tt>label</tt> argument.</p>
1314 <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
1315 href="#i_ret">ret</a> int 1<br>IfUnequal:<br> <a href="#i_ret">ret</a> int 0<br></pre>
1317 <!-- _______________________________________________________________________ -->
1318 <div class="doc_subsubsection">
1319 <a name="i_switch">'<tt>switch</tt>' Instruction</a>
1322 <div class="doc_text">
1326 switch <intty> <value>, label <defaultdest> [ <intty> <val>, label <dest> ... ]
1331 <p>The '<tt>switch</tt>' instruction is used to transfer control flow to one of
1332 several different places. It is a generalization of the '<tt>br</tt>'
1333 instruction, allowing a branch to occur to one of many possible
1339 <p>The '<tt>switch</tt>' instruction uses three parameters: an integer
1340 comparison value '<tt>value</tt>', a default '<tt>label</tt>' destination, and
1341 an array of pairs of comparison value constants and '<tt>label</tt>'s. The
1342 table is not allowed to contain duplicate constant entries.</p>
1346 <p>The <tt>switch</tt> instruction specifies a table of values and
1347 destinations. When the '<tt>switch</tt>' instruction is executed, this
1348 table is searched for the given value. If the value is found, control flow is
1349 transfered to the corresponding destination; otherwise, control flow is
1350 transfered to the default destination.</p>
1352 <h5>Implementation:</h5>
1354 <p>Depending on properties of the target machine and the particular
1355 <tt>switch</tt> instruction, this instruction may be code generated in different
1356 ways. For example, it could be generated as a series of chained conditional
1357 branches or with a lookup table.</p>
1362 <i>; Emulate a conditional br instruction</i>
1363 %Val = <a href="#i_cast">cast</a> bool %value to int
1364 switch int %Val, label %truedest [int 0, label %falsedest ]
1366 <i>; Emulate an unconditional br instruction</i>
1367 switch uint 0, label %dest [ ]
1369 <i>; Implement a jump table:</i>
1370 switch uint %val, label %otherwise [ uint 0, label %onzero
1371 uint 1, label %onone
1372 uint 2, label %ontwo ]
1376 <!-- _______________________________________________________________________ -->
1377 <div class="doc_subsubsection">
1378 <a name="i_invoke">'<tt>invoke</tt>' Instruction</a>
1381 <div class="doc_text">
1386 <result> = invoke [<a href="#callingconv">cconv</a>] <ptr to function ty> %<function ptr val>(<function args>)
1387 to label <normal label> unwind label <exception label>
1392 <p>The '<tt>invoke</tt>' instruction causes control to transfer to a specified
1393 function, with the possibility of control flow transfer to either the
1394 '<tt>normal</tt>' label or the
1395 '<tt>exception</tt>' label. If the callee function returns with the
1396 "<tt><a href="#i_ret">ret</a></tt>" instruction, control flow will return to the
1397 "normal" label. If the callee (or any indirect callees) returns with the "<a
1398 href="#i_unwind"><tt>unwind</tt></a>" instruction, control is interrupted and
1399 continued at the dynamically nearest "exception" label.</p>
1403 <p>This instruction requires several arguments:</p>
1407 The optional "cconv" marker indicates which <a href="callingconv">calling
1408 convention</a> the call should use. If none is specified, the call defaults
1409 to using C calling conventions.
1411 <li>'<tt>ptr to function ty</tt>': shall be the signature of the pointer to
1412 function value being invoked. In most cases, this is a direct function
1413 invocation, but indirect <tt>invoke</tt>s are just as possible, branching off
1414 an arbitrary pointer to function value.
1417 <li>'<tt>function ptr val</tt>': An LLVM value containing a pointer to a
1418 function to be invoked. </li>
1420 <li>'<tt>function args</tt>': argument list whose types match the function
1421 signature argument types. If the function signature indicates the function
1422 accepts a variable number of arguments, the extra arguments can be
1425 <li>'<tt>normal label</tt>': the label reached when the called function
1426 executes a '<tt><a href="#i_ret">ret</a></tt>' instruction. </li>
1428 <li>'<tt>exception label</tt>': the label reached when a callee returns with
1429 the <a href="#i_unwind"><tt>unwind</tt></a> instruction. </li>
1435 <p>This instruction is designed to operate as a standard '<tt><a
1436 href="#i_call">call</a></tt>' instruction in most regards. The primary
1437 difference is that it establishes an association with a label, which is used by
1438 the runtime library to unwind the stack.</p>
1440 <p>This instruction is used in languages with destructors to ensure that proper
1441 cleanup is performed in the case of either a <tt>longjmp</tt> or a thrown
1442 exception. Additionally, this is important for implementation of
1443 '<tt>catch</tt>' clauses in high-level languages that support them.</p>
1447 %retval = invoke int %Test(int 15) to label %Continue
1448 unwind label %TestCleanup <i>; {int}:retval set</i>
1449 %retval = invoke <a href="#callingconv">coldcc</a> int %Test(int 15) to label %Continue
1450 unwind label %TestCleanup <i>; {int}:retval set</i>
1455 <!-- _______________________________________________________________________ -->
1457 <div class="doc_subsubsection"> <a name="i_unwind">'<tt>unwind</tt>'
1458 Instruction</a> </div>
1460 <div class="doc_text">
1469 <p>The '<tt>unwind</tt>' instruction unwinds the stack, continuing control flow
1470 at the first callee in the dynamic call stack which used an <a
1471 href="#i_invoke"><tt>invoke</tt></a> instruction to perform the call. This is
1472 primarily used to implement exception handling.</p>
1476 <p>The '<tt>unwind</tt>' intrinsic causes execution of the current function to
1477 immediately halt. The dynamic call stack is then searched for the first <a
1478 href="#i_invoke"><tt>invoke</tt></a> instruction on the call stack. Once found,
1479 execution continues at the "exceptional" destination block specified by the
1480 <tt>invoke</tt> instruction. If there is no <tt>invoke</tt> instruction in the
1481 dynamic call chain, undefined behavior results.</p>
1484 <!-- _______________________________________________________________________ -->
1486 <div class="doc_subsubsection"> <a name="i_unreachable">'<tt>unreachable</tt>'
1487 Instruction</a> </div>
1489 <div class="doc_text">
1498 <p>The '<tt>unreachable</tt>' instruction has no defined semantics. This
1499 instruction is used to inform the optimizer that a particular portion of the
1500 code is not reachable. This can be used to indicate that the code after a
1501 no-return function cannot be reached, and other facts.</p>
1505 <p>The '<tt>unreachable</tt>' instruction has no defined semantics.</p>
1510 <!-- ======================================================================= -->
1511 <div class="doc_subsection"> <a name="binaryops">Binary Operations</a> </div>
1512 <div class="doc_text">
1513 <p>Binary operators are used to do most of the computation in a
1514 program. They require two operands, execute an operation on them, and
1515 produce a single value. The operands might represent
1516 multiple data, as is the case with the <a href="#t_packed">packed</a> data type.
1517 The result value of a binary operator is not
1518 necessarily the same type as its operands.</p>
1519 <p>There are several different binary operators:</p>
1521 <!-- _______________________________________________________________________ -->
1522 <div class="doc_subsubsection"> <a name="i_add">'<tt>add</tt>'
1523 Instruction</a> </div>
1524 <div class="doc_text">
1526 <pre> <result> = add <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1529 <p>The '<tt>add</tt>' instruction returns the sum of its two operands.</p>
1531 <p>The two arguments to the '<tt>add</tt>' instruction must be either <a
1532 href="#t_integer">integer</a> or <a href="#t_floating">floating point</a> values.
1533 This instruction can also take <a href="#t_packed">packed</a> versions of the values.
1534 Both arguments must have identical types.</p>
1536 <p>The value produced is the integer or floating point sum of the two
1539 <pre> <result> = add int 4, %var <i>; yields {int}:result = 4 + %var</i>
1542 <!-- _______________________________________________________________________ -->
1543 <div class="doc_subsubsection"> <a name="i_sub">'<tt>sub</tt>'
1544 Instruction</a> </div>
1545 <div class="doc_text">
1547 <pre> <result> = sub <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1550 <p>The '<tt>sub</tt>' instruction returns the difference of its two
1552 <p>Note that the '<tt>sub</tt>' instruction is used to represent the '<tt>neg</tt>'
1553 instruction present in most other intermediate representations.</p>
1555 <p>The two arguments to the '<tt>sub</tt>' instruction must be either <a
1556 href="#t_integer">integer</a> or <a href="#t_floating">floating point</a>
1558 This instruction can also take <a href="#t_packed">packed</a> versions of the values.
1559 Both arguments must have identical types.</p>
1561 <p>The value produced is the integer or floating point difference of
1562 the two operands.</p>
1564 <pre> <result> = sub int 4, %var <i>; yields {int}:result = 4 - %var</i>
1565 <result> = sub int 0, %val <i>; yields {int}:result = -%var</i>
1568 <!-- _______________________________________________________________________ -->
1569 <div class="doc_subsubsection"> <a name="i_mul">'<tt>mul</tt>'
1570 Instruction</a> </div>
1571 <div class="doc_text">
1573 <pre> <result> = mul <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1576 <p>The '<tt>mul</tt>' instruction returns the product of its two
1579 <p>The two arguments to the '<tt>mul</tt>' instruction must be either <a
1580 href="#t_integer">integer</a> or <a href="#t_floating">floating point</a>
1582 This instruction can also take <a href="#t_packed">packed</a> versions of the values.
1583 Both arguments must have identical types.</p>
1585 <p>The value produced is the integer or floating point product of the
1587 <p>There is no signed vs unsigned multiplication. The appropriate
1588 action is taken based on the type of the operand.</p>
1590 <pre> <result> = mul int 4, %var <i>; yields {int}:result = 4 * %var</i>
1593 <!-- _______________________________________________________________________ -->
1594 <div class="doc_subsubsection"> <a name="i_div">'<tt>div</tt>'
1595 Instruction</a> </div>
1596 <div class="doc_text">
1598 <pre> <result> = div <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1601 <p>The '<tt>div</tt>' instruction returns the quotient of its two
1604 <p>The two arguments to the '<tt>div</tt>' instruction must be either <a
1605 href="#t_integer">integer</a> or <a href="#t_floating">floating point</a>
1607 This instruction can also take <a href="#t_packed">packed</a> versions of the values.
1608 Both arguments must have identical types.</p>
1610 <p>The value produced is the integer or floating point quotient of the
1613 <pre> <result> = div int 4, %var <i>; yields {int}:result = 4 / %var</i>
1616 <!-- _______________________________________________________________________ -->
1617 <div class="doc_subsubsection"> <a name="i_rem">'<tt>rem</tt>'
1618 Instruction</a> </div>
1619 <div class="doc_text">
1621 <pre> <result> = rem <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1624 <p>The '<tt>rem</tt>' instruction returns the remainder from the
1625 division of its two operands.</p>
1627 <p>The two arguments to the '<tt>rem</tt>' instruction must be either <a
1628 href="#t_integer">integer</a> or <a href="#t_floating">floating point</a>
1630 This instruction can also take <a href="#t_packed">packed</a> versions of the values.
1631 Both arguments must have identical types.</p>
1633 <p>This returns the <i>remainder</i> of a division (where the result
1634 has the same sign as the divisor), not the <i>modulus</i> (where the
1635 result has the same sign as the dividend) of a value. For more
1636 information about the difference, see <a
1637 href="http://mathforum.org/dr.math/problems/anne.4.28.99.html">The
1640 <pre> <result> = rem int 4, %var <i>; yields {int}:result = 4 % %var</i>
1644 <!-- _______________________________________________________________________ -->
1645 <div class="doc_subsubsection"> <a name="i_setcc">'<tt>set<i>cc</i></tt>'
1646 Instructions</a> </div>
1647 <div class="doc_text">
1649 <pre> <result> = seteq <ty> <var1>, <var2> <i>; yields {bool}:result</i>
1650 <result> = setne <ty> <var1>, <var2> <i>; yields {bool}:result</i>
1651 <result> = setlt <ty> <var1>, <var2> <i>; yields {bool}:result</i>
1652 <result> = setgt <ty> <var1>, <var2> <i>; yields {bool}:result</i>
1653 <result> = setle <ty> <var1>, <var2> <i>; yields {bool}:result</i>
1654 <result> = setge <ty> <var1>, <var2> <i>; yields {bool}:result</i>
1657 <p>The '<tt>set<i>cc</i></tt>' family of instructions returns a boolean
1658 value based on a comparison of their two operands.</p>
1660 <p>The two arguments to the '<tt>set<i>cc</i></tt>' instructions must
1661 be of <a href="#t_firstclass">first class</a> type (it is not possible
1662 to compare '<tt>label</tt>'s, '<tt>array</tt>'s, '<tt>structure</tt>'
1663 or '<tt>void</tt>' values, etc...). Both arguments must have identical
1666 <p>The '<tt>seteq</tt>' instruction yields a <tt>true</tt> '<tt>bool</tt>'
1667 value if both operands are equal.<br>
1668 The '<tt>setne</tt>' instruction yields a <tt>true</tt> '<tt>bool</tt>'
1669 value if both operands are unequal.<br>
1670 The '<tt>setlt</tt>' instruction yields a <tt>true</tt> '<tt>bool</tt>'
1671 value if the first operand is less than the second operand.<br>
1672 The '<tt>setgt</tt>' instruction yields a <tt>true</tt> '<tt>bool</tt>'
1673 value if the first operand is greater than the second operand.<br>
1674 The '<tt>setle</tt>' instruction yields a <tt>true</tt> '<tt>bool</tt>'
1675 value if the first operand is less than or equal to the second operand.<br>
1676 The '<tt>setge</tt>' instruction yields a <tt>true</tt> '<tt>bool</tt>'
1677 value if the first operand is greater than or equal to the second
1680 <pre> <result> = seteq int 4, 5 <i>; yields {bool}:result = false</i>
1681 <result> = setne float 4, 5 <i>; yields {bool}:result = true</i>
1682 <result> = setlt uint 4, 5 <i>; yields {bool}:result = true</i>
1683 <result> = setgt sbyte 4, 5 <i>; yields {bool}:result = false</i>
1684 <result> = setle sbyte 4, 5 <i>; yields {bool}:result = true</i>
1685 <result> = setge sbyte 4, 5 <i>; yields {bool}:result = false</i>
1689 <!-- ======================================================================= -->
1690 <div class="doc_subsection"> <a name="bitwiseops">Bitwise Binary
1691 Operations</a> </div>
1692 <div class="doc_text">
1693 <p>Bitwise binary operators are used to do various forms of
1694 bit-twiddling in a program. They are generally very efficient
1695 instructions and can commonly be strength reduced from other
1696 instructions. They require two operands, execute an operation on them,
1697 and produce a single value. The resulting value of the bitwise binary
1698 operators is always the same type as its first operand.</p>
1700 <!-- _______________________________________________________________________ -->
1701 <div class="doc_subsubsection"> <a name="i_and">'<tt>and</tt>'
1702 Instruction</a> </div>
1703 <div class="doc_text">
1705 <pre> <result> = and <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1708 <p>The '<tt>and</tt>' instruction returns the bitwise logical and of
1709 its two operands.</p>
1711 <p>The two arguments to the '<tt>and</tt>' instruction must be <a
1712 href="#t_integral">integral</a> values. Both arguments must have
1713 identical types.</p>
1715 <p>The truth table used for the '<tt>and</tt>' instruction is:</p>
1717 <div style="align: center">
1718 <table border="1" cellspacing="0" cellpadding="4">
1749 <pre> <result> = and int 4, %var <i>; yields {int}:result = 4 & %var</i>
1750 <result> = and int 15, 40 <i>; yields {int}:result = 8</i>
1751 <result> = and int 4, 8 <i>; yields {int}:result = 0</i>
1754 <!-- _______________________________________________________________________ -->
1755 <div class="doc_subsubsection"> <a name="i_or">'<tt>or</tt>' Instruction</a> </div>
1756 <div class="doc_text">
1758 <pre> <result> = or <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1761 <p>The '<tt>or</tt>' instruction returns the bitwise logical inclusive
1762 or of its two operands.</p>
1764 <p>The two arguments to the '<tt>or</tt>' instruction must be <a
1765 href="#t_integral">integral</a> values. Both arguments must have
1766 identical types.</p>
1768 <p>The truth table used for the '<tt>or</tt>' instruction is:</p>
1770 <div style="align: center">
1771 <table border="1" cellspacing="0" cellpadding="4">
1802 <pre> <result> = or int 4, %var <i>; yields {int}:result = 4 | %var</i>
1803 <result> = or int 15, 40 <i>; yields {int}:result = 47</i>
1804 <result> = or int 4, 8 <i>; yields {int}:result = 12</i>
1807 <!-- _______________________________________________________________________ -->
1808 <div class="doc_subsubsection"> <a name="i_xor">'<tt>xor</tt>'
1809 Instruction</a> </div>
1810 <div class="doc_text">
1812 <pre> <result> = xor <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1815 <p>The '<tt>xor</tt>' instruction returns the bitwise logical exclusive
1816 or of its two operands. The <tt>xor</tt> is used to implement the
1817 "one's complement" operation, which is the "~" operator in C.</p>
1819 <p>The two arguments to the '<tt>xor</tt>' instruction must be <a
1820 href="#t_integral">integral</a> values. Both arguments must have
1821 identical types.</p>
1823 <p>The truth table used for the '<tt>xor</tt>' instruction is:</p>
1825 <div style="align: center">
1826 <table border="1" cellspacing="0" cellpadding="4">
1858 <pre> <result> = xor int 4, %var <i>; yields {int}:result = 4 ^ %var</i>
1859 <result> = xor int 15, 40 <i>; yields {int}:result = 39</i>
1860 <result> = xor int 4, 8 <i>; yields {int}:result = 12</i>
1861 <result> = xor int %V, -1 <i>; yields {int}:result = ~%V</i>
1864 <!-- _______________________________________________________________________ -->
1865 <div class="doc_subsubsection"> <a name="i_shl">'<tt>shl</tt>'
1866 Instruction</a> </div>
1867 <div class="doc_text">
1869 <pre> <result> = shl <ty> <var1>, ubyte <var2> <i>; yields {ty}:result</i>
1872 <p>The '<tt>shl</tt>' instruction returns the first operand shifted to
1873 the left a specified number of bits.</p>
1875 <p>The first argument to the '<tt>shl</tt>' instruction must be an <a
1876 href="#t_integer">integer</a> type. The second argument must be an '<tt>ubyte</tt>'
1879 <p>The value produced is <tt>var1</tt> * 2<sup><tt>var2</tt></sup>.</p>
1881 <pre> <result> = shl int 4, ubyte %var <i>; yields {int}:result = 4 << %var</i>
1882 <result> = shl int 4, ubyte 2 <i>; yields {int}:result = 16</i>
1883 <result> = shl int 1, ubyte 10 <i>; yields {int}:result = 1024</i>
1886 <!-- _______________________________________________________________________ -->
1887 <div class="doc_subsubsection"> <a name="i_shr">'<tt>shr</tt>'
1888 Instruction</a> </div>
1889 <div class="doc_text">
1891 <pre> <result> = shr <ty> <var1>, ubyte <var2> <i>; yields {ty}:result</i>
1894 <p>The '<tt>shr</tt>' instruction returns the first operand shifted to
1895 the right a specified number of bits.</p>
1897 <p>The first argument to the '<tt>shr</tt>' instruction must be an <a
1898 href="#t_integer">integer</a> type. The second argument must be an '<tt>ubyte</tt>'
1901 <p>If the first argument is a <a href="#t_signed">signed</a> type, the
1902 most significant bit is duplicated in the newly free'd bit positions.
1903 If the first argument is unsigned, zero bits shall fill the empty
1906 <pre> <result> = shr int 4, ubyte %var <i>; yields {int}:result = 4 >> %var</i>
1907 <result> = shr uint 4, ubyte 1 <i>; yields {uint}:result = 2</i>
1908 <result> = shr int 4, ubyte 2 <i>; yields {int}:result = 1</i>
1909 <result> = shr sbyte 4, ubyte 3 <i>; yields {sbyte}:result = 0</i>
1910 <result> = shr sbyte -2, ubyte 1 <i>; yields {sbyte}:result = -1</i>
1914 <!-- ======================================================================= -->
1915 <div class="doc_subsection">
1916 <a name="vectorops">Vector Operations</a>
1919 <div class="doc_text">
1921 <p>LLVM supports several instructions to represent vector operations in a
1922 target-independent manner. This instructions cover the element-access and
1923 vector-specific operations needed to process vectors effectively. While LLVM
1924 does directly support these vector operations, many sophisticated algorithms
1925 will want to use target-specific intrinsics to take full advantage of a specific
1930 <!-- _______________________________________________________________________ -->
1931 <div class="doc_subsubsection">
1932 <a name="i_extractelement">'<tt>extractelement</tt>' Instruction</a>
1935 <div class="doc_text">
1940 <result> = extractelement <n x <ty>> <val>, uint <idx> <i>; yields <ty></i>
1946 The '<tt>extractelement</tt>' instruction extracts a single scalar
1947 element from a packed vector at a specified index.
1954 The first operand of an '<tt>extractelement</tt>' instruction is a
1955 value of <a href="#t_packed">packed</a> type. The second operand is
1956 an index indicating the position from which to extract the element.
1957 The index may be a variable.</p>
1962 The result is a scalar of the same type as the element type of
1963 <tt>val</tt>. Its value is the value at position <tt>idx</tt> of
1964 <tt>val</tt>. If <tt>idx</tt> exceeds the length of <tt>val</tt>, the
1965 results are undefined.
1971 %result = extractelement <4 x int> %vec, uint 0 <i>; yields int</i>
1976 <!-- _______________________________________________________________________ -->
1977 <div class="doc_subsubsection">
1978 <a name="i_insertelement">'<tt>insertelement</tt>' Instruction</a>
1981 <div class="doc_text">
1986 <result> = insertelement <n x <ty>> <val>, <ty> <elt>, uint <idx> <i>; yields <n x <ty>></i>
1992 The '<tt>insertelement</tt>' instruction inserts a scalar
1993 element into a packed vector at a specified index.
2000 The first operand of an '<tt>insertelement</tt>' instruction is a
2001 value of <a href="#t_packed">packed</a> type. The second operand is a
2002 scalar value whose type must equal the element type of the first
2003 operand. The third operand is an index indicating the position at
2004 which to insert the value. The index may be a variable.</p>
2009 The result is a packed vector of the same type as <tt>val</tt>. Its
2010 element values are those of <tt>val</tt> except at position
2011 <tt>idx</tt>, where it gets the value <tt>elt</tt>. If <tt>idx</tt>
2012 exceeds the length of <tt>val</tt>, the results are undefined.
2018 %result = insertelement <4 x int> %vec, int 1, uint 0 <i>; yields <4 x int></i>
2022 <!-- _______________________________________________________________________ -->
2023 <div class="doc_subsubsection">
2024 <a name="i_shufflevector">'<tt>shufflevector</tt>' Instruction</a>
2027 <div class="doc_text">
2032 <result> = shufflevector <n x <ty>> <v1>, <n x <ty>> <v2>, <n x uint> <mask> <i>; yields <n x <ty>></i>
2038 The '<tt>shufflevector</tt>' instruction constructs a permutation of elements
2039 from two input vectors, returning a vector of the same type.
2045 The first two operands of a '<tt>shufflevector</tt>' instruction are vectors
2046 with types that match each other and types that match the result of the
2047 instruction. The third argument is a shuffle mask, which has the same number
2048 of elements as the other vector type, but whose element type is always 'uint'.
2052 The shuffle mask operand is required to be a constant vector with either
2053 constant integer or undef values.
2059 The elements of the two input vectors are numbered from left to right across
2060 both of the vectors. The shuffle mask operand specifies, for each element of
2061 the result vector, which element of the two input registers the result element
2062 gets. The element selector may be undef (meaning "don't care") and the second
2063 operand may be undef if performing a shuffle from only one vector.
2069 %result = shufflevector <4 x int> %v1, <4 x int> %v2,
2070 <4 x uint> <uint 0, uint 4, uint 1, uint 5> <i>; yields <4 x int></i>
2071 %result = shufflevector <4 x int> %v1, <4 x int> undef,
2072 <4 x uint> <uint 0, uint 1, uint 2, uint 3> <i>; yields <4 x int></i> - Identity shuffle.
2077 <!-- _______________________________________________________________________ -->
2078 <div class="doc_subsubsection"> <a name="i_vsetint">'<tt>vsetint</tt>'
2079 Instruction</a> </div>
2080 <div class="doc_text">
2082 <pre><result> = vsetint <op>, <n x <ty>> <var1>, <var2> <i>; yields <n x bool></i>
2087 <p>The '<tt>vsetint</tt>' instruction takes two integer vectors and
2088 returns a vector of boolean values representing, at each position, the
2089 result of the comparison between the values at that position in the
2094 <p>The arguments to a '<tt>vsetint</tt>' instruction are a comparison
2095 operation and two value arguments. The value arguments must be of <a
2096 href="#t_integral">integral</a> <a href="#t_packed">packed</a> type,
2097 and they must have identical types. The operation argument must be
2098 one of <tt>eq</tt>, <tt>ne</tt>, <tt>slt</tt>, <tt>sgt</tt>,
2099 <tt>sle</tt>, <tt>sge</tt>, <tt>ult</tt>, <tt>ugt</tt>, <tt>ule</tt>,
2100 <tt>uge</tt>, <tt>true</tt>, and <tt>false</tt>. The result is a
2101 packed <tt>bool</tt> value with the same length as each operand.</p>
2105 <p>The following table shows the semantics of '<tt>vsetint</tt>'. For
2106 each position of the result, the comparison is done on the
2107 corresponding positions of the two value arguments. Note that the
2108 signedness of the comparison depends on the comparison opcode and
2109 <i>not</i> on the signedness of the value operands. E.g., <tt>vsetint
2110 slt <4 x unsigned> %x, %y</tt> does an elementwise <i>signed</i>
2111 comparison of <tt>%x</tt> and <tt>%y</tt>.</p>
2113 <table border="1" cellspacing="0" cellpadding="4">
2115 <tr><th>Operation</th><th>Result is true iff</th><th>Comparison is</th></tr>
2116 <tr><td><tt>eq</tt></td><td>var1 == var2</td><td>--</td></tr>
2117 <tr><td><tt>ne</tt></td><td>var1 != var2</td><td>--</td></tr>
2118 <tr><td><tt>slt</tt></td><td>var1 < var2</td><td>signed</td></tr>
2119 <tr><td><tt>sgt</tt></td><td>var1 > var2</td><td>signed</td></tr>
2120 <tr><td><tt>sle</tt></td><td>var1 <= var2</td><td>signed</td></tr>
2121 <tr><td><tt>sge</tt></td><td>var1 >= var2</td><td>signed</td></tr>
2122 <tr><td><tt>ult</tt></td><td>var1 < var2</td><td>unsigned</td></tr>
2123 <tr><td><tt>ugt</tt></td><td>var1 > var2</td><td>unsigned</td></tr>
2124 <tr><td><tt>ule</tt></td><td>var1 <= var2</td><td>unsigned</td></tr>
2125 <tr><td><tt>uge</tt></td><td>var1 >= var2</td><td>unsigned</td></tr>
2126 <tr><td><tt>true</tt></td><td>always</td><td>--</td></tr>
2127 <tr><td><tt>false</tt></td><td>never</td><td>--</td></tr>
2132 <pre> <result> = vsetint eq <2 x int> <int 0, int 1>, <int 1, int 0> <i>; yields {<2 x bool>}:result = false, false</i>
2133 <result> = vsetint ne <2 x int> <int 0, int 1>, <int 1, int 0> <i>; yields {<2 x bool>}:result = true, true</i>
2134 <result> = vsetint slt <2 x int> <int 0, int 1>, <int 1, int 0> <i>; yields {<2 x bool>}:result = true, false</i>
2135 <result> = vsetint sgt <2 x int> <int 0, int 1>, <int 1, int 0> <i>; yields {<2 x bool>}:result = false, true</i>
2136 <result> = vsetint sle <2 x int> <int 0, int 1>, <int 1, int 0> <i>; yields {<2 x bool>}:result = true, false</i>
2137 <result> = vsetint sge <2 x int> <int 0, int 1>, <int 1, int 0> <i>; yields {<2 x bool>}:result = false, true</i>
2141 <!-- _______________________________________________________________________ -->
2142 <div class="doc_subsubsection"> <a name="i_vsetfp">'<tt>vsetfp</tt>'
2143 Instruction</a> </div>
2144 <div class="doc_text">
2146 <pre><result> = vsetfp <op>, <n x <ty>> <var1>, <var2> <i>; yields <n x bool></i>
2151 <p>The '<tt>vsetfp</tt>' instruction takes two floating point vector
2152 arguments and returns a vector of boolean values representing, at each
2153 position, the result of the comparison between the values at that
2154 position in the two operands.</p>
2158 <p>The arguments to a '<tt>vsetfp</tt>' instruction are a comparison
2159 operation and two value arguments. The value arguments must be of <a
2160 href="t_floating">floating point</a> <a href="#t_packed">packed</a>
2161 type, and they must have identical types. The operation argument must
2162 be one of <tt>eq</tt>, <tt>ne</tt>, <tt>lt</tt>, <tt>gt</tt>,
2163 <tt>le</tt>, <tt>ge</tt>, <tt>oeq</tt>, <tt>one</tt>, <tt>olt</tt>,
2164 <tt>ogt</tt>, <tt>ole</tt>, <tt>oge</tt>, <tt>ueq</tt>, <tt>une</tt>,
2165 <tt>ult</tt>, <tt>ugt</tt>, <tt>ule</tt>, <tt>uge</tt>, <tt>o</tt>,
2166 <tt>u</tt>, <tt>true</tt>, and <tt>false</tt>. The result is a packed
2167 <tt>bool</tt> value with the same length as each operand.</p>
2171 <p>The following table shows the semantics of '<tt>vsetfp</tt>' for
2172 floating point types. If either operand is a floating point Not a
2173 Number (NaN) value, the operation is unordered, and the value in the
2174 first column below is produced at that position. Otherwise, the
2175 operation is ordered, and the value in the second column is
2178 <table border="1" cellspacing="0" cellpadding="4">
2180 <tr><th>Operation</th><th>If unordered<th>Otherwise true iff</th></tr>
2181 <tr><td><tt>eq</tt></td><td>undefined</td><td>var1 == var2</td></tr>
2182 <tr><td><tt>ne</tt></td><td>undefined</td><td>var1 != var2</td></tr>
2183 <tr><td><tt>lt</tt></td><td>undefined</td><td>var1 < var2</td></tr>
2184 <tr><td><tt>gt</tt></td><td>undefined</td><td>var1 > var2</td></tr>
2185 <tr><td><tt>le</tt></td><td>undefined</td><td>var1 <= var2</td></tr>
2186 <tr><td><tt>ge</tt></td><td>undefined</td><td>var1 >= var2</td></tr>
2187 <tr><td><tt>oeq</tt></td><td>false</td><td>var1 == var2</td></tr>
2188 <tr><td><tt>one</tt></td><td>false</td><td>var1 != var2</td></tr>
2189 <tr><td><tt>olt</tt></td><td>false</td><td>var1 < var2</td></tr>
2190 <tr><td><tt>ogt</tt></td><td>false</td><td>var1 > var2</td></tr>
2191 <tr><td><tt>ole</tt></td><td>false</td><td>var1 <= var2</td></tr>
2192 <tr><td><tt>oge</tt></td><td>false</td><td>var1 >= var2</td></tr>
2193 <tr><td><tt>ueq</tt></td><td>true</td><td>var1 == var2</td></tr>
2194 <tr><td><tt>une</tt></td><td>true</td><td>var1 != var2</td></tr>
2195 <tr><td><tt>ult</tt></td><td>true</td><td>var1 < var2</td></tr>
2196 <tr><td><tt>ugt</tt></td><td>true</td><td>var1 > var2</td></tr>
2197 <tr><td><tt>ule</tt></td><td>true</td><td>var1 <= var2</td></tr>
2198 <tr><td><tt>uge</tt></td><td>true</td><td>var1 >= var2</td></tr>
2199 <tr><td><tt>o</tt></td><td>false</td><td>always</td></tr>
2200 <tr><td><tt>u</tt></td><td>true</td><td>never</td></tr>
2201 <tr><td><tt>true</tt></td><td>true</td><td>always</td></tr>
2202 <tr><td><tt>false</tt></td><td>false</td><td>never</td></tr>
2207 <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>
2208 <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>
2209 <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>
2210 <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>
2211 <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>
2212 <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>
2216 <!-- _______________________________________________________________________ -->
2217 <div class="doc_subsubsection">
2218 <a name="i_vselect">'<tt>vselect</tt>' Instruction</a>
2221 <div class="doc_text">
2226 <result> = vselect <n x bool> <cond>, <n x <ty>> <val1>, <n x <ty>> <val2> <i>; yields <n x <ty>></i>
2232 The '<tt>vselect</tt>' instruction chooses one value at each position
2233 of a vector based on a condition.
2240 The '<tt>vselect</tt>' instruction requires a <a
2241 href="#t_packed">packed</a> <tt>bool</tt> value indicating the
2242 condition at each vector position, and two values of the same packed
2243 type. All three operands must have the same length. The type of the
2244 result is the same as the type of the two value operands.</p>
2249 At each position where the <tt>bool</tt> vector is true, that position
2250 of the result gets its value from the first value argument; otherwise,
2251 it gets its value from the second value argument.
2257 %X = vselect bool <2 x bool> <bool true, bool false>, <2 x ubyte> <ubyte 17, ubyte 17>,
2258 <2 x ubyte> <ubyte 42, ubyte 42> <i>; yields <2 x ubyte>:17, 42</i>
2264 <!-- ======================================================================= -->
2265 <div class="doc_subsection">
2266 <a name="memoryops">Memory Access Operations</a>
2269 <div class="doc_text">
2271 <p>A key design point of an SSA-based representation is how it
2272 represents memory. In LLVM, no memory locations are in SSA form, which
2273 makes things very simple. This section describes how to read, write,
2274 allocate, and free memory in LLVM.</p>
2278 <!-- _______________________________________________________________________ -->
2279 <div class="doc_subsubsection">
2280 <a name="i_malloc">'<tt>malloc</tt>' Instruction</a>
2283 <div class="doc_text">
2288 <result> = malloc <type>[, uint <NumElements>][, align <alignment>] <i>; yields {type*}:result</i>
2293 <p>The '<tt>malloc</tt>' instruction allocates memory from the system
2294 heap and returns a pointer to it.</p>
2298 <p>The '<tt>malloc</tt>' instruction allocates
2299 <tt>sizeof(<type>)*NumElements</tt>
2300 bytes of memory from the operating system and returns a pointer of the
2301 appropriate type to the program. If "NumElements" is specified, it is the
2302 number of elements allocated. If an alignment is specified, the value result
2303 of the allocation is guaranteed to be aligned to at least that boundary. If
2304 not specified, or if zero, the target can choose to align the allocation on any
2305 convenient boundary.</p>
2307 <p>'<tt>type</tt>' must be a sized type.</p>
2311 <p>Memory is allocated using the system "<tt>malloc</tt>" function, and
2312 a pointer is returned.</p>
2317 %array = malloc [4 x ubyte ] <i>; yields {[%4 x ubyte]*}:array</i>
2319 %size = <a href="#i_add">add</a> uint 2, 2 <i>; yields {uint}:size = uint 4</i>
2320 %array1 = malloc ubyte, uint 4 <i>; yields {ubyte*}:array1</i>
2321 %array2 = malloc [12 x ubyte], uint %size <i>; yields {[12 x ubyte]*}:array2</i>
2322 %array3 = malloc int, uint 4, align 1024 <i>; yields {int*}:array3</i>
2323 %array4 = malloc int, align 1024 <i>; yields {int*}:array4</i>
2327 <!-- _______________________________________________________________________ -->
2328 <div class="doc_subsubsection">
2329 <a name="i_free">'<tt>free</tt>' Instruction</a>
2332 <div class="doc_text">
2337 free <type> <value> <i>; yields {void}</i>
2342 <p>The '<tt>free</tt>' instruction returns memory back to the unused
2343 memory heap to be reallocated in the future.</p>
2347 <p>'<tt>value</tt>' shall be a pointer value that points to a value
2348 that was allocated with the '<tt><a href="#i_malloc">malloc</a></tt>'
2353 <p>Access to the memory pointed to by the pointer is no longer defined
2354 after this instruction executes.</p>
2359 %array = <a href="#i_malloc">malloc</a> [4 x ubyte] <i>; yields {[4 x ubyte]*}:array</i>
2360 free [4 x ubyte]* %array
2364 <!-- _______________________________________________________________________ -->
2365 <div class="doc_subsubsection">
2366 <a name="i_alloca">'<tt>alloca</tt>' Instruction</a>
2369 <div class="doc_text">
2374 <result> = alloca <type>[, uint <NumElements>][, align <alignment>] <i>; yields {type*}:result</i>
2379 <p>The '<tt>alloca</tt>' instruction allocates memory on the current
2380 stack frame of the procedure that is live until the current function
2381 returns to its caller.</p>
2385 <p>The '<tt>alloca</tt>' instruction allocates <tt>sizeof(<type>)*NumElements</tt>
2386 bytes of memory on the runtime stack, returning a pointer of the
2387 appropriate type to the program. If "NumElements" is specified, it is the
2388 number of elements allocated. If an alignment is specified, the value result
2389 of the allocation is guaranteed to be aligned to at least that boundary. If
2390 not specified, or if zero, the target can choose to align the allocation on any
2391 convenient boundary.</p>
2393 <p>'<tt>type</tt>' may be any sized type.</p>
2397 <p>Memory is allocated; a pointer is returned. '<tt>alloca</tt>'d
2398 memory is automatically released when the function returns. The '<tt>alloca</tt>'
2399 instruction is commonly used to represent automatic variables that must
2400 have an address available. When the function returns (either with the <tt><a
2401 href="#i_ret">ret</a></tt> or <tt><a href="#i_unwind">unwind</a></tt>
2402 instructions), the memory is reclaimed.</p>
2407 %ptr = alloca int <i>; yields {int*}:ptr</i>
2408 %ptr = alloca int, uint 4 <i>; yields {int*}:ptr</i>
2409 %ptr = alloca int, uint 4, align 1024 <i>; yields {int*}:ptr</i>
2410 %ptr = alloca int, align 1024 <i>; yields {int*}:ptr</i>
2414 <!-- _______________________________________________________________________ -->
2415 <div class="doc_subsubsection"> <a name="i_load">'<tt>load</tt>'
2416 Instruction</a> </div>
2417 <div class="doc_text">
2419 <pre> <result> = load <ty>* <pointer><br> <result> = volatile load <ty>* <pointer><br></pre>
2421 <p>The '<tt>load</tt>' instruction is used to read from memory.</p>
2423 <p>The argument to the '<tt>load</tt>' instruction specifies the memory
2424 address from which to load. The pointer must point to a <a
2425 href="#t_firstclass">first class</a> type. If the <tt>load</tt> is
2426 marked as <tt>volatile</tt>, then the optimizer is not allowed to modify
2427 the number or order of execution of this <tt>load</tt> with other
2428 volatile <tt>load</tt> and <tt><a href="#i_store">store</a></tt>
2431 <p>The location of memory pointed to is loaded.</p>
2433 <pre> %ptr = <a href="#i_alloca">alloca</a> int <i>; yields {int*}:ptr</i>
2435 href="#i_store">store</a> int 3, int* %ptr <i>; yields {void}</i>
2436 %val = load int* %ptr <i>; yields {int}:val = int 3</i>
2439 <!-- _______________________________________________________________________ -->
2440 <div class="doc_subsubsection"> <a name="i_store">'<tt>store</tt>'
2441 Instruction</a> </div>
2443 <pre> store <ty> <value>, <ty>* <pointer> <i>; yields {void}</i>
2444 volatile store <ty> <value>, <ty>* <pointer> <i>; yields {void}</i>
2447 <p>The '<tt>store</tt>' instruction is used to write to memory.</p>
2449 <p>There are two arguments to the '<tt>store</tt>' instruction: a value
2450 to store and an address in which to store it. The type of the '<tt><pointer></tt>'
2451 operand must be a pointer to the type of the '<tt><value></tt>'
2452 operand. If the <tt>store</tt> is marked as <tt>volatile</tt>, then the
2453 optimizer is not allowed to modify the number or order of execution of
2454 this <tt>store</tt> with other volatile <tt>load</tt> and <tt><a
2455 href="#i_store">store</a></tt> instructions.</p>
2457 <p>The contents of memory are updated to contain '<tt><value></tt>'
2458 at the location specified by the '<tt><pointer></tt>' operand.</p>
2460 <pre> %ptr = <a href="#i_alloca">alloca</a> int <i>; yields {int*}:ptr</i>
2462 href="#i_store">store</a> int 3, int* %ptr <i>; yields {void}</i>
2463 %val = load int* %ptr <i>; yields {int}:val = int 3</i>
2465 <!-- _______________________________________________________________________ -->
2466 <div class="doc_subsubsection">
2467 <a name="i_getelementptr">'<tt>getelementptr</tt>' Instruction</a>
2470 <div class="doc_text">
2473 <result> = getelementptr <ty>* <ptrval>{, <ty> <idx>}*
2479 The '<tt>getelementptr</tt>' instruction is used to get the address of a
2480 subelement of an aggregate data structure.</p>
2484 <p>This instruction takes a list of integer constants that indicate what
2485 elements of the aggregate object to index to. The actual types of the arguments
2486 provided depend on the type of the first pointer argument. The
2487 '<tt>getelementptr</tt>' instruction is used to index down through the type
2488 levels of a structure or to a specific index in an array. When indexing into a
2489 structure, only <tt>uint</tt>
2490 integer constants are allowed. When indexing into an array or pointer,
2491 <tt>int</tt> and <tt>long</tt> indexes are allowed of any sign.</p>
2493 <p>For example, let's consider a C code fragment and how it gets
2494 compiled to LLVM:</p>
2508 int *foo(struct ST *s) {
2509 return &s[1].Z.B[5][13];
2513 <p>The LLVM code generated by the GCC frontend is:</p>
2516 %RT = type { sbyte, [10 x [20 x int]], sbyte }
2517 %ST = type { int, double, %RT }
2521 int* %foo(%ST* %s) {
2523 %reg = getelementptr %ST* %s, int 1, uint 2, uint 1, int 5, int 13
2530 <p>The index types specified for the '<tt>getelementptr</tt>' instruction depend
2531 on the pointer type that is being indexed into. <a href="#t_pointer">Pointer</a>
2532 and <a href="#t_array">array</a> types require <tt>uint</tt>, <tt>int</tt>,
2533 <tt>ulong</tt>, or <tt>long</tt> values, and <a href="#t_struct">structure</a>
2534 types require <tt>uint</tt> <b>constants</b>.</p>
2536 <p>In the example above, the first index is indexing into the '<tt>%ST*</tt>'
2537 type, which is a pointer, yielding a '<tt>%ST</tt>' = '<tt>{ int, double, %RT
2538 }</tt>' type, a structure. The second index indexes into the third element of
2539 the structure, yielding a '<tt>%RT</tt>' = '<tt>{ sbyte, [10 x [20 x int]],
2540 sbyte }</tt>' type, another structure. The third index indexes into the second
2541 element of the structure, yielding a '<tt>[10 x [20 x int]]</tt>' type, an
2542 array. The two dimensions of the array are subscripted into, yielding an
2543 '<tt>int</tt>' type. The '<tt>getelementptr</tt>' instruction returns a pointer
2544 to this element, thus computing a value of '<tt>int*</tt>' type.</p>
2546 <p>Note that it is perfectly legal to index partially through a
2547 structure, returning a pointer to an inner element. Because of this,
2548 the LLVM code for the given testcase is equivalent to:</p>
2551 int* %foo(%ST* %s) {
2552 %t1 = getelementptr %ST* %s, int 1 <i>; yields %ST*:%t1</i>
2553 %t2 = getelementptr %ST* %t1, int 0, uint 2 <i>; yields %RT*:%t2</i>
2554 %t3 = getelementptr %RT* %t2, int 0, uint 1 <i>; yields [10 x [20 x int]]*:%t3</i>
2555 %t4 = getelementptr [10 x [20 x int]]* %t3, int 0, int 5 <i>; yields [20 x int]*:%t4</i>
2556 %t5 = getelementptr [20 x int]* %t4, int 0, int 13 <i>; yields int*:%t5</i>
2561 <p>Note that it is undefined to access an array out of bounds: array and
2562 pointer indexes must always be within the defined bounds of the array type.
2563 The one exception for this rules is zero length arrays. These arrays are
2564 defined to be accessible as variable length arrays, which requires access
2565 beyond the zero'th element.</p>
2570 <i>; yields [12 x ubyte]*:aptr</i>
2571 %aptr = getelementptr {int, [12 x ubyte]}* %sptr, long 0, uint 1
2575 <!-- ======================================================================= -->
2576 <div class="doc_subsection"> <a name="otherops">Other Operations</a> </div>
2577 <div class="doc_text">
2578 <p>The instructions in this category are the "miscellaneous"
2579 instructions, which defy better classification.</p>
2581 <!-- _______________________________________________________________________ -->
2582 <div class="doc_subsubsection"> <a name="i_phi">'<tt>phi</tt>'
2583 Instruction</a> </div>
2584 <div class="doc_text">
2586 <pre> <result> = phi <ty> [ <val0>, <label0>], ...<br></pre>
2588 <p>The '<tt>phi</tt>' instruction is used to implement the φ node in
2589 the SSA graph representing the function.</p>
2591 <p>The type of the incoming values are specified with the first type
2592 field. After this, the '<tt>phi</tt>' instruction takes a list of pairs
2593 as arguments, with one pair for each predecessor basic block of the
2594 current block. Only values of <a href="#t_firstclass">first class</a>
2595 type may be used as the value arguments to the PHI node. Only labels
2596 may be used as the label arguments.</p>
2597 <p>There must be no non-phi instructions between the start of a basic
2598 block and the PHI instructions: i.e. PHI instructions must be first in
2601 <p>At runtime, the '<tt>phi</tt>' instruction logically takes on the
2602 value specified by the parameter, depending on which basic block we
2603 came from in the last <a href="#terminators">terminator</a> instruction.</p>
2605 <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>
2608 <!-- _______________________________________________________________________ -->
2609 <div class="doc_subsubsection">
2610 <a name="i_cast">'<tt>cast .. to</tt>' Instruction</a>
2613 <div class="doc_text">
2618 <result> = cast <ty> <value> to <ty2> <i>; yields ty2</i>
2624 The '<tt>cast</tt>' instruction is used as the primitive means to convert
2625 integers to floating point, change data type sizes, and break type safety (by
2633 The '<tt>cast</tt>' instruction takes a value to cast, which must be a first
2634 class value, and a type to cast it to, which must also be a <a
2635 href="#t_firstclass">first class</a> type.
2641 This instruction follows the C rules for explicit casts when determining how the
2642 data being cast must change to fit in its new container.
2646 When casting to bool, any value that would be considered true in the context of
2647 a C '<tt>if</tt>' condition is converted to the boolean '<tt>true</tt>' values,
2648 all else are '<tt>false</tt>'.
2652 When extending an integral value from a type of one signness to another (for
2653 example '<tt>sbyte</tt>' to '<tt>ulong</tt>'), the value is sign-extended if the
2654 <b>source</b> value is signed, and zero-extended if the source value is
2655 unsigned. <tt>bool</tt> values are always zero extended into either zero or
2662 %X = cast int 257 to ubyte <i>; yields ubyte:1</i>
2663 %Y = cast int 123 to bool <i>; yields bool:true</i>
2667 <!-- _______________________________________________________________________ -->
2668 <div class="doc_subsubsection">
2669 <a name="i_select">'<tt>select</tt>' Instruction</a>
2672 <div class="doc_text">
2677 <result> = select bool <cond>, <ty> <val1>, <ty> <val2> <i>; yields ty</i>
2683 The '<tt>select</tt>' instruction is used to choose one value based on a
2684 condition, without branching.
2691 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.
2697 If the boolean condition evaluates to true, the instruction returns the first
2698 value argument; otherwise, it returns the second value argument.
2704 %X = select bool true, ubyte 17, ubyte 42 <i>; yields ubyte:17</i>
2709 <!-- _______________________________________________________________________ -->
2710 <div class="doc_subsubsection">
2711 <a name="i_call">'<tt>call</tt>' Instruction</a>
2714 <div class="doc_text">
2718 <result> = [tail] call [<a href="#callingconv">cconv</a>] <ty>* <fnptrval>(<param list>)
2723 <p>The '<tt>call</tt>' instruction represents a simple function call.</p>
2727 <p>This instruction requires several arguments:</p>
2731 <p>The optional "tail" marker indicates whether the callee function accesses
2732 any allocas or varargs in the caller. If the "tail" marker is present, the
2733 function call is eligible for tail call optimization. Note that calls may
2734 be marked "tail" even if they do not occur before a <a
2735 href="#i_ret"><tt>ret</tt></a> instruction.
2738 <p>The optional "cconv" marker indicates which <a href="callingconv">calling
2739 convention</a> the call should use. If none is specified, the call defaults
2740 to using C calling conventions.
2743 <p>'<tt>ty</tt>': shall be the signature of the pointer to function value
2744 being invoked. The argument types must match the types implied by this
2745 signature. This type can be omitted if the function is not varargs and
2746 if the function type does not return a pointer to a function.</p>
2749 <p>'<tt>fnptrval</tt>': An LLVM value containing a pointer to a function to
2750 be invoked. In most cases, this is a direct function invocation, but
2751 indirect <tt>call</tt>s are just as possible, calling an arbitrary pointer
2752 to function value.</p>
2755 <p>'<tt>function args</tt>': argument list whose types match the
2756 function signature argument types. All arguments must be of
2757 <a href="#t_firstclass">first class</a> type. If the function signature
2758 indicates the function accepts a variable number of arguments, the extra
2759 arguments can be specified.</p>
2765 <p>The '<tt>call</tt>' instruction is used to cause control flow to
2766 transfer to a specified function, with its incoming arguments bound to
2767 the specified values. Upon a '<tt><a href="#i_ret">ret</a></tt>'
2768 instruction in the called function, control flow continues with the
2769 instruction after the function call, and the return value of the
2770 function is bound to the result argument. This is a simpler case of
2771 the <a href="#i_invoke">invoke</a> instruction.</p>
2776 %retval = call int %test(int %argc)
2777 call int(sbyte*, ...) *%printf(sbyte* %msg, int 12, sbyte 42);
2778 %X = tail call int %foo()
2779 %Y = tail call <a href="#callingconv">fastcc</a> int %foo()
2784 <!-- _______________________________________________________________________ -->
2785 <div class="doc_subsubsection">
2786 <a name="i_va_arg">'<tt>va_arg</tt>' Instruction</a>
2789 <div class="doc_text">
2794 <resultval> = va_arg <va_list*> <arglist>, <argty>
2799 <p>The '<tt>va_arg</tt>' instruction is used to access arguments passed through
2800 the "variable argument" area of a function call. It is used to implement the
2801 <tt>va_arg</tt> macro in C.</p>
2805 <p>This instruction takes a <tt>va_list*</tt> value and the type of
2806 the argument. It returns a value of the specified argument type and
2807 increments the <tt>va_list</tt> to point to the next argument. Again, the
2808 actual type of <tt>va_list</tt> is target specific.</p>
2812 <p>The '<tt>va_arg</tt>' instruction loads an argument of the specified
2813 type from the specified <tt>va_list</tt> and causes the
2814 <tt>va_list</tt> to point to the next argument. For more information,
2815 see the variable argument handling <a href="#int_varargs">Intrinsic
2818 <p>It is legal for this instruction to be called in a function which does not
2819 take a variable number of arguments, for example, the <tt>vfprintf</tt>
2822 <p><tt>va_arg</tt> is an LLVM instruction instead of an <a
2823 href="#intrinsics">intrinsic function</a> because it takes a type as an
2828 <p>See the <a href="#int_varargs">variable argument processing</a> section.</p>
2832 <!-- *********************************************************************** -->
2833 <div class="doc_section"> <a name="intrinsics">Intrinsic Functions</a> </div>
2834 <!-- *********************************************************************** -->
2836 <div class="doc_text">
2838 <p>LLVM supports the notion of an "intrinsic function". These functions have
2839 well known names and semantics and are required to follow certain
2840 restrictions. Overall, these instructions represent an extension mechanism for
2841 the LLVM language that does not require changing all of the transformations in
2842 LLVM to add to the language (or the bytecode reader/writer, the parser,
2845 <p>Intrinsic function names must all start with an "<tt>llvm.</tt>" prefix. This
2846 prefix is reserved in LLVM for intrinsic names; thus, functions may not be named
2847 this. Intrinsic functions must always be external functions: you cannot define
2848 the body of intrinsic functions. Intrinsic functions may only be used in call
2849 or invoke instructions: it is illegal to take the address of an intrinsic
2850 function. Additionally, because intrinsic functions are part of the LLVM
2851 language, it is required that they all be documented here if any are added.</p>
2854 <p>To learn how to add an intrinsic function, please see the <a
2855 href="ExtendingLLVM.html">Extending LLVM Guide</a>.
2860 <!-- ======================================================================= -->
2861 <div class="doc_subsection">
2862 <a name="int_varargs">Variable Argument Handling Intrinsics</a>
2865 <div class="doc_text">
2867 <p>Variable argument support is defined in LLVM with the <a
2868 href="#i_va_arg"><tt>va_arg</tt></a> instruction and these three
2869 intrinsic functions. These functions are related to the similarly
2870 named macros defined in the <tt><stdarg.h></tt> header file.</p>
2872 <p>All of these functions operate on arguments that use a
2873 target-specific value type "<tt>va_list</tt>". The LLVM assembly
2874 language reference manual does not define what this type is, so all
2875 transformations should be prepared to handle intrinsics with any type
2878 <p>This example shows how the <a href="#i_va_arg"><tt>va_arg</tt></a>
2879 instruction and the variable argument handling intrinsic functions are
2883 int %test(int %X, ...) {
2884 ; Initialize variable argument processing
2886 call void %<a href="#i_va_start">llvm.va_start</a>(sbyte** %ap)
2888 ; Read a single integer argument
2889 %tmp = va_arg sbyte** %ap, int
2891 ; Demonstrate usage of llvm.va_copy and llvm.va_end
2893 call void %<a href="#i_va_copy">llvm.va_copy</a>(sbyte** %aq, sbyte** %ap)
2894 call void %<a href="#i_va_end">llvm.va_end</a>(sbyte** %aq)
2896 ; Stop processing of arguments.
2897 call void %<a href="#i_va_end">llvm.va_end</a>(sbyte** %ap)
2903 <!-- _______________________________________________________________________ -->
2904 <div class="doc_subsubsection">
2905 <a name="i_va_start">'<tt>llvm.va_start</tt>' Intrinsic</a>
2909 <div class="doc_text">
2911 <pre> declare void %llvm.va_start(<va_list>* <arglist>)<br></pre>
2913 <P>The '<tt>llvm.va_start</tt>' intrinsic initializes
2914 <tt>*<arglist></tt> for subsequent use by <tt><a
2915 href="#i_va_arg">va_arg</a></tt>.</p>
2919 <P>The argument is a pointer to a <tt>va_list</tt> element to initialize.</p>
2923 <P>The '<tt>llvm.va_start</tt>' intrinsic works just like the <tt>va_start</tt>
2924 macro available in C. In a target-dependent way, it initializes the
2925 <tt>va_list</tt> element the argument points to, so that the next call to
2926 <tt>va_arg</tt> will produce the first variable argument passed to the function.
2927 Unlike the C <tt>va_start</tt> macro, this intrinsic does not need to know the
2928 last argument of the function, the compiler can figure that out.</p>
2932 <!-- _______________________________________________________________________ -->
2933 <div class="doc_subsubsection">
2934 <a name="i_va_end">'<tt>llvm.va_end</tt>' Intrinsic</a>
2937 <div class="doc_text">
2939 <pre> declare void %llvm.va_end(<va_list*> <arglist>)<br></pre>
2941 <p>The '<tt>llvm.va_end</tt>' intrinsic destroys <tt><arglist></tt>
2942 which has been initialized previously with <tt><a href="#i_va_start">llvm.va_start</a></tt>
2943 or <tt><a href="#i_va_copy">llvm.va_copy</a></tt>.</p>
2945 <p>The argument is a <tt>va_list</tt> to destroy.</p>
2947 <p>The '<tt>llvm.va_end</tt>' intrinsic works just like the <tt>va_end</tt>
2948 macro available in C. In a target-dependent way, it destroys the <tt>va_list</tt>.
2949 Calls to <a href="#i_va_start"><tt>llvm.va_start</tt></a> and <a
2950 href="#i_va_copy"><tt>llvm.va_copy</tt></a> must be matched exactly
2951 with calls to <tt>llvm.va_end</tt>.</p>
2954 <!-- _______________________________________________________________________ -->
2955 <div class="doc_subsubsection">
2956 <a name="i_va_copy">'<tt>llvm.va_copy</tt>' Intrinsic</a>
2959 <div class="doc_text">
2964 declare void %llvm.va_copy(<va_list>* <destarglist>,
2965 <va_list>* <srcarglist>)
2970 <p>The '<tt>llvm.va_copy</tt>' intrinsic copies the current argument position from
2971 the source argument list to the destination argument list.</p>
2975 <p>The first argument is a pointer to a <tt>va_list</tt> element to initialize.
2976 The second argument is a pointer to a <tt>va_list</tt> element to copy from.</p>
2981 <p>The '<tt>llvm.va_copy</tt>' intrinsic works just like the <tt>va_copy</tt> macro
2982 available in C. In a target-dependent way, it copies the source
2983 <tt>va_list</tt> element into the destination list. This intrinsic is necessary
2984 because the <tt><a href="i_va_begin">llvm.va_begin</a></tt> intrinsic may be
2985 arbitrarily complex and require memory allocation, for example.</p>
2989 <!-- ======================================================================= -->
2990 <div class="doc_subsection">
2991 <a name="int_gc">Accurate Garbage Collection Intrinsics</a>
2994 <div class="doc_text">
2997 LLVM support for <a href="GarbageCollection.html">Accurate Garbage
2998 Collection</a> requires the implementation and generation of these intrinsics.
2999 These intrinsics allow identification of <a href="#i_gcroot">GC roots on the
3000 stack</a>, as well as garbage collector implementations that require <a
3001 href="#i_gcread">read</a> and <a href="#i_gcwrite">write</a> barriers.
3002 Front-ends for type-safe garbage collected languages should generate these
3003 intrinsics to make use of the LLVM garbage collectors. For more details, see <a
3004 href="GarbageCollection.html">Accurate Garbage Collection with LLVM</a>.
3008 <!-- _______________________________________________________________________ -->
3009 <div class="doc_subsubsection">
3010 <a name="i_gcroot">'<tt>llvm.gcroot</tt>' Intrinsic</a>
3013 <div class="doc_text">
3018 declare void %llvm.gcroot(<ty>** %ptrloc, <ty2>* %metadata)
3023 <p>The '<tt>llvm.gcroot</tt>' intrinsic declares the existence of a GC root to
3024 the code generator, and allows some metadata to be associated with it.</p>
3028 <p>The first argument specifies the address of a stack object that contains the
3029 root pointer. The second pointer (which must be either a constant or a global
3030 value address) contains the meta-data to be associated with the root.</p>
3034 <p>At runtime, a call to this intrinsics stores a null pointer into the "ptrloc"
3035 location. At compile-time, the code generator generates information to allow
3036 the runtime to find the pointer at GC safe points.
3042 <!-- _______________________________________________________________________ -->
3043 <div class="doc_subsubsection">
3044 <a name="i_gcread">'<tt>llvm.gcread</tt>' Intrinsic</a>
3047 <div class="doc_text">
3052 declare sbyte* %llvm.gcread(sbyte* %ObjPtr, sbyte** %Ptr)
3057 <p>The '<tt>llvm.gcread</tt>' intrinsic identifies reads of references from heap
3058 locations, allowing garbage collector implementations that require read
3063 <p>The second argument is the address to read from, which should be an address
3064 allocated from the garbage collector. The first object is a pointer to the
3065 start of the referenced object, if needed by the language runtime (otherwise
3070 <p>The '<tt>llvm.gcread</tt>' intrinsic has the same semantics as a load
3071 instruction, but may be replaced with substantially more complex code by the
3072 garbage collector runtime, as needed.</p>
3077 <!-- _______________________________________________________________________ -->
3078 <div class="doc_subsubsection">
3079 <a name="i_gcwrite">'<tt>llvm.gcwrite</tt>' Intrinsic</a>
3082 <div class="doc_text">
3087 declare void %llvm.gcwrite(sbyte* %P1, sbyte* %Obj, sbyte** %P2)
3092 <p>The '<tt>llvm.gcwrite</tt>' intrinsic identifies writes of references to heap
3093 locations, allowing garbage collector implementations that require write
3094 barriers (such as generational or reference counting collectors).</p>
3098 <p>The first argument is the reference to store, the second is the start of the
3099 object to store it to, and the third is the address of the field of Obj to
3100 store to. If the runtime does not require a pointer to the object, Obj may be
3105 <p>The '<tt>llvm.gcwrite</tt>' intrinsic has the same semantics as a store
3106 instruction, but may be replaced with substantially more complex code by the
3107 garbage collector runtime, as needed.</p>
3113 <!-- ======================================================================= -->
3114 <div class="doc_subsection">
3115 <a name="int_codegen">Code Generator Intrinsics</a>
3118 <div class="doc_text">
3120 These intrinsics are provided by LLVM to expose special features that may only
3121 be implemented with code generator support.
3126 <!-- _______________________________________________________________________ -->
3127 <div class="doc_subsubsection">
3128 <a name="i_returnaddress">'<tt>llvm.returnaddress</tt>' Intrinsic</a>
3131 <div class="doc_text">
3135 declare sbyte *%llvm.returnaddress(uint <level>)
3141 The '<tt>llvm.returnaddress</tt>' intrinsic returns a target-specific value
3142 indicating the return address of the current function or one of its callers.
3148 The argument to this intrinsic indicates which function to return the address
3149 for. Zero indicates the calling function, one indicates its caller, etc. The
3150 argument is <b>required</b> to be a constant integer value.
3156 The '<tt>llvm.returnaddress</tt>' intrinsic either returns a pointer indicating
3157 the return address of the specified call frame, or zero if it cannot be
3158 identified. The value returned by this intrinsic is likely to be incorrect or 0
3159 for arguments other than zero, so it should only be used for debugging purposes.
3163 Note that calling this intrinsic does not prevent function inlining or other
3164 aggressive transformations, so the value returned may not be that of the obvious
3165 source-language caller.
3170 <!-- _______________________________________________________________________ -->
3171 <div class="doc_subsubsection">
3172 <a name="i_frameaddress">'<tt>llvm.frameaddress</tt>' Intrinsic</a>
3175 <div class="doc_text">
3179 declare sbyte *%llvm.frameaddress(uint <level>)
3185 The '<tt>llvm.frameaddress</tt>' intrinsic returns the target-specific frame
3186 pointer value for the specified stack frame.
3192 The argument to this intrinsic indicates which function to return the frame
3193 pointer for. Zero indicates the calling function, one indicates its caller,
3194 etc. The argument is <b>required</b> to be a constant integer value.
3200 The '<tt>llvm.frameaddress</tt>' intrinsic either returns a pointer indicating
3201 the frame 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.
3213 <!-- _______________________________________________________________________ -->
3214 <div class="doc_subsubsection">
3215 <a name="i_stacksave">'<tt>llvm.stacksave</tt>' Intrinsic</a>
3218 <div class="doc_text">
3222 declare sbyte *%llvm.stacksave()
3228 The '<tt>llvm.stacksave</tt>' intrinsic is used to remember the current state of
3229 the function stack, for use with <a href="#i_stackrestore">
3230 <tt>llvm.stackrestore</tt></a>. This is useful for implementing language
3231 features like scoped automatic variable sized arrays in C99.
3237 This intrinsic returns a opaque pointer value that can be passed to <a
3238 href="#i_stackrestore"><tt>llvm.stackrestore</tt></a>. When an
3239 <tt>llvm.stackrestore</tt> intrinsic is executed with a value saved from
3240 <tt>llvm.stacksave</tt>, it effectively restores the state of the stack to the
3241 state it was in when the <tt>llvm.stacksave</tt> intrinsic executed. In
3242 practice, this pops any <a href="#i_alloca">alloca</a> blocks from the stack
3243 that were allocated after the <tt>llvm.stacksave</tt> was executed.
3248 <!-- _______________________________________________________________________ -->
3249 <div class="doc_subsubsection">
3250 <a name="i_stackrestore">'<tt>llvm.stackrestore</tt>' Intrinsic</a>
3253 <div class="doc_text">
3257 declare void %llvm.stackrestore(sbyte* %ptr)
3263 The '<tt>llvm.stackrestore</tt>' intrinsic is used to restore the state of
3264 the function stack to the state it was in when the corresponding <a
3265 href="#llvm.stacksave"><tt>llvm.stacksave</tt></a> intrinsic executed. This is
3266 useful for implementing language features like scoped automatic variable sized
3273 See the description for <a href="#i_stacksave"><tt>llvm.stacksave</tt></a>.
3279 <!-- _______________________________________________________________________ -->
3280 <div class="doc_subsubsection">
3281 <a name="i_prefetch">'<tt>llvm.prefetch</tt>' Intrinsic</a>
3284 <div class="doc_text">
3288 declare void %llvm.prefetch(sbyte * <address>,
3289 uint <rw>, uint <locality>)
3296 The '<tt>llvm.prefetch</tt>' intrinsic is a hint to the code generator to insert
3297 a prefetch instruction if supported; otherwise, it is a noop. Prefetches have
3299 effect on the behavior of the program but can change its performance
3306 <tt>address</tt> is the address to be prefetched, <tt>rw</tt> is the specifier
3307 determining if the fetch should be for a read (0) or write (1), and
3308 <tt>locality</tt> is a temporal locality specifier ranging from (0) - no
3309 locality, to (3) - extremely local keep in cache. The <tt>rw</tt> and
3310 <tt>locality</tt> arguments must be constant integers.
3316 This intrinsic does not modify the behavior of the program. In particular,
3317 prefetches cannot trap and do not produce a value. On targets that support this
3318 intrinsic, the prefetch can provide hints to the processor cache for better
3324 <!-- _______________________________________________________________________ -->
3325 <div class="doc_subsubsection">
3326 <a name="i_pcmarker">'<tt>llvm.pcmarker</tt>' Intrinsic</a>
3329 <div class="doc_text">
3333 declare void %llvm.pcmarker( uint <id> )
3340 The '<tt>llvm.pcmarker</tt>' intrinsic is a method to export a Program Counter
3342 code to simulators and other tools. The method is target specific, but it is
3343 expected that the marker will use exported symbols to transmit the PC of the marker.
3344 The marker makes no guarantees that it will remain with any specific instruction
3345 after optimizations. It is possible that the presence of a marker will inhibit
3346 optimizations. The intended use is to be inserted after optimizations to allow
3347 correlations of simulation runs.
3353 <tt>id</tt> is a numerical id identifying the marker.
3359 This intrinsic does not modify the behavior of the program. Backends that do not
3360 support this intrinisic may ignore it.
3365 <!-- _______________________________________________________________________ -->
3366 <div class="doc_subsubsection">
3367 <a name="i_readcyclecounter">'<tt>llvm.readcyclecounter</tt>' Intrinsic</a>
3370 <div class="doc_text">
3374 declare ulong %llvm.readcyclecounter( )
3381 The '<tt>llvm.readcyclecounter</tt>' intrinsic provides access to the cycle
3382 counter register (or similar low latency, high accuracy clocks) on those targets
3383 that support it. On X86, it should map to RDTSC. On Alpha, it should map to RPCC.
3384 As the backing counters overflow quickly (on the order of 9 seconds on alpha), this
3385 should only be used for small timings.
3391 When directly supported, reading the cycle counter should not modify any memory.
3392 Implementations are allowed to either return a application specific value or a
3393 system wide value. On backends without support, this is lowered to a constant 0.
3398 <!-- ======================================================================= -->
3399 <div class="doc_subsection">
3400 <a name="int_libc">Standard C Library Intrinsics</a>
3403 <div class="doc_text">
3405 LLVM provides intrinsics for a few important standard C library functions.
3406 These intrinsics allow source-language front-ends to pass information about the
3407 alignment of the pointer arguments to the code generator, providing opportunity
3408 for more efficient code generation.
3413 <!-- _______________________________________________________________________ -->
3414 <div class="doc_subsubsection">
3415 <a name="i_memcpy">'<tt>llvm.memcpy</tt>' Intrinsic</a>
3418 <div class="doc_text">
3422 declare void %llvm.memcpy.i32(sbyte* <dest>, sbyte* <src>,
3423 uint <len>, uint <align>)
3424 declare void %llvm.memcpy.i64(sbyte* <dest>, sbyte* <src>,
3425 ulong <len>, uint <align>)
3431 The '<tt>llvm.memcpy.*</tt>' intrinsics copy a block of memory from the source
3432 location to the destination location.
3436 Note that, unlike the standard libc function, the <tt>llvm.memcpy.*</tt>
3437 intrinsics do not return a value, and takes an extra alignment argument.
3443 The first argument is a pointer to the destination, the second is a pointer to
3444 the source. The third argument is an integer argument
3445 specifying the number of bytes to copy, and the fourth argument is the alignment
3446 of the source and destination locations.
3450 If the call to this intrinisic has an alignment value that is not 0 or 1, then
3451 the caller guarantees that both the source and destination pointers are aligned
3458 The '<tt>llvm.memcpy.*</tt>' intrinsics copy a block of memory from the source
3459 location to the destination location, which are not allowed to overlap. It
3460 copies "len" bytes of memory over. If the argument is known to be aligned to
3461 some boundary, this can be specified as the fourth argument, otherwise it should
3467 <!-- _______________________________________________________________________ -->
3468 <div class="doc_subsubsection">
3469 <a name="i_memmove">'<tt>llvm.memmove</tt>' Intrinsic</a>
3472 <div class="doc_text">
3476 declare void %llvm.memmove.i32(sbyte* <dest>, sbyte* <src>,
3477 uint <len>, uint <align>)
3478 declare void %llvm.memmove.i64(sbyte* <dest>, sbyte* <src>,
3479 ulong <len>, uint <align>)
3485 The '<tt>llvm.memmove.*</tt>' intrinsics move a block of memory from the source
3486 location to the destination location. It is similar to the
3487 '<tt>llvm.memcmp</tt>' intrinsic but allows the two memory locations to overlap.
3491 Note that, unlike the standard libc function, the <tt>llvm.memmove.*</tt>
3492 intrinsics do not return a value, and takes an extra alignment argument.
3498 The first argument is a pointer to the destination, the second is a pointer to
3499 the source. The third argument is an integer argument
3500 specifying the number of bytes to copy, and the fourth argument is the alignment
3501 of the source and destination locations.
3505 If the call to this intrinisic has an alignment value that is not 0 or 1, then
3506 the caller guarantees that the source and destination pointers are aligned to
3513 The '<tt>llvm.memmove.*</tt>' intrinsics copy a block of memory from the source
3514 location to the destination location, which may overlap. It
3515 copies "len" bytes of memory over. If the argument is known to be aligned to
3516 some boundary, this can be specified as the fourth argument, otherwise it should
3522 <!-- _______________________________________________________________________ -->
3523 <div class="doc_subsubsection">
3524 <a name="i_memset">'<tt>llvm.memset.*</tt>' Intrinsics</a>
3527 <div class="doc_text">
3531 declare void %llvm.memset.i32(sbyte* <dest>, ubyte <val>,
3532 uint <len>, uint <align>)
3533 declare void %llvm.memset.i64(sbyte* <dest>, ubyte <val>,
3534 ulong <len>, uint <align>)
3540 The '<tt>llvm.memset.*</tt>' intrinsics fill a block of memory with a particular
3545 Note that, unlike the standard libc function, the <tt>llvm.memset</tt> intrinsic
3546 does not return a value, and takes an extra alignment argument.
3552 The first argument is a pointer to the destination to fill, the second is the
3553 byte value to fill it with, the third argument is an integer
3554 argument specifying the number of bytes to fill, and the fourth argument is the
3555 known alignment of destination location.
3559 If the call to this intrinisic has an alignment value that is not 0 or 1, then
3560 the caller guarantees that the destination pointer is aligned to that boundary.
3566 The '<tt>llvm.memset.*</tt>' intrinsics fill "len" bytes of memory starting at
3568 destination location. If the argument is known to be aligned to some boundary,
3569 this can be specified as the fourth argument, otherwise it should be set to 0 or
3575 <!-- _______________________________________________________________________ -->
3576 <div class="doc_subsubsection">
3577 <a name="i_isunordered">'<tt>llvm.isunordered.*</tt>' Intrinsic</a>
3580 <div class="doc_text">
3584 declare bool %llvm.isunordered.f32(float Val1, float Val2)
3585 declare bool %llvm.isunordered.f64(double Val1, double Val2)
3591 The '<tt>llvm.isunordered</tt>' intrinsics return true if either or both of the
3592 specified floating point values is a NAN.
3598 The arguments are floating point numbers of the same type.
3604 If either or both of the arguments is a SNAN or QNAN, it returns true, otherwise
3610 <!-- _______________________________________________________________________ -->
3611 <div class="doc_subsubsection">
3612 <a name="i_sqrt">'<tt>llvm.sqrt.*</tt>' Intrinsic</a>
3615 <div class="doc_text">
3619 declare double %llvm.sqrt.f32(float Val)
3620 declare double %llvm.sqrt.f64(double Val)
3626 The '<tt>llvm.sqrt</tt>' intrinsics return the sqrt of the specified operand,
3627 returning the same value as the libm '<tt>sqrt</tt>' function would. Unlike
3628 <tt>sqrt</tt> in libm, however, <tt>llvm.sqrt</tt> has undefined behavior for
3629 negative numbers (which allows for better optimization).
3635 The argument and return value are floating point numbers of the same type.
3641 This function returns the sqrt of the specified operand if it is a positive
3642 floating point number.
3646 <!-- ======================================================================= -->
3647 <div class="doc_subsection">
3648 <a name="int_manip">Bit Manipulation Intrinsics</a>
3651 <div class="doc_text">
3653 LLVM provides intrinsics for a few important bit manipulation operations.
3654 These allow efficient code generation for some algorithms.
3659 <!-- _______________________________________________________________________ -->
3660 <div class="doc_subsubsection">
3661 <a name="i_bswap">'<tt>llvm.bswap.*</tt>' Intrinsics</a>
3664 <div class="doc_text">
3668 declare ushort %llvm.bswap.i16(ushort <id>)
3669 declare uint %llvm.bswap.i32(uint <id>)
3670 declare ulong %llvm.bswap.i64(ulong <id>)
3676 The '<tt>llvm.bwsap</tt>' family of intrinsics is used to byteswap a 16, 32 or
3677 64 bit quantity. These are useful for performing operations on data that is not
3678 in the target's native byte order.
3684 The <tt>llvm.bswap.16</tt> intrinsic returns a ushort value that has the high and low
3685 byte of the input ushort swapped. Similarly, the <tt>llvm.bswap.i32</tt> intrinsic
3686 returns a uint value that has the four bytes of the input uint swapped, so that
3687 if the input bytes are numbered 0, 1, 2, 3 then the returned uint will have its
3688 bytes in 3, 2, 1, 0 order. The <tt>llvm.bswap.i64</tt> intrinsic extends this concept
3694 <!-- _______________________________________________________________________ -->
3695 <div class="doc_subsubsection">
3696 <a name="int_ctpop">'<tt>llvm.ctpop.*</tt>' Intrinsic</a>
3699 <div class="doc_text">
3703 declare ubyte %llvm.ctpop.i8 (ubyte <src>)
3704 declare ushort %llvm.ctpop.i16(ushort <src>)
3705 declare uint %llvm.ctpop.i32(uint <src>)
3706 declare ulong %llvm.ctpop.i64(ulong <src>)
3712 The '<tt>llvm.ctpop</tt>' family of intrinsics counts the number of bits set in a
3719 The only argument is the value to be counted. The argument may be of any
3720 unsigned integer type. The return type must match the argument type.
3726 The '<tt>llvm.ctpop</tt>' intrinsic counts the 1's in a variable.
3730 <!-- _______________________________________________________________________ -->
3731 <div class="doc_subsubsection">
3732 <a name="int_ctlz">'<tt>llvm.ctlz.*</tt>' Intrinsic</a>
3735 <div class="doc_text">
3739 declare ubyte %llvm.ctlz.i8 (ubyte <src>)
3740 declare ushort %llvm.ctlz.i16(ushort <src>)
3741 declare uint %llvm.ctlz.i32(uint <src>)
3742 declare ulong %llvm.ctlz.i64(ulong <src>)
3748 The '<tt>llvm.ctlz</tt>' family of intrinsic functions counts the number of
3749 leading zeros in a variable.
3755 The only argument is the value to be counted. The argument may be of any
3756 unsigned integer type. The return type must match the argument type.
3762 The '<tt>llvm.ctlz</tt>' intrinsic counts the leading (most significant) zeros
3763 in a variable. If the src == 0 then the result is the size in bits of the type
3764 of src. For example, <tt>llvm.ctlz(int 2) = 30</tt>.
3770 <!-- _______________________________________________________________________ -->
3771 <div class="doc_subsubsection">
3772 <a name="int_cttz">'<tt>llvm.cttz.*</tt>' Intrinsic</a>
3775 <div class="doc_text">
3779 declare ubyte %llvm.cttz.i8 (ubyte <src>)
3780 declare ushort %llvm.cttz.i16(ushort <src>)
3781 declare uint %llvm.cttz.i32(uint <src>)
3782 declare ulong %llvm.cttz.i64(ulong <src>)
3788 The '<tt>llvm.cttz</tt>' family of intrinsic functions counts the number of
3795 The only argument is the value to be counted. The argument may be of any
3796 unsigned integer type. The return type must match the argument type.
3802 The '<tt>llvm.cttz</tt>' intrinsic counts the trailing (least significant) zeros
3803 in a variable. If the src == 0 then the result is the size in bits of the type
3804 of src. For example, <tt>llvm.cttz(2) = 1</tt>.
3808 <!-- ======================================================================= -->
3809 <div class="doc_subsection">
3810 <a name="int_debugger">Debugger Intrinsics</a>
3813 <div class="doc_text">
3815 The LLVM debugger intrinsics (which all start with <tt>llvm.dbg.</tt> prefix),
3816 are described in the <a
3817 href="SourceLevelDebugging.html#format_common_intrinsics">LLVM Source Level
3818 Debugging</a> document.
3823 <!-- *********************************************************************** -->
3826 <a href="http://jigsaw.w3.org/css-validator/check/referer"><img
3827 src="http://jigsaw.w3.org/css-validator/images/vcss" alt="Valid CSS!"></a>
3828 <a href="http://validator.w3.org/check/referer"><img
3829 src="http://www.w3.org/Icons/valid-html401" alt="Valid HTML 4.01!" /></a>
3831 <a href="mailto:sabre@nondot.org">Chris Lattner</a><br>
3832 <a href="http://llvm.org">The LLVM Compiler Infrastructure</a><br>
3833 Last modified: $Date$