1 <!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01//EN"
2 "http://www.w3.org/TR/html4/strict.dtd">
5 <title>LLVM Assembly Language Reference Manual</title>
6 <meta http-equiv="Content-Type" content="text/html; charset=utf-8">
7 <meta name="author" content="Chris Lattner">
8 <meta name="description"
9 content="LLVM Assembly Language Reference Manual.">
10 <link rel="stylesheet" href="llvm.css" type="text/css">
15 <div class="doc_title"> LLVM Language Reference Manual </div>
17 <li><a href="#abstract">Abstract</a></li>
18 <li><a href="#introduction">Introduction</a></li>
19 <li><a href="#identifiers">Identifiers</a></li>
20 <li><a href="#highlevel">High Level Structure</a>
22 <li><a href="#modulestructure">Module Structure</a></li>
23 <li><a href="#linkage">Linkage Types</a></li>
24 <li><a href="#callingconv">Calling Conventions</a></li>
25 <li><a href="#globalvars">Global Variables</a></li>
26 <li><a href="#functionstructure">Functions</a></li>
27 <li><a href="#moduleasm">Module-Level Inline Assembly</a></li>
30 <li><a href="#typesystem">Type System</a>
32 <li><a href="#t_primitive">Primitive Types</a>
34 <li><a href="#t_classifications">Type Classifications</a></li>
37 <li><a href="#t_derived">Derived Types</a>
39 <li><a href="#t_array">Array Type</a></li>
40 <li><a href="#t_function">Function Type</a></li>
41 <li><a href="#t_pointer">Pointer Type</a></li>
42 <li><a href="#t_struct">Structure Type</a></li>
43 <li><a href="#t_packed">Packed Type</a></li>
44 <li><a href="#t_opaque">Opaque Type</a></li>
49 <li><a href="#constants">Constants</a>
51 <li><a href="#simpleconstants">Simple Constants</a>
52 <li><a href="#aggregateconstants">Aggregate Constants</a>
53 <li><a href="#globalconstants">Global Variable and Function Addresses</a>
54 <li><a href="#undefvalues">Undefined Values</a>
55 <li><a href="#constantexprs">Constant Expressions</a>
58 <li><a href="#othervalues">Other Values</a>
60 <li><a href="#inlineasm">Inline Assembler Expressions</a>
63 <li><a href="#instref">Instruction Reference</a>
65 <li><a href="#terminators">Terminator Instructions</a>
67 <li><a href="#i_ret">'<tt>ret</tt>' Instruction</a></li>
68 <li><a href="#i_br">'<tt>br</tt>' Instruction</a></li>
69 <li><a href="#i_switch">'<tt>switch</tt>' Instruction</a></li>
70 <li><a href="#i_invoke">'<tt>invoke</tt>' Instruction</a></li>
71 <li><a href="#i_unwind">'<tt>unwind</tt>' Instruction</a></li>
72 <li><a href="#i_unreachable">'<tt>unreachable</tt>' Instruction</a></li>
75 <li><a href="#binaryops">Binary Operations</a>
77 <li><a href="#i_add">'<tt>add</tt>' Instruction</a></li>
78 <li><a href="#i_sub">'<tt>sub</tt>' Instruction</a></li>
79 <li><a href="#i_mul">'<tt>mul</tt>' Instruction</a></li>
80 <li><a href="#i_udiv">'<tt>udiv</tt>' Instruction</a></li>
81 <li><a href="#i_sdiv">'<tt>sdiv</tt>' Instruction</a></li>
82 <li><a href="#i_fdiv">'<tt>fdiv</tt>' Instruction</a></li>
83 <li><a href="#i_urem">'<tt>urem</tt>' Instruction</a></li>
84 <li><a href="#i_srem">'<tt>srem</tt>' Instruction</a></li>
85 <li><a href="#i_frem">'<tt>frem</tt>' Instruction</a></li>
88 <li><a href="#bitwiseops">Bitwise Binary Operations</a>
90 <li><a href="#i_and">'<tt>and</tt>' Instruction</a></li>
91 <li><a href="#i_or">'<tt>or</tt>' Instruction</a></li>
92 <li><a href="#i_xor">'<tt>xor</tt>' Instruction</a></li>
93 <li><a href="#i_shl">'<tt>shl</tt>' Instruction</a></li>
94 <li><a href="#i_lshr">'<tt>lshr</tt>' Instruction</a></li>
95 <li><a href="#i_ashr">'<tt>ashr</tt>' Instruction</a></li>
98 <li><a href="#vectorops">Vector Operations</a>
100 <li><a href="#i_extractelement">'<tt>extractelement</tt>' Instruction</a></li>
101 <li><a href="#i_insertelement">'<tt>insertelement</tt>' Instruction</a></li>
102 <li><a href="#i_shufflevector">'<tt>shufflevector</tt>' Instruction</a></li>
105 <li><a href="#memoryops">Memory Access and Addressing Operations</a>
107 <li><a href="#i_malloc">'<tt>malloc</tt>' Instruction</a></li>
108 <li><a href="#i_free">'<tt>free</tt>' Instruction</a></li>
109 <li><a href="#i_alloca">'<tt>alloca</tt>' Instruction</a></li>
110 <li><a href="#i_load">'<tt>load</tt>' Instruction</a></li>
111 <li><a href="#i_store">'<tt>store</tt>' Instruction</a></li>
112 <li><a href="#i_getelementptr">'<tt>getelementptr</tt>' Instruction</a></li>
115 <li><a href="#convertops">Conversion Operations</a>
117 <li><a href="#i_trunc">'<tt>trunc .. to</tt>' Instruction</a></li>
118 <li><a href="#i_zext">'<tt>zext .. to</tt>' Instruction</a></li>
119 <li><a href="#i_sext">'<tt>sext .. to</tt>' Instruction</a></li>
120 <li><a href="#i_fptrunc">'<tt>fptrunc .. to</tt>' Instruction</a></li>
121 <li><a href="#i_fpext">'<tt>fpext .. to</tt>' Instruction</a></li>
122 <li><a href="#i_fptoui">'<tt>fptoui .. to</tt>' Instruction</a></li>
123 <li><a href="#i_fptosi">'<tt>fptosi .. to</tt>' Instruction</a></li>
124 <li><a href="#i_uitofp">'<tt>uitofp .. to</tt>' Instruction</a></li>
125 <li><a href="#i_sitofp">'<tt>sitofp .. to</tt>' Instruction</a></li>
126 <li><a href="#i_ptrtoint">'<tt>ptrtoint .. to</tt>' Instruction</a></li>
127 <li><a href="#i_inttoptr">'<tt>inttoptr .. to</tt>' Instruction</a></li>
128 <li><a href="#i_bitcast">'<tt>bitcast .. to</tt>' Instruction</a></li>
130 <li><a href="#otherops">Other Operations</a>
132 <li><a href="#i_icmp">'<tt>icmp</tt>' Instruction</a></li>
133 <li><a href="#i_fcmp">'<tt>fcmp</tt>' Instruction</a></li>
134 <li><a href="#i_phi">'<tt>phi</tt>' Instruction</a></li>
135 <li><a href="#i_select">'<tt>select</tt>' Instruction</a></li>
136 <li><a href="#i_call">'<tt>call</tt>' Instruction</a></li>
137 <li><a href="#i_va_arg">'<tt>va_arg</tt>' Instruction</a></li>
142 <li><a href="#intrinsics">Intrinsic Functions</a>
144 <li><a href="#int_varargs">Variable Argument Handling Intrinsics</a>
146 <li><a href="#i_va_start">'<tt>llvm.va_start</tt>' Intrinsic</a></li>
147 <li><a href="#i_va_end">'<tt>llvm.va_end</tt>' Intrinsic</a></li>
148 <li><a href="#i_va_copy">'<tt>llvm.va_copy</tt>' Intrinsic</a></li>
151 <li><a href="#int_gc">Accurate Garbage Collection Intrinsics</a>
153 <li><a href="#i_gcroot">'<tt>llvm.gcroot</tt>' Intrinsic</a></li>
154 <li><a href="#i_gcread">'<tt>llvm.gcread</tt>' Intrinsic</a></li>
155 <li><a href="#i_gcwrite">'<tt>llvm.gcwrite</tt>' Intrinsic</a></li>
158 <li><a href="#int_codegen">Code Generator Intrinsics</a>
160 <li><a href="#i_returnaddress">'<tt>llvm.returnaddress</tt>' Intrinsic</a></li>
161 <li><a href="#i_frameaddress">'<tt>llvm.frameaddress</tt>' Intrinsic</a></li>
162 <li><a href="#i_stacksave">'<tt>llvm.stacksave</tt>' Intrinsic</a></li>
163 <li><a href="#i_stackrestore">'<tt>llvm.stackrestore</tt>' Intrinsic</a></li>
164 <li><a href="#i_prefetch">'<tt>llvm.prefetch</tt>' Intrinsic</a></li>
165 <li><a href="#i_pcmarker">'<tt>llvm.pcmarker</tt>' Intrinsic</a></li>
166 <li><a href="#i_readcyclecounter"><tt>llvm.readcyclecounter</tt>' Intrinsic</a></li>
169 <li><a href="#int_libc">Standard C Library Intrinsics</a>
171 <li><a href="#i_memcpy">'<tt>llvm.memcpy.*</tt>' Intrinsic</a></li>
172 <li><a href="#i_memmove">'<tt>llvm.memmove.*</tt>' Intrinsic</a></li>
173 <li><a href="#i_memset">'<tt>llvm.memset.*</tt>' Intrinsic</a></li>
174 <li><a href="#i_isunordered">'<tt>llvm.isunordered.*</tt>' Intrinsic</a></li>
175 <li><a href="#i_sqrt">'<tt>llvm.sqrt.*</tt>' Intrinsic</a></li>
176 <li><a href="#i_powi">'<tt>llvm.powi.*</tt>' Intrinsic</a></li>
179 <li><a href="#int_manip">Bit Manipulation Intrinsics</a>
181 <li><a href="#i_bswap">'<tt>llvm.bswap.*</tt>' Intrinsics</a></li>
182 <li><a href="#int_ctpop">'<tt>llvm.ctpop.*</tt>' Intrinsic </a></li>
183 <li><a href="#int_ctlz">'<tt>llvm.ctlz.*</tt>' Intrinsic </a></li>
184 <li><a href="#int_cttz">'<tt>llvm.cttz.*</tt>' Intrinsic </a></li>
187 <li><a href="#int_debugger">Debugger intrinsics</a></li>
192 <div class="doc_author">
193 <p>Written by <a href="mailto:sabre@nondot.org">Chris Lattner</a>
194 and <a href="mailto:vadve@cs.uiuc.edu">Vikram Adve</a></p>
197 <!-- *********************************************************************** -->
198 <div class="doc_section"> <a name="abstract">Abstract </a></div>
199 <!-- *********************************************************************** -->
201 <div class="doc_text">
202 <p>This document is a reference manual for the LLVM assembly language.
203 LLVM is an SSA based representation that provides type safety,
204 low-level operations, flexibility, and the capability of representing
205 'all' high-level languages cleanly. It is the common code
206 representation used throughout all phases of the LLVM compilation
210 <!-- *********************************************************************** -->
211 <div class="doc_section"> <a name="introduction">Introduction</a> </div>
212 <!-- *********************************************************************** -->
214 <div class="doc_text">
216 <p>The LLVM code representation is designed to be used in three
217 different forms: as an in-memory compiler IR, as an on-disk bytecode
218 representation (suitable for fast loading by a Just-In-Time compiler),
219 and as a human readable assembly language representation. This allows
220 LLVM to provide a powerful intermediate representation for efficient
221 compiler transformations and analysis, while providing a natural means
222 to debug and visualize the transformations. The three different forms
223 of LLVM are all equivalent. This document describes the human readable
224 representation and notation.</p>
226 <p>The LLVM representation aims to be light-weight and low-level
227 while being expressive, typed, and extensible at the same time. It
228 aims to be a "universal IR" of sorts, by being at a low enough level
229 that high-level ideas may be cleanly mapped to it (similar to how
230 microprocessors are "universal IR's", allowing many source languages to
231 be mapped to them). By providing type information, LLVM can be used as
232 the target of optimizations: for example, through pointer analysis, it
233 can be proven that a C automatic variable is never accessed outside of
234 the current function... allowing it to be promoted to a simple SSA
235 value instead of a memory location.</p>
239 <!-- _______________________________________________________________________ -->
240 <div class="doc_subsubsection"> <a name="wellformed">Well-Formedness</a> </div>
242 <div class="doc_text">
244 <p>It is important to note that this document describes 'well formed'
245 LLVM assembly language. There is a difference between what the parser
246 accepts and what is considered 'well formed'. For example, the
247 following instruction is syntactically okay, but not well formed:</p>
250 %x = <a href="#i_add">add</a> int 1, %x
253 <p>...because the definition of <tt>%x</tt> does not dominate all of
254 its uses. The LLVM infrastructure provides a verification pass that may
255 be used to verify that an LLVM module is well formed. This pass is
256 automatically run by the parser after parsing input assembly and by
257 the optimizer before it outputs bytecode. The violations pointed out
258 by the verifier pass indicate bugs in transformation passes or input to
261 <!-- Describe the typesetting conventions here. --> </div>
263 <!-- *********************************************************************** -->
264 <div class="doc_section"> <a name="identifiers">Identifiers</a> </div>
265 <!-- *********************************************************************** -->
267 <div class="doc_text">
269 <p>LLVM uses three different forms of identifiers, for different
273 <li>Named values are represented as a string of characters with a '%' prefix.
274 For example, %foo, %DivisionByZero, %a.really.long.identifier. The actual
275 regular expression used is '<tt>%[a-zA-Z$._][a-zA-Z$._0-9]*</tt>'.
276 Identifiers which require other characters in their names can be surrounded
277 with quotes. In this way, anything except a <tt>"</tt> character can be used
280 <li>Unnamed values are represented as an unsigned numeric value with a '%'
281 prefix. For example, %12, %2, %44.</li>
283 <li>Constants, which are described in a <a href="#constants">section about
284 constants</a>, below.</li>
287 <p>LLVM requires that values start with a '%' sign for two reasons: Compilers
288 don't need to worry about name clashes with reserved words, and the set of
289 reserved words may be expanded in the future without penalty. Additionally,
290 unnamed identifiers allow a compiler to quickly come up with a temporary
291 variable without having to avoid symbol table conflicts.</p>
293 <p>Reserved words in LLVM are very similar to reserved words in other
294 languages. There are keywords for different opcodes
295 ('<tt><a href="#i_add">add</a></tt>',
296 '<tt><a href="#i_bitcast">bitcast</a></tt>',
297 '<tt><a href="#i_ret">ret</a></tt>', etc...), for primitive type names ('<tt><a
298 href="#t_void">void</a></tt>', '<tt><a href="#t_uint">uint</a></tt>', etc...),
299 and others. These reserved words cannot conflict with variable names, because
300 none of them start with a '%' character.</p>
302 <p>Here is an example of LLVM code to multiply the integer variable
303 '<tt>%X</tt>' by 8:</p>
308 %result = <a href="#i_mul">mul</a> uint %X, 8
311 <p>After strength reduction:</p>
314 %result = <a href="#i_shl">shl</a> uint %X, ubyte 3
317 <p>And the hard way:</p>
320 <a href="#i_add">add</a> uint %X, %X <i>; yields {uint}:%0</i>
321 <a href="#i_add">add</a> uint %0, %0 <i>; yields {uint}:%1</i>
322 %result = <a href="#i_add">add</a> uint %1, %1
325 <p>This last way of multiplying <tt>%X</tt> by 8 illustrates several
326 important lexical features of LLVM:</p>
330 <li>Comments are delimited with a '<tt>;</tt>' and go until the end of
333 <li>Unnamed temporaries are created when the result of a computation is not
334 assigned to a named value.</li>
336 <li>Unnamed temporaries are numbered sequentially</li>
340 <p>...and it also shows a convention that we follow in this document. When
341 demonstrating instructions, we will follow an instruction with a comment that
342 defines the type and name of value produced. Comments are shown in italic
347 <!-- *********************************************************************** -->
348 <div class="doc_section"> <a name="highlevel">High Level Structure</a> </div>
349 <!-- *********************************************************************** -->
351 <!-- ======================================================================= -->
352 <div class="doc_subsection"> <a name="modulestructure">Module Structure</a>
355 <div class="doc_text">
357 <p>LLVM programs are composed of "Module"s, each of which is a
358 translation unit of the input programs. Each module consists of
359 functions, global variables, and symbol table entries. Modules may be
360 combined together with the LLVM linker, which merges function (and
361 global variable) definitions, resolves forward declarations, and merges
362 symbol table entries. Here is an example of the "hello world" module:</p>
364 <pre><i>; Declare the string constant as a global constant...</i>
365 <a href="#identifiers">%.LC0</a> = <a href="#linkage_internal">internal</a> <a
366 href="#globalvars">constant</a> <a href="#t_array">[13 x sbyte]</a> c"hello world\0A\00" <i>; [13 x sbyte]*</i>
368 <i>; External declaration of the puts function</i>
369 <a href="#functionstructure">declare</a> int %puts(sbyte*) <i>; int(sbyte*)* </i>
371 <i>; Global variable / Function body section separator</i>
374 <i>; Definition of main function</i>
375 int %main() { <i>; int()* </i>
376 <i>; Convert [13x sbyte]* to sbyte *...</i>
378 href="#i_getelementptr">getelementptr</a> [13 x sbyte]* %.LC0, long 0, long 0 <i>; sbyte*</i>
380 <i>; Call puts function to write out the string to stdout...</i>
382 href="#i_call">call</a> int %puts(sbyte* %cast210) <i>; int</i>
384 href="#i_ret">ret</a> int 0<br>}<br></pre>
386 <p>This example is made up of a <a href="#globalvars">global variable</a>
387 named "<tt>.LC0</tt>", an external declaration of the "<tt>puts</tt>"
388 function, and a <a href="#functionstructure">function definition</a>
389 for "<tt>main</tt>".</p>
391 <p>In general, a module is made up of a list of global values,
392 where both functions and global variables are global values. Global values are
393 represented by a pointer to a memory location (in this case, a pointer to an
394 array of char, and a pointer to a function), and have one of the following <a
395 href="#linkage">linkage types</a>.</p>
397 <p>Due to a limitation in the current LLVM assembly parser (it is limited by
398 one-token lookahead), modules are split into two pieces by the "implementation"
399 keyword. Global variable prototypes and definitions must occur before the
400 keyword, and function definitions must occur after it. Function prototypes may
401 occur either before or after it. In the future, the implementation keyword may
402 become a noop, if the parser gets smarter.</p>
406 <!-- ======================================================================= -->
407 <div class="doc_subsection">
408 <a name="linkage">Linkage Types</a>
411 <div class="doc_text">
414 All Global Variables and Functions have one of the following types of linkage:
419 <dt><tt><b><a name="linkage_internal">internal</a></b></tt> </dt>
421 <dd>Global values with internal linkage are only directly accessible by
422 objects in the current module. In particular, linking code into a module with
423 an internal global value may cause the internal to be renamed as necessary to
424 avoid collisions. Because the symbol is internal to the module, all
425 references can be updated. This corresponds to the notion of the
426 '<tt>static</tt>' keyword in C, or the idea of "anonymous namespaces" in C++.
429 <dt><tt><b><a name="linkage_linkonce">linkonce</a></b></tt>: </dt>
431 <dd>"<tt>linkonce</tt>" linkage is similar to <tt>internal</tt> linkage, with
432 the twist that linking together two modules defining the same
433 <tt>linkonce</tt> globals will cause one of the globals to be discarded. This
434 is typically used to implement inline functions. Unreferenced
435 <tt>linkonce</tt> globals are allowed to be discarded.
438 <dt><tt><b><a name="linkage_weak">weak</a></b></tt>: </dt>
440 <dd>"<tt>weak</tt>" linkage is exactly the same as <tt>linkonce</tt> linkage,
441 except that unreferenced <tt>weak</tt> globals may not be discarded. This is
442 used to implement constructs in C such as "<tt>int X;</tt>" at global scope.
445 <dt><tt><b><a name="linkage_appending">appending</a></b></tt>: </dt>
447 <dd>"<tt>appending</tt>" linkage may only be applied to global variables of
448 pointer to array type. When two global variables with appending linkage are
449 linked together, the two global arrays are appended together. This is the
450 LLVM, typesafe, equivalent of having the system linker append together
451 "sections" with identical names when .o files are linked.
454 <dt><tt><b><a name="linkage_external">externally visible</a></b></tt>:</dt>
456 <dd>If none of the above identifiers are used, the global is externally
457 visible, meaning that it participates in linkage and can be used to resolve
458 external symbol references.
461 <dt><tt><b><a name="linkage_externweak">extern_weak</a></b></tt>: </dt>
463 <dd>"<tt>extern_weak</tt>" TBD
467 The next two types of linkage are targeted for Microsoft Windows platform
468 only. They are designed to support importing (exporting) symbols from (to)
472 <dt><tt><b><a name="linkage_dllimport">dllimport</a></b></tt>: </dt>
474 <dd>"<tt>dllimport</tt>" linkage causes the compiler to reference a function
475 or variable via a global pointer to a pointer that is set up by the DLL
476 exporting the symbol. On Microsoft Windows targets, the pointer name is
477 formed by combining <code>_imp__</code> and the function or variable name.
480 <dt><tt><b><a name="linkage_dllexport">dllexport</a></b></tt>: </dt>
482 <dd>"<tt>dllexport</tt>" linkage causes the compiler to provide a global
483 pointer to a pointer in a DLL, so that it can be referenced with the
484 <tt>dllimport</tt> attribute. On Microsoft Windows targets, the pointer
485 name is formed by combining <code>_imp__</code> and the function or variable
491 <p><a name="linkage_external">For example, since the "<tt>.LC0</tt>"
492 variable is defined to be internal, if another module defined a "<tt>.LC0</tt>"
493 variable and was linked with this one, one of the two would be renamed,
494 preventing a collision. Since "<tt>main</tt>" and "<tt>puts</tt>" are
495 external (i.e., lacking any linkage declarations), they are accessible
496 outside of the current module. It is illegal for a function <i>declaration</i>
497 to have any linkage type other than "externally visible".</a></p>
501 <!-- ======================================================================= -->
502 <div class="doc_subsection">
503 <a name="callingconv">Calling Conventions</a>
506 <div class="doc_text">
508 <p>LLVM <a href="#functionstructure">functions</a>, <a href="#i_call">calls</a>
509 and <a href="#i_invoke">invokes</a> can all have an optional calling convention
510 specified for the call. The calling convention of any pair of dynamic
511 caller/callee must match, or the behavior of the program is undefined. The
512 following calling conventions are supported by LLVM, and more may be added in
516 <dt><b>"<tt>ccc</tt>" - The C calling convention</b>:</dt>
518 <dd>This calling convention (the default if no other calling convention is
519 specified) matches the target C calling conventions. This calling convention
520 supports varargs function calls and tolerates some mismatch in the declared
521 prototype and implemented declaration of the function (as does normal C).
524 <dt><b>"<tt>csretcc</tt>" - The C struct return calling convention</b>:</dt>
526 <dd>This calling convention matches the target C calling conventions, except
527 that functions with this convention are required to take a pointer as their
528 first argument, and the return type of the function must be void. This is
529 used for C functions that return aggregates by-value. In this case, the
530 function has been transformed to take a pointer to the struct as the first
531 argument to the function. For targets where the ABI specifies specific
532 behavior for structure-return calls, the calling convention can be used to
533 distinguish between struct return functions and other functions that take a
534 pointer to a struct as the first argument.
537 <dt><b>"<tt>fastcc</tt>" - The fast calling convention</b>:</dt>
539 <dd>This calling convention attempts to make calls as fast as possible
540 (e.g. by passing things in registers). This calling convention allows the
541 target to use whatever tricks it wants to produce fast code for the target,
542 without having to conform to an externally specified ABI. Implementations of
543 this convention should allow arbitrary tail call optimization to be supported.
544 This calling convention does not support varargs and requires the prototype of
545 all callees to exactly match the prototype of the function definition.
548 <dt><b>"<tt>coldcc</tt>" - The cold calling convention</b>:</dt>
550 <dd>This calling convention attempts to make code in the caller as efficient
551 as possible under the assumption that the call is not commonly executed. As
552 such, these calls often preserve all registers so that the call does not break
553 any live ranges in the caller side. This calling convention does not support
554 varargs and requires the prototype of all callees to exactly match the
555 prototype of the function definition.
558 <dt><b>"<tt>cc <<em>n</em>></tt>" - Numbered convention</b>:</dt>
560 <dd>Any calling convention may be specified by number, allowing
561 target-specific calling conventions to be used. Target specific calling
562 conventions start at 64.
566 <p>More calling conventions can be added/defined on an as-needed basis, to
567 support pascal conventions or any other well-known target-independent
572 <!-- ======================================================================= -->
573 <div class="doc_subsection">
574 <a name="globalvars">Global Variables</a>
577 <div class="doc_text">
579 <p>Global variables define regions of memory allocated at compilation time
580 instead of run-time. Global variables may optionally be initialized, may have
581 an explicit section to be placed in, and may
582 have an optional explicit alignment specified. A
583 variable may be defined as a global "constant," which indicates that the
584 contents of the variable will <b>never</b> be modified (enabling better
585 optimization, allowing the global data to be placed in the read-only section of
586 an executable, etc). Note that variables that need runtime initialization
587 cannot be marked "constant" as there is a store to the variable.</p>
590 LLVM explicitly allows <em>declarations</em> of global variables to be marked
591 constant, even if the final definition of the global is not. This capability
592 can be used to enable slightly better optimization of the program, but requires
593 the language definition to guarantee that optimizations based on the
594 'constantness' are valid for the translation units that do not include the
598 <p>As SSA values, global variables define pointer values that are in
599 scope (i.e. they dominate) all basic blocks in the program. Global
600 variables always define a pointer to their "content" type because they
601 describe a region of memory, and all memory objects in LLVM are
602 accessed through pointers.</p>
604 <p>LLVM allows an explicit section to be specified for globals. If the target
605 supports it, it will emit globals to the section specified.</p>
607 <p>An explicit alignment may be specified for a global. If not present, or if
608 the alignment is set to zero, the alignment of the global is set by the target
609 to whatever it feels convenient. If an explicit alignment is specified, the
610 global is forced to have at least that much alignment. All alignments must be
616 <!-- ======================================================================= -->
617 <div class="doc_subsection">
618 <a name="functionstructure">Functions</a>
621 <div class="doc_text">
623 <p>LLVM function definitions consist of an optional <a href="#linkage">linkage
624 type</a>, an optional <a href="#callingconv">calling convention</a>, a return
625 type, a function name, a (possibly empty) argument list, an optional section,
626 an optional alignment, an opening curly brace,
627 a list of basic blocks, and a closing curly brace. LLVM function declarations
628 are defined with the "<tt>declare</tt>" keyword, an optional <a
629 href="#callingconv">calling convention</a>, a return type, a function name,
630 a possibly empty list of arguments, and an optional alignment.</p>
632 <p>A function definition contains a list of basic blocks, forming the CFG for
633 the function. Each basic block may optionally start with a label (giving the
634 basic block a symbol table entry), contains a list of instructions, and ends
635 with a <a href="#terminators">terminator</a> instruction (such as a branch or
636 function return).</p>
638 <p>The first basic block in a program is special in two ways: it is immediately
639 executed on entrance to the function, and it is not allowed to have predecessor
640 basic blocks (i.e. there can not be any branches to the entry block of a
641 function). Because the block can have no predecessors, it also cannot have any
642 <a href="#i_phi">PHI nodes</a>.</p>
644 <p>LLVM functions are identified by their name and type signature. Hence, two
645 functions with the same name but different parameter lists or return values are
646 considered different functions, and LLVM will resolve references to each
649 <p>LLVM allows an explicit section to be specified for functions. If the target
650 supports it, it will emit functions to the section specified.</p>
652 <p>An explicit alignment may be specified for a function. If not present, or if
653 the alignment is set to zero, the alignment of the function is set by the target
654 to whatever it feels convenient. If an explicit alignment is specified, the
655 function is forced to have at least that much alignment. All alignments must be
660 <!-- ======================================================================= -->
661 <div class="doc_subsection">
662 <a name="moduleasm">Module-Level Inline Assembly</a>
665 <div class="doc_text">
667 Modules may contain "module-level inline asm" blocks, which corresponds to the
668 GCC "file scope inline asm" blocks. These blocks are internally concatenated by
669 LLVM and treated as a single unit, but may be separated in the .ll file if
670 desired. The syntax is very simple:
673 <div class="doc_code"><pre>
674 module asm "inline asm code goes here"
675 module asm "more can go here"
678 <p>The strings can contain any character by escaping non-printable characters.
679 The escape sequence used is simply "\xx" where "xx" is the two digit hex code
684 The inline asm code is simply printed to the machine code .s file when
685 assembly code is generated.
690 <!-- *********************************************************************** -->
691 <div class="doc_section"> <a name="typesystem">Type System</a> </div>
692 <!-- *********************************************************************** -->
694 <div class="doc_text">
696 <p>The LLVM type system is one of the most important features of the
697 intermediate representation. Being typed enables a number of
698 optimizations to be performed on the IR directly, without having to do
699 extra analyses on the side before the transformation. A strong type
700 system makes it easier to read the generated code and enables novel
701 analyses and transformations that are not feasible to perform on normal
702 three address code representations.</p>
706 <!-- ======================================================================= -->
707 <div class="doc_subsection"> <a name="t_primitive">Primitive Types</a> </div>
708 <div class="doc_text">
709 <p>The primitive types are the fundamental building blocks of the LLVM
710 system. The current set of primitive types is as follows:</p>
712 <table class="layout">
717 <tr><th>Type</th><th>Description</th></tr>
718 <tr><td><tt>void</tt></td><td>No value</td></tr>
719 <tr><td><tt>ubyte</tt></td><td>Unsigned 8-bit value</td></tr>
720 <tr><td><tt>ushort</tt></td><td>Unsigned 16-bit value</td></tr>
721 <tr><td><tt>uint</tt></td><td>Unsigned 32-bit value</td></tr>
722 <tr><td><tt>ulong</tt></td><td>Unsigned 64-bit value</td></tr>
723 <tr><td><tt>float</tt></td><td>32-bit floating point value</td></tr>
724 <tr><td><tt>label</tt></td><td>Branch destination</td></tr>
731 <tr><th>Type</th><th>Description</th></tr>
732 <tr><td><tt>bool</tt></td><td>True or False value</td></tr>
733 <tr><td><tt>sbyte</tt></td><td>Signed 8-bit value</td></tr>
734 <tr><td><tt>short</tt></td><td>Signed 16-bit value</td></tr>
735 <tr><td><tt>int</tt></td><td>Signed 32-bit value</td></tr>
736 <tr><td><tt>long</tt></td><td>Signed 64-bit value</td></tr>
737 <tr><td><tt>double</tt></td><td>64-bit floating point value</td></tr>
745 <!-- _______________________________________________________________________ -->
746 <div class="doc_subsubsection"> <a name="t_classifications">Type
747 Classifications</a> </div>
748 <div class="doc_text">
749 <p>These different primitive types fall into a few useful
752 <table border="1" cellspacing="0" cellpadding="4">
754 <tr><th>Classification</th><th>Types</th></tr>
756 <td><a name="t_signed">signed</a></td>
757 <td><tt>sbyte, short, int, long, float, double</tt></td>
760 <td><a name="t_unsigned">unsigned</a></td>
761 <td><tt>ubyte, ushort, uint, ulong</tt></td>
764 <td><a name="t_integer">integer</a></td>
765 <td><tt>ubyte, sbyte, ushort, short, uint, int, ulong, long</tt></td>
768 <td><a name="t_integral">integral</a></td>
769 <td><tt>bool, ubyte, sbyte, ushort, short, uint, int, ulong, long</tt>
773 <td><a name="t_floating">floating point</a></td>
774 <td><tt>float, double</tt></td>
777 <td><a name="t_firstclass">first class</a></td>
778 <td><tt>bool, ubyte, sbyte, ushort, short, uint, int, ulong, long,<br>
779 float, double, <a href="#t_pointer">pointer</a>,
780 <a href="#t_packed">packed</a></tt></td>
785 <p>The <a href="#t_firstclass">first class</a> types are perhaps the
786 most important. Values of these types are the only ones which can be
787 produced by instructions, passed as arguments, or used as operands to
788 instructions. This means that all structures and arrays must be
789 manipulated either by pointer or by component.</p>
792 <!-- ======================================================================= -->
793 <div class="doc_subsection"> <a name="t_derived">Derived Types</a> </div>
795 <div class="doc_text">
797 <p>The real power in LLVM comes from the derived types in the system.
798 This is what allows a programmer to represent arrays, functions,
799 pointers, and other useful types. Note that these derived types may be
800 recursive: For example, it is possible to have a two dimensional array.</p>
804 <!-- _______________________________________________________________________ -->
805 <div class="doc_subsubsection"> <a name="t_array">Array Type</a> </div>
807 <div class="doc_text">
811 <p>The array type is a very simple derived type that arranges elements
812 sequentially in memory. The array type requires a size (number of
813 elements) and an underlying data type.</p>
818 [<# elements> x <elementtype>]
821 <p>The number of elements is a constant integer value; elementtype may
822 be any type with a size.</p>
825 <table class="layout">
828 <tt>[40 x int ]</tt><br/>
829 <tt>[41 x int ]</tt><br/>
830 <tt>[40 x uint]</tt><br/>
833 Array of 40 integer values.<br/>
834 Array of 41 integer values.<br/>
835 Array of 40 unsigned integer values.<br/>
839 <p>Here are some examples of multidimensional arrays:</p>
840 <table class="layout">
843 <tt>[3 x [4 x int]]</tt><br/>
844 <tt>[12 x [10 x float]]</tt><br/>
845 <tt>[2 x [3 x [4 x uint]]]</tt><br/>
848 3x4 array of integer values.<br/>
849 12x10 array of single precision floating point values.<br/>
850 2x3x4 array of unsigned integer values.<br/>
855 <p>Note that 'variable sized arrays' can be implemented in LLVM with a zero
856 length array. Normally, accesses past the end of an array are undefined in
857 LLVM (e.g. it is illegal to access the 5th element of a 3 element array).
858 As a special case, however, zero length arrays are recognized to be variable
859 length. This allows implementation of 'pascal style arrays' with the LLVM
860 type "{ int, [0 x float]}", for example.</p>
864 <!-- _______________________________________________________________________ -->
865 <div class="doc_subsubsection"> <a name="t_function">Function Type</a> </div>
866 <div class="doc_text">
868 <p>The function type can be thought of as a function signature. It
869 consists of a return type and a list of formal parameter types.
870 Function types are usually used to build virtual function tables
871 (which are structures of pointers to functions), for indirect function
872 calls, and when defining a function.</p>
874 The return type of a function type cannot be an aggregate type.
877 <pre> <returntype> (<parameter list>)<br></pre>
878 <p>...where '<tt><parameter list></tt>' is a comma-separated list of type
879 specifiers. Optionally, the parameter list may include a type <tt>...</tt>,
880 which indicates that the function takes a variable number of arguments.
881 Variable argument functions can access their arguments with the <a
882 href="#int_varargs">variable argument handling intrinsic</a> functions.</p>
884 <table class="layout">
887 <tt>int (int)</tt> <br/>
888 <tt>float (int, int *) *</tt><br/>
889 <tt>int (sbyte *, ...)</tt><br/>
892 function taking an <tt>int</tt>, returning an <tt>int</tt><br/>
893 <a href="#t_pointer">Pointer</a> to a function that takes an
894 <tt>int</tt> and a <a href="#t_pointer">pointer</a> to <tt>int</tt>,
895 returning <tt>float</tt>.<br/>
896 A vararg function that takes at least one <a href="#t_pointer">pointer</a>
897 to <tt>sbyte</tt> (signed char in C), which returns an integer. This is
898 the signature for <tt>printf</tt> in LLVM.<br/>
904 <!-- _______________________________________________________________________ -->
905 <div class="doc_subsubsection"> <a name="t_struct">Structure Type</a> </div>
906 <div class="doc_text">
908 <p>The structure type is used to represent a collection of data members
909 together in memory. The packing of the field types is defined to match
910 the ABI of the underlying processor. The elements of a structure may
911 be any type that has a size.</p>
912 <p>Structures are accessed using '<tt><a href="#i_load">load</a></tt>
913 and '<tt><a href="#i_store">store</a></tt>' by getting a pointer to a
914 field with the '<tt><a href="#i_getelementptr">getelementptr</a></tt>'
917 <pre> { <type list> }<br></pre>
919 <table class="layout">
922 <tt>{ int, int, int }</tt><br/>
923 <tt>{ float, int (int) * }</tt><br/>
926 a triple of three <tt>int</tt> values<br/>
927 A pair, where the first element is a <tt>float</tt> and the second element
928 is a <a href="#t_pointer">pointer</a> to a <a href="#t_function">function</a>
929 that takes an <tt>int</tt>, returning an <tt>int</tt>.<br/>
935 <!-- _______________________________________________________________________ -->
936 <div class="doc_subsubsection"> <a name="t_pointer">Pointer Type</a> </div>
937 <div class="doc_text">
939 <p>As in many languages, the pointer type represents a pointer or
940 reference to another object, which must live in memory.</p>
942 <pre> <type> *<br></pre>
944 <table class="layout">
947 <tt>[4x int]*</tt><br/>
948 <tt>int (int *) *</tt><br/>
951 A <a href="#t_pointer">pointer</a> to <a href="#t_array">array</a> of
952 four <tt>int</tt> values<br/>
953 A <a href="#t_pointer">pointer</a> to a <a
954 href="#t_function">function</a> that takes an <tt>int*</tt>, returning an
961 <!-- _______________________________________________________________________ -->
962 <div class="doc_subsubsection"> <a name="t_packed">Packed Type</a> </div>
963 <div class="doc_text">
967 <p>A packed type is a simple derived type that represents a vector
968 of elements. Packed types are used when multiple primitive data
969 are operated in parallel using a single instruction (SIMD).
970 A packed type requires a size (number of
971 elements) and an underlying primitive data type. Vectors must have a power
972 of two length (1, 2, 4, 8, 16 ...). Packed types are
973 considered <a href="#t_firstclass">first class</a>.</p>
978 < <# elements> x <elementtype> >
981 <p>The number of elements is a constant integer value; elementtype may
982 be any integral or floating point type.</p>
986 <table class="layout">
989 <tt><4 x int></tt><br/>
990 <tt><8 x float></tt><br/>
991 <tt><2 x uint></tt><br/>
994 Packed vector of 4 integer values.<br/>
995 Packed vector of 8 floating-point values.<br/>
996 Packed vector of 2 unsigned integer values.<br/>
1002 <!-- _______________________________________________________________________ -->
1003 <div class="doc_subsubsection"> <a name="t_opaque">Opaque Type</a> </div>
1004 <div class="doc_text">
1008 <p>Opaque types are used to represent unknown types in the system. This
1009 corresponds (for example) to the C notion of a foward declared structure type.
1010 In LLVM, opaque types can eventually be resolved to any type (not just a
1011 structure type).</p>
1021 <table class="layout">
1027 An opaque type.<br/>
1034 <!-- *********************************************************************** -->
1035 <div class="doc_section"> <a name="constants">Constants</a> </div>
1036 <!-- *********************************************************************** -->
1038 <div class="doc_text">
1040 <p>LLVM has several different basic types of constants. This section describes
1041 them all and their syntax.</p>
1045 <!-- ======================================================================= -->
1046 <div class="doc_subsection"><a name="simpleconstants">Simple Constants</a></div>
1048 <div class="doc_text">
1051 <dt><b>Boolean constants</b></dt>
1053 <dd>The two strings '<tt>true</tt>' and '<tt>false</tt>' are both valid
1054 constants of the <tt><a href="#t_primitive">bool</a></tt> type.
1057 <dt><b>Integer constants</b></dt>
1059 <dd>Standard integers (such as '4') are constants of the <a
1060 href="#t_integer">integer</a> type. Negative numbers may be used with signed
1064 <dt><b>Floating point constants</b></dt>
1066 <dd>Floating point constants use standard decimal notation (e.g. 123.421),
1067 exponential notation (e.g. 1.23421e+2), or a more precise hexadecimal
1068 notation (see below). Floating point constants must have a <a
1069 href="#t_floating">floating point</a> type. </dd>
1071 <dt><b>Null pointer constants</b></dt>
1073 <dd>The identifier '<tt>null</tt>' is recognized as a null pointer constant
1074 and must be of <a href="#t_pointer">pointer type</a>.</dd>
1078 <p>The one non-intuitive notation for constants is the optional hexadecimal form
1079 of floating point constants. For example, the form '<tt>double
1080 0x432ff973cafa8000</tt>' is equivalent to (but harder to read than) '<tt>double
1081 4.5e+15</tt>'. The only time hexadecimal floating point constants are required
1082 (and the only time that they are generated by the disassembler) is when a
1083 floating point constant must be emitted but it cannot be represented as a
1084 decimal floating point number. For example, NaN's, infinities, and other
1085 special values are represented in their IEEE hexadecimal format so that
1086 assembly and disassembly do not cause any bits to change in the constants.</p>
1090 <!-- ======================================================================= -->
1091 <div class="doc_subsection"><a name="aggregateconstants">Aggregate Constants</a>
1094 <div class="doc_text">
1095 <p>Aggregate constants arise from aggregation of simple constants
1096 and smaller aggregate constants.</p>
1099 <dt><b>Structure constants</b></dt>
1101 <dd>Structure constants are represented with notation similar to structure
1102 type definitions (a comma separated list of elements, surrounded by braces
1103 (<tt>{}</tt>)). For example: "<tt>{ int 4, float 17.0, int* %G }</tt>",
1104 where "<tt>%G</tt>" is declared as "<tt>%G = external global int</tt>". Structure constants
1105 must have <a href="#t_struct">structure type</a>, and the number and
1106 types of elements must match those specified by the type.
1109 <dt><b>Array constants</b></dt>
1111 <dd>Array constants are represented with notation similar to array type
1112 definitions (a comma separated list of elements, surrounded by square brackets
1113 (<tt>[]</tt>)). For example: "<tt>[ int 42, int 11, int 74 ]</tt>". Array
1114 constants must have <a href="#t_array">array type</a>, and the number and
1115 types of elements must match those specified by the type.
1118 <dt><b>Packed constants</b></dt>
1120 <dd>Packed constants are represented with notation similar to packed type
1121 definitions (a comma separated list of elements, surrounded by
1122 less-than/greater-than's (<tt><></tt>)). For example: "<tt>< int 42,
1123 int 11, int 74, int 100 ></tt>". Packed constants must have <a
1124 href="#t_packed">packed type</a>, and the number and types of elements must
1125 match those specified by the type.
1128 <dt><b>Zero initialization</b></dt>
1130 <dd>The string '<tt>zeroinitializer</tt>' can be used to zero initialize a
1131 value to zero of <em>any</em> type, including scalar and aggregate types.
1132 This is often used to avoid having to print large zero initializers (e.g. for
1133 large arrays) and is always exactly equivalent to using explicit zero
1140 <!-- ======================================================================= -->
1141 <div class="doc_subsection">
1142 <a name="globalconstants">Global Variable and Function Addresses</a>
1145 <div class="doc_text">
1147 <p>The addresses of <a href="#globalvars">global variables</a> and <a
1148 href="#functionstructure">functions</a> are always implicitly valid (link-time)
1149 constants. These constants are explicitly referenced when the <a
1150 href="#identifiers">identifier for the global</a> is used and always have <a
1151 href="#t_pointer">pointer</a> type. For example, the following is a legal LLVM
1157 %Z = global [2 x int*] [ int* %X, int* %Y ]
1162 <!-- ======================================================================= -->
1163 <div class="doc_subsection"><a name="undefvalues">Undefined Values</a></div>
1164 <div class="doc_text">
1165 <p>The string '<tt>undef</tt>' is recognized as a type-less constant that has
1166 no specific value. Undefined values may be of any type and be used anywhere
1167 a constant is permitted.</p>
1169 <p>Undefined values indicate to the compiler that the program is well defined
1170 no matter what value is used, giving the compiler more freedom to optimize.
1174 <!-- ======================================================================= -->
1175 <div class="doc_subsection"><a name="constantexprs">Constant Expressions</a>
1178 <div class="doc_text">
1180 <p>Constant expressions are used to allow expressions involving other constants
1181 to be used as constants. Constant expressions may be of any <a
1182 href="#t_firstclass">first class</a> type and may involve any LLVM operation
1183 that does not have side effects (e.g. load and call are not supported). The
1184 following is the syntax for constant expressions:</p>
1187 <dt><b><tt>trunc ( CST to TYPE )</tt></b></dt>
1188 <dd>Truncate a constant to another type. The bit size of CST must be larger
1189 than the bit size of TYPE. Both types must be integral.</dd>
1191 <dt><b><tt>zext ( CST to TYPE )</tt></b></dt>
1192 <dd>Zero extend a constant to another type. The bit size of CST must be
1193 smaller or equal to the bit size of TYPE. Both types must be integral.</dd>
1195 <dt><b><tt>sext ( CST to TYPE )</tt></b></dt>
1196 <dd>Sign extend a constant to another type. The bit size of CST must be
1197 smaller or equal to the bit size of TYPE. Both types must be integral.</dd>
1199 <dt><b><tt>fptrunc ( CST to TYPE )</tt></b></dt>
1200 <dd>Truncate a floating point constant to another floating point type. The
1201 size of CST must be larger than the size of TYPE. Both types must be
1202 floating point.</dd>
1204 <dt><b><tt>fpext ( CST to TYPE )</tt></b></dt>
1205 <dd>Floating point extend a constant to another type. The size of CST must be
1206 smaller or equal to the size of TYPE. Both types must be floating point.</dd>
1208 <dt><b><tt>fp2uint ( CST to TYPE )</tt></b></dt>
1209 <dd>Convert a floating point constant to the corresponding unsigned integer
1210 constant. TYPE must be an integer type. CST must be floating point. If the
1211 value won't fit in the integer type, the results are undefined.</dd>
1213 <dt><b><tt>fptosi ( CST to TYPE )</tt></b></dt>
1214 <dd>Convert a floating point constant to the corresponding signed integer
1215 constant. TYPE must be an integer type. CST must be floating point. If the
1216 value won't fit in the integer type, the results are undefined.</dd>
1218 <dt><b><tt>uitofp ( CST to TYPE )</tt></b></dt>
1219 <dd>Convert an unsigned integer constant to the corresponding floating point
1220 constant. TYPE must be floating point. CST must be of integer type. If the
1221 value won't fit in the floating point type, the results are undefined.</dd>
1223 <dt><b><tt>sitofp ( CST to TYPE )</tt></b></dt>
1224 <dd>Convert a signed integer constant to the corresponding floating point
1225 constant. TYPE must be floating point. CST must be of integer type. If the
1226 value won't fit in the floating point type, the results are undefined.</dd>
1228 <dt><b><tt>ptrtoint ( CST to TYPE )</tt></b></dt>
1229 <dd>Convert a pointer typed constant to the corresponding integer constant
1230 TYPE must be an integer type. CST must be of pointer type. The CST value is
1231 zero extended, truncated, or unchanged to make it fit in TYPE.</dd>
1233 <dt><b><tt>inttoptr ( CST to TYPE )</tt></b></dt>
1234 <dd>Convert a integer constant to a pointer constant. TYPE must be a
1235 pointer type. CST must be of integer type. The CST value is zero extended,
1236 truncated, or unchanged to make it fit in a pointer size. This one is
1237 <i>really</i> dangerous!</dd>
1239 <dt><b><tt>bitcast ( CST to TYPE )</tt></b></dt>
1240 <dd>Convert a constant, CST, to another TYPE. The size of CST and TYPE must be
1241 identical (same number of bits). The conversion is done as if the CST value
1242 was stored to memory and read back as TYPE. In other words, no bits change
1243 with this operator, just the type. This can be used for conversion of
1244 packed types to any other type, as long as they have the same bit width. For
1245 pointers it is only valid to cast to another pointer type.
1248 <dt><b><tt>getelementptr ( CSTPTR, IDX0, IDX1, ... )</tt></b></dt>
1250 <dd>Perform the <a href="#i_getelementptr">getelementptr operation</a> on
1251 constants. As with the <a href="#i_getelementptr">getelementptr</a>
1252 instruction, the index list may have zero or more indexes, which are required
1253 to make sense for the type of "CSTPTR".</dd>
1255 <dt><b><tt>select ( COND, VAL1, VAL2 )</tt></b></dt>
1257 <dd>Perform the <a href="#i_select">select operation</a> on
1260 <dt><b><tt>icmp COND ( VAL1, VAL2 )</tt></b></dt>
1261 <dd>Performs the <a href="#i_icmp">icmp operation</a> on constants.</dd>
1263 <dt><b><tt>fcmp COND ( VAL1, VAL2 )</tt></b></dt>
1264 <dd>Performs the <a href="#i_fcmp">fcmp operation</a> on constants.</dd>
1266 <dt><b><tt>extractelement ( VAL, IDX )</tt></b></dt>
1268 <dd>Perform the <a href="#i_extractelement">extractelement
1269 operation</a> on constants.
1271 <dt><b><tt>insertelement ( VAL, ELT, IDX )</tt></b></dt>
1273 <dd>Perform the <a href="#i_insertelement">insertelement
1274 operation</a> on constants.</dd>
1277 <dt><b><tt>shufflevector ( VEC1, VEC2, IDXMASK )</tt></b></dt>
1279 <dd>Perform the <a href="#i_shufflevector">shufflevector
1280 operation</a> on constants.</dd>
1282 <dt><b><tt>OPCODE ( LHS, RHS )</tt></b></dt>
1284 <dd>Perform the specified operation of the LHS and RHS constants. OPCODE may
1285 be any of the <a href="#binaryops">binary</a> or <a href="#bitwiseops">bitwise
1286 binary</a> operations. The constraints on operands are the same as those for
1287 the corresponding instruction (e.g. no bitwise operations on floating point
1288 values are allowed).</dd>
1292 <!-- *********************************************************************** -->
1293 <div class="doc_section"> <a name="othervalues">Other Values</a> </div>
1294 <!-- *********************************************************************** -->
1296 <!-- ======================================================================= -->
1297 <div class="doc_subsection">
1298 <a name="inlineasm">Inline Assembler Expressions</a>
1301 <div class="doc_text">
1304 LLVM supports inline assembler expressions (as opposed to <a href="#moduleasm">
1305 Module-Level Inline Assembly</a>) through the use of a special value. This
1306 value represents the inline assembler as a string (containing the instructions
1307 to emit), a list of operand constraints (stored as a string), and a flag that
1308 indicates whether or not the inline asm expression has side effects. An example
1309 inline assembler expression is:
1313 int(int) asm "bswap $0", "=r,r"
1317 Inline assembler expressions may <b>only</b> be used as the callee operand of
1318 a <a href="#i_call"><tt>call</tt> instruction</a>. Thus, typically we have:
1322 %X = call int asm "<a href="#i_bswap">bswap</a> $0", "=r,r"(int %Y)
1326 Inline asms with side effects not visible in the constraint list must be marked
1327 as having side effects. This is done through the use of the
1328 '<tt>sideeffect</tt>' keyword, like so:
1332 call void asm sideeffect "eieio", ""()
1335 <p>TODO: The format of the asm and constraints string still need to be
1336 documented here. Constraints on what can be done (e.g. duplication, moving, etc
1337 need to be documented).
1342 <!-- *********************************************************************** -->
1343 <div class="doc_section"> <a name="instref">Instruction Reference</a> </div>
1344 <!-- *********************************************************************** -->
1346 <div class="doc_text">
1348 <p>The LLVM instruction set consists of several different
1349 classifications of instructions: <a href="#terminators">terminator
1350 instructions</a>, <a href="#binaryops">binary instructions</a>,
1351 <a href="#bitwiseops">bitwise binary instructions</a>, <a
1352 href="#memoryops">memory instructions</a>, and <a href="#otherops">other
1353 instructions</a>.</p>
1357 <!-- ======================================================================= -->
1358 <div class="doc_subsection"> <a name="terminators">Terminator
1359 Instructions</a> </div>
1361 <div class="doc_text">
1363 <p>As mentioned <a href="#functionstructure">previously</a>, every
1364 basic block in a program ends with a "Terminator" instruction, which
1365 indicates which block should be executed after the current block is
1366 finished. These terminator instructions typically yield a '<tt>void</tt>'
1367 value: they produce control flow, not values (the one exception being
1368 the '<a href="#i_invoke"><tt>invoke</tt></a>' instruction).</p>
1369 <p>There are six different terminator instructions: the '<a
1370 href="#i_ret"><tt>ret</tt></a>' instruction, the '<a href="#i_br"><tt>br</tt></a>'
1371 instruction, the '<a href="#i_switch"><tt>switch</tt></a>' instruction,
1372 the '<a href="#i_invoke"><tt>invoke</tt></a>' instruction, the '<a
1373 href="#i_unwind"><tt>unwind</tt></a>' instruction, and the '<a
1374 href="#i_unreachable"><tt>unreachable</tt></a>' instruction.</p>
1378 <!-- _______________________________________________________________________ -->
1379 <div class="doc_subsubsection"> <a name="i_ret">'<tt>ret</tt>'
1380 Instruction</a> </div>
1381 <div class="doc_text">
1383 <pre> ret <type> <value> <i>; Return a value from a non-void function</i>
1384 ret void <i>; Return from void function</i>
1387 <p>The '<tt>ret</tt>' instruction is used to return control flow (and a
1388 value) from a function back to the caller.</p>
1389 <p>There are two forms of the '<tt>ret</tt>' instruction: one that
1390 returns a value and then causes control flow, and one that just causes
1391 control flow to occur.</p>
1393 <p>The '<tt>ret</tt>' instruction may return any '<a
1394 href="#t_firstclass">first class</a>' type. Notice that a function is
1395 not <a href="#wellformed">well formed</a> if there exists a '<tt>ret</tt>'
1396 instruction inside of the function that returns a value that does not
1397 match the return type of the function.</p>
1399 <p>When the '<tt>ret</tt>' instruction is executed, control flow
1400 returns back to the calling function's context. If the caller is a "<a
1401 href="#i_call"><tt>call</tt></a>" instruction, execution continues at
1402 the instruction after the call. If the caller was an "<a
1403 href="#i_invoke"><tt>invoke</tt></a>" instruction, execution continues
1404 at the beginning of the "normal" destination block. If the instruction
1405 returns a value, that value shall set the call or invoke instruction's
1408 <pre> ret int 5 <i>; Return an integer value of 5</i>
1409 ret void <i>; Return from a void function</i>
1412 <!-- _______________________________________________________________________ -->
1413 <div class="doc_subsubsection"> <a name="i_br">'<tt>br</tt>' Instruction</a> </div>
1414 <div class="doc_text">
1416 <pre> br bool <cond>, label <iftrue>, label <iffalse><br> br label <dest> <i>; Unconditional branch</i>
1419 <p>The '<tt>br</tt>' instruction is used to cause control flow to
1420 transfer to a different basic block in the current function. There are
1421 two forms of this instruction, corresponding to a conditional branch
1422 and an unconditional branch.</p>
1424 <p>The conditional branch form of the '<tt>br</tt>' instruction takes a
1425 single '<tt>bool</tt>' value and two '<tt>label</tt>' values. The
1426 unconditional form of the '<tt>br</tt>' instruction takes a single '<tt>label</tt>'
1427 value as a target.</p>
1429 <p>Upon execution of a conditional '<tt>br</tt>' instruction, the '<tt>bool</tt>'
1430 argument is evaluated. If the value is <tt>true</tt>, control flows
1431 to the '<tt>iftrue</tt>' <tt>label</tt> argument. If "cond" is <tt>false</tt>,
1432 control flows to the '<tt>iffalse</tt>' <tt>label</tt> argument.</p>
1434 <pre>Test:<br> %cond = <a href="#i_icmp">icmp</a> eq, int %a, %b<br> br bool %cond, label %IfEqual, label %IfUnequal<br>IfEqual:<br> <a
1435 href="#i_ret">ret</a> int 1<br>IfUnequal:<br> <a href="#i_ret">ret</a> int 0<br></pre>
1437 <!-- _______________________________________________________________________ -->
1438 <div class="doc_subsubsection">
1439 <a name="i_switch">'<tt>switch</tt>' Instruction</a>
1442 <div class="doc_text">
1446 switch <intty> <value>, label <defaultdest> [ <intty> <val>, label <dest> ... ]
1451 <p>The '<tt>switch</tt>' instruction is used to transfer control flow to one of
1452 several different places. It is a generalization of the '<tt>br</tt>'
1453 instruction, allowing a branch to occur to one of many possible
1459 <p>The '<tt>switch</tt>' instruction uses three parameters: an integer
1460 comparison value '<tt>value</tt>', a default '<tt>label</tt>' destination, and
1461 an array of pairs of comparison value constants and '<tt>label</tt>'s. The
1462 table is not allowed to contain duplicate constant entries.</p>
1466 <p>The <tt>switch</tt> instruction specifies a table of values and
1467 destinations. When the '<tt>switch</tt>' instruction is executed, this
1468 table is searched for the given value. If the value is found, control flow is
1469 transfered to the corresponding destination; otherwise, control flow is
1470 transfered to the default destination.</p>
1472 <h5>Implementation:</h5>
1474 <p>Depending on properties of the target machine and the particular
1475 <tt>switch</tt> instruction, this instruction may be code generated in different
1476 ways. For example, it could be generated as a series of chained conditional
1477 branches or with a lookup table.</p>
1482 <i>; Emulate a conditional br instruction</i>
1483 %Val = <a href="#i_zext">zext</a> bool %value to int
1484 switch int %Val, label %truedest [int 0, label %falsedest ]
1486 <i>; Emulate an unconditional br instruction</i>
1487 switch uint 0, label %dest [ ]
1489 <i>; Implement a jump table:</i>
1490 switch uint %val, label %otherwise [ uint 0, label %onzero
1491 uint 1, label %onone
1492 uint 2, label %ontwo ]
1496 <!-- _______________________________________________________________________ -->
1497 <div class="doc_subsubsection">
1498 <a name="i_invoke">'<tt>invoke</tt>' Instruction</a>
1501 <div class="doc_text">
1506 <result> = invoke [<a href="#callingconv">cconv</a>] <ptr to function ty> %<function ptr val>(<function args>)
1507 to label <normal label> unwind label <exception label>
1512 <p>The '<tt>invoke</tt>' instruction causes control to transfer to a specified
1513 function, with the possibility of control flow transfer to either the
1514 '<tt>normal</tt>' label or the
1515 '<tt>exception</tt>' label. If the callee function returns with the
1516 "<tt><a href="#i_ret">ret</a></tt>" instruction, control flow will return to the
1517 "normal" label. If the callee (or any indirect callees) returns with the "<a
1518 href="#i_unwind"><tt>unwind</tt></a>" instruction, control is interrupted and
1519 continued at the dynamically nearest "exception" label.</p>
1523 <p>This instruction requires several arguments:</p>
1527 The optional "cconv" marker indicates which <a href="callingconv">calling
1528 convention</a> the call should use. If none is specified, the call defaults
1529 to using C calling conventions.
1531 <li>'<tt>ptr to function ty</tt>': shall be the signature of the pointer to
1532 function value being invoked. In most cases, this is a direct function
1533 invocation, but indirect <tt>invoke</tt>s are just as possible, branching off
1534 an arbitrary pointer to function value.
1537 <li>'<tt>function ptr val</tt>': An LLVM value containing a pointer to a
1538 function to be invoked. </li>
1540 <li>'<tt>function args</tt>': argument list whose types match the function
1541 signature argument types. If the function signature indicates the function
1542 accepts a variable number of arguments, the extra arguments can be
1545 <li>'<tt>normal label</tt>': the label reached when the called function
1546 executes a '<tt><a href="#i_ret">ret</a></tt>' instruction. </li>
1548 <li>'<tt>exception label</tt>': the label reached when a callee returns with
1549 the <a href="#i_unwind"><tt>unwind</tt></a> instruction. </li>
1555 <p>This instruction is designed to operate as a standard '<tt><a
1556 href="#i_call">call</a></tt>' instruction in most regards. The primary
1557 difference is that it establishes an association with a label, which is used by
1558 the runtime library to unwind the stack.</p>
1560 <p>This instruction is used in languages with destructors to ensure that proper
1561 cleanup is performed in the case of either a <tt>longjmp</tt> or a thrown
1562 exception. Additionally, this is important for implementation of
1563 '<tt>catch</tt>' clauses in high-level languages that support them.</p>
1567 %retval = invoke int %Test(int 15) to label %Continue
1568 unwind label %TestCleanup <i>; {int}:retval set</i>
1569 %retval = invoke <a href="#callingconv">coldcc</a> int %Test(int 15) to label %Continue
1570 unwind label %TestCleanup <i>; {int}:retval set</i>
1575 <!-- _______________________________________________________________________ -->
1577 <div class="doc_subsubsection"> <a name="i_unwind">'<tt>unwind</tt>'
1578 Instruction</a> </div>
1580 <div class="doc_text">
1589 <p>The '<tt>unwind</tt>' instruction unwinds the stack, continuing control flow
1590 at the first callee in the dynamic call stack which used an <a
1591 href="#i_invoke"><tt>invoke</tt></a> instruction to perform the call. This is
1592 primarily used to implement exception handling.</p>
1596 <p>The '<tt>unwind</tt>' intrinsic causes execution of the current function to
1597 immediately halt. The dynamic call stack is then searched for the first <a
1598 href="#i_invoke"><tt>invoke</tt></a> instruction on the call stack. Once found,
1599 execution continues at the "exceptional" destination block specified by the
1600 <tt>invoke</tt> instruction. If there is no <tt>invoke</tt> instruction in the
1601 dynamic call chain, undefined behavior results.</p>
1604 <!-- _______________________________________________________________________ -->
1606 <div class="doc_subsubsection"> <a name="i_unreachable">'<tt>unreachable</tt>'
1607 Instruction</a> </div>
1609 <div class="doc_text">
1618 <p>The '<tt>unreachable</tt>' instruction has no defined semantics. This
1619 instruction is used to inform the optimizer that a particular portion of the
1620 code is not reachable. This can be used to indicate that the code after a
1621 no-return function cannot be reached, and other facts.</p>
1625 <p>The '<tt>unreachable</tt>' instruction has no defined semantics.</p>
1630 <!-- ======================================================================= -->
1631 <div class="doc_subsection"> <a name="binaryops">Binary Operations</a> </div>
1632 <div class="doc_text">
1633 <p>Binary operators are used to do most of the computation in a
1634 program. They require two operands, execute an operation on them, and
1635 produce a single value. The operands might represent
1636 multiple data, as is the case with the <a href="#t_packed">packed</a> data type.
1637 The result value of a binary operator is not
1638 necessarily the same type as its operands.</p>
1639 <p>There are several different binary operators:</p>
1641 <!-- _______________________________________________________________________ -->
1642 <div class="doc_subsubsection"> <a name="i_add">'<tt>add</tt>'
1643 Instruction</a> </div>
1644 <div class="doc_text">
1646 <pre> <result> = add <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1649 <p>The '<tt>add</tt>' instruction returns the sum of its two operands.</p>
1651 <p>The two arguments to the '<tt>add</tt>' instruction must be either <a
1652 href="#t_integer">integer</a> or <a href="#t_floating">floating point</a> values.
1653 This instruction can also take <a href="#t_packed">packed</a> versions of the values.
1654 Both arguments must have identical types.</p>
1656 <p>The value produced is the integer or floating point sum of the two
1659 <pre> <result> = add int 4, %var <i>; yields {int}:result = 4 + %var</i>
1662 <!-- _______________________________________________________________________ -->
1663 <div class="doc_subsubsection"> <a name="i_sub">'<tt>sub</tt>'
1664 Instruction</a> </div>
1665 <div class="doc_text">
1667 <pre> <result> = sub <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1670 <p>The '<tt>sub</tt>' instruction returns the difference of its two
1672 <p>Note that the '<tt>sub</tt>' instruction is used to represent the '<tt>neg</tt>'
1673 instruction present in most other intermediate representations.</p>
1675 <p>The two arguments to the '<tt>sub</tt>' instruction must be either <a
1676 href="#t_integer">integer</a> or <a href="#t_floating">floating point</a>
1678 This instruction can also take <a href="#t_packed">packed</a> versions of the values.
1679 Both arguments must have identical types.</p>
1681 <p>The value produced is the integer or floating point difference of
1682 the two operands.</p>
1684 <pre> <result> = sub int 4, %var <i>; yields {int}:result = 4 - %var</i>
1685 <result> = sub int 0, %val <i>; yields {int}:result = -%var</i>
1688 <!-- _______________________________________________________________________ -->
1689 <div class="doc_subsubsection"> <a name="i_mul">'<tt>mul</tt>'
1690 Instruction</a> </div>
1691 <div class="doc_text">
1693 <pre> <result> = mul <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1696 <p>The '<tt>mul</tt>' instruction returns the product of its two
1699 <p>The two arguments to the '<tt>mul</tt>' instruction must be either <a
1700 href="#t_integer">integer</a> or <a href="#t_floating">floating point</a>
1702 This instruction can also take <a href="#t_packed">packed</a> versions of the values.
1703 Both arguments must have identical types.</p>
1705 <p>The value produced is the integer or floating point product of the
1707 <p>There is no signed vs unsigned multiplication. The appropriate
1708 action is taken based on the type of the operand.</p>
1710 <pre> <result> = mul int 4, %var <i>; yields {int}:result = 4 * %var</i>
1713 <!-- _______________________________________________________________________ -->
1714 <div class="doc_subsubsection"> <a name="i_udiv">'<tt>udiv</tt>' Instruction
1716 <div class="doc_text">
1718 <pre> <result> = udiv <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1721 <p>The '<tt>udiv</tt>' instruction returns the quotient of its two
1724 <p>The two arguments to the '<tt>udiv</tt>' instruction must be
1725 <a href="#t_integer">integer</a> values. Both arguments must have identical
1726 types. This instruction can also take <a href="#t_packed">packed</a> versions
1727 of the values in which case the elements must be integers.</p>
1729 <p>The value produced is the unsigned integer quotient of the two operands. This
1730 instruction always performs an unsigned division operation, regardless of
1731 whether the arguments are unsigned or not.</p>
1733 <pre> <result> = udiv uint 4, %var <i>; yields {uint}:result = 4 / %var</i>
1736 <!-- _______________________________________________________________________ -->
1737 <div class="doc_subsubsection"> <a name="i_sdiv">'<tt>sdiv</tt>' Instruction
1739 <div class="doc_text">
1741 <pre> <result> = sdiv <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1744 <p>The '<tt>sdiv</tt>' instruction returns the quotient of its two
1747 <p>The two arguments to the '<tt>sdiv</tt>' instruction must be
1748 <a href="#t_integer">integer</a> values. Both arguments must have identical
1749 types. This instruction can also take <a href="#t_packed">packed</a> versions
1750 of the values in which case the elements must be integers.</p>
1752 <p>The value produced is the signed integer quotient of the two operands. This
1753 instruction always performs a signed division operation, regardless of whether
1754 the arguments are signed or not.</p>
1756 <pre> <result> = sdiv int 4, %var <i>; yields {int}:result = 4 / %var</i>
1759 <!-- _______________________________________________________________________ -->
1760 <div class="doc_subsubsection"> <a name="i_fdiv">'<tt>fdiv</tt>'
1761 Instruction</a> </div>
1762 <div class="doc_text">
1764 <pre> <result> = fdiv <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1767 <p>The '<tt>fdiv</tt>' instruction returns the quotient of its two
1770 <p>The two arguments to the '<tt>div</tt>' instruction must be
1771 <a href="#t_floating">floating point</a> values. Both arguments must have
1772 identical types. This instruction can also take <a href="#t_packed">packed</a>
1773 versions of the values in which case the elements must be floating point.</p>
1775 <p>The value produced is the floating point quotient of the two operands.</p>
1777 <pre> <result> = fdiv float 4.0, %var <i>; yields {float}:result = 4.0 / %var</i>
1780 <!-- _______________________________________________________________________ -->
1781 <div class="doc_subsubsection"> <a name="i_urem">'<tt>urem</tt>' Instruction</a>
1783 <div class="doc_text">
1785 <pre> <result> = urem <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1788 <p>The '<tt>urem</tt>' instruction returns the remainder from the
1789 unsigned division of its two arguments.</p>
1791 <p>The two arguments to the '<tt>urem</tt>' instruction must be
1792 <a href="#t_integer">integer</a> values. Both arguments must have identical
1795 <p>This instruction returns the unsigned integer <i>remainder</i> of a division.
1796 This instruction always performs an unsigned division to get the remainder,
1797 regardless of whether the arguments are unsigned or not.</p>
1799 <pre> <result> = urem uint 4, %var <i>; yields {uint}:result = 4 % %var</i>
1803 <!-- _______________________________________________________________________ -->
1804 <div class="doc_subsubsection"> <a name="i_srem">'<tt>srem</tt>'
1805 Instruction</a> </div>
1806 <div class="doc_text">
1808 <pre> <result> = srem <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1811 <p>The '<tt>srem</tt>' instruction returns the remainder from the
1812 signed division of its two operands.</p>
1814 <p>The two arguments to the '<tt>srem</tt>' instruction must be
1815 <a href="#t_integer">integer</a> values. Both arguments must have identical
1818 <p>This instruction returns the <i>remainder</i> of a division (where the result
1819 has the same sign as the divisor), not the <i>modulus</i> (where the
1820 result has the same sign as the dividend) of a value. For more
1821 information about the difference, see <a
1822 href="http://mathforum.org/dr.math/problems/anne.4.28.99.html">The
1825 <pre> <result> = srem int 4, %var <i>; yields {int}:result = 4 % %var</i>
1829 <!-- _______________________________________________________________________ -->
1830 <div class="doc_subsubsection"> <a name="i_frem">'<tt>frem</tt>'
1831 Instruction</a> </div>
1832 <div class="doc_text">
1834 <pre> <result> = frem <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1837 <p>The '<tt>frem</tt>' instruction returns the remainder from the
1838 division of its two operands.</p>
1840 <p>The two arguments to the '<tt>frem</tt>' instruction must be
1841 <a href="#t_floating">floating point</a> values. Both arguments must have
1842 identical types.</p>
1844 <p>This instruction returns the <i>remainder</i> of a division.</p>
1846 <pre> <result> = frem float 4.0, %var <i>; yields {float}:result = 4.0 % %var</i>
1850 <!-- ======================================================================= -->
1851 <div class="doc_subsection"> <a name="bitwiseops">Bitwise Binary
1852 Operations</a> </div>
1853 <div class="doc_text">
1854 <p>Bitwise binary operators are used to do various forms of
1855 bit-twiddling in a program. They are generally very efficient
1856 instructions and can commonly be strength reduced from other
1857 instructions. They require two operands, execute an operation on them,
1858 and produce a single value. The resulting value of the bitwise binary
1859 operators is always the same type as its first operand.</p>
1861 <!-- _______________________________________________________________________ -->
1862 <div class="doc_subsubsection"> <a name="i_and">'<tt>and</tt>'
1863 Instruction</a> </div>
1864 <div class="doc_text">
1866 <pre> <result> = and <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1869 <p>The '<tt>and</tt>' instruction returns the bitwise logical and of
1870 its two operands.</p>
1872 <p>The two arguments to the '<tt>and</tt>' instruction must be <a
1873 href="#t_integral">integral</a> values. Both arguments must have
1874 identical types.</p>
1876 <p>The truth table used for the '<tt>and</tt>' instruction is:</p>
1878 <div style="align: center">
1879 <table border="1" cellspacing="0" cellpadding="4">
1910 <pre> <result> = and int 4, %var <i>; yields {int}:result = 4 & %var</i>
1911 <result> = and int 15, 40 <i>; yields {int}:result = 8</i>
1912 <result> = and int 4, 8 <i>; yields {int}:result = 0</i>
1915 <!-- _______________________________________________________________________ -->
1916 <div class="doc_subsubsection"> <a name="i_or">'<tt>or</tt>' Instruction</a> </div>
1917 <div class="doc_text">
1919 <pre> <result> = or <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1922 <p>The '<tt>or</tt>' instruction returns the bitwise logical inclusive
1923 or of its two operands.</p>
1925 <p>The two arguments to the '<tt>or</tt>' instruction must be <a
1926 href="#t_integral">integral</a> values. Both arguments must have
1927 identical types.</p>
1929 <p>The truth table used for the '<tt>or</tt>' instruction is:</p>
1931 <div style="align: center">
1932 <table border="1" cellspacing="0" cellpadding="4">
1963 <pre> <result> = or int 4, %var <i>; yields {int}:result = 4 | %var</i>
1964 <result> = or int 15, 40 <i>; yields {int}:result = 47</i>
1965 <result> = or int 4, 8 <i>; yields {int}:result = 12</i>
1968 <!-- _______________________________________________________________________ -->
1969 <div class="doc_subsubsection"> <a name="i_xor">'<tt>xor</tt>'
1970 Instruction</a> </div>
1971 <div class="doc_text">
1973 <pre> <result> = xor <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1976 <p>The '<tt>xor</tt>' instruction returns the bitwise logical exclusive
1977 or of its two operands. The <tt>xor</tt> is used to implement the
1978 "one's complement" operation, which is the "~" operator in C.</p>
1980 <p>The two arguments to the '<tt>xor</tt>' instruction must be <a
1981 href="#t_integral">integral</a> values. Both arguments must have
1982 identical types.</p>
1984 <p>The truth table used for the '<tt>xor</tt>' instruction is:</p>
1986 <div style="align: center">
1987 <table border="1" cellspacing="0" cellpadding="4">
2019 <pre> <result> = xor int 4, %var <i>; yields {int}:result = 4 ^ %var</i>
2020 <result> = xor int 15, 40 <i>; yields {int}:result = 39</i>
2021 <result> = xor int 4, 8 <i>; yields {int}:result = 12</i>
2022 <result> = xor int %V, -1 <i>; yields {int}:result = ~%V</i>
2025 <!-- _______________________________________________________________________ -->
2026 <div class="doc_subsubsection"> <a name="i_shl">'<tt>shl</tt>'
2027 Instruction</a> </div>
2028 <div class="doc_text">
2030 <pre> <result> = shl <ty> <var1>, ubyte <var2> <i>; yields {ty}:result</i>
2033 <p>The '<tt>shl</tt>' instruction returns the first operand shifted to
2034 the left a specified number of bits.</p>
2036 <p>The first argument to the '<tt>shl</tt>' instruction must be an <a
2037 href="#t_integer">integer</a> type. The second argument must be an '<tt>ubyte</tt>'
2040 <p>The value produced is <tt>var1</tt> * 2<sup><tt>var2</tt></sup>.</p>
2042 <pre> <result> = shl int 4, ubyte %var <i>; yields {int}:result = 4 << %var</i>
2043 <result> = shl int 4, ubyte 2 <i>; yields {int}:result = 16</i>
2044 <result> = shl int 1, ubyte 10 <i>; yields {int}:result = 1024</i>
2047 <!-- _______________________________________________________________________ -->
2048 <div class="doc_subsubsection"> <a name="i_lshr">'<tt>lshr</tt>'
2049 Instruction</a> </div>
2050 <div class="doc_text">
2052 <pre> <result> = lshr <ty> <var1>, ubyte <var2> <i>; yields {ty}:result</i>
2056 <p>The '<tt>lshr</tt>' instruction (logical shift right) returns the first
2057 operand shifted to the right a specified number of bits.</p>
2060 <p>The first argument to the '<tt>lshr</tt>' instruction must be an <a
2061 href="#t_integer">integer</a> type. The second argument must be an '<tt>ubyte</tt>' type.</p>
2064 <p>This instruction always performs a logical shift right operation, regardless
2065 of whether the arguments are unsigned or not. The <tt>var2</tt> most significant
2066 bits will be filled with zero bits after the shift.</p>
2070 <result> = lshr uint 4, ubyte 1 <i>; yields {uint}:result = 2</i>
2071 <result> = lshr int 4, ubyte 2 <i>; yields {uint}:result = 1</i>
2072 <result> = lshr sbyte 4, ubyte 3 <i>; yields {sbyte}:result = 0</i>
2073 <result> = lshr sbyte -2, ubyte 1 <i>; yields {sbyte}:result = 0x7FFFFFFF </i>
2077 <!-- ======================================================================= -->
2078 <div class="doc_subsubsection"> <a name="i_ashr">'<tt>ashr</tt>'
2079 Instruction</a> </div>
2080 <div class="doc_text">
2083 <pre> <result> = ashr <ty> <var1>, ubyte <var2> <i>; yields {ty}:result</i>
2087 <p>The '<tt>ashr</tt>' instruction (arithmetic shift right) returns the first
2088 operand shifted to the right a specified number of bits.</p>
2091 <p>The first argument to the '<tt>ashr</tt>' instruction must be an
2092 <a href="#t_integer">integer</a> type. The second argument must be an
2093 '<tt>ubyte</tt>' type.</p>
2096 <p>This instruction always performs an arithmetic shift right operation,
2097 regardless of whether the arguments are signed or not. The <tt>var2</tt> most
2098 significant bits will be filled with the sign bit of <tt>var1</tt>.</p>
2102 <result> = ashr uint 4, ubyte 1 <i>; yields {uint}:result = 2</i>
2103 <result> = ashr int 4, ubyte 2 <i>; yields {int}:result = 1</i>
2104 <result> = ashr ubyte 4, ubyte 3 <i>; yields {ubyte}:result = 0</i>
2105 <result> = ashr sbyte -2, ubyte 1 <i>; yields {sbyte}:result = -1</i>
2109 <!-- ======================================================================= -->
2110 <div class="doc_subsection">
2111 <a name="vectorops">Vector Operations</a>
2114 <div class="doc_text">
2116 <p>LLVM supports several instructions to represent vector operations in a
2117 target-independent manner. This instructions cover the element-access and
2118 vector-specific operations needed to process vectors effectively. While LLVM
2119 does directly support these vector operations, many sophisticated algorithms
2120 will want to use target-specific intrinsics to take full advantage of a specific
2125 <!-- _______________________________________________________________________ -->
2126 <div class="doc_subsubsection">
2127 <a name="i_extractelement">'<tt>extractelement</tt>' Instruction</a>
2130 <div class="doc_text">
2135 <result> = extractelement <n x <ty>> <val>, uint <idx> <i>; yields <ty></i>
2141 The '<tt>extractelement</tt>' instruction extracts a single scalar
2142 element from a packed vector at a specified index.
2149 The first operand of an '<tt>extractelement</tt>' instruction is a
2150 value of <a href="#t_packed">packed</a> type. The second operand is
2151 an index indicating the position from which to extract the element.
2152 The index may be a variable.</p>
2157 The result is a scalar of the same type as the element type of
2158 <tt>val</tt>. Its value is the value at position <tt>idx</tt> of
2159 <tt>val</tt>. If <tt>idx</tt> exceeds the length of <tt>val</tt>, the
2160 results are undefined.
2166 %result = extractelement <4 x int> %vec, uint 0 <i>; yields int</i>
2171 <!-- _______________________________________________________________________ -->
2172 <div class="doc_subsubsection">
2173 <a name="i_insertelement">'<tt>insertelement</tt>' Instruction</a>
2176 <div class="doc_text">
2181 <result> = insertelement <n x <ty>> <val>, <ty> <elt>, uint <idx> <i>; yields <n x <ty>></i>
2187 The '<tt>insertelement</tt>' instruction inserts a scalar
2188 element into a packed vector at a specified index.
2195 The first operand of an '<tt>insertelement</tt>' instruction is a
2196 value of <a href="#t_packed">packed</a> type. The second operand is a
2197 scalar value whose type must equal the element type of the first
2198 operand. The third operand is an index indicating the position at
2199 which to insert the value. The index may be a variable.</p>
2204 The result is a packed vector of the same type as <tt>val</tt>. Its
2205 element values are those of <tt>val</tt> except at position
2206 <tt>idx</tt>, where it gets the value <tt>elt</tt>. If <tt>idx</tt>
2207 exceeds the length of <tt>val</tt>, the results are undefined.
2213 %result = insertelement <4 x int> %vec, int 1, uint 0 <i>; yields <4 x int></i>
2217 <!-- _______________________________________________________________________ -->
2218 <div class="doc_subsubsection">
2219 <a name="i_shufflevector">'<tt>shufflevector</tt>' Instruction</a>
2222 <div class="doc_text">
2227 <result> = shufflevector <n x <ty>> <v1>, <n x <ty>> <v2>, <n x uint> <mask> <i>; yields <n x <ty>></i>
2233 The '<tt>shufflevector</tt>' instruction constructs a permutation of elements
2234 from two input vectors, returning a vector of the same type.
2240 The first two operands of a '<tt>shufflevector</tt>' instruction are vectors
2241 with types that match each other and types that match the result of the
2242 instruction. The third argument is a shuffle mask, which has the same number
2243 of elements as the other vector type, but whose element type is always 'uint'.
2247 The shuffle mask operand is required to be a constant vector with either
2248 constant integer or undef values.
2254 The elements of the two input vectors are numbered from left to right across
2255 both of the vectors. The shuffle mask operand specifies, for each element of
2256 the result vector, which element of the two input registers the result element
2257 gets. The element selector may be undef (meaning "don't care") and the second
2258 operand may be undef if performing a shuffle from only one vector.
2264 %result = shufflevector <4 x int> %v1, <4 x int> %v2,
2265 <4 x uint> <uint 0, uint 4, uint 1, uint 5> <i>; yields <4 x int></i>
2266 %result = shufflevector <4 x int> %v1, <4 x int> undef,
2267 <4 x uint> <uint 0, uint 1, uint 2, uint 3> <i>; yields <4 x int></i> - Identity shuffle.
2272 <!-- ======================================================================= -->
2273 <div class="doc_subsection">
2274 <a name="memoryops">Memory Access and Addressing Operations</a>
2277 <div class="doc_text">
2279 <p>A key design point of an SSA-based representation is how it
2280 represents memory. In LLVM, no memory locations are in SSA form, which
2281 makes things very simple. This section describes how to read, write,
2282 allocate, and free memory in LLVM.</p>
2286 <!-- _______________________________________________________________________ -->
2287 <div class="doc_subsubsection">
2288 <a name="i_malloc">'<tt>malloc</tt>' Instruction</a>
2291 <div class="doc_text">
2296 <result> = malloc <type>[, uint <NumElements>][, align <alignment>] <i>; yields {type*}:result</i>
2301 <p>The '<tt>malloc</tt>' instruction allocates memory from the system
2302 heap and returns a pointer to it.</p>
2306 <p>The '<tt>malloc</tt>' instruction allocates
2307 <tt>sizeof(<type>)*NumElements</tt>
2308 bytes of memory from the operating system and returns a pointer of the
2309 appropriate type to the program. If "NumElements" is specified, it is the
2310 number of elements allocated. If an alignment is specified, the value result
2311 of the allocation is guaranteed to be aligned to at least that boundary. If
2312 not specified, or if zero, the target can choose to align the allocation on any
2313 convenient boundary.</p>
2315 <p>'<tt>type</tt>' must be a sized type.</p>
2319 <p>Memory is allocated using the system "<tt>malloc</tt>" function, and
2320 a pointer is returned.</p>
2325 %array = malloc [4 x ubyte ] <i>; yields {[%4 x ubyte]*}:array</i>
2327 %size = <a href="#i_add">add</a> uint 2, 2 <i>; yields {uint}:size = uint 4</i>
2328 %array1 = malloc ubyte, uint 4 <i>; yields {ubyte*}:array1</i>
2329 %array2 = malloc [12 x ubyte], uint %size <i>; yields {[12 x ubyte]*}:array2</i>
2330 %array3 = malloc int, uint 4, align 1024 <i>; yields {int*}:array3</i>
2331 %array4 = malloc int, align 1024 <i>; yields {int*}:array4</i>
2335 <!-- _______________________________________________________________________ -->
2336 <div class="doc_subsubsection">
2337 <a name="i_free">'<tt>free</tt>' Instruction</a>
2340 <div class="doc_text">
2345 free <type> <value> <i>; yields {void}</i>
2350 <p>The '<tt>free</tt>' instruction returns memory back to the unused
2351 memory heap to be reallocated in the future.</p>
2355 <p>'<tt>value</tt>' shall be a pointer value that points to a value
2356 that was allocated with the '<tt><a href="#i_malloc">malloc</a></tt>'
2361 <p>Access to the memory pointed to by the pointer is no longer defined
2362 after this instruction executes.</p>
2367 %array = <a href="#i_malloc">malloc</a> [4 x ubyte] <i>; yields {[4 x ubyte]*}:array</i>
2368 free [4 x ubyte]* %array
2372 <!-- _______________________________________________________________________ -->
2373 <div class="doc_subsubsection">
2374 <a name="i_alloca">'<tt>alloca</tt>' Instruction</a>
2377 <div class="doc_text">
2382 <result> = alloca <type>[, uint <NumElements>][, align <alignment>] <i>; yields {type*}:result</i>
2387 <p>The '<tt>alloca</tt>' instruction allocates memory on the current
2388 stack frame of the procedure that is live until the current function
2389 returns to its caller.</p>
2393 <p>The '<tt>alloca</tt>' instruction allocates <tt>sizeof(<type>)*NumElements</tt>
2394 bytes of memory on the runtime stack, returning a pointer of the
2395 appropriate type to the program. If "NumElements" is specified, it is the
2396 number of elements allocated. If an alignment is specified, the value result
2397 of the allocation is guaranteed to be aligned to at least that boundary. If
2398 not specified, or if zero, the target can choose to align the allocation on any
2399 convenient boundary.</p>
2401 <p>'<tt>type</tt>' may be any sized type.</p>
2405 <p>Memory is allocated; a pointer is returned. '<tt>alloca</tt>'d
2406 memory is automatically released when the function returns. The '<tt>alloca</tt>'
2407 instruction is commonly used to represent automatic variables that must
2408 have an address available. When the function returns (either with the <tt><a
2409 href="#i_ret">ret</a></tt> or <tt><a href="#i_unwind">unwind</a></tt>
2410 instructions), the memory is reclaimed.</p>
2415 %ptr = alloca int <i>; yields {int*}:ptr</i>
2416 %ptr = alloca int, uint 4 <i>; yields {int*}:ptr</i>
2417 %ptr = alloca int, uint 4, align 1024 <i>; yields {int*}:ptr</i>
2418 %ptr = alloca int, align 1024 <i>; yields {int*}:ptr</i>
2422 <!-- _______________________________________________________________________ -->
2423 <div class="doc_subsubsection"> <a name="i_load">'<tt>load</tt>'
2424 Instruction</a> </div>
2425 <div class="doc_text">
2427 <pre> <result> = load <ty>* <pointer><br> <result> = volatile load <ty>* <pointer><br></pre>
2429 <p>The '<tt>load</tt>' instruction is used to read from memory.</p>
2431 <p>The argument to the '<tt>load</tt>' instruction specifies the memory
2432 address from which to load. The pointer must point to a <a
2433 href="#t_firstclass">first class</a> type. If the <tt>load</tt> is
2434 marked as <tt>volatile</tt>, then the optimizer is not allowed to modify
2435 the number or order of execution of this <tt>load</tt> with other
2436 volatile <tt>load</tt> and <tt><a href="#i_store">store</a></tt>
2439 <p>The location of memory pointed to is loaded.</p>
2441 <pre> %ptr = <a href="#i_alloca">alloca</a> int <i>; yields {int*}:ptr</i>
2443 href="#i_store">store</a> int 3, int* %ptr <i>; yields {void}</i>
2444 %val = load int* %ptr <i>; yields {int}:val = int 3</i>
2447 <!-- _______________________________________________________________________ -->
2448 <div class="doc_subsubsection"> <a name="i_store">'<tt>store</tt>'
2449 Instruction</a> </div>
2450 <div class="doc_text">
2452 <pre> store <ty> <value>, <ty>* <pointer> <i>; yields {void}</i>
2453 volatile store <ty> <value>, <ty>* <pointer> <i>; yields {void}</i>
2456 <p>The '<tt>store</tt>' instruction is used to write to memory.</p>
2458 <p>There are two arguments to the '<tt>store</tt>' instruction: a value
2459 to store and an address in which to store it. The type of the '<tt><pointer></tt>'
2460 operand must be a pointer to the type of the '<tt><value></tt>'
2461 operand. If the <tt>store</tt> is marked as <tt>volatile</tt>, then the
2462 optimizer is not allowed to modify the number or order of execution of
2463 this <tt>store</tt> with other volatile <tt>load</tt> and <tt><a
2464 href="#i_store">store</a></tt> instructions.</p>
2466 <p>The contents of memory are updated to contain '<tt><value></tt>'
2467 at the location specified by the '<tt><pointer></tt>' operand.</p>
2469 <pre> %ptr = <a href="#i_alloca">alloca</a> int <i>; yields {int*}:ptr</i>
2471 href="#i_store">store</a> int 3, int* %ptr <i>; yields {void}</i>
2472 %val = load int* %ptr <i>; yields {int}:val = int 3</i>
2476 <!-- _______________________________________________________________________ -->
2477 <div class="doc_subsubsection">
2478 <a name="i_getelementptr">'<tt>getelementptr</tt>' Instruction</a>
2481 <div class="doc_text">
2484 <result> = getelementptr <ty>* <ptrval>{, <ty> <idx>}*
2490 The '<tt>getelementptr</tt>' instruction is used to get the address of a
2491 subelement of an aggregate data structure.</p>
2495 <p>This instruction takes a list of integer operands that indicate what
2496 elements of the aggregate object to index to. The actual types of the arguments
2497 provided depend on the type of the first pointer argument. The
2498 '<tt>getelementptr</tt>' instruction is used to index down through the type
2499 levels of a structure or to a specific index in an array. When indexing into a
2500 structure, only <tt>uint</tt> integer constants are allowed. When indexing
2501 into an array or pointer, only integers of 32 or 64 bits are allowed, and will
2502 be sign extended to 64-bit values.</p>
2504 <p>For example, let's consider a C code fragment and how it gets
2505 compiled to LLVM:</p>
2519 int *foo(struct ST *s) {
2520 return &s[1].Z.B[5][13];
2524 <p>The LLVM code generated by the GCC frontend is:</p>
2527 %RT = type { sbyte, [10 x [20 x int]], sbyte }
2528 %ST = type { int, double, %RT }
2532 int* %foo(%ST* %s) {
2534 %reg = getelementptr %ST* %s, int 1, uint 2, uint 1, int 5, int 13
2541 <p>The index types specified for the '<tt>getelementptr</tt>' instruction depend
2542 on the pointer type that is being indexed into. <a href="#t_pointer">Pointer</a>
2543 and <a href="#t_array">array</a> types can use a 32-bit or 64-bit
2544 <a href="#t_integer">integer</a> type but the value will always be sign extended
2545 to 64-bits. <a href="#t_struct">Structure</a> types, require <tt>uint</tt>
2546 <b>constants</b>.</p>
2548 <p>In the example above, the first index is indexing into the '<tt>%ST*</tt>'
2549 type, which is a pointer, yielding a '<tt>%ST</tt>' = '<tt>{ int, double, %RT
2550 }</tt>' type, a structure. The second index indexes into the third element of
2551 the structure, yielding a '<tt>%RT</tt>' = '<tt>{ sbyte, [10 x [20 x int]],
2552 sbyte }</tt>' type, another structure. The third index indexes into the second
2553 element of the structure, yielding a '<tt>[10 x [20 x int]]</tt>' type, an
2554 array. The two dimensions of the array are subscripted into, yielding an
2555 '<tt>int</tt>' type. The '<tt>getelementptr</tt>' instruction returns a pointer
2556 to this element, thus computing a value of '<tt>int*</tt>' type.</p>
2558 <p>Note that it is perfectly legal to index partially through a
2559 structure, returning a pointer to an inner element. Because of this,
2560 the LLVM code for the given testcase is equivalent to:</p>
2563 int* %foo(%ST* %s) {
2564 %t1 = getelementptr %ST* %s, int 1 <i>; yields %ST*:%t1</i>
2565 %t2 = getelementptr %ST* %t1, int 0, uint 2 <i>; yields %RT*:%t2</i>
2566 %t3 = getelementptr %RT* %t2, int 0, uint 1 <i>; yields [10 x [20 x int]]*:%t3</i>
2567 %t4 = getelementptr [10 x [20 x int]]* %t3, int 0, int 5 <i>; yields [20 x int]*:%t4</i>
2568 %t5 = getelementptr [20 x int]* %t4, int 0, int 13 <i>; yields int*:%t5</i>
2573 <p>Note that it is undefined to access an array out of bounds: array and
2574 pointer indexes must always be within the defined bounds of the array type.
2575 The one exception for this rules is zero length arrays. These arrays are
2576 defined to be accessible as variable length arrays, which requires access
2577 beyond the zero'th element.</p>
2579 <p>The getelementptr instruction is often confusing. For some more insight
2580 into how it works, see <a href="GetElementPtr.html">the getelementptr
2586 <i>; yields [12 x ubyte]*:aptr</i>
2587 %aptr = getelementptr {int, [12 x ubyte]}* %sptr, long 0, uint 1
2591 <!-- ======================================================================= -->
2592 <div class="doc_subsection"> <a name="convertops">Conversion Operations</a>
2594 <div class="doc_text">
2595 <p>The instructions in this category are the conversion instructions (casting)
2596 which all take a single operand and a type. They perform various bit conversions
2600 <!-- _______________________________________________________________________ -->
2601 <div class="doc_subsubsection">
2602 <a name="i_trunc">'<tt>trunc .. to</tt>' Instruction</a>
2604 <div class="doc_text">
2608 <result> = trunc <ty> <value> to <ty2> <i>; yields ty2</i>
2613 The '<tt>trunc</tt>' instruction truncates its operand to the type <tt>ty2</tt>.
2618 The '<tt>trunc</tt>' instruction takes a <tt>value</tt> to trunc, which must
2619 be an <a href="#t_integer">integer</a> type, and a type that specifies the size
2620 and type of the result, which must be an <a href="#t_integral">integral</a>
2621 type. The bit size of <tt>value</tt> must be larger than the bit size of
2622 <tt>ty2</tt>. Equal sized types are not allowed.</p>
2626 The '<tt>trunc</tt>' instruction truncates the high order bits in <tt>value</tt>
2627 and converts the remaining bits to <tt>ty2</tt>. Since the source size must be
2628 larger than the destination size, <tt>trunc</tt> cannot be a <i>no-op cast</i>.
2629 It will always truncate bits.</p>
2633 %X = trunc int 257 to ubyte <i>; yields ubyte:1</i>
2634 %Y = trunc int 123 to bool <i>; yields bool:true</i>
2638 <!-- _______________________________________________________________________ -->
2639 <div class="doc_subsubsection">
2640 <a name="i_zext">'<tt>zext .. to</tt>' Instruction</a>
2642 <div class="doc_text">
2646 <result> = zext <ty> <value> to <ty2> <i>; yields ty2</i>
2650 <p>The '<tt>zext</tt>' instruction zero extends its operand to type
2655 <p>The '<tt>zext</tt>' instruction takes a value to cast, which must be of
2656 <a href="#t_integral">integral</a> type, and a type to cast it to, which must
2657 also be of <a href="#t_integral">integral</a> type. The bit size of the
2658 <tt>value</tt> must be smaller than the bit size of the destination type,
2662 <p>The <tt>zext</tt> fills the high order bits of the <tt>value</tt> with zero
2663 bits until it reaches the size of the destination type, <tt>ty2</tt>. When the
2664 the operand and the type are the same size, no bit filling is done and the
2665 cast is considered a <i>no-op cast</i> because no bits change (only the type
2668 <p>When zero extending from bool, the result will alwasy be either 0 or 1.</p>
2672 %X = zext int 257 to ulong <i>; yields ulong:257</i>
2673 %Y = zext bool true to int <i>; yields int:1</i>
2677 <!-- _______________________________________________________________________ -->
2678 <div class="doc_subsubsection">
2679 <a name="i_sext">'<tt>sext .. to</tt>' Instruction</a>
2681 <div class="doc_text">
2685 <result> = sext <ty> <value> to <ty2> <i>; yields ty2</i>
2689 <p>The '<tt>sext</tt>' sign extends <tt>value</tt> to the type <tt>ty2</tt>.</p>
2693 The '<tt>sext</tt>' instruction takes a value to cast, which must be of
2694 <a href="#t_integral">integral</a> type, and a type to cast it to, which must
2695 also be of <a href="#t_integral">integral</a> type. The bit size of the
2696 <tt>value</tt> must be smaller than the bit size of the destination type,
2701 The '<tt>sext</tt>' instruction performs a sign extension by copying the sign
2702 bit (highest order bit) of the <tt>value</tt> until it reaches the bit size of
2703 the type <tt>ty2</tt>. When the the operand and the type are the same size,
2704 no bit filling is done and the cast is considered a <i>no-op cast</i> because
2705 no bits change (only the type changes).</p>
2707 <p>When sign extending from bool, the extension always results in -1 or 0.</p>
2711 %X = sext sbyte -1 to ushort <i>; yields ushort:65535</i>
2712 %Y = sext bool true to int <i>; yields int:-1</i>
2716 <!-- _______________________________________________________________________ -->
2717 <div class="doc_subsubsection">
2718 <a name="i_fptrunc">'<tt>fptrunc .. to</tt>' Instruction</a>
2721 <div class="doc_text">
2726 <result> = fptrunc <ty> <value> to <ty2> <i>; yields ty2</i>
2730 <p>The '<tt>fptrunc</tt>' instruction truncates <tt>value</tt> to type
2735 <p>The '<tt>fptrunc</tt>' instruction takes a <a href="#t_floating">floating
2736 point</a> value to cast and a <a href="#t_floating">floating point</a> type to
2737 cast it to. The size of <tt>value</tt> must be larger than the size of
2738 <tt>ty2</tt>. This implies that <tt>fptrunc</tt> cannot be used to make a
2739 <i>no-op cast</i>.</p>
2742 <p> The '<tt>fptrunc</tt>' instruction truncates a <tt>value</tt> from a larger
2743 <a href="#t_floating">floating point</a> type to a smaller
2744 <a href="#t_floating">floating point</a> type. If the value cannot fit within
2745 the destination type, <tt>ty2</tt>, then the results are undefined.</p>
2749 %X = fptrunc double 123.0 to float <i>; yields float:123.0</i>
2750 %Y = fptrunc double 1.0E+300 to float <i>; yields undefined</i>
2754 <!-- _______________________________________________________________________ -->
2755 <div class="doc_subsubsection">
2756 <a name="i_fpext">'<tt>fpext .. to</tt>' Instruction</a>
2758 <div class="doc_text">
2762 <result> = fpext <ty> <value> to <ty2> <i>; yields ty2</i>
2766 <p>The '<tt>fpext</tt>' extends a floating point <tt>value</tt> to a larger
2767 floating point value.</p>
2770 <p>The '<tt>fpext</tt>' instruction takes a
2771 <a href="#t_floating">floating point</a> <tt>value</tt> to cast,
2772 and a <a href="#t_floating">floating point</a> type to cast it to. The source
2773 type must be smaller than the destination type.</p>
2776 <p>The '<tt>fpext</tt>' instruction extends the <tt>value</tt> from a smaller
2777 <a href="t_floating">floating point</a> type to a larger
2778 <a href="t_floating">floating point</a> type. The <tt>fpext</tt> cannot be
2779 used to make a <i>no-op cast</i> because it always changes bits. Use
2780 <tt>bitcast</tt> to make a <i>no-op cast</i> for a floating point cast.</p>
2784 %X = fpext float 3.1415 to double <i>; yields double:3.1415</i>
2785 %Y = fpext float 1.0 to float <i>; yields float:1.0 (no-op)</i>
2789 <!-- _______________________________________________________________________ -->
2790 <div class="doc_subsubsection">
2791 <a name="i_fp2uint">'<tt>fptoui .. to</tt>' Instruction</a>
2793 <div class="doc_text">
2797 <result> = fp2uint <ty> <value> to <ty2> <i>; yields ty2</i>
2801 <p>The '<tt>fp2uint</tt>' converts a floating point <tt>value</tt> to its
2802 unsigned integer equivalent of type <tt>ty2</tt>.
2806 <p>The '<tt>fp2uint</tt>' instruction takes a value to cast, which must be a
2807 <a href="#t_floating">floating point</a> value, and a type to cast it to, which
2808 must be an <a href="#t_integral">integral</a> type.</p>
2811 <p> The '<tt>fp2uint</tt>' instruction converts its
2812 <a href="#t_floating">floating point</a> operand into the nearest (rounding
2813 towards zero) unsigned integer value. If the value cannot fit in <tt>ty2</tt>,
2814 the results are undefined.</p>
2816 <p>When converting to bool, the conversion is done as a comparison against
2817 zero. If the <tt>value</tt> was zero, the bool result will be <tt>false</tt>.
2818 If the <tt>value</tt> was non-zero, the bool result will be <tt>true</tt>.</p>
2822 %X = fp2uint double 123.0 to int <i>; yields int:123</i>
2823 %Y = fp2uint float 1.0E+300 to bool <i>; yields bool:true</i>
2824 %X = fp2uint float 1.04E+17 to ubyte <i>; yields undefined:1</i>
2828 <!-- _______________________________________________________________________ -->
2829 <div class="doc_subsubsection">
2830 <a name="i_fptosi">'<tt>fptosi .. to</tt>' Instruction</a>
2832 <div class="doc_text">
2836 <result> = fptosi <ty> <value> to <ty2> <i>; yields ty2</i>
2840 <p>The '<tt>fptosi</tt>' instruction converts
2841 <a href="#t_floating">floating point</a> <tt>value</tt> to type <tt>ty2</tt>.
2846 <p> The '<tt>fptosi</tt>' instruction takes a value to cast, which must be a
2847 <a href="#t_floating">floating point</a> value, and a type to cast it to, which
2848 must also be an <a href="#t_integral">integral</a> type.</p>
2851 <p>The '<tt>fptosi</tt>' instruction converts its
2852 <a href="#t_floating">floating point</a> operand into the nearest (rounding
2853 towards zero) signed integer value. If the value cannot fit in <tt>ty2</tt>,
2854 the results are undefined.</p>
2856 <p>When converting to bool, the conversion is done as a comparison against
2857 zero. If the <tt>value</tt> was zero, the bool result will be <tt>false</tt>.
2858 If the <tt>value</tt> was non-zero, the bool result will be <tt>true</tt>.</p>
2862 %X = fptosi double -123.0 to int <i>; yields int:-123</i>
2863 %Y = fptosi float 1.0E-247 to bool <i>; yields bool:true</i>
2864 %X = fptosi float 1.04E+17 to sbyte <i>; yields undefined:1</i>
2868 <!-- _______________________________________________________________________ -->
2869 <div class="doc_subsubsection">
2870 <a name="i_uitofp">'<tt>uitofp .. to</tt>' Instruction</a>
2872 <div class="doc_text">
2876 <result> = uitofp <ty> <value> to <ty2> <i>; yields ty2</i>
2880 <p>The '<tt>uitofp</tt>' instruction regards <tt>value</tt> as an unsigned
2881 integer and converts that value to the <tt>ty2</tt> type.</p>
2885 <p>The '<tt>uitofp</tt>' instruction takes a value to cast, which must be an
2886 <a href="#t_integral">integral</a> value, and a type to cast it to, which must
2887 be a <a href="#t_floating">floating point</a> type.</p>
2890 <p>The '<tt>uitofp</tt>' instruction interprets its operand as an unsigned
2891 integer quantity and converts it to the corresponding floating point value. If
2892 the value cannot fit in the floating point value, the results are undefined.</p>
2897 %X = uitofp int 257 to float <i>; yields float:257.0</i>
2898 %Y = uitofp sbyte -1 to double <i>; yields double:255.0</i>
2902 <!-- _______________________________________________________________________ -->
2903 <div class="doc_subsubsection">
2904 <a name="i_sitofp">'<tt>sitofp .. to</tt>' Instruction</a>
2906 <div class="doc_text">
2910 <result> = sitofp <ty> <value> to <ty2> <i>; yields ty2</i>
2914 <p>The '<tt>sitofp</tt>' instruction regards <tt>value</tt> as a signed
2915 integer and converts that value to the <tt>ty2</tt> type.</p>
2918 <p>The '<tt>sitofp</tt>' instruction takes a value to cast, which must be an
2919 <a href="#t_integral">integral</a> value, and a type to cast it to, which must be
2920 a <a href="#t_floating">floating point</a> type.</p>
2923 <p>The '<tt>sitofp</tt>' instruction interprets its operand as a signed
2924 integer quantity and converts it to the corresponding floating point value. If
2925 the value cannot fit in the floating point value, the results are undefined.</p>
2929 %X = sitofp int 257 to float <i>; yields float:257.0</i>
2930 %Y = sitofp sbyte -1 to double <i>; yields double:-1.0</i>
2934 <!-- _______________________________________________________________________ -->
2935 <div class="doc_subsubsection">
2936 <a name="i_ptrtoint">'<tt>ptrtoint .. to</tt>' Instruction</a>
2938 <div class="doc_text">
2942 <result> = ptrtoint <ty> <value> to <ty2> <i>; yields ty2</i>
2946 <p>The '<tt>ptrtoint</tt>' instruction converts the pointer <tt>value</tt> to
2947 the integer type <tt>ty2</tt>.</p>
2950 <p>The '<tt>ptrtoint</tt>' instruction takes a <tt>value</tt> to cast, which
2951 must be a <a href="t_pointer">pointer</a> value, and a type to cast it to
2952 <tt>ty2</tt>, which must be an <a href="#t_integer">integer</a> type.
2955 <p>The '<tt>ptrtoint</tt>' instruction converts <tt>value</tt> to integer type
2956 <tt>ty2</tt> by interpreting the pointer value as an integer and either
2957 truncating or zero extending that value to the size of the integer type. If
2958 <tt>value</tt> is smaller than <tt>ty2</tt> then a zero extension is done. If
2959 <tt>value</tt> is larger than <tt>ty2</tt> then a truncation is done. If they
2960 are the same size, then nothing is done (<i>no-op cast</i>).</p>
2964 %X = ptrtoint int* %X to sbyte <i>; yields truncation on 32-bit</i>
2965 %Y = ptrtoint int* %x to ulong <i>; yields zero extend on 32-bit</i>
2969 <!-- _______________________________________________________________________ -->
2970 <div class="doc_subsubsection">
2971 <a name="i_inttoptr">'<tt>inttoptr .. to</tt>' Instruction</a>
2973 <div class="doc_text">
2977 <result> = inttoptr <ty> <value> to <ty2> <i>; yields ty2</i>
2981 <p>The '<tt>inttoptr</tt>' instruction converts an integer <tt>value</tt> to
2982 a pointer type, <tt>ty2</tt>.</p>
2985 <p>The '<tt>inttoptr</tt>' instruction takes an <a href="i_integer">integer</a>
2986 value to cast, and a type to cast it to, which must be a
2987 <a href="#t_pointer">pointer</a> type. </tt>
2990 <p>The '<tt>inttoptr</tt>' instruction converts <tt>value</tt> to type
2991 <tt>ty2</tt> by applying either a zero extension or a truncation depending on
2992 the size of the integer <tt>value</tt>. If <tt>value</tt> is larger than the
2993 size of a pointer then a truncation is done. If <tt>value</tt> is smaller than
2994 the size of a pointer then a zero extension is done. If they are the same size,
2995 nothing is done (<i>no-op cast</i>).</p>
2999 %X = inttoptr int 255 to int* <i>; yields zero extend on 64-bit</i>
3000 %X = inttoptr int 255 to int* <i>; yields no-op on 32-bit </i>
3001 %Y = inttoptr short 0 to int* <i>; yields zero extend on 32-bit</i>
3005 <!-- _______________________________________________________________________ -->
3006 <div class="doc_subsubsection">
3007 <a name="i_bitcast">'<tt>bitcast .. to</tt>' Instruction</a>
3009 <div class="doc_text">
3013 <result> = bitcast <ty> <value> to <ty2> <i>; yields ty2</i>
3017 <p>The '<tt>bitcast</tt>' instruction converts <tt>value</tt> to type
3018 <tt>ty2</tt> without changing any bits.</p>
3021 <p>The '<tt>bitcast</tt>' instruction takes a value to cast, which must be
3022 a first class value, and a type to cast it to, which must also be a <a
3023 href="#t_firstclass">first class</a> type. The bit sizes of <tt>value</tt>
3024 and the destination type, <tt>ty2</tt>, must be identical.</p>
3027 <p>The '<tt>bitcast</tt>' instruction converts <tt>value</tt> to type
3028 <tt>ty2</tt>. It is always a <i>no-op cast</i> because no bits change with
3029 this conversion. The conversion is done as if the <tt>value</tt> had been
3030 stored to memory and read back as type <tt>ty2</tt>. Pointer types may only be
3031 converted to other pointer types with this instruction. To convert pointers to
3032 other types, use the <a href="#i_inttoptr">inttoptr</a> or
3033 <a href="#i_ptrtoint">ptrtoint</a> instructions first.</p>
3037 %X = bitcast ubyte 255 to sbyte <i>; yields sbyte:-1</i>
3038 %Y = bitcast uint* %x to sint* <i>; yields sint*:%x</i>
3039 %Z = bitcast <2xint> %V to long; <i>; yields long: %V</i>
3043 <!-- ======================================================================= -->
3044 <div class="doc_subsection"> <a name="otherops">Other Operations</a> </div>
3045 <div class="doc_text">
3046 <p>The instructions in this category are the "miscellaneous"
3047 instructions, which defy better classification.</p>
3050 <!-- _______________________________________________________________________ -->
3051 <div class="doc_subsubsection"><a name="i_icmp">'<tt>icmp</tt>' Instruction</a>
3053 <div class="doc_text">
3055 <pre> <result> = icmp <cond> <ty> <var1>, <var2> <i>; yields {bool}:result</i>
3058 <p>The '<tt>icmp</tt>' instruction returns a boolean value based on comparison
3059 of its two integer operands.</p>
3061 <p>The '<tt>icmp</tt>' instruction takes three operands. The first operand is
3062 the condition code which indicates the kind of comparison to perform. It is not
3063 a value, just a keyword. The possibilities for the condition code are:
3065 <li><tt>eq</tt>: equal</li>
3066 <li><tt>ne</tt>: not equal </li>
3067 <li><tt>ugt</tt>: unsigned greater than</li>
3068 <li><tt>uge</tt>: unsigned greater or equal</li>
3069 <li><tt>ult</tt>: unsigned less than</li>
3070 <li><tt>ule</tt>: unsigned less or equal</li>
3071 <li><tt>sgt</tt>: signed greater than</li>
3072 <li><tt>sge</tt>: signed greater or equal</li>
3073 <li><tt>slt</tt>: signed less than</li>
3074 <li><tt>sle</tt>: signed less or equal</li>
3076 <p>The remaining two arguments must be of <a href="#t_integral">integral</a>,
3077 <a href="#t_pointer">pointer</a> or a <a href="#t_packed">packed</a> integral
3078 type. They must have identical types.</p>
3080 <p>The '<tt>icmp</tt>' compares <tt>var1</tt> and <tt>var2</tt> according to
3081 the condition code given as <tt>cond</tt>. The comparison performed always
3082 yields a <a href="#t_bool">bool</a> result, as follows:
3084 <li><tt>eq</tt>: yields <tt>true</tt> if the operands are equal,
3085 <tt>false</tt> otherwise. No sign interpretation is necessary or performed.
3087 <li><tt>ne</tt>: yields <tt>true</tt> if the operands are unequal,
3088 <tt>false</tt> otherwise. No sign interpretation is necessary or performed.
3089 <li><tt>ugt</tt>: interprets the operands as unsigned values and yields
3090 <tt>true</tt> if <tt>var1</tt> is greater than <tt>var2</tt>.</li>
3091 <li><tt>uge</tt>: interprets the operands as unsigned values and yields
3092 <tt>true</tt> if <tt>var1</tt> is greater than or equal to <tt>var2</tt>.</li>
3093 <li><tt>ult</tt>: interprets the operands as unsigned values and yields
3094 <tt>true</tt> if <tt>var1</tt> is less than <tt>var2</tt>.</li>
3095 <li><tt>ule</tt>: interprets the operands as unsigned values and yields
3096 <tt>true</tt> if <tt>var1</tt> is less than or equal to <tt>var2</tt>.</li>
3097 <li><tt>sgt</tt>: interprets the operands as signed values and yields
3098 <tt>true</tt> if <tt>var1</tt> is greater than <tt>var2</tt>.</li>
3099 <li><tt>sge</tt>: interprets the operands as signed values and yields
3100 <tt>true</tt> if <tt>var1</tt> is greater than or equal to <tt>var2</tt>.</li>
3101 <li><tt>slt</tt>: interprets the operands as signed values and yields
3102 <tt>true</tt> if <tt>var1</tt> is less than <tt>var2</tt>.</li>
3103 <li><tt>sle</tt>: interprets the operands as signed values and yields
3104 <tt>true</tt> if <tt>var1</tt> is less than or equal to <tt>var2</tt>.</li>
3107 <p>If the operands are <a href="#t_pointer">pointer</a> typed, the pointer
3108 values are treated as integers and then compared.</p>
3109 <p>If the operands are <a href="#t_packed">packed</a> typed, the elements of
3110 the vector are compared in turn and the predicate must hold for all
3114 <pre> <result> = icmp eq int 4, 5 <i>; yields: result=false</i>
3115 <result> = icmp ne float* %X, %X <i>; yields: result=false</i>
3116 <result> = icmp ult short 4, 5 <i>; yields: result=true</i>
3117 <result> = icmp sgt sbyte 4, 5 <i>; yields: result=false</i>
3118 <result> = icmp ule sbyte -4, 5 <i>; yields: result=false</i>
3119 <result> = icmp sge sbyte 4, 5 <i>; yields: result=false</i>
3123 <!-- _______________________________________________________________________ -->
3124 <div class="doc_subsubsection"><a name="i_fcmp">'<tt>fcmp</tt>' Instruction</a>
3126 <div class="doc_text">
3128 <pre> <result> = fcmp <cond> <ty> <var1>, <var2> <i>; yields {bool}:result</i>
3131 <p>The '<tt>fcmp</tt>' instruction returns a boolean value based on comparison
3132 of its floating point operands.</p>
3134 <p>The '<tt>fcmp</tt>' instruction takes three operands. The first operand is
3135 the condition code which indicates the kind of comparison to perform. It is not
3136 a value, just a keyword. The possibilities for the condition code are:
3138 <li><tt>false</tt>: no comparison, always returns false</li>
3139 <li><tt>oeq</tt>: ordered and equal</li>
3140 <li><tt>ogt</tt>: ordered and greater than </li>
3141 <li><tt>oge</tt>: ordered and greater than or equal</li>
3142 <li><tt>olt</tt>: ordered and less than </li>
3143 <li><tt>ole</tt>: ordered and less than or equal</li>
3144 <li><tt>one</tt>: ordered and not equal</li>
3145 <li><tt>ord</tt>: ordered (no nans)</li>
3146 <li><tt>ueq</tt>: unordered or equal</li>
3147 <li><tt>ugt</tt>: unordered or greater than </li>
3148 <li><tt>uge</tt>: unordered or greater than or equal</li>
3149 <li><tt>ult</tt>: unordered or less than </li>
3150 <li><tt>ule</tt>: unordered or less than or equal</li>
3151 <li><tt>une</tt>: unordered or not equal</li>
3152 <li><tt>uno</tt>: unordered (either nans)</li>
3153 <li><tt>true</tt>: no comparison, always returns true</li>
3155 <p>In the preceding, <i>ordered</i> means that neither operand is a QNAN while
3156 <i>unordered</i> means that either operand may be a QNAN.</p>
3157 <p>The <tt>val1</tt> and <tt>val2</tt> arguments must be of
3158 <a href="#t_floating">floating point</a>, or a <a href="#t_packed">packed</a>
3159 floating point type. They must have identical types.</p>
3160 <p>In the foregoing, <i>ordered</i> means that neither operand is a QNAN and
3161 <i>unordered</i> means that either operand is a QNAN.</p>
3163 <p>The '<tt>fcmp</tt>' compares <tt>var1</tt> and <tt>var2</tt> according to
3164 the condition code given as <tt>cond</tt>. The comparison performed always
3165 yields a <a href="#t_bool">bool</a> result, as follows:
3167 <li><tt>false</tt>: always yields <tt>false</tt>, regardless of operands.</li>
3168 <li><tt>oeq</tt>: yields <tt>true</tt> if both operands are not a QNAN and
3169 <tt>var1</tt> is equal to <tt>var2</tt>.</li>
3170 <li><tt>ogt</tt>: yields <tt>true</tt> if both operands are not a QNAN and
3171 <tt>var1</tt> is greather than <tt>var2</tt>.</li>
3172 <li><tt>oge</tt>: yields <tt>true</tt> if both operands are not a QNAN and
3173 <tt>var1</tt> is greater than or equal to <tt>var2</tt>.</li>
3174 <li><tt>olt</tt>: yields <tt>true</tt> if both operands are not a QNAN and
3175 <tt>var1</tt> is less than <tt>var2</tt>.</li>
3176 <li><tt>ole</tt>: yields <tt>true</tt> if both operands are not a QNAN and
3177 <tt>var1</tt> is less than or equal to <tt>var2</tt>.</li>
3178 <li><tt>one</tt>: yields <tt>true</tt> if both operands are not a QNAN and
3179 <tt>var1</tt> is not equal to <tt>var2</tt>.</li>
3180 <li><tt>ord</tt>: yields <tt>true</tt> if both operands are not a QNAN.</li>
3181 <li><tt>ueq</tt>: yields <tt>true</tt> if either operand is a QNAN or
3182 <tt>var1</tt> is equal to <tt>var2</tt>.</li>
3183 <li><tt>ugt</tt>: yields <tt>true</tt> if either operand is a QNAN or
3184 <tt>var1</tt> is greater than <tt>var2</tt>.</li>
3185 <li><tt>uge</tt>: yields <tt>true</tt> if either operand is a QNAN or
3186 <tt>var1</tt> is greater than or equal to <tt>var2</tt>.</li>
3187 <li><tt>ult</tt>: yields <tt>true</tt> if either operand is a QNAN or
3188 <tt>var1</tt> is less than <tt>var2</tt>.</li>
3189 <li><tt>ule</tt>: yields <tt>true</tt> if either operand is a QNAN or
3190 <tt>var1</tt> is less than or equal to <tt>var2</tt>.</li>
3191 <li><tt>une</tt>: yields <tt>true</tt> if either operand is a QNAN or
3192 <tt>var1</tt> is not equal to <tt>var2</tt>.</li>
3193 <li><tt>uno</tt>: yields <tt>true</tt> if either operand is a QNAN.</li>
3194 <li><tt>true</tt>: always yields <tt>true</tt>, regardless of operands.</li>
3196 <p>If the operands are <a href="#t_packed">packed</a> typed, the elements of
3197 the vector are compared in turn and the predicate must hold for all elements.
3201 <pre> <result> = fcmp oeq float 4.0, 5.0 <i>; yields: result=false</i>
3202 <result> = icmp one float 4.0, 5.0 <i>; yields: result=true</i>
3203 <result> = icmp olt float 4.0, 5.0 <i>; yields: result=true</i>
3204 <result> = icmp ueq double 1.0, 2.0 <i>; yields: result=false</i>
3208 <!-- _______________________________________________________________________ -->
3209 <div class="doc_subsubsection"> <a name="i_phi">'<tt>phi</tt>'
3210 Instruction</a> </div>
3211 <div class="doc_text">
3213 <pre> <result> = phi <ty> [ <val0>, <label0>], ...<br></pre>
3215 <p>The '<tt>phi</tt>' instruction is used to implement the φ node in
3216 the SSA graph representing the function.</p>
3218 <p>The type of the incoming values are specified with the first type
3219 field. After this, the '<tt>phi</tt>' instruction takes a list of pairs
3220 as arguments, with one pair for each predecessor basic block of the
3221 current block. Only values of <a href="#t_firstclass">first class</a>
3222 type may be used as the value arguments to the PHI node. Only labels
3223 may be used as the label arguments.</p>
3224 <p>There must be no non-phi instructions between the start of a basic
3225 block and the PHI instructions: i.e. PHI instructions must be first in
3228 <p>At runtime, the '<tt>phi</tt>' instruction logically takes on the
3229 value specified by the parameter, depending on which basic block we
3230 came from in the last <a href="#terminators">terminator</a> instruction.</p>
3232 <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>
3235 <!-- _______________________________________________________________________ -->
3236 <div class="doc_subsubsection">
3237 <a name="i_select">'<tt>select</tt>' Instruction</a>
3240 <div class="doc_text">
3245 <result> = select bool <cond>, <ty> <val1>, <ty> <val2> <i>; yields ty</i>
3251 The '<tt>select</tt>' instruction is used to choose one value based on a
3252 condition, without branching.
3259 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.
3265 If the boolean condition evaluates to true, the instruction returns the first
3266 value argument; otherwise, it returns the second value argument.
3272 %X = select bool true, ubyte 17, ubyte 42 <i>; yields ubyte:17</i>
3277 <!-- _______________________________________________________________________ -->
3278 <div class="doc_subsubsection">
3279 <a name="i_call">'<tt>call</tt>' Instruction</a>
3282 <div class="doc_text">
3286 <result> = [tail] call [<a href="#callingconv">cconv</a>] <ty>* <fnptrval>(<param list>)
3291 <p>The '<tt>call</tt>' instruction represents a simple function call.</p>
3295 <p>This instruction requires several arguments:</p>
3299 <p>The optional "tail" marker indicates whether the callee function accesses
3300 any allocas or varargs in the caller. If the "tail" marker is present, the
3301 function call is eligible for tail call optimization. Note that calls may
3302 be marked "tail" even if they do not occur before a <a
3303 href="#i_ret"><tt>ret</tt></a> instruction.
3306 <p>The optional "cconv" marker indicates which <a href="callingconv">calling
3307 convention</a> the call should use. If none is specified, the call defaults
3308 to using C calling conventions.
3311 <p>'<tt>ty</tt>': shall be the signature of the pointer to function value
3312 being invoked. The argument types must match the types implied by this
3313 signature. This type can be omitted if the function is not varargs and
3314 if the function type does not return a pointer to a function.</p>
3317 <p>'<tt>fnptrval</tt>': An LLVM value containing a pointer to a function to
3318 be invoked. In most cases, this is a direct function invocation, but
3319 indirect <tt>call</tt>s are just as possible, calling an arbitrary pointer
3320 to function value.</p>
3323 <p>'<tt>function args</tt>': argument list whose types match the
3324 function signature argument types. All arguments must be of
3325 <a href="#t_firstclass">first class</a> type. If the function signature
3326 indicates the function accepts a variable number of arguments, the extra
3327 arguments can be specified.</p>
3333 <p>The '<tt>call</tt>' instruction is used to cause control flow to
3334 transfer to a specified function, with its incoming arguments bound to
3335 the specified values. Upon a '<tt><a href="#i_ret">ret</a></tt>'
3336 instruction in the called function, control flow continues with the
3337 instruction after the function call, and the return value of the
3338 function is bound to the result argument. This is a simpler case of
3339 the <a href="#i_invoke">invoke</a> instruction.</p>
3344 %retval = call int %test(int %argc)
3345 call int(sbyte*, ...) *%printf(sbyte* %msg, int 12, sbyte 42);
3346 %X = tail call int %foo()
3347 %Y = tail call <a href="#callingconv">fastcc</a> int %foo()
3352 <!-- _______________________________________________________________________ -->
3353 <div class="doc_subsubsection">
3354 <a name="i_va_arg">'<tt>va_arg</tt>' Instruction</a>
3357 <div class="doc_text">
3362 <resultval> = va_arg <va_list*> <arglist>, <argty>
3367 <p>The '<tt>va_arg</tt>' instruction is used to access arguments passed through
3368 the "variable argument" area of a function call. It is used to implement the
3369 <tt>va_arg</tt> macro in C.</p>
3373 <p>This instruction takes a <tt>va_list*</tt> value and the type of
3374 the argument. It returns a value of the specified argument type and
3375 increments the <tt>va_list</tt> to point to the next argument. Again, the
3376 actual type of <tt>va_list</tt> is target specific.</p>
3380 <p>The '<tt>va_arg</tt>' instruction loads an argument of the specified
3381 type from the specified <tt>va_list</tt> and causes the
3382 <tt>va_list</tt> to point to the next argument. For more information,
3383 see the variable argument handling <a href="#int_varargs">Intrinsic
3386 <p>It is legal for this instruction to be called in a function which does not
3387 take a variable number of arguments, for example, the <tt>vfprintf</tt>
3390 <p><tt>va_arg</tt> is an LLVM instruction instead of an <a
3391 href="#intrinsics">intrinsic function</a> because it takes a type as an
3396 <p>See the <a href="#int_varargs">variable argument processing</a> section.</p>
3400 <!-- *********************************************************************** -->
3401 <div class="doc_section"> <a name="intrinsics">Intrinsic Functions</a> </div>
3402 <!-- *********************************************************************** -->
3404 <div class="doc_text">
3406 <p>LLVM supports the notion of an "intrinsic function". These functions have
3407 well known names and semantics and are required to follow certain
3408 restrictions. Overall, these instructions represent an extension mechanism for
3409 the LLVM language that does not require changing all of the transformations in
3410 LLVM to add to the language (or the bytecode reader/writer, the parser,
3413 <p>Intrinsic function names must all start with an "<tt>llvm.</tt>" prefix. This
3414 prefix is reserved in LLVM for intrinsic names; thus, functions may not be named
3415 this. Intrinsic functions must always be external functions: you cannot define
3416 the body of intrinsic functions. Intrinsic functions may only be used in call
3417 or invoke instructions: it is illegal to take the address of an intrinsic
3418 function. Additionally, because intrinsic functions are part of the LLVM
3419 language, it is required that they all be documented here if any are added.</p>
3422 <p>To learn how to add an intrinsic function, please see the <a
3423 href="ExtendingLLVM.html">Extending LLVM Guide</a>.
3428 <!-- ======================================================================= -->
3429 <div class="doc_subsection">
3430 <a name="int_varargs">Variable Argument Handling Intrinsics</a>
3433 <div class="doc_text">
3435 <p>Variable argument support is defined in LLVM with the <a
3436 href="#i_va_arg"><tt>va_arg</tt></a> instruction and these three
3437 intrinsic functions. These functions are related to the similarly
3438 named macros defined in the <tt><stdarg.h></tt> header file.</p>
3440 <p>All of these functions operate on arguments that use a
3441 target-specific value type "<tt>va_list</tt>". The LLVM assembly
3442 language reference manual does not define what this type is, so all
3443 transformations should be prepared to handle intrinsics with any type
3446 <p>This example shows how the <a href="#i_va_arg"><tt>va_arg</tt></a>
3447 instruction and the variable argument handling intrinsic functions are
3451 int %test(int %X, ...) {
3452 ; Initialize variable argument processing
3454 call void %<a href="#i_va_start">llvm.va_start</a>(sbyte** %ap)
3456 ; Read a single integer argument
3457 %tmp = va_arg sbyte** %ap, int
3459 ; Demonstrate usage of llvm.va_copy and llvm.va_end
3461 call void %<a href="#i_va_copy">llvm.va_copy</a>(sbyte** %aq, sbyte** %ap)
3462 call void %<a href="#i_va_end">llvm.va_end</a>(sbyte** %aq)
3464 ; Stop processing of arguments.
3465 call void %<a href="#i_va_end">llvm.va_end</a>(sbyte** %ap)
3471 <!-- _______________________________________________________________________ -->
3472 <div class="doc_subsubsection">
3473 <a name="i_va_start">'<tt>llvm.va_start</tt>' Intrinsic</a>
3477 <div class="doc_text">
3479 <pre> declare void %llvm.va_start(<va_list>* <arglist>)<br></pre>
3481 <P>The '<tt>llvm.va_start</tt>' intrinsic initializes
3482 <tt>*<arglist></tt> for subsequent use by <tt><a
3483 href="#i_va_arg">va_arg</a></tt>.</p>
3487 <P>The argument is a pointer to a <tt>va_list</tt> element to initialize.</p>
3491 <P>The '<tt>llvm.va_start</tt>' intrinsic works just like the <tt>va_start</tt>
3492 macro available in C. In a target-dependent way, it initializes the
3493 <tt>va_list</tt> element the argument points to, so that the next call to
3494 <tt>va_arg</tt> will produce the first variable argument passed to the function.
3495 Unlike the C <tt>va_start</tt> macro, this intrinsic does not need to know the
3496 last argument of the function, the compiler can figure that out.</p>
3500 <!-- _______________________________________________________________________ -->
3501 <div class="doc_subsubsection">
3502 <a name="i_va_end">'<tt>llvm.va_end</tt>' Intrinsic</a>
3505 <div class="doc_text">
3507 <pre> declare void %llvm.va_end(<va_list*> <arglist>)<br></pre>
3509 <p>The '<tt>llvm.va_end</tt>' intrinsic destroys <tt><arglist></tt>
3510 which has been initialized previously with <tt><a href="#i_va_start">llvm.va_start</a></tt>
3511 or <tt><a href="#i_va_copy">llvm.va_copy</a></tt>.</p>
3513 <p>The argument is a <tt>va_list</tt> to destroy.</p>
3515 <p>The '<tt>llvm.va_end</tt>' intrinsic works just like the <tt>va_end</tt>
3516 macro available in C. In a target-dependent way, it destroys the <tt>va_list</tt>.
3517 Calls to <a href="#i_va_start"><tt>llvm.va_start</tt></a> and <a
3518 href="#i_va_copy"><tt>llvm.va_copy</tt></a> must be matched exactly
3519 with calls to <tt>llvm.va_end</tt>.</p>
3522 <!-- _______________________________________________________________________ -->
3523 <div class="doc_subsubsection">
3524 <a name="i_va_copy">'<tt>llvm.va_copy</tt>' Intrinsic</a>
3527 <div class="doc_text">
3532 declare void %llvm.va_copy(<va_list>* <destarglist>,
3533 <va_list>* <srcarglist>)
3538 <p>The '<tt>llvm.va_copy</tt>' intrinsic copies the current argument position from
3539 the source argument list to the destination argument list.</p>
3543 <p>The first argument is a pointer to a <tt>va_list</tt> element to initialize.
3544 The second argument is a pointer to a <tt>va_list</tt> element to copy from.</p>
3549 <p>The '<tt>llvm.va_copy</tt>' intrinsic works just like the <tt>va_copy</tt> macro
3550 available in C. In a target-dependent way, it copies the source
3551 <tt>va_list</tt> element into the destination list. This intrinsic is necessary
3552 because the <tt><a href="i_va_begin">llvm.va_begin</a></tt> intrinsic may be
3553 arbitrarily complex and require memory allocation, for example.</p>
3557 <!-- ======================================================================= -->
3558 <div class="doc_subsection">
3559 <a name="int_gc">Accurate Garbage Collection Intrinsics</a>
3562 <div class="doc_text">
3565 LLVM support for <a href="GarbageCollection.html">Accurate Garbage
3566 Collection</a> requires the implementation and generation of these intrinsics.
3567 These intrinsics allow identification of <a href="#i_gcroot">GC roots on the
3568 stack</a>, as well as garbage collector implementations that require <a
3569 href="#i_gcread">read</a> and <a href="#i_gcwrite">write</a> barriers.
3570 Front-ends for type-safe garbage collected languages should generate these
3571 intrinsics to make use of the LLVM garbage collectors. For more details, see <a
3572 href="GarbageCollection.html">Accurate Garbage Collection with LLVM</a>.
3576 <!-- _______________________________________________________________________ -->
3577 <div class="doc_subsubsection">
3578 <a name="i_gcroot">'<tt>llvm.gcroot</tt>' Intrinsic</a>
3581 <div class="doc_text">
3586 declare void %llvm.gcroot(<ty>** %ptrloc, <ty2>* %metadata)
3591 <p>The '<tt>llvm.gcroot</tt>' intrinsic declares the existence of a GC root to
3592 the code generator, and allows some metadata to be associated with it.</p>
3596 <p>The first argument specifies the address of a stack object that contains the
3597 root pointer. The second pointer (which must be either a constant or a global
3598 value address) contains the meta-data to be associated with the root.</p>
3602 <p>At runtime, a call to this intrinsics stores a null pointer into the "ptrloc"
3603 location. At compile-time, the code generator generates information to allow
3604 the runtime to find the pointer at GC safe points.
3610 <!-- _______________________________________________________________________ -->
3611 <div class="doc_subsubsection">
3612 <a name="i_gcread">'<tt>llvm.gcread</tt>' Intrinsic</a>
3615 <div class="doc_text">
3620 declare sbyte* %llvm.gcread(sbyte* %ObjPtr, sbyte** %Ptr)
3625 <p>The '<tt>llvm.gcread</tt>' intrinsic identifies reads of references from heap
3626 locations, allowing garbage collector implementations that require read
3631 <p>The second argument is the address to read from, which should be an address
3632 allocated from the garbage collector. The first object is a pointer to the
3633 start of the referenced object, if needed by the language runtime (otherwise
3638 <p>The '<tt>llvm.gcread</tt>' intrinsic has the same semantics as a load
3639 instruction, but may be replaced with substantially more complex code by the
3640 garbage collector runtime, as needed.</p>
3645 <!-- _______________________________________________________________________ -->
3646 <div class="doc_subsubsection">
3647 <a name="i_gcwrite">'<tt>llvm.gcwrite</tt>' Intrinsic</a>
3650 <div class="doc_text">
3655 declare void %llvm.gcwrite(sbyte* %P1, sbyte* %Obj, sbyte** %P2)
3660 <p>The '<tt>llvm.gcwrite</tt>' intrinsic identifies writes of references to heap
3661 locations, allowing garbage collector implementations that require write
3662 barriers (such as generational or reference counting collectors).</p>
3666 <p>The first argument is the reference to store, the second is the start of the
3667 object to store it to, and the third is the address of the field of Obj to
3668 store to. If the runtime does not require a pointer to the object, Obj may be
3673 <p>The '<tt>llvm.gcwrite</tt>' intrinsic has the same semantics as a store
3674 instruction, but may be replaced with substantially more complex code by the
3675 garbage collector runtime, as needed.</p>
3681 <!-- ======================================================================= -->
3682 <div class="doc_subsection">
3683 <a name="int_codegen">Code Generator Intrinsics</a>
3686 <div class="doc_text">
3688 These intrinsics are provided by LLVM to expose special features that may only
3689 be implemented with code generator support.
3694 <!-- _______________________________________________________________________ -->
3695 <div class="doc_subsubsection">
3696 <a name="i_returnaddress">'<tt>llvm.returnaddress</tt>' Intrinsic</a>
3699 <div class="doc_text">
3703 declare sbyte *%llvm.returnaddress(uint <level>)
3709 The '<tt>llvm.returnaddress</tt>' intrinsic attempts to compute a
3710 target-specific value indicating the return address of the current function
3711 or one of its callers.
3717 The argument to this intrinsic indicates which function to return the address
3718 for. Zero indicates the calling function, one indicates its caller, etc. The
3719 argument is <b>required</b> to be a constant integer value.
3725 The '<tt>llvm.returnaddress</tt>' intrinsic either returns a pointer indicating
3726 the return address of the specified call frame, or zero if it cannot be
3727 identified. The value returned by this intrinsic is likely to be incorrect or 0
3728 for arguments other than zero, so it should only be used for debugging purposes.
3732 Note that calling this intrinsic does not prevent function inlining or other
3733 aggressive transformations, so the value returned may not be that of the obvious
3734 source-language caller.
3739 <!-- _______________________________________________________________________ -->
3740 <div class="doc_subsubsection">
3741 <a name="i_frameaddress">'<tt>llvm.frameaddress</tt>' Intrinsic</a>
3744 <div class="doc_text">
3748 declare sbyte *%llvm.frameaddress(uint <level>)
3754 The '<tt>llvm.frameaddress</tt>' intrinsic attempts to return the
3755 target-specific frame pointer value for the specified stack frame.
3761 The argument to this intrinsic indicates which function to return the frame
3762 pointer for. Zero indicates the calling function, one indicates its caller,
3763 etc. The argument is <b>required</b> to be a constant integer value.
3769 The '<tt>llvm.frameaddress</tt>' intrinsic either returns a pointer indicating
3770 the frame address of the specified call frame, or zero if it cannot be
3771 identified. The value returned by this intrinsic is likely to be incorrect or 0
3772 for arguments other than zero, so it should only be used for debugging purposes.
3776 Note that calling this intrinsic does not prevent function inlining or other
3777 aggressive transformations, so the value returned may not be that of the obvious
3778 source-language caller.
3782 <!-- _______________________________________________________________________ -->
3783 <div class="doc_subsubsection">
3784 <a name="i_stacksave">'<tt>llvm.stacksave</tt>' Intrinsic</a>
3787 <div class="doc_text">
3791 declare sbyte *%llvm.stacksave()
3797 The '<tt>llvm.stacksave</tt>' intrinsic is used to remember the current state of
3798 the function stack, for use with <a href="#i_stackrestore">
3799 <tt>llvm.stackrestore</tt></a>. This is useful for implementing language
3800 features like scoped automatic variable sized arrays in C99.
3806 This intrinsic returns a opaque pointer value that can be passed to <a
3807 href="#i_stackrestore"><tt>llvm.stackrestore</tt></a>. When an
3808 <tt>llvm.stackrestore</tt> intrinsic is executed with a value saved from
3809 <tt>llvm.stacksave</tt>, it effectively restores the state of the stack to the
3810 state it was in when the <tt>llvm.stacksave</tt> intrinsic executed. In
3811 practice, this pops any <a href="#i_alloca">alloca</a> blocks from the stack
3812 that were allocated after the <tt>llvm.stacksave</tt> was executed.
3817 <!-- _______________________________________________________________________ -->
3818 <div class="doc_subsubsection">
3819 <a name="i_stackrestore">'<tt>llvm.stackrestore</tt>' Intrinsic</a>
3822 <div class="doc_text">
3826 declare void %llvm.stackrestore(sbyte* %ptr)
3832 The '<tt>llvm.stackrestore</tt>' intrinsic is used to restore the state of
3833 the function stack to the state it was in when the corresponding <a
3834 href="#llvm.stacksave"><tt>llvm.stacksave</tt></a> intrinsic executed. This is
3835 useful for implementing language features like scoped automatic variable sized
3842 See the description for <a href="#i_stacksave"><tt>llvm.stacksave</tt></a>.
3848 <!-- _______________________________________________________________________ -->
3849 <div class="doc_subsubsection">
3850 <a name="i_prefetch">'<tt>llvm.prefetch</tt>' Intrinsic</a>
3853 <div class="doc_text">
3857 declare void %llvm.prefetch(sbyte * <address>,
3858 uint <rw>, uint <locality>)
3865 The '<tt>llvm.prefetch</tt>' intrinsic is a hint to the code generator to insert
3866 a prefetch instruction if supported; otherwise, it is a noop. Prefetches have
3868 effect on the behavior of the program but can change its performance
3875 <tt>address</tt> is the address to be prefetched, <tt>rw</tt> is the specifier
3876 determining if the fetch should be for a read (0) or write (1), and
3877 <tt>locality</tt> is a temporal locality specifier ranging from (0) - no
3878 locality, to (3) - extremely local keep in cache. The <tt>rw</tt> and
3879 <tt>locality</tt> arguments must be constant integers.
3885 This intrinsic does not modify the behavior of the program. In particular,
3886 prefetches cannot trap and do not produce a value. On targets that support this
3887 intrinsic, the prefetch can provide hints to the processor cache for better
3893 <!-- _______________________________________________________________________ -->
3894 <div class="doc_subsubsection">
3895 <a name="i_pcmarker">'<tt>llvm.pcmarker</tt>' Intrinsic</a>
3898 <div class="doc_text">
3902 declare void %llvm.pcmarker( uint <id> )
3909 The '<tt>llvm.pcmarker</tt>' intrinsic is a method to export a Program Counter
3911 code to simulators and other tools. The method is target specific, but it is
3912 expected that the marker will use exported symbols to transmit the PC of the marker.
3913 The marker makes no guarantees that it will remain with any specific instruction
3914 after optimizations. It is possible that the presence of a marker will inhibit
3915 optimizations. The intended use is to be inserted after optimizations to allow
3916 correlations of simulation runs.
3922 <tt>id</tt> is a numerical id identifying the marker.
3928 This intrinsic does not modify the behavior of the program. Backends that do not
3929 support this intrinisic may ignore it.
3934 <!-- _______________________________________________________________________ -->
3935 <div class="doc_subsubsection">
3936 <a name="i_readcyclecounter">'<tt>llvm.readcyclecounter</tt>' Intrinsic</a>
3939 <div class="doc_text">
3943 declare ulong %llvm.readcyclecounter( )
3950 The '<tt>llvm.readcyclecounter</tt>' intrinsic provides access to the cycle
3951 counter register (or similar low latency, high accuracy clocks) on those targets
3952 that support it. On X86, it should map to RDTSC. On Alpha, it should map to RPCC.
3953 As the backing counters overflow quickly (on the order of 9 seconds on alpha), this
3954 should only be used for small timings.
3960 When directly supported, reading the cycle counter should not modify any memory.
3961 Implementations are allowed to either return a application specific value or a
3962 system wide value. On backends without support, this is lowered to a constant 0.
3967 <!-- ======================================================================= -->
3968 <div class="doc_subsection">
3969 <a name="int_libc">Standard C Library Intrinsics</a>
3972 <div class="doc_text">
3974 LLVM provides intrinsics for a few important standard C library functions.
3975 These intrinsics allow source-language front-ends to pass information about the
3976 alignment of the pointer arguments to the code generator, providing opportunity
3977 for more efficient code generation.
3982 <!-- _______________________________________________________________________ -->
3983 <div class="doc_subsubsection">
3984 <a name="i_memcpy">'<tt>llvm.memcpy</tt>' Intrinsic</a>
3987 <div class="doc_text">
3991 declare void %llvm.memcpy.i32(sbyte* <dest>, sbyte* <src>,
3992 uint <len>, uint <align>)
3993 declare void %llvm.memcpy.i64(sbyte* <dest>, sbyte* <src>,
3994 ulong <len>, uint <align>)
4000 The '<tt>llvm.memcpy.*</tt>' intrinsics copy a block of memory from the source
4001 location to the destination location.
4005 Note that, unlike the standard libc function, the <tt>llvm.memcpy.*</tt>
4006 intrinsics do not return a value, and takes an extra alignment argument.
4012 The first argument is a pointer to the destination, the second is a pointer to
4013 the source. The third argument is an integer argument
4014 specifying the number of bytes to copy, and the fourth argument is the alignment
4015 of the source and destination locations.
4019 If the call to this intrinisic has an alignment value that is not 0 or 1, then
4020 the caller guarantees that both the source and destination pointers are aligned
4027 The '<tt>llvm.memcpy.*</tt>' intrinsics copy a block of memory from the source
4028 location to the destination location, which are not allowed to overlap. It
4029 copies "len" bytes of memory over. If the argument is known to be aligned to
4030 some boundary, this can be specified as the fourth argument, otherwise it should
4036 <!-- _______________________________________________________________________ -->
4037 <div class="doc_subsubsection">
4038 <a name="i_memmove">'<tt>llvm.memmove</tt>' Intrinsic</a>
4041 <div class="doc_text">
4045 declare void %llvm.memmove.i32(sbyte* <dest>, sbyte* <src>,
4046 uint <len>, uint <align>)
4047 declare void %llvm.memmove.i64(sbyte* <dest>, sbyte* <src>,
4048 ulong <len>, uint <align>)
4054 The '<tt>llvm.memmove.*</tt>' intrinsics move a block of memory from the source
4055 location to the destination location. It is similar to the
4056 '<tt>llvm.memcmp</tt>' intrinsic but allows the two memory locations to overlap.
4060 Note that, unlike the standard libc function, the <tt>llvm.memmove.*</tt>
4061 intrinsics do not return a value, and takes an extra alignment argument.
4067 The first argument is a pointer to the destination, the second is a pointer to
4068 the source. The third argument is an integer argument
4069 specifying the number of bytes to copy, and the fourth argument is the alignment
4070 of the source and destination locations.
4074 If the call to this intrinisic has an alignment value that is not 0 or 1, then
4075 the caller guarantees that the source and destination pointers are aligned to
4082 The '<tt>llvm.memmove.*</tt>' intrinsics copy a block of memory from the source
4083 location to the destination location, which may overlap. It
4084 copies "len" bytes of memory over. If the argument is known to be aligned to
4085 some boundary, this can be specified as the fourth argument, otherwise it should
4091 <!-- _______________________________________________________________________ -->
4092 <div class="doc_subsubsection">
4093 <a name="i_memset">'<tt>llvm.memset.*</tt>' Intrinsics</a>
4096 <div class="doc_text">
4100 declare void %llvm.memset.i32(sbyte* <dest>, ubyte <val>,
4101 uint <len>, uint <align>)
4102 declare void %llvm.memset.i64(sbyte* <dest>, ubyte <val>,
4103 ulong <len>, uint <align>)
4109 The '<tt>llvm.memset.*</tt>' intrinsics fill a block of memory with a particular
4114 Note that, unlike the standard libc function, the <tt>llvm.memset</tt> intrinsic
4115 does not return a value, and takes an extra alignment argument.
4121 The first argument is a pointer to the destination to fill, the second is the
4122 byte value to fill it with, the third argument is an integer
4123 argument specifying the number of bytes to fill, and the fourth argument is the
4124 known alignment of destination location.
4128 If the call to this intrinisic has an alignment value that is not 0 or 1, then
4129 the caller guarantees that the destination pointer is aligned to that boundary.
4135 The '<tt>llvm.memset.*</tt>' intrinsics fill "len" bytes of memory starting at
4137 destination location. If the argument is known to be aligned to some boundary,
4138 this can be specified as the fourth argument, otherwise it should be set to 0 or
4144 <!-- _______________________________________________________________________ -->
4145 <div class="doc_subsubsection">
4146 <a name="i_isunordered">'<tt>llvm.isunordered.*</tt>' Intrinsic</a>
4149 <div class="doc_text">
4153 declare bool %llvm.isunordered.f32(float Val1, float Val2)
4154 declare bool %llvm.isunordered.f64(double Val1, double Val2)
4160 The '<tt>llvm.isunordered</tt>' intrinsics return true if either or both of the
4161 specified floating point values is a NAN.
4167 The arguments are floating point numbers of the same type.
4173 If either or both of the arguments is a SNAN or QNAN, it returns true, otherwise
4179 <!-- _______________________________________________________________________ -->
4180 <div class="doc_subsubsection">
4181 <a name="i_sqrt">'<tt>llvm.sqrt.*</tt>' Intrinsic</a>
4184 <div class="doc_text">
4188 declare float %llvm.sqrt.f32(float %Val)
4189 declare double %llvm.sqrt.f64(double %Val)
4195 The '<tt>llvm.sqrt</tt>' intrinsics return the sqrt of the specified operand,
4196 returning the same value as the libm '<tt>sqrt</tt>' function would. Unlike
4197 <tt>sqrt</tt> in libm, however, <tt>llvm.sqrt</tt> has undefined behavior for
4198 negative numbers (which allows for better optimization).
4204 The argument and return value are floating point numbers of the same type.
4210 This function returns the sqrt of the specified operand if it is a positive
4211 floating point number.
4215 <!-- _______________________________________________________________________ -->
4216 <div class="doc_subsubsection">
4217 <a name="i_powi">'<tt>llvm.powi.*</tt>' Intrinsic</a>
4220 <div class="doc_text">
4224 declare float %llvm.powi.f32(float %Val, int %power)
4225 declare double %llvm.powi.f64(double %Val, int %power)
4231 The '<tt>llvm.powi.*</tt>' intrinsics return the first operand raised to the
4232 specified (positive or negative) power. The order of evaluation of
4233 multiplications is not defined.
4239 The second argument is an integer power, and the first is a value to raise to
4246 This function returns the first value raised to the second power with an
4247 unspecified sequence of rounding operations.</p>
4251 <!-- ======================================================================= -->
4252 <div class="doc_subsection">
4253 <a name="int_manip">Bit Manipulation Intrinsics</a>
4256 <div class="doc_text">
4258 LLVM provides intrinsics for a few important bit manipulation operations.
4259 These allow efficient code generation for some algorithms.
4264 <!-- _______________________________________________________________________ -->
4265 <div class="doc_subsubsection">
4266 <a name="i_bswap">'<tt>llvm.bswap.*</tt>' Intrinsics</a>
4269 <div class="doc_text">
4273 declare ushort %llvm.bswap.i16(ushort <id>)
4274 declare uint %llvm.bswap.i32(uint <id>)
4275 declare ulong %llvm.bswap.i64(ulong <id>)
4281 The '<tt>llvm.bwsap</tt>' family of intrinsics is used to byteswap a 16, 32 or
4282 64 bit quantity. These are useful for performing operations on data that is not
4283 in the target's native byte order.
4289 The <tt>llvm.bswap.16</tt> intrinsic returns a ushort value that has the high and low
4290 byte of the input ushort swapped. Similarly, the <tt>llvm.bswap.i32</tt> intrinsic
4291 returns a uint value that has the four bytes of the input uint swapped, so that
4292 if the input bytes are numbered 0, 1, 2, 3 then the returned uint will have its
4293 bytes in 3, 2, 1, 0 order. The <tt>llvm.bswap.i64</tt> intrinsic extends this concept
4299 <!-- _______________________________________________________________________ -->
4300 <div class="doc_subsubsection">
4301 <a name="int_ctpop">'<tt>llvm.ctpop.*</tt>' Intrinsic</a>
4304 <div class="doc_text">
4308 declare ubyte %llvm.ctpop.i8 (ubyte <src>)
4309 declare ushort %llvm.ctpop.i16(ushort <src>)
4310 declare uint %llvm.ctpop.i32(uint <src>)
4311 declare ulong %llvm.ctpop.i64(ulong <src>)
4317 The '<tt>llvm.ctpop</tt>' family of intrinsics counts the number of bits set in a
4324 The only argument is the value to be counted. The argument may be of any
4325 unsigned integer type. The return type must match the argument type.
4331 The '<tt>llvm.ctpop</tt>' intrinsic counts the 1's in a variable.
4335 <!-- _______________________________________________________________________ -->
4336 <div class="doc_subsubsection">
4337 <a name="int_ctlz">'<tt>llvm.ctlz.*</tt>' Intrinsic</a>
4340 <div class="doc_text">
4344 declare ubyte %llvm.ctlz.i8 (ubyte <src>)
4345 declare ushort %llvm.ctlz.i16(ushort <src>)
4346 declare uint %llvm.ctlz.i32(uint <src>)
4347 declare ulong %llvm.ctlz.i64(ulong <src>)
4353 The '<tt>llvm.ctlz</tt>' family of intrinsic functions counts the number of
4354 leading zeros in a variable.
4360 The only argument is the value to be counted. The argument may be of any
4361 unsigned integer type. The return type must match the argument type.
4367 The '<tt>llvm.ctlz</tt>' intrinsic counts the leading (most significant) zeros
4368 in a variable. If the src == 0 then the result is the size in bits of the type
4369 of src. For example, <tt>llvm.ctlz(int 2) = 30</tt>.
4375 <!-- _______________________________________________________________________ -->
4376 <div class="doc_subsubsection">
4377 <a name="int_cttz">'<tt>llvm.cttz.*</tt>' Intrinsic</a>
4380 <div class="doc_text">
4384 declare ubyte %llvm.cttz.i8 (ubyte <src>)
4385 declare ushort %llvm.cttz.i16(ushort <src>)
4386 declare uint %llvm.cttz.i32(uint <src>)
4387 declare ulong %llvm.cttz.i64(ulong <src>)
4393 The '<tt>llvm.cttz</tt>' family of intrinsic functions counts the number of
4400 The only argument is the value to be counted. The argument may be of any
4401 unsigned integer type. The return type must match the argument type.
4407 The '<tt>llvm.cttz</tt>' intrinsic counts the trailing (least significant) zeros
4408 in a variable. If the src == 0 then the result is the size in bits of the type
4409 of src. For example, <tt>llvm.cttz(2) = 1</tt>.
4413 <!-- ======================================================================= -->
4414 <div class="doc_subsection">
4415 <a name="int_debugger">Debugger Intrinsics</a>
4418 <div class="doc_text">
4420 The LLVM debugger intrinsics (which all start with <tt>llvm.dbg.</tt> prefix),
4421 are described in the <a
4422 href="SourceLevelDebugging.html#format_common_intrinsics">LLVM Source Level
4423 Debugging</a> document.
4428 <!-- *********************************************************************** -->
4431 <a href="http://jigsaw.w3.org/css-validator/check/referer"><img
4432 src="http://jigsaw.w3.org/css-validator/images/vcss" alt="Valid CSS!"></a>
4433 <a href="http://validator.w3.org/check/referer"><img
4434 src="http://www.w3.org/Icons/valid-html401" alt="Valid HTML 4.01!" /></a>
4436 <a href="mailto:sabre@nondot.org">Chris Lattner</a><br>
4437 <a href="http://llvm.org">The LLVM Compiler Infrastructure</a><br>
4438 Last modified: $Date$