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="#memoryops">Memory Access Operations</a>
96 <li><a href="#i_malloc">'<tt>malloc</tt>' Instruction</a></li>
97 <li><a href="#i_free">'<tt>free</tt>' Instruction</a></li>
98 <li><a href="#i_alloca">'<tt>alloca</tt>' Instruction</a></li>
99 <li><a href="#i_load">'<tt>load</tt>' Instruction</a></li>
100 <li><a href="#i_store">'<tt>store</tt>' Instruction</a></li>
101 <li><a href="#i_getelementptr">'<tt>getelementptr</tt>' Instruction</a></li>
104 <li><a href="#otherops">Other Operations</a>
106 <li><a href="#i_phi">'<tt>phi</tt>' Instruction</a></li>
107 <li><a href="#i_cast">'<tt>cast .. to</tt>' Instruction</a></li>
108 <li><a href="#i_select">'<tt>select</tt>' Instruction</a></li>
109 <li><a href="#i_vsetint">'<tt>vsetint</tt>' Instruction</a></li>
110 <li><a href="#i_vsetfp">'<tt>vsetfp</tt>' Instruction</a></li>
111 <li><a href="#i_vselect">'<tt>vselect</tt>' Instruction</a></li>
112 <li><a href="#i_extractelement">'<tt>extractelement</tt>' Instruction</a></li>
113 <li><a href="#i_insertelement">'<tt>insertelement</tt>' Instruction</a></li>
114 <li><a href="#i_call">'<tt>call</tt>' Instruction</a></li>
115 <li><a href="#i_va_arg">'<tt>va_arg</tt>' Instruction</a></li>
120 <li><a href="#intrinsics">Intrinsic Functions</a>
122 <li><a href="#int_varargs">Variable Argument Handling Intrinsics</a>
124 <li><a href="#i_va_start">'<tt>llvm.va_start</tt>' Intrinsic</a></li>
125 <li><a href="#i_va_end">'<tt>llvm.va_end</tt>' Intrinsic</a></li>
126 <li><a href="#i_va_copy">'<tt>llvm.va_copy</tt>' Intrinsic</a></li>
129 <li><a href="#int_gc">Accurate Garbage Collection Intrinsics</a>
131 <li><a href="#i_gcroot">'<tt>llvm.gcroot</tt>' Intrinsic</a></li>
132 <li><a href="#i_gcread">'<tt>llvm.gcread</tt>' Intrinsic</a></li>
133 <li><a href="#i_gcwrite">'<tt>llvm.gcwrite</tt>' Intrinsic</a></li>
136 <li><a href="#int_codegen">Code Generator Intrinsics</a>
138 <li><a href="#i_returnaddress">'<tt>llvm.returnaddress</tt>' Intrinsic</a></li>
139 <li><a href="#i_frameaddress">'<tt>llvm.frameaddress</tt>' Intrinsic</a></li>
140 <li><a href="#i_stacksave">'<tt>llvm.stacksave</tt>' Intrinsic</a></li>
141 <li><a href="#i_stackrestore">'<tt>llvm.stackrestore</tt>' Intrinsic</a></li>
142 <li><a href="#i_prefetch">'<tt>llvm.prefetch</tt>' Intrinsic</a></li>
143 <li><a href="#i_pcmarker">'<tt>llvm.pcmarker</tt>' Intrinsic</a></li>
144 <li><a href="#i_readcyclecounter"><tt>llvm.readcyclecounter</tt>' Intrinsic</a></li>
147 <li><a href="#int_libc">Standard C Library Intrinsics</a>
149 <li><a href="#i_memcpy">'<tt>llvm.memcpy.*</tt>' Intrinsic</a></li>
150 <li><a href="#i_memmove">'<tt>llvm.memmove.*</tt>' Intrinsic</a></li>
151 <li><a href="#i_memset">'<tt>llvm.memset.*</tt>' Intrinsic</a></li>
152 <li><a href="#i_isunordered">'<tt>llvm.isunordered.*</tt>' Intrinsic</a></li>
153 <li><a href="#i_sqrt">'<tt>llvm.sqrt.*</tt>' Intrinsic</a></li>
157 <li><a href="#int_manip">Bit Manipulation Intrinsics</a>
159 <li><a href="#i_bswap">'<tt>llvm.bswap.*</tt>' Intrinsics</a></li>
160 <li><a href="#int_ctpop">'<tt>llvm.ctpop.*</tt>' Intrinsic </a></li>
161 <li><a href="#int_ctlz">'<tt>llvm.ctlz.*</tt>' Intrinsic </a></li>
162 <li><a href="#int_cttz">'<tt>llvm.cttz.*</tt>' Intrinsic </a></li>
165 <li><a href="#int_debugger">Debugger intrinsics</a></li>
170 <div class="doc_author">
171 <p>Written by <a href="mailto:sabre@nondot.org">Chris Lattner</a>
172 and <a href="mailto:vadve@cs.uiuc.edu">Vikram Adve</a></p>
175 <!-- *********************************************************************** -->
176 <div class="doc_section"> <a name="abstract">Abstract </a></div>
177 <!-- *********************************************************************** -->
179 <div class="doc_text">
180 <p>This document is a reference manual for the LLVM assembly language.
181 LLVM is an SSA based representation that provides type safety,
182 low-level operations, flexibility, and the capability of representing
183 'all' high-level languages cleanly. It is the common code
184 representation used throughout all phases of the LLVM compilation
188 <!-- *********************************************************************** -->
189 <div class="doc_section"> <a name="introduction">Introduction</a> </div>
190 <!-- *********************************************************************** -->
192 <div class="doc_text">
194 <p>The LLVM code representation is designed to be used in three
195 different forms: as an in-memory compiler IR, as an on-disk bytecode
196 representation (suitable for fast loading by a Just-In-Time compiler),
197 and as a human readable assembly language representation. This allows
198 LLVM to provide a powerful intermediate representation for efficient
199 compiler transformations and analysis, while providing a natural means
200 to debug and visualize the transformations. The three different forms
201 of LLVM are all equivalent. This document describes the human readable
202 representation and notation.</p>
204 <p>The LLVM representation aims to be light-weight and low-level
205 while being expressive, typed, and extensible at the same time. It
206 aims to be a "universal IR" of sorts, by being at a low enough level
207 that high-level ideas may be cleanly mapped to it (similar to how
208 microprocessors are "universal IR's", allowing many source languages to
209 be mapped to them). By providing type information, LLVM can be used as
210 the target of optimizations: for example, through pointer analysis, it
211 can be proven that a C automatic variable is never accessed outside of
212 the current function... allowing it to be promoted to a simple SSA
213 value instead of a memory location.</p>
217 <!-- _______________________________________________________________________ -->
218 <div class="doc_subsubsection"> <a name="wellformed">Well-Formedness</a> </div>
220 <div class="doc_text">
222 <p>It is important to note that this document describes 'well formed'
223 LLVM assembly language. There is a difference between what the parser
224 accepts and what is considered 'well formed'. For example, the
225 following instruction is syntactically okay, but not well formed:</p>
228 %x = <a href="#i_add">add</a> int 1, %x
231 <p>...because the definition of <tt>%x</tt> does not dominate all of
232 its uses. The LLVM infrastructure provides a verification pass that may
233 be used to verify that an LLVM module is well formed. This pass is
234 automatically run by the parser after parsing input assembly and by
235 the optimizer before it outputs bytecode. The violations pointed out
236 by the verifier pass indicate bugs in transformation passes or input to
239 <!-- Describe the typesetting conventions here. --> </div>
241 <!-- *********************************************************************** -->
242 <div class="doc_section"> <a name="identifiers">Identifiers</a> </div>
243 <!-- *********************************************************************** -->
245 <div class="doc_text">
247 <p>LLVM uses three different forms of identifiers, for different
251 <li>Named values are represented as a string of characters with a '%' prefix.
252 For example, %foo, %DivisionByZero, %a.really.long.identifier. The actual
253 regular expression used is '<tt>%[a-zA-Z$._][a-zA-Z$._0-9]*</tt>'.
254 Identifiers which require other characters in their names can be surrounded
255 with quotes. In this way, anything except a <tt>"</tt> character can be used
258 <li>Unnamed values are represented as an unsigned numeric value with a '%'
259 prefix. For example, %12, %2, %44.</li>
261 <li>Constants, which are described in a <a href="#constants">section about
262 constants</a>, below.</li>
265 <p>LLVM requires that values start with a '%' sign for two reasons: Compilers
266 don't need to worry about name clashes with reserved words, and the set of
267 reserved words may be expanded in the future without penalty. Additionally,
268 unnamed identifiers allow a compiler to quickly come up with a temporary
269 variable without having to avoid symbol table conflicts.</p>
271 <p>Reserved words in LLVM are very similar to reserved words in other
272 languages. There are keywords for different opcodes ('<tt><a
273 href="#i_add">add</a></tt>', '<tt><a href="#i_cast">cast</a></tt>', '<tt><a
274 href="#i_ret">ret</a></tt>', etc...), for primitive type names ('<tt><a
275 href="#t_void">void</a></tt>', '<tt><a href="#t_uint">uint</a></tt>', etc...),
276 and others. These reserved words cannot conflict with variable names, because
277 none of them start with a '%' character.</p>
279 <p>Here is an example of LLVM code to multiply the integer variable
280 '<tt>%X</tt>' by 8:</p>
285 %result = <a href="#i_mul">mul</a> uint %X, 8
288 <p>After strength reduction:</p>
291 %result = <a href="#i_shl">shl</a> uint %X, ubyte 3
294 <p>And the hard way:</p>
297 <a href="#i_add">add</a> uint %X, %X <i>; yields {uint}:%0</i>
298 <a href="#i_add">add</a> uint %0, %0 <i>; yields {uint}:%1</i>
299 %result = <a href="#i_add">add</a> uint %1, %1
302 <p>This last way of multiplying <tt>%X</tt> by 8 illustrates several
303 important lexical features of LLVM:</p>
307 <li>Comments are delimited with a '<tt>;</tt>' and go until the end of
310 <li>Unnamed temporaries are created when the result of a computation is not
311 assigned to a named value.</li>
313 <li>Unnamed temporaries are numbered sequentially</li>
317 <p>...and it also shows a convention that we follow in this document. When
318 demonstrating instructions, we will follow an instruction with a comment that
319 defines the type and name of value produced. Comments are shown in italic
324 <!-- *********************************************************************** -->
325 <div class="doc_section"> <a name="highlevel">High Level Structure</a> </div>
326 <!-- *********************************************************************** -->
328 <!-- ======================================================================= -->
329 <div class="doc_subsection"> <a name="modulestructure">Module Structure</a>
332 <div class="doc_text">
334 <p>LLVM programs are composed of "Module"s, each of which is a
335 translation unit of the input programs. Each module consists of
336 functions, global variables, and symbol table entries. Modules may be
337 combined together with the LLVM linker, which merges function (and
338 global variable) definitions, resolves forward declarations, and merges
339 symbol table entries. Here is an example of the "hello world" module:</p>
341 <pre><i>; Declare the string constant as a global constant...</i>
342 <a href="#identifiers">%.LC0</a> = <a href="#linkage_internal">internal</a> <a
343 href="#globalvars">constant</a> <a href="#t_array">[13 x sbyte]</a> c"hello world\0A\00" <i>; [13 x sbyte]*</i>
345 <i>; External declaration of the puts function</i>
346 <a href="#functionstructure">declare</a> int %puts(sbyte*) <i>; int(sbyte*)* </i>
348 <i>; Definition of main function</i>
349 int %main() { <i>; int()* </i>
350 <i>; Convert [13x sbyte]* to sbyte *...</i>
352 href="#i_getelementptr">getelementptr</a> [13 x sbyte]* %.LC0, long 0, long 0 <i>; sbyte*</i>
354 <i>; Call puts function to write out the string to stdout...</i>
356 href="#i_call">call</a> int %puts(sbyte* %cast210) <i>; int</i>
358 href="#i_ret">ret</a> int 0<br>}<br></pre>
360 <p>This example is made up of a <a href="#globalvars">global variable</a>
361 named "<tt>.LC0</tt>", an external declaration of the "<tt>puts</tt>"
362 function, and a <a href="#functionstructure">function definition</a>
363 for "<tt>main</tt>".</p>
365 <p>In general, a module is made up of a list of global values,
366 where both functions and global variables are global values. Global values are
367 represented by a pointer to a memory location (in this case, a pointer to an
368 array of char, and a pointer to a function), and have one of the following <a
369 href="#linkage">linkage types</a>.</p>
373 <!-- ======================================================================= -->
374 <div class="doc_subsection">
375 <a name="linkage">Linkage Types</a>
378 <div class="doc_text">
381 All Global Variables and Functions have one of the following types of linkage:
386 <dt><tt><b><a name="linkage_internal">internal</a></b></tt> </dt>
388 <dd>Global values with internal linkage are only directly accessible by
389 objects in the current module. In particular, linking code into a module with
390 an internal global value may cause the internal to be renamed as necessary to
391 avoid collisions. Because the symbol is internal to the module, all
392 references can be updated. This corresponds to the notion of the
393 '<tt>static</tt>' keyword in C, or the idea of "anonymous namespaces" in C++.
396 <dt><tt><b><a name="linkage_linkonce">linkonce</a></b></tt>: </dt>
398 <dd>"<tt>linkonce</tt>" linkage is similar to <tt>internal</tt> linkage, with
399 the twist that linking together two modules defining the same
400 <tt>linkonce</tt> globals will cause one of the globals to be discarded. This
401 is typically used to implement inline functions. Unreferenced
402 <tt>linkonce</tt> globals are allowed to be discarded.
405 <dt><tt><b><a name="linkage_weak">weak</a></b></tt>: </dt>
407 <dd>"<tt>weak</tt>" linkage is exactly the same as <tt>linkonce</tt> linkage,
408 except that unreferenced <tt>weak</tt> globals may not be discarded. This is
409 used to implement constructs in C such as "<tt>int X;</tt>" at global scope.
412 <dt><tt><b><a name="linkage_appending">appending</a></b></tt>: </dt>
414 <dd>"<tt>appending</tt>" linkage may only be applied to global variables of
415 pointer to array type. When two global variables with appending linkage are
416 linked together, the two global arrays are appended together. This is the
417 LLVM, typesafe, equivalent of having the system linker append together
418 "sections" with identical names when .o files are linked.
421 <dt><tt><b><a name="linkage_external">externally visible</a></b></tt>:</dt>
423 <dd>If none of the above identifiers are used, the global is externally
424 visible, meaning that it participates in linkage and can be used to resolve
425 external symbol references.
429 <p><a name="linkage_external">For example, since the "<tt>.LC0</tt>"
430 variable is defined to be internal, if another module defined a "<tt>.LC0</tt>"
431 variable and was linked with this one, one of the two would be renamed,
432 preventing a collision. Since "<tt>main</tt>" and "<tt>puts</tt>" are
433 external (i.e., lacking any linkage declarations), they are accessible
434 outside of the current module. It is illegal for a function <i>declaration</i>
435 to have any linkage type other than "externally visible".</a></p>
439 <!-- ======================================================================= -->
440 <div class="doc_subsection">
441 <a name="callingconv">Calling Conventions</a>
444 <div class="doc_text">
446 <p>LLVM <a href="#functionstructure">functions</a>, <a href="#i_call">calls</a>
447 and <a href="#i_invoke">invokes</a> can all have an optional calling convention
448 specified for the call. The calling convention of any pair of dynamic
449 caller/callee must match, or the behavior of the program is undefined. The
450 following calling conventions are supported by LLVM, and more may be added in
454 <dt><b>"<tt>ccc</tt>" - The C calling convention</b>:</dt>
456 <dd>This calling convention (the default if no other calling convention is
457 specified) matches the target C calling conventions. This calling convention
458 supports varargs function calls and tolerates some mismatch in the declared
459 prototype and implemented declaration of the function (as does normal C).
462 <dt><b>"<tt>fastcc</tt>" - The fast calling convention</b>:</dt>
464 <dd>This calling convention attempts to make calls as fast as possible
465 (e.g. by passing things in registers). This calling convention allows the
466 target to use whatever tricks it wants to produce fast code for the target,
467 without having to conform to an externally specified ABI. Implementations of
468 this convention should allow arbitrary tail call optimization to be supported.
469 This calling convention does not support varargs and requires the prototype of
470 all callees to exactly match the prototype of the function definition.
473 <dt><b>"<tt>coldcc</tt>" - The cold calling convention</b>:</dt>
475 <dd>This calling convention attempts to make code in the caller as efficient
476 as possible under the assumption that the call is not commonly executed. As
477 such, these calls often preserve all registers so that the call does not break
478 any live ranges in the caller side. This calling convention does not support
479 varargs and requires the prototype of all callees to exactly match the
480 prototype of the function definition.
483 <dt><b>"<tt>cc <<em>n</em>></tt>" - Numbered convention</b>:</dt>
485 <dd>Any calling convention may be specified by number, allowing
486 target-specific calling conventions to be used. Target specific calling
487 conventions start at 64.
491 <p>More calling conventions can be added/defined on an as-needed basis, to
492 support pascal conventions or any other well-known target-independent
497 <!-- ======================================================================= -->
498 <div class="doc_subsection">
499 <a name="globalvars">Global Variables</a>
502 <div class="doc_text">
504 <p>Global variables define regions of memory allocated at compilation time
505 instead of run-time. Global variables may optionally be initialized, may have
506 an explicit section to be placed in, and may
507 have an optional explicit alignment specified. A
508 variable may be defined as a global "constant," which indicates that the
509 contents of the variable will <b>never</b> be modified (enabling better
510 optimization, allowing the global data to be placed in the read-only section of
511 an executable, etc). Note that variables that need runtime initialization
512 cannot be marked "constant" as there is a store to the variable.</p>
515 LLVM explicitly allows <em>declarations</em> of global variables to be marked
516 constant, even if the final definition of the global is not. This capability
517 can be used to enable slightly better optimization of the program, but requires
518 the language definition to guarantee that optimizations based on the
519 'constantness' are valid for the translation units that do not include the
523 <p>As SSA values, global variables define pointer values that are in
524 scope (i.e. they dominate) all basic blocks in the program. Global
525 variables always define a pointer to their "content" type because they
526 describe a region of memory, and all memory objects in LLVM are
527 accessed through pointers.</p>
529 <p>LLVM allows an explicit section to be specified for globals. If the target
530 supports it, it will emit globals to the section specified.</p>
532 <p>An explicit alignment may be specified for a global. If not present, or if
533 the alignment is set to zero, the alignment of the global is set by the target
534 to whatever it feels convenient. If an explicit alignment is specified, the
535 global is forced to have at least that much alignment. All alignments must be
541 <!-- ======================================================================= -->
542 <div class="doc_subsection">
543 <a name="functionstructure">Functions</a>
546 <div class="doc_text">
548 <p>LLVM function definitions consist of an optional <a href="#linkage">linkage
549 type</a>, an optional <a href="#callingconv">calling convention</a>, a return
550 type, a function name, a (possibly empty) argument list, an optional section,
551 an optional alignment, an opening curly brace,
552 a list of basic blocks, and a closing curly brace. LLVM function declarations
553 are defined with the "<tt>declare</tt>" keyword, an optional <a
554 href="#callingconv">calling convention</a>, a return type, a function name,
555 a possibly empty list of arguments, and an optional alignment.</p>
557 <p>A function definition contains a list of basic blocks, forming the CFG for
558 the function. Each basic block may optionally start with a label (giving the
559 basic block a symbol table entry), contains a list of instructions, and ends
560 with a <a href="#terminators">terminator</a> instruction (such as a branch or
561 function return).</p>
563 <p>The first basic block in a program is special in two ways: it is immediately
564 executed on entrance to the function, and it is not allowed to have predecessor
565 basic blocks (i.e. there can not be any branches to the entry block of a
566 function). Because the block can have no predecessors, it also cannot have any
567 <a href="#i_phi">PHI nodes</a>.</p>
569 <p>LLVM functions are identified by their name and type signature. Hence, two
570 functions with the same name but different parameter lists or return values are
571 considered different functions, and LLVM will resolve references to each
574 <p>LLVM allows an explicit section to be specified for functions. If the target
575 supports it, it will emit functions to the section specified.</p>
577 <p>An explicit alignment may be specified for a function. If not present, or if
578 the alignment is set to zero, the alignment of the function is set by the target
579 to whatever it feels convenient. If an explicit alignment is specified, the
580 function is forced to have at least that much alignment. All alignments must be
585 <!-- ======================================================================= -->
586 <div class="doc_subsection">
587 <a name="moduleasm">Module-Level Inline Assembly</a></li>
590 <div class="doc_text">
592 Modules may contain "module-level inline asm" blocks, which corresponds to the
593 GCC "file scope inline asm" blocks. These blocks are internally concatenated by
594 LLVM and treated as a single unit, but may be separated in the .ll file if
595 desired. The syntax is very simple:
598 <div class="doc_code"><pre>
599 module asm "inline asm code goes here"
600 module asm "more can go here"
603 <p>The strings can contain any character by escaping non-printable characters.
604 The escape sequence used is simply "\xx" where "xx" is the two digit hex code
609 The inline asm code is simply printed to the machine code .s file when
610 assembly code is generated.
615 <!-- *********************************************************************** -->
616 <div class="doc_section"> <a name="typesystem">Type System</a> </div>
617 <!-- *********************************************************************** -->
619 <div class="doc_text">
621 <p>The LLVM type system is one of the most important features of the
622 intermediate representation. Being typed enables a number of
623 optimizations to be performed on the IR directly, without having to do
624 extra analyses on the side before the transformation. A strong type
625 system makes it easier to read the generated code and enables novel
626 analyses and transformations that are not feasible to perform on normal
627 three address code representations.</p>
631 <!-- ======================================================================= -->
632 <div class="doc_subsection"> <a name="t_primitive">Primitive Types</a> </div>
633 <div class="doc_text">
634 <p>The primitive types are the fundamental building blocks of the LLVM
635 system. The current set of primitive types is as follows:</p>
637 <table class="layout">
642 <tr><th>Type</th><th>Description</th></tr>
643 <tr><td><tt>void</tt></td><td>No value</td></tr>
644 <tr><td><tt>ubyte</tt></td><td>Unsigned 8-bit value</td></tr>
645 <tr><td><tt>ushort</tt></td><td>Unsigned 16-bit value</td></tr>
646 <tr><td><tt>uint</tt></td><td>Unsigned 32-bit value</td></tr>
647 <tr><td><tt>ulong</tt></td><td>Unsigned 64-bit value</td></tr>
648 <tr><td><tt>float</tt></td><td>32-bit floating point value</td></tr>
649 <tr><td><tt>label</tt></td><td>Branch destination</td></tr>
656 <tr><th>Type</th><th>Description</th></tr>
657 <tr><td><tt>bool</tt></td><td>True or False value</td></tr>
658 <tr><td><tt>sbyte</tt></td><td>Signed 8-bit value</td></tr>
659 <tr><td><tt>short</tt></td><td>Signed 16-bit value</td></tr>
660 <tr><td><tt>int</tt></td><td>Signed 32-bit value</td></tr>
661 <tr><td><tt>long</tt></td><td>Signed 64-bit value</td></tr>
662 <tr><td><tt>double</tt></td><td>64-bit floating point value</td></tr>
670 <!-- _______________________________________________________________________ -->
671 <div class="doc_subsubsection"> <a name="t_classifications">Type
672 Classifications</a> </div>
673 <div class="doc_text">
674 <p>These different primitive types fall into a few useful
677 <table border="1" cellspacing="0" cellpadding="4">
679 <tr><th>Classification</th><th>Types</th></tr>
681 <td><a name="t_signed">signed</a></td>
682 <td><tt>sbyte, short, int, long, float, double</tt></td>
685 <td><a name="t_unsigned">unsigned</a></td>
686 <td><tt>ubyte, ushort, uint, ulong</tt></td>
689 <td><a name="t_integer">integer</a></td>
690 <td><tt>ubyte, sbyte, ushort, short, uint, int, ulong, long</tt></td>
693 <td><a name="t_integral">integral</a></td>
694 <td><tt>bool, ubyte, sbyte, ushort, short, uint, int, ulong, long</tt>
698 <td><a name="t_floating">floating point</a></td>
699 <td><tt>float, double</tt></td>
702 <td><a name="t_firstclass">first class</a></td>
703 <td><tt>bool, ubyte, sbyte, ushort, short, uint, int, ulong, long,<br>
704 float, double, <a href="#t_pointer">pointer</a>,
705 <a href="#t_packed">packed</a></tt></td>
710 <p>The <a href="#t_firstclass">first class</a> types are perhaps the
711 most important. Values of these types are the only ones which can be
712 produced by instructions, passed as arguments, or used as operands to
713 instructions. This means that all structures and arrays must be
714 manipulated either by pointer or by component.</p>
717 <!-- ======================================================================= -->
718 <div class="doc_subsection"> <a name="t_derived">Derived Types</a> </div>
720 <div class="doc_text">
722 <p>The real power in LLVM comes from the derived types in the system.
723 This is what allows a programmer to represent arrays, functions,
724 pointers, and other useful types. Note that these derived types may be
725 recursive: For example, it is possible to have a two dimensional array.</p>
729 <!-- _______________________________________________________________________ -->
730 <div class="doc_subsubsection"> <a name="t_array">Array Type</a> </div>
732 <div class="doc_text">
736 <p>The array type is a very simple derived type that arranges elements
737 sequentially in memory. The array type requires a size (number of
738 elements) and an underlying data type.</p>
743 [<# elements> x <elementtype>]
746 <p>The number of elements is a constant integer value; elementtype may
747 be any type with a size.</p>
750 <table class="layout">
753 <tt>[40 x int ]</tt><br/>
754 <tt>[41 x int ]</tt><br/>
755 <tt>[40 x uint]</tt><br/>
758 Array of 40 integer values.<br/>
759 Array of 41 integer values.<br/>
760 Array of 40 unsigned integer values.<br/>
764 <p>Here are some examples of multidimensional arrays:</p>
765 <table class="layout">
768 <tt>[3 x [4 x int]]</tt><br/>
769 <tt>[12 x [10 x float]]</tt><br/>
770 <tt>[2 x [3 x [4 x uint]]]</tt><br/>
773 3x4 array of integer values.<br/>
774 12x10 array of single precision floating point values.<br/>
775 2x3x4 array of unsigned integer values.<br/>
780 <p>Note that 'variable sized arrays' can be implemented in LLVM with a zero
781 length array. Normally, accesses past the end of an array are undefined in
782 LLVM (e.g. it is illegal to access the 5th element of a 3 element array).
783 As a special case, however, zero length arrays are recognized to be variable
784 length. This allows implementation of 'pascal style arrays' with the LLVM
785 type "{ int, [0 x float]}", for example.</p>
789 <!-- _______________________________________________________________________ -->
790 <div class="doc_subsubsection"> <a name="t_function">Function Type</a> </div>
791 <div class="doc_text">
793 <p>The function type can be thought of as a function signature. It
794 consists of a return type and a list of formal parameter types.
795 Function types are usually used to build virtual function tables
796 (which are structures of pointers to functions), for indirect function
797 calls, and when defining a function.</p>
799 The return type of a function type cannot be an aggregate type.
802 <pre> <returntype> (<parameter list>)<br></pre>
803 <p>...where '<tt><parameter list></tt>' is a comma-separated list of type
804 specifiers. Optionally, the parameter list may include a type <tt>...</tt>,
805 which indicates that the function takes a variable number of arguments.
806 Variable argument functions can access their arguments with the <a
807 href="#int_varargs">variable argument handling intrinsic</a> functions.</p>
809 <table class="layout">
812 <tt>int (int)</tt> <br/>
813 <tt>float (int, int *) *</tt><br/>
814 <tt>int (sbyte *, ...)</tt><br/>
817 function taking an <tt>int</tt>, returning an <tt>int</tt><br/>
818 <a href="#t_pointer">Pointer</a> to a function that takes an
819 <tt>int</tt> and a <a href="#t_pointer">pointer</a> to <tt>int</tt>,
820 returning <tt>float</tt>.<br/>
821 A vararg function that takes at least one <a href="#t_pointer">pointer</a>
822 to <tt>sbyte</tt> (signed char in C), which returns an integer. This is
823 the signature for <tt>printf</tt> in LLVM.<br/>
829 <!-- _______________________________________________________________________ -->
830 <div class="doc_subsubsection"> <a name="t_struct">Structure Type</a> </div>
831 <div class="doc_text">
833 <p>The structure type is used to represent a collection of data members
834 together in memory. The packing of the field types is defined to match
835 the ABI of the underlying processor. The elements of a structure may
836 be any type that has a size.</p>
837 <p>Structures are accessed using '<tt><a href="#i_load">load</a></tt>
838 and '<tt><a href="#i_store">store</a></tt>' by getting a pointer to a
839 field with the '<tt><a href="#i_getelementptr">getelementptr</a></tt>'
842 <pre> { <type list> }<br></pre>
844 <table class="layout">
847 <tt>{ int, int, int }</tt><br/>
848 <tt>{ float, int (int) * }</tt><br/>
851 a triple of three <tt>int</tt> values<br/>
852 A pair, where the first element is a <tt>float</tt> and the second element
853 is a <a href="#t_pointer">pointer</a> to a <a href="#t_function">function</a>
854 that takes an <tt>int</tt>, returning an <tt>int</tt>.<br/>
860 <!-- _______________________________________________________________________ -->
861 <div class="doc_subsubsection"> <a name="t_pointer">Pointer Type</a> </div>
862 <div class="doc_text">
864 <p>As in many languages, the pointer type represents a pointer or
865 reference to another object, which must live in memory.</p>
867 <pre> <type> *<br></pre>
869 <table class="layout">
872 <tt>[4x int]*</tt><br/>
873 <tt>int (int *) *</tt><br/>
876 A <a href="#t_pointer">pointer</a> to <a href="#t_array">array</a> of
877 four <tt>int</tt> values<br/>
878 A <a href="#t_pointer">pointer</a> to a <a
879 href="#t_function">function</a> that takes an <tt>int*</tt>, returning an
886 <!-- _______________________________________________________________________ -->
887 <div class="doc_subsubsection"> <a name="t_packed">Packed Type</a> </div>
888 <div class="doc_text">
892 <p>A packed type is a simple derived type that represents a vector
893 of elements. Packed types are used when multiple primitive data
894 are operated in parallel using a single instruction (SIMD).
895 A packed type requires a size (number of
896 elements) and an underlying primitive data type. Vectors must have a power
897 of two length (1, 2, 4, 8, 16 ...). Packed types are
898 considered <a href="#t_firstclass">first class</a>.</p>
903 < <# elements> x <elementtype> >
906 <p>The number of elements is a constant integer value; elementtype may
907 be any integral or floating point type.</p>
911 <table class="layout">
914 <tt><4 x int></tt><br/>
915 <tt><8 x float></tt><br/>
916 <tt><2 x uint></tt><br/>
919 Packed vector of 4 integer values.<br/>
920 Packed vector of 8 floating-point values.<br/>
921 Packed vector of 2 unsigned integer values.<br/>
927 <!-- _______________________________________________________________________ -->
928 <div class="doc_subsubsection"> <a name="t_opaque">Opaque Type</a> </div>
929 <div class="doc_text">
933 <p>Opaque types are used to represent unknown types in the system. This
934 corresponds (for example) to the C notion of a foward declared structure type.
935 In LLVM, opaque types can eventually be resolved to any type (not just a
946 <table class="layout">
959 <!-- *********************************************************************** -->
960 <div class="doc_section"> <a name="constants">Constants</a> </div>
961 <!-- *********************************************************************** -->
963 <div class="doc_text">
965 <p>LLVM has several different basic types of constants. This section describes
966 them all and their syntax.</p>
970 <!-- ======================================================================= -->
971 <div class="doc_subsection"><a name="simpleconstants">Simple Constants</a></div>
973 <div class="doc_text">
976 <dt><b>Boolean constants</b></dt>
978 <dd>The two strings '<tt>true</tt>' and '<tt>false</tt>' are both valid
979 constants of the <tt><a href="#t_primitive">bool</a></tt> type.
982 <dt><b>Integer constants</b></dt>
984 <dd>Standard integers (such as '4') are constants of the <a
985 href="#t_integer">integer</a> type. Negative numbers may be used with signed
989 <dt><b>Floating point constants</b></dt>
991 <dd>Floating point constants use standard decimal notation (e.g. 123.421),
992 exponential notation (e.g. 1.23421e+2), or a more precise hexadecimal
993 notation (see below). Floating point constants must have a <a
994 href="#t_floating">floating point</a> type. </dd>
996 <dt><b>Null pointer constants</b></dt>
998 <dd>The identifier '<tt>null</tt>' is recognized as a null pointer constant
999 and must be of <a href="#t_pointer">pointer type</a>.</dd>
1003 <p>The one non-intuitive notation for constants is the optional hexadecimal form
1004 of floating point constants. For example, the form '<tt>double
1005 0x432ff973cafa8000</tt>' is equivalent to (but harder to read than) '<tt>double
1006 4.5e+15</tt>'. The only time hexadecimal floating point constants are required
1007 (and the only time that they are generated by the disassembler) is when a
1008 floating point constant must be emitted but it cannot be represented as a
1009 decimal floating point number. For example, NaN's, infinities, and other
1010 special values are represented in their IEEE hexadecimal format so that
1011 assembly and disassembly do not cause any bits to change in the constants.</p>
1015 <!-- ======================================================================= -->
1016 <div class="doc_subsection"><a name="aggregateconstants">Aggregate Constants</a>
1019 <div class="doc_text">
1020 <p>Aggregate constants arise from aggregation of simple constants
1021 and smaller aggregate constants.</p>
1024 <dt><b>Structure constants</b></dt>
1026 <dd>Structure constants are represented with notation similar to structure
1027 type definitions (a comma separated list of elements, surrounded by braces
1028 (<tt>{}</tt>)). For example: "<tt>{ int 4, float 17.0, int* %G }</tt>",
1029 where "<tt>%G</tt>" is declared as "<tt>%G = external global int</tt>". Structure constants
1030 must have <a href="#t_struct">structure type</a>, and the number and
1031 types of elements must match those specified by the type.
1034 <dt><b>Array constants</b></dt>
1036 <dd>Array constants are represented with notation similar to array type
1037 definitions (a comma separated list of elements, surrounded by square brackets
1038 (<tt>[]</tt>)). For example: "<tt>[ int 42, int 11, int 74 ]</tt>". Array
1039 constants must have <a href="#t_array">array type</a>, and the number and
1040 types of elements must match those specified by the type.
1043 <dt><b>Packed constants</b></dt>
1045 <dd>Packed constants are represented with notation similar to packed type
1046 definitions (a comma separated list of elements, surrounded by
1047 less-than/greater-than's (<tt><></tt>)). For example: "<tt>< int 42,
1048 int 11, int 74, int 100 ></tt>". Packed constants must have <a
1049 href="#t_packed">packed type</a>, and the number and types of elements must
1050 match those specified by the type.
1053 <dt><b>Zero initialization</b></dt>
1055 <dd>The string '<tt>zeroinitializer</tt>' can be used to zero initialize a
1056 value to zero of <em>any</em> type, including scalar and aggregate types.
1057 This is often used to avoid having to print large zero initializers (e.g. for
1058 large arrays) and is always exactly equivalent to using explicit zero
1065 <!-- ======================================================================= -->
1066 <div class="doc_subsection">
1067 <a name="globalconstants">Global Variable and Function Addresses</a>
1070 <div class="doc_text">
1072 <p>The addresses of <a href="#globalvars">global variables</a> and <a
1073 href="#functionstructure">functions</a> are always implicitly valid (link-time)
1074 constants. These constants are explicitly referenced when the <a
1075 href="#identifiers">identifier for the global</a> is used and always have <a
1076 href="#t_pointer">pointer</a> type. For example, the following is a legal LLVM
1082 %Z = global [2 x int*] [ int* %X, int* %Y ]
1087 <!-- ======================================================================= -->
1088 <div class="doc_subsection"><a name="undefvalues">Undefined Values</a></div>
1089 <div class="doc_text">
1090 <p>The string '<tt>undef</tt>' is recognized as a type-less constant that has
1091 no specific value. Undefined values may be of any type and be used anywhere
1092 a constant is permitted.</p>
1094 <p>Undefined values indicate to the compiler that the program is well defined
1095 no matter what value is used, giving the compiler more freedom to optimize.
1099 <!-- ======================================================================= -->
1100 <div class="doc_subsection"><a name="constantexprs">Constant Expressions</a>
1103 <div class="doc_text">
1105 <p>Constant expressions are used to allow expressions involving other constants
1106 to be used as constants. Constant expressions may be of any <a
1107 href="#t_firstclass">first class</a> type and may involve any LLVM operation
1108 that does not have side effects (e.g. load and call are not supported). The
1109 following is the syntax for constant expressions:</p>
1112 <dt><b><tt>cast ( CST to TYPE )</tt></b></dt>
1114 <dd>Cast a constant to another type.</dd>
1116 <dt><b><tt>getelementptr ( CSTPTR, IDX0, IDX1, ... )</tt></b></dt>
1118 <dd>Perform the <a href="#i_getelementptr">getelementptr operation</a> on
1119 constants. As with the <a href="#i_getelementptr">getelementptr</a>
1120 instruction, the index list may have zero or more indexes, which are required
1121 to make sense for the type of "CSTPTR".</dd>
1123 <dt><b><tt>select ( COND, VAL1, VAL2 )</tt></b></dt>
1125 <dd>Perform the <a href="#i_select">select operation</a> on
1128 <dt><b><tt>extractelement ( VAL, IDX )</tt></b></dt>
1130 <dd>Perform the <a href="#i_extractelement">extractelement
1131 operation</a> on constants.
1133 <dt><b><tt>insertelement ( VAL, ELT, IDX )</tt></b></dt>
1135 <dd>Perform the <a href="#i_insertelement">insertelement
1136 operation</a> on constants.
1138 <dt><b><tt>OPCODE ( LHS, RHS )</tt></b></dt>
1140 <dd>Perform the specified operation of the LHS and RHS constants. OPCODE may
1141 be any of the <a href="#binaryops">binary</a> or <a href="#bitwiseops">bitwise
1142 binary</a> operations. The constraints on operands are the same as those for
1143 the corresponding instruction (e.g. no bitwise operations on floating point
1144 values are allowed).</dd>
1148 <!-- *********************************************************************** -->
1149 <div class="doc_section"> <a name="othervalues">Other Values</a> </div>
1150 <!-- *********************************************************************** -->
1152 <!-- ======================================================================= -->
1153 <div class="doc_subsection">
1154 <a name="inlineasm">Inline Assembler Expressions</a>
1157 <div class="doc_text">
1160 LLVM supports inline assembler expressions (as opposed to <a href="#moduleasm">
1161 Module-Level Inline Assembly</a>) through the use of a special value. This
1162 value represents the inline assembler as a string (containing the instructions
1163 to emit), a list of operand constraints (stored as a string), and a flag that
1164 indicates whether or not the inline asm expression has side effects. An example
1165 inline assembler expression is:
1169 int(int) asm "bswap $0", "=r,r"
1173 Inline assembler expressions may <b>only</b> be used as the callee operand of
1174 a <a href="#i_call"><tt>call</tt> instruction</a>. Thus, typically we have:
1178 %X = call int asm "<a href="#i_bswap">bswap</a> $0", "=r,r"(int %Y)
1182 Inline asms with side effects not visible in the constraint list must be marked
1183 as having side effects. This is done through the use of the
1184 '<tt>sideeffect</tt>' keyword, like so:
1188 call void asm sideeffect "eieio", ""()
1191 <p>TODO: The format of the asm and constraints string still need to be
1192 documented here. Constraints on what can be done (e.g. duplication, moving, etc
1193 need to be documented).
1198 <!-- *********************************************************************** -->
1199 <div class="doc_section"> <a name="instref">Instruction Reference</a> </div>
1200 <!-- *********************************************************************** -->
1202 <div class="doc_text">
1204 <p>The LLVM instruction set consists of several different
1205 classifications of instructions: <a href="#terminators">terminator
1206 instructions</a>, <a href="#binaryops">binary instructions</a>,
1207 <a href="#bitwiseops">bitwise binary instructions</a>, <a
1208 href="#memoryops">memory instructions</a>, and <a href="#otherops">other
1209 instructions</a>.</p>
1213 <!-- ======================================================================= -->
1214 <div class="doc_subsection"> <a name="terminators">Terminator
1215 Instructions</a> </div>
1217 <div class="doc_text">
1219 <p>As mentioned <a href="#functionstructure">previously</a>, every
1220 basic block in a program ends with a "Terminator" instruction, which
1221 indicates which block should be executed after the current block is
1222 finished. These terminator instructions typically yield a '<tt>void</tt>'
1223 value: they produce control flow, not values (the one exception being
1224 the '<a href="#i_invoke"><tt>invoke</tt></a>' instruction).</p>
1225 <p>There are six different terminator instructions: the '<a
1226 href="#i_ret"><tt>ret</tt></a>' instruction, the '<a href="#i_br"><tt>br</tt></a>'
1227 instruction, the '<a href="#i_switch"><tt>switch</tt></a>' instruction,
1228 the '<a href="#i_invoke"><tt>invoke</tt></a>' instruction, the '<a
1229 href="#i_unwind"><tt>unwind</tt></a>' instruction, and the '<a
1230 href="#i_unreachable"><tt>unreachable</tt></a>' instruction.</p>
1234 <!-- _______________________________________________________________________ -->
1235 <div class="doc_subsubsection"> <a name="i_ret">'<tt>ret</tt>'
1236 Instruction</a> </div>
1237 <div class="doc_text">
1239 <pre> ret <type> <value> <i>; Return a value from a non-void function</i>
1240 ret void <i>; Return from void function</i>
1243 <p>The '<tt>ret</tt>' instruction is used to return control flow (and a
1244 value) from a function back to the caller.</p>
1245 <p>There are two forms of the '<tt>ret</tt>' instruction: one that
1246 returns a value and then causes control flow, and one that just causes
1247 control flow to occur.</p>
1249 <p>The '<tt>ret</tt>' instruction may return any '<a
1250 href="#t_firstclass">first class</a>' type. Notice that a function is
1251 not <a href="#wellformed">well formed</a> if there exists a '<tt>ret</tt>'
1252 instruction inside of the function that returns a value that does not
1253 match the return type of the function.</p>
1255 <p>When the '<tt>ret</tt>' instruction is executed, control flow
1256 returns back to the calling function's context. If the caller is a "<a
1257 href="#i_call"><tt>call</tt></a>" instruction, execution continues at
1258 the instruction after the call. If the caller was an "<a
1259 href="#i_invoke"><tt>invoke</tt></a>" instruction, execution continues
1260 at the beginning of the "normal" destination block. If the instruction
1261 returns a value, that value shall set the call or invoke instruction's
1264 <pre> ret int 5 <i>; Return an integer value of 5</i>
1265 ret void <i>; Return from a void function</i>
1268 <!-- _______________________________________________________________________ -->
1269 <div class="doc_subsubsection"> <a name="i_br">'<tt>br</tt>' Instruction</a> </div>
1270 <div class="doc_text">
1272 <pre> br bool <cond>, label <iftrue>, label <iffalse><br> br label <dest> <i>; Unconditional branch</i>
1275 <p>The '<tt>br</tt>' instruction is used to cause control flow to
1276 transfer to a different basic block in the current function. There are
1277 two forms of this instruction, corresponding to a conditional branch
1278 and an unconditional branch.</p>
1280 <p>The conditional branch form of the '<tt>br</tt>' instruction takes a
1281 single '<tt>bool</tt>' value and two '<tt>label</tt>' values. The
1282 unconditional form of the '<tt>br</tt>' instruction takes a single '<tt>label</tt>'
1283 value as a target.</p>
1285 <p>Upon execution of a conditional '<tt>br</tt>' instruction, the '<tt>bool</tt>'
1286 argument is evaluated. If the value is <tt>true</tt>, control flows
1287 to the '<tt>iftrue</tt>' <tt>label</tt> argument. If "cond" is <tt>false</tt>,
1288 control flows to the '<tt>iffalse</tt>' <tt>label</tt> argument.</p>
1290 <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
1291 href="#i_ret">ret</a> int 1<br>IfUnequal:<br> <a href="#i_ret">ret</a> int 0<br></pre>
1293 <!-- _______________________________________________________________________ -->
1294 <div class="doc_subsubsection">
1295 <a name="i_switch">'<tt>switch</tt>' Instruction</a>
1298 <div class="doc_text">
1302 switch <intty> <value>, label <defaultdest> [ <intty> <val>, label <dest> ... ]
1307 <p>The '<tt>switch</tt>' instruction is used to transfer control flow to one of
1308 several different places. It is a generalization of the '<tt>br</tt>'
1309 instruction, allowing a branch to occur to one of many possible
1315 <p>The '<tt>switch</tt>' instruction uses three parameters: an integer
1316 comparison value '<tt>value</tt>', a default '<tt>label</tt>' destination, and
1317 an array of pairs of comparison value constants and '<tt>label</tt>'s. The
1318 table is not allowed to contain duplicate constant entries.</p>
1322 <p>The <tt>switch</tt> instruction specifies a table of values and
1323 destinations. When the '<tt>switch</tt>' instruction is executed, this
1324 table is searched for the given value. If the value is found, control flow is
1325 transfered to the corresponding destination; otherwise, control flow is
1326 transfered to the default destination.</p>
1328 <h5>Implementation:</h5>
1330 <p>Depending on properties of the target machine and the particular
1331 <tt>switch</tt> instruction, this instruction may be code generated in different
1332 ways. For example, it could be generated as a series of chained conditional
1333 branches or with a lookup table.</p>
1338 <i>; Emulate a conditional br instruction</i>
1339 %Val = <a href="#i_cast">cast</a> bool %value to int
1340 switch int %Val, label %truedest [int 0, label %falsedest ]
1342 <i>; Emulate an unconditional br instruction</i>
1343 switch uint 0, label %dest [ ]
1345 <i>; Implement a jump table:</i>
1346 switch uint %val, label %otherwise [ uint 0, label %onzero
1347 uint 1, label %onone
1348 uint 2, label %ontwo ]
1352 <!-- _______________________________________________________________________ -->
1353 <div class="doc_subsubsection">
1354 <a name="i_invoke">'<tt>invoke</tt>' Instruction</a>
1357 <div class="doc_text">
1362 <result> = invoke [<a href="#callingconv">cconv</a>] <ptr to function ty> %<function ptr val>(<function args>)
1363 to label <normal label> except label <exception label>
1368 <p>The '<tt>invoke</tt>' instruction causes control to transfer to a specified
1369 function, with the possibility of control flow transfer to either the
1370 '<tt>normal</tt>' label or the
1371 '<tt>exception</tt>' label. If the callee function returns with the
1372 "<tt><a href="#i_ret">ret</a></tt>" instruction, control flow will return to the
1373 "normal" label. If the callee (or any indirect callees) returns with the "<a
1374 href="#i_unwind"><tt>unwind</tt></a>" instruction, control is interrupted and
1375 continued at the dynamically nearest "exception" label.</p>
1379 <p>This instruction requires several arguments:</p>
1383 The optional "cconv" marker indicates which <a href="callingconv">calling
1384 convention</a> the call should use. If none is specified, the call defaults
1385 to using C calling conventions.
1387 <li>'<tt>ptr to function ty</tt>': shall be the signature of the pointer to
1388 function value being invoked. In most cases, this is a direct function
1389 invocation, but indirect <tt>invoke</tt>s are just as possible, branching off
1390 an arbitrary pointer to function value.
1393 <li>'<tt>function ptr val</tt>': An LLVM value containing a pointer to a
1394 function to be invoked. </li>
1396 <li>'<tt>function args</tt>': argument list whose types match the function
1397 signature argument types. If the function signature indicates the function
1398 accepts a variable number of arguments, the extra arguments can be
1401 <li>'<tt>normal label</tt>': the label reached when the called function
1402 executes a '<tt><a href="#i_ret">ret</a></tt>' instruction. </li>
1404 <li>'<tt>exception label</tt>': the label reached when a callee returns with
1405 the <a href="#i_unwind"><tt>unwind</tt></a> instruction. </li>
1411 <p>This instruction is designed to operate as a standard '<tt><a
1412 href="#i_call">call</a></tt>' instruction in most regards. The primary
1413 difference is that it establishes an association with a label, which is used by
1414 the runtime library to unwind the stack.</p>
1416 <p>This instruction is used in languages with destructors to ensure that proper
1417 cleanup is performed in the case of either a <tt>longjmp</tt> or a thrown
1418 exception. Additionally, this is important for implementation of
1419 '<tt>catch</tt>' clauses in high-level languages that support them.</p>
1423 %retval = invoke int %Test(int 15) to label %Continue
1424 except label %TestCleanup <i>; {int}:retval set</i>
1425 %retval = invoke <a href="#callingconv">coldcc</a> int %Test(int 15) to label %Continue
1426 except label %TestCleanup <i>; {int}:retval set</i>
1431 <!-- _______________________________________________________________________ -->
1433 <div class="doc_subsubsection"> <a name="i_unwind">'<tt>unwind</tt>'
1434 Instruction</a> </div>
1436 <div class="doc_text">
1445 <p>The '<tt>unwind</tt>' instruction unwinds the stack, continuing control flow
1446 at the first callee in the dynamic call stack which used an <a
1447 href="#i_invoke"><tt>invoke</tt></a> instruction to perform the call. This is
1448 primarily used to implement exception handling.</p>
1452 <p>The '<tt>unwind</tt>' intrinsic causes execution of the current function to
1453 immediately halt. The dynamic call stack is then searched for the first <a
1454 href="#i_invoke"><tt>invoke</tt></a> instruction on the call stack. Once found,
1455 execution continues at the "exceptional" destination block specified by the
1456 <tt>invoke</tt> instruction. If there is no <tt>invoke</tt> instruction in the
1457 dynamic call chain, undefined behavior results.</p>
1460 <!-- _______________________________________________________________________ -->
1462 <div class="doc_subsubsection"> <a name="i_unreachable">'<tt>unreachable</tt>'
1463 Instruction</a> </div>
1465 <div class="doc_text">
1474 <p>The '<tt>unreachable</tt>' instruction has no defined semantics. This
1475 instruction is used to inform the optimizer that a particular portion of the
1476 code is not reachable. This can be used to indicate that the code after a
1477 no-return function cannot be reached, and other facts.</p>
1481 <p>The '<tt>unreachable</tt>' instruction has no defined semantics.</p>
1486 <!-- ======================================================================= -->
1487 <div class="doc_subsection"> <a name="binaryops">Binary Operations</a> </div>
1488 <div class="doc_text">
1489 <p>Binary operators are used to do most of the computation in a
1490 program. They require two operands, execute an operation on them, and
1491 produce a single value. The operands might represent
1492 multiple data, as is the case with the <a href="#t_packed">packed</a> data type.
1493 The result value of a binary operator is not
1494 necessarily the same type as its operands.</p>
1495 <p>There are several different binary operators:</p>
1497 <!-- _______________________________________________________________________ -->
1498 <div class="doc_subsubsection"> <a name="i_add">'<tt>add</tt>'
1499 Instruction</a> </div>
1500 <div class="doc_text">
1502 <pre> <result> = add <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1505 <p>The '<tt>add</tt>' instruction returns the sum of its two operands.</p>
1507 <p>The two arguments to the '<tt>add</tt>' instruction must be either <a
1508 href="#t_integer">integer</a> or <a href="#t_floating">floating point</a> values.
1509 This instruction can also take <a href="#t_packed">packed</a> versions of the values.
1510 Both arguments must have identical types.</p>
1512 <p>The value produced is the integer or floating point sum of the two
1515 <pre> <result> = add int 4, %var <i>; yields {int}:result = 4 + %var</i>
1518 <!-- _______________________________________________________________________ -->
1519 <div class="doc_subsubsection"> <a name="i_sub">'<tt>sub</tt>'
1520 Instruction</a> </div>
1521 <div class="doc_text">
1523 <pre> <result> = sub <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1526 <p>The '<tt>sub</tt>' instruction returns the difference of its two
1528 <p>Note that the '<tt>sub</tt>' instruction is used to represent the '<tt>neg</tt>'
1529 instruction present in most other intermediate representations.</p>
1531 <p>The two arguments to the '<tt>sub</tt>' instruction must be either <a
1532 href="#t_integer">integer</a> or <a href="#t_floating">floating point</a>
1534 This instruction can also take <a href="#t_packed">packed</a> versions of the values.
1535 Both arguments must have identical types.</p>
1537 <p>The value produced is the integer or floating point difference of
1538 the two operands.</p>
1540 <pre> <result> = sub int 4, %var <i>; yields {int}:result = 4 - %var</i>
1541 <result> = sub int 0, %val <i>; yields {int}:result = -%var</i>
1544 <!-- _______________________________________________________________________ -->
1545 <div class="doc_subsubsection"> <a name="i_mul">'<tt>mul</tt>'
1546 Instruction</a> </div>
1547 <div class="doc_text">
1549 <pre> <result> = mul <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1552 <p>The '<tt>mul</tt>' instruction returns the product of its two
1555 <p>The two arguments to the '<tt>mul</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 product of the
1563 <p>There is no signed vs unsigned multiplication. The appropriate
1564 action is taken based on the type of the operand.</p>
1566 <pre> <result> = mul int 4, %var <i>; yields {int}:result = 4 * %var</i>
1569 <!-- _______________________________________________________________________ -->
1570 <div class="doc_subsubsection"> <a name="i_div">'<tt>div</tt>'
1571 Instruction</a> </div>
1572 <div class="doc_text">
1574 <pre> <result> = div <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1577 <p>The '<tt>div</tt>' instruction returns the quotient of its two
1580 <p>The two arguments to the '<tt>div</tt>' instruction must be either <a
1581 href="#t_integer">integer</a> or <a href="#t_floating">floating point</a>
1583 This instruction can also take <a href="#t_packed">packed</a> versions of the values.
1584 Both arguments must have identical types.</p>
1586 <p>The value produced is the integer or floating point quotient of the
1589 <pre> <result> = div int 4, %var <i>; yields {int}:result = 4 / %var</i>
1592 <!-- _______________________________________________________________________ -->
1593 <div class="doc_subsubsection"> <a name="i_rem">'<tt>rem</tt>'
1594 Instruction</a> </div>
1595 <div class="doc_text">
1597 <pre> <result> = rem <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1600 <p>The '<tt>rem</tt>' instruction returns the remainder from the
1601 division of its two operands.</p>
1603 <p>The two arguments to the '<tt>rem</tt>' instruction must be either <a
1604 href="#t_integer">integer</a> or <a href="#t_floating">floating point</a>
1606 This instruction can also take <a href="#t_packed">packed</a> versions of the values.
1607 Both arguments must have identical types.</p>
1609 <p>This returns the <i>remainder</i> of a division (where the result
1610 has the same sign as the divisor), not the <i>modulus</i> (where the
1611 result has the same sign as the dividend) of a value. For more
1612 information about the difference, see <a
1613 href="http://mathforum.org/dr.math/problems/anne.4.28.99.html">The
1616 <pre> <result> = rem int 4, %var <i>; yields {int}:result = 4 % %var</i>
1620 <!-- _______________________________________________________________________ -->
1621 <div class="doc_subsubsection"> <a name="i_setcc">'<tt>set<i>cc</i></tt>'
1622 Instructions</a> </div>
1623 <div class="doc_text">
1625 <pre> <result> = seteq <ty> <var1>, <var2> <i>; yields {bool}:result</i>
1626 <result> = setne <ty> <var1>, <var2> <i>; yields {bool}:result</i>
1627 <result> = setlt <ty> <var1>, <var2> <i>; yields {bool}:result</i>
1628 <result> = setgt <ty> <var1>, <var2> <i>; yields {bool}:result</i>
1629 <result> = setle <ty> <var1>, <var2> <i>; yields {bool}:result</i>
1630 <result> = setge <ty> <var1>, <var2> <i>; yields {bool}:result</i>
1633 <p>The '<tt>set<i>cc</i></tt>' family of instructions returns a boolean
1634 value based on a comparison of their two operands.</p>
1636 <p>The two arguments to the '<tt>set<i>cc</i></tt>' instructions must
1637 be of <a href="#t_firstclass">first class</a> type (it is not possible
1638 to compare '<tt>label</tt>'s, '<tt>array</tt>'s, '<tt>structure</tt>'
1639 or '<tt>void</tt>' values, etc...). Both arguments must have identical
1642 <p>The '<tt>seteq</tt>' instruction yields a <tt>true</tt> '<tt>bool</tt>'
1643 value if both operands are equal.<br>
1644 The '<tt>setne</tt>' instruction yields a <tt>true</tt> '<tt>bool</tt>'
1645 value if both operands are unequal.<br>
1646 The '<tt>setlt</tt>' instruction yields a <tt>true</tt> '<tt>bool</tt>'
1647 value if the first operand is less than the second operand.<br>
1648 The '<tt>setgt</tt>' instruction yields a <tt>true</tt> '<tt>bool</tt>'
1649 value if the first operand is greater than the second operand.<br>
1650 The '<tt>setle</tt>' instruction yields a <tt>true</tt> '<tt>bool</tt>'
1651 value if the first operand is less than or equal to the second operand.<br>
1652 The '<tt>setge</tt>' instruction yields a <tt>true</tt> '<tt>bool</tt>'
1653 value if the first operand is greater than or equal to the second
1656 <pre> <result> = seteq int 4, 5 <i>; yields {bool}:result = false</i>
1657 <result> = setne float 4, 5 <i>; yields {bool}:result = true</i>
1658 <result> = setlt uint 4, 5 <i>; yields {bool}:result = true</i>
1659 <result> = setgt sbyte 4, 5 <i>; yields {bool}:result = false</i>
1660 <result> = setle sbyte 4, 5 <i>; yields {bool}:result = true</i>
1661 <result> = setge sbyte 4, 5 <i>; yields {bool}:result = false</i>
1665 <!-- ======================================================================= -->
1666 <div class="doc_subsection"> <a name="bitwiseops">Bitwise Binary
1667 Operations</a> </div>
1668 <div class="doc_text">
1669 <p>Bitwise binary operators are used to do various forms of
1670 bit-twiddling in a program. They are generally very efficient
1671 instructions and can commonly be strength reduced from other
1672 instructions. They require two operands, execute an operation on them,
1673 and produce a single value. The resulting value of the bitwise binary
1674 operators is always the same type as its first operand.</p>
1676 <!-- _______________________________________________________________________ -->
1677 <div class="doc_subsubsection"> <a name="i_and">'<tt>and</tt>'
1678 Instruction</a> </div>
1679 <div class="doc_text">
1681 <pre> <result> = and <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1684 <p>The '<tt>and</tt>' instruction returns the bitwise logical and of
1685 its two operands.</p>
1687 <p>The two arguments to the '<tt>and</tt>' instruction must be <a
1688 href="#t_integral">integral</a> values. Both arguments must have
1689 identical types.</p>
1691 <p>The truth table used for the '<tt>and</tt>' instruction is:</p>
1693 <div style="align: center">
1694 <table border="1" cellspacing="0" cellpadding="4">
1725 <pre> <result> = and int 4, %var <i>; yields {int}:result = 4 & %var</i>
1726 <result> = and int 15, 40 <i>; yields {int}:result = 8</i>
1727 <result> = and int 4, 8 <i>; yields {int}:result = 0</i>
1730 <!-- _______________________________________________________________________ -->
1731 <div class="doc_subsubsection"> <a name="i_or">'<tt>or</tt>' Instruction</a> </div>
1732 <div class="doc_text">
1734 <pre> <result> = or <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1737 <p>The '<tt>or</tt>' instruction returns the bitwise logical inclusive
1738 or of its two operands.</p>
1740 <p>The two arguments to the '<tt>or</tt>' instruction must be <a
1741 href="#t_integral">integral</a> values. Both arguments must have
1742 identical types.</p>
1744 <p>The truth table used for the '<tt>or</tt>' instruction is:</p>
1746 <div style="align: center">
1747 <table border="1" cellspacing="0" cellpadding="4">
1778 <pre> <result> = or int 4, %var <i>; yields {int}:result = 4 | %var</i>
1779 <result> = or int 15, 40 <i>; yields {int}:result = 47</i>
1780 <result> = or int 4, 8 <i>; yields {int}:result = 12</i>
1783 <!-- _______________________________________________________________________ -->
1784 <div class="doc_subsubsection"> <a name="i_xor">'<tt>xor</tt>'
1785 Instruction</a> </div>
1786 <div class="doc_text">
1788 <pre> <result> = xor <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1791 <p>The '<tt>xor</tt>' instruction returns the bitwise logical exclusive
1792 or of its two operands. The <tt>xor</tt> is used to implement the
1793 "one's complement" operation, which is the "~" operator in C.</p>
1795 <p>The two arguments to the '<tt>xor</tt>' instruction must be <a
1796 href="#t_integral">integral</a> values. Both arguments must have
1797 identical types.</p>
1799 <p>The truth table used for the '<tt>xor</tt>' instruction is:</p>
1801 <div style="align: center">
1802 <table border="1" cellspacing="0" cellpadding="4">
1834 <pre> <result> = xor int 4, %var <i>; yields {int}:result = 4 ^ %var</i>
1835 <result> = xor int 15, 40 <i>; yields {int}:result = 39</i>
1836 <result> = xor int 4, 8 <i>; yields {int}:result = 12</i>
1837 <result> = xor int %V, -1 <i>; yields {int}:result = ~%V</i>
1840 <!-- _______________________________________________________________________ -->
1841 <div class="doc_subsubsection"> <a name="i_shl">'<tt>shl</tt>'
1842 Instruction</a> </div>
1843 <div class="doc_text">
1845 <pre> <result> = shl <ty> <var1>, ubyte <var2> <i>; yields {ty}:result</i>
1848 <p>The '<tt>shl</tt>' instruction returns the first operand shifted to
1849 the left a specified number of bits.</p>
1851 <p>The first argument to the '<tt>shl</tt>' instruction must be an <a
1852 href="#t_integer">integer</a> type. The second argument must be an '<tt>ubyte</tt>'
1855 <p>The value produced is <tt>var1</tt> * 2<sup><tt>var2</tt></sup>.</p>
1857 <pre> <result> = shl int 4, ubyte %var <i>; yields {int}:result = 4 << %var</i>
1858 <result> = shl int 4, ubyte 2 <i>; yields {int}:result = 16</i>
1859 <result> = shl int 1, ubyte 10 <i>; yields {int}:result = 1024</i>
1862 <!-- _______________________________________________________________________ -->
1863 <div class="doc_subsubsection"> <a name="i_shr">'<tt>shr</tt>'
1864 Instruction</a> </div>
1865 <div class="doc_text">
1867 <pre> <result> = shr <ty> <var1>, ubyte <var2> <i>; yields {ty}:result</i>
1870 <p>The '<tt>shr</tt>' instruction returns the first operand shifted to
1871 the right a specified number of bits.</p>
1873 <p>The first argument to the '<tt>shr</tt>' instruction must be an <a
1874 href="#t_integer">integer</a> type. The second argument must be an '<tt>ubyte</tt>'
1877 <p>If the first argument is a <a href="#t_signed">signed</a> type, the
1878 most significant bit is duplicated in the newly free'd bit positions.
1879 If the first argument is unsigned, zero bits shall fill the empty
1882 <pre> <result> = shr int 4, ubyte %var <i>; yields {int}:result = 4 >> %var</i>
1883 <result> = shr uint 4, ubyte 1 <i>; yields {uint}:result = 2</i>
1884 <result> = shr int 4, ubyte 2 <i>; yields {int}:result = 1</i>
1885 <result> = shr sbyte 4, ubyte 3 <i>; yields {sbyte}:result = 0</i>
1886 <result> = shr sbyte -2, ubyte 1 <i>; yields {sbyte}:result = -1</i>
1890 <!-- ======================================================================= -->
1891 <div class="doc_subsection">
1892 <a name="memoryops">Memory Access Operations</a>
1895 <div class="doc_text">
1897 <p>A key design point of an SSA-based representation is how it
1898 represents memory. In LLVM, no memory locations are in SSA form, which
1899 makes things very simple. This section describes how to read, write,
1900 allocate, and free memory in LLVM.</p>
1904 <!-- _______________________________________________________________________ -->
1905 <div class="doc_subsubsection">
1906 <a name="i_malloc">'<tt>malloc</tt>' Instruction</a>
1909 <div class="doc_text">
1914 <result> = malloc <type>[, uint <NumElements>][, align <alignment>] <i>; yields {type*}:result</i>
1919 <p>The '<tt>malloc</tt>' instruction allocates memory from the system
1920 heap and returns a pointer to it.</p>
1924 <p>The '<tt>malloc</tt>' instruction allocates
1925 <tt>sizeof(<type>)*NumElements</tt>
1926 bytes of memory from the operating system and returns a pointer of the
1927 appropriate type to the program. If "NumElements" is specified, it is the
1928 number of elements allocated. If an alignment is specified, the value result
1929 of the allocation is guaranteed to be aligned to at least that boundary. If
1930 not specified, or if zero, the target can choose to align the allocation on any
1931 convenient boundary.</p>
1933 <p>'<tt>type</tt>' must be a sized type.</p>
1937 <p>Memory is allocated using the system "<tt>malloc</tt>" function, and
1938 a pointer is returned.</p>
1943 %array = malloc [4 x ubyte ] <i>; yields {[%4 x ubyte]*}:array</i>
1945 %size = <a href="#i_add">add</a> uint 2, 2 <i>; yields {uint}:size = uint 4</i>
1946 %array1 = malloc ubyte, uint 4 <i>; yields {ubyte*}:array1</i>
1947 %array2 = malloc [12 x ubyte], uint %size <i>; yields {[12 x ubyte]*}:array2</i>
1948 %array3 = malloc int, uint 4, align 1024 <i>; yields {int*}:array3</i>
1949 %array4 = malloc int, align 1024 <i>; yields {int*}:array4</i>
1953 <!-- _______________________________________________________________________ -->
1954 <div class="doc_subsubsection">
1955 <a name="i_free">'<tt>free</tt>' Instruction</a>
1958 <div class="doc_text">
1963 free <type> <value> <i>; yields {void}</i>
1968 <p>The '<tt>free</tt>' instruction returns memory back to the unused
1969 memory heap to be reallocated in the future.</p>
1973 <p>'<tt>value</tt>' shall be a pointer value that points to a value
1974 that was allocated with the '<tt><a href="#i_malloc">malloc</a></tt>'
1979 <p>Access to the memory pointed to by the pointer is no longer defined
1980 after this instruction executes.</p>
1985 %array = <a href="#i_malloc">malloc</a> [4 x ubyte] <i>; yields {[4 x ubyte]*}:array</i>
1986 free [4 x ubyte]* %array
1990 <!-- _______________________________________________________________________ -->
1991 <div class="doc_subsubsection">
1992 <a name="i_alloca">'<tt>alloca</tt>' Instruction</a>
1995 <div class="doc_text">
2000 <result> = alloca <type>[, uint <NumElements>][, align <alignment>] <i>; yields {type*}:result</i>
2005 <p>The '<tt>alloca</tt>' instruction allocates memory on the current
2006 stack frame of the procedure that is live until the current function
2007 returns to its caller.</p>
2011 <p>The '<tt>alloca</tt>' instruction allocates <tt>sizeof(<type>)*NumElements</tt>
2012 bytes of memory on the runtime stack, returning a pointer of the
2013 appropriate type to the program. If "NumElements" is specified, it is the
2014 number of elements allocated. If an alignment is specified, the value result
2015 of the allocation is guaranteed to be aligned to at least that boundary. If
2016 not specified, or if zero, the target can choose to align the allocation on any
2017 convenient boundary.</p>
2019 <p>'<tt>type</tt>' may be any sized type.</p>
2023 <p>Memory is allocated; a pointer is returned. '<tt>alloca</tt>'d
2024 memory is automatically released when the function returns. The '<tt>alloca</tt>'
2025 instruction is commonly used to represent automatic variables that must
2026 have an address available. When the function returns (either with the <tt><a
2027 href="#i_ret">ret</a></tt> or <tt><a href="#i_unwind">unwind</a></tt>
2028 instructions), the memory is reclaimed.</p>
2033 %ptr = alloca int <i>; yields {int*}:ptr</i>
2034 %ptr = alloca int, uint 4 <i>; yields {int*}:ptr</i>
2035 %ptr = alloca int, uint 4, align 1024 <i>; yields {int*}:ptr</i>
2036 %ptr = alloca int, align 1024 <i>; yields {int*}:ptr</i>
2040 <!-- _______________________________________________________________________ -->
2041 <div class="doc_subsubsection"> <a name="i_load">'<tt>load</tt>'
2042 Instruction</a> </div>
2043 <div class="doc_text">
2045 <pre> <result> = load <ty>* <pointer><br> <result> = volatile load <ty>* <pointer><br></pre>
2047 <p>The '<tt>load</tt>' instruction is used to read from memory.</p>
2049 <p>The argument to the '<tt>load</tt>' instruction specifies the memory
2050 address from which to load. The pointer must point to a <a
2051 href="#t_firstclass">first class</a> type. If the <tt>load</tt> is
2052 marked as <tt>volatile</tt>, then the optimizer is not allowed to modify
2053 the number or order of execution of this <tt>load</tt> with other
2054 volatile <tt>load</tt> and <tt><a href="#i_store">store</a></tt>
2057 <p>The location of memory pointed to is loaded.</p>
2059 <pre> %ptr = <a href="#i_alloca">alloca</a> int <i>; yields {int*}:ptr</i>
2061 href="#i_store">store</a> int 3, int* %ptr <i>; yields {void}</i>
2062 %val = load int* %ptr <i>; yields {int}:val = int 3</i>
2065 <!-- _______________________________________________________________________ -->
2066 <div class="doc_subsubsection"> <a name="i_store">'<tt>store</tt>'
2067 Instruction</a> </div>
2069 <pre> store <ty> <value>, <ty>* <pointer> <i>; yields {void}</i>
2070 volatile store <ty> <value>, <ty>* <pointer> <i>; yields {void}</i>
2073 <p>The '<tt>store</tt>' instruction is used to write to memory.</p>
2075 <p>There are two arguments to the '<tt>store</tt>' instruction: a value
2076 to store and an address in which to store it. The type of the '<tt><pointer></tt>'
2077 operand must be a pointer to the type of the '<tt><value></tt>'
2078 operand. If the <tt>store</tt> is marked as <tt>volatile</tt>, then the
2079 optimizer is not allowed to modify the number or order of execution of
2080 this <tt>store</tt> with other volatile <tt>load</tt> and <tt><a
2081 href="#i_store">store</a></tt> instructions.</p>
2083 <p>The contents of memory are updated to contain '<tt><value></tt>'
2084 at the location specified by the '<tt><pointer></tt>' operand.</p>
2086 <pre> %ptr = <a href="#i_alloca">alloca</a> int <i>; yields {int*}:ptr</i>
2088 href="#i_store">store</a> int 3, int* %ptr <i>; yields {void}</i>
2089 %val = load int* %ptr <i>; yields {int}:val = int 3</i>
2091 <!-- _______________________________________________________________________ -->
2092 <div class="doc_subsubsection">
2093 <a name="i_getelementptr">'<tt>getelementptr</tt>' Instruction</a>
2096 <div class="doc_text">
2099 <result> = getelementptr <ty>* <ptrval>{, <ty> <idx>}*
2105 The '<tt>getelementptr</tt>' instruction is used to get the address of a
2106 subelement of an aggregate data structure.</p>
2110 <p>This instruction takes a list of integer constants that indicate what
2111 elements of the aggregate object to index to. The actual types of the arguments
2112 provided depend on the type of the first pointer argument. The
2113 '<tt>getelementptr</tt>' instruction is used to index down through the type
2114 levels of a structure or to a specific index in an array. When indexing into a
2115 structure, only <tt>uint</tt>
2116 integer constants are allowed. When indexing into an array or pointer,
2117 <tt>int</tt> and <tt>long</tt> indexes are allowed of any sign.</p>
2119 <p>For example, let's consider a C code fragment and how it gets
2120 compiled to LLVM:</p>
2134 int *foo(struct ST *s) {
2135 return &s[1].Z.B[5][13];
2139 <p>The LLVM code generated by the GCC frontend is:</p>
2142 %RT = type { sbyte, [10 x [20 x int]], sbyte }
2143 %ST = type { int, double, %RT }
2147 int* %foo(%ST* %s) {
2149 %reg = getelementptr %ST* %s, int 1, uint 2, uint 1, int 5, int 13
2156 <p>The index types specified for the '<tt>getelementptr</tt>' instruction depend
2157 on the pointer type that is being indexed into. <a href="#t_pointer">Pointer</a>
2158 and <a href="#t_array">array</a> types require <tt>uint</tt>, <tt>int</tt>,
2159 <tt>ulong</tt>, or <tt>long</tt> values, and <a href="#t_struct">structure</a>
2160 types require <tt>uint</tt> <b>constants</b>.</p>
2162 <p>In the example above, the first index is indexing into the '<tt>%ST*</tt>'
2163 type, which is a pointer, yielding a '<tt>%ST</tt>' = '<tt>{ int, double, %RT
2164 }</tt>' type, a structure. The second index indexes into the third element of
2165 the structure, yielding a '<tt>%RT</tt>' = '<tt>{ sbyte, [10 x [20 x int]],
2166 sbyte }</tt>' type, another structure. The third index indexes into the second
2167 element of the structure, yielding a '<tt>[10 x [20 x int]]</tt>' type, an
2168 array. The two dimensions of the array are subscripted into, yielding an
2169 '<tt>int</tt>' type. The '<tt>getelementptr</tt>' instruction returns a pointer
2170 to this element, thus computing a value of '<tt>int*</tt>' type.</p>
2172 <p>Note that it is perfectly legal to index partially through a
2173 structure, returning a pointer to an inner element. Because of this,
2174 the LLVM code for the given testcase is equivalent to:</p>
2177 int* %foo(%ST* %s) {
2178 %t1 = getelementptr %ST* %s, int 1 <i>; yields %ST*:%t1</i>
2179 %t2 = getelementptr %ST* %t1, int 0, uint 2 <i>; yields %RT*:%t2</i>
2180 %t3 = getelementptr %RT* %t2, int 0, uint 1 <i>; yields [10 x [20 x int]]*:%t3</i>
2181 %t4 = getelementptr [10 x [20 x int]]* %t3, int 0, int 5 <i>; yields [20 x int]*:%t4</i>
2182 %t5 = getelementptr [20 x int]* %t4, int 0, int 13 <i>; yields int*:%t5</i>
2187 <p>Note that it is undefined to access an array out of bounds: array and
2188 pointer indexes must always be within the defined bounds of the array type.
2189 The one exception for this rules is zero length arrays. These arrays are
2190 defined to be accessible as variable length arrays, which requires access
2191 beyond the zero'th element.</p>
2196 <i>; yields [12 x ubyte]*:aptr</i>
2197 %aptr = getelementptr {int, [12 x ubyte]}* %sptr, long 0, uint 1
2201 <!-- ======================================================================= -->
2202 <div class="doc_subsection"> <a name="otherops">Other Operations</a> </div>
2203 <div class="doc_text">
2204 <p>The instructions in this category are the "miscellaneous"
2205 instructions, which defy better classification.</p>
2207 <!-- _______________________________________________________________________ -->
2208 <div class="doc_subsubsection"> <a name="i_phi">'<tt>phi</tt>'
2209 Instruction</a> </div>
2210 <div class="doc_text">
2212 <pre> <result> = phi <ty> [ <val0>, <label0>], ...<br></pre>
2214 <p>The '<tt>phi</tt>' instruction is used to implement the φ node in
2215 the SSA graph representing the function.</p>
2217 <p>The type of the incoming values are specified with the first type
2218 field. After this, the '<tt>phi</tt>' instruction takes a list of pairs
2219 as arguments, with one pair for each predecessor basic block of the
2220 current block. Only values of <a href="#t_firstclass">first class</a>
2221 type may be used as the value arguments to the PHI node. Only labels
2222 may be used as the label arguments.</p>
2223 <p>There must be no non-phi instructions between the start of a basic
2224 block and the PHI instructions: i.e. PHI instructions must be first in
2227 <p>At runtime, the '<tt>phi</tt>' instruction logically takes on the
2228 value specified by the parameter, depending on which basic block we
2229 came from in the last <a href="#terminators">terminator</a> instruction.</p>
2231 <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>
2234 <!-- _______________________________________________________________________ -->
2235 <div class="doc_subsubsection">
2236 <a name="i_cast">'<tt>cast .. to</tt>' Instruction</a>
2239 <div class="doc_text">
2244 <result> = cast <ty> <value> to <ty2> <i>; yields ty2</i>
2250 The '<tt>cast</tt>' instruction is used as the primitive means to convert
2251 integers to floating point, change data type sizes, and break type safety (by
2259 The '<tt>cast</tt>' instruction takes a value to cast, which must be a first
2260 class value, and a type to cast it to, which must also be a <a
2261 href="#t_firstclass">first class</a> type.
2267 This instruction follows the C rules for explicit casts when determining how the
2268 data being cast must change to fit in its new container.
2272 When casting to bool, any value that would be considered true in the context of
2273 a C '<tt>if</tt>' condition is converted to the boolean '<tt>true</tt>' values,
2274 all else are '<tt>false</tt>'.
2278 When extending an integral value from a type of one signness to another (for
2279 example '<tt>sbyte</tt>' to '<tt>ulong</tt>'), the value is sign-extended if the
2280 <b>source</b> value is signed, and zero-extended if the source value is
2281 unsigned. <tt>bool</tt> values are always zero extended into either zero or
2288 %X = cast int 257 to ubyte <i>; yields ubyte:1</i>
2289 %Y = cast int 123 to bool <i>; yields bool:true</i>
2293 <!-- _______________________________________________________________________ -->
2294 <div class="doc_subsubsection">
2295 <a name="i_select">'<tt>select</tt>' Instruction</a>
2298 <div class="doc_text">
2303 <result> = select bool <cond>, <ty> <val1>, <ty> <val2> <i>; yields ty</i>
2309 The '<tt>select</tt>' instruction is used to choose one value based on a
2310 condition, without branching.
2317 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.
2323 If the boolean condition evaluates to true, the instruction returns the first
2324 value argument; otherwise, it returns the second value argument.
2330 %X = select bool true, ubyte 17, ubyte 42 <i>; yields ubyte:17</i>
2334 <!-- _______________________________________________________________________ -->
2335 <div class="doc_subsubsection"> <a name="i_vsetint">'<tt>vsetint</tt>'
2336 Instruction</a> </div>
2337 <div class="doc_text">
2339 <pre><result> = vsetint <op>, <n x <ty>> <var1>, <var2> <i>; yields <n x bool></i>
2344 <p>The '<tt>vsetint</tt>' instruction takes two integer vectors and
2345 returns a vector of boolean values representing, at each position, the
2346 result of the comparison between the values at that position in the
2351 <p>The arguments to a '<tt>vsetint</tt>' instruction are a comparison
2352 operation and two value arguments. The value arguments must be of <a
2353 href="#t_integral">integral</a> <a href="#t_packed">packed</a> type,
2354 and they must have identical types. The operation argument must be
2355 one of <tt>eq</tt>, <tt>ne</tt>, <tt>slt</tt>, <tt>sgt</tt>,
2356 <tt>sle</tt>, <tt>sge</tt>, <tt>ult</tt>, <tt>ugt</tt>, <tt>ule</tt>,
2357 <tt>uge</tt>, <tt>true</tt>, and <tt>false</tt>. The result is a
2358 packed <tt>bool</tt> value with the same length as each operand.</p>
2362 <p>The following table shows the semantics of '<tt>vsetint</tt>'. For
2363 each position of the result, the comparison is done on the
2364 corresponding positions of the two value arguments. Note that the
2365 signedness of the comparison depends on the comparison opcode and
2366 <i>not</i> on the signedness of the value operands. E.g., <tt>vsetint
2367 slt <4 x unsigned> %x, %y</tt> does an elementwise <i>signed</i>
2368 comparison of <tt>%x</tt> and <tt>%y</tt>.</p>
2370 <table border="1" cellspacing="0" cellpadding="4">
2372 <tr><th>Operation</th><th>Result is true iff</th><th>Comparison is</th></tr>
2373 <tr><td><tt>eq</tt></td><td>var1 == var2</td><td>--</td></tr>
2374 <tr><td><tt>ne</tt></td><td>var1 != var2</td><td>--</td></tr>
2375 <tr><td><tt>slt</tt></td><td>var1 < var2</td><td>signed</td></tr>
2376 <tr><td><tt>sgt</tt></td><td>var1 > var2</td><td>signed</td></tr>
2377 <tr><td><tt>sle</tt></td><td>var1 <= var2</td><td>signed</td></tr>
2378 <tr><td><tt>sge</tt></td><td>var1 >= var2</td><td>signed</td></tr>
2379 <tr><td><tt>ult</tt></td><td>var1 < var2</td><td>unsigned</td></tr>
2380 <tr><td><tt>ugt</tt></td><td>var1 > var2</td><td>unsigned</td></tr>
2381 <tr><td><tt>ule</tt></td><td>var1 <= var2</td><td>unsigned</td></tr>
2382 <tr><td><tt>uge</tt></td><td>var1 >= var2</td><td>unsigned</td></tr>
2383 <tr><td><tt>true</tt></td><td>always</td><td>--</td></tr>
2384 <tr><td><tt>false</tt></td><td>never</td><td>--</td></tr>
2389 <pre> <result> = vsetint eq <2 x int> <int 0, int 1>, <int 1, int 0> <i>; yields {<2 x bool>}:result = false, false</i>
2390 <result> = vsetint ne <2 x int> <int 0, int 1>, <int 1, int 0> <i>; yields {<2 x bool>}:result = true, true</i>
2391 <result> = vsetint slt <2 x int> <int 0, int 1>, <int 1, int 0> <i>; yields {<2 x bool>}:result = true, false</i>
2392 <result> = vsetint sgt <2 x int> <int 0, int 1>, <int 1, int 0> <i>; yields {<2 x bool>}:result = false, true</i>
2393 <result> = vsetint sle <2 x int> <int 0, int 1>, <int 1, int 0> <i>; yields {<2 x bool>}:result = true, false</i>
2394 <result> = vsetint sge <2 x int> <int 0, int 1>, <int 1, int 0> <i>; yields {<2 x bool>}:result = false, true</i>
2398 <!-- _______________________________________________________________________ -->
2399 <div class="doc_subsubsection"> <a name="i_vsetfp">'<tt>vsetfp</tt>'
2400 Instruction</a> </div>
2401 <div class="doc_text">
2403 <pre><result> = vsetfp <op>, <n x <ty>> <var1>, <var2> <i>; yields <n x bool></i>
2408 <p>The '<tt>vsetfp</tt>' instruction takes two floating point vector
2409 arguments and returns a vector of boolean values representing, at each
2410 position, the result of the comparison between the values at that
2411 position in the two operands.</p>
2415 <p>The arguments to a '<tt>vsetfp</tt>' instruction are a comparison
2416 operation and two value arguments. The value arguments must be of <a
2417 href="t_floating">floating point</a> <a href="#t_packed">packed</a>
2418 type, and they must have identical types. The operation argument must
2419 be one of <tt>eq</tt>, <tt>ne</tt>, <tt>lt</tt>, <tt>gt</tt>,
2420 <tt>le</tt>, <tt>ge</tt>, <tt>oeq</tt>, <tt>one</tt>, <tt>olt</tt>,
2421 <tt>ogt</tt>, <tt>ole</tt>, <tt>oge</tt>, <tt>ueq</tt>, <tt>une</tt>,
2422 <tt>ult</tt>, <tt>ugt</tt>, <tt>ule</tt>, <tt>uge</tt>, <tt>o</tt>,
2423 <tt>u</tt>, <tt>true</tt>, and <tt>false</tt>. The result is a packed
2424 <tt>bool</tt> value with the same length as each operand.</p>
2428 <p>The following table shows the semantics of '<tt>vsetfp</tt>' for
2429 floating point types. If either operand is a floating point Not a
2430 Number (NaN) value, the operation is unordered, and the value in the
2431 first column below is produced at that position. Otherwise, the
2432 operation is ordered, and the value in the second column is
2435 <table border="1" cellspacing="0" cellpadding="4">
2437 <tr><th>Operation</th><th>If unordered<th>Otherwise true iff</th></tr>
2438 <tr><td><tt>eq</tt></td><td>undefined</td><td>var1 == var2</td></tr>
2439 <tr><td><tt>ne</tt></td><td>undefined</td><td>var1 != var2</td></tr>
2440 <tr><td><tt>lt</tt></td><td>undefined</td><td>var1 < var2</td></tr>
2441 <tr><td><tt>gt</tt></td><td>undefined</td><td>var1 > var2</td></tr>
2442 <tr><td><tt>le</tt></td><td>undefined</td><td>var1 <= var2</td></tr>
2443 <tr><td><tt>ge</tt></td><td>undefined</td><td>var1 >= var2</td></tr>
2444 <tr><td><tt>oeq</tt></td><td>false</td><td>var1 == var2</td></tr>
2445 <tr><td><tt>one</tt></td><td>false</td><td>var1 != var2</td></tr>
2446 <tr><td><tt>olt</tt></td><td>false</td><td>var1 < var2</td></tr>
2447 <tr><td><tt>ogt</tt></td><td>false</td><td>var1 > var2</td></tr>
2448 <tr><td><tt>ole</tt></td><td>false</td><td>var1 <= var2</td></tr>
2449 <tr><td><tt>oge</tt></td><td>false</td><td>var1 >= var2</td></tr>
2450 <tr><td><tt>ueq</tt></td><td>true</td><td>var1 == var2</td></tr>
2451 <tr><td><tt>une</tt></td><td>true</td><td>var1 != var2</td></tr>
2452 <tr><td><tt>ult</tt></td><td>true</td><td>var1 < var2</td></tr>
2453 <tr><td><tt>ugt</tt></td><td>true</td><td>var1 > var2</td></tr>
2454 <tr><td><tt>ule</tt></td><td>true</td><td>var1 <= var2</td></tr>
2455 <tr><td><tt>uge</tt></td><td>true</td><td>var1 >= var2</td></tr>
2456 <tr><td><tt>o</tt></td><td>false</td><td>always</td></tr>
2457 <tr><td><tt>u</tt></td><td>true</td><td>never</td></tr>
2458 <tr><td><tt>true</tt></td><td>true</td><td>always</td></tr>
2459 <tr><td><tt>false</tt></td><td>false</td><td>never</td></tr>
2464 <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>
2465 <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>
2466 <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>
2467 <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>
2468 <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>
2469 <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>
2473 <!-- _______________________________________________________________________ -->
2474 <div class="doc_subsubsection">
2475 <a name="i_vselect">'<tt>vselect</tt>' Instruction</a>
2478 <div class="doc_text">
2483 <result> = vselect <n x bool> <cond>, <n x <ty>> <val1>, <n x <ty>> <val2> <i>; yields <n x <ty>></i>
2489 The '<tt>vselect</tt>' instruction chooses one value at each position
2490 of a vector based on a condition.
2497 The '<tt>vselect</tt>' instruction requires a <a
2498 href="#t_packed">packed</a> <tt>bool</tt> value indicating the
2499 condition at each vector position, and two values of the same packed
2500 type. All three operands must have the same length. The type of the
2501 result is the same as the type of the two value operands.</p>
2506 At each position where the <tt>bool</tt> vector is true, that position
2507 of the result gets its value from the first value argument; otherwise,
2508 it gets its value from the second value argument.
2514 %X = vselect bool <2 x bool> <bool true, bool false>, <2 x ubyte> <ubyte 17, ubyte 17>,
2515 <2 x ubyte> <ubyte 42, ubyte 42> <i>; yields <2 x ubyte>:17, 42</i>
2519 <!-- _______________________________________________________________________ -->
2520 <div class="doc_subsubsection">
2521 <a name="i_extractelement">'<tt>extractelement</tt>' Instruction</a>
2524 <div class="doc_text">
2529 <result> = extractelement <n x <ty>> <val>, uint <idx> <i>; yields <ty></i>
2535 The '<tt>extractelement</tt>' instruction extracts a single scalar
2536 element from a packed vector at a specified index.
2543 The first operand of an '<tt>extractelement</tt>' instruction is a
2544 value of <a href="#t_packed">packed</a> type. The second operand is
2545 an index indicating the position from which to extract the element.
2546 The index may be a variable.</p>
2551 The result is a scalar of the same type as the element type of
2552 <tt>val</tt>. Its value is the value at position <tt>idx</tt> of
2553 <tt>val</tt>. If <tt>idx</tt> exceeds the length of <tt>val</tt>, the
2554 results are undefined.
2560 %result = extractelement <4 x int> %vec, uint 0 <i>; yields int</i>
2565 <!-- _______________________________________________________________________ -->
2566 <div class="doc_subsubsection">
2567 <a name="i_insertelement">'<tt>insertelement</tt>' Instruction</a>
2570 <div class="doc_text">
2575 <result> = insertelement <n x <ty>> <val>, <ty> <elt>, uint <idx> <i>; yields <n x <ty>></i>
2581 The '<tt>insertelement</tt>' instruction inserts a scalar
2582 element into a packed vector at a specified index.
2589 The first operand of an '<tt>insertelement</tt>' instruction is a
2590 value of <a href="#t_packed">packed</a> type. The second operand is a
2591 scalar value whose type must equal the element type of the first
2592 operand. The third operand is an index indicating the position at
2593 which to insert the value. The index may be a variable.</p>
2598 The result is a packed vector of the same type as <tt>val</tt>. Its
2599 element values are those of <tt>val</tt> except at position
2600 <tt>idx</tt>, where it gets the value <tt>elt</tt>. If <tt>idx</tt>
2601 exceeds the length of <tt>val</tt>, the results are undefined.
2607 %result = insertelement <4 x int> %vec, int 1, uint 0 <i>; yields <4 x int></i>
2612 <!-- _______________________________________________________________________ -->
2613 <div class="doc_subsubsection">
2614 <a name="i_call">'<tt>call</tt>' Instruction</a>
2617 <div class="doc_text">
2621 <result> = [tail] call [<a href="#callingconv">cconv</a>] <ty>* <fnptrval>(<param list>)
2626 <p>The '<tt>call</tt>' instruction represents a simple function call.</p>
2630 <p>This instruction requires several arguments:</p>
2634 <p>The optional "tail" marker indicates whether the callee function accesses
2635 any allocas or varargs in the caller. If the "tail" marker is present, the
2636 function call is eligible for tail call optimization. Note that calls may
2637 be marked "tail" even if they do not occur before a <a
2638 href="#i_ret"><tt>ret</tt></a> instruction.
2641 <p>The optional "cconv" marker indicates which <a href="callingconv">calling
2642 convention</a> the call should use. If none is specified, the call defaults
2643 to using C calling conventions.
2646 <p>'<tt>ty</tt>': shall be the signature of the pointer to function value
2647 being invoked. The argument types must match the types implied by this
2648 signature. This type can be omitted if the function is not varargs and
2649 if the function type does not return a pointer to a function.</p>
2652 <p>'<tt>fnptrval</tt>': An LLVM value containing a pointer to a function to
2653 be invoked. In most cases, this is a direct function invocation, but
2654 indirect <tt>call</tt>s are just as possible, calling an arbitrary pointer
2655 to function value.</p>
2658 <p>'<tt>function args</tt>': argument list whose types match the
2659 function signature argument types. All arguments must be of
2660 <a href="#t_firstclass">first class</a> type. If the function signature
2661 indicates the function accepts a variable number of arguments, the extra
2662 arguments can be specified.</p>
2668 <p>The '<tt>call</tt>' instruction is used to cause control flow to
2669 transfer to a specified function, with its incoming arguments bound to
2670 the specified values. Upon a '<tt><a href="#i_ret">ret</a></tt>'
2671 instruction in the called function, control flow continues with the
2672 instruction after the function call, and the return value of the
2673 function is bound to the result argument. This is a simpler case of
2674 the <a href="#i_invoke">invoke</a> instruction.</p>
2679 %retval = call int %test(int %argc)
2680 call int(sbyte*, ...) *%printf(sbyte* %msg, int 12, sbyte 42);
2681 %X = tail call int %foo()
2682 %Y = tail call <a href="#callingconv">fastcc</a> int %foo()
2687 <!-- _______________________________________________________________________ -->
2688 <div class="doc_subsubsection">
2689 <a name="i_va_arg">'<tt>va_arg</tt>' Instruction</a>
2692 <div class="doc_text">
2697 <resultval> = va_arg <va_list*> <arglist>, <argty>
2702 <p>The '<tt>va_arg</tt>' instruction is used to access arguments passed through
2703 the "variable argument" area of a function call. It is used to implement the
2704 <tt>va_arg</tt> macro in C.</p>
2708 <p>This instruction takes a <tt>va_list*</tt> value and the type of
2709 the argument. It returns a value of the specified argument type and
2710 increments the <tt>va_list</tt> to point to the next argument. Again, the
2711 actual type of <tt>va_list</tt> is target specific.</p>
2715 <p>The '<tt>va_arg</tt>' instruction loads an argument of the specified
2716 type from the specified <tt>va_list</tt> and causes the
2717 <tt>va_list</tt> to point to the next argument. For more information,
2718 see the variable argument handling <a href="#int_varargs">Intrinsic
2721 <p>It is legal for this instruction to be called in a function which does not
2722 take a variable number of arguments, for example, the <tt>vfprintf</tt>
2725 <p><tt>va_arg</tt> is an LLVM instruction instead of an <a
2726 href="#intrinsics">intrinsic function</a> because it takes a type as an
2731 <p>See the <a href="#int_varargs">variable argument processing</a> section.</p>
2735 <!-- *********************************************************************** -->
2736 <div class="doc_section"> <a name="intrinsics">Intrinsic Functions</a> </div>
2737 <!-- *********************************************************************** -->
2739 <div class="doc_text">
2741 <p>LLVM supports the notion of an "intrinsic function". These functions have
2742 well known names and semantics and are required to follow certain
2743 restrictions. Overall, these instructions represent an extension mechanism for
2744 the LLVM language that does not require changing all of the transformations in
2745 LLVM to add to the language (or the bytecode reader/writer, the parser,
2748 <p>Intrinsic function names must all start with an "<tt>llvm.</tt>" prefix. This
2749 prefix is reserved in LLVM for intrinsic names; thus, functions may not be named
2750 this. Intrinsic functions must always be external functions: you cannot define
2751 the body of intrinsic functions. Intrinsic functions may only be used in call
2752 or invoke instructions: it is illegal to take the address of an intrinsic
2753 function. Additionally, because intrinsic functions are part of the LLVM
2754 language, it is required that they all be documented here if any are added.</p>
2757 <p>To learn how to add an intrinsic function, please see the <a
2758 href="ExtendingLLVM.html">Extending LLVM Guide</a>.
2763 <!-- ======================================================================= -->
2764 <div class="doc_subsection">
2765 <a name="int_varargs">Variable Argument Handling Intrinsics</a>
2768 <div class="doc_text">
2770 <p>Variable argument support is defined in LLVM with the <a
2771 href="#i_va_arg"><tt>va_arg</tt></a> instruction and these three
2772 intrinsic functions. These functions are related to the similarly
2773 named macros defined in the <tt><stdarg.h></tt> header file.</p>
2775 <p>All of these functions operate on arguments that use a
2776 target-specific value type "<tt>va_list</tt>". The LLVM assembly
2777 language reference manual does not define what this type is, so all
2778 transformations should be prepared to handle intrinsics with any type
2781 <p>This example shows how the <a href="#i_vanext"><tt>vanext</tt></a>
2782 instruction and the variable argument handling intrinsic functions are
2786 int %test(int %X, ...) {
2787 ; Initialize variable argument processing
2789 call void %<a href="#i_va_start">llvm.va_start</a>(sbyte** %ap)
2791 ; Read a single integer argument
2792 %tmp = va_arg sbyte** %ap, int
2794 ; Demonstrate usage of llvm.va_copy and llvm.va_end
2796 call void %<a href="#i_va_copy">llvm.va_copy</a>(sbyte** %aq, sbyte** %ap)
2797 call void %<a href="#i_va_end">llvm.va_end</a>(sbyte** %aq)
2799 ; Stop processing of arguments.
2800 call void %<a href="#i_va_end">llvm.va_end</a>(sbyte** %ap)
2806 <!-- _______________________________________________________________________ -->
2807 <div class="doc_subsubsection">
2808 <a name="i_va_start">'<tt>llvm.va_start</tt>' Intrinsic</a>
2812 <div class="doc_text">
2814 <pre> declare void %llvm.va_start(<va_list>* <arglist>)<br></pre>
2816 <P>The '<tt>llvm.va_start</tt>' intrinsic initializes
2817 <tt>*<arglist></tt> for subsequent use by <tt><a
2818 href="#i_va_arg">va_arg</a></tt>.</p>
2822 <P>The argument is a pointer to a <tt>va_list</tt> element to initialize.</p>
2826 <P>The '<tt>llvm.va_start</tt>' intrinsic works just like the <tt>va_start</tt>
2827 macro available in C. In a target-dependent way, it initializes the
2828 <tt>va_list</tt> element the argument points to, so that the next call to
2829 <tt>va_arg</tt> will produce the first variable argument passed to the function.
2830 Unlike the C <tt>va_start</tt> macro, this intrinsic does not need to know the
2831 last argument of the function, the compiler can figure that out.</p>
2835 <!-- _______________________________________________________________________ -->
2836 <div class="doc_subsubsection">
2837 <a name="i_va_end">'<tt>llvm.va_end</tt>' Intrinsic</a>
2840 <div class="doc_text">
2842 <pre> declare void %llvm.va_end(<va_list*> <arglist>)<br></pre>
2844 <p>The '<tt>llvm.va_end</tt>' intrinsic destroys <tt><arglist></tt>
2845 which has been initialized previously with <tt><a href="#i_va_start">llvm.va_start</a></tt>
2846 or <tt><a href="#i_va_copy">llvm.va_copy</a></tt>.</p>
2848 <p>The argument is a <tt>va_list</tt> to destroy.</p>
2850 <p>The '<tt>llvm.va_end</tt>' intrinsic works just like the <tt>va_end</tt>
2851 macro available in C. In a target-dependent way, it destroys the <tt>va_list</tt>.
2852 Calls to <a href="#i_va_start"><tt>llvm.va_start</tt></a> and <a
2853 href="#i_va_copy"><tt>llvm.va_copy</tt></a> must be matched exactly
2854 with calls to <tt>llvm.va_end</tt>.</p>
2857 <!-- _______________________________________________________________________ -->
2858 <div class="doc_subsubsection">
2859 <a name="i_va_copy">'<tt>llvm.va_copy</tt>' Intrinsic</a>
2862 <div class="doc_text">
2867 declare void %llvm.va_copy(<va_list>* <destarglist>,
2868 <va_list>* <srcarglist>)
2873 <p>The '<tt>llvm.va_copy</tt>' intrinsic copies the current argument position from
2874 the source argument list to the destination argument list.</p>
2878 <p>The first argument is a pointer to a <tt>va_list</tt> element to initialize.
2879 The second argument is a pointer to a <tt>va_list</tt> element to copy from.</p>
2884 <p>The '<tt>llvm.va_copy</tt>' intrinsic works just like the <tt>va_copy</tt> macro
2885 available in C. In a target-dependent way, it copies the source
2886 <tt>va_list</tt> element into the destination list. This intrinsic is necessary
2887 because the <tt><a href="i_va_begin">llvm.va_begin</a></tt> intrinsic may be
2888 arbitrarily complex and require memory allocation, for example.</p>
2892 <!-- ======================================================================= -->
2893 <div class="doc_subsection">
2894 <a name="int_gc">Accurate Garbage Collection Intrinsics</a>
2897 <div class="doc_text">
2900 LLVM support for <a href="GarbageCollection.html">Accurate Garbage
2901 Collection</a> requires the implementation and generation of these intrinsics.
2902 These intrinsics allow identification of <a href="#i_gcroot">GC roots on the
2903 stack</a>, as well as garbage collector implementations that require <a
2904 href="#i_gcread">read</a> and <a href="#i_gcwrite">write</a> barriers.
2905 Front-ends for type-safe garbage collected languages should generate these
2906 intrinsics to make use of the LLVM garbage collectors. For more details, see <a
2907 href="GarbageCollection.html">Accurate Garbage Collection with LLVM</a>.
2911 <!-- _______________________________________________________________________ -->
2912 <div class="doc_subsubsection">
2913 <a name="i_gcroot">'<tt>llvm.gcroot</tt>' Intrinsic</a>
2916 <div class="doc_text">
2921 declare void %llvm.gcroot(<ty>** %ptrloc, <ty2>* %metadata)
2926 <p>The '<tt>llvm.gcroot</tt>' intrinsic declares the existence of a GC root to
2927 the code generator, and allows some metadata to be associated with it.</p>
2931 <p>The first argument specifies the address of a stack object that contains the
2932 root pointer. The second pointer (which must be either a constant or a global
2933 value address) contains the meta-data to be associated with the root.</p>
2937 <p>At runtime, a call to this intrinsics stores a null pointer into the "ptrloc"
2938 location. At compile-time, the code generator generates information to allow
2939 the runtime to find the pointer at GC safe points.
2945 <!-- _______________________________________________________________________ -->
2946 <div class="doc_subsubsection">
2947 <a name="i_gcread">'<tt>llvm.gcread</tt>' Intrinsic</a>
2950 <div class="doc_text">
2955 declare sbyte* %llvm.gcread(sbyte* %ObjPtr, sbyte** %Ptr)
2960 <p>The '<tt>llvm.gcread</tt>' intrinsic identifies reads of references from heap
2961 locations, allowing garbage collector implementations that require read
2966 <p>The second argument is the address to read from, which should be an address
2967 allocated from the garbage collector. The first object is a pointer to the
2968 start of the referenced object, if needed by the language runtime (otherwise
2973 <p>The '<tt>llvm.gcread</tt>' intrinsic has the same semantics as a load
2974 instruction, but may be replaced with substantially more complex code by the
2975 garbage collector runtime, as needed.</p>
2980 <!-- _______________________________________________________________________ -->
2981 <div class="doc_subsubsection">
2982 <a name="i_gcwrite">'<tt>llvm.gcwrite</tt>' Intrinsic</a>
2985 <div class="doc_text">
2990 declare void %llvm.gcwrite(sbyte* %P1, sbyte* %Obj, sbyte** %P2)
2995 <p>The '<tt>llvm.gcwrite</tt>' intrinsic identifies writes of references to heap
2996 locations, allowing garbage collector implementations that require write
2997 barriers (such as generational or reference counting collectors).</p>
3001 <p>The first argument is the reference to store, the second is the start of the
3002 object to store it to, and the third is the address of the field of Obj to
3003 store to. If the runtime does not require a pointer to the object, Obj may be
3008 <p>The '<tt>llvm.gcwrite</tt>' intrinsic has the same semantics as a store
3009 instruction, but may be replaced with substantially more complex code by the
3010 garbage collector runtime, as needed.</p>
3016 <!-- ======================================================================= -->
3017 <div class="doc_subsection">
3018 <a name="int_codegen">Code Generator Intrinsics</a>
3021 <div class="doc_text">
3023 These intrinsics are provided by LLVM to expose special features that may only
3024 be implemented with code generator support.
3029 <!-- _______________________________________________________________________ -->
3030 <div class="doc_subsubsection">
3031 <a name="i_returnaddress">'<tt>llvm.returnaddress</tt>' Intrinsic</a>
3034 <div class="doc_text">
3038 declare sbyte *%llvm.returnaddress(uint <level>)
3044 The '<tt>llvm.returnaddress</tt>' intrinsic returns a target-specific value
3045 indicating the return address of the current function or one of its callers.
3051 The argument to this intrinsic indicates which function to return the address
3052 for. Zero indicates the calling function, one indicates its caller, etc. The
3053 argument is <b>required</b> to be a constant integer value.
3059 The '<tt>llvm.returnaddress</tt>' intrinsic either returns a pointer indicating
3060 the return address of the specified call frame, or zero if it cannot be
3061 identified. The value returned by this intrinsic is likely to be incorrect or 0
3062 for arguments other than zero, so it should only be used for debugging purposes.
3066 Note that calling this intrinsic does not prevent function inlining or other
3067 aggressive transformations, so the value returned may not be that of the obvious
3068 source-language caller.
3073 <!-- _______________________________________________________________________ -->
3074 <div class="doc_subsubsection">
3075 <a name="i_frameaddress">'<tt>llvm.frameaddress</tt>' Intrinsic</a>
3078 <div class="doc_text">
3082 declare sbyte *%llvm.frameaddress(uint <level>)
3088 The '<tt>llvm.frameaddress</tt>' intrinsic returns the target-specific frame
3089 pointer value for the specified stack frame.
3095 The argument to this intrinsic indicates which function to return the frame
3096 pointer for. Zero indicates the calling function, one indicates its caller,
3097 etc. The argument is <b>required</b> to be a constant integer value.
3103 The '<tt>llvm.frameaddress</tt>' intrinsic either returns a pointer indicating
3104 the frame address of the specified call frame, or zero if it cannot be
3105 identified. The value returned by this intrinsic is likely to be incorrect or 0
3106 for arguments other than zero, so it should only be used for debugging purposes.
3110 Note that calling this intrinsic does not prevent function inlining or other
3111 aggressive transformations, so the value returned may not be that of the obvious
3112 source-language caller.
3116 <!-- _______________________________________________________________________ -->
3117 <div class="doc_subsubsection">
3118 <a name="i_stacksave">'<tt>llvm.stacksave</tt>' Intrinsic</a>
3121 <div class="doc_text">
3125 declare sbyte *%llvm.stacksave()
3131 The '<tt>llvm.stacksave</tt>' intrinsic is used to remember the current state of
3132 the function stack, for use with <a href="#i_stackrestore">
3133 <tt>llvm.stackrestore</tt></a>. This is useful for implementing language
3134 features like scoped automatic variable sized arrays in C99.
3140 This intrinsic returns a opaque pointer value that can be passed to <a
3141 href="#i_stackrestore"><tt>llvm.stackrestore</tt></a>. When an
3142 <tt>llvm.stackrestore</tt> intrinsic is executed with a value saved from
3143 <tt>llvm.stacksave</tt>, it effectively restores the state of the stack to the
3144 state it was in when the <tt>llvm.stacksave</tt> intrinsic executed. In
3145 practice, this pops any <a href="#i_alloca">alloca</a> blocks from the stack
3146 that were allocated after the <tt>llvm.stacksave</tt> was executed.
3151 <!-- _______________________________________________________________________ -->
3152 <div class="doc_subsubsection">
3153 <a name="i_stackrestore">'<tt>llvm.stackrestore</tt>' Intrinsic</a>
3156 <div class="doc_text">
3160 declare void %llvm.stackrestore(sbyte* %ptr)
3166 The '<tt>llvm.stackrestore</tt>' intrinsic is used to restore the state of
3167 the function stack to the state it was in when the corresponding <a
3168 href="#llvm.stacksave"><tt>llvm.stacksave</tt></a> intrinsic executed. This is
3169 useful for implementing language features like scoped automatic variable sized
3176 See the description for <a href="#i_stacksave"><tt>llvm.stacksave</tt></a>.
3182 <!-- _______________________________________________________________________ -->
3183 <div class="doc_subsubsection">
3184 <a name="i_prefetch">'<tt>llvm.prefetch</tt>' Intrinsic</a>
3187 <div class="doc_text">
3191 declare void %llvm.prefetch(sbyte * <address>,
3192 uint <rw>, uint <locality>)
3199 The '<tt>llvm.prefetch</tt>' intrinsic is a hint to the code generator to insert
3200 a prefetch instruction if supported; otherwise, it is a noop. Prefetches have
3202 effect on the behavior of the program but can change its performance
3209 <tt>address</tt> is the address to be prefetched, <tt>rw</tt> is the specifier
3210 determining if the fetch should be for a read (0) or write (1), and
3211 <tt>locality</tt> is a temporal locality specifier ranging from (0) - no
3212 locality, to (3) - extremely local keep in cache. The <tt>rw</tt> and
3213 <tt>locality</tt> arguments must be constant integers.
3219 This intrinsic does not modify the behavior of the program. In particular,
3220 prefetches cannot trap and do not produce a value. On targets that support this
3221 intrinsic, the prefetch can provide hints to the processor cache for better
3227 <!-- _______________________________________________________________________ -->
3228 <div class="doc_subsubsection">
3229 <a name="i_pcmarker">'<tt>llvm.pcmarker</tt>' Intrinsic</a>
3232 <div class="doc_text">
3236 declare void %llvm.pcmarker( uint <id> )
3243 The '<tt>llvm.pcmarker</tt>' intrinsic is a method to export a Program Counter
3245 code to simulators and other tools. The method is target specific, but it is
3246 expected that the marker will use exported symbols to transmit the PC of the marker.
3247 The marker makes no guarantees that it will remain with any specific instruction
3248 after optimizations. It is possible that the presence of a marker will inhibit
3249 optimizations. The intended use is to be inserted after optimizations to allow
3250 correlations of simulation runs.
3256 <tt>id</tt> is a numerical id identifying the marker.
3262 This intrinsic does not modify the behavior of the program. Backends that do not
3263 support this intrinisic may ignore it.
3268 <!-- _______________________________________________________________________ -->
3269 <div class="doc_subsubsection">
3270 <a name="i_readcyclecounter">'<tt>llvm.readcyclecounter</tt>' Intrinsic</a>
3273 <div class="doc_text">
3277 declare ulong %llvm.readcyclecounter( )
3284 The '<tt>llvm.readcyclecounter</tt>' intrinsic provides access to the cycle
3285 counter register (or similar low latency, high accuracy clocks) on those targets
3286 that support it. On X86, it should map to RDTSC. On Alpha, it should map to RPCC.
3287 As the backing counters overflow quickly (on the order of 9 seconds on alpha), this
3288 should only be used for small timings.
3294 When directly supported, reading the cycle counter should not modify any memory.
3295 Implementations are allowed to either return a application specific value or a
3296 system wide value. On backends without support, this is lowered to a constant 0.
3301 <!-- ======================================================================= -->
3302 <div class="doc_subsection">
3303 <a name="int_libc">Standard C Library Intrinsics</a>
3306 <div class="doc_text">
3308 LLVM provides intrinsics for a few important standard C library functions.
3309 These intrinsics allow source-language front-ends to pass information about the
3310 alignment of the pointer arguments to the code generator, providing opportunity
3311 for more efficient code generation.
3316 <!-- _______________________________________________________________________ -->
3317 <div class="doc_subsubsection">
3318 <a name="i_memcpy">'<tt>llvm.memcpy</tt>' Intrinsic</a>
3321 <div class="doc_text">
3325 declare void %llvm.memcpy.i32(sbyte* <dest>, sbyte* <src>,
3326 uint <len>, uint <align>)
3327 declare void %llvm.memcpy.i64(sbyte* <dest>, sbyte* <src>,
3328 ulong <len>, uint <align>)
3334 The '<tt>llvm.memcpy.*</tt>' intrinsics copy a block of memory from the source
3335 location to the destination location.
3339 Note that, unlike the standard libc function, the <tt>llvm.memcpy.*</tt>
3340 intrinsics do not return a value, and takes an extra alignment argument.
3346 The first argument is a pointer to the destination, the second is a pointer to
3347 the source. The third argument is an integer argument
3348 specifying the number of bytes to copy, and the fourth argument is the alignment
3349 of the source and destination locations.
3353 If the call to this intrinisic has an alignment value that is not 0 or 1, then
3354 the caller guarantees that both the source and destination pointers are aligned
3361 The '<tt>llvm.memcpy.*</tt>' intrinsics copy a block of memory from the source
3362 location to the destination location, which are not allowed to overlap. It
3363 copies "len" bytes of memory over. If the argument is known to be aligned to
3364 some boundary, this can be specified as the fourth argument, otherwise it should
3370 <!-- _______________________________________________________________________ -->
3371 <div class="doc_subsubsection">
3372 <a name="i_memmove">'<tt>llvm.memmove</tt>' Intrinsic</a>
3375 <div class="doc_text">
3379 declare void %llvm.memmove.i32(sbyte* <dest>, sbyte* <src>,
3380 uint <len>, uint <align>)
3381 declare void %llvm.memmove.i64(sbyte* <dest>, sbyte* <src>,
3382 ulong <len>, uint <align>)
3388 The '<tt>llvm.memmove.*</tt>' intrinsics move a block of memory from the source
3389 location to the destination location. It is similar to the
3390 '<tt>llvm.memcmp</tt>' intrinsic but allows the two memory locations to overlap.
3394 Note that, unlike the standard libc function, the <tt>llvm.memmove.*</tt>
3395 intrinsics do not return a value, and takes an extra alignment argument.
3401 The first argument is a pointer to the destination, the second is a pointer to
3402 the source. The third argument is an integer argument
3403 specifying the number of bytes to copy, and the fourth argument is the alignment
3404 of the source and destination locations.
3408 If the call to this intrinisic has an alignment value that is not 0 or 1, then
3409 the caller guarantees that the source and destination pointers are aligned to
3416 The '<tt>llvm.memmove.*</tt>' intrinsics copy a block of memory from the source
3417 location to the destination location, which may overlap. It
3418 copies "len" bytes of memory over. If the argument is known to be aligned to
3419 some boundary, this can be specified as the fourth argument, otherwise it should
3425 <!-- _______________________________________________________________________ -->
3426 <div class="doc_subsubsection">
3427 <a name="i_memset">'<tt>llvm.memset.*</tt>' Intrinsics</a>
3430 <div class="doc_text">
3434 declare void %llvm.memset.i32(sbyte* <dest>, ubyte <val>,
3435 uint <len>, uint <align>)
3436 declare void %llvm.memset.i64(sbyte* <dest>, ubyte <val>,
3437 ulong <len>, uint <align>)
3443 The '<tt>llvm.memset.*</tt>' intrinsics fill a block of memory with a particular
3448 Note that, unlike the standard libc function, the <tt>llvm.memset</tt> intrinsic
3449 does not return a value, and takes an extra alignment argument.
3455 The first argument is a pointer to the destination to fill, the second is the
3456 byte value to fill it with, the third argument is an integer
3457 argument specifying the number of bytes to fill, and the fourth argument is the
3458 known alignment of destination location.
3462 If the call to this intrinisic has an alignment value that is not 0 or 1, then
3463 the caller guarantees that the destination pointer is aligned to that boundary.
3469 The '<tt>llvm.memset.*</tt>' intrinsics fill "len" bytes of memory starting at
3471 destination location. If the argument is known to be aligned to some boundary,
3472 this can be specified as the fourth argument, otherwise it should be set to 0 or
3478 <!-- _______________________________________________________________________ -->
3479 <div class="doc_subsubsection">
3480 <a name="i_isunordered">'<tt>llvm.isunordered.*</tt>' Intrinsic</a>
3483 <div class="doc_text">
3487 declare bool %llvm.isunordered.f32(float Val1, float Val2)
3488 declare bool %llvm.isunordered.f64(double Val1, double Val2)
3494 The '<tt>llvm.isunordered</tt>' intrinsics return true if either or both of the
3495 specified floating point values is a NAN.
3501 The arguments are floating point numbers of the same type.
3507 If either or both of the arguments is a SNAN or QNAN, it returns true, otherwise
3513 <!-- _______________________________________________________________________ -->
3514 <div class="doc_subsubsection">
3515 <a name="i_sqrt">'<tt>llvm.sqrt.*</tt>' Intrinsic</a>
3518 <div class="doc_text">
3522 declare double %llvm.sqrt.f32(float Val)
3523 declare double %llvm.sqrt.f64(double Val)
3529 The '<tt>llvm.sqrt</tt>' intrinsics return the sqrt of the specified operand,
3530 returning the same value as the libm '<tt>sqrt</tt>' function would. Unlike
3531 <tt>sqrt</tt> in libm, however, <tt>llvm.sqrt</tt> has undefined behavior for
3532 negative numbers (which allows for better optimization).
3538 The argument and return value are floating point numbers of the same type.
3544 This function returns the sqrt of the specified operand if it is a positive
3545 floating point number.
3549 <!-- ======================================================================= -->
3550 <div class="doc_subsection">
3551 <a name="int_manip">Bit Manipulation Intrinsics</a>
3554 <div class="doc_text">
3556 LLVM provides intrinsics for a few important bit manipulation operations.
3557 These allow efficient code generation for some algorithms.
3562 <!-- _______________________________________________________________________ -->
3563 <div class="doc_subsubsection">
3564 <a name="i_bswap">'<tt>llvm.bswap.*</tt>' Intrinsics</a>
3567 <div class="doc_text">
3571 declare ushort %llvm.bswap.i16(ushort <id>)
3572 declare uint %llvm.bswap.i32(uint <id>)
3573 declare ulong %llvm.bswap.i64(ulong <id>)
3579 The '<tt>llvm.bwsap</tt>' family of intrinsics is used to byteswap a 16, 32 or
3580 64 bit quantity. These are useful for performing operations on data that is not
3581 in the target's native byte order.
3587 The <tt>llvm.bswap.16</tt> intrinsic returns a ushort value that has the high and low
3588 byte of the input ushort swapped. Similarly, the <tt>llvm.bswap.i32</tt> intrinsic
3589 returns a uint value that has the four bytes of the input uint swapped, so that
3590 if the input bytes are numbered 0, 1, 2, 3 then the returned uint will have its
3591 bytes in 3, 2, 1, 0 order. The <tt>llvm.bswap.i64</tt> intrinsic extends this concept
3597 <!-- _______________________________________________________________________ -->
3598 <div class="doc_subsubsection">
3599 <a name="int_ctpop">'<tt>llvm.ctpop.*</tt>' Intrinsic</a>
3602 <div class="doc_text">
3606 declare ubyte %llvm.ctpop.i8 (ubyte <src>)
3607 declare ushort %llvm.ctpop.i16(ushort <src>)
3608 declare uint %llvm.ctpop.i32(uint <src>)
3609 declare ulong %llvm.ctpop.i64(ulong <src>)
3615 The '<tt>llvm.ctpop</tt>' family of intrinsics counts the number of bits set in a
3622 The only argument is the value to be counted. The argument may be of any
3623 unsigned integer type. The return type must match the argument type.
3629 The '<tt>llvm.ctpop</tt>' intrinsic counts the 1's in a variable.
3633 <!-- _______________________________________________________________________ -->
3634 <div class="doc_subsubsection">
3635 <a name="int_ctlz">'<tt>llvm.ctlz.*</tt>' Intrinsic</a>
3638 <div class="doc_text">
3642 declare ubyte %llvm.ctlz.i8 (ubyte <src>)
3643 declare ushort %llvm.ctlz.i16(ushort <src>)
3644 declare uint %llvm.ctlz.i32(uint <src>)
3645 declare ulong %llvm.ctlz.i64(ulong <src>)
3651 The '<tt>llvm.ctlz</tt>' family of intrinsic functions counts the number of
3652 leading zeros in a variable.
3658 The only argument is the value to be counted. The argument may be of any
3659 unsigned integer type. The return type must match the argument type.
3665 The '<tt>llvm.ctlz</tt>' intrinsic counts the leading (most significant) zeros
3666 in a variable. If the src == 0 then the result is the size in bits of the type
3667 of src. For example, <tt>llvm.cttz(int 2) = 30</tt>.
3673 <!-- _______________________________________________________________________ -->
3674 <div class="doc_subsubsection">
3675 <a name="int_cttz">'<tt>llvm.cttz.*</tt>' Intrinsic</a>
3678 <div class="doc_text">
3682 declare ubyte %llvm.cttz.i8 (ubyte <src>)
3683 declare ushort %llvm.cttz.i16(ushort <src>)
3684 declare uint %llvm.cttz.i32(uint <src>)
3685 declare ulong %llvm.cttz.i64(ulong <src>)
3691 The '<tt>llvm.cttz</tt>' family of intrinsic functions counts the number of
3698 The only argument is the value to be counted. The argument may be of any
3699 unsigned integer type. The return type must match the argument type.
3705 The '<tt>llvm.cttz</tt>' intrinsic counts the trailing (least significant) zeros
3706 in a variable. If the src == 0 then the result is the size in bits of the type
3707 of src. For example, <tt>llvm.cttz(2) = 1</tt>.
3711 <!-- ======================================================================= -->
3712 <div class="doc_subsection">
3713 <a name="int_debugger">Debugger Intrinsics</a>
3716 <div class="doc_text">
3718 The LLVM debugger intrinsics (which all start with <tt>llvm.dbg.</tt> prefix),
3719 are described in the <a
3720 href="SourceLevelDebugging.html#format_common_intrinsics">LLVM Source Level
3721 Debugging</a> document.
3726 <!-- *********************************************************************** -->
3729 <a href="http://jigsaw.w3.org/css-validator/check/referer"><img
3730 src="http://jigsaw.w3.org/css-validator/images/vcss" alt="Valid CSS!"></a>
3731 <a href="http://validator.w3.org/check/referer"><img
3732 src="http://www.w3.org/Icons/valid-html401" alt="Valid HTML 4.01!" /></a>
3734 <a href="mailto:sabre@nondot.org">Chris Lattner</a><br>
3735 <a href="http://llvm.org">The LLVM Compiler Infrastructure</a><br>
3736 Last modified: $Date$