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="#paramattrs">Parameter Attributes</a></li>
28 <li><a href="#moduleasm">Module-Level Inline Assembly</a></li>
29 <li><a href="#datalayout">Data Layout</a></li>
32 <li><a href="#typesystem">Type System</a>
34 <li><a href="#t_primitive">Primitive Types</a>
36 <li><a href="#t_classifications">Type Classifications</a></li>
39 <li><a href="#t_derived">Derived Types</a>
41 <li><a href="#t_array">Array Type</a></li>
42 <li><a href="#t_function">Function Type</a></li>
43 <li><a href="#t_pointer">Pointer Type</a></li>
44 <li><a href="#t_struct">Structure Type</a></li>
45 <li><a href="#t_pstruct">Packed Structure Type</a></li>
46 <li><a href="#t_vector">Vector Type</a></li>
47 <li><a href="#t_opaque">Opaque Type</a></li>
52 <li><a href="#constants">Constants</a>
54 <li><a href="#simpleconstants">Simple Constants</a>
55 <li><a href="#aggregateconstants">Aggregate Constants</a>
56 <li><a href="#globalconstants">Global Variable and Function Addresses</a>
57 <li><a href="#undefvalues">Undefined Values</a>
58 <li><a href="#constantexprs">Constant Expressions</a>
61 <li><a href="#othervalues">Other Values</a>
63 <li><a href="#inlineasm">Inline Assembler Expressions</a>
66 <li><a href="#instref">Instruction Reference</a>
68 <li><a href="#terminators">Terminator Instructions</a>
70 <li><a href="#i_ret">'<tt>ret</tt>' Instruction</a></li>
71 <li><a href="#i_br">'<tt>br</tt>' Instruction</a></li>
72 <li><a href="#i_switch">'<tt>switch</tt>' Instruction</a></li>
73 <li><a href="#i_invoke">'<tt>invoke</tt>' Instruction</a></li>
74 <li><a href="#i_unwind">'<tt>unwind</tt>' Instruction</a></li>
75 <li><a href="#i_unreachable">'<tt>unreachable</tt>' Instruction</a></li>
78 <li><a href="#binaryops">Binary Operations</a>
80 <li><a href="#i_add">'<tt>add</tt>' Instruction</a></li>
81 <li><a href="#i_sub">'<tt>sub</tt>' Instruction</a></li>
82 <li><a href="#i_mul">'<tt>mul</tt>' Instruction</a></li>
83 <li><a href="#i_udiv">'<tt>udiv</tt>' Instruction</a></li>
84 <li><a href="#i_sdiv">'<tt>sdiv</tt>' Instruction</a></li>
85 <li><a href="#i_fdiv">'<tt>fdiv</tt>' Instruction</a></li>
86 <li><a href="#i_urem">'<tt>urem</tt>' Instruction</a></li>
87 <li><a href="#i_srem">'<tt>srem</tt>' Instruction</a></li>
88 <li><a href="#i_frem">'<tt>frem</tt>' Instruction</a></li>
91 <li><a href="#bitwiseops">Bitwise Binary Operations</a>
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>
96 <li><a href="#i_and">'<tt>and</tt>' Instruction</a></li>
97 <li><a href="#i_or">'<tt>or</tt>' Instruction</a></li>
98 <li><a href="#i_xor">'<tt>xor</tt>' Instruction</a></li>
101 <li><a href="#vectorops">Vector Operations</a>
103 <li><a href="#i_extractelement">'<tt>extractelement</tt>' Instruction</a></li>
104 <li><a href="#i_insertelement">'<tt>insertelement</tt>' Instruction</a></li>
105 <li><a href="#i_shufflevector">'<tt>shufflevector</tt>' Instruction</a></li>
108 <li><a href="#memoryops">Memory Access and Addressing Operations</a>
110 <li><a href="#i_malloc">'<tt>malloc</tt>' Instruction</a></li>
111 <li><a href="#i_free">'<tt>free</tt>' Instruction</a></li>
112 <li><a href="#i_alloca">'<tt>alloca</tt>' Instruction</a></li>
113 <li><a href="#i_load">'<tt>load</tt>' Instruction</a></li>
114 <li><a href="#i_store">'<tt>store</tt>' Instruction</a></li>
115 <li><a href="#i_getelementptr">'<tt>getelementptr</tt>' Instruction</a></li>
118 <li><a href="#convertops">Conversion Operations</a>
120 <li><a href="#i_trunc">'<tt>trunc .. to</tt>' Instruction</a></li>
121 <li><a href="#i_zext">'<tt>zext .. to</tt>' Instruction</a></li>
122 <li><a href="#i_sext">'<tt>sext .. to</tt>' Instruction</a></li>
123 <li><a href="#i_fptrunc">'<tt>fptrunc .. to</tt>' Instruction</a></li>
124 <li><a href="#i_fpext">'<tt>fpext .. to</tt>' Instruction</a></li>
125 <li><a href="#i_fptoui">'<tt>fptoui .. to</tt>' Instruction</a></li>
126 <li><a href="#i_fptosi">'<tt>fptosi .. to</tt>' Instruction</a></li>
127 <li><a href="#i_uitofp">'<tt>uitofp .. to</tt>' Instruction</a></li>
128 <li><a href="#i_sitofp">'<tt>sitofp .. to</tt>' Instruction</a></li>
129 <li><a href="#i_ptrtoint">'<tt>ptrtoint .. to</tt>' Instruction</a></li>
130 <li><a href="#i_inttoptr">'<tt>inttoptr .. to</tt>' Instruction</a></li>
131 <li><a href="#i_bitcast">'<tt>bitcast .. to</tt>' Instruction</a></li>
133 <li><a href="#otherops">Other Operations</a>
135 <li><a href="#i_icmp">'<tt>icmp</tt>' Instruction</a></li>
136 <li><a href="#i_fcmp">'<tt>fcmp</tt>' Instruction</a></li>
137 <li><a href="#i_phi">'<tt>phi</tt>' Instruction</a></li>
138 <li><a href="#i_select">'<tt>select</tt>' Instruction</a></li>
139 <li><a href="#i_call">'<tt>call</tt>' Instruction</a></li>
140 <li><a href="#i_va_arg">'<tt>va_arg</tt>' Instruction</a></li>
145 <li><a href="#intrinsics">Intrinsic Functions</a>
147 <li><a href="#int_varargs">Variable Argument Handling Intrinsics</a>
149 <li><a href="#i_va_start">'<tt>llvm.va_start</tt>' Intrinsic</a></li>
150 <li><a href="#i_va_end">'<tt>llvm.va_end</tt>' Intrinsic</a></li>
151 <li><a href="#i_va_copy">'<tt>llvm.va_copy</tt>' Intrinsic</a></li>
154 <li><a href="#int_gc">Accurate Garbage Collection Intrinsics</a>
156 <li><a href="#i_gcroot">'<tt>llvm.gcroot</tt>' Intrinsic</a></li>
157 <li><a href="#i_gcread">'<tt>llvm.gcread</tt>' Intrinsic</a></li>
158 <li><a href="#i_gcwrite">'<tt>llvm.gcwrite</tt>' Intrinsic</a></li>
161 <li><a href="#int_codegen">Code Generator Intrinsics</a>
163 <li><a href="#i_returnaddress">'<tt>llvm.returnaddress</tt>' Intrinsic</a></li>
164 <li><a href="#i_frameaddress">'<tt>llvm.frameaddress</tt>' Intrinsic</a></li>
165 <li><a href="#i_stacksave">'<tt>llvm.stacksave</tt>' Intrinsic</a></li>
166 <li><a href="#i_stackrestore">'<tt>llvm.stackrestore</tt>' Intrinsic</a></li>
167 <li><a href="#i_prefetch">'<tt>llvm.prefetch</tt>' Intrinsic</a></li>
168 <li><a href="#i_pcmarker">'<tt>llvm.pcmarker</tt>' Intrinsic</a></li>
169 <li><a href="#i_readcyclecounter"><tt>llvm.readcyclecounter</tt>' Intrinsic</a></li>
172 <li><a href="#int_libc">Standard C Library Intrinsics</a>
174 <li><a href="#i_memcpy">'<tt>llvm.memcpy.*</tt>' Intrinsic</a></li>
175 <li><a href="#i_memmove">'<tt>llvm.memmove.*</tt>' Intrinsic</a></li>
176 <li><a href="#i_memset">'<tt>llvm.memset.*</tt>' Intrinsic</a></li>
177 <li><a href="#i_sqrt">'<tt>llvm.sqrt.*</tt>' Intrinsic</a></li>
178 <li><a href="#i_powi">'<tt>llvm.powi.*</tt>' Intrinsic</a></li>
181 <li><a href="#int_manip">Bit Manipulation Intrinsics</a>
183 <li><a href="#i_bswap">'<tt>llvm.bswap.*</tt>' Intrinsics</a></li>
184 <li><a href="#int_ctpop">'<tt>llvm.ctpop.*</tt>' Intrinsic </a></li>
185 <li><a href="#int_ctlz">'<tt>llvm.ctlz.*</tt>' Intrinsic </a></li>
186 <li><a href="#int_cttz">'<tt>llvm.cttz.*</tt>' Intrinsic </a></li>
189 <li><a href="#int_debugger">Debugger intrinsics</a></li>
190 <li><a href="#int_eh">Exception Handling intrinsics</a></li>
195 <div class="doc_author">
196 <p>Written by <a href="mailto:sabre@nondot.org">Chris Lattner</a>
197 and <a href="mailto:vadve@cs.uiuc.edu">Vikram Adve</a></p>
200 <!-- *********************************************************************** -->
201 <div class="doc_section"> <a name="abstract">Abstract </a></div>
202 <!-- *********************************************************************** -->
204 <div class="doc_text">
205 <p>This document is a reference manual for the LLVM assembly language.
206 LLVM is an SSA based representation that provides type safety,
207 low-level operations, flexibility, and the capability of representing
208 'all' high-level languages cleanly. It is the common code
209 representation used throughout all phases of the LLVM compilation
213 <!-- *********************************************************************** -->
214 <div class="doc_section"> <a name="introduction">Introduction</a> </div>
215 <!-- *********************************************************************** -->
217 <div class="doc_text">
219 <p>The LLVM code representation is designed to be used in three
220 different forms: as an in-memory compiler IR, as an on-disk bytecode
221 representation (suitable for fast loading by a Just-In-Time compiler),
222 and as a human readable assembly language representation. This allows
223 LLVM to provide a powerful intermediate representation for efficient
224 compiler transformations and analysis, while providing a natural means
225 to debug and visualize the transformations. The three different forms
226 of LLVM are all equivalent. This document describes the human readable
227 representation and notation.</p>
229 <p>The LLVM representation aims to be light-weight and low-level
230 while being expressive, typed, and extensible at the same time. It
231 aims to be a "universal IR" of sorts, by being at a low enough level
232 that high-level ideas may be cleanly mapped to it (similar to how
233 microprocessors are "universal IR's", allowing many source languages to
234 be mapped to them). By providing type information, LLVM can be used as
235 the target of optimizations: for example, through pointer analysis, it
236 can be proven that a C automatic variable is never accessed outside of
237 the current function... allowing it to be promoted to a simple SSA
238 value instead of a memory location.</p>
242 <!-- _______________________________________________________________________ -->
243 <div class="doc_subsubsection"> <a name="wellformed">Well-Formedness</a> </div>
245 <div class="doc_text">
247 <p>It is important to note that this document describes 'well formed'
248 LLVM assembly language. There is a difference between what the parser
249 accepts and what is considered 'well formed'. For example, the
250 following instruction is syntactically okay, but not well formed:</p>
253 %x = <a href="#i_add">add</a> i32 1, %x
256 <p>...because the definition of <tt>%x</tt> does not dominate all of
257 its uses. The LLVM infrastructure provides a verification pass that may
258 be used to verify that an LLVM module is well formed. This pass is
259 automatically run by the parser after parsing input assembly and by
260 the optimizer before it outputs bytecode. The violations pointed out
261 by the verifier pass indicate bugs in transformation passes or input to
264 <!-- Describe the typesetting conventions here. --> </div>
266 <!-- *********************************************************************** -->
267 <div class="doc_section"> <a name="identifiers">Identifiers</a> </div>
268 <!-- *********************************************************************** -->
270 <div class="doc_text">
272 <p>LLVM uses three different forms of identifiers, for different
276 <li>Named values are represented as a string of characters with a '%' prefix.
277 For example, %foo, %DivisionByZero, %a.really.long.identifier. The actual
278 regular expression used is '<tt>%[a-zA-Z$._][a-zA-Z$._0-9]*</tt>'.
279 Identifiers which require other characters in their names can be surrounded
280 with quotes. In this way, anything except a <tt>"</tt> character can be used
283 <li>Unnamed values are represented as an unsigned numeric value with a '%'
284 prefix. For example, %12, %2, %44.</li>
286 <li>Constants, which are described in a <a href="#constants">section about
287 constants</a>, below.</li>
290 <p>LLVM requires that values start with a '%' sign for two reasons: Compilers
291 don't need to worry about name clashes with reserved words, and the set of
292 reserved words may be expanded in the future without penalty. Additionally,
293 unnamed identifiers allow a compiler to quickly come up with a temporary
294 variable without having to avoid symbol table conflicts.</p>
296 <p>Reserved words in LLVM are very similar to reserved words in other
297 languages. There are keywords for different opcodes
298 ('<tt><a href="#i_add">add</a></tt>',
299 '<tt><a href="#i_bitcast">bitcast</a></tt>',
300 '<tt><a href="#i_ret">ret</a></tt>', etc...), for primitive type names ('<tt><a
301 href="#t_void">void</a></tt>', '<tt><a href="#t_primitive">i32</a></tt>', etc...),
302 and others. These reserved words cannot conflict with variable names, because
303 none of them start with a '%' character.</p>
305 <p>Here is an example of LLVM code to multiply the integer variable
306 '<tt>%X</tt>' by 8:</p>
311 %result = <a href="#i_mul">mul</a> i32 %X, 8
314 <p>After strength reduction:</p>
317 %result = <a href="#i_shl">shl</a> i32 %X, i8 3
320 <p>And the hard way:</p>
323 <a href="#i_add">add</a> i32 %X, %X <i>; yields {i32}:%0</i>
324 <a href="#i_add">add</a> i32 %0, %0 <i>; yields {i32}:%1</i>
325 %result = <a href="#i_add">add</a> i32 %1, %1
328 <p>This last way of multiplying <tt>%X</tt> by 8 illustrates several
329 important lexical features of LLVM:</p>
333 <li>Comments are delimited with a '<tt>;</tt>' and go until the end of
336 <li>Unnamed temporaries are created when the result of a computation is not
337 assigned to a named value.</li>
339 <li>Unnamed temporaries are numbered sequentially</li>
343 <p>...and it also shows a convention that we follow in this document. When
344 demonstrating instructions, we will follow an instruction with a comment that
345 defines the type and name of value produced. Comments are shown in italic
350 <!-- *********************************************************************** -->
351 <div class="doc_section"> <a name="highlevel">High Level Structure</a> </div>
352 <!-- *********************************************************************** -->
354 <!-- ======================================================================= -->
355 <div class="doc_subsection"> <a name="modulestructure">Module Structure</a>
358 <div class="doc_text">
360 <p>LLVM programs are composed of "Module"s, each of which is a
361 translation unit of the input programs. Each module consists of
362 functions, global variables, and symbol table entries. Modules may be
363 combined together with the LLVM linker, which merges function (and
364 global variable) definitions, resolves forward declarations, and merges
365 symbol table entries. Here is an example of the "hello world" module:</p>
367 <pre><i>; Declare the string constant as a global constant...</i>
368 <a href="#identifiers">%.LC0</a> = <a href="#linkage_internal">internal</a> <a
369 href="#globalvars">constant</a> <a href="#t_array">[13 x i8 ]</a> c"hello world\0A\00" <i>; [13 x i8 ]*</i>
371 <i>; External declaration of the puts function</i>
372 <a href="#functionstructure">declare</a> i32 %puts(i8 *) <i>; i32(i8 *)* </i>
374 <i>; Definition of main function</i>
375 define i32 %main() { <i>; i32()* </i>
376 <i>; Convert [13x i8 ]* to i8 *...</i>
378 href="#i_getelementptr">getelementptr</a> [13 x i8 ]* %.LC0, i64 0, i64 0 <i>; i8 *</i>
380 <i>; Call puts function to write out the string to stdout...</i>
382 href="#i_call">call</a> i32 %puts(i8 * %cast210) <i>; i32</i>
384 href="#i_ret">ret</a> i32 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>
399 <!-- ======================================================================= -->
400 <div class="doc_subsection">
401 <a name="linkage">Linkage Types</a>
404 <div class="doc_text">
407 All Global Variables and Functions have one of the following types of linkage:
412 <dt><tt><b><a name="linkage_internal">internal</a></b></tt> </dt>
414 <dd>Global values with internal linkage are only directly accessible by
415 objects in the current module. In particular, linking code into a module with
416 an internal global value may cause the internal to be renamed as necessary to
417 avoid collisions. Because the symbol is internal to the module, all
418 references can be updated. This corresponds to the notion of the
419 '<tt>static</tt>' keyword in C.
422 <dt><tt><b><a name="linkage_linkonce">linkonce</a></b></tt>: </dt>
424 <dd>Globals with "<tt>linkonce</tt>" linkage are merged with other globals of
425 the same name when linkage occurs. This is typically used to implement
426 inline functions, templates, or other code which must be generated in each
427 translation unit that uses it. Unreferenced <tt>linkonce</tt> globals are
428 allowed to be discarded.
431 <dt><tt><b><a name="linkage_weak">weak</a></b></tt>: </dt>
433 <dd>"<tt>weak</tt>" linkage is exactly the same as <tt>linkonce</tt> linkage,
434 except that unreferenced <tt>weak</tt> globals may not be discarded. This is
435 used for globals that may be emitted in multiple translation units, but that
436 are not guaranteed to be emitted into every translation unit that uses them.
437 One example of this are common globals in C, such as "<tt>int X;</tt>" at
441 <dt><tt><b><a name="linkage_appending">appending</a></b></tt>: </dt>
443 <dd>"<tt>appending</tt>" linkage may only be applied to global variables of
444 pointer to array type. When two global variables with appending linkage are
445 linked together, the two global arrays are appended together. This is the
446 LLVM, typesafe, equivalent of having the system linker append together
447 "sections" with identical names when .o files are linked.
450 <dt><tt><b><a name="linkage_externweak">extern_weak</a></b></tt>: </dt>
451 <dd>The semantics of this linkage follow the ELF model: the symbol is weak
452 until linked, if not linked, the symbol becomes null instead of being an
457 <dt><tt><b><a name="linkage_external">externally visible</a></b></tt>:</dt>
459 <dd>If none of the above identifiers are used, the global is externally
460 visible, meaning that it participates in linkage and can be used to resolve
461 external symbol references.
465 The next two types of linkage are targeted for Microsoft Windows platform
466 only. They are designed to support importing (exporting) symbols from (to)
471 <dt><tt><b><a name="linkage_dllimport">dllimport</a></b></tt>: </dt>
473 <dd>"<tt>dllimport</tt>" linkage causes the compiler to reference a function
474 or variable via a global pointer to a pointer that is set up by the DLL
475 exporting the symbol. On Microsoft Windows targets, the pointer name is
476 formed by combining <code>_imp__</code> and the function or variable name.
479 <dt><tt><b><a name="linkage_dllexport">dllexport</a></b></tt>: </dt>
481 <dd>"<tt>dllexport</tt>" linkage causes the compiler to provide a global
482 pointer to a pointer in a DLL, so that it can be referenced with the
483 <tt>dllimport</tt> attribute. On Microsoft Windows targets, the pointer
484 name is formed by combining <code>_imp__</code> and the function or variable
490 <p><a name="linkage_external"></a>For example, since the "<tt>.LC0</tt>"
491 variable is defined to be internal, if another module defined a "<tt>.LC0</tt>"
492 variable and was linked with this one, one of the two would be renamed,
493 preventing a collision. Since "<tt>main</tt>" and "<tt>puts</tt>" are
494 external (i.e., lacking any linkage declarations), they are accessible
495 outside of the current module.</p>
496 <p>It is illegal for a function <i>declaration</i>
497 to have any linkage type other than "externally visible", <tt>dllimport</tt>,
498 or <tt>extern_weak</tt>.</p>
502 <!-- ======================================================================= -->
503 <div class="doc_subsection">
504 <a name="callingconv">Calling Conventions</a>
507 <div class="doc_text">
509 <p>LLVM <a href="#functionstructure">functions</a>, <a href="#i_call">calls</a>
510 and <a href="#i_invoke">invokes</a> can all have an optional calling convention
511 specified for the call. The calling convention of any pair of dynamic
512 caller/callee must match, or the behavior of the program is undefined. The
513 following calling conventions are supported by LLVM, and more may be added in
517 <dt><b>"<tt>ccc</tt>" - The C calling convention</b>:</dt>
519 <dd>This calling convention (the default if no other calling convention is
520 specified) matches the target C calling conventions. This calling convention
521 supports varargs function calls and tolerates some mismatch in the declared
522 prototype and implemented declaration of the function (as does normal C).
525 <dt><b>"<tt>fastcc</tt>" - The fast calling convention</b>:</dt>
527 <dd>This calling convention attempts to make calls as fast as possible
528 (e.g. by passing things in registers). This calling convention allows the
529 target to use whatever tricks it wants to produce fast code for the target,
530 without having to conform to an externally specified ABI. Implementations of
531 this convention should allow arbitrary tail call optimization to be supported.
532 This calling convention does not support varargs and requires the prototype of
533 all callees to exactly match the prototype of the function definition.
536 <dt><b>"<tt>coldcc</tt>" - The cold calling convention</b>:</dt>
538 <dd>This calling convention attempts to make code in the caller as efficient
539 as possible under the assumption that the call is not commonly executed. As
540 such, these calls often preserve all registers so that the call does not break
541 any live ranges in the caller side. This calling convention does not support
542 varargs and requires the prototype of all callees to exactly match the
543 prototype of the function definition.
546 <dt><b>"<tt>cc <<em>n</em>></tt>" - Numbered convention</b>:</dt>
548 <dd>Any calling convention may be specified by number, allowing
549 target-specific calling conventions to be used. Target specific calling
550 conventions start at 64.
554 <p>More calling conventions can be added/defined on an as-needed basis, to
555 support pascal conventions or any other well-known target-independent
560 <!-- ======================================================================= -->
561 <div class="doc_subsection">
562 <a name="visibility">Visibility Styles</a>
565 <div class="doc_text">
568 All Global Variables and Functions have one of the following visibility styles:
572 <dt><b>"<tt>default</tt>" - Default style</b>:</dt>
574 <dd>On ELF, default visibility means that the declaration is visible to other
575 modules and, in shared libraries, means that the declared entity may be
576 overridden. On Darwin, default visibility means that the declaration is
577 visible to other modules. Default visibility corresponds to "external
578 linkage" in the language.
581 <dt><b>"<tt>hidden</tt>" - Hidden style</b>:</dt>
583 <dd>Two declarations of an object with hidden visibility refer to the same
584 object if they are in the same shared object. Usually, hidden visibility
585 indicates that the symbol will not be placed into the dynamic symbol table,
586 so no other module (executable or shared library) can reference it
594 <!-- ======================================================================= -->
595 <div class="doc_subsection">
596 <a name="globalvars">Global Variables</a>
599 <div class="doc_text">
601 <p>Global variables define regions of memory allocated at compilation time
602 instead of run-time. Global variables may optionally be initialized, may have
603 an explicit section to be placed in, and may
604 have an optional explicit alignment specified. A
605 variable may be defined as a global "constant," which indicates that the
606 contents of the variable will <b>never</b> be modified (enabling better
607 optimization, allowing the global data to be placed in the read-only section of
608 an executable, etc). Note that variables that need runtime initialization
609 cannot be marked "constant" as there is a store to the variable.</p>
612 LLVM explicitly allows <em>declarations</em> of global variables to be marked
613 constant, even if the final definition of the global is not. This capability
614 can be used to enable slightly better optimization of the program, but requires
615 the language definition to guarantee that optimizations based on the
616 'constantness' are valid for the translation units that do not include the
620 <p>As SSA values, global variables define pointer values that are in
621 scope (i.e. they dominate) all basic blocks in the program. Global
622 variables always define a pointer to their "content" type because they
623 describe a region of memory, and all memory objects in LLVM are
624 accessed through pointers.</p>
626 <p>LLVM allows an explicit section to be specified for globals. If the target
627 supports it, it will emit globals to the section specified.</p>
629 <p>An explicit alignment may be specified for a global. If not present, or if
630 the alignment is set to zero, the alignment of the global is set by the target
631 to whatever it feels convenient. If an explicit alignment is specified, the
632 global is forced to have at least that much alignment. All alignments must be
635 <p>For example, the following defines a global with an initializer, section,
639 %G = constant float 1.0, section "foo", align 4
645 <!-- ======================================================================= -->
646 <div class="doc_subsection">
647 <a name="functionstructure">Functions</a>
650 <div class="doc_text">
652 <p>LLVM function definitions consist of the "<tt>define</tt>" keyord,
653 an optional <a href="#linkage">linkage type</a>, an optional
654 <a href="#visibility">visibility style</a>, an optional
655 <a href="#callingconv">calling convention</a>, a return type, an optional
656 <a href="#paramattrs">parameter attribute</a> for the return type, a function
657 name, a (possibly empty) argument list (each with optional
658 <a href="#paramattrs">parameter attributes</a>), an optional section, an
659 optional alignment, an opening curly brace, a list of basic blocks, and a
662 LLVM function declarations consist of the "<tt>declare</tt>" keyword, an
663 optional <a href="#linkage">linkage type</a>, an optional
664 <a href="#visibility">visibility style</a>, an optional
665 <a href="#callingconv">calling convention</a>, a return type, an optional
666 <a href="#paramattrs">parameter attribute</a> for the return type, a function
667 name, a possibly empty list of arguments, and an optional alignment.</p>
669 <p>A function definition contains a list of basic blocks, forming the CFG for
670 the function. Each basic block may optionally start with a label (giving the
671 basic block a symbol table entry), contains a list of instructions, and ends
672 with a <a href="#terminators">terminator</a> instruction (such as a branch or
673 function return).</p>
675 <p>The first basic block in a program is special in two ways: it is immediately
676 executed on entrance to the function, and it is not allowed to have predecessor
677 basic blocks (i.e. there can not be any branches to the entry block of a
678 function). Because the block can have no predecessors, it also cannot have any
679 <a href="#i_phi">PHI nodes</a>.</p>
681 <p>LLVM functions are identified by their name and type signature. Hence, two
682 functions with the same name but different parameter lists or return values are
683 considered different functions, and LLVM will resolve references to each
686 <p>LLVM allows an explicit section to be specified for functions. If the target
687 supports it, it will emit functions to the section specified.</p>
689 <p>An explicit alignment may be specified for a function. If not present, or if
690 the alignment is set to zero, the alignment of the function is set by the target
691 to whatever it feels convenient. If an explicit alignment is specified, the
692 function is forced to have at least that much alignment. All alignments must be
697 <!-- ======================================================================= -->
698 <div class="doc_subsection"><a name="paramattrs">Parameter Attributes</a></div>
699 <div class="doc_text">
700 <p>The return type and each parameter of a function type may have a set of
701 <i>parameter attributes</i> associated with them. Parameter attributes are
702 used to communicate additional information about the result or parameters of
703 a function. Parameter attributes are considered to be part of the function
704 type so two functions types that differ only by the parameter attributes
705 are different function types.</p>
707 <p>Parameter attributes are simple keywords that follow the type specified. If
708 multiple parameter attributes are needed, they are space separated. For
710 %someFunc = i16 (i8 sext %someParam) zext
711 %someFunc = i16 (i8 zext %someParam) zext</pre>
712 <p>Note that the two function types above are unique because the parameter has
713 a different attribute (sext in the first one, zext in the second). Also note
714 that the attribute for the function result (zext) comes immediately after the
717 <p>Currently, only the following parameter attributes are defined:</p>
719 <dt><tt>zext</tt></dt>
720 <dd>This indicates that the parameter should be zero extended just before
721 a call to this function.</dd>
722 <dt><tt>sext</tt></dt>
723 <dd>This indicates that the parameter should be sign extended just before
724 a call to this function.</dd>
725 <dt><tt>inreg</tt></dt>
726 <dd>This indicates that the parameter should be placed in register (if
727 possible) during assembling function call. Support for this attribute is
729 <dt><tt>sret</tt></dt>
730 <dd>This indicates that the parameter specifies the address of a structure
731 that is the return value of the function in the source program.</dd>
732 <dt><tt>noreturn</tt></dt>
733 <dd>This function attribute indicates that the function never returns. This
734 indicates to LLVM that every call to this function should be treated as if
735 an <tt>unreachable</tt> instruction immediately followed the call.</dd>
736 <dt><tt>nounwind</tt></dt>
737 <dd>This function attribute indicates that the function type does not use
738 the unwind instruction and does not allow stack unwinding to propagate
744 <!-- ======================================================================= -->
745 <div class="doc_subsection">
746 <a name="moduleasm">Module-Level Inline Assembly</a>
749 <div class="doc_text">
751 Modules may contain "module-level inline asm" blocks, which corresponds to the
752 GCC "file scope inline asm" blocks. These blocks are internally concatenated by
753 LLVM and treated as a single unit, but may be separated in the .ll file if
754 desired. The syntax is very simple:
757 <div class="doc_code"><pre>
758 module asm "inline asm code goes here"
759 module asm "more can go here"
762 <p>The strings can contain any character by escaping non-printable characters.
763 The escape sequence used is simply "\xx" where "xx" is the two digit hex code
768 The inline asm code is simply printed to the machine code .s file when
769 assembly code is generated.
773 <!-- ======================================================================= -->
774 <div class="doc_subsection">
775 <a name="datalayout">Data Layout</a>
778 <div class="doc_text">
779 <p>A module may specify a target specific data layout string that specifies how
780 data is to be laid out in memory. The syntax for the data layout is simply:<br/>
781 <pre> target datalayout = "<i>layout specification</i>"
783 The <i>layout specification</i> consists of a list of specifications separated
784 by the minus sign character ('-'). Each specification starts with a letter
785 and may include other information after the letter to define some aspect of the
786 data layout. The specifications accepted are as follows: </p>
789 <dd>Specifies that the target lays out data in big-endian form. That is, the
790 bits with the most significance have the lowest address location.</dd>
792 <dd>Specifies that hte target lays out data in little-endian form. That is,
793 the bits with the least significance have the lowest address location.</dd>
794 <dt><tt>p:<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
795 <dd>This specifies the <i>size</i> of a pointer and its <i>abi</i> and
796 <i>preferred</i> alignments. All sizes are in bits. Specifying the <i>pref</i>
797 alignment is optional. If omitted, the preceding <tt>:</tt> should be omitted
799 <dt><tt>i<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
800 <dd>This specifies the alignment for an integer type of a given bit
801 <i>size</i>. The value of <i>size</i> must be in the range [1,2^23).</dd>
802 <dt><tt>v<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
803 <dd>This specifies the alignment for a vector type of a given bit
805 <dt><tt>f<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
806 <dd>This specifies the alignment for a floating point type of a given bit
807 <i>size</i>. The value of <i>size</i> must be either 32 (float) or 64
809 <dt><tt>a<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
810 <dd>This specifies the alignment for an aggregate type of a given bit
813 <p>When constructing the data layout for a given target, LLVM starts with a
814 default set of specifications which are then (possibly) overriden by the
815 specifications in the <tt>datalayout</tt> keyword. The default specifications
816 are given in this list:</p>
818 <li><tt>E</tt> - big endian</li>
819 <li><tt>p:32:64:64</tt> - 32-bit pointers with 64-bit alignment</li>
820 <li><tt>i1:8:8</tt> - i1 is 8-bit (byte) aligned</li>
821 <li><tt>i8:8:8</tt> - i8 is 8-bit (byte) aligned</li>
822 <li><tt>i16:16:16</tt> - i16 is 16-bit aligned</li>
823 <li><tt>i32:32:32</tt> - i32 is 32-bit aligned</li>
824 <li><tt>i64:32:64</tt> - i64 has abi alignment of 32-bits but preferred
825 alignment of 64-bits</li>
826 <li><tt>f32:32:32</tt> - float is 32-bit aligned</li>
827 <li><tt>f64:64:64</tt> - double is 64-bit aligned</li>
828 <li><tt>v64:64:64</tt> - 64-bit vector is 64-bit aligned</li>
829 <li><tt>v128:128:128</tt> - 128-bit vector is 128-bit aligned</li>
830 <li><tt>a0:0:1</tt> - aggregates are 8-bit aligned</li>
832 <p>When llvm is determining the alignment for a given type, it uses the
835 <li>If the type sought is an exact match for one of the specifications, that
836 specification is used.</li>
837 <li>If no match is found, and the type sought is an integer type, then the
838 smallest integer type that is larger than the bitwidth of the sought type is
839 used. If none of the specifications are larger than the bitwidth then the the
840 largest integer type is used. For example, given the default specifications
841 above, the i7 type will use the alignment of i8 (next largest) while both
842 i65 and i256 will use the alignment of i64 (largest specified).</li>
843 <li>If no match is found, and the type sought is a vector type, then the
844 largest vector type that is smaller than the sought vector type will be used
845 as a fall back. This happens because <128 x double> can be implemented in
846 terms of 64 <2 x double>, for example.</li>
850 <!-- *********************************************************************** -->
851 <div class="doc_section"> <a name="typesystem">Type System</a> </div>
852 <!-- *********************************************************************** -->
854 <div class="doc_text">
856 <p>The LLVM type system is one of the most important features of the
857 intermediate representation. Being typed enables a number of
858 optimizations to be performed on the IR directly, without having to do
859 extra analyses on the side before the transformation. A strong type
860 system makes it easier to read the generated code and enables novel
861 analyses and transformations that are not feasible to perform on normal
862 three address code representations.</p>
866 <!-- ======================================================================= -->
867 <div class="doc_subsection"> <a name="t_primitive">Primitive Types</a> </div>
868 <div class="doc_text">
869 <p>The primitive types are the fundamental building blocks of the LLVM
870 system. The current set of primitive types is as follows:</p>
872 <table class="layout">
877 <tr><th>Type</th><th>Description</th></tr>
878 <tr><td><tt><a name="t_void">void</a></tt></td><td>No value</td></tr>
879 <tr><td><tt>i8</tt></td><td>8-bit value</td></tr>
880 <tr><td><tt>i32</tt></td><td>32-bit value</td></tr>
881 <tr><td><tt>float</tt></td><td>32-bit floating point value</td></tr>
882 <tr><td><tt>label</tt></td><td>Branch destination</td></tr>
889 <tr><th>Type</th><th>Description</th></tr>
890 <tr><td><tt>i1</tt></td><td>True or False value</td></tr>
891 <tr><td><tt>i16</tt></td><td>16-bit value</td></tr>
892 <tr><td><tt>i64</tt></td><td>64-bit value</td></tr>
893 <tr><td><tt>double</tt></td><td>64-bit floating point value</td></tr>
901 <!-- _______________________________________________________________________ -->
902 <div class="doc_subsubsection"> <a name="t_classifications">Type
903 Classifications</a> </div>
904 <div class="doc_text">
905 <p>These different primitive types fall into a few useful
908 <table border="1" cellspacing="0" cellpadding="4">
910 <tr><th>Classification</th><th>Types</th></tr>
912 <td><a name="t_integer">integer</a></td>
913 <td><tt>i1, i8, i16, i32, i64</tt></td>
916 <td><a name="t_floating">floating point</a></td>
917 <td><tt>float, double</tt></td>
920 <td><a name="t_firstclass">first class</a></td>
921 <td><tt>i1, i8, i16, i32, i64, float, double, <br/>
922 <a href="#t_pointer">pointer</a>,<a href="#t_vector">vector</a></tt>
928 <p>The <a href="#t_firstclass">first class</a> types are perhaps the
929 most important. Values of these types are the only ones which can be
930 produced by instructions, passed as arguments, or used as operands to
931 instructions. This means that all structures and arrays must be
932 manipulated either by pointer or by component.</p>
935 <!-- ======================================================================= -->
936 <div class="doc_subsection"> <a name="t_derived">Derived Types</a> </div>
938 <div class="doc_text">
940 <p>The real power in LLVM comes from the derived types in the system.
941 This is what allows a programmer to represent arrays, functions,
942 pointers, and other useful types. Note that these derived types may be
943 recursive: For example, it is possible to have a two dimensional array.</p>
947 <!-- _______________________________________________________________________ -->
948 <div class="doc_subsubsection"> <a name="t_array">Array Type</a> </div>
950 <div class="doc_text">
954 <p>The array type is a very simple derived type that arranges elements
955 sequentially in memory. The array type requires a size (number of
956 elements) and an underlying data type.</p>
961 [<# elements> x <elementtype>]
964 <p>The number of elements is a constant integer value; elementtype may
965 be any type with a size.</p>
968 <table class="layout">
971 <tt>[40 x i32 ]</tt><br/>
972 <tt>[41 x i32 ]</tt><br/>
973 <tt>[40 x i8]</tt><br/>
976 Array of 40 32-bit integer values.<br/>
977 Array of 41 32-bit integer values.<br/>
978 Array of 40 8-bit integer values.<br/>
982 <p>Here are some examples of multidimensional arrays:</p>
983 <table class="layout">
986 <tt>[3 x [4 x i32]]</tt><br/>
987 <tt>[12 x [10 x float]]</tt><br/>
988 <tt>[2 x [3 x [4 x i16]]]</tt><br/>
991 3x4 array of 32-bit integer values.<br/>
992 12x10 array of single precision floating point values.<br/>
993 2x3x4 array of 16-bit integer values.<br/>
998 <p>Note that 'variable sized arrays' can be implemented in LLVM with a zero
999 length array. Normally, accesses past the end of an array are undefined in
1000 LLVM (e.g. it is illegal to access the 5th element of a 3 element array).
1001 As a special case, however, zero length arrays are recognized to be variable
1002 length. This allows implementation of 'pascal style arrays' with the LLVM
1003 type "{ i32, [0 x float]}", for example.</p>
1007 <!-- _______________________________________________________________________ -->
1008 <div class="doc_subsubsection"> <a name="t_function">Function Type</a> </div>
1009 <div class="doc_text">
1011 <p>The function type can be thought of as a function signature. It
1012 consists of a return type and a list of formal parameter types.
1013 Function types are usually used to build virtual function tables
1014 (which are structures of pointers to functions), for indirect function
1015 calls, and when defining a function.</p>
1017 The return type of a function type cannot be an aggregate type.
1020 <pre> <returntype> (<parameter list>)<br></pre>
1021 <p>...where '<tt><parameter list></tt>' is a comma-separated list of type
1022 specifiers. Optionally, the parameter list may include a type <tt>...</tt>,
1023 which indicates that the function takes a variable number of arguments.
1024 Variable argument functions can access their arguments with the <a
1025 href="#int_varargs">variable argument handling intrinsic</a> functions.</p>
1027 <table class="layout">
1029 <td class="left"><tt>i32 (i32)</tt></td>
1030 <td class="left">function taking an <tt>i32</tt>, returning an <tt>i32</tt>
1032 </tr><tr class="layout">
1033 <td class="left"><tt>float (i16 sext, i32 *) *
1035 <td class="left"><a href="#t_pointer">Pointer</a> to a function that takes
1036 an <tt>i16</tt> that should be sign extended and a
1037 <a href="#t_pointer">pointer</a> to <tt>i32</tt>, returning
1040 </tr><tr class="layout">
1041 <td class="left"><tt>i32 (i8*, ...)</tt></td>
1042 <td class="left">A vararg function that takes at least one
1043 <a href="#t_pointer">pointer</a> to <tt>i8 </tt> (char in C),
1044 which returns an integer. This is the signature for <tt>printf</tt> in
1051 <!-- _______________________________________________________________________ -->
1052 <div class="doc_subsubsection"> <a name="t_struct">Structure Type</a> </div>
1053 <div class="doc_text">
1055 <p>The structure type is used to represent a collection of data members
1056 together in memory. The packing of the field types is defined to match
1057 the ABI of the underlying processor. The elements of a structure may
1058 be any type that has a size.</p>
1059 <p>Structures are accessed using '<tt><a href="#i_load">load</a></tt>
1060 and '<tt><a href="#i_store">store</a></tt>' by getting a pointer to a
1061 field with the '<tt><a href="#i_getelementptr">getelementptr</a></tt>'
1064 <pre> { <type list> }<br></pre>
1066 <table class="layout">
1069 <tt>{ i32, i32, i32 }</tt><br/>
1070 <tt>{ float, i32 (i32) * }</tt><br/>
1073 a triple of three <tt>i32</tt> values<br/>
1074 A pair, where the first element is a <tt>float</tt> and the second element
1075 is a <a href="#t_pointer">pointer</a> to a <a href="#t_function">function</a>
1076 that takes an <tt>i32</tt>, returning an <tt>i32</tt>.<br/>
1082 <!-- _______________________________________________________________________ -->
1083 <div class="doc_subsubsection"> <a name="t_pstruct">Packed Structure Type</a>
1085 <div class="doc_text">
1087 <p>The packed structure type is used to represent a collection of data members
1088 together in memory. There is no padding between fields. Further, the alignment
1089 of a packed structure is 1 byte. The elements of a packed structure may
1090 be any type that has a size.</p>
1091 <p>Structures are accessed using '<tt><a href="#i_load">load</a></tt>
1092 and '<tt><a href="#i_store">store</a></tt>' by getting a pointer to a
1093 field with the '<tt><a href="#i_getelementptr">getelementptr</a></tt>'
1096 <pre> < { <type list> } > <br></pre>
1098 <table class="layout">
1101 <tt> < { i32, i32, i32 } > </tt><br/>
1102 <tt> < { float, i32 (i32) * } > </tt><br/>
1105 a triple of three <tt>i32</tt> values<br/>
1106 A pair, where the first element is a <tt>float</tt> and the second element
1107 is a <a href="#t_pointer">pointer</a> to a <a href="#t_function">function</a>
1108 that takes an <tt>i32</tt>, returning an <tt>i32</tt>.<br/>
1114 <!-- _______________________________________________________________________ -->
1115 <div class="doc_subsubsection"> <a name="t_pointer">Pointer Type</a> </div>
1116 <div class="doc_text">
1118 <p>As in many languages, the pointer type represents a pointer or
1119 reference to another object, which must live in memory.</p>
1121 <pre> <type> *<br></pre>
1123 <table class="layout">
1126 <tt>[4x i32]*</tt><br/>
1127 <tt>i32 (i32 *) *</tt><br/>
1130 A <a href="#t_pointer">pointer</a> to <a href="#t_array">array</a> of
1131 four <tt>i32</tt> values<br/>
1132 A <a href="#t_pointer">pointer</a> to a <a
1133 href="#t_function">function</a> that takes an <tt>i32*</tt>, returning an
1140 <!-- _______________________________________________________________________ -->
1141 <div class="doc_subsubsection"> <a name="t_vector">Vector Type</a> </div>
1142 <div class="doc_text">
1146 <p>A vector type is a simple derived type that represents a vector
1147 of elements. Vector types are used when multiple primitive data
1148 are operated in parallel using a single instruction (SIMD).
1149 A vector type requires a size (number of
1150 elements) and an underlying primitive data type. Vectors must have a power
1151 of two length (1, 2, 4, 8, 16 ...). Vector types are
1152 considered <a href="#t_firstclass">first class</a>.</p>
1157 < <# elements> x <elementtype> >
1160 <p>The number of elements is a constant integer value; elementtype may
1161 be any integer or floating point type.</p>
1165 <table class="layout">
1168 <tt><4 x i32></tt><br/>
1169 <tt><8 x float></tt><br/>
1170 <tt><2 x i64></tt><br/>
1173 Vector of 4 32-bit integer values.<br/>
1174 Vector of 8 floating-point values.<br/>
1175 Vector of 2 64-bit integer values.<br/>
1181 <!-- _______________________________________________________________________ -->
1182 <div class="doc_subsubsection"> <a name="t_opaque">Opaque Type</a> </div>
1183 <div class="doc_text">
1187 <p>Opaque types are used to represent unknown types in the system. This
1188 corresponds (for example) to the C notion of a foward declared structure type.
1189 In LLVM, opaque types can eventually be resolved to any type (not just a
1190 structure type).</p>
1200 <table class="layout">
1206 An opaque type.<br/>
1213 <!-- *********************************************************************** -->
1214 <div class="doc_section"> <a name="constants">Constants</a> </div>
1215 <!-- *********************************************************************** -->
1217 <div class="doc_text">
1219 <p>LLVM has several different basic types of constants. This section describes
1220 them all and their syntax.</p>
1224 <!-- ======================================================================= -->
1225 <div class="doc_subsection"><a name="simpleconstants">Simple Constants</a></div>
1227 <div class="doc_text">
1230 <dt><b>Boolean constants</b></dt>
1232 <dd>The two strings '<tt>true</tt>' and '<tt>false</tt>' are both valid
1233 constants of the <tt><a href="#t_primitive">i1</a></tt> type.
1236 <dt><b>Integer constants</b></dt>
1238 <dd>Standard integers (such as '4') are constants of the <a
1239 href="#t_integer">integer</a> type. Negative numbers may be used with
1243 <dt><b>Floating point constants</b></dt>
1245 <dd>Floating point constants use standard decimal notation (e.g. 123.421),
1246 exponential notation (e.g. 1.23421e+2), or a more precise hexadecimal
1247 notation (see below). Floating point constants must have a <a
1248 href="#t_floating">floating point</a> type. </dd>
1250 <dt><b>Null pointer constants</b></dt>
1252 <dd>The identifier '<tt>null</tt>' is recognized as a null pointer constant
1253 and must be of <a href="#t_pointer">pointer type</a>.</dd>
1257 <p>The one non-intuitive notation for constants is the optional hexadecimal form
1258 of floating point constants. For example, the form '<tt>double
1259 0x432ff973cafa8000</tt>' is equivalent to (but harder to read than) '<tt>double
1260 4.5e+15</tt>'. The only time hexadecimal floating point constants are required
1261 (and the only time that they are generated by the disassembler) is when a
1262 floating point constant must be emitted but it cannot be represented as a
1263 decimal floating point number. For example, NaN's, infinities, and other
1264 special values are represented in their IEEE hexadecimal format so that
1265 assembly and disassembly do not cause any bits to change in the constants.</p>
1269 <!-- ======================================================================= -->
1270 <div class="doc_subsection"><a name="aggregateconstants">Aggregate Constants</a>
1273 <div class="doc_text">
1274 <p>Aggregate constants arise from aggregation of simple constants
1275 and smaller aggregate constants.</p>
1278 <dt><b>Structure constants</b></dt>
1280 <dd>Structure constants are represented with notation similar to structure
1281 type definitions (a comma separated list of elements, surrounded by braces
1282 (<tt>{}</tt>)). For example: "<tt>{ i32 4, float 17.0, i32* %G }</tt>",
1283 where "<tt>%G</tt>" is declared as "<tt>%G = external global i32</tt>". Structure constants
1284 must have <a href="#t_struct">structure type</a>, and the number and
1285 types of elements must match those specified by the type.
1288 <dt><b>Array constants</b></dt>
1290 <dd>Array constants are represented with notation similar to array type
1291 definitions (a comma separated list of elements, surrounded by square brackets
1292 (<tt>[]</tt>)). For example: "<tt>[ i32 42, i32 11, i32 74 ]</tt>". Array
1293 constants must have <a href="#t_array">array type</a>, and the number and
1294 types of elements must match those specified by the type.
1297 <dt><b>Vector constants</b></dt>
1299 <dd>Vector constants are represented with notation similar to vector type
1300 definitions (a comma separated list of elements, surrounded by
1301 less-than/greater-than's (<tt><></tt>)). For example: "<tt>< i32 42,
1302 i32 11, i32 74, i32 100 ></tt>". VEctor constants must have <a
1303 href="#t_vector">vector type</a>, and the number and types of elements must
1304 match those specified by the type.
1307 <dt><b>Zero initialization</b></dt>
1309 <dd>The string '<tt>zeroinitializer</tt>' can be used to zero initialize a
1310 value to zero of <em>any</em> type, including scalar and aggregate types.
1311 This is often used to avoid having to print large zero initializers (e.g. for
1312 large arrays) and is always exactly equivalent to using explicit zero
1319 <!-- ======================================================================= -->
1320 <div class="doc_subsection">
1321 <a name="globalconstants">Global Variable and Function Addresses</a>
1324 <div class="doc_text">
1326 <p>The addresses of <a href="#globalvars">global variables</a> and <a
1327 href="#functionstructure">functions</a> are always implicitly valid (link-time)
1328 constants. These constants are explicitly referenced when the <a
1329 href="#identifiers">identifier for the global</a> is used and always have <a
1330 href="#t_pointer">pointer</a> type. For example, the following is a legal LLVM
1336 %Z = global [2 x i32*] [ i32* %X, i32* %Y ]
1341 <!-- ======================================================================= -->
1342 <div class="doc_subsection"><a name="undefvalues">Undefined Values</a></div>
1343 <div class="doc_text">
1344 <p>The string '<tt>undef</tt>' is recognized as a type-less constant that has
1345 no specific value. Undefined values may be of any type and be used anywhere
1346 a constant is permitted.</p>
1348 <p>Undefined values indicate to the compiler that the program is well defined
1349 no matter what value is used, giving the compiler more freedom to optimize.
1353 <!-- ======================================================================= -->
1354 <div class="doc_subsection"><a name="constantexprs">Constant Expressions</a>
1357 <div class="doc_text">
1359 <p>Constant expressions are used to allow expressions involving other constants
1360 to be used as constants. Constant expressions may be of any <a
1361 href="#t_firstclass">first class</a> type and may involve any LLVM operation
1362 that does not have side effects (e.g. load and call are not supported). The
1363 following is the syntax for constant expressions:</p>
1366 <dt><b><tt>trunc ( CST to TYPE )</tt></b></dt>
1367 <dd>Truncate a constant to another type. The bit size of CST must be larger
1368 than the bit size of TYPE. Both types must be integers.</dd>
1370 <dt><b><tt>zext ( CST to TYPE )</tt></b></dt>
1371 <dd>Zero extend a constant to another type. The bit size of CST must be
1372 smaller or equal to the bit size of TYPE. Both types must be integers.</dd>
1374 <dt><b><tt>sext ( CST to TYPE )</tt></b></dt>
1375 <dd>Sign extend a constant to another type. The bit size of CST must be
1376 smaller or equal to the bit size of TYPE. Both types must be integers.</dd>
1378 <dt><b><tt>fptrunc ( CST to TYPE )</tt></b></dt>
1379 <dd>Truncate a floating point constant to another floating point type. The
1380 size of CST must be larger than the size of TYPE. Both types must be
1381 floating point.</dd>
1383 <dt><b><tt>fpext ( CST to TYPE )</tt></b></dt>
1384 <dd>Floating point extend a constant to another type. The size of CST must be
1385 smaller or equal to the size of TYPE. Both types must be floating point.</dd>
1387 <dt><b><tt>fp2uint ( CST to TYPE )</tt></b></dt>
1388 <dd>Convert a floating point constant to the corresponding unsigned integer
1389 constant. TYPE must be an integer type. CST must be floating point. If the
1390 value won't fit in the integer type, the results are undefined.</dd>
1392 <dt><b><tt>fptosi ( CST to TYPE )</tt></b></dt>
1393 <dd>Convert a floating point constant to the corresponding signed integer
1394 constant. TYPE must be an integer type. CST must be floating point. If the
1395 value won't fit in the integer type, the results are undefined.</dd>
1397 <dt><b><tt>uitofp ( CST to TYPE )</tt></b></dt>
1398 <dd>Convert an unsigned integer constant to the corresponding floating point
1399 constant. TYPE must be floating point. CST must be of integer type. If the
1400 value won't fit in the floating point type, the results are undefined.</dd>
1402 <dt><b><tt>sitofp ( CST to TYPE )</tt></b></dt>
1403 <dd>Convert a signed integer constant to the corresponding floating point
1404 constant. TYPE must be floating point. CST must be of integer type. If the
1405 value won't fit in the floating point type, the results are undefined.</dd>
1407 <dt><b><tt>ptrtoint ( CST to TYPE )</tt></b></dt>
1408 <dd>Convert a pointer typed constant to the corresponding integer constant
1409 TYPE must be an integer type. CST must be of pointer type. The CST value is
1410 zero extended, truncated, or unchanged to make it fit in TYPE.</dd>
1412 <dt><b><tt>inttoptr ( CST to TYPE )</tt></b></dt>
1413 <dd>Convert a integer constant to a pointer constant. TYPE must be a
1414 pointer type. CST must be of integer type. The CST value is zero extended,
1415 truncated, or unchanged to make it fit in a pointer size. This one is
1416 <i>really</i> dangerous!</dd>
1418 <dt><b><tt>bitcast ( CST to TYPE )</tt></b></dt>
1419 <dd>Convert a constant, CST, to another TYPE. The size of CST and TYPE must be
1420 identical (same number of bits). The conversion is done as if the CST value
1421 was stored to memory and read back as TYPE. In other words, no bits change
1422 with this operator, just the type. This can be used for conversion of
1423 vector types to any other type, as long as they have the same bit width. For
1424 pointers it is only valid to cast to another pointer type.
1427 <dt><b><tt>getelementptr ( CSTPTR, IDX0, IDX1, ... )</tt></b></dt>
1429 <dd>Perform the <a href="#i_getelementptr">getelementptr operation</a> on
1430 constants. As with the <a href="#i_getelementptr">getelementptr</a>
1431 instruction, the index list may have zero or more indexes, which are required
1432 to make sense for the type of "CSTPTR".</dd>
1434 <dt><b><tt>select ( COND, VAL1, VAL2 )</tt></b></dt>
1436 <dd>Perform the <a href="#i_select">select operation</a> on
1439 <dt><b><tt>icmp COND ( VAL1, VAL2 )</tt></b></dt>
1440 <dd>Performs the <a href="#i_icmp">icmp operation</a> on constants.</dd>
1442 <dt><b><tt>fcmp COND ( VAL1, VAL2 )</tt></b></dt>
1443 <dd>Performs the <a href="#i_fcmp">fcmp operation</a> on constants.</dd>
1445 <dt><b><tt>extractelement ( VAL, IDX )</tt></b></dt>
1447 <dd>Perform the <a href="#i_extractelement">extractelement
1448 operation</a> on constants.
1450 <dt><b><tt>insertelement ( VAL, ELT, IDX )</tt></b></dt>
1452 <dd>Perform the <a href="#i_insertelement">insertelement
1453 operation</a> on constants.</dd>
1456 <dt><b><tt>shufflevector ( VEC1, VEC2, IDXMASK )</tt></b></dt>
1458 <dd>Perform the <a href="#i_shufflevector">shufflevector
1459 operation</a> on constants.</dd>
1461 <dt><b><tt>OPCODE ( LHS, RHS )</tt></b></dt>
1463 <dd>Perform the specified operation of the LHS and RHS constants. OPCODE may
1464 be any of the <a href="#binaryops">binary</a> or <a href="#bitwiseops">bitwise
1465 binary</a> operations. The constraints on operands are the same as those for
1466 the corresponding instruction (e.g. no bitwise operations on floating point
1467 values are allowed).</dd>
1471 <!-- *********************************************************************** -->
1472 <div class="doc_section"> <a name="othervalues">Other Values</a> </div>
1473 <!-- *********************************************************************** -->
1475 <!-- ======================================================================= -->
1476 <div class="doc_subsection">
1477 <a name="inlineasm">Inline Assembler Expressions</a>
1480 <div class="doc_text">
1483 LLVM supports inline assembler expressions (as opposed to <a href="#moduleasm">
1484 Module-Level Inline Assembly</a>) through the use of a special value. This
1485 value represents the inline assembler as a string (containing the instructions
1486 to emit), a list of operand constraints (stored as a string), and a flag that
1487 indicates whether or not the inline asm expression has side effects. An example
1488 inline assembler expression is:
1492 i32 (i32) asm "bswap $0", "=r,r"
1496 Inline assembler expressions may <b>only</b> be used as the callee operand of
1497 a <a href="#i_call"><tt>call</tt> instruction</a>. Thus, typically we have:
1501 %X = call i32 asm "<a href="#i_bswap">bswap</a> $0", "=r,r"(i32 %Y)
1505 Inline asms with side effects not visible in the constraint list must be marked
1506 as having side effects. This is done through the use of the
1507 '<tt>sideeffect</tt>' keyword, like so:
1511 call void asm sideeffect "eieio", ""()
1514 <p>TODO: The format of the asm and constraints string still need to be
1515 documented here. Constraints on what can be done (e.g. duplication, moving, etc
1516 need to be documented).
1521 <!-- *********************************************************************** -->
1522 <div class="doc_section"> <a name="instref">Instruction Reference</a> </div>
1523 <!-- *********************************************************************** -->
1525 <div class="doc_text">
1527 <p>The LLVM instruction set consists of several different
1528 classifications of instructions: <a href="#terminators">terminator
1529 instructions</a>, <a href="#binaryops">binary instructions</a>,
1530 <a href="#bitwiseops">bitwise binary instructions</a>, <a
1531 href="#memoryops">memory instructions</a>, and <a href="#otherops">other
1532 instructions</a>.</p>
1536 <!-- ======================================================================= -->
1537 <div class="doc_subsection"> <a name="terminators">Terminator
1538 Instructions</a> </div>
1540 <div class="doc_text">
1542 <p>As mentioned <a href="#functionstructure">previously</a>, every
1543 basic block in a program ends with a "Terminator" instruction, which
1544 indicates which block should be executed after the current block is
1545 finished. These terminator instructions typically yield a '<tt>void</tt>'
1546 value: they produce control flow, not values (the one exception being
1547 the '<a href="#i_invoke"><tt>invoke</tt></a>' instruction).</p>
1548 <p>There are six different terminator instructions: the '<a
1549 href="#i_ret"><tt>ret</tt></a>' instruction, the '<a href="#i_br"><tt>br</tt></a>'
1550 instruction, the '<a href="#i_switch"><tt>switch</tt></a>' instruction,
1551 the '<a href="#i_invoke"><tt>invoke</tt></a>' instruction, the '<a
1552 href="#i_unwind"><tt>unwind</tt></a>' instruction, and the '<a
1553 href="#i_unreachable"><tt>unreachable</tt></a>' instruction.</p>
1557 <!-- _______________________________________________________________________ -->
1558 <div class="doc_subsubsection"> <a name="i_ret">'<tt>ret</tt>'
1559 Instruction</a> </div>
1560 <div class="doc_text">
1562 <pre> ret <type> <value> <i>; Return a value from a non-void function</i>
1563 ret void <i>; Return from void function</i>
1566 <p>The '<tt>ret</tt>' instruction is used to return control flow (and a
1567 value) from a function back to the caller.</p>
1568 <p>There are two forms of the '<tt>ret</tt>' instruction: one that
1569 returns a value and then causes control flow, and one that just causes
1570 control flow to occur.</p>
1572 <p>The '<tt>ret</tt>' instruction may return any '<a
1573 href="#t_firstclass">first class</a>' type. Notice that a function is
1574 not <a href="#wellformed">well formed</a> if there exists a '<tt>ret</tt>'
1575 instruction inside of the function that returns a value that does not
1576 match the return type of the function.</p>
1578 <p>When the '<tt>ret</tt>' instruction is executed, control flow
1579 returns back to the calling function's context. If the caller is a "<a
1580 href="#i_call"><tt>call</tt></a>" instruction, execution continues at
1581 the instruction after the call. If the caller was an "<a
1582 href="#i_invoke"><tt>invoke</tt></a>" instruction, execution continues
1583 at the beginning of the "normal" destination block. If the instruction
1584 returns a value, that value shall set the call or invoke instruction's
1587 <pre> ret i32 5 <i>; Return an integer value of 5</i>
1588 ret void <i>; Return from a void function</i>
1591 <!-- _______________________________________________________________________ -->
1592 <div class="doc_subsubsection"> <a name="i_br">'<tt>br</tt>' Instruction</a> </div>
1593 <div class="doc_text">
1595 <pre> br i1 <cond>, label <iftrue>, label <iffalse><br> br label <dest> <i>; Unconditional branch</i>
1598 <p>The '<tt>br</tt>' instruction is used to cause control flow to
1599 transfer to a different basic block in the current function. There are
1600 two forms of this instruction, corresponding to a conditional branch
1601 and an unconditional branch.</p>
1603 <p>The conditional branch form of the '<tt>br</tt>' instruction takes a
1604 single '<tt>i1</tt>' value and two '<tt>label</tt>' values. The
1605 unconditional form of the '<tt>br</tt>' instruction takes a single
1606 '<tt>label</tt>' value as a target.</p>
1608 <p>Upon execution of a conditional '<tt>br</tt>' instruction, the '<tt>i1</tt>'
1609 argument is evaluated. If the value is <tt>true</tt>, control flows
1610 to the '<tt>iftrue</tt>' <tt>label</tt> argument. If "cond" is <tt>false</tt>,
1611 control flows to the '<tt>iffalse</tt>' <tt>label</tt> argument.</p>
1613 <pre>Test:<br> %cond = <a href="#i_icmp">icmp</a> eq, i32 %a, %b<br> br i1 %cond, label %IfEqual, label %IfUnequal<br>IfEqual:<br> <a
1614 href="#i_ret">ret</a> i32 1<br>IfUnequal:<br> <a href="#i_ret">ret</a> i32 0<br></pre>
1616 <!-- _______________________________________________________________________ -->
1617 <div class="doc_subsubsection">
1618 <a name="i_switch">'<tt>switch</tt>' Instruction</a>
1621 <div class="doc_text">
1625 switch <intty> <value>, label <defaultdest> [ <intty> <val>, label <dest> ... ]
1630 <p>The '<tt>switch</tt>' instruction is used to transfer control flow to one of
1631 several different places. It is a generalization of the '<tt>br</tt>'
1632 instruction, allowing a branch to occur to one of many possible
1638 <p>The '<tt>switch</tt>' instruction uses three parameters: an integer
1639 comparison value '<tt>value</tt>', a default '<tt>label</tt>' destination, and
1640 an array of pairs of comparison value constants and '<tt>label</tt>'s. The
1641 table is not allowed to contain duplicate constant entries.</p>
1645 <p>The <tt>switch</tt> instruction specifies a table of values and
1646 destinations. When the '<tt>switch</tt>' instruction is executed, this
1647 table is searched for the given value. If the value is found, control flow is
1648 transfered to the corresponding destination; otherwise, control flow is
1649 transfered to the default destination.</p>
1651 <h5>Implementation:</h5>
1653 <p>Depending on properties of the target machine and the particular
1654 <tt>switch</tt> instruction, this instruction may be code generated in different
1655 ways. For example, it could be generated as a series of chained conditional
1656 branches or with a lookup table.</p>
1661 <i>; Emulate a conditional br instruction</i>
1662 %Val = <a href="#i_zext">zext</a> i1 %value to i32
1663 switch i32 %Val, label %truedest [i32 0, label %falsedest ]
1665 <i>; Emulate an unconditional br instruction</i>
1666 switch i32 0, label %dest [ ]
1668 <i>; Implement a jump table:</i>
1669 switch i32 %val, label %otherwise [ i32 0, label %onzero
1671 i32 2, label %ontwo ]
1675 <!-- _______________________________________________________________________ -->
1676 <div class="doc_subsubsection">
1677 <a name="i_invoke">'<tt>invoke</tt>' Instruction</a>
1680 <div class="doc_text">
1685 <result> = invoke [<a href="#callingconv">cconv</a>] <ptr to function ty> %<function ptr val>(<function args>)
1686 to label <normal label> unwind label <exception label>
1691 <p>The '<tt>invoke</tt>' instruction causes control to transfer to a specified
1692 function, with the possibility of control flow transfer to either the
1693 '<tt>normal</tt>' label or the
1694 '<tt>exception</tt>' label. If the callee function returns with the
1695 "<tt><a href="#i_ret">ret</a></tt>" instruction, control flow will return to the
1696 "normal" label. If the callee (or any indirect callees) returns with the "<a
1697 href="#i_unwind"><tt>unwind</tt></a>" instruction, control is interrupted and
1698 continued at the dynamically nearest "exception" label.</p>
1702 <p>This instruction requires several arguments:</p>
1706 The optional "cconv" marker indicates which <a href="#callingconv">calling
1707 convention</a> the call should use. If none is specified, the call defaults
1708 to using C calling conventions.
1710 <li>'<tt>ptr to function ty</tt>': shall be the signature of the pointer to
1711 function value being invoked. In most cases, this is a direct function
1712 invocation, but indirect <tt>invoke</tt>s are just as possible, branching off
1713 an arbitrary pointer to function value.
1716 <li>'<tt>function ptr val</tt>': An LLVM value containing a pointer to a
1717 function to be invoked. </li>
1719 <li>'<tt>function args</tt>': argument list whose types match the function
1720 signature argument types. If the function signature indicates the function
1721 accepts a variable number of arguments, the extra arguments can be
1724 <li>'<tt>normal label</tt>': the label reached when the called function
1725 executes a '<tt><a href="#i_ret">ret</a></tt>' instruction. </li>
1727 <li>'<tt>exception label</tt>': the label reached when a callee returns with
1728 the <a href="#i_unwind"><tt>unwind</tt></a> instruction. </li>
1734 <p>This instruction is designed to operate as a standard '<tt><a
1735 href="#i_call">call</a></tt>' instruction in most regards. The primary
1736 difference is that it establishes an association with a label, which is used by
1737 the runtime library to unwind the stack.</p>
1739 <p>This instruction is used in languages with destructors to ensure that proper
1740 cleanup is performed in the case of either a <tt>longjmp</tt> or a thrown
1741 exception. Additionally, this is important for implementation of
1742 '<tt>catch</tt>' clauses in high-level languages that support them.</p>
1746 %retval = invoke i32 %Test(i32 15) to label %Continue
1747 unwind label %TestCleanup <i>; {i32}:retval set</i>
1748 %retval = invoke <a href="#callingconv">coldcc</a> i32 %Test(i32 15) to label %Continue
1749 unwind label %TestCleanup <i>; {i32}:retval set</i>
1754 <!-- _______________________________________________________________________ -->
1756 <div class="doc_subsubsection"> <a name="i_unwind">'<tt>unwind</tt>'
1757 Instruction</a> </div>
1759 <div class="doc_text">
1768 <p>The '<tt>unwind</tt>' instruction unwinds the stack, continuing control flow
1769 at the first callee in the dynamic call stack which used an <a
1770 href="#i_invoke"><tt>invoke</tt></a> instruction to perform the call. This is
1771 primarily used to implement exception handling.</p>
1775 <p>The '<tt>unwind</tt>' intrinsic causes execution of the current function to
1776 immediately halt. The dynamic call stack is then searched for the first <a
1777 href="#i_invoke"><tt>invoke</tt></a> instruction on the call stack. Once found,
1778 execution continues at the "exceptional" destination block specified by the
1779 <tt>invoke</tt> instruction. If there is no <tt>invoke</tt> instruction in the
1780 dynamic call chain, undefined behavior results.</p>
1783 <!-- _______________________________________________________________________ -->
1785 <div class="doc_subsubsection"> <a name="i_unreachable">'<tt>unreachable</tt>'
1786 Instruction</a> </div>
1788 <div class="doc_text">
1797 <p>The '<tt>unreachable</tt>' instruction has no defined semantics. This
1798 instruction is used to inform the optimizer that a particular portion of the
1799 code is not reachable. This can be used to indicate that the code after a
1800 no-return function cannot be reached, and other facts.</p>
1804 <p>The '<tt>unreachable</tt>' instruction has no defined semantics.</p>
1809 <!-- ======================================================================= -->
1810 <div class="doc_subsection"> <a name="binaryops">Binary Operations</a> </div>
1811 <div class="doc_text">
1812 <p>Binary operators are used to do most of the computation in a
1813 program. They require two operands, execute an operation on them, and
1814 produce a single value. The operands might represent
1815 multiple data, as is the case with the <a href="#t_vector">vector</a> data type.
1816 The result value of a binary operator is not
1817 necessarily the same type as its operands.</p>
1818 <p>There are several different binary operators:</p>
1820 <!-- _______________________________________________________________________ -->
1821 <div class="doc_subsubsection"> <a name="i_add">'<tt>add</tt>'
1822 Instruction</a> </div>
1823 <div class="doc_text">
1825 <pre> <result> = add <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1828 <p>The '<tt>add</tt>' instruction returns the sum of its two operands.</p>
1830 <p>The two arguments to the '<tt>add</tt>' instruction must be either <a
1831 href="#t_integer">integer</a> or <a href="#t_floating">floating point</a> values.
1832 This instruction can also take <a href="#t_vector">vector</a> versions of the values.
1833 Both arguments must have identical types.</p>
1835 <p>The value produced is the integer or floating point sum of the two
1838 <pre> <result> = add i32 4, %var <i>; yields {i32}:result = 4 + %var</i>
1841 <!-- _______________________________________________________________________ -->
1842 <div class="doc_subsubsection"> <a name="i_sub">'<tt>sub</tt>'
1843 Instruction</a> </div>
1844 <div class="doc_text">
1846 <pre> <result> = sub <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1849 <p>The '<tt>sub</tt>' instruction returns the difference of its two
1851 <p>Note that the '<tt>sub</tt>' instruction is used to represent the '<tt>neg</tt>'
1852 instruction present in most other intermediate representations.</p>
1854 <p>The two arguments to the '<tt>sub</tt>' instruction must be either <a
1855 href="#t_integer">integer</a> or <a href="#t_floating">floating point</a>
1857 This instruction can also take <a href="#t_vector">vector</a> versions of the values.
1858 Both arguments must have identical types.</p>
1860 <p>The value produced is the integer or floating point difference of
1861 the two operands.</p>
1863 <pre> <result> = sub i32 4, %var <i>; yields {i32}:result = 4 - %var</i>
1864 <result> = sub i32 0, %val <i>; yields {i32}:result = -%var</i>
1867 <!-- _______________________________________________________________________ -->
1868 <div class="doc_subsubsection"> <a name="i_mul">'<tt>mul</tt>'
1869 Instruction</a> </div>
1870 <div class="doc_text">
1872 <pre> <result> = mul <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1875 <p>The '<tt>mul</tt>' instruction returns the product of its two
1878 <p>The two arguments to the '<tt>mul</tt>' instruction must be either <a
1879 href="#t_integer">integer</a> or <a href="#t_floating">floating point</a>
1881 This instruction can also take <a href="#t_vector">vector</a> versions of the values.
1882 Both arguments must have identical types.</p>
1884 <p>The value produced is the integer or floating point product of the
1886 <p>Because the operands are the same width, the result of an integer
1887 multiplication is the same whether the operands should be deemed unsigned or
1890 <pre> <result> = mul i32 4, %var <i>; yields {i32}:result = 4 * %var</i>
1893 <!-- _______________________________________________________________________ -->
1894 <div class="doc_subsubsection"> <a name="i_udiv">'<tt>udiv</tt>' Instruction
1896 <div class="doc_text">
1898 <pre> <result> = udiv <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1901 <p>The '<tt>udiv</tt>' instruction returns the quotient of its two
1904 <p>The two arguments to the '<tt>udiv</tt>' instruction must be
1905 <a href="#t_integer">integer</a> values. Both arguments must have identical
1906 types. This instruction can also take <a href="#t_vector">vector</a> versions
1907 of the values in which case the elements must be integers.</p>
1909 <p>The value produced is the unsigned integer quotient of the two operands. This
1910 instruction always performs an unsigned division operation, regardless of
1911 whether the arguments are unsigned or not.</p>
1913 <pre> <result> = udiv i32 4, %var <i>; yields {i32}:result = 4 / %var</i>
1916 <!-- _______________________________________________________________________ -->
1917 <div class="doc_subsubsection"> <a name="i_sdiv">'<tt>sdiv</tt>' Instruction
1919 <div class="doc_text">
1921 <pre> <result> = sdiv <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1924 <p>The '<tt>sdiv</tt>' instruction returns the quotient of its two
1927 <p>The two arguments to the '<tt>sdiv</tt>' instruction must be
1928 <a href="#t_integer">integer</a> values. Both arguments must have identical
1929 types. This instruction can also take <a href="#t_vector">vector</a> versions
1930 of the values in which case the elements must be integers.</p>
1932 <p>The value produced is the signed integer quotient of the two operands. This
1933 instruction always performs a signed division operation, regardless of whether
1934 the arguments are signed or not.</p>
1936 <pre> <result> = sdiv i32 4, %var <i>; yields {i32}:result = 4 / %var</i>
1939 <!-- _______________________________________________________________________ -->
1940 <div class="doc_subsubsection"> <a name="i_fdiv">'<tt>fdiv</tt>'
1941 Instruction</a> </div>
1942 <div class="doc_text">
1944 <pre> <result> = fdiv <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1947 <p>The '<tt>fdiv</tt>' instruction returns the quotient of its two
1950 <p>The two arguments to the '<tt>div</tt>' instruction must be
1951 <a href="#t_floating">floating point</a> values. Both arguments must have
1952 identical types. This instruction can also take <a href="#t_vector">vector</a>
1953 versions of the values in which case the elements must be floating point.</p>
1955 <p>The value produced is the floating point quotient of the two operands.</p>
1957 <pre> <result> = fdiv float 4.0, %var <i>; yields {float}:result = 4.0 / %var</i>
1960 <!-- _______________________________________________________________________ -->
1961 <div class="doc_subsubsection"> <a name="i_urem">'<tt>urem</tt>' Instruction</a>
1963 <div class="doc_text">
1965 <pre> <result> = urem <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1968 <p>The '<tt>urem</tt>' instruction returns the remainder from the
1969 unsigned division of its two arguments.</p>
1971 <p>The two arguments to the '<tt>urem</tt>' instruction must be
1972 <a href="#t_integer">integer</a> values. Both arguments must have identical
1975 <p>This instruction returns the unsigned integer <i>remainder</i> of a division.
1976 This instruction always performs an unsigned division to get the remainder,
1977 regardless of whether the arguments are unsigned or not.</p>
1979 <pre> <result> = urem i32 4, %var <i>; yields {i32}:result = 4 % %var</i>
1983 <!-- _______________________________________________________________________ -->
1984 <div class="doc_subsubsection"> <a name="i_srem">'<tt>srem</tt>'
1985 Instruction</a> </div>
1986 <div class="doc_text">
1988 <pre> <result> = srem <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1991 <p>The '<tt>srem</tt>' instruction returns the remainder from the
1992 signed division of its two operands.</p>
1994 <p>The two arguments to the '<tt>srem</tt>' instruction must be
1995 <a href="#t_integer">integer</a> values. Both arguments must have identical
1998 <p>This instruction returns the <i>remainder</i> of a division (where the result
1999 has the same sign as the dividend, <tt>var1</tt>), not the <i>modulo</i>
2000 operator (where the result has the same sign as the divisor, <tt>var2</tt>) of
2001 a value. For more information about the difference, see <a
2002 href="http://mathforum.org/dr.math/problems/anne.4.28.99.html">The
2003 Math Forum</a>. For a table of how this is implemented in various languages,
2004 please see <a href="http://en.wikipedia.org/wiki/Modulo_operation">
2005 Wikipedia: modulo operation</a>.</p>
2007 <pre> <result> = srem i32 4, %var <i>; yields {i32}:result = 4 % %var</i>
2011 <!-- _______________________________________________________________________ -->
2012 <div class="doc_subsubsection"> <a name="i_frem">'<tt>frem</tt>'
2013 Instruction</a> </div>
2014 <div class="doc_text">
2016 <pre> <result> = frem <ty> <var1>, <var2> <i>; yields {ty}:result</i>
2019 <p>The '<tt>frem</tt>' instruction returns the remainder from the
2020 division of its two operands.</p>
2022 <p>The two arguments to the '<tt>frem</tt>' instruction must be
2023 <a href="#t_floating">floating point</a> values. Both arguments must have
2024 identical types.</p>
2026 <p>This instruction returns the <i>remainder</i> of a division.</p>
2028 <pre> <result> = frem float 4.0, %var <i>; yields {float}:result = 4.0 % %var</i>
2032 <!-- ======================================================================= -->
2033 <div class="doc_subsection"> <a name="bitwiseops">Bitwise Binary
2034 Operations</a> </div>
2035 <div class="doc_text">
2036 <p>Bitwise binary operators are used to do various forms of
2037 bit-twiddling in a program. They are generally very efficient
2038 instructions and can commonly be strength reduced from other
2039 instructions. They require two operands, execute an operation on them,
2040 and produce a single value. The resulting value of the bitwise binary
2041 operators is always the same type as its first operand.</p>
2044 <!-- _______________________________________________________________________ -->
2045 <div class="doc_subsubsection"> <a name="i_shl">'<tt>shl</tt>'
2046 Instruction</a> </div>
2047 <div class="doc_text">
2049 <pre> <result> = shl <ty> <var1>, <var2> <i>; yields {ty}:result</i>
2052 <p>The '<tt>shl</tt>' instruction returns the first operand shifted to
2053 the left a specified number of bits.</p>
2055 <p>Both arguments to the '<tt>shl</tt>' instruction must be the same <a
2056 href="#t_integer">integer</a> type.</p>
2058 <p>The value produced is <tt>var1</tt> * 2<sup><tt>var2</tt></sup>.</p>
2059 <h5>Example:</h5><pre>
2060 <result> = shl i32 4, %var <i>; yields {i32}: 4 << %var</i>
2061 <result> = shl i32 4, 2 <i>; yields {i32}: 16</i>
2062 <result> = shl i32 1, 10 <i>; yields {i32}: 1024</i>
2065 <!-- _______________________________________________________________________ -->
2066 <div class="doc_subsubsection"> <a name="i_lshr">'<tt>lshr</tt>'
2067 Instruction</a> </div>
2068 <div class="doc_text">
2070 <pre> <result> = lshr <ty> <var1>, <var2> <i>; yields {ty}:result</i>
2074 <p>The '<tt>lshr</tt>' instruction (logical shift right) returns the first
2075 operand shifted to the right a specified number of bits.</p>
2078 <p>Both arguments to the '<tt>lshr</tt>' instruction must be the same
2079 <a href="#t_integer">integer</a> type.</p>
2082 <p>This instruction always performs a logical shift right operation. The most
2083 significant bits of the result will be filled with zero bits after the
2088 <result> = lshr i32 4, 1 <i>; yields {i32}:result = 2</i>
2089 <result> = lshr i32 4, 2 <i>; yields {i32}:result = 1</i>
2090 <result> = lshr i8 4, 3 <i>; yields {i8}:result = 0</i>
2091 <result> = lshr i8 -2, 1 <i>; yields {i8}:result = 0x7FFFFFFF </i>
2095 <!-- _______________________________________________________________________ -->
2096 <div class="doc_subsubsection"> <a name="i_ashr">'<tt>ashr</tt>'
2097 Instruction</a> </div>
2098 <div class="doc_text">
2101 <pre> <result> = ashr <ty> <var1>, <var2> <i>; yields {ty}:result</i>
2105 <p>The '<tt>ashr</tt>' instruction (arithmetic shift right) returns the first
2106 operand shifted to the right a specified number of bits.</p>
2109 <p>Both arguments to the '<tt>ashr</tt>' instruction must be the same
2110 <a href="#t_integer">integer</a> type.</p>
2113 <p>This instruction always performs an arithmetic shift right operation,
2114 The most significant bits of the result will be filled with the sign bit
2115 of <tt>var1</tt>.</p>
2119 <result> = ashr i32 4, 1 <i>; yields {i32}:result = 2</i>
2120 <result> = ashr i32 4, 2 <i>; yields {i32}:result = 1</i>
2121 <result> = ashr i8 4, 3 <i>; yields {i8}:result = 0</i>
2122 <result> = ashr i8 -2, 1 <i>; yields {i8}:result = -1</i>
2126 <!-- _______________________________________________________________________ -->
2127 <div class="doc_subsubsection"> <a name="i_and">'<tt>and</tt>'
2128 Instruction</a> </div>
2129 <div class="doc_text">
2131 <pre> <result> = and <ty> <var1>, <var2> <i>; yields {ty}:result</i>
2134 <p>The '<tt>and</tt>' instruction returns the bitwise logical and of
2135 its two operands.</p>
2137 <p>The two arguments to the '<tt>and</tt>' instruction must be <a
2138 href="#t_integer">integer</a> values. Both arguments must have
2139 identical types.</p>
2141 <p>The truth table used for the '<tt>and</tt>' instruction is:</p>
2143 <div style="align: center">
2144 <table border="1" cellspacing="0" cellpadding="4">
2175 <pre> <result> = and i32 4, %var <i>; yields {i32}:result = 4 & %var</i>
2176 <result> = and i32 15, 40 <i>; yields {i32}:result = 8</i>
2177 <result> = and i32 4, 8 <i>; yields {i32}:result = 0</i>
2180 <!-- _______________________________________________________________________ -->
2181 <div class="doc_subsubsection"> <a name="i_or">'<tt>or</tt>' Instruction</a> </div>
2182 <div class="doc_text">
2184 <pre> <result> = or <ty> <var1>, <var2> <i>; yields {ty}:result</i>
2187 <p>The '<tt>or</tt>' instruction returns the bitwise logical inclusive
2188 or of its two operands.</p>
2190 <p>The two arguments to the '<tt>or</tt>' instruction must be <a
2191 href="#t_integer">integer</a> values. Both arguments must have
2192 identical types.</p>
2194 <p>The truth table used for the '<tt>or</tt>' instruction is:</p>
2196 <div style="align: center">
2197 <table border="1" cellspacing="0" cellpadding="4">
2228 <pre> <result> = or i32 4, %var <i>; yields {i32}:result = 4 | %var</i>
2229 <result> = or i32 15, 40 <i>; yields {i32}:result = 47</i>
2230 <result> = or i32 4, 8 <i>; yields {i32}:result = 12</i>
2233 <!-- _______________________________________________________________________ -->
2234 <div class="doc_subsubsection"> <a name="i_xor">'<tt>xor</tt>'
2235 Instruction</a> </div>
2236 <div class="doc_text">
2238 <pre> <result> = xor <ty> <var1>, <var2> <i>; yields {ty}:result</i>
2241 <p>The '<tt>xor</tt>' instruction returns the bitwise logical exclusive
2242 or of its two operands. The <tt>xor</tt> is used to implement the
2243 "one's complement" operation, which is the "~" operator in C.</p>
2245 <p>The two arguments to the '<tt>xor</tt>' instruction must be <a
2246 href="#t_integer">integer</a> values. Both arguments must have
2247 identical types.</p>
2249 <p>The truth table used for the '<tt>xor</tt>' instruction is:</p>
2251 <div style="align: center">
2252 <table border="1" cellspacing="0" cellpadding="4">
2284 <pre> <result> = xor i32 4, %var <i>; yields {i32}:result = 4 ^ %var</i>
2285 <result> = xor i32 15, 40 <i>; yields {i32}:result = 39</i>
2286 <result> = xor i32 4, 8 <i>; yields {i32}:result = 12</i>
2287 <result> = xor i32 %V, -1 <i>; yields {i32}:result = ~%V</i>
2291 <!-- ======================================================================= -->
2292 <div class="doc_subsection">
2293 <a name="vectorops">Vector Operations</a>
2296 <div class="doc_text">
2298 <p>LLVM supports several instructions to represent vector operations in a
2299 target-independent manner. This instructions cover the element-access and
2300 vector-specific operations needed to process vectors effectively. While LLVM
2301 does directly support these vector operations, many sophisticated algorithms
2302 will want to use target-specific intrinsics to take full advantage of a specific
2307 <!-- _______________________________________________________________________ -->
2308 <div class="doc_subsubsection">
2309 <a name="i_extractelement">'<tt>extractelement</tt>' Instruction</a>
2312 <div class="doc_text">
2317 <result> = extractelement <n x <ty>> <val>, i32 <idx> <i>; yields <ty></i>
2323 The '<tt>extractelement</tt>' instruction extracts a single scalar
2324 element from a vector at a specified index.
2331 The first operand of an '<tt>extractelement</tt>' instruction is a
2332 value of <a href="#t_vector">vector</a> type. The second operand is
2333 an index indicating the position from which to extract the element.
2334 The index may be a variable.</p>
2339 The result is a scalar of the same type as the element type of
2340 <tt>val</tt>. Its value is the value at position <tt>idx</tt> of
2341 <tt>val</tt>. If <tt>idx</tt> exceeds the length of <tt>val</tt>, the
2342 results are undefined.
2348 %result = extractelement <4 x i32> %vec, i32 0 <i>; yields i32</i>
2353 <!-- _______________________________________________________________________ -->
2354 <div class="doc_subsubsection">
2355 <a name="i_insertelement">'<tt>insertelement</tt>' Instruction</a>
2358 <div class="doc_text">
2363 <result> = insertelement <n x <ty>> <val>, <ty> <elt>, i32 <idx> <i>; yields <n x <ty>></i>
2369 The '<tt>insertelement</tt>' instruction inserts a scalar
2370 element into a vector at a specified index.
2377 The first operand of an '<tt>insertelement</tt>' instruction is a
2378 value of <a href="#t_vector">vector</a> type. The second operand is a
2379 scalar value whose type must equal the element type of the first
2380 operand. The third operand is an index indicating the position at
2381 which to insert the value. The index may be a variable.</p>
2386 The result is a vector of the same type as <tt>val</tt>. Its
2387 element values are those of <tt>val</tt> except at position
2388 <tt>idx</tt>, where it gets the value <tt>elt</tt>. If <tt>idx</tt>
2389 exceeds the length of <tt>val</tt>, the results are undefined.
2395 %result = insertelement <4 x i32> %vec, i32 1, i32 0 <i>; yields <4 x i32></i>
2399 <!-- _______________________________________________________________________ -->
2400 <div class="doc_subsubsection">
2401 <a name="i_shufflevector">'<tt>shufflevector</tt>' Instruction</a>
2404 <div class="doc_text">
2409 <result> = shufflevector <n x <ty>> <v1>, <n x <ty>> <v2>, <n x i32> <mask> <i>; yields <n x <ty>></i>
2415 The '<tt>shufflevector</tt>' instruction constructs a permutation of elements
2416 from two input vectors, returning a vector of the same type.
2422 The first two operands of a '<tt>shufflevector</tt>' instruction are vectors
2423 with types that match each other and types that match the result of the
2424 instruction. The third argument is a shuffle mask, which has the same number
2425 of elements as the other vector type, but whose element type is always 'i32'.
2429 The shuffle mask operand is required to be a constant vector with either
2430 constant integer or undef values.
2436 The elements of the two input vectors are numbered from left to right across
2437 both of the vectors. The shuffle mask operand specifies, for each element of
2438 the result vector, which element of the two input registers the result element
2439 gets. The element selector may be undef (meaning "don't care") and the second
2440 operand may be undef if performing a shuffle from only one vector.
2446 %result = shufflevector <4 x i32> %v1, <4 x i32> %v2,
2447 <4 x i32> <i32 0, i32 4, i32 1, i32 5> <i>; yields <4 x i32></i>
2448 %result = shufflevector <4 x i32> %v1, <4 x i32> undef,
2449 <4 x i32> <i32 0, i32 1, i32 2, i32 3> <i>; yields <4 x i32></i> - Identity shuffle.
2454 <!-- ======================================================================= -->
2455 <div class="doc_subsection">
2456 <a name="memoryops">Memory Access and Addressing Operations</a>
2459 <div class="doc_text">
2461 <p>A key design point of an SSA-based representation is how it
2462 represents memory. In LLVM, no memory locations are in SSA form, which
2463 makes things very simple. This section describes how to read, write,
2464 allocate, and free memory in LLVM.</p>
2468 <!-- _______________________________________________________________________ -->
2469 <div class="doc_subsubsection">
2470 <a name="i_malloc">'<tt>malloc</tt>' Instruction</a>
2473 <div class="doc_text">
2478 <result> = malloc <type>[, i32 <NumElements>][, align <alignment>] <i>; yields {type*}:result</i>
2483 <p>The '<tt>malloc</tt>' instruction allocates memory from the system
2484 heap and returns a pointer to it.</p>
2488 <p>The '<tt>malloc</tt>' instruction allocates
2489 <tt>sizeof(<type>)*NumElements</tt>
2490 bytes of memory from the operating system and returns a pointer of the
2491 appropriate type to the program. If "NumElements" is specified, it is the
2492 number of elements allocated. If an alignment is specified, the value result
2493 of the allocation is guaranteed to be aligned to at least that boundary. If
2494 not specified, or if zero, the target can choose to align the allocation on any
2495 convenient boundary.</p>
2497 <p>'<tt>type</tt>' must be a sized type.</p>
2501 <p>Memory is allocated using the system "<tt>malloc</tt>" function, and
2502 a pointer is returned.</p>
2507 %array = malloc [4 x i8 ] <i>; yields {[%4 x i8]*}:array</i>
2509 %size = <a href="#i_add">add</a> i32 2, 2 <i>; yields {i32}:size = i32 4</i>
2510 %array1 = malloc i8, i32 4 <i>; yields {i8*}:array1</i>
2511 %array2 = malloc [12 x i8], i32 %size <i>; yields {[12 x i8]*}:array2</i>
2512 %array3 = malloc i32, i32 4, align 1024 <i>; yields {i32*}:array3</i>
2513 %array4 = malloc i32, align 1024 <i>; yields {i32*}:array4</i>
2517 <!-- _______________________________________________________________________ -->
2518 <div class="doc_subsubsection">
2519 <a name="i_free">'<tt>free</tt>' Instruction</a>
2522 <div class="doc_text">
2527 free <type> <value> <i>; yields {void}</i>
2532 <p>The '<tt>free</tt>' instruction returns memory back to the unused
2533 memory heap to be reallocated in the future.</p>
2537 <p>'<tt>value</tt>' shall be a pointer value that points to a value
2538 that was allocated with the '<tt><a href="#i_malloc">malloc</a></tt>'
2543 <p>Access to the memory pointed to by the pointer is no longer defined
2544 after this instruction executes.</p>
2549 %array = <a href="#i_malloc">malloc</a> [4 x i8] <i>; yields {[4 x i8]*}:array</i>
2550 free [4 x i8]* %array
2554 <!-- _______________________________________________________________________ -->
2555 <div class="doc_subsubsection">
2556 <a name="i_alloca">'<tt>alloca</tt>' Instruction</a>
2559 <div class="doc_text">
2564 <result> = alloca <type>[, i32 <NumElements>][, align <alignment>] <i>; yields {type*}:result</i>
2569 <p>The '<tt>alloca</tt>' instruction allocates memory on the current
2570 stack frame of the procedure that is live until the current function
2571 returns to its caller.</p>
2575 <p>The '<tt>alloca</tt>' instruction allocates <tt>sizeof(<type>)*NumElements</tt>
2576 bytes of memory on the runtime stack, returning a pointer of the
2577 appropriate type to the program. If "NumElements" is specified, it is the
2578 number of elements allocated. If an alignment is specified, the value result
2579 of the allocation is guaranteed to be aligned to at least that boundary. If
2580 not specified, or if zero, the target can choose to align the allocation on any
2581 convenient boundary.</p>
2583 <p>'<tt>type</tt>' may be any sized type.</p>
2587 <p>Memory is allocated; a pointer is returned. '<tt>alloca</tt>'d
2588 memory is automatically released when the function returns. The '<tt>alloca</tt>'
2589 instruction is commonly used to represent automatic variables that must
2590 have an address available. When the function returns (either with the <tt><a
2591 href="#i_ret">ret</a></tt> or <tt><a href="#i_unwind">unwind</a></tt>
2592 instructions), the memory is reclaimed.</p>
2597 %ptr = alloca i32 <i>; yields {i32*}:ptr</i>
2598 %ptr = alloca i32, i32 4 <i>; yields {i32*}:ptr</i>
2599 %ptr = alloca i32, i32 4, align 1024 <i>; yields {i32*}:ptr</i>
2600 %ptr = alloca i32, align 1024 <i>; yields {i32*}:ptr</i>
2604 <!-- _______________________________________________________________________ -->
2605 <div class="doc_subsubsection"> <a name="i_load">'<tt>load</tt>'
2606 Instruction</a> </div>
2607 <div class="doc_text">
2609 <pre> <result> = load <ty>* <pointer><br> <result> = volatile load <ty>* <pointer><br></pre>
2611 <p>The '<tt>load</tt>' instruction is used to read from memory.</p>
2613 <p>The argument to the '<tt>load</tt>' instruction specifies the memory
2614 address from which to load. The pointer must point to a <a
2615 href="#t_firstclass">first class</a> type. If the <tt>load</tt> is
2616 marked as <tt>volatile</tt>, then the optimizer is not allowed to modify
2617 the number or order of execution of this <tt>load</tt> with other
2618 volatile <tt>load</tt> and <tt><a href="#i_store">store</a></tt>
2621 <p>The location of memory pointed to is loaded.</p>
2623 <pre> %ptr = <a href="#i_alloca">alloca</a> i32 <i>; yields {i32*}:ptr</i>
2625 href="#i_store">store</a> i32 3, i32* %ptr <i>; yields {void}</i>
2626 %val = load i32* %ptr <i>; yields {i32}:val = i32 3</i>
2629 <!-- _______________________________________________________________________ -->
2630 <div class="doc_subsubsection"> <a name="i_store">'<tt>store</tt>'
2631 Instruction</a> </div>
2632 <div class="doc_text">
2634 <pre> store <ty> <value>, <ty>* <pointer> <i>; yields {void}</i>
2635 volatile store <ty> <value>, <ty>* <pointer> <i>; yields {void}</i>
2638 <p>The '<tt>store</tt>' instruction is used to write to memory.</p>
2640 <p>There are two arguments to the '<tt>store</tt>' instruction: a value
2641 to store and an address in which to store it. The type of the '<tt><pointer></tt>'
2642 operand must be a pointer to the type of the '<tt><value></tt>'
2643 operand. If the <tt>store</tt> is marked as <tt>volatile</tt>, then the
2644 optimizer is not allowed to modify the number or order of execution of
2645 this <tt>store</tt> with other volatile <tt>load</tt> and <tt><a
2646 href="#i_store">store</a></tt> instructions.</p>
2648 <p>The contents of memory are updated to contain '<tt><value></tt>'
2649 at the location specified by the '<tt><pointer></tt>' operand.</p>
2651 <pre> %ptr = <a href="#i_alloca">alloca</a> i32 <i>; yields {i32*}:ptr</i>
2653 href="#i_store">store</a> i32 3, i32* %ptr <i>; yields {void}</i>
2654 %val = load i32* %ptr <i>; yields {i32}:val = i32 3</i>
2658 <!-- _______________________________________________________________________ -->
2659 <div class="doc_subsubsection">
2660 <a name="i_getelementptr">'<tt>getelementptr</tt>' Instruction</a>
2663 <div class="doc_text">
2666 <result> = getelementptr <ty>* <ptrval>{, <ty> <idx>}*
2672 The '<tt>getelementptr</tt>' instruction is used to get the address of a
2673 subelement of an aggregate data structure.</p>
2677 <p>This instruction takes a list of integer operands that indicate what
2678 elements of the aggregate object to index to. The actual types of the arguments
2679 provided depend on the type of the first pointer argument. The
2680 '<tt>getelementptr</tt>' instruction is used to index down through the type
2681 levels of a structure or to a specific index in an array. When indexing into a
2682 structure, only <tt>i32</tt> integer constants are allowed. When indexing
2683 into an array or pointer, only integers of 32 or 64 bits are allowed, and will
2684 be sign extended to 64-bit values.</p>
2686 <p>For example, let's consider a C code fragment and how it gets
2687 compiled to LLVM:</p>
2701 define i32 *foo(struct ST *s) {
2702 return &s[1].Z.B[5][13];
2706 <p>The LLVM code generated by the GCC frontend is:</p>
2709 %RT = type { i8 , [10 x [20 x i32]], i8 }
2710 %ST = type { i32, double, %RT }
2712 define i32* %foo(%ST* %s) {
2714 %reg = getelementptr %ST* %s, i32 1, i32 2, i32 1, i32 5, i32 13
2721 <p>The index types specified for the '<tt>getelementptr</tt>' instruction depend
2722 on the pointer type that is being indexed into. <a href="#t_pointer">Pointer</a>
2723 and <a href="#t_array">array</a> types can use a 32-bit or 64-bit
2724 <a href="#t_integer">integer</a> type but the value will always be sign extended
2725 to 64-bits. <a href="#t_struct">Structure</a> types, require <tt>i32</tt>
2726 <b>constants</b>.</p>
2728 <p>In the example above, the first index is indexing into the '<tt>%ST*</tt>'
2729 type, which is a pointer, yielding a '<tt>%ST</tt>' = '<tt>{ i32, double, %RT
2730 }</tt>' type, a structure. The second index indexes into the third element of
2731 the structure, yielding a '<tt>%RT</tt>' = '<tt>{ i8 , [10 x [20 x i32]],
2732 i8 }</tt>' type, another structure. The third index indexes into the second
2733 element of the structure, yielding a '<tt>[10 x [20 x i32]]</tt>' type, an
2734 array. The two dimensions of the array are subscripted into, yielding an
2735 '<tt>i32</tt>' type. The '<tt>getelementptr</tt>' instruction returns a pointer
2736 to this element, thus computing a value of '<tt>i32*</tt>' type.</p>
2738 <p>Note that it is perfectly legal to index partially through a
2739 structure, returning a pointer to an inner element. Because of this,
2740 the LLVM code for the given testcase is equivalent to:</p>
2743 define i32* %foo(%ST* %s) {
2744 %t1 = getelementptr %ST* %s, i32 1 <i>; yields %ST*:%t1</i>
2745 %t2 = getelementptr %ST* %t1, i32 0, i32 2 <i>; yields %RT*:%t2</i>
2746 %t3 = getelementptr %RT* %t2, i32 0, i32 1 <i>; yields [10 x [20 x i32]]*:%t3</i>
2747 %t4 = getelementptr [10 x [20 x i32]]* %t3, i32 0, i32 5 <i>; yields [20 x i32]*:%t4</i>
2748 %t5 = getelementptr [20 x i32]* %t4, i32 0, i32 13 <i>; yields i32*:%t5</i>
2753 <p>Note that it is undefined to access an array out of bounds: array and
2754 pointer indexes must always be within the defined bounds of the array type.
2755 The one exception for this rules is zero length arrays. These arrays are
2756 defined to be accessible as variable length arrays, which requires access
2757 beyond the zero'th element.</p>
2759 <p>The getelementptr instruction is often confusing. For some more insight
2760 into how it works, see <a href="GetElementPtr.html">the getelementptr
2766 <i>; yields [12 x i8]*:aptr</i>
2767 %aptr = getelementptr {i32, [12 x i8]}* %sptr, i64 0, i32 1
2771 <!-- ======================================================================= -->
2772 <div class="doc_subsection"> <a name="convertops">Conversion Operations</a>
2774 <div class="doc_text">
2775 <p>The instructions in this category are the conversion instructions (casting)
2776 which all take a single operand and a type. They perform various bit conversions
2780 <!-- _______________________________________________________________________ -->
2781 <div class="doc_subsubsection">
2782 <a name="i_trunc">'<tt>trunc .. to</tt>' Instruction</a>
2784 <div class="doc_text">
2788 <result> = trunc <ty> <value> to <ty2> <i>; yields ty2</i>
2793 The '<tt>trunc</tt>' instruction truncates its operand to the type <tt>ty2</tt>.
2798 The '<tt>trunc</tt>' instruction takes a <tt>value</tt> to trunc, which must
2799 be an <a href="#t_integer">integer</a> type, and a type that specifies the size
2800 and type of the result, which must be an <a href="#t_integer">integer</a>
2801 type. The bit size of <tt>value</tt> must be larger than the bit size of
2802 <tt>ty2</tt>. Equal sized types are not allowed.</p>
2806 The '<tt>trunc</tt>' instruction truncates the high order bits in <tt>value</tt>
2807 and converts the remaining bits to <tt>ty2</tt>. Since the source size must be
2808 larger than the destination size, <tt>trunc</tt> cannot be a <i>no-op cast</i>.
2809 It will always truncate bits.</p>
2813 %X = trunc i32 257 to i8 <i>; yields i8:1</i>
2814 %Y = trunc i32 123 to i1 <i>; yields i1:true</i>
2815 %Y = trunc i32 122 to i1 <i>; yields i1:false</i>
2819 <!-- _______________________________________________________________________ -->
2820 <div class="doc_subsubsection">
2821 <a name="i_zext">'<tt>zext .. to</tt>' Instruction</a>
2823 <div class="doc_text">
2827 <result> = zext <ty> <value> to <ty2> <i>; yields ty2</i>
2831 <p>The '<tt>zext</tt>' instruction zero extends its operand to type
2836 <p>The '<tt>zext</tt>' instruction takes a value to cast, which must be of
2837 <a href="#t_integer">integer</a> type, and a type to cast it to, which must
2838 also be of <a href="#t_integer">integer</a> type. The bit size of the
2839 <tt>value</tt> must be smaller than the bit size of the destination type,
2843 <p>The <tt>zext</tt> fills the high order bits of the <tt>value</tt> with zero
2844 bits until it reaches the size of the destination type, <tt>ty2</tt>. When the
2845 the operand and the type are the same size, no bit filling is done and the
2846 cast is considered a <i>no-op cast</i> because no bits change (only the type
2849 <p>When zero extending from i1, the result will always be either 0 or 1.</p>
2853 %X = zext i32 257 to i64 <i>; yields i64:257</i>
2854 %Y = zext i1 true to i32 <i>; yields i32:1</i>
2858 <!-- _______________________________________________________________________ -->
2859 <div class="doc_subsubsection">
2860 <a name="i_sext">'<tt>sext .. to</tt>' Instruction</a>
2862 <div class="doc_text">
2866 <result> = sext <ty> <value> to <ty2> <i>; yields ty2</i>
2870 <p>The '<tt>sext</tt>' sign extends <tt>value</tt> to the type <tt>ty2</tt>.</p>
2874 The '<tt>sext</tt>' instruction takes a value to cast, which must be of
2875 <a href="#t_integer">integer</a> type, and a type to cast it to, which must
2876 also be of <a href="#t_integer">integer</a> type. The bit size of the
2877 <tt>value</tt> must be smaller than the bit size of the destination type,
2882 The '<tt>sext</tt>' instruction performs a sign extension by copying the sign
2883 bit (highest order bit) of the <tt>value</tt> until it reaches the bit size of
2884 the type <tt>ty2</tt>. When the the operand and the type are the same size,
2885 no bit filling is done and the cast is considered a <i>no-op cast</i> because
2886 no bits change (only the type changes).</p>
2888 <p>When sign extending from i1, the extension always results in -1 or 0.</p>
2892 %X = sext i8 -1 to i16 <i>; yields i16 :65535</i>
2893 %Y = sext i1 true to i32 <i>; yields i32:-1</i>
2897 <!-- _______________________________________________________________________ -->
2898 <div class="doc_subsubsection">
2899 <a name="i_fptrunc">'<tt>fptrunc .. to</tt>' Instruction</a>
2902 <div class="doc_text">
2907 <result> = fptrunc <ty> <value> to <ty2> <i>; yields ty2</i>
2911 <p>The '<tt>fptrunc</tt>' instruction truncates <tt>value</tt> to type
2916 <p>The '<tt>fptrunc</tt>' instruction takes a <a href="#t_floating">floating
2917 point</a> value to cast and a <a href="#t_floating">floating point</a> type to
2918 cast it to. The size of <tt>value</tt> must be larger than the size of
2919 <tt>ty2</tt>. This implies that <tt>fptrunc</tt> cannot be used to make a
2920 <i>no-op cast</i>.</p>
2923 <p> The '<tt>fptrunc</tt>' instruction truncates a <tt>value</tt> from a larger
2924 <a href="#t_floating">floating point</a> type to a smaller
2925 <a href="#t_floating">floating point</a> type. If the value cannot fit within
2926 the destination type, <tt>ty2</tt>, then the results are undefined.</p>
2930 %X = fptrunc double 123.0 to float <i>; yields float:123.0</i>
2931 %Y = fptrunc double 1.0E+300 to float <i>; yields undefined</i>
2935 <!-- _______________________________________________________________________ -->
2936 <div class="doc_subsubsection">
2937 <a name="i_fpext">'<tt>fpext .. to</tt>' Instruction</a>
2939 <div class="doc_text">
2943 <result> = fpext <ty> <value> to <ty2> <i>; yields ty2</i>
2947 <p>The '<tt>fpext</tt>' extends a floating point <tt>value</tt> to a larger
2948 floating point value.</p>
2951 <p>The '<tt>fpext</tt>' instruction takes a
2952 <a href="#t_floating">floating point</a> <tt>value</tt> to cast,
2953 and a <a href="#t_floating">floating point</a> type to cast it to. The source
2954 type must be smaller than the destination type.</p>
2957 <p>The '<tt>fpext</tt>' instruction extends the <tt>value</tt> from a smaller
2958 <a href="#t_floating">floating point</a> type to a larger
2959 <a href="#t_floating">floating point</a> type. The <tt>fpext</tt> cannot be
2960 used to make a <i>no-op cast</i> because it always changes bits. Use
2961 <tt>bitcast</tt> to make a <i>no-op cast</i> for a floating point cast.</p>
2965 %X = fpext float 3.1415 to double <i>; yields double:3.1415</i>
2966 %Y = fpext float 1.0 to float <i>; yields float:1.0 (no-op)</i>
2970 <!-- _______________________________________________________________________ -->
2971 <div class="doc_subsubsection">
2972 <a name="i_fptoui">'<tt>fptoui .. to</tt>' Instruction</a>
2974 <div class="doc_text">
2978 <result> = fp2uint <ty> <value> to <ty2> <i>; yields ty2</i>
2982 <p>The '<tt>fp2uint</tt>' converts a floating point <tt>value</tt> to its
2983 unsigned integer equivalent of type <tt>ty2</tt>.
2987 <p>The '<tt>fp2uint</tt>' instruction takes a value to cast, which must be a
2988 <a href="#t_floating">floating point</a> value, and a type to cast it to, which
2989 must be an <a href="#t_integer">integer</a> type.</p>
2992 <p> The '<tt>fp2uint</tt>' instruction converts its
2993 <a href="#t_floating">floating point</a> operand into the nearest (rounding
2994 towards zero) unsigned integer value. If the value cannot fit in <tt>ty2</tt>,
2995 the results are undefined.</p>
2997 <p>When converting to i1, the conversion is done as a comparison against
2998 zero. If the <tt>value</tt> was zero, the i1 result will be <tt>false</tt>.
2999 If the <tt>value</tt> was non-zero, the i1 result will be <tt>true</tt>.</p>
3003 %X = fp2uint double 123.0 to i32 <i>; yields i32:123</i>
3004 %Y = fp2uint float 1.0E+300 to i1 <i>; yields i1:true</i>
3005 %X = fp2uint float 1.04E+17 to i8 <i>; yields undefined:1</i>
3009 <!-- _______________________________________________________________________ -->
3010 <div class="doc_subsubsection">
3011 <a name="i_fptosi">'<tt>fptosi .. to</tt>' Instruction</a>
3013 <div class="doc_text">
3017 <result> = fptosi <ty> <value> to <ty2> <i>; yields ty2</i>
3021 <p>The '<tt>fptosi</tt>' instruction converts
3022 <a href="#t_floating">floating point</a> <tt>value</tt> to type <tt>ty2</tt>.
3027 <p> The '<tt>fptosi</tt>' instruction takes a value to cast, which must be a
3028 <a href="#t_floating">floating point</a> value, and a type to cast it to, which
3029 must also be an <a href="#t_integer">integer</a> type.</p>
3032 <p>The '<tt>fptosi</tt>' instruction converts its
3033 <a href="#t_floating">floating point</a> operand into the nearest (rounding
3034 towards zero) signed integer value. If the value cannot fit in <tt>ty2</tt>,
3035 the results are undefined.</p>
3037 <p>When converting to i1, the conversion is done as a comparison against
3038 zero. If the <tt>value</tt> was zero, the i1 result will be <tt>false</tt>.
3039 If the <tt>value</tt> was non-zero, the i1 result will be <tt>true</tt>.</p>
3043 %X = fptosi double -123.0 to i32 <i>; yields i32:-123</i>
3044 %Y = fptosi float 1.0E-247 to i1 <i>; yields i1:true</i>
3045 %X = fptosi float 1.04E+17 to i8 <i>; yields undefined:1</i>
3049 <!-- _______________________________________________________________________ -->
3050 <div class="doc_subsubsection">
3051 <a name="i_uitofp">'<tt>uitofp .. to</tt>' Instruction</a>
3053 <div class="doc_text">
3057 <result> = uitofp <ty> <value> to <ty2> <i>; yields ty2</i>
3061 <p>The '<tt>uitofp</tt>' instruction regards <tt>value</tt> as an unsigned
3062 integer and converts that value to the <tt>ty2</tt> type.</p>
3066 <p>The '<tt>uitofp</tt>' instruction takes a value to cast, which must be an
3067 <a href="#t_integer">integer</a> value, and a type to cast it to, which must
3068 be a <a href="#t_floating">floating point</a> type.</p>
3071 <p>The '<tt>uitofp</tt>' instruction interprets its operand as an unsigned
3072 integer quantity and converts it to the corresponding floating point value. If
3073 the value cannot fit in the floating point value, the results are undefined.</p>
3078 %X = uitofp i32 257 to float <i>; yields float:257.0</i>
3079 %Y = uitofp i8 -1 to double <i>; yields double:255.0</i>
3083 <!-- _______________________________________________________________________ -->
3084 <div class="doc_subsubsection">
3085 <a name="i_sitofp">'<tt>sitofp .. to</tt>' Instruction</a>
3087 <div class="doc_text">
3091 <result> = sitofp <ty> <value> to <ty2> <i>; yields ty2</i>
3095 <p>The '<tt>sitofp</tt>' instruction regards <tt>value</tt> as a signed
3096 integer and converts that value to the <tt>ty2</tt> type.</p>
3099 <p>The '<tt>sitofp</tt>' instruction takes a value to cast, which must be an
3100 <a href="#t_integer">integer</a> value, and a type to cast it to, which must be
3101 a <a href="#t_floating">floating point</a> type.</p>
3104 <p>The '<tt>sitofp</tt>' instruction interprets its operand as a signed
3105 integer quantity and converts it to the corresponding floating point value. If
3106 the value cannot fit in the floating point value, the results are undefined.</p>
3110 %X = sitofp i32 257 to float <i>; yields float:257.0</i>
3111 %Y = sitofp i8 -1 to double <i>; yields double:-1.0</i>
3115 <!-- _______________________________________________________________________ -->
3116 <div class="doc_subsubsection">
3117 <a name="i_ptrtoint">'<tt>ptrtoint .. to</tt>' Instruction</a>
3119 <div class="doc_text">
3123 <result> = ptrtoint <ty> <value> to <ty2> <i>; yields ty2</i>
3127 <p>The '<tt>ptrtoint</tt>' instruction converts the pointer <tt>value</tt> to
3128 the integer type <tt>ty2</tt>.</p>
3131 <p>The '<tt>ptrtoint</tt>' instruction takes a <tt>value</tt> to cast, which
3132 must be a <a href="#t_pointer">pointer</a> value, and a type to cast it to
3133 <tt>ty2</tt>, which must be an <a href="#t_integer">integer</a> type.
3136 <p>The '<tt>ptrtoint</tt>' instruction converts <tt>value</tt> to integer type
3137 <tt>ty2</tt> by interpreting the pointer value as an integer and either
3138 truncating or zero extending that value to the size of the integer type. If
3139 <tt>value</tt> is smaller than <tt>ty2</tt> then a zero extension is done. If
3140 <tt>value</tt> is larger than <tt>ty2</tt> then a truncation is done. If they
3141 are the same size, then nothing is done (<i>no-op cast</i>).</p>
3145 %X = ptrtoint i32* %X to i8 <i>; yields truncation on 32-bit</i>
3146 %Y = ptrtoint i32* %x to i64 <i>; yields zero extend on 32-bit</i>
3150 <!-- _______________________________________________________________________ -->
3151 <div class="doc_subsubsection">
3152 <a name="i_inttoptr">'<tt>inttoptr .. to</tt>' Instruction</a>
3154 <div class="doc_text">
3158 <result> = inttoptr <ty> <value> to <ty2> <i>; yields ty2</i>
3162 <p>The '<tt>inttoptr</tt>' instruction converts an integer <tt>value</tt> to
3163 a pointer type, <tt>ty2</tt>.</p>
3166 <p>The '<tt>inttoptr</tt>' instruction takes an <a href="#t_integer">integer</a>
3167 value to cast, and a type to cast it to, which must be a
3168 <a href="#t_pointer">pointer</a> type.
3171 <p>The '<tt>inttoptr</tt>' instruction converts <tt>value</tt> to type
3172 <tt>ty2</tt> by applying either a zero extension or a truncation depending on
3173 the size of the integer <tt>value</tt>. If <tt>value</tt> is larger than the
3174 size of a pointer then a truncation is done. If <tt>value</tt> is smaller than
3175 the size of a pointer then a zero extension is done. If they are the same size,
3176 nothing is done (<i>no-op cast</i>).</p>
3180 %X = inttoptr i32 255 to i32* <i>; yields zero extend on 64-bit</i>
3181 %X = inttoptr i32 255 to i32* <i>; yields no-op on 32-bit </i>
3182 %Y = inttoptr i16 0 to i32* <i>; yields zero extend on 32-bit</i>
3186 <!-- _______________________________________________________________________ -->
3187 <div class="doc_subsubsection">
3188 <a name="i_bitcast">'<tt>bitcast .. to</tt>' Instruction</a>
3190 <div class="doc_text">
3194 <result> = bitcast <ty> <value> to <ty2> <i>; yields ty2</i>
3198 <p>The '<tt>bitcast</tt>' instruction converts <tt>value</tt> to type
3199 <tt>ty2</tt> without changing any bits.</p>
3202 <p>The '<tt>bitcast</tt>' instruction takes a value to cast, which must be
3203 a first class value, and a type to cast it to, which must also be a <a
3204 href="#t_firstclass">first class</a> type. The bit sizes of <tt>value</tt>
3205 and the destination type, <tt>ty2</tt>, must be identical. If the source
3206 type is a pointer, the destination type must also be a pointer.</p>
3209 <p>The '<tt>bitcast</tt>' instruction converts <tt>value</tt> to type
3210 <tt>ty2</tt>. It is always a <i>no-op cast</i> because no bits change with
3211 this conversion. The conversion is done as if the <tt>value</tt> had been
3212 stored to memory and read back as type <tt>ty2</tt>. Pointer types may only be
3213 converted to other pointer types with this instruction. To convert pointers to
3214 other types, use the <a href="#i_inttoptr">inttoptr</a> or
3215 <a href="#i_ptrtoint">ptrtoint</a> instructions first.</p>
3219 %X = bitcast i8 255 to i8 <i>; yields i8 :-1</i>
3220 %Y = bitcast i32* %x to sint* <i>; yields sint*:%x</i>
3221 %Z = bitcast <2xint> %V to i64; <i>; yields i64: %V</i>
3225 <!-- ======================================================================= -->
3226 <div class="doc_subsection"> <a name="otherops">Other Operations</a> </div>
3227 <div class="doc_text">
3228 <p>The instructions in this category are the "miscellaneous"
3229 instructions, which defy better classification.</p>
3232 <!-- _______________________________________________________________________ -->
3233 <div class="doc_subsubsection"><a name="i_icmp">'<tt>icmp</tt>' Instruction</a>
3235 <div class="doc_text">
3237 <pre> <result> = icmp <cond> <ty> <var1>, <var2>
3238 <i>; yields {i1}:result</i>
3241 <p>The '<tt>icmp</tt>' instruction returns a boolean value based on comparison
3242 of its two integer operands.</p>
3244 <p>The '<tt>icmp</tt>' instruction takes three operands. The first operand is
3245 the condition code which indicates the kind of comparison to perform. It is not
3246 a value, just a keyword. The possibilities for the condition code are:
3248 <li><tt>eq</tt>: equal</li>
3249 <li><tt>ne</tt>: not equal </li>
3250 <li><tt>ugt</tt>: unsigned greater than</li>
3251 <li><tt>uge</tt>: unsigned greater or equal</li>
3252 <li><tt>ult</tt>: unsigned less than</li>
3253 <li><tt>ule</tt>: unsigned less or equal</li>
3254 <li><tt>sgt</tt>: signed greater than</li>
3255 <li><tt>sge</tt>: signed greater or equal</li>
3256 <li><tt>slt</tt>: signed less than</li>
3257 <li><tt>sle</tt>: signed less or equal</li>
3259 <p>The remaining two arguments must be <a href="#t_integer">integer</a> or
3260 <a href="#t_pointer">pointer</a> typed. They must also be identical types.</p>
3262 <p>The '<tt>icmp</tt>' compares <tt>var1</tt> and <tt>var2</tt> according to
3263 the condition code given as <tt>cond</tt>. The comparison performed always
3264 yields a <a href="#t_primitive">i1</a> result, as follows:
3266 <li><tt>eq</tt>: yields <tt>true</tt> if the operands are equal,
3267 <tt>false</tt> otherwise. No sign interpretation is necessary or performed.
3269 <li><tt>ne</tt>: yields <tt>true</tt> if the operands are unequal,
3270 <tt>false</tt> otherwise. No sign interpretation is necessary or performed.
3271 <li><tt>ugt</tt>: interprets the operands as unsigned values and yields
3272 <tt>true</tt> if <tt>var1</tt> is greater than <tt>var2</tt>.</li>
3273 <li><tt>uge</tt>: interprets the operands as unsigned values and yields
3274 <tt>true</tt> if <tt>var1</tt> is greater than or equal to <tt>var2</tt>.</li>
3275 <li><tt>ult</tt>: interprets the operands as unsigned values and yields
3276 <tt>true</tt> if <tt>var1</tt> is less than <tt>var2</tt>.</li>
3277 <li><tt>ule</tt>: interprets the operands as unsigned values and yields
3278 <tt>true</tt> if <tt>var1</tt> is less than or equal to <tt>var2</tt>.</li>
3279 <li><tt>sgt</tt>: interprets the operands as signed values and yields
3280 <tt>true</tt> if <tt>var1</tt> is greater than <tt>var2</tt>.</li>
3281 <li><tt>sge</tt>: interprets the operands as signed values and yields
3282 <tt>true</tt> if <tt>var1</tt> is greater than or equal to <tt>var2</tt>.</li>
3283 <li><tt>slt</tt>: interprets the operands as signed values and yields
3284 <tt>true</tt> if <tt>var1</tt> is less than <tt>var2</tt>.</li>
3285 <li><tt>sle</tt>: interprets the operands as signed values and yields
3286 <tt>true</tt> if <tt>var1</tt> is less than or equal to <tt>var2</tt>.</li>
3288 <p>If the operands are <a href="#t_pointer">pointer</a> typed, the pointer
3289 values are treated as integers and then compared.</p>
3292 <pre> <result> = icmp eq i32 4, 5 <i>; yields: result=false</i>
3293 <result> = icmp ne float* %X, %X <i>; yields: result=false</i>
3294 <result> = icmp ult i16 4, 5 <i>; yields: result=true</i>
3295 <result> = icmp sgt i16 4, 5 <i>; yields: result=false</i>
3296 <result> = icmp ule i16 -4, 5 <i>; yields: result=false</i>
3297 <result> = icmp sge i16 4, 5 <i>; yields: result=false</i>
3301 <!-- _______________________________________________________________________ -->
3302 <div class="doc_subsubsection"><a name="i_fcmp">'<tt>fcmp</tt>' Instruction</a>
3304 <div class="doc_text">
3306 <pre> <result> = fcmp <cond> <ty> <var1>, <var2>
3307 <i>; yields {i1}:result</i>
3310 <p>The '<tt>fcmp</tt>' instruction returns a boolean value based on comparison
3311 of its floating point operands.</p>
3313 <p>The '<tt>fcmp</tt>' instruction takes three operands. The first operand is
3314 the condition code which indicates the kind of comparison to perform. It is not
3315 a value, just a keyword. The possibilities for the condition code are:
3317 <li><tt>false</tt>: no comparison, always returns false</li>
3318 <li><tt>oeq</tt>: ordered and equal</li>
3319 <li><tt>ogt</tt>: ordered and greater than </li>
3320 <li><tt>oge</tt>: ordered and greater than or equal</li>
3321 <li><tt>olt</tt>: ordered and less than </li>
3322 <li><tt>ole</tt>: ordered and less than or equal</li>
3323 <li><tt>one</tt>: ordered and not equal</li>
3324 <li><tt>ord</tt>: ordered (no nans)</li>
3325 <li><tt>ueq</tt>: unordered or equal</li>
3326 <li><tt>ugt</tt>: unordered or greater than </li>
3327 <li><tt>uge</tt>: unordered or greater than or equal</li>
3328 <li><tt>ult</tt>: unordered or less than </li>
3329 <li><tt>ule</tt>: unordered or less than or equal</li>
3330 <li><tt>une</tt>: unordered or not equal</li>
3331 <li><tt>uno</tt>: unordered (either nans)</li>
3332 <li><tt>true</tt>: no comparison, always returns true</li>
3334 <p>In the preceding, <i>ordered</i> means that neither operand is a QNAN while
3335 <i>unordered</i> means that either operand may be a QNAN.</p>
3336 <p>The <tt>val1</tt> and <tt>val2</tt> arguments must be
3337 <a href="#t_floating">floating point</a> typed. They must have identical
3339 <p>In the foregoing, <i>ordered</i> means that neither operand is a QNAN and
3340 <i>unordered</i> means that either operand is a QNAN.</p>
3342 <p>The '<tt>fcmp</tt>' compares <tt>var1</tt> and <tt>var2</tt> according to
3343 the condition code given as <tt>cond</tt>. The comparison performed always
3344 yields a <a href="#t_primitive">i1</a> result, as follows:
3346 <li><tt>false</tt>: always yields <tt>false</tt>, regardless of operands.</li>
3347 <li><tt>oeq</tt>: yields <tt>true</tt> if both operands are not a QNAN and
3348 <tt>var1</tt> is equal to <tt>var2</tt>.</li>
3349 <li><tt>ogt</tt>: yields <tt>true</tt> if both operands are not a QNAN and
3350 <tt>var1</tt> is greather than <tt>var2</tt>.</li>
3351 <li><tt>oge</tt>: yields <tt>true</tt> if both operands are not a QNAN and
3352 <tt>var1</tt> is greater than or equal to <tt>var2</tt>.</li>
3353 <li><tt>olt</tt>: yields <tt>true</tt> if both operands are not a QNAN and
3354 <tt>var1</tt> is less than <tt>var2</tt>.</li>
3355 <li><tt>ole</tt>: yields <tt>true</tt> if both operands are not a QNAN and
3356 <tt>var1</tt> is less than or equal to <tt>var2</tt>.</li>
3357 <li><tt>one</tt>: yields <tt>true</tt> if both operands are not a QNAN and
3358 <tt>var1</tt> is not equal to <tt>var2</tt>.</li>
3359 <li><tt>ord</tt>: yields <tt>true</tt> if both operands are not a QNAN.</li>
3360 <li><tt>ueq</tt>: yields <tt>true</tt> if either operand is a QNAN or
3361 <tt>var1</tt> is equal to <tt>var2</tt>.</li>
3362 <li><tt>ugt</tt>: yields <tt>true</tt> if either operand is a QNAN or
3363 <tt>var1</tt> is greater than <tt>var2</tt>.</li>
3364 <li><tt>uge</tt>: yields <tt>true</tt> if either operand is a QNAN or
3365 <tt>var1</tt> is greater than or equal to <tt>var2</tt>.</li>
3366 <li><tt>ult</tt>: yields <tt>true</tt> if either operand is a QNAN or
3367 <tt>var1</tt> is less than <tt>var2</tt>.</li>
3368 <li><tt>ule</tt>: yields <tt>true</tt> if either operand is a QNAN or
3369 <tt>var1</tt> is less than or equal to <tt>var2</tt>.</li>
3370 <li><tt>une</tt>: yields <tt>true</tt> if either operand is a QNAN or
3371 <tt>var1</tt> is not equal to <tt>var2</tt>.</li>
3372 <li><tt>uno</tt>: yields <tt>true</tt> if either operand is a QNAN.</li>
3373 <li><tt>true</tt>: always yields <tt>true</tt>, regardless of operands.</li>
3377 <pre> <result> = fcmp oeq float 4.0, 5.0 <i>; yields: result=false</i>
3378 <result> = icmp one float 4.0, 5.0 <i>; yields: result=true</i>
3379 <result> = icmp olt float 4.0, 5.0 <i>; yields: result=true</i>
3380 <result> = icmp ueq double 1.0, 2.0 <i>; yields: result=false</i>
3384 <!-- _______________________________________________________________________ -->
3385 <div class="doc_subsubsection"> <a name="i_phi">'<tt>phi</tt>'
3386 Instruction</a> </div>
3387 <div class="doc_text">
3389 <pre> <result> = phi <ty> [ <val0>, <label0>], ...<br></pre>
3391 <p>The '<tt>phi</tt>' instruction is used to implement the φ node in
3392 the SSA graph representing the function.</p>
3394 <p>The type of the incoming values are specified with the first type
3395 field. After this, the '<tt>phi</tt>' instruction takes a list of pairs
3396 as arguments, with one pair for each predecessor basic block of the
3397 current block. Only values of <a href="#t_firstclass">first class</a>
3398 type may be used as the value arguments to the PHI node. Only labels
3399 may be used as the label arguments.</p>
3400 <p>There must be no non-phi instructions between the start of a basic
3401 block and the PHI instructions: i.e. PHI instructions must be first in
3404 <p>At runtime, the '<tt>phi</tt>' instruction logically takes on the
3405 value specified by the parameter, depending on which basic block we
3406 came from in the last <a href="#terminators">terminator</a> instruction.</p>
3408 <pre>Loop: ; Infinite loop that counts from 0 on up...<br> %indvar = phi i32 [ 0, %LoopHeader ], [ %nextindvar, %Loop ]<br> %nextindvar = add i32 %indvar, 1<br> br label %Loop<br></pre>
3411 <!-- _______________________________________________________________________ -->
3412 <div class="doc_subsubsection">
3413 <a name="i_select">'<tt>select</tt>' Instruction</a>
3416 <div class="doc_text">
3421 <result> = select i1 <cond>, <ty> <val1>, <ty> <val2> <i>; yields ty</i>
3427 The '<tt>select</tt>' instruction is used to choose one value based on a
3428 condition, without branching.
3435 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.
3441 If the boolean condition evaluates to true, the instruction returns the first
3442 value argument; otherwise, it returns the second value argument.
3448 %X = select i1 true, i8 17, i8 42 <i>; yields i8:17</i>
3453 <!-- _______________________________________________________________________ -->
3454 <div class="doc_subsubsection">
3455 <a name="i_call">'<tt>call</tt>' Instruction</a>
3458 <div class="doc_text">
3462 <result> = [tail] call [<a href="#callingconv">cconv</a>] <ty>* <fnptrval>(<param list>)
3467 <p>The '<tt>call</tt>' instruction represents a simple function call.</p>
3471 <p>This instruction requires several arguments:</p>
3475 <p>The optional "tail" marker indicates whether the callee function accesses
3476 any allocas or varargs in the caller. If the "tail" marker is present, the
3477 function call is eligible for tail call optimization. Note that calls may
3478 be marked "tail" even if they do not occur before a <a
3479 href="#i_ret"><tt>ret</tt></a> instruction.
3482 <p>The optional "cconv" marker indicates which <a href="#callingconv">calling
3483 convention</a> the call should use. If none is specified, the call defaults
3484 to using C calling conventions.
3487 <p>'<tt>ty</tt>': shall be the signature of the pointer to function value
3488 being invoked. The argument types must match the types implied by this
3489 signature. This type can be omitted if the function is not varargs and
3490 if the function type does not return a pointer to a function.</p>
3493 <p>'<tt>fnptrval</tt>': An LLVM value containing a pointer to a function to
3494 be invoked. In most cases, this is a direct function invocation, but
3495 indirect <tt>call</tt>s are just as possible, calling an arbitrary pointer
3496 to function value.</p>
3499 <p>'<tt>function args</tt>': argument list whose types match the
3500 function signature argument types. All arguments must be of
3501 <a href="#t_firstclass">first class</a> type. If the function signature
3502 indicates the function accepts a variable number of arguments, the extra
3503 arguments can be specified.</p>
3509 <p>The '<tt>call</tt>' instruction is used to cause control flow to
3510 transfer to a specified function, with its incoming arguments bound to
3511 the specified values. Upon a '<tt><a href="#i_ret">ret</a></tt>'
3512 instruction in the called function, control flow continues with the
3513 instruction after the function call, and the return value of the
3514 function is bound to the result argument. This is a simpler case of
3515 the <a href="#i_invoke">invoke</a> instruction.</p>
3520 %retval = call i32 %test(i32 %argc)
3521 call i32(i8 *, ...) *%printf(i8 * %msg, i32 12, i8 42);
3522 %X = tail call i32 %foo()
3523 %Y = tail call <a href="#callingconv">fastcc</a> i32 %foo()
3528 <!-- _______________________________________________________________________ -->
3529 <div class="doc_subsubsection">
3530 <a name="i_va_arg">'<tt>va_arg</tt>' Instruction</a>
3533 <div class="doc_text">
3538 <resultval> = va_arg <va_list*> <arglist>, <argty>
3543 <p>The '<tt>va_arg</tt>' instruction is used to access arguments passed through
3544 the "variable argument" area of a function call. It is used to implement the
3545 <tt>va_arg</tt> macro in C.</p>
3549 <p>This instruction takes a <tt>va_list*</tt> value and the type of
3550 the argument. It returns a value of the specified argument type and
3551 increments the <tt>va_list</tt> to point to the next argument. Again, the
3552 actual type of <tt>va_list</tt> is target specific.</p>
3556 <p>The '<tt>va_arg</tt>' instruction loads an argument of the specified
3557 type from the specified <tt>va_list</tt> and causes the
3558 <tt>va_list</tt> to point to the next argument. For more information,
3559 see the variable argument handling <a href="#int_varargs">Intrinsic
3562 <p>It is legal for this instruction to be called in a function which does not
3563 take a variable number of arguments, for example, the <tt>vfprintf</tt>
3566 <p><tt>va_arg</tt> is an LLVM instruction instead of an <a
3567 href="#intrinsics">intrinsic function</a> because it takes a type as an
3572 <p>See the <a href="#int_varargs">variable argument processing</a> section.</p>
3576 <!-- *********************************************************************** -->
3577 <div class="doc_section"> <a name="intrinsics">Intrinsic Functions</a> </div>
3578 <!-- *********************************************************************** -->
3580 <div class="doc_text">
3582 <p>LLVM supports the notion of an "intrinsic function". These functions have
3583 well known names and semantics and are required to follow certain restrictions.
3584 Overall, these intrinsics represent an extension mechanism for the LLVM
3585 language that does not require changing all of the transformations in LLVM to
3586 add to the language (or the bytecode reader/writer, the parser,
3589 <p>Intrinsic function names must all start with an "<tt>llvm.</tt>" prefix. This
3590 prefix is reserved in LLVM for intrinsic names; thus, functions may not be named
3591 this. Intrinsic functions must always be external functions: you cannot define
3592 the body of intrinsic functions. Intrinsic functions may only be used in call
3593 or invoke instructions: it is illegal to take the address of an intrinsic
3594 function. Additionally, because intrinsic functions are part of the LLVM
3595 language, it is required that they all be documented here if any are added.</p>
3597 <p>Some intrinsic functions can be overloaded. That is, the intrinsic represents
3598 a family of functions that perform the same operation but on different data
3599 types. This is most frequent with the integer types. Since LLVM can represent
3600 over 8 million different integer types, there is a way to declare an intrinsic
3601 that can be overloaded based on its arguments. Such intrinsics will have the
3602 names of the arbitrary types encoded into the intrinsic function name, each
3603 preceded by a period. For example, the <tt>llvm.ctpop</tt> function can take an
3604 integer of any width. This leads to a family of functions such as
3605 <tt>i32 @llvm.ctpop.i8(i8 %val)</tt> and <tt>i32 @llvm.ctpop.i29(i29 %val)</tt>.
3609 <p>To learn how to add an intrinsic function, please see the
3610 <a href="ExtendingLLVM.html">Extending LLVM Guide</a>.
3615 <!-- ======================================================================= -->
3616 <div class="doc_subsection">
3617 <a name="int_varargs">Variable Argument Handling Intrinsics</a>
3620 <div class="doc_text">
3622 <p>Variable argument support is defined in LLVM with the <a
3623 href="#i_va_arg"><tt>va_arg</tt></a> instruction and these three
3624 intrinsic functions. These functions are related to the similarly
3625 named macros defined in the <tt><stdarg.h></tt> header file.</p>
3627 <p>All of these functions operate on arguments that use a
3628 target-specific value type "<tt>va_list</tt>". The LLVM assembly
3629 language reference manual does not define what this type is, so all
3630 transformations should be prepared to handle intrinsics with any type
3633 <p>This example shows how the <a href="#i_va_arg"><tt>va_arg</tt></a>
3634 instruction and the variable argument handling intrinsic functions are
3638 define i32 @test(i32 %X, ...) {
3639 ; Initialize variable argument processing
3641 %ap2 = bitcast i8** %ap to i8*
3642 call void @llvm.va_start(i8* %ap2)
3644 ; Read a single integer argument
3645 %tmp = va_arg i8 ** %ap, i32
3647 ; Demonstrate usage of llvm.va_copy and llvm.va_end
3649 %aq2 = bitcast i8** %aq to i8*
3650 call void @llvm.va_copy(i8 *%aq2, i8* %ap2)
3651 call void @llvm.va_end(i8* %aq2)
3653 ; Stop processing of arguments.
3654 call void @llvm.va_end(i8* %ap2)
3658 declare void @llvm.va_start(i8*)
3659 declare void @llvm.va_copy(i8*, i8*)
3660 declare void @llvm.va_end(i8*)
3664 <!-- _______________________________________________________________________ -->
3665 <div class="doc_subsubsection">
3666 <a name="i_va_start">'<tt>llvm.va_start</tt>' Intrinsic</a>
3670 <div class="doc_text">
3672 <pre> declare void %llvm.va_start(i8* <arglist>)<br></pre>
3674 <P>The '<tt>llvm.va_start</tt>' intrinsic initializes
3675 <tt>*<arglist></tt> for subsequent use by <tt><a
3676 href="#i_va_arg">va_arg</a></tt>.</p>
3680 <P>The argument is a pointer to a <tt>va_list</tt> element to initialize.</p>
3684 <P>The '<tt>llvm.va_start</tt>' intrinsic works just like the <tt>va_start</tt>
3685 macro available in C. In a target-dependent way, it initializes the
3686 <tt>va_list</tt> element the argument points to, so that the next call to
3687 <tt>va_arg</tt> will produce the first variable argument passed to the function.
3688 Unlike the C <tt>va_start</tt> macro, this intrinsic does not need to know the
3689 last argument of the function, the compiler can figure that out.</p>
3693 <!-- _______________________________________________________________________ -->
3694 <div class="doc_subsubsection">
3695 <a name="i_va_end">'<tt>llvm.va_end</tt>' Intrinsic</a>
3698 <div class="doc_text">
3700 <pre> declare void @llvm.va_end(i8* <arglist>)<br></pre>
3703 <p>The '<tt>llvm.va_end</tt>' intrinsic destroys <tt><arglist></tt>
3704 which has been initialized previously with <tt><a href="#i_va_start">llvm.va_start</a></tt>
3705 or <tt><a href="#i_va_copy">llvm.va_copy</a></tt>.</p>
3709 <p>The argument is a <tt>va_list</tt> to destroy.</p>
3713 <p>The '<tt>llvm.va_end</tt>' intrinsic works just like the <tt>va_end</tt>
3714 macro available in C. In a target-dependent way, it destroys the <tt>va_list</tt>.
3715 Calls to <a href="#i_va_start"><tt>llvm.va_start</tt></a> and <a
3716 href="#i_va_copy"><tt>llvm.va_copy</tt></a> must be matched exactly
3717 with calls to <tt>llvm.va_end</tt>.</p>
3721 <!-- _______________________________________________________________________ -->
3722 <div class="doc_subsubsection">
3723 <a name="i_va_copy">'<tt>llvm.va_copy</tt>' Intrinsic</a>
3726 <div class="doc_text">
3731 declare void @llvm.va_copy(i8* <destarglist>, i8* <srcarglist>)
3736 <p>The '<tt>llvm.va_copy</tt>' intrinsic copies the current argument position from
3737 the source argument list to the destination argument list.</p>
3741 <p>The first argument is a pointer to a <tt>va_list</tt> element to initialize.
3742 The second argument is a pointer to a <tt>va_list</tt> element to copy from.</p>
3747 <p>The '<tt>llvm.va_copy</tt>' intrinsic works just like the <tt>va_copy</tt> macro
3748 available in C. In a target-dependent way, it copies the source
3749 <tt>va_list</tt> element into the destination list. This intrinsic is necessary
3750 because the <tt><a href="#i_va_start">llvm.va_start</a></tt> intrinsic may be
3751 arbitrarily complex and require memory allocation, for example.</p>
3755 <!-- ======================================================================= -->
3756 <div class="doc_subsection">
3757 <a name="int_gc">Accurate Garbage Collection Intrinsics</a>
3760 <div class="doc_text">
3763 LLVM support for <a href="GarbageCollection.html">Accurate Garbage
3764 Collection</a> requires the implementation and generation of these intrinsics.
3765 These intrinsics allow identification of <a href="#i_gcroot">GC roots on the
3766 stack</a>, as well as garbage collector implementations that require <a
3767 href="#i_gcread">read</a> and <a href="#i_gcwrite">write</a> barriers.
3768 Front-ends for type-safe garbage collected languages should generate these
3769 intrinsics to make use of the LLVM garbage collectors. For more details, see <a
3770 href="GarbageCollection.html">Accurate Garbage Collection with LLVM</a>.
3774 <!-- _______________________________________________________________________ -->
3775 <div class="doc_subsubsection">
3776 <a name="i_gcroot">'<tt>llvm.gcroot</tt>' Intrinsic</a>
3779 <div class="doc_text">
3784 declare void @llvm.gcroot(<ty>** %ptrloc, <ty2>* %metadata)
3789 <p>The '<tt>llvm.gcroot</tt>' intrinsic declares the existence of a GC root to
3790 the code generator, and allows some metadata to be associated with it.</p>
3794 <p>The first argument specifies the address of a stack object that contains the
3795 root pointer. The second pointer (which must be either a constant or a global
3796 value address) contains the meta-data to be associated with the root.</p>
3800 <p>At runtime, a call to this intrinsics stores a null pointer into the "ptrloc"
3801 location. At compile-time, the code generator generates information to allow
3802 the runtime to find the pointer at GC safe points.
3808 <!-- _______________________________________________________________________ -->
3809 <div class="doc_subsubsection">
3810 <a name="i_gcread">'<tt>llvm.gcread</tt>' Intrinsic</a>
3813 <div class="doc_text">
3818 declare i8 * @llvm.gcread(i8 * %ObjPtr, i8 ** %Ptr)
3823 <p>The '<tt>llvm.gcread</tt>' intrinsic identifies reads of references from heap
3824 locations, allowing garbage collector implementations that require read
3829 <p>The second argument is the address to read from, which should be an address
3830 allocated from the garbage collector. The first object is a pointer to the
3831 start of the referenced object, if needed by the language runtime (otherwise
3836 <p>The '<tt>llvm.gcread</tt>' intrinsic has the same semantics as a load
3837 instruction, but may be replaced with substantially more complex code by the
3838 garbage collector runtime, as needed.</p>
3843 <!-- _______________________________________________________________________ -->
3844 <div class="doc_subsubsection">
3845 <a name="i_gcwrite">'<tt>llvm.gcwrite</tt>' Intrinsic</a>
3848 <div class="doc_text">
3853 declare void @llvm.gcwrite(i8 * %P1, i8 * %Obj, i8 ** %P2)
3858 <p>The '<tt>llvm.gcwrite</tt>' intrinsic identifies writes of references to heap
3859 locations, allowing garbage collector implementations that require write
3860 barriers (such as generational or reference counting collectors).</p>
3864 <p>The first argument is the reference to store, the second is the start of the
3865 object to store it to, and the third is the address of the field of Obj to
3866 store to. If the runtime does not require a pointer to the object, Obj may be
3871 <p>The '<tt>llvm.gcwrite</tt>' intrinsic has the same semantics as a store
3872 instruction, but may be replaced with substantially more complex code by the
3873 garbage collector runtime, as needed.</p>
3879 <!-- ======================================================================= -->
3880 <div class="doc_subsection">
3881 <a name="int_codegen">Code Generator Intrinsics</a>
3884 <div class="doc_text">
3886 These intrinsics are provided by LLVM to expose special features that may only
3887 be implemented with code generator support.
3892 <!-- _______________________________________________________________________ -->
3893 <div class="doc_subsubsection">
3894 <a name="i_returnaddress">'<tt>llvm.returnaddress</tt>' Intrinsic</a>
3897 <div class="doc_text">
3901 declare i8 *@llvm.returnaddress(i32 <level>)
3907 The '<tt>llvm.returnaddress</tt>' intrinsic attempts to compute a
3908 target-specific value indicating the return address of the current function
3909 or one of its callers.
3915 The argument to this intrinsic indicates which function to return the address
3916 for. Zero indicates the calling function, one indicates its caller, etc. The
3917 argument is <b>required</b> to be a constant integer value.
3923 The '<tt>llvm.returnaddress</tt>' intrinsic either returns a pointer indicating
3924 the return address of the specified call frame, or zero if it cannot be
3925 identified. The value returned by this intrinsic is likely to be incorrect or 0
3926 for arguments other than zero, so it should only be used for debugging purposes.
3930 Note that calling this intrinsic does not prevent function inlining or other
3931 aggressive transformations, so the value returned may not be that of the obvious
3932 source-language caller.
3937 <!-- _______________________________________________________________________ -->
3938 <div class="doc_subsubsection">
3939 <a name="i_frameaddress">'<tt>llvm.frameaddress</tt>' Intrinsic</a>
3942 <div class="doc_text">
3946 declare i8 *@llvm.frameaddress(i32 <level>)
3952 The '<tt>llvm.frameaddress</tt>' intrinsic attempts to return the
3953 target-specific frame pointer value for the specified stack frame.
3959 The argument to this intrinsic indicates which function to return the frame
3960 pointer for. Zero indicates the calling function, one indicates its caller,
3961 etc. The argument is <b>required</b> to be a constant integer value.
3967 The '<tt>llvm.frameaddress</tt>' intrinsic either returns a pointer indicating
3968 the frame address of the specified call frame, or zero if it cannot be
3969 identified. The value returned by this intrinsic is likely to be incorrect or 0
3970 for arguments other than zero, so it should only be used for debugging purposes.
3974 Note that calling this intrinsic does not prevent function inlining or other
3975 aggressive transformations, so the value returned may not be that of the obvious
3976 source-language caller.
3980 <!-- _______________________________________________________________________ -->
3981 <div class="doc_subsubsection">
3982 <a name="i_stacksave">'<tt>llvm.stacksave</tt>' Intrinsic</a>
3985 <div class="doc_text">
3989 declare i8 *@llvm.stacksave()
3995 The '<tt>llvm.stacksave</tt>' intrinsic is used to remember the current state of
3996 the function stack, for use with <a href="#i_stackrestore">
3997 <tt>llvm.stackrestore</tt></a>. This is useful for implementing language
3998 features like scoped automatic variable sized arrays in C99.
4004 This intrinsic returns a opaque pointer value that can be passed to <a
4005 href="#i_stackrestore"><tt>llvm.stackrestore</tt></a>. When an
4006 <tt>llvm.stackrestore</tt> intrinsic is executed with a value saved from
4007 <tt>llvm.stacksave</tt>, it effectively restores the state of the stack to the
4008 state it was in when the <tt>llvm.stacksave</tt> intrinsic executed. In
4009 practice, this pops any <a href="#i_alloca">alloca</a> blocks from the stack
4010 that were allocated after the <tt>llvm.stacksave</tt> was executed.
4015 <!-- _______________________________________________________________________ -->
4016 <div class="doc_subsubsection">
4017 <a name="i_stackrestore">'<tt>llvm.stackrestore</tt>' Intrinsic</a>
4020 <div class="doc_text">
4024 declare void @llvm.stackrestore(i8 * %ptr)
4030 The '<tt>llvm.stackrestore</tt>' intrinsic is used to restore the state of
4031 the function stack to the state it was in when the corresponding <a
4032 href="#i_stacksave"><tt>llvm.stacksave</tt></a> intrinsic executed. This is
4033 useful for implementing language features like scoped automatic variable sized
4040 See the description for <a href="#i_stacksave"><tt>llvm.stacksave</tt></a>.
4046 <!-- _______________________________________________________________________ -->
4047 <div class="doc_subsubsection">
4048 <a name="i_prefetch">'<tt>llvm.prefetch</tt>' Intrinsic</a>
4051 <div class="doc_text">
4055 declare void @llvm.prefetch(i8 * <address>,
4056 i32 <rw>, i32 <locality>)
4063 The '<tt>llvm.prefetch</tt>' intrinsic is a hint to the code generator to insert
4064 a prefetch instruction if supported; otherwise, it is a noop. Prefetches have
4066 effect on the behavior of the program but can change its performance
4073 <tt>address</tt> is the address to be prefetched, <tt>rw</tt> is the specifier
4074 determining if the fetch should be for a read (0) or write (1), and
4075 <tt>locality</tt> is a temporal locality specifier ranging from (0) - no
4076 locality, to (3) - extremely local keep in cache. The <tt>rw</tt> and
4077 <tt>locality</tt> arguments must be constant integers.
4083 This intrinsic does not modify the behavior of the program. In particular,
4084 prefetches cannot trap and do not produce a value. On targets that support this
4085 intrinsic, the prefetch can provide hints to the processor cache for better
4091 <!-- _______________________________________________________________________ -->
4092 <div class="doc_subsubsection">
4093 <a name="i_pcmarker">'<tt>llvm.pcmarker</tt>' Intrinsic</a>
4096 <div class="doc_text">
4100 declare void @llvm.pcmarker( i32 <id> )
4107 The '<tt>llvm.pcmarker</tt>' intrinsic is a method to export a Program Counter
4109 code to simulators and other tools. The method is target specific, but it is
4110 expected that the marker will use exported symbols to transmit the PC of the marker.
4111 The marker makes no guarantees that it will remain with any specific instruction
4112 after optimizations. It is possible that the presence of a marker will inhibit
4113 optimizations. The intended use is to be inserted after optimizations to allow
4114 correlations of simulation runs.
4120 <tt>id</tt> is a numerical id identifying the marker.
4126 This intrinsic does not modify the behavior of the program. Backends that do not
4127 support this intrinisic may ignore it.
4132 <!-- _______________________________________________________________________ -->
4133 <div class="doc_subsubsection">
4134 <a name="i_readcyclecounter">'<tt>llvm.readcyclecounter</tt>' Intrinsic</a>
4137 <div class="doc_text">
4141 declare i64 @llvm.readcyclecounter( )
4148 The '<tt>llvm.readcyclecounter</tt>' intrinsic provides access to the cycle
4149 counter register (or similar low latency, high accuracy clocks) on those targets
4150 that support it. On X86, it should map to RDTSC. On Alpha, it should map to RPCC.
4151 As the backing counters overflow quickly (on the order of 9 seconds on alpha), this
4152 should only be used for small timings.
4158 When directly supported, reading the cycle counter should not modify any memory.
4159 Implementations are allowed to either return a application specific value or a
4160 system wide value. On backends without support, this is lowered to a constant 0.
4165 <!-- ======================================================================= -->
4166 <div class="doc_subsection">
4167 <a name="int_libc">Standard C Library Intrinsics</a>
4170 <div class="doc_text">
4172 LLVM provides intrinsics for a few important standard C library functions.
4173 These intrinsics allow source-language front-ends to pass information about the
4174 alignment of the pointer arguments to the code generator, providing opportunity
4175 for more efficient code generation.
4180 <!-- _______________________________________________________________________ -->
4181 <div class="doc_subsubsection">
4182 <a name="i_memcpy">'<tt>llvm.memcpy</tt>' Intrinsic</a>
4185 <div class="doc_text">
4189 declare void @llvm.memcpy.i32(i8 * <dest>, i8 * <src>,
4190 i32 <len>, i32 <align>)
4191 declare void @llvm.memcpy.i64(i8 * <dest>, i8 * <src>,
4192 i64 <len>, i32 <align>)
4198 The '<tt>llvm.memcpy.*</tt>' intrinsics copy a block of memory from the source
4199 location to the destination location.
4203 Note that, unlike the standard libc function, the <tt>llvm.memcpy.*</tt>
4204 intrinsics do not return a value, and takes an extra alignment argument.
4210 The first argument is a pointer to the destination, the second is a pointer to
4211 the source. The third argument is an integer argument
4212 specifying the number of bytes to copy, and the fourth argument is the alignment
4213 of the source and destination locations.
4217 If the call to this intrinisic has an alignment value that is not 0 or 1, then
4218 the caller guarantees that both the source and destination pointers are aligned
4225 The '<tt>llvm.memcpy.*</tt>' intrinsics copy a block of memory from the source
4226 location to the destination location, which are not allowed to overlap. It
4227 copies "len" bytes of memory over. If the argument is known to be aligned to
4228 some boundary, this can be specified as the fourth argument, otherwise it should
4234 <!-- _______________________________________________________________________ -->
4235 <div class="doc_subsubsection">
4236 <a name="i_memmove">'<tt>llvm.memmove</tt>' Intrinsic</a>
4239 <div class="doc_text">
4243 declare void @llvm.memmove.i32(i8 * <dest>, i8 * <src>,
4244 i32 <len>, i32 <align>)
4245 declare void @llvm.memmove.i64(i8 * <dest>, i8 * <src>,
4246 i64 <len>, i32 <align>)
4252 The '<tt>llvm.memmove.*</tt>' intrinsics move a block of memory from the source
4253 location to the destination location. It is similar to the
4254 '<tt>llvm.memcmp</tt>' intrinsic but allows the two memory locations to overlap.
4258 Note that, unlike the standard libc function, the <tt>llvm.memmove.*</tt>
4259 intrinsics do not return a value, and takes an extra alignment argument.
4265 The first argument is a pointer to the destination, the second is a pointer to
4266 the source. The third argument is an integer argument
4267 specifying the number of bytes to copy, and the fourth argument is the alignment
4268 of the source and destination locations.
4272 If the call to this intrinisic has an alignment value that is not 0 or 1, then
4273 the caller guarantees that the source and destination pointers are aligned to
4280 The '<tt>llvm.memmove.*</tt>' intrinsics copy a block of memory from the source
4281 location to the destination location, which may overlap. It
4282 copies "len" bytes of memory over. If the argument is known to be aligned to
4283 some boundary, this can be specified as the fourth argument, otherwise it should
4289 <!-- _______________________________________________________________________ -->
4290 <div class="doc_subsubsection">
4291 <a name="i_memset">'<tt>llvm.memset.*</tt>' Intrinsics</a>
4294 <div class="doc_text">
4298 declare void @llvm.memset.i32(i8 * <dest>, i8 <val>,
4299 i32 <len>, i32 <align>)
4300 declare void @llvm.memset.i64(i8 * <dest>, i8 <val>,
4301 i64 <len>, i32 <align>)
4307 The '<tt>llvm.memset.*</tt>' intrinsics fill a block of memory with a particular
4312 Note that, unlike the standard libc function, the <tt>llvm.memset</tt> intrinsic
4313 does not return a value, and takes an extra alignment argument.
4319 The first argument is a pointer to the destination to fill, the second is the
4320 byte value to fill it with, the third argument is an integer
4321 argument specifying the number of bytes to fill, and the fourth argument is the
4322 known alignment of destination location.
4326 If the call to this intrinisic has an alignment value that is not 0 or 1, then
4327 the caller guarantees that the destination pointer is aligned to that boundary.
4333 The '<tt>llvm.memset.*</tt>' intrinsics fill "len" bytes of memory starting at
4335 destination location. If the argument is known to be aligned to some boundary,
4336 this can be specified as the fourth argument, otherwise it should be set to 0 or
4342 <!-- _______________________________________________________________________ -->
4343 <div class="doc_subsubsection">
4344 <a name="i_sqrt">'<tt>llvm.sqrt.*</tt>' Intrinsic</a>
4347 <div class="doc_text">
4351 declare float @llvm.sqrt.f32(float %Val)
4352 declare double @llvm.sqrt.f64(double %Val)
4358 The '<tt>llvm.sqrt</tt>' intrinsics return the sqrt of the specified operand,
4359 returning the same value as the libm '<tt>sqrt</tt>' function would. Unlike
4360 <tt>sqrt</tt> in libm, however, <tt>llvm.sqrt</tt> has undefined behavior for
4361 negative numbers (which allows for better optimization).
4367 The argument and return value are floating point numbers of the same type.
4373 This function returns the sqrt of the specified operand if it is a positive
4374 floating point number.
4378 <!-- _______________________________________________________________________ -->
4379 <div class="doc_subsubsection">
4380 <a name="i_powi">'<tt>llvm.powi.*</tt>' Intrinsic</a>
4383 <div class="doc_text">
4387 declare float @llvm.powi.f32(float %Val, i32 %power)
4388 declare double @llvm.powi.f64(double %Val, i32 %power)
4394 The '<tt>llvm.powi.*</tt>' intrinsics return the first operand raised to the
4395 specified (positive or negative) power. The order of evaluation of
4396 multiplications is not defined.
4402 The second argument is an integer power, and the first is a value to raise to
4409 This function returns the first value raised to the second power with an
4410 unspecified sequence of rounding operations.</p>
4414 <!-- ======================================================================= -->
4415 <div class="doc_subsection">
4416 <a name="int_manip">Bit Manipulation Intrinsics</a>
4419 <div class="doc_text">
4421 LLVM provides intrinsics for a few important bit manipulation operations.
4422 These allow efficient code generation for some algorithms.
4427 <!-- _______________________________________________________________________ -->
4428 <div class="doc_subsubsection">
4429 <a name="i_bswap">'<tt>llvm.bswap.*</tt>' Intrinsics</a>
4432 <div class="doc_text">
4435 <p>This is an overloaded intrinsic function. You can use bswap on any integer
4436 type that is an even number of bytes (i.e. BitWidth % 16 == 0). Note the suffix
4437 that includes the type for the result and the operand.
4439 declare i16 @llvm.bswap.i16.i16(i16 <id>)
4440 declare i32 @llvm.bswap.i32.i32(i32 <id>)
4441 declare i64 @llvm.bswap.i64.i32(i64 <id>)
4447 The '<tt>llvm.bwsap</tt>' family of intrinsics is used to byteswap integer
4448 values with an even number of bytes (positive multiple of 16 bits). These are
4449 useful for performing operations on data that is not in the target's native
4456 The <tt>llvm.bswap.16.i16</tt> intrinsic returns an i16 value that has the high
4457 and low byte of the input i16 swapped. Similarly, the <tt>llvm.bswap.i32</tt>
4458 intrinsic returns an i32 value that has the four bytes of the input i32
4459 swapped, so that if the input bytes are numbered 0, 1, 2, 3 then the returned
4460 i32 will have its bytes in 3, 2, 1, 0 order. The <tt>llvm.bswap.i48.i48</tt>,
4461 <tt>llvm.bswap.i64.i64</tt> and other intrinsics extend this concept to
4462 additional even-byte lengths (6 bytes, 8 bytes and more, respectively).
4467 <!-- _______________________________________________________________________ -->
4468 <div class="doc_subsubsection">
4469 <a name="int_ctpop">'<tt>llvm.ctpop.*</tt>' Intrinsic</a>
4472 <div class="doc_text">
4475 <p>This is an overloaded intrinsic. You can use llvm.ctpop on any integer bit
4476 width. Not all targets support all bit widths however.
4478 declare i32 @llvm.ctpop.i8 (i8 <src>)
4479 declare i32 @llvm.ctpop.i16(i16 <src>)
4480 declare i32 @llvm.ctpop.i32(i32 <src>)
4481 declare i32 @llvm.ctpop.i64(i64 <src>)
4482 declare i32 @llvm.ctpop.i256(i256 <src>)
4488 The '<tt>llvm.ctpop</tt>' family of intrinsics counts the number of bits set in a
4495 The only argument is the value to be counted. The argument may be of any
4496 integer type. The return type must match the argument type.
4502 The '<tt>llvm.ctpop</tt>' intrinsic counts the 1's in a variable.
4506 <!-- _______________________________________________________________________ -->
4507 <div class="doc_subsubsection">
4508 <a name="int_ctlz">'<tt>llvm.ctlz.*</tt>' Intrinsic</a>
4511 <div class="doc_text">
4514 <p>This is an overloaded intrinsic. You can use <tt>llvm.ctlz</tt> on any
4515 integer bit width. Not all targets support all bit widths however.
4517 declare i32 @llvm.ctlz.i8 (i8 <src>)
4518 declare i32 @llvm.ctlz.i16(i16 <src>)
4519 declare i32 @llvm.ctlz.i32(i32 <src>)
4520 declare i32 @llvm.ctlz.i64(i64 <src>)
4521 declare i32 @llvm.ctlz.i256(i256 <src>)
4527 The '<tt>llvm.ctlz</tt>' family of intrinsic functions counts the number of
4528 leading zeros in a variable.
4534 The only argument is the value to be counted. The argument may be of any
4535 integer type. The return type must match the argument type.
4541 The '<tt>llvm.ctlz</tt>' intrinsic counts the leading (most significant) zeros
4542 in a variable. If the src == 0 then the result is the size in bits of the type
4543 of src. For example, <tt>llvm.ctlz(i32 2) = 30</tt>.
4549 <!-- _______________________________________________________________________ -->
4550 <div class="doc_subsubsection">
4551 <a name="int_cttz">'<tt>llvm.cttz.*</tt>' Intrinsic</a>
4554 <div class="doc_text">
4557 <p>This is an overloaded intrinsic. You can use <tt>llvm.cttz</tt> on any
4558 integer bit width. Not all targets support all bit widths however.
4560 declare i32 @llvm.cttz.i8 (i8 <src>)
4561 declare i32 @llvm.cttz.i16(i16 <src>)
4562 declare i32 @llvm.cttz.i32(i32 <src>)
4563 declare i32 @llvm.cttz.i64(i64 <src>)
4564 declare i32 @llvm.cttz.i256(i256 <src>)
4570 The '<tt>llvm.cttz</tt>' family of intrinsic functions counts the number of
4577 The only argument is the value to be counted. The argument may be of any
4578 integer type. The return type must match the argument type.
4584 The '<tt>llvm.cttz</tt>' intrinsic counts the trailing (least significant) zeros
4585 in a variable. If the src == 0 then the result is the size in bits of the type
4586 of src. For example, <tt>llvm.cttz(2) = 1</tt>.
4590 <!-- _______________________________________________________________________ -->
4591 <div class="doc_subsubsection">
4592 <a name="int_cttz">'<tt>llvm.bit.concat.*</tt>' Intrinsic</a>
4595 <div class="doc_text">
4598 <p>This is an overloaded intrinsic. You can use <tt>llvm.bit.concat</tt> on any
4601 declare i32 @llvm.bit.concat.i32.i17.i15 (i17 %hi, i15 %lo)
4602 declare i29 @llvm.bit.concat.i29(i16 %lo, i13 %lo)
4607 The '<tt>llvm.bit.concat</tt>' family of intrinsic functions concatenates two
4608 integer values to produce a longer one.
4614 The two arguments may be any bit width. The result must be an integer whose bit
4615 width is the sum of the arguments' bit widths. The first argument represents the
4616 bits that will occupy the high order bit locations in the concatenated result.
4617 THe second argument will occupy the lower order bit locations in the result.
4623 The '<tt>llvm.bit.concat</tt>' intrinsic is the equivalent of two <tt>zext</tt>
4624 instructions, a <tt>shl</tt> and an <tt>or</tt>. This sequence can be
4625 implemented in hardware so this intrinsic assists with recognizing the sequence
4626 for code generation purposes. The operation proceeds as follows:</p>
4628 <li>Each of the arguments is <tt>zext</tt>'d to the result bit width.</li>
4629 <li>The <tt>%hi</tt> argument is shift left by the width of the <tt>%lo</tt>
4630 argument (shifted into to high order bits).</li>
4631 <li>The shifted <tt>%hi</tt> value and <tt>%lo</tt> are <tt>or</tt>'d together
4632 to form the result.</li>
4636 <!-- ======================================================================= -->
4637 <div class="doc_subsection">
4638 <a name="int_debugger">Debugger Intrinsics</a>
4641 <div class="doc_text">
4643 The LLVM debugger intrinsics (which all start with <tt>llvm.dbg.</tt> prefix),
4644 are described in the <a
4645 href="SourceLevelDebugging.html#format_common_intrinsics">LLVM Source Level
4646 Debugging</a> document.
4651 <!-- ======================================================================= -->
4652 <div class="doc_subsection">
4653 <a name="int_eh">Exception Handling Intrinsics</a>
4656 <div class="doc_text">
4657 <p> The LLVM exception handling intrinsics (which all start with
4658 <tt>llvm.eh.</tt> prefix), are described in the <a
4659 href="ExceptionHandling.html#format_common_intrinsics">LLVM Exception
4660 Handling</a> document. </p>
4664 <!-- *********************************************************************** -->
4667 <a href="http://jigsaw.w3.org/css-validator/check/referer"><img
4668 src="http://jigsaw.w3.org/css-validator/images/vcss" alt="Valid CSS!"></a>
4669 <a href="http://validator.w3.org/check/referer"><img
4670 src="http://www.w3.org/Icons/valid-html401" alt="Valid HTML 4.01!" /></a>
4672 <a href="mailto:sabre@nondot.org">Chris Lattner</a><br>
4673 <a href="http://llvm.org">The LLVM Compiler Infrastructure</a><br>
4674 Last modified: $Date$