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="#aliasstructure">Aliases</a>
28 <li><a href="#paramattrs">Parameter Attributes</a></li>
29 <li><a href="#moduleasm">Module-Level Inline Assembly</a></li>
30 <li><a href="#datalayout">Data Layout</a></li>
33 <li><a href="#typesystem">Type System</a>
35 <li><a href="#t_primitive">Primitive Types</a>
37 <li><a href="#t_classifications">Type Classifications</a></li>
40 <li><a href="#t_derived">Derived Types</a>
42 <li><a href="#t_array">Array Type</a></li>
43 <li><a href="#t_function">Function Type</a></li>
44 <li><a href="#t_pointer">Pointer Type</a></li>
45 <li><a href="#t_struct">Structure Type</a></li>
46 <li><a href="#t_pstruct">Packed Structure Type</a></li>
47 <li><a href="#t_vector">Vector Type</a></li>
48 <li><a href="#t_opaque">Opaque Type</a></li>
53 <li><a href="#constants">Constants</a>
55 <li><a href="#simpleconstants">Simple Constants</a>
56 <li><a href="#aggregateconstants">Aggregate Constants</a>
57 <li><a href="#globalconstants">Global Variable and Function Addresses</a>
58 <li><a href="#undefvalues">Undefined Values</a>
59 <li><a href="#constantexprs">Constant Expressions</a>
62 <li><a href="#othervalues">Other Values</a>
64 <li><a href="#inlineasm">Inline Assembler Expressions</a>
67 <li><a href="#instref">Instruction Reference</a>
69 <li><a href="#terminators">Terminator Instructions</a>
71 <li><a href="#i_ret">'<tt>ret</tt>' Instruction</a></li>
72 <li><a href="#i_br">'<tt>br</tt>' Instruction</a></li>
73 <li><a href="#i_switch">'<tt>switch</tt>' Instruction</a></li>
74 <li><a href="#i_invoke">'<tt>invoke</tt>' Instruction</a></li>
75 <li><a href="#i_unwind">'<tt>unwind</tt>' Instruction</a></li>
76 <li><a href="#i_unreachable">'<tt>unreachable</tt>' Instruction</a></li>
79 <li><a href="#binaryops">Binary Operations</a>
81 <li><a href="#i_add">'<tt>add</tt>' Instruction</a></li>
82 <li><a href="#i_sub">'<tt>sub</tt>' Instruction</a></li>
83 <li><a href="#i_mul">'<tt>mul</tt>' Instruction</a></li>
84 <li><a href="#i_udiv">'<tt>udiv</tt>' Instruction</a></li>
85 <li><a href="#i_sdiv">'<tt>sdiv</tt>' Instruction</a></li>
86 <li><a href="#i_fdiv">'<tt>fdiv</tt>' Instruction</a></li>
87 <li><a href="#i_urem">'<tt>urem</tt>' Instruction</a></li>
88 <li><a href="#i_srem">'<tt>srem</tt>' Instruction</a></li>
89 <li><a href="#i_frem">'<tt>frem</tt>' Instruction</a></li>
92 <li><a href="#bitwiseops">Bitwise Binary Operations</a>
94 <li><a href="#i_shl">'<tt>shl</tt>' Instruction</a></li>
95 <li><a href="#i_lshr">'<tt>lshr</tt>' Instruction</a></li>
96 <li><a href="#i_ashr">'<tt>ashr</tt>' Instruction</a></li>
97 <li><a href="#i_and">'<tt>and</tt>' Instruction</a></li>
98 <li><a href="#i_or">'<tt>or</tt>' Instruction</a></li>
99 <li><a href="#i_xor">'<tt>xor</tt>' Instruction</a></li>
102 <li><a href="#vectorops">Vector Operations</a>
104 <li><a href="#i_extractelement">'<tt>extractelement</tt>' Instruction</a></li>
105 <li><a href="#i_insertelement">'<tt>insertelement</tt>' Instruction</a></li>
106 <li><a href="#i_shufflevector">'<tt>shufflevector</tt>' Instruction</a></li>
109 <li><a href="#memoryops">Memory Access and Addressing Operations</a>
111 <li><a href="#i_malloc">'<tt>malloc</tt>' Instruction</a></li>
112 <li><a href="#i_free">'<tt>free</tt>' Instruction</a></li>
113 <li><a href="#i_alloca">'<tt>alloca</tt>' Instruction</a></li>
114 <li><a href="#i_load">'<tt>load</tt>' Instruction</a></li>
115 <li><a href="#i_store">'<tt>store</tt>' Instruction</a></li>
116 <li><a href="#i_getelementptr">'<tt>getelementptr</tt>' Instruction</a></li>
119 <li><a href="#convertops">Conversion Operations</a>
121 <li><a href="#i_trunc">'<tt>trunc .. to</tt>' Instruction</a></li>
122 <li><a href="#i_zext">'<tt>zext .. to</tt>' Instruction</a></li>
123 <li><a href="#i_sext">'<tt>sext .. to</tt>' Instruction</a></li>
124 <li><a href="#i_fptrunc">'<tt>fptrunc .. to</tt>' Instruction</a></li>
125 <li><a href="#i_fpext">'<tt>fpext .. to</tt>' Instruction</a></li>
126 <li><a href="#i_fptoui">'<tt>fptoui .. to</tt>' Instruction</a></li>
127 <li><a href="#i_fptosi">'<tt>fptosi .. to</tt>' Instruction</a></li>
128 <li><a href="#i_uitofp">'<tt>uitofp .. to</tt>' Instruction</a></li>
129 <li><a href="#i_sitofp">'<tt>sitofp .. to</tt>' Instruction</a></li>
130 <li><a href="#i_ptrtoint">'<tt>ptrtoint .. to</tt>' Instruction</a></li>
131 <li><a href="#i_inttoptr">'<tt>inttoptr .. to</tt>' Instruction</a></li>
132 <li><a href="#i_bitcast">'<tt>bitcast .. to</tt>' Instruction</a></li>
134 <li><a href="#otherops">Other Operations</a>
136 <li><a href="#i_icmp">'<tt>icmp</tt>' Instruction</a></li>
137 <li><a href="#i_fcmp">'<tt>fcmp</tt>' Instruction</a></li>
138 <li><a href="#i_phi">'<tt>phi</tt>' Instruction</a></li>
139 <li><a href="#i_select">'<tt>select</tt>' Instruction</a></li>
140 <li><a href="#i_call">'<tt>call</tt>' Instruction</a></li>
141 <li><a href="#i_va_arg">'<tt>va_arg</tt>' Instruction</a></li>
146 <li><a href="#intrinsics">Intrinsic Functions</a>
148 <li><a href="#int_varargs">Variable Argument Handling Intrinsics</a>
150 <li><a href="#int_va_start">'<tt>llvm.va_start</tt>' Intrinsic</a></li>
151 <li><a href="#int_va_end">'<tt>llvm.va_end</tt>' Intrinsic</a></li>
152 <li><a href="#int_va_copy">'<tt>llvm.va_copy</tt>' Intrinsic</a></li>
155 <li><a href="#int_gc">Accurate Garbage Collection Intrinsics</a>
157 <li><a href="#int_gcroot">'<tt>llvm.gcroot</tt>' Intrinsic</a></li>
158 <li><a href="#int_gcread">'<tt>llvm.gcread</tt>' Intrinsic</a></li>
159 <li><a href="#int_gcwrite">'<tt>llvm.gcwrite</tt>' Intrinsic</a></li>
162 <li><a href="#int_codegen">Code Generator Intrinsics</a>
164 <li><a href="#int_returnaddress">'<tt>llvm.returnaddress</tt>' Intrinsic</a></li>
165 <li><a href="#int_frameaddress">'<tt>llvm.frameaddress</tt>' Intrinsic</a></li>
166 <li><a href="#int_stacksave">'<tt>llvm.stacksave</tt>' Intrinsic</a></li>
167 <li><a href="#int_stackrestore">'<tt>llvm.stackrestore</tt>' Intrinsic</a></li>
168 <li><a href="#int_prefetch">'<tt>llvm.prefetch</tt>' Intrinsic</a></li>
169 <li><a href="#int_pcmarker">'<tt>llvm.pcmarker</tt>' Intrinsic</a></li>
170 <li><a href="#int_readcyclecounter"><tt>llvm.readcyclecounter</tt>' Intrinsic</a></li>
173 <li><a href="#int_libc">Standard C Library Intrinsics</a>
175 <li><a href="#int_memcpy">'<tt>llvm.memcpy.*</tt>' Intrinsic</a></li>
176 <li><a href="#int_memmove">'<tt>llvm.memmove.*</tt>' Intrinsic</a></li>
177 <li><a href="#int_memset">'<tt>llvm.memset.*</tt>' Intrinsic</a></li>
178 <li><a href="#int_sqrt">'<tt>llvm.sqrt.*</tt>' Intrinsic</a></li>
179 <li><a href="#int_powi">'<tt>llvm.powi.*</tt>' Intrinsic</a></li>
182 <li><a href="#int_manip">Bit Manipulation Intrinsics</a>
184 <li><a href="#int_bswap">'<tt>llvm.bswap.*</tt>' Intrinsics</a></li>
185 <li><a href="#int_ctpop">'<tt>llvm.ctpop.*</tt>' Intrinsic </a></li>
186 <li><a href="#int_ctlz">'<tt>llvm.ctlz.*</tt>' Intrinsic </a></li>
187 <li><a href="#int_cttz">'<tt>llvm.cttz.*</tt>' Intrinsic </a></li>
188 <li><a href="#int_part_select">'<tt>llvm.part.select.*</tt>' Intrinsic </a></li>
189 <li><a href="#int_part_set">'<tt>llvm.part.set.*</tt>' Intrinsic </a></li>
192 <li><a href="#int_debugger">Debugger intrinsics</a></li>
193 <li><a href="#int_eh">Exception Handling intrinsics</a></li>
194 <li><a href="#int_trampoline">Trampoline Intrinsic</a>
196 <li><a href="#int_it">'<tt>llvm.init.trampoline</tt>' Intrinsic</a></li>
199 <li><a href="#int_general">General intrinsics</a>
201 <li><a href="#int_var_annotation">
202 <tt>llvm.var.annotation</tt>' Intrinsic</a></li>
205 <li><a href="#int_annotation">
206 <tt>llvm.annotation.*</tt>' Intrinsic</a></li>
213 <div class="doc_author">
214 <p>Written by <a href="mailto:sabre@nondot.org">Chris Lattner</a>
215 and <a href="mailto:vadve@cs.uiuc.edu">Vikram Adve</a></p>
218 <!-- *********************************************************************** -->
219 <div class="doc_section"> <a name="abstract">Abstract </a></div>
220 <!-- *********************************************************************** -->
222 <div class="doc_text">
223 <p>This document is a reference manual for the LLVM assembly language.
224 LLVM is an SSA based representation that provides type safety,
225 low-level operations, flexibility, and the capability of representing
226 'all' high-level languages cleanly. It is the common code
227 representation used throughout all phases of the LLVM compilation
231 <!-- *********************************************************************** -->
232 <div class="doc_section"> <a name="introduction">Introduction</a> </div>
233 <!-- *********************************************************************** -->
235 <div class="doc_text">
237 <p>The LLVM code representation is designed to be used in three
238 different forms: as an in-memory compiler IR, as an on-disk bitcode
239 representation (suitable for fast loading by a Just-In-Time compiler),
240 and as a human readable assembly language representation. This allows
241 LLVM to provide a powerful intermediate representation for efficient
242 compiler transformations and analysis, while providing a natural means
243 to debug and visualize the transformations. The three different forms
244 of LLVM are all equivalent. This document describes the human readable
245 representation and notation.</p>
247 <p>The LLVM representation aims to be light-weight and low-level
248 while being expressive, typed, and extensible at the same time. It
249 aims to be a "universal IR" of sorts, by being at a low enough level
250 that high-level ideas may be cleanly mapped to it (similar to how
251 microprocessors are "universal IR's", allowing many source languages to
252 be mapped to them). By providing type information, LLVM can be used as
253 the target of optimizations: for example, through pointer analysis, it
254 can be proven that a C automatic variable is never accessed outside of
255 the current function... allowing it to be promoted to a simple SSA
256 value instead of a memory location.</p>
260 <!-- _______________________________________________________________________ -->
261 <div class="doc_subsubsection"> <a name="wellformed">Well-Formedness</a> </div>
263 <div class="doc_text">
265 <p>It is important to note that this document describes 'well formed'
266 LLVM assembly language. There is a difference between what the parser
267 accepts and what is considered 'well formed'. For example, the
268 following instruction is syntactically okay, but not well formed:</p>
270 <div class="doc_code">
272 %x = <a href="#i_add">add</a> i32 1, %x
276 <p>...because the definition of <tt>%x</tt> does not dominate all of
277 its uses. The LLVM infrastructure provides a verification pass that may
278 be used to verify that an LLVM module is well formed. This pass is
279 automatically run by the parser after parsing input assembly and by
280 the optimizer before it outputs bitcode. The violations pointed out
281 by the verifier pass indicate bugs in transformation passes or input to
285 <!-- Describe the typesetting conventions here. -->
287 <!-- *********************************************************************** -->
288 <div class="doc_section"> <a name="identifiers">Identifiers</a> </div>
289 <!-- *********************************************************************** -->
291 <div class="doc_text">
293 <p>LLVM identifiers come in two basic types: global and local. Global
294 identifiers (functions, global variables) begin with the @ character. Local
295 identifiers (register names, types) begin with the % character. Additionally,
296 there are three different formats for identifiers, for different purposes:
299 <li>Named values are represented as a string of characters with their prefix.
300 For example, %foo, @DivisionByZero, %a.really.long.identifier. The actual
301 regular expression used is '<tt>[%@][a-zA-Z$._][a-zA-Z$._0-9]*</tt>'.
302 Identifiers which require other characters in their names can be surrounded
303 with quotes. In this way, anything except a <tt>"</tt> character can
304 be used in a named value.</li>
306 <li>Unnamed values are represented as an unsigned numeric value with their
307 prefix. For example, %12, @2, %44.</li>
309 <li>Constants, which are described in a <a href="#constants">section about
310 constants</a>, below.</li>
313 <p>LLVM requires that values start with a prefix for two reasons: Compilers
314 don't need to worry about name clashes with reserved words, and the set of
315 reserved words may be expanded in the future without penalty. Additionally,
316 unnamed identifiers allow a compiler to quickly come up with a temporary
317 variable without having to avoid symbol table conflicts.</p>
319 <p>Reserved words in LLVM are very similar to reserved words in other
320 languages. There are keywords for different opcodes
321 ('<tt><a href="#i_add">add</a></tt>',
322 '<tt><a href="#i_bitcast">bitcast</a></tt>',
323 '<tt><a href="#i_ret">ret</a></tt>', etc...), for primitive type names ('<tt><a
324 href="#t_void">void</a></tt>', '<tt><a href="#t_primitive">i32</a></tt>', etc...),
325 and others. These reserved words cannot conflict with variable names, because
326 none of them start with a prefix character ('%' or '@').</p>
328 <p>Here is an example of LLVM code to multiply the integer variable
329 '<tt>%X</tt>' by 8:</p>
333 <div class="doc_code">
335 %result = <a href="#i_mul">mul</a> i32 %X, 8
339 <p>After strength reduction:</p>
341 <div class="doc_code">
343 %result = <a href="#i_shl">shl</a> i32 %X, i8 3
347 <p>And the hard way:</p>
349 <div class="doc_code">
351 <a href="#i_add">add</a> i32 %X, %X <i>; yields {i32}:%0</i>
352 <a href="#i_add">add</a> i32 %0, %0 <i>; yields {i32}:%1</i>
353 %result = <a href="#i_add">add</a> i32 %1, %1
357 <p>This last way of multiplying <tt>%X</tt> by 8 illustrates several
358 important lexical features of LLVM:</p>
362 <li>Comments are delimited with a '<tt>;</tt>' and go until the end of
365 <li>Unnamed temporaries are created when the result of a computation is not
366 assigned to a named value.</li>
368 <li>Unnamed temporaries are numbered sequentially</li>
372 <p>...and it also shows a convention that we follow in this document. When
373 demonstrating instructions, we will follow an instruction with a comment that
374 defines the type and name of value produced. Comments are shown in italic
379 <!-- *********************************************************************** -->
380 <div class="doc_section"> <a name="highlevel">High Level Structure</a> </div>
381 <!-- *********************************************************************** -->
383 <!-- ======================================================================= -->
384 <div class="doc_subsection"> <a name="modulestructure">Module Structure</a>
387 <div class="doc_text">
389 <p>LLVM programs are composed of "Module"s, each of which is a
390 translation unit of the input programs. Each module consists of
391 functions, global variables, and symbol table entries. Modules may be
392 combined together with the LLVM linker, which merges function (and
393 global variable) definitions, resolves forward declarations, and merges
394 symbol table entries. Here is an example of the "hello world" module:</p>
396 <div class="doc_code">
397 <pre><i>; Declare the string constant as a global constant...</i>
398 <a href="#identifiers">@.LC0</a> = <a href="#linkage_internal">internal</a> <a
399 href="#globalvars">constant</a> <a href="#t_array">[13 x i8]</a> c"hello world\0A\00" <i>; [13 x i8]*</i>
401 <i>; External declaration of the puts function</i>
402 <a href="#functionstructure">declare</a> i32 @puts(i8 *) <i>; i32(i8 *)* </i>
404 <i>; Definition of main function</i>
405 define i32 @main() { <i>; i32()* </i>
406 <i>; Convert [13x i8 ]* to i8 *...</i>
408 href="#i_getelementptr">getelementptr</a> [13 x i8 ]* @.LC0, i64 0, i64 0 <i>; i8 *</i>
410 <i>; Call puts function to write out the string to stdout...</i>
412 href="#i_call">call</a> i32 @puts(i8 * %cast210) <i>; i32</i>
414 href="#i_ret">ret</a> i32 0<br>}<br>
418 <p>This example is made up of a <a href="#globalvars">global variable</a>
419 named "<tt>.LC0</tt>", an external declaration of the "<tt>puts</tt>"
420 function, and a <a href="#functionstructure">function definition</a>
421 for "<tt>main</tt>".</p>
423 <p>In general, a module is made up of a list of global values,
424 where both functions and global variables are global values. Global values are
425 represented by a pointer to a memory location (in this case, a pointer to an
426 array of char, and a pointer to a function), and have one of the following <a
427 href="#linkage">linkage types</a>.</p>
431 <!-- ======================================================================= -->
432 <div class="doc_subsection">
433 <a name="linkage">Linkage Types</a>
436 <div class="doc_text">
439 All Global Variables and Functions have one of the following types of linkage:
444 <dt><tt><b><a name="linkage_internal">internal</a></b></tt> </dt>
446 <dd>Global values with internal linkage are only directly accessible by
447 objects in the current module. In particular, linking code into a module with
448 an internal global value may cause the internal to be renamed as necessary to
449 avoid collisions. Because the symbol is internal to the module, all
450 references can be updated. This corresponds to the notion of the
451 '<tt>static</tt>' keyword in C.
454 <dt><tt><b><a name="linkage_linkonce">linkonce</a></b></tt>: </dt>
456 <dd>Globals with "<tt>linkonce</tt>" linkage are merged with other globals of
457 the same name when linkage occurs. This is typically used to implement
458 inline functions, templates, or other code which must be generated in each
459 translation unit that uses it. Unreferenced <tt>linkonce</tt> globals are
460 allowed to be discarded.
463 <dt><tt><b><a name="linkage_weak">weak</a></b></tt>: </dt>
465 <dd>"<tt>weak</tt>" linkage is exactly the same as <tt>linkonce</tt> linkage,
466 except that unreferenced <tt>weak</tt> globals may not be discarded. This is
467 used for globals that may be emitted in multiple translation units, but that
468 are not guaranteed to be emitted into every translation unit that uses them.
469 One example of this are common globals in C, such as "<tt>int X;</tt>" at
473 <dt><tt><b><a name="linkage_appending">appending</a></b></tt>: </dt>
475 <dd>"<tt>appending</tt>" linkage may only be applied to global variables of
476 pointer to array type. When two global variables with appending linkage are
477 linked together, the two global arrays are appended together. This is the
478 LLVM, typesafe, equivalent of having the system linker append together
479 "sections" with identical names when .o files are linked.
482 <dt><tt><b><a name="linkage_externweak">extern_weak</a></b></tt>: </dt>
483 <dd>The semantics of this linkage follow the ELF model: the symbol is weak
484 until linked, if not linked, the symbol becomes null instead of being an
488 <dt><tt><b><a name="linkage_external">externally visible</a></b></tt>:</dt>
490 <dd>If none of the above identifiers are used, the global is externally
491 visible, meaning that it participates in linkage and can be used to resolve
492 external symbol references.
497 The next two types of linkage are targeted for Microsoft Windows platform
498 only. They are designed to support importing (exporting) symbols from (to)
503 <dt><tt><b><a name="linkage_dllimport">dllimport</a></b></tt>: </dt>
505 <dd>"<tt>dllimport</tt>" linkage causes the compiler to reference a function
506 or variable via a global pointer to a pointer that is set up by the DLL
507 exporting the symbol. On Microsoft Windows targets, the pointer name is
508 formed by combining <code>_imp__</code> and the function or variable name.
511 <dt><tt><b><a name="linkage_dllexport">dllexport</a></b></tt>: </dt>
513 <dd>"<tt>dllexport</tt>" linkage causes the compiler to provide a global
514 pointer to a pointer in a DLL, so that it can be referenced with the
515 <tt>dllimport</tt> attribute. On Microsoft Windows targets, the pointer
516 name is formed by combining <code>_imp__</code> and the function or variable
522 <p><a name="linkage_external"></a>For example, since the "<tt>.LC0</tt>"
523 variable is defined to be internal, if another module defined a "<tt>.LC0</tt>"
524 variable and was linked with this one, one of the two would be renamed,
525 preventing a collision. Since "<tt>main</tt>" and "<tt>puts</tt>" are
526 external (i.e., lacking any linkage declarations), they are accessible
527 outside of the current module.</p>
528 <p>It is illegal for a function <i>declaration</i>
529 to have any linkage type other than "externally visible", <tt>dllimport</tt>,
530 or <tt>extern_weak</tt>.</p>
531 <p>Aliases can have only <tt>external</tt>, <tt>internal</tt> and <tt>weak</tt>
535 <!-- ======================================================================= -->
536 <div class="doc_subsection">
537 <a name="callingconv">Calling Conventions</a>
540 <div class="doc_text">
542 <p>LLVM <a href="#functionstructure">functions</a>, <a href="#i_call">calls</a>
543 and <a href="#i_invoke">invokes</a> can all have an optional calling convention
544 specified for the call. The calling convention of any pair of dynamic
545 caller/callee must match, or the behavior of the program is undefined. The
546 following calling conventions are supported by LLVM, and more may be added in
550 <dt><b>"<tt>ccc</tt>" - The C calling convention</b>:</dt>
552 <dd>This calling convention (the default if no other calling convention is
553 specified) matches the target C calling conventions. This calling convention
554 supports varargs function calls and tolerates some mismatch in the declared
555 prototype and implemented declaration of the function (as does normal C).
558 <dt><b>"<tt>fastcc</tt>" - The fast calling convention</b>:</dt>
560 <dd>This calling convention attempts to make calls as fast as possible
561 (e.g. by passing things in registers). This calling convention allows the
562 target to use whatever tricks it wants to produce fast code for the target,
563 without having to conform to an externally specified ABI. Implementations of
564 this convention should allow arbitrary tail call optimization to be supported.
565 This calling convention does not support varargs and requires the prototype of
566 all callees to exactly match the prototype of the function definition.
569 <dt><b>"<tt>coldcc</tt>" - The cold calling convention</b>:</dt>
571 <dd>This calling convention attempts to make code in the caller as efficient
572 as possible under the assumption that the call is not commonly executed. As
573 such, these calls often preserve all registers so that the call does not break
574 any live ranges in the caller side. This calling convention does not support
575 varargs and requires the prototype of all callees to exactly match the
576 prototype of the function definition.
579 <dt><b>"<tt>cc <<em>n</em>></tt>" - Numbered convention</b>:</dt>
581 <dd>Any calling convention may be specified by number, allowing
582 target-specific calling conventions to be used. Target specific calling
583 conventions start at 64.
587 <p>More calling conventions can be added/defined on an as-needed basis, to
588 support pascal conventions or any other well-known target-independent
593 <!-- ======================================================================= -->
594 <div class="doc_subsection">
595 <a name="visibility">Visibility Styles</a>
598 <div class="doc_text">
601 All Global Variables and Functions have one of the following visibility styles:
605 <dt><b>"<tt>default</tt>" - Default style</b>:</dt>
607 <dd>On ELF, default visibility means that the declaration is visible to other
608 modules and, in shared libraries, means that the declared entity may be
609 overridden. On Darwin, default visibility means that the declaration is
610 visible to other modules. Default visibility corresponds to "external
611 linkage" in the language.
614 <dt><b>"<tt>hidden</tt>" - Hidden style</b>:</dt>
616 <dd>Two declarations of an object with hidden visibility refer to the same
617 object if they are in the same shared object. Usually, hidden visibility
618 indicates that the symbol will not be placed into the dynamic symbol table,
619 so no other module (executable or shared library) can reference it
623 <dt><b>"<tt>protected</tt>" - Protected style</b>:</dt>
625 <dd>On ELF, protected visibility indicates that the symbol will be placed in
626 the dynamic symbol table, but that references within the defining module will
627 bind to the local symbol. That is, the symbol cannot be overridden by another
634 <!-- ======================================================================= -->
635 <div class="doc_subsection">
636 <a name="globalvars">Global Variables</a>
639 <div class="doc_text">
641 <p>Global variables define regions of memory allocated at compilation time
642 instead of run-time. Global variables may optionally be initialized, may have
643 an explicit section to be placed in, and may have an optional explicit alignment
644 specified. A variable may be defined as "thread_local", which means that it
645 will not be shared by threads (each thread will have a separated copy of the
646 variable). A variable may be defined as a global "constant," which indicates
647 that the contents of the variable will <b>never</b> be modified (enabling better
648 optimization, allowing the global data to be placed in the read-only section of
649 an executable, etc). Note that variables that need runtime initialization
650 cannot be marked "constant" as there is a store to the variable.</p>
653 LLVM explicitly allows <em>declarations</em> of global variables to be marked
654 constant, even if the final definition of the global is not. This capability
655 can be used to enable slightly better optimization of the program, but requires
656 the language definition to guarantee that optimizations based on the
657 'constantness' are valid for the translation units that do not include the
661 <p>As SSA values, global variables define pointer values that are in
662 scope (i.e. they dominate) all basic blocks in the program. Global
663 variables always define a pointer to their "content" type because they
664 describe a region of memory, and all memory objects in LLVM are
665 accessed through pointers.</p>
667 <p>LLVM allows an explicit section to be specified for globals. If the target
668 supports it, it will emit globals to the section specified.</p>
670 <p>An explicit alignment may be specified for a global. If not present, or if
671 the alignment is set to zero, the alignment of the global is set by the target
672 to whatever it feels convenient. If an explicit alignment is specified, the
673 global is forced to have at least that much alignment. All alignments must be
676 <p>For example, the following defines a global with an initializer, section,
679 <div class="doc_code">
681 @G = constant float 1.0, section "foo", align 4
688 <!-- ======================================================================= -->
689 <div class="doc_subsection">
690 <a name="functionstructure">Functions</a>
693 <div class="doc_text">
695 <p>LLVM function definitions consist of the "<tt>define</tt>" keyord,
696 an optional <a href="#linkage">linkage type</a>, an optional
697 <a href="#visibility">visibility style</a>, an optional
698 <a href="#callingconv">calling convention</a>, a return type, an optional
699 <a href="#paramattrs">parameter attribute</a> for the return type, a function
700 name, a (possibly empty) argument list (each with optional
701 <a href="#paramattrs">parameter attributes</a>), an optional section, an
702 optional alignment, an opening curly brace, a list of basic blocks, and a
705 LLVM function declarations consist of the "<tt>declare</tt>" keyword, an
706 optional <a href="#linkage">linkage type</a>, an optional
707 <a href="#visibility">visibility style</a>, an optional
708 <a href="#callingconv">calling convention</a>, a return type, an optional
709 <a href="#paramattrs">parameter attribute</a> for the return type, a function
710 name, a possibly empty list of arguments, and an optional alignment.</p>
712 <p>A function definition contains a list of basic blocks, forming the CFG for
713 the function. Each basic block may optionally start with a label (giving the
714 basic block a symbol table entry), contains a list of instructions, and ends
715 with a <a href="#terminators">terminator</a> instruction (such as a branch or
716 function return).</p>
718 <p>The first basic block in a function is special in two ways: it is immediately
719 executed on entrance to the function, and it is not allowed to have predecessor
720 basic blocks (i.e. there can not be any branches to the entry block of a
721 function). Because the block can have no predecessors, it also cannot have any
722 <a href="#i_phi">PHI nodes</a>.</p>
724 <p>LLVM allows an explicit section to be specified for functions. If the target
725 supports it, it will emit functions to the section specified.</p>
727 <p>An explicit alignment may be specified for a function. If not present, or if
728 the alignment is set to zero, the alignment of the function is set by the target
729 to whatever it feels convenient. If an explicit alignment is specified, the
730 function is forced to have at least that much alignment. All alignments must be
736 <!-- ======================================================================= -->
737 <div class="doc_subsection">
738 <a name="aliasstructure">Aliases</a>
740 <div class="doc_text">
741 <p>Aliases act as "second name" for the aliasee value (which can be either
742 function or global variable or bitcast of global value). Aliases may have an
743 optional <a href="#linkage">linkage type</a>, and an
744 optional <a href="#visibility">visibility style</a>.</p>
748 <div class="doc_code">
750 @<Name> = [Linkage] [Visibility] alias <AliaseeTy> @<Aliasee>
758 <!-- ======================================================================= -->
759 <div class="doc_subsection"><a name="paramattrs">Parameter Attributes</a></div>
760 <div class="doc_text">
761 <p>The return type and each parameter of a function type may have a set of
762 <i>parameter attributes</i> associated with them. Parameter attributes are
763 used to communicate additional information about the result or parameters of
764 a function. Parameter attributes are considered to be part of the function
765 type so two functions types that differ only by the parameter attributes
766 are different function types.</p>
768 <p>Parameter attributes are simple keywords that follow the type specified. If
769 multiple parameter attributes are needed, they are space separated. For
772 <div class="doc_code">
774 %someFunc = i16 (i8 signext %someParam) zeroext
775 %someFunc = i16 (i8 zeroext %someParam) zeroext
779 <p>Note that the two function types above are unique because the parameter has
780 a different attribute (<tt>signext</tt> in the first one, <tt>zeroext</tt> in
781 the second). Also note that the attribute for the function result
782 (<tt>zeroext</tt>) comes immediately after the argument list.</p>
784 <p>Currently, only the following parameter attributes are defined:</p>
786 <dt><tt>zeroext</tt></dt>
787 <dd>This indicates that the parameter should be zero extended just before
788 a call to this function.</dd>
789 <dt><tt>signext</tt></dt>
790 <dd>This indicates that the parameter should be sign extended just before
791 a call to this function.</dd>
792 <dt><tt>inreg</tt></dt>
793 <dd>This indicates that the parameter should be placed in register (if
794 possible) during assembling function call. Support for this attribute is
796 <dt><tt>sret</tt></dt>
797 <dd>This indicates that the parameter specifies the address of a structure
798 that is the return value of the function in the source program.</dd>
799 <dt><tt>noalias</tt></dt>
800 <dd>This indicates that the parameter not alias any other object or any
801 other "noalias" objects during the function call.
802 <dt><tt>noreturn</tt></dt>
803 <dd>This function attribute indicates that the function never returns. This
804 indicates to LLVM that every call to this function should be treated as if
805 an <tt>unreachable</tt> instruction immediately followed the call.</dd>
806 <dt><tt>nounwind</tt></dt>
807 <dd>This function attribute indicates that the function type does not use
808 the unwind instruction and does not allow stack unwinding to propagate
810 <dt><tt>nest</tt></dt>
811 <dd>This indicates that the parameter can be excised using the
812 <a href="#int_trampoline">trampoline intrinsics</a>.</dd>
817 <!-- ======================================================================= -->
818 <div class="doc_subsection">
819 <a name="moduleasm">Module-Level Inline Assembly</a>
822 <div class="doc_text">
824 Modules may contain "module-level inline asm" blocks, which corresponds to the
825 GCC "file scope inline asm" blocks. These blocks are internally concatenated by
826 LLVM and treated as a single unit, but may be separated in the .ll file if
827 desired. The syntax is very simple:
830 <div class="doc_code">
832 module asm "inline asm code goes here"
833 module asm "more can go here"
837 <p>The strings can contain any character by escaping non-printable characters.
838 The escape sequence used is simply "\xx" where "xx" is the two digit hex code
843 The inline asm code is simply printed to the machine code .s file when
844 assembly code is generated.
848 <!-- ======================================================================= -->
849 <div class="doc_subsection">
850 <a name="datalayout">Data Layout</a>
853 <div class="doc_text">
854 <p>A module may specify a target specific data layout string that specifies how
855 data is to be laid out in memory. The syntax for the data layout is simply:</p>
856 <pre> target datalayout = "<i>layout specification</i>"</pre>
857 <p>The <i>layout specification</i> consists of a list of specifications
858 separated by the minus sign character ('-'). Each specification starts with a
859 letter and may include other information after the letter to define some
860 aspect of the data layout. The specifications accepted are as follows: </p>
863 <dd>Specifies that the target lays out data in big-endian form. That is, the
864 bits with the most significance have the lowest address location.</dd>
866 <dd>Specifies that hte target lays out data in little-endian form. That is,
867 the bits with the least significance have the lowest address location.</dd>
868 <dt><tt>p:<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
869 <dd>This specifies the <i>size</i> of a pointer and its <i>abi</i> and
870 <i>preferred</i> alignments. All sizes are in bits. Specifying the <i>pref</i>
871 alignment is optional. If omitted, the preceding <tt>:</tt> should be omitted
873 <dt><tt>i<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
874 <dd>This specifies the alignment for an integer type of a given bit
875 <i>size</i>. The value of <i>size</i> must be in the range [1,2^23).</dd>
876 <dt><tt>v<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
877 <dd>This specifies the alignment for a vector type of a given bit
879 <dt><tt>f<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
880 <dd>This specifies the alignment for a floating point type of a given bit
881 <i>size</i>. The value of <i>size</i> must be either 32 (float) or 64
883 <dt><tt>a<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
884 <dd>This specifies the alignment for an aggregate type of a given bit
887 <p>When constructing the data layout for a given target, LLVM starts with a
888 default set of specifications which are then (possibly) overriden by the
889 specifications in the <tt>datalayout</tt> keyword. The default specifications
890 are given in this list:</p>
892 <li><tt>E</tt> - big endian</li>
893 <li><tt>p:32:64:64</tt> - 32-bit pointers with 64-bit alignment</li>
894 <li><tt>i1:8:8</tt> - i1 is 8-bit (byte) aligned</li>
895 <li><tt>i8:8:8</tt> - i8 is 8-bit (byte) aligned</li>
896 <li><tt>i16:16:16</tt> - i16 is 16-bit aligned</li>
897 <li><tt>i32:32:32</tt> - i32 is 32-bit aligned</li>
898 <li><tt>i64:32:64</tt> - i64 has abi alignment of 32-bits but preferred
899 alignment of 64-bits</li>
900 <li><tt>f32:32:32</tt> - float is 32-bit aligned</li>
901 <li><tt>f64:64:64</tt> - double is 64-bit aligned</li>
902 <li><tt>v64:64:64</tt> - 64-bit vector is 64-bit aligned</li>
903 <li><tt>v128:128:128</tt> - 128-bit vector is 128-bit aligned</li>
904 <li><tt>a0:0:1</tt> - aggregates are 8-bit aligned</li>
906 <p>When llvm is determining the alignment for a given type, it uses the
909 <li>If the type sought is an exact match for one of the specifications, that
910 specification is used.</li>
911 <li>If no match is found, and the type sought is an integer type, then the
912 smallest integer type that is larger than the bitwidth of the sought type is
913 used. If none of the specifications are larger than the bitwidth then the the
914 largest integer type is used. For example, given the default specifications
915 above, the i7 type will use the alignment of i8 (next largest) while both
916 i65 and i256 will use the alignment of i64 (largest specified).</li>
917 <li>If no match is found, and the type sought is a vector type, then the
918 largest vector type that is smaller than the sought vector type will be used
919 as a fall back. This happens because <128 x double> can be implemented in
920 terms of 64 <2 x double>, for example.</li>
924 <!-- *********************************************************************** -->
925 <div class="doc_section"> <a name="typesystem">Type System</a> </div>
926 <!-- *********************************************************************** -->
928 <div class="doc_text">
930 <p>The LLVM type system is one of the most important features of the
931 intermediate representation. Being typed enables a number of
932 optimizations to be performed on the IR directly, without having to do
933 extra analyses on the side before the transformation. A strong type
934 system makes it easier to read the generated code and enables novel
935 analyses and transformations that are not feasible to perform on normal
936 three address code representations.</p>
940 <!-- ======================================================================= -->
941 <div class="doc_subsection"> <a name="t_primitive">Primitive Types</a> </div>
942 <div class="doc_text">
943 <p>The primitive types are the fundamental building blocks of the LLVM
944 system. The current set of primitive types is as follows:</p>
946 <table class="layout">
951 <tr><th>Type</th><th>Description</th></tr>
952 <tr><td><tt><a name="t_void">void</a></tt></td><td>No value</td></tr>
953 <tr><td><tt>label</tt></td><td>Branch destination</td></tr>
960 <tr><th>Type</th><th>Description</th></tr>
961 <tr><td><tt>float</tt></td><td>32-bit floating point value</td></tr>
962 <tr><td><tt>double</tt></td><td>64-bit floating point value</td></tr>
970 <!-- _______________________________________________________________________ -->
971 <div class="doc_subsubsection"> <a name="t_classifications">Type
972 Classifications</a> </div>
973 <div class="doc_text">
974 <p>These different primitive types fall into a few useful
977 <table border="1" cellspacing="0" cellpadding="4">
979 <tr><th>Classification</th><th>Types</th></tr>
981 <td><a name="t_integer">integer</a></td>
982 <td><tt>i1, i2, i3, ... i8, ... i16, ... i32, ... i64, ... </tt></td>
985 <td><a name="t_floating">floating point</a></td>
986 <td><tt>float, double</tt></td>
989 <td><a name="t_firstclass">first class</a></td>
990 <td><tt>i1, ..., float, double, <br/>
991 <a href="#t_pointer">pointer</a>,<a href="#t_vector">vector</a></tt>
997 <p>The <a href="#t_firstclass">first class</a> types are perhaps the
998 most important. Values of these types are the only ones which can be
999 produced by instructions, passed as arguments, or used as operands to
1000 instructions. This means that all structures and arrays must be
1001 manipulated either by pointer or by component.</p>
1004 <!-- ======================================================================= -->
1005 <div class="doc_subsection"> <a name="t_derived">Derived Types</a> </div>
1007 <div class="doc_text">
1009 <p>The real power in LLVM comes from the derived types in the system.
1010 This is what allows a programmer to represent arrays, functions,
1011 pointers, and other useful types. Note that these derived types may be
1012 recursive: For example, it is possible to have a two dimensional array.</p>
1016 <!-- _______________________________________________________________________ -->
1017 <div class="doc_subsubsection"> <a name="t_integer">Integer Type</a> </div>
1019 <div class="doc_text">
1022 <p>The integer type is a very simple derived type that simply specifies an
1023 arbitrary bit width for the integer type desired. Any bit width from 1 bit to
1024 2^23-1 (about 8 million) can be specified.</p>
1032 <p>The number of bits the integer will occupy is specified by the <tt>N</tt>
1036 <table class="layout">
1046 <tt>i1942652</tt><br/>
1049 A boolean integer of 1 bit<br/>
1050 A nibble sized integer of 4 bits.<br/>
1051 A byte sized integer of 8 bits.<br/>
1052 A half word sized integer of 16 bits.<br/>
1053 A word sized integer of 32 bits.<br/>
1054 An integer whose bit width is the answer. <br/>
1055 A double word sized integer of 64 bits.<br/>
1056 A really big integer of over 1 million bits.<br/>
1062 <!-- _______________________________________________________________________ -->
1063 <div class="doc_subsubsection"> <a name="t_array">Array Type</a> </div>
1065 <div class="doc_text">
1069 <p>The array type is a very simple derived type that arranges elements
1070 sequentially in memory. The array type requires a size (number of
1071 elements) and an underlying data type.</p>
1076 [<# elements> x <elementtype>]
1079 <p>The number of elements is a constant integer value; elementtype may
1080 be any type with a size.</p>
1083 <table class="layout">
1086 <tt>[40 x i32 ]</tt><br/>
1087 <tt>[41 x i32 ]</tt><br/>
1088 <tt>[40 x i8]</tt><br/>
1091 Array of 40 32-bit integer values.<br/>
1092 Array of 41 32-bit integer values.<br/>
1093 Array of 40 8-bit integer values.<br/>
1097 <p>Here are some examples of multidimensional arrays:</p>
1098 <table class="layout">
1101 <tt>[3 x [4 x i32]]</tt><br/>
1102 <tt>[12 x [10 x float]]</tt><br/>
1103 <tt>[2 x [3 x [4 x i16]]]</tt><br/>
1106 3x4 array of 32-bit integer values.<br/>
1107 12x10 array of single precision floating point values.<br/>
1108 2x3x4 array of 16-bit integer values.<br/>
1113 <p>Note that 'variable sized arrays' can be implemented in LLVM with a zero
1114 length array. Normally, accesses past the end of an array are undefined in
1115 LLVM (e.g. it is illegal to access the 5th element of a 3 element array).
1116 As a special case, however, zero length arrays are recognized to be variable
1117 length. This allows implementation of 'pascal style arrays' with the LLVM
1118 type "{ i32, [0 x float]}", for example.</p>
1122 <!-- _______________________________________________________________________ -->
1123 <div class="doc_subsubsection"> <a name="t_function">Function Type</a> </div>
1124 <div class="doc_text">
1126 <p>The function type can be thought of as a function signature. It
1127 consists of a return type and a list of formal parameter types.
1128 Function types are usually used to build virtual function tables
1129 (which are structures of pointers to functions), for indirect function
1130 calls, and when defining a function.</p>
1132 The return type of a function type cannot be an aggregate type.
1135 <pre> <returntype> (<parameter list>)<br></pre>
1136 <p>...where '<tt><parameter list></tt>' is a comma-separated list of type
1137 specifiers. Optionally, the parameter list may include a type <tt>...</tt>,
1138 which indicates that the function takes a variable number of arguments.
1139 Variable argument functions can access their arguments with the <a
1140 href="#int_varargs">variable argument handling intrinsic</a> functions.</p>
1142 <table class="layout">
1144 <td class="left"><tt>i32 (i32)</tt></td>
1145 <td class="left">function taking an <tt>i32</tt>, returning an <tt>i32</tt>
1147 </tr><tr class="layout">
1148 <td class="left"><tt>float (i16 signext, i32 *) *
1150 <td class="left"><a href="#t_pointer">Pointer</a> to a function that takes
1151 an <tt>i16</tt> that should be sign extended and a
1152 <a href="#t_pointer">pointer</a> to <tt>i32</tt>, returning
1155 </tr><tr class="layout">
1156 <td class="left"><tt>i32 (i8*, ...)</tt></td>
1157 <td class="left">A vararg function that takes at least one
1158 <a href="#t_pointer">pointer</a> to <tt>i8 </tt> (char in C),
1159 which returns an integer. This is the signature for <tt>printf</tt> in
1166 <!-- _______________________________________________________________________ -->
1167 <div class="doc_subsubsection"> <a name="t_struct">Structure Type</a> </div>
1168 <div class="doc_text">
1170 <p>The structure type is used to represent a collection of data members
1171 together in memory. The packing of the field types is defined to match
1172 the ABI of the underlying processor. The elements of a structure may
1173 be any type that has a size.</p>
1174 <p>Structures are accessed using '<tt><a href="#i_load">load</a></tt>
1175 and '<tt><a href="#i_store">store</a></tt>' by getting a pointer to a
1176 field with the '<tt><a href="#i_getelementptr">getelementptr</a></tt>'
1179 <pre> { <type list> }<br></pre>
1181 <table class="layout">
1183 <td class="left"><tt>{ i32, i32, i32 }</tt></td>
1184 <td class="left">A triple of three <tt>i32</tt> values</td>
1185 </tr><tr class="layout">
1186 <td class="left"><tt>{ float, i32 (i32) * }</tt></td>
1187 <td class="left">A pair, where the first element is a <tt>float</tt> and the
1188 second element is a <a href="#t_pointer">pointer</a> to a
1189 <a href="#t_function">function</a> that takes an <tt>i32</tt>, returning
1190 an <tt>i32</tt>.</td>
1195 <!-- _______________________________________________________________________ -->
1196 <div class="doc_subsubsection"> <a name="t_pstruct">Packed Structure Type</a>
1198 <div class="doc_text">
1200 <p>The packed structure type is used to represent a collection of data members
1201 together in memory. There is no padding between fields. Further, the alignment
1202 of a packed structure is 1 byte. The elements of a packed structure may
1203 be any type that has a size.</p>
1204 <p>Structures are accessed using '<tt><a href="#i_load">load</a></tt>
1205 and '<tt><a href="#i_store">store</a></tt>' by getting a pointer to a
1206 field with the '<tt><a href="#i_getelementptr">getelementptr</a></tt>'
1209 <pre> < { <type list> } > <br></pre>
1211 <table class="layout">
1213 <td class="left"><tt>< { i32, i32, i32 } ></tt></td>
1214 <td class="left">A triple of three <tt>i32</tt> values</td>
1215 </tr><tr class="layout">
1216 <td class="left"><tt>< { float, i32 (i32) * } ></tt></td>
1217 <td class="left">A pair, where the first element is a <tt>float</tt> and the
1218 second element is a <a href="#t_pointer">pointer</a> to a
1219 <a href="#t_function">function</a> that takes an <tt>i32</tt>, returning
1220 an <tt>i32</tt>.</td>
1225 <!-- _______________________________________________________________________ -->
1226 <div class="doc_subsubsection"> <a name="t_pointer">Pointer Type</a> </div>
1227 <div class="doc_text">
1229 <p>As in many languages, the pointer type represents a pointer or
1230 reference to another object, which must live in memory.</p>
1232 <pre> <type> *<br></pre>
1234 <table class="layout">
1237 <tt>[4x i32]*</tt><br/>
1238 <tt>i32 (i32 *) *</tt><br/>
1241 A <a href="#t_pointer">pointer</a> to <a href="#t_array">array</a> of
1242 four <tt>i32</tt> values<br/>
1243 A <a href="#t_pointer">pointer</a> to a <a
1244 href="#t_function">function</a> that takes an <tt>i32*</tt>, returning an
1251 <!-- _______________________________________________________________________ -->
1252 <div class="doc_subsubsection"> <a name="t_vector">Vector Type</a> </div>
1253 <div class="doc_text">
1257 <p>A vector type is a simple derived type that represents a vector
1258 of elements. Vector types are used when multiple primitive data
1259 are operated in parallel using a single instruction (SIMD).
1260 A vector type requires a size (number of
1261 elements) and an underlying primitive data type. Vectors must have a power
1262 of two length (1, 2, 4, 8, 16 ...). Vector types are
1263 considered <a href="#t_firstclass">first class</a>.</p>
1268 < <# elements> x <elementtype> >
1271 <p>The number of elements is a constant integer value; elementtype may
1272 be any integer or floating point type.</p>
1276 <table class="layout">
1279 <tt><4 x i32></tt><br/>
1280 <tt><8 x float></tt><br/>
1281 <tt><2 x i64></tt><br/>
1284 Vector of 4 32-bit integer values.<br/>
1285 Vector of 8 floating-point values.<br/>
1286 Vector of 2 64-bit integer values.<br/>
1292 <!-- _______________________________________________________________________ -->
1293 <div class="doc_subsubsection"> <a name="t_opaque">Opaque Type</a> </div>
1294 <div class="doc_text">
1298 <p>Opaque types are used to represent unknown types in the system. This
1299 corresponds (for example) to the C notion of a forward declared structure type.
1300 In LLVM, opaque types can eventually be resolved to any type (not just a
1301 structure type).</p>
1311 <table class="layout">
1317 An opaque type.<br/>
1324 <!-- *********************************************************************** -->
1325 <div class="doc_section"> <a name="constants">Constants</a> </div>
1326 <!-- *********************************************************************** -->
1328 <div class="doc_text">
1330 <p>LLVM has several different basic types of constants. This section describes
1331 them all and their syntax.</p>
1335 <!-- ======================================================================= -->
1336 <div class="doc_subsection"><a name="simpleconstants">Simple Constants</a></div>
1338 <div class="doc_text">
1341 <dt><b>Boolean constants</b></dt>
1343 <dd>The two strings '<tt>true</tt>' and '<tt>false</tt>' are both valid
1344 constants of the <tt><a href="#t_primitive">i1</a></tt> type.
1347 <dt><b>Integer constants</b></dt>
1349 <dd>Standard integers (such as '4') are constants of the <a
1350 href="#t_integer">integer</a> type. Negative numbers may be used with
1354 <dt><b>Floating point constants</b></dt>
1356 <dd>Floating point constants use standard decimal notation (e.g. 123.421),
1357 exponential notation (e.g. 1.23421e+2), or a more precise hexadecimal
1358 notation (see below). Floating point constants must have a <a
1359 href="#t_floating">floating point</a> type. </dd>
1361 <dt><b>Null pointer constants</b></dt>
1363 <dd>The identifier '<tt>null</tt>' is recognized as a null pointer constant
1364 and must be of <a href="#t_pointer">pointer type</a>.</dd>
1368 <p>The one non-intuitive notation for constants is the optional hexadecimal form
1369 of floating point constants. For example, the form '<tt>double
1370 0x432ff973cafa8000</tt>' is equivalent to (but harder to read than) '<tt>double
1371 4.5e+15</tt>'. The only time hexadecimal floating point constants are required
1372 (and the only time that they are generated by the disassembler) is when a
1373 floating point constant must be emitted but it cannot be represented as a
1374 decimal floating point number. For example, NaN's, infinities, and other
1375 special values are represented in their IEEE hexadecimal format so that
1376 assembly and disassembly do not cause any bits to change in the constants.</p>
1380 <!-- ======================================================================= -->
1381 <div class="doc_subsection"><a name="aggregateconstants">Aggregate Constants</a>
1384 <div class="doc_text">
1385 <p>Aggregate constants arise from aggregation of simple constants
1386 and smaller aggregate constants.</p>
1389 <dt><b>Structure constants</b></dt>
1391 <dd>Structure constants are represented with notation similar to structure
1392 type definitions (a comma separated list of elements, surrounded by braces
1393 (<tt>{}</tt>)). For example: "<tt>{ i32 4, float 17.0, i32* %G }</tt>",
1394 where "<tt>%G</tt>" is declared as "<tt>@G = external global i32</tt>". Structure constants
1395 must have <a href="#t_struct">structure type</a>, and the number and
1396 types of elements must match those specified by the type.
1399 <dt><b>Array constants</b></dt>
1401 <dd>Array constants are represented with notation similar to array type
1402 definitions (a comma separated list of elements, surrounded by square brackets
1403 (<tt>[]</tt>)). For example: "<tt>[ i32 42, i32 11, i32 74 ]</tt>". Array
1404 constants must have <a href="#t_array">array type</a>, and the number and
1405 types of elements must match those specified by the type.
1408 <dt><b>Vector constants</b></dt>
1410 <dd>Vector constants are represented with notation similar to vector type
1411 definitions (a comma separated list of elements, surrounded by
1412 less-than/greater-than's (<tt><></tt>)). For example: "<tt>< i32 42,
1413 i32 11, i32 74, i32 100 ></tt>". Vector constants must have <a
1414 href="#t_vector">vector type</a>, and the number and types of elements must
1415 match those specified by the type.
1418 <dt><b>Zero initialization</b></dt>
1420 <dd>The string '<tt>zeroinitializer</tt>' can be used to zero initialize a
1421 value to zero of <em>any</em> type, including scalar and aggregate types.
1422 This is often used to avoid having to print large zero initializers (e.g. for
1423 large arrays) and is always exactly equivalent to using explicit zero
1430 <!-- ======================================================================= -->
1431 <div class="doc_subsection">
1432 <a name="globalconstants">Global Variable and Function Addresses</a>
1435 <div class="doc_text">
1437 <p>The addresses of <a href="#globalvars">global variables</a> and <a
1438 href="#functionstructure">functions</a> are always implicitly valid (link-time)
1439 constants. These constants are explicitly referenced when the <a
1440 href="#identifiers">identifier for the global</a> is used and always have <a
1441 href="#t_pointer">pointer</a> type. For example, the following is a legal LLVM
1444 <div class="doc_code">
1448 @Z = global [2 x i32*] [ i32* @X, i32* @Y ]
1454 <!-- ======================================================================= -->
1455 <div class="doc_subsection"><a name="undefvalues">Undefined Values</a></div>
1456 <div class="doc_text">
1457 <p>The string '<tt>undef</tt>' is recognized as a type-less constant that has
1458 no specific value. Undefined values may be of any type and be used anywhere
1459 a constant is permitted.</p>
1461 <p>Undefined values indicate to the compiler that the program is well defined
1462 no matter what value is used, giving the compiler more freedom to optimize.
1466 <!-- ======================================================================= -->
1467 <div class="doc_subsection"><a name="constantexprs">Constant Expressions</a>
1470 <div class="doc_text">
1472 <p>Constant expressions are used to allow expressions involving other constants
1473 to be used as constants. Constant expressions may be of any <a
1474 href="#t_firstclass">first class</a> type and may involve any LLVM operation
1475 that does not have side effects (e.g. load and call are not supported). The
1476 following is the syntax for constant expressions:</p>
1479 <dt><b><tt>trunc ( CST to TYPE )</tt></b></dt>
1480 <dd>Truncate a constant to another type. The bit size of CST must be larger
1481 than the bit size of TYPE. Both types must be integers.</dd>
1483 <dt><b><tt>zext ( CST to TYPE )</tt></b></dt>
1484 <dd>Zero extend a constant to another type. The bit size of CST must be
1485 smaller or equal to the bit size of TYPE. Both types must be integers.</dd>
1487 <dt><b><tt>sext ( CST to TYPE )</tt></b></dt>
1488 <dd>Sign extend a constant to another type. The bit size of CST must be
1489 smaller or equal to the bit size of TYPE. Both types must be integers.</dd>
1491 <dt><b><tt>fptrunc ( CST to TYPE )</tt></b></dt>
1492 <dd>Truncate a floating point constant to another floating point type. The
1493 size of CST must be larger than the size of TYPE. Both types must be
1494 floating point.</dd>
1496 <dt><b><tt>fpext ( CST to TYPE )</tt></b></dt>
1497 <dd>Floating point extend a constant to another type. The size of CST must be
1498 smaller or equal to the size of TYPE. Both types must be floating point.</dd>
1500 <dt><b><tt>fptoui ( CST to TYPE )</tt></b></dt>
1501 <dd>Convert a floating point constant to the corresponding unsigned integer
1502 constant. TYPE must be an integer type. CST must be floating point. If the
1503 value won't fit in the integer type, the results are undefined.</dd>
1505 <dt><b><tt>fptosi ( CST to TYPE )</tt></b></dt>
1506 <dd>Convert a floating point constant to the corresponding signed integer
1507 constant. TYPE must be an integer type. CST must be floating point. If the
1508 value won't fit in the integer type, the results are undefined.</dd>
1510 <dt><b><tt>uitofp ( CST to TYPE )</tt></b></dt>
1511 <dd>Convert an unsigned integer constant to the corresponding floating point
1512 constant. TYPE must be floating point. CST must be of integer type. If the
1513 value won't fit in the floating point type, the results are undefined.</dd>
1515 <dt><b><tt>sitofp ( CST to TYPE )</tt></b></dt>
1516 <dd>Convert a signed integer constant to the corresponding floating point
1517 constant. TYPE must be floating point. CST must be of integer type. If the
1518 value won't fit in the floating point type, the results are undefined.</dd>
1520 <dt><b><tt>ptrtoint ( CST to TYPE )</tt></b></dt>
1521 <dd>Convert a pointer typed constant to the corresponding integer constant
1522 TYPE must be an integer type. CST must be of pointer type. The CST value is
1523 zero extended, truncated, or unchanged to make it fit in TYPE.</dd>
1525 <dt><b><tt>inttoptr ( CST to TYPE )</tt></b></dt>
1526 <dd>Convert a integer constant to a pointer constant. TYPE must be a
1527 pointer type. CST must be of integer type. The CST value is zero extended,
1528 truncated, or unchanged to make it fit in a pointer size. This one is
1529 <i>really</i> dangerous!</dd>
1531 <dt><b><tt>bitcast ( CST to TYPE )</tt></b></dt>
1532 <dd>Convert a constant, CST, to another TYPE. The size of CST and TYPE must be
1533 identical (same number of bits). The conversion is done as if the CST value
1534 was stored to memory and read back as TYPE. In other words, no bits change
1535 with this operator, just the type. This can be used for conversion of
1536 vector types to any other type, as long as they have the same bit width. For
1537 pointers it is only valid to cast to another pointer type.
1540 <dt><b><tt>getelementptr ( CSTPTR, IDX0, IDX1, ... )</tt></b></dt>
1542 <dd>Perform the <a href="#i_getelementptr">getelementptr operation</a> on
1543 constants. As with the <a href="#i_getelementptr">getelementptr</a>
1544 instruction, the index list may have zero or more indexes, which are required
1545 to make sense for the type of "CSTPTR".</dd>
1547 <dt><b><tt>select ( COND, VAL1, VAL2 )</tt></b></dt>
1549 <dd>Perform the <a href="#i_select">select operation</a> on
1552 <dt><b><tt>icmp COND ( VAL1, VAL2 )</tt></b></dt>
1553 <dd>Performs the <a href="#i_icmp">icmp operation</a> on constants.</dd>
1555 <dt><b><tt>fcmp COND ( VAL1, VAL2 )</tt></b></dt>
1556 <dd>Performs the <a href="#i_fcmp">fcmp operation</a> on constants.</dd>
1558 <dt><b><tt>extractelement ( VAL, IDX )</tt></b></dt>
1560 <dd>Perform the <a href="#i_extractelement">extractelement
1561 operation</a> on constants.
1563 <dt><b><tt>insertelement ( VAL, ELT, IDX )</tt></b></dt>
1565 <dd>Perform the <a href="#i_insertelement">insertelement
1566 operation</a> on constants.</dd>
1569 <dt><b><tt>shufflevector ( VEC1, VEC2, IDXMASK )</tt></b></dt>
1571 <dd>Perform the <a href="#i_shufflevector">shufflevector
1572 operation</a> on constants.</dd>
1574 <dt><b><tt>OPCODE ( LHS, RHS )</tt></b></dt>
1576 <dd>Perform the specified operation of the LHS and RHS constants. OPCODE may
1577 be any of the <a href="#binaryops">binary</a> or <a href="#bitwiseops">bitwise
1578 binary</a> operations. The constraints on operands are the same as those for
1579 the corresponding instruction (e.g. no bitwise operations on floating point
1580 values are allowed).</dd>
1584 <!-- *********************************************************************** -->
1585 <div class="doc_section"> <a name="othervalues">Other Values</a> </div>
1586 <!-- *********************************************************************** -->
1588 <!-- ======================================================================= -->
1589 <div class="doc_subsection">
1590 <a name="inlineasm">Inline Assembler Expressions</a>
1593 <div class="doc_text">
1596 LLVM supports inline assembler expressions (as opposed to <a href="#moduleasm">
1597 Module-Level Inline Assembly</a>) through the use of a special value. This
1598 value represents the inline assembler as a string (containing the instructions
1599 to emit), a list of operand constraints (stored as a string), and a flag that
1600 indicates whether or not the inline asm expression has side effects. An example
1601 inline assembler expression is:
1604 <div class="doc_code">
1606 i32 (i32) asm "bswap $0", "=r,r"
1611 Inline assembler expressions may <b>only</b> be used as the callee operand of
1612 a <a href="#i_call"><tt>call</tt> instruction</a>. Thus, typically we have:
1615 <div class="doc_code">
1617 %X = call i32 asm "<a href="#int_bswap">bswap</a> $0", "=r,r"(i32 %Y)
1622 Inline asms with side effects not visible in the constraint list must be marked
1623 as having side effects. This is done through the use of the
1624 '<tt>sideeffect</tt>' keyword, like so:
1627 <div class="doc_code">
1629 call void asm sideeffect "eieio", ""()
1633 <p>TODO: The format of the asm and constraints string still need to be
1634 documented here. Constraints on what can be done (e.g. duplication, moving, etc
1635 need to be documented).
1640 <!-- *********************************************************************** -->
1641 <div class="doc_section"> <a name="instref">Instruction Reference</a> </div>
1642 <!-- *********************************************************************** -->
1644 <div class="doc_text">
1646 <p>The LLVM instruction set consists of several different
1647 classifications of instructions: <a href="#terminators">terminator
1648 instructions</a>, <a href="#binaryops">binary instructions</a>,
1649 <a href="#bitwiseops">bitwise binary instructions</a>, <a
1650 href="#memoryops">memory instructions</a>, and <a href="#otherops">other
1651 instructions</a>.</p>
1655 <!-- ======================================================================= -->
1656 <div class="doc_subsection"> <a name="terminators">Terminator
1657 Instructions</a> </div>
1659 <div class="doc_text">
1661 <p>As mentioned <a href="#functionstructure">previously</a>, every
1662 basic block in a program ends with a "Terminator" instruction, which
1663 indicates which block should be executed after the current block is
1664 finished. These terminator instructions typically yield a '<tt>void</tt>'
1665 value: they produce control flow, not values (the one exception being
1666 the '<a href="#i_invoke"><tt>invoke</tt></a>' instruction).</p>
1667 <p>There are six different terminator instructions: the '<a
1668 href="#i_ret"><tt>ret</tt></a>' instruction, the '<a href="#i_br"><tt>br</tt></a>'
1669 instruction, the '<a href="#i_switch"><tt>switch</tt></a>' instruction,
1670 the '<a href="#i_invoke"><tt>invoke</tt></a>' instruction, the '<a
1671 href="#i_unwind"><tt>unwind</tt></a>' instruction, and the '<a
1672 href="#i_unreachable"><tt>unreachable</tt></a>' instruction.</p>
1676 <!-- _______________________________________________________________________ -->
1677 <div class="doc_subsubsection"> <a name="i_ret">'<tt>ret</tt>'
1678 Instruction</a> </div>
1679 <div class="doc_text">
1681 <pre> ret <type> <value> <i>; Return a value from a non-void function</i>
1682 ret void <i>; Return from void function</i>
1685 <p>The '<tt>ret</tt>' instruction is used to return control flow (and a
1686 value) from a function back to the caller.</p>
1687 <p>There are two forms of the '<tt>ret</tt>' instruction: one that
1688 returns a value and then causes control flow, and one that just causes
1689 control flow to occur.</p>
1691 <p>The '<tt>ret</tt>' instruction may return any '<a
1692 href="#t_firstclass">first class</a>' type. Notice that a function is
1693 not <a href="#wellformed">well formed</a> if there exists a '<tt>ret</tt>'
1694 instruction inside of the function that returns a value that does not
1695 match the return type of the function.</p>
1697 <p>When the '<tt>ret</tt>' instruction is executed, control flow
1698 returns back to the calling function's context. If the caller is a "<a
1699 href="#i_call"><tt>call</tt></a>" instruction, execution continues at
1700 the instruction after the call. If the caller was an "<a
1701 href="#i_invoke"><tt>invoke</tt></a>" instruction, execution continues
1702 at the beginning of the "normal" destination block. If the instruction
1703 returns a value, that value shall set the call or invoke instruction's
1706 <pre> ret i32 5 <i>; Return an integer value of 5</i>
1707 ret void <i>; Return from a void function</i>
1710 <!-- _______________________________________________________________________ -->
1711 <div class="doc_subsubsection"> <a name="i_br">'<tt>br</tt>' Instruction</a> </div>
1712 <div class="doc_text">
1714 <pre> br i1 <cond>, label <iftrue>, label <iffalse><br> br label <dest> <i>; Unconditional branch</i>
1717 <p>The '<tt>br</tt>' instruction is used to cause control flow to
1718 transfer to a different basic block in the current function. There are
1719 two forms of this instruction, corresponding to a conditional branch
1720 and an unconditional branch.</p>
1722 <p>The conditional branch form of the '<tt>br</tt>' instruction takes a
1723 single '<tt>i1</tt>' value and two '<tt>label</tt>' values. The
1724 unconditional form of the '<tt>br</tt>' instruction takes a single
1725 '<tt>label</tt>' value as a target.</p>
1727 <p>Upon execution of a conditional '<tt>br</tt>' instruction, the '<tt>i1</tt>'
1728 argument is evaluated. If the value is <tt>true</tt>, control flows
1729 to the '<tt>iftrue</tt>' <tt>label</tt> argument. If "cond" is <tt>false</tt>,
1730 control flows to the '<tt>iffalse</tt>' <tt>label</tt> argument.</p>
1732 <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
1733 href="#i_ret">ret</a> i32 1<br>IfUnequal:<br> <a href="#i_ret">ret</a> i32 0<br></pre>
1735 <!-- _______________________________________________________________________ -->
1736 <div class="doc_subsubsection">
1737 <a name="i_switch">'<tt>switch</tt>' Instruction</a>
1740 <div class="doc_text">
1744 switch <intty> <value>, label <defaultdest> [ <intty> <val>, label <dest> ... ]
1749 <p>The '<tt>switch</tt>' instruction is used to transfer control flow to one of
1750 several different places. It is a generalization of the '<tt>br</tt>'
1751 instruction, allowing a branch to occur to one of many possible
1757 <p>The '<tt>switch</tt>' instruction uses three parameters: an integer
1758 comparison value '<tt>value</tt>', a default '<tt>label</tt>' destination, and
1759 an array of pairs of comparison value constants and '<tt>label</tt>'s. The
1760 table is not allowed to contain duplicate constant entries.</p>
1764 <p>The <tt>switch</tt> instruction specifies a table of values and
1765 destinations. When the '<tt>switch</tt>' instruction is executed, this
1766 table is searched for the given value. If the value is found, control flow is
1767 transfered to the corresponding destination; otherwise, control flow is
1768 transfered to the default destination.</p>
1770 <h5>Implementation:</h5>
1772 <p>Depending on properties of the target machine and the particular
1773 <tt>switch</tt> instruction, this instruction may be code generated in different
1774 ways. For example, it could be generated as a series of chained conditional
1775 branches or with a lookup table.</p>
1780 <i>; Emulate a conditional br instruction</i>
1781 %Val = <a href="#i_zext">zext</a> i1 %value to i32
1782 switch i32 %Val, label %truedest [i32 0, label %falsedest ]
1784 <i>; Emulate an unconditional br instruction</i>
1785 switch i32 0, label %dest [ ]
1787 <i>; Implement a jump table:</i>
1788 switch i32 %val, label %otherwise [ i32 0, label %onzero
1790 i32 2, label %ontwo ]
1794 <!-- _______________________________________________________________________ -->
1795 <div class="doc_subsubsection">
1796 <a name="i_invoke">'<tt>invoke</tt>' Instruction</a>
1799 <div class="doc_text">
1804 <result> = invoke [<a href="#callingconv">cconv</a>] <ptr to function ty> %<function ptr val>(<function args>)
1805 to label <normal label> unwind label <exception label>
1810 <p>The '<tt>invoke</tt>' instruction causes control to transfer to a specified
1811 function, with the possibility of control flow transfer to either the
1812 '<tt>normal</tt>' label or the
1813 '<tt>exception</tt>' label. If the callee function returns with the
1814 "<tt><a href="#i_ret">ret</a></tt>" instruction, control flow will return to the
1815 "normal" label. If the callee (or any indirect callees) returns with the "<a
1816 href="#i_unwind"><tt>unwind</tt></a>" instruction, control is interrupted and
1817 continued at the dynamically nearest "exception" label.</p>
1821 <p>This instruction requires several arguments:</p>
1825 The optional "cconv" marker indicates which <a href="#callingconv">calling
1826 convention</a> the call should use. If none is specified, the call defaults
1827 to using C calling conventions.
1829 <li>'<tt>ptr to function ty</tt>': shall be the signature of the pointer to
1830 function value being invoked. In most cases, this is a direct function
1831 invocation, but indirect <tt>invoke</tt>s are just as possible, branching off
1832 an arbitrary pointer to function value.
1835 <li>'<tt>function ptr val</tt>': An LLVM value containing a pointer to a
1836 function to be invoked. </li>
1838 <li>'<tt>function args</tt>': argument list whose types match the function
1839 signature argument types. If the function signature indicates the function
1840 accepts a variable number of arguments, the extra arguments can be
1843 <li>'<tt>normal label</tt>': the label reached when the called function
1844 executes a '<tt><a href="#i_ret">ret</a></tt>' instruction. </li>
1846 <li>'<tt>exception label</tt>': the label reached when a callee returns with
1847 the <a href="#i_unwind"><tt>unwind</tt></a> instruction. </li>
1853 <p>This instruction is designed to operate as a standard '<tt><a
1854 href="#i_call">call</a></tt>' instruction in most regards. The primary
1855 difference is that it establishes an association with a label, which is used by
1856 the runtime library to unwind the stack.</p>
1858 <p>This instruction is used in languages with destructors to ensure that proper
1859 cleanup is performed in the case of either a <tt>longjmp</tt> or a thrown
1860 exception. Additionally, this is important for implementation of
1861 '<tt>catch</tt>' clauses in high-level languages that support them.</p>
1865 %retval = invoke i32 %Test(i32 15) to label %Continue
1866 unwind label %TestCleanup <i>; {i32}:retval set</i>
1867 %retval = invoke <a href="#callingconv">coldcc</a> i32 %Test(i32 15) to label %Continue
1868 unwind label %TestCleanup <i>; {i32}:retval set</i>
1873 <!-- _______________________________________________________________________ -->
1875 <div class="doc_subsubsection"> <a name="i_unwind">'<tt>unwind</tt>'
1876 Instruction</a> </div>
1878 <div class="doc_text">
1887 <p>The '<tt>unwind</tt>' instruction unwinds the stack, continuing control flow
1888 at the first callee in the dynamic call stack which used an <a
1889 href="#i_invoke"><tt>invoke</tt></a> instruction to perform the call. This is
1890 primarily used to implement exception handling.</p>
1894 <p>The '<tt>unwind</tt>' intrinsic causes execution of the current function to
1895 immediately halt. The dynamic call stack is then searched for the first <a
1896 href="#i_invoke"><tt>invoke</tt></a> instruction on the call stack. Once found,
1897 execution continues at the "exceptional" destination block specified by the
1898 <tt>invoke</tt> instruction. If there is no <tt>invoke</tt> instruction in the
1899 dynamic call chain, undefined behavior results.</p>
1902 <!-- _______________________________________________________________________ -->
1904 <div class="doc_subsubsection"> <a name="i_unreachable">'<tt>unreachable</tt>'
1905 Instruction</a> </div>
1907 <div class="doc_text">
1916 <p>The '<tt>unreachable</tt>' instruction has no defined semantics. This
1917 instruction is used to inform the optimizer that a particular portion of the
1918 code is not reachable. This can be used to indicate that the code after a
1919 no-return function cannot be reached, and other facts.</p>
1923 <p>The '<tt>unreachable</tt>' instruction has no defined semantics.</p>
1928 <!-- ======================================================================= -->
1929 <div class="doc_subsection"> <a name="binaryops">Binary Operations</a> </div>
1930 <div class="doc_text">
1931 <p>Binary operators are used to do most of the computation in a
1932 program. They require two operands, execute an operation on them, and
1933 produce a single value. The operands might represent
1934 multiple data, as is the case with the <a href="#t_vector">vector</a> data type.
1935 The result value of a binary operator is not
1936 necessarily the same type as its operands.</p>
1937 <p>There are several different binary operators:</p>
1939 <!-- _______________________________________________________________________ -->
1940 <div class="doc_subsubsection"> <a name="i_add">'<tt>add</tt>'
1941 Instruction</a> </div>
1942 <div class="doc_text">
1944 <pre> <result> = add <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1947 <p>The '<tt>add</tt>' instruction returns the sum of its two operands.</p>
1949 <p>The two arguments to the '<tt>add</tt>' instruction must be either <a
1950 href="#t_integer">integer</a> or <a href="#t_floating">floating point</a> values.
1951 This instruction can also take <a href="#t_vector">vector</a> versions of the values.
1952 Both arguments must have identical types.</p>
1954 <p>The value produced is the integer or floating point sum of the two
1957 <pre> <result> = add i32 4, %var <i>; yields {i32}:result = 4 + %var</i>
1960 <!-- _______________________________________________________________________ -->
1961 <div class="doc_subsubsection"> <a name="i_sub">'<tt>sub</tt>'
1962 Instruction</a> </div>
1963 <div class="doc_text">
1965 <pre> <result> = sub <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1968 <p>The '<tt>sub</tt>' instruction returns the difference of its two
1970 <p>Note that the '<tt>sub</tt>' instruction is used to represent the '<tt>neg</tt>'
1971 instruction present in most other intermediate representations.</p>
1973 <p>The two arguments to the '<tt>sub</tt>' instruction must be either <a
1974 href="#t_integer">integer</a> or <a href="#t_floating">floating point</a>
1976 This instruction can also take <a href="#t_vector">vector</a> versions of the values.
1977 Both arguments must have identical types.</p>
1979 <p>The value produced is the integer or floating point difference of
1980 the two operands.</p>
1983 <result> = sub i32 4, %var <i>; yields {i32}:result = 4 - %var</i>
1984 <result> = sub i32 0, %val <i>; yields {i32}:result = -%var</i>
1987 <!-- _______________________________________________________________________ -->
1988 <div class="doc_subsubsection"> <a name="i_mul">'<tt>mul</tt>'
1989 Instruction</a> </div>
1990 <div class="doc_text">
1992 <pre> <result> = mul <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1995 <p>The '<tt>mul</tt>' instruction returns the product of its two
1998 <p>The two arguments to the '<tt>mul</tt>' instruction must be either <a
1999 href="#t_integer">integer</a> or <a href="#t_floating">floating point</a>
2001 This instruction can also take <a href="#t_vector">vector</a> versions of the values.
2002 Both arguments must have identical types.</p>
2004 <p>The value produced is the integer or floating point product of the
2006 <p>Because the operands are the same width, the result of an integer
2007 multiplication is the same whether the operands should be deemed unsigned or
2010 <pre> <result> = mul i32 4, %var <i>; yields {i32}:result = 4 * %var</i>
2013 <!-- _______________________________________________________________________ -->
2014 <div class="doc_subsubsection"> <a name="i_udiv">'<tt>udiv</tt>' Instruction
2016 <div class="doc_text">
2018 <pre> <result> = udiv <ty> <var1>, <var2> <i>; yields {ty}:result</i>
2021 <p>The '<tt>udiv</tt>' instruction returns the quotient of its two
2024 <p>The two arguments to the '<tt>udiv</tt>' instruction must be
2025 <a href="#t_integer">integer</a> values. Both arguments must have identical
2026 types. This instruction can also take <a href="#t_vector">vector</a> versions
2027 of the values in which case the elements must be integers.</p>
2029 <p>The value produced is the unsigned integer quotient of the two operands. This
2030 instruction always performs an unsigned division operation, regardless of
2031 whether the arguments are unsigned or not.</p>
2033 <pre> <result> = udiv i32 4, %var <i>; yields {i32}:result = 4 / %var</i>
2036 <!-- _______________________________________________________________________ -->
2037 <div class="doc_subsubsection"> <a name="i_sdiv">'<tt>sdiv</tt>' Instruction
2039 <div class="doc_text">
2041 <pre> <result> = sdiv <ty> <var1>, <var2> <i>; yields {ty}:result</i>
2044 <p>The '<tt>sdiv</tt>' instruction returns the quotient of its two
2047 <p>The two arguments to the '<tt>sdiv</tt>' instruction must be
2048 <a href="#t_integer">integer</a> values. Both arguments must have identical
2049 types. This instruction can also take <a href="#t_vector">vector</a> versions
2050 of the values in which case the elements must be integers.</p>
2052 <p>The value produced is the signed integer quotient of the two operands. This
2053 instruction always performs a signed division operation, regardless of whether
2054 the arguments are signed or not.</p>
2056 <pre> <result> = sdiv i32 4, %var <i>; yields {i32}:result = 4 / %var</i>
2059 <!-- _______________________________________________________________________ -->
2060 <div class="doc_subsubsection"> <a name="i_fdiv">'<tt>fdiv</tt>'
2061 Instruction</a> </div>
2062 <div class="doc_text">
2064 <pre> <result> = fdiv <ty> <var1>, <var2> <i>; yields {ty}:result</i>
2067 <p>The '<tt>fdiv</tt>' instruction returns the quotient of its two
2070 <p>The two arguments to the '<tt>fdiv</tt>' instruction must be
2071 <a href="#t_floating">floating point</a> values. Both arguments must have
2072 identical types. This instruction can also take <a href="#t_vector">vector</a>
2073 versions of floating point values.</p>
2075 <p>The value produced is the floating point quotient of the two operands.</p>
2077 <pre> <result> = fdiv float 4.0, %var <i>; yields {float}:result = 4.0 / %var</i>
2080 <!-- _______________________________________________________________________ -->
2081 <div class="doc_subsubsection"> <a name="i_urem">'<tt>urem</tt>' Instruction</a>
2083 <div class="doc_text">
2085 <pre> <result> = urem <ty> <var1>, <var2> <i>; yields {ty}:result</i>
2088 <p>The '<tt>urem</tt>' instruction returns the remainder from the
2089 unsigned division of its two arguments.</p>
2091 <p>The two arguments to the '<tt>urem</tt>' instruction must be
2092 <a href="#t_integer">integer</a> values. Both arguments must have identical
2095 <p>This instruction returns the unsigned integer <i>remainder</i> of a division.
2096 This instruction always performs an unsigned division to get the remainder,
2097 regardless of whether the arguments are unsigned or not.</p>
2099 <pre> <result> = urem i32 4, %var <i>; yields {i32}:result = 4 % %var</i>
2103 <!-- _______________________________________________________________________ -->
2104 <div class="doc_subsubsection"> <a name="i_srem">'<tt>srem</tt>'
2105 Instruction</a> </div>
2106 <div class="doc_text">
2108 <pre> <result> = srem <ty> <var1>, <var2> <i>; yields {ty}:result</i>
2111 <p>The '<tt>srem</tt>' instruction returns the remainder from the
2112 signed division of its two operands.</p>
2114 <p>The two arguments to the '<tt>srem</tt>' instruction must be
2115 <a href="#t_integer">integer</a> values. Both arguments must have identical
2118 <p>This instruction returns the <i>remainder</i> of a division (where the result
2119 has the same sign as the dividend, <tt>var1</tt>), not the <i>modulo</i>
2120 operator (where the result has the same sign as the divisor, <tt>var2</tt>) of
2121 a value. For more information about the difference, see <a
2122 href="http://mathforum.org/dr.math/problems/anne.4.28.99.html">The
2123 Math Forum</a>. For a table of how this is implemented in various languages,
2124 please see <a href="http://en.wikipedia.org/wiki/Modulo_operation">
2125 Wikipedia: modulo operation</a>.</p>
2127 <pre> <result> = srem i32 4, %var <i>; yields {i32}:result = 4 % %var</i>
2131 <!-- _______________________________________________________________________ -->
2132 <div class="doc_subsubsection"> <a name="i_frem">'<tt>frem</tt>'
2133 Instruction</a> </div>
2134 <div class="doc_text">
2136 <pre> <result> = frem <ty> <var1>, <var2> <i>; yields {ty}:result</i>
2139 <p>The '<tt>frem</tt>' instruction returns the remainder from the
2140 division of its two operands.</p>
2142 <p>The two arguments to the '<tt>frem</tt>' instruction must be
2143 <a href="#t_floating">floating point</a> values. Both arguments must have
2144 identical types.</p>
2146 <p>This instruction returns the <i>remainder</i> of a division.</p>
2148 <pre> <result> = frem float 4.0, %var <i>; yields {float}:result = 4.0 % %var</i>
2152 <!-- ======================================================================= -->
2153 <div class="doc_subsection"> <a name="bitwiseops">Bitwise Binary
2154 Operations</a> </div>
2155 <div class="doc_text">
2156 <p>Bitwise binary operators are used to do various forms of
2157 bit-twiddling in a program. They are generally very efficient
2158 instructions and can commonly be strength reduced from other
2159 instructions. They require two operands, execute an operation on them,
2160 and produce a single value. The resulting value of the bitwise binary
2161 operators is always the same type as its first operand.</p>
2164 <!-- _______________________________________________________________________ -->
2165 <div class="doc_subsubsection"> <a name="i_shl">'<tt>shl</tt>'
2166 Instruction</a> </div>
2167 <div class="doc_text">
2169 <pre> <result> = shl <ty> <var1>, <var2> <i>; yields {ty}:result</i>
2174 <p>The '<tt>shl</tt>' instruction returns the first operand shifted to
2175 the left a specified number of bits.</p>
2179 <p>Both arguments to the '<tt>shl</tt>' instruction must be the same <a
2180 href="#t_integer">integer</a> type.</p>
2184 <p>The value produced is <tt>var1</tt> * 2<sup><tt>var2</tt></sup>. If
2185 <tt>var2</tt> is (statically or dynamically) equal to or larger than the number
2186 of bits in <tt>var1</tt>, the result is undefined.</p>
2188 <h5>Example:</h5><pre>
2189 <result> = shl i32 4, %var <i>; yields {i32}: 4 << %var</i>
2190 <result> = shl i32 4, 2 <i>; yields {i32}: 16</i>
2191 <result> = shl i32 1, 10 <i>; yields {i32}: 1024</i>
2192 <result> = shl i32 1, 32 <i>; undefined</i>
2195 <!-- _______________________________________________________________________ -->
2196 <div class="doc_subsubsection"> <a name="i_lshr">'<tt>lshr</tt>'
2197 Instruction</a> </div>
2198 <div class="doc_text">
2200 <pre> <result> = lshr <ty> <var1>, <var2> <i>; yields {ty}:result</i>
2204 <p>The '<tt>lshr</tt>' instruction (logical shift right) returns the first
2205 operand shifted to the right a specified number of bits with zero fill.</p>
2208 <p>Both arguments to the '<tt>lshr</tt>' instruction must be the same
2209 <a href="#t_integer">integer</a> type.</p>
2213 <p>This instruction always performs a logical shift right operation. The most
2214 significant bits of the result will be filled with zero bits after the
2215 shift. If <tt>var2</tt> is (statically or dynamically) equal to or larger than
2216 the number of bits in <tt>var1</tt>, the result is undefined.</p>
2220 <result> = lshr i32 4, 1 <i>; yields {i32}:result = 2</i>
2221 <result> = lshr i32 4, 2 <i>; yields {i32}:result = 1</i>
2222 <result> = lshr i8 4, 3 <i>; yields {i8}:result = 0</i>
2223 <result> = lshr i8 -2, 1 <i>; yields {i8}:result = 0x7FFFFFFF </i>
2224 <result> = lshr i32 1, 32 <i>; undefined</i>
2228 <!-- _______________________________________________________________________ -->
2229 <div class="doc_subsubsection"> <a name="i_ashr">'<tt>ashr</tt>'
2230 Instruction</a> </div>
2231 <div class="doc_text">
2234 <pre> <result> = ashr <ty> <var1>, <var2> <i>; yields {ty}:result</i>
2238 <p>The '<tt>ashr</tt>' instruction (arithmetic shift right) returns the first
2239 operand shifted to the right a specified number of bits with sign extension.</p>
2242 <p>Both arguments to the '<tt>ashr</tt>' instruction must be the same
2243 <a href="#t_integer">integer</a> type.</p>
2246 <p>This instruction always performs an arithmetic shift right operation,
2247 The most significant bits of the result will be filled with the sign bit
2248 of <tt>var1</tt>. If <tt>var2</tt> is (statically or dynamically) equal to or
2249 larger than the number of bits in <tt>var1</tt>, the result is undefined.
2254 <result> = ashr i32 4, 1 <i>; yields {i32}:result = 2</i>
2255 <result> = ashr i32 4, 2 <i>; yields {i32}:result = 1</i>
2256 <result> = ashr i8 4, 3 <i>; yields {i8}:result = 0</i>
2257 <result> = ashr i8 -2, 1 <i>; yields {i8}:result = -1</i>
2258 <result> = ashr i32 1, 32 <i>; undefined</i>
2262 <!-- _______________________________________________________________________ -->
2263 <div class="doc_subsubsection"> <a name="i_and">'<tt>and</tt>'
2264 Instruction</a> </div>
2265 <div class="doc_text">
2267 <pre> <result> = and <ty> <var1>, <var2> <i>; yields {ty}:result</i>
2270 <p>The '<tt>and</tt>' instruction returns the bitwise logical and of
2271 its two operands.</p>
2273 <p>The two arguments to the '<tt>and</tt>' instruction must be <a
2274 href="#t_integer">integer</a> values. Both arguments must have
2275 identical types.</p>
2277 <p>The truth table used for the '<tt>and</tt>' instruction is:</p>
2279 <div style="align: center">
2280 <table border="1" cellspacing="0" cellpadding="4">
2311 <pre> <result> = and i32 4, %var <i>; yields {i32}:result = 4 & %var</i>
2312 <result> = and i32 15, 40 <i>; yields {i32}:result = 8</i>
2313 <result> = and i32 4, 8 <i>; yields {i32}:result = 0</i>
2316 <!-- _______________________________________________________________________ -->
2317 <div class="doc_subsubsection"> <a name="i_or">'<tt>or</tt>' Instruction</a> </div>
2318 <div class="doc_text">
2320 <pre> <result> = or <ty> <var1>, <var2> <i>; yields {ty}:result</i>
2323 <p>The '<tt>or</tt>' instruction returns the bitwise logical inclusive
2324 or of its two operands.</p>
2326 <p>The two arguments to the '<tt>or</tt>' instruction must be <a
2327 href="#t_integer">integer</a> values. Both arguments must have
2328 identical types.</p>
2330 <p>The truth table used for the '<tt>or</tt>' instruction is:</p>
2332 <div style="align: center">
2333 <table border="1" cellspacing="0" cellpadding="4">
2364 <pre> <result> = or i32 4, %var <i>; yields {i32}:result = 4 | %var</i>
2365 <result> = or i32 15, 40 <i>; yields {i32}:result = 47</i>
2366 <result> = or i32 4, 8 <i>; yields {i32}:result = 12</i>
2369 <!-- _______________________________________________________________________ -->
2370 <div class="doc_subsubsection"> <a name="i_xor">'<tt>xor</tt>'
2371 Instruction</a> </div>
2372 <div class="doc_text">
2374 <pre> <result> = xor <ty> <var1>, <var2> <i>; yields {ty}:result</i>
2377 <p>The '<tt>xor</tt>' instruction returns the bitwise logical exclusive
2378 or of its two operands. The <tt>xor</tt> is used to implement the
2379 "one's complement" operation, which is the "~" operator in C.</p>
2381 <p>The two arguments to the '<tt>xor</tt>' instruction must be <a
2382 href="#t_integer">integer</a> values. Both arguments must have
2383 identical types.</p>
2385 <p>The truth table used for the '<tt>xor</tt>' instruction is:</p>
2387 <div style="align: center">
2388 <table border="1" cellspacing="0" cellpadding="4">
2420 <pre> <result> = xor i32 4, %var <i>; yields {i32}:result = 4 ^ %var</i>
2421 <result> = xor i32 15, 40 <i>; yields {i32}:result = 39</i>
2422 <result> = xor i32 4, 8 <i>; yields {i32}:result = 12</i>
2423 <result> = xor i32 %V, -1 <i>; yields {i32}:result = ~%V</i>
2427 <!-- ======================================================================= -->
2428 <div class="doc_subsection">
2429 <a name="vectorops">Vector Operations</a>
2432 <div class="doc_text">
2434 <p>LLVM supports several instructions to represent vector operations in a
2435 target-independent manner. These instructions cover the element-access and
2436 vector-specific operations needed to process vectors effectively. While LLVM
2437 does directly support these vector operations, many sophisticated algorithms
2438 will want to use target-specific intrinsics to take full advantage of a specific
2443 <!-- _______________________________________________________________________ -->
2444 <div class="doc_subsubsection">
2445 <a name="i_extractelement">'<tt>extractelement</tt>' Instruction</a>
2448 <div class="doc_text">
2453 <result> = extractelement <n x <ty>> <val>, i32 <idx> <i>; yields <ty></i>
2459 The '<tt>extractelement</tt>' instruction extracts a single scalar
2460 element from a vector at a specified index.
2467 The first operand of an '<tt>extractelement</tt>' instruction is a
2468 value of <a href="#t_vector">vector</a> type. The second operand is
2469 an index indicating the position from which to extract the element.
2470 The index may be a variable.</p>
2475 The result is a scalar of the same type as the element type of
2476 <tt>val</tt>. Its value is the value at position <tt>idx</tt> of
2477 <tt>val</tt>. If <tt>idx</tt> exceeds the length of <tt>val</tt>, the
2478 results are undefined.
2484 %result = extractelement <4 x i32> %vec, i32 0 <i>; yields i32</i>
2489 <!-- _______________________________________________________________________ -->
2490 <div class="doc_subsubsection">
2491 <a name="i_insertelement">'<tt>insertelement</tt>' Instruction</a>
2494 <div class="doc_text">
2499 <result> = insertelement <n x <ty>> <val>, <ty> <elt>, i32 <idx> <i>; yields <n x <ty>></i>
2505 The '<tt>insertelement</tt>' instruction inserts a scalar
2506 element into a vector at a specified index.
2513 The first operand of an '<tt>insertelement</tt>' instruction is a
2514 value of <a href="#t_vector">vector</a> type. The second operand is a
2515 scalar value whose type must equal the element type of the first
2516 operand. The third operand is an index indicating the position at
2517 which to insert the value. The index may be a variable.</p>
2522 The result is a vector of the same type as <tt>val</tt>. Its
2523 element values are those of <tt>val</tt> except at position
2524 <tt>idx</tt>, where it gets the value <tt>elt</tt>. If <tt>idx</tt>
2525 exceeds the length of <tt>val</tt>, the results are undefined.
2531 %result = insertelement <4 x i32> %vec, i32 1, i32 0 <i>; yields <4 x i32></i>
2535 <!-- _______________________________________________________________________ -->
2536 <div class="doc_subsubsection">
2537 <a name="i_shufflevector">'<tt>shufflevector</tt>' Instruction</a>
2540 <div class="doc_text">
2545 <result> = shufflevector <n x <ty>> <v1>, <n x <ty>> <v2>, <n x i32> <mask> <i>; yields <n x <ty>></i>
2551 The '<tt>shufflevector</tt>' instruction constructs a permutation of elements
2552 from two input vectors, returning a vector of the same type.
2558 The first two operands of a '<tt>shufflevector</tt>' instruction are vectors
2559 with types that match each other and types that match the result of the
2560 instruction. The third argument is a shuffle mask, which has the same number
2561 of elements as the other vector type, but whose element type is always 'i32'.
2565 The shuffle mask operand is required to be a constant vector with either
2566 constant integer or undef values.
2572 The elements of the two input vectors are numbered from left to right across
2573 both of the vectors. The shuffle mask operand specifies, for each element of
2574 the result vector, which element of the two input registers the result element
2575 gets. The element selector may be undef (meaning "don't care") and the second
2576 operand may be undef if performing a shuffle from only one vector.
2582 %result = shufflevector <4 x i32> %v1, <4 x i32> %v2,
2583 <4 x i32> <i32 0, i32 4, i32 1, i32 5> <i>; yields <4 x i32></i>
2584 %result = shufflevector <4 x i32> %v1, <4 x i32> undef,
2585 <4 x i32> <i32 0, i32 1, i32 2, i32 3> <i>; yields <4 x i32></i> - Identity shuffle.
2590 <!-- ======================================================================= -->
2591 <div class="doc_subsection">
2592 <a name="memoryops">Memory Access and Addressing Operations</a>
2595 <div class="doc_text">
2597 <p>A key design point of an SSA-based representation is how it
2598 represents memory. In LLVM, no memory locations are in SSA form, which
2599 makes things very simple. This section describes how to read, write,
2600 allocate, and free memory in LLVM.</p>
2604 <!-- _______________________________________________________________________ -->
2605 <div class="doc_subsubsection">
2606 <a name="i_malloc">'<tt>malloc</tt>' Instruction</a>
2609 <div class="doc_text">
2614 <result> = malloc <type>[, i32 <NumElements>][, align <alignment>] <i>; yields {type*}:result</i>
2619 <p>The '<tt>malloc</tt>' instruction allocates memory from the system
2620 heap and returns a pointer to it.</p>
2624 <p>The '<tt>malloc</tt>' instruction allocates
2625 <tt>sizeof(<type>)*NumElements</tt>
2626 bytes of memory from the operating system and returns a pointer of the
2627 appropriate type to the program. If "NumElements" is specified, it is the
2628 number of elements allocated. If an alignment is specified, the value result
2629 of the allocation is guaranteed to be aligned to at least that boundary. If
2630 not specified, or if zero, the target can choose to align the allocation on any
2631 convenient boundary.</p>
2633 <p>'<tt>type</tt>' must be a sized type.</p>
2637 <p>Memory is allocated using the system "<tt>malloc</tt>" function, and
2638 a pointer is returned.</p>
2643 %array = malloc [4 x i8 ] <i>; yields {[%4 x i8]*}:array</i>
2645 %size = <a href="#i_add">add</a> i32 2, 2 <i>; yields {i32}:size = i32 4</i>
2646 %array1 = malloc i8, i32 4 <i>; yields {i8*}:array1</i>
2647 %array2 = malloc [12 x i8], i32 %size <i>; yields {[12 x i8]*}:array2</i>
2648 %array3 = malloc i32, i32 4, align 1024 <i>; yields {i32*}:array3</i>
2649 %array4 = malloc i32, align 1024 <i>; yields {i32*}:array4</i>
2653 <!-- _______________________________________________________________________ -->
2654 <div class="doc_subsubsection">
2655 <a name="i_free">'<tt>free</tt>' Instruction</a>
2658 <div class="doc_text">
2663 free <type> <value> <i>; yields {void}</i>
2668 <p>The '<tt>free</tt>' instruction returns memory back to the unused
2669 memory heap to be reallocated in the future.</p>
2673 <p>'<tt>value</tt>' shall be a pointer value that points to a value
2674 that was allocated with the '<tt><a href="#i_malloc">malloc</a></tt>'
2679 <p>Access to the memory pointed to by the pointer is no longer defined
2680 after this instruction executes.</p>
2685 %array = <a href="#i_malloc">malloc</a> [4 x i8] <i>; yields {[4 x i8]*}:array</i>
2686 free [4 x i8]* %array
2690 <!-- _______________________________________________________________________ -->
2691 <div class="doc_subsubsection">
2692 <a name="i_alloca">'<tt>alloca</tt>' Instruction</a>
2695 <div class="doc_text">
2700 <result> = alloca <type>[, i32 <NumElements>][, align <alignment>] <i>; yields {type*}:result</i>
2705 <p>The '<tt>alloca</tt>' instruction allocates memory on the stack frame of the
2706 currently executing function, to be automatically released when this function
2707 returns to its caller.</p>
2711 <p>The '<tt>alloca</tt>' instruction allocates <tt>sizeof(<type>)*NumElements</tt>
2712 bytes of memory on the runtime stack, returning a pointer of the
2713 appropriate type to the program. If "NumElements" is specified, it is the
2714 number of elements allocated. If an alignment is specified, the value result
2715 of the allocation is guaranteed to be aligned to at least that boundary. If
2716 not specified, or if zero, the target can choose to align the allocation on any
2717 convenient boundary.</p>
2719 <p>'<tt>type</tt>' may be any sized type.</p>
2723 <p>Memory is allocated; a pointer is returned. '<tt>alloca</tt>'d
2724 memory is automatically released when the function returns. The '<tt>alloca</tt>'
2725 instruction is commonly used to represent automatic variables that must
2726 have an address available. When the function returns (either with the <tt><a
2727 href="#i_ret">ret</a></tt> or <tt><a href="#i_unwind">unwind</a></tt>
2728 instructions), the memory is reclaimed.</p>
2733 %ptr = alloca i32 <i>; yields {i32*}:ptr</i>
2734 %ptr = alloca i32, i32 4 <i>; yields {i32*}:ptr</i>
2735 %ptr = alloca i32, i32 4, align 1024 <i>; yields {i32*}:ptr</i>
2736 %ptr = alloca i32, align 1024 <i>; yields {i32*}:ptr</i>
2740 <!-- _______________________________________________________________________ -->
2741 <div class="doc_subsubsection"> <a name="i_load">'<tt>load</tt>'
2742 Instruction</a> </div>
2743 <div class="doc_text">
2745 <pre> <result> = load <ty>* <pointer>[, align <alignment>]<br> <result> = volatile load <ty>* <pointer>[, align <alignment>]<br></pre>
2747 <p>The '<tt>load</tt>' instruction is used to read from memory.</p>
2749 <p>The argument to the '<tt>load</tt>' instruction specifies the memory
2750 address from which to load. The pointer must point to a <a
2751 href="#t_firstclass">first class</a> type. If the <tt>load</tt> is
2752 marked as <tt>volatile</tt>, then the optimizer is not allowed to modify
2753 the number or order of execution of this <tt>load</tt> with other
2754 volatile <tt>load</tt> and <tt><a href="#i_store">store</a></tt>
2757 <p>The location of memory pointed to is loaded.</p>
2759 <pre> %ptr = <a href="#i_alloca">alloca</a> i32 <i>; yields {i32*}:ptr</i>
2761 href="#i_store">store</a> i32 3, i32* %ptr <i>; yields {void}</i>
2762 %val = load i32* %ptr <i>; yields {i32}:val = i32 3</i>
2765 <!-- _______________________________________________________________________ -->
2766 <div class="doc_subsubsection"> <a name="i_store">'<tt>store</tt>'
2767 Instruction</a> </div>
2768 <div class="doc_text">
2770 <pre> store <ty> <value>, <ty>* <pointer>[, align <alignment>] <i>; yields {void}</i>
2771 volatile store <ty> <value>, <ty>* <pointer>[, align <alignment>] <i>; yields {void}</i>
2774 <p>The '<tt>store</tt>' instruction is used to write to memory.</p>
2776 <p>There are two arguments to the '<tt>store</tt>' instruction: a value
2777 to store and an address at which to store it. The type of the '<tt><pointer></tt>'
2778 operand must be a pointer to the type of the '<tt><value></tt>'
2779 operand. If the <tt>store</tt> is marked as <tt>volatile</tt>, then the
2780 optimizer is not allowed to modify the number or order of execution of
2781 this <tt>store</tt> with other volatile <tt>load</tt> and <tt><a
2782 href="#i_store">store</a></tt> instructions.</p>
2784 <p>The contents of memory are updated to contain '<tt><value></tt>'
2785 at the location specified by the '<tt><pointer></tt>' operand.</p>
2787 <pre> %ptr = <a href="#i_alloca">alloca</a> i32 <i>; yields {i32*}:ptr</i>
2789 href="#i_store">store</a> i32 3, i32* %ptr <i>; yields {void}</i>
2790 %val = load i32* %ptr <i>; yields {i32}:val = i32 3</i>
2794 <!-- _______________________________________________________________________ -->
2795 <div class="doc_subsubsection">
2796 <a name="i_getelementptr">'<tt>getelementptr</tt>' Instruction</a>
2799 <div class="doc_text">
2802 <result> = getelementptr <ty>* <ptrval>{, <ty> <idx>}*
2808 The '<tt>getelementptr</tt>' instruction is used to get the address of a
2809 subelement of an aggregate data structure.</p>
2813 <p>This instruction takes a list of integer operands that indicate what
2814 elements of the aggregate object to index to. The actual types of the arguments
2815 provided depend on the type of the first pointer argument. The
2816 '<tt>getelementptr</tt>' instruction is used to index down through the type
2817 levels of a structure or to a specific index in an array. When indexing into a
2818 structure, only <tt>i32</tt> integer constants are allowed. When indexing
2819 into an array or pointer, only integers of 32 or 64 bits are allowed, and will
2820 be sign extended to 64-bit values.</p>
2822 <p>For example, let's consider a C code fragment and how it gets
2823 compiled to LLVM:</p>
2825 <div class="doc_code">
2838 int *foo(struct ST *s) {
2839 return &s[1].Z.B[5][13];
2844 <p>The LLVM code generated by the GCC frontend is:</p>
2846 <div class="doc_code">
2848 %RT = type { i8 , [10 x [20 x i32]], i8 }
2849 %ST = type { i32, double, %RT }
2851 define i32* %foo(%ST* %s) {
2853 %reg = getelementptr %ST* %s, i32 1, i32 2, i32 1, i32 5, i32 13
2861 <p>The index types specified for the '<tt>getelementptr</tt>' instruction depend
2862 on the pointer type that is being indexed into. <a href="#t_pointer">Pointer</a>
2863 and <a href="#t_array">array</a> types can use a 32-bit or 64-bit
2864 <a href="#t_integer">integer</a> type but the value will always be sign extended
2865 to 64-bits. <a href="#t_struct">Structure</a> types require <tt>i32</tt>
2866 <b>constants</b>.</p>
2868 <p>In the example above, the first index is indexing into the '<tt>%ST*</tt>'
2869 type, which is a pointer, yielding a '<tt>%ST</tt>' = '<tt>{ i32, double, %RT
2870 }</tt>' type, a structure. The second index indexes into the third element of
2871 the structure, yielding a '<tt>%RT</tt>' = '<tt>{ i8 , [10 x [20 x i32]],
2872 i8 }</tt>' type, another structure. The third index indexes into the second
2873 element of the structure, yielding a '<tt>[10 x [20 x i32]]</tt>' type, an
2874 array. The two dimensions of the array are subscripted into, yielding an
2875 '<tt>i32</tt>' type. The '<tt>getelementptr</tt>' instruction returns a pointer
2876 to this element, thus computing a value of '<tt>i32*</tt>' type.</p>
2878 <p>Note that it is perfectly legal to index partially through a
2879 structure, returning a pointer to an inner element. Because of this,
2880 the LLVM code for the given testcase is equivalent to:</p>
2883 define i32* %foo(%ST* %s) {
2884 %t1 = getelementptr %ST* %s, i32 1 <i>; yields %ST*:%t1</i>
2885 %t2 = getelementptr %ST* %t1, i32 0, i32 2 <i>; yields %RT*:%t2</i>
2886 %t3 = getelementptr %RT* %t2, i32 0, i32 1 <i>; yields [10 x [20 x i32]]*:%t3</i>
2887 %t4 = getelementptr [10 x [20 x i32]]* %t3, i32 0, i32 5 <i>; yields [20 x i32]*:%t4</i>
2888 %t5 = getelementptr [20 x i32]* %t4, i32 0, i32 13 <i>; yields i32*:%t5</i>
2893 <p>Note that it is undefined to access an array out of bounds: array and
2894 pointer indexes must always be within the defined bounds of the array type.
2895 The one exception for this rules is zero length arrays. These arrays are
2896 defined to be accessible as variable length arrays, which requires access
2897 beyond the zero'th element.</p>
2899 <p>The getelementptr instruction is often confusing. For some more insight
2900 into how it works, see <a href="GetElementPtr.html">the getelementptr
2906 <i>; yields [12 x i8]*:aptr</i>
2907 %aptr = getelementptr {i32, [12 x i8]}* %sptr, i64 0, i32 1
2911 <!-- ======================================================================= -->
2912 <div class="doc_subsection"> <a name="convertops">Conversion Operations</a>
2914 <div class="doc_text">
2915 <p>The instructions in this category are the conversion instructions (casting)
2916 which all take a single operand and a type. They perform various bit conversions
2920 <!-- _______________________________________________________________________ -->
2921 <div class="doc_subsubsection">
2922 <a name="i_trunc">'<tt>trunc .. to</tt>' Instruction</a>
2924 <div class="doc_text">
2928 <result> = trunc <ty> <value> to <ty2> <i>; yields ty2</i>
2933 The '<tt>trunc</tt>' instruction truncates its operand to the type <tt>ty2</tt>.
2938 The '<tt>trunc</tt>' instruction takes a <tt>value</tt> to trunc, which must
2939 be an <a href="#t_integer">integer</a> type, and a type that specifies the size
2940 and type of the result, which must be an <a href="#t_integer">integer</a>
2941 type. The bit size of <tt>value</tt> must be larger than the bit size of
2942 <tt>ty2</tt>. Equal sized types are not allowed.</p>
2946 The '<tt>trunc</tt>' instruction truncates the high order bits in <tt>value</tt>
2947 and converts the remaining bits to <tt>ty2</tt>. Since the source size must be
2948 larger than the destination size, <tt>trunc</tt> cannot be a <i>no-op cast</i>.
2949 It will always truncate bits.</p>
2953 %X = trunc i32 257 to i8 <i>; yields i8:1</i>
2954 %Y = trunc i32 123 to i1 <i>; yields i1:true</i>
2955 %Y = trunc i32 122 to i1 <i>; yields i1:false</i>
2959 <!-- _______________________________________________________________________ -->
2960 <div class="doc_subsubsection">
2961 <a name="i_zext">'<tt>zext .. to</tt>' Instruction</a>
2963 <div class="doc_text">
2967 <result> = zext <ty> <value> to <ty2> <i>; yields ty2</i>
2971 <p>The '<tt>zext</tt>' instruction zero extends its operand to type
2976 <p>The '<tt>zext</tt>' instruction takes a value to cast, which must be of
2977 <a href="#t_integer">integer</a> type, and a type to cast it to, which must
2978 also be of <a href="#t_integer">integer</a> type. The bit size of the
2979 <tt>value</tt> must be smaller than the bit size of the destination type,
2983 <p>The <tt>zext</tt> fills the high order bits of the <tt>value</tt> with zero
2984 bits until it reaches the size of the destination type, <tt>ty2</tt>.</p>
2986 <p>When zero extending from i1, the result will always be either 0 or 1.</p>
2990 %X = zext i32 257 to i64 <i>; yields i64:257</i>
2991 %Y = zext i1 true to i32 <i>; yields i32:1</i>
2995 <!-- _______________________________________________________________________ -->
2996 <div class="doc_subsubsection">
2997 <a name="i_sext">'<tt>sext .. to</tt>' Instruction</a>
2999 <div class="doc_text">
3003 <result> = sext <ty> <value> to <ty2> <i>; yields ty2</i>
3007 <p>The '<tt>sext</tt>' sign extends <tt>value</tt> to the type <tt>ty2</tt>.</p>
3011 The '<tt>sext</tt>' instruction takes a value to cast, which must be of
3012 <a href="#t_integer">integer</a> type, and a type to cast it to, which must
3013 also be of <a href="#t_integer">integer</a> type. The bit size of the
3014 <tt>value</tt> must be smaller than the bit size of the destination type,
3019 The '<tt>sext</tt>' instruction performs a sign extension by copying the sign
3020 bit (highest order bit) of the <tt>value</tt> until it reaches the bit size of
3021 the type <tt>ty2</tt>.</p>
3023 <p>When sign extending from i1, the extension always results in -1 or 0.</p>
3027 %X = sext i8 -1 to i16 <i>; yields i16 :65535</i>
3028 %Y = sext i1 true to i32 <i>; yields i32:-1</i>
3032 <!-- _______________________________________________________________________ -->
3033 <div class="doc_subsubsection">
3034 <a name="i_fptrunc">'<tt>fptrunc .. to</tt>' Instruction</a>
3037 <div class="doc_text">
3042 <result> = fptrunc <ty> <value> to <ty2> <i>; yields ty2</i>
3046 <p>The '<tt>fptrunc</tt>' instruction truncates <tt>value</tt> to type
3051 <p>The '<tt>fptrunc</tt>' instruction takes a <a href="#t_floating">floating
3052 point</a> value to cast and a <a href="#t_floating">floating point</a> type to
3053 cast it to. The size of <tt>value</tt> must be larger than the size of
3054 <tt>ty2</tt>. This implies that <tt>fptrunc</tt> cannot be used to make a
3055 <i>no-op cast</i>.</p>
3058 <p> The '<tt>fptrunc</tt>' instruction truncates a <tt>value</tt> from a larger
3059 <a href="#t_floating">floating point</a> type to a smaller
3060 <a href="#t_floating">floating point</a> type. If the value cannot fit within
3061 the destination type, <tt>ty2</tt>, then the results are undefined.</p>
3065 %X = fptrunc double 123.0 to float <i>; yields float:123.0</i>
3066 %Y = fptrunc double 1.0E+300 to float <i>; yields undefined</i>
3070 <!-- _______________________________________________________________________ -->
3071 <div class="doc_subsubsection">
3072 <a name="i_fpext">'<tt>fpext .. to</tt>' Instruction</a>
3074 <div class="doc_text">
3078 <result> = fpext <ty> <value> to <ty2> <i>; yields ty2</i>
3082 <p>The '<tt>fpext</tt>' extends a floating point <tt>value</tt> to a larger
3083 floating point value.</p>
3086 <p>The '<tt>fpext</tt>' instruction takes a
3087 <a href="#t_floating">floating point</a> <tt>value</tt> to cast,
3088 and a <a href="#t_floating">floating point</a> type to cast it to. The source
3089 type must be smaller than the destination type.</p>
3092 <p>The '<tt>fpext</tt>' instruction extends the <tt>value</tt> from a smaller
3093 <a href="#t_floating">floating point</a> type to a larger
3094 <a href="#t_floating">floating point</a> type. The <tt>fpext</tt> cannot be
3095 used to make a <i>no-op cast</i> because it always changes bits. Use
3096 <tt>bitcast</tt> to make a <i>no-op cast</i> for a floating point cast.</p>
3100 %X = fpext float 3.1415 to double <i>; yields double:3.1415</i>
3101 %Y = fpext float 1.0 to float <i>; yields float:1.0 (no-op)</i>
3105 <!-- _______________________________________________________________________ -->
3106 <div class="doc_subsubsection">
3107 <a name="i_fptoui">'<tt>fptoui .. to</tt>' Instruction</a>
3109 <div class="doc_text">
3113 <result> = fptoui <ty> <value> to <ty2> <i>; yields ty2</i>
3117 <p>The '<tt>fptoui</tt>' converts a floating point <tt>value</tt> to its
3118 unsigned integer equivalent of type <tt>ty2</tt>.
3122 <p>The '<tt>fptoui</tt>' instruction takes a value to cast, which must be a
3123 <a href="#t_floating">floating point</a> value, and a type to cast it to, which
3124 must be an <a href="#t_integer">integer</a> type.</p>
3127 <p> The '<tt>fptoui</tt>' instruction converts its
3128 <a href="#t_floating">floating point</a> operand into the nearest (rounding
3129 towards zero) unsigned integer value. If the value cannot fit in <tt>ty2</tt>,
3130 the results are undefined.</p>
3134 %X = fptoui double 123.0 to i32 <i>; yields i32:123</i>
3135 %Y = fptoui float 1.0E+300 to i1 <i>; yields undefined:1</i>
3136 %X = fptoui float 1.04E+17 to i8 <i>; yields undefined:1</i>
3140 <!-- _______________________________________________________________________ -->
3141 <div class="doc_subsubsection">
3142 <a name="i_fptosi">'<tt>fptosi .. to</tt>' Instruction</a>
3144 <div class="doc_text">
3148 <result> = fptosi <ty> <value> to <ty2> <i>; yields ty2</i>
3152 <p>The '<tt>fptosi</tt>' instruction converts
3153 <a href="#t_floating">floating point</a> <tt>value</tt> to type <tt>ty2</tt>.
3158 <p> The '<tt>fptosi</tt>' instruction takes a value to cast, which must be a
3159 <a href="#t_floating">floating point</a> value, and a type to cast it to, which
3160 must also be an <a href="#t_integer">integer</a> type.</p>
3163 <p>The '<tt>fptosi</tt>' instruction converts its
3164 <a href="#t_floating">floating point</a> operand into the nearest (rounding
3165 towards zero) signed integer value. If the value cannot fit in <tt>ty2</tt>,
3166 the results are undefined.</p>
3170 %X = fptosi double -123.0 to i32 <i>; yields i32:-123</i>
3171 %Y = fptosi float 1.0E-247 to i1 <i>; yields undefined:1</i>
3172 %X = fptosi float 1.04E+17 to i8 <i>; yields undefined:1</i>
3176 <!-- _______________________________________________________________________ -->
3177 <div class="doc_subsubsection">
3178 <a name="i_uitofp">'<tt>uitofp .. to</tt>' Instruction</a>
3180 <div class="doc_text">
3184 <result> = uitofp <ty> <value> to <ty2> <i>; yields ty2</i>
3188 <p>The '<tt>uitofp</tt>' instruction regards <tt>value</tt> as an unsigned
3189 integer and converts that value to the <tt>ty2</tt> type.</p>
3193 <p>The '<tt>uitofp</tt>' instruction takes a value to cast, which must be an
3194 <a href="#t_integer">integer</a> value, and a type to cast it to, which must
3195 be a <a href="#t_floating">floating point</a> type.</p>
3198 <p>The '<tt>uitofp</tt>' instruction interprets its operand as an unsigned
3199 integer quantity and converts it to the corresponding floating point value. If
3200 the value cannot fit in the floating point value, the results are undefined.</p>
3205 %X = uitofp i32 257 to float <i>; yields float:257.0</i>
3206 %Y = uitofp i8 -1 to double <i>; yields double:255.0</i>
3210 <!-- _______________________________________________________________________ -->
3211 <div class="doc_subsubsection">
3212 <a name="i_sitofp">'<tt>sitofp .. to</tt>' Instruction</a>
3214 <div class="doc_text">
3218 <result> = sitofp <ty> <value> to <ty2> <i>; yields ty2</i>
3222 <p>The '<tt>sitofp</tt>' instruction regards <tt>value</tt> as a signed
3223 integer and converts that value to the <tt>ty2</tt> type.</p>
3226 <p>The '<tt>sitofp</tt>' instruction takes a value to cast, which must be an
3227 <a href="#t_integer">integer</a> value, and a type to cast it to, which must be
3228 a <a href="#t_floating">floating point</a> type.</p>
3231 <p>The '<tt>sitofp</tt>' instruction interprets its operand as a signed
3232 integer quantity and converts it to the corresponding floating point value. If
3233 the value cannot fit in the floating point value, the results are undefined.</p>
3237 %X = sitofp i32 257 to float <i>; yields float:257.0</i>
3238 %Y = sitofp i8 -1 to double <i>; yields double:-1.0</i>
3242 <!-- _______________________________________________________________________ -->
3243 <div class="doc_subsubsection">
3244 <a name="i_ptrtoint">'<tt>ptrtoint .. to</tt>' Instruction</a>
3246 <div class="doc_text">
3250 <result> = ptrtoint <ty> <value> to <ty2> <i>; yields ty2</i>
3254 <p>The '<tt>ptrtoint</tt>' instruction converts the pointer <tt>value</tt> to
3255 the integer type <tt>ty2</tt>.</p>
3258 <p>The '<tt>ptrtoint</tt>' instruction takes a <tt>value</tt> to cast, which
3259 must be a <a href="#t_pointer">pointer</a> value, and a type to cast it to
3260 <tt>ty2</tt>, which must be an <a href="#t_integer">integer</a> type.
3263 <p>The '<tt>ptrtoint</tt>' instruction converts <tt>value</tt> to integer type
3264 <tt>ty2</tt> by interpreting the pointer value as an integer and either
3265 truncating or zero extending that value to the size of the integer type. If
3266 <tt>value</tt> is smaller than <tt>ty2</tt> then a zero extension is done. If
3267 <tt>value</tt> is larger than <tt>ty2</tt> then a truncation is done. If they
3268 are the same size, then nothing is done (<i>no-op cast</i>) other than a type
3273 %X = ptrtoint i32* %X to i8 <i>; yields truncation on 32-bit architecture</i>
3274 %Y = ptrtoint i32* %x to i64 <i>; yields zero extension on 32-bit architecture</i>
3278 <!-- _______________________________________________________________________ -->
3279 <div class="doc_subsubsection">
3280 <a name="i_inttoptr">'<tt>inttoptr .. to</tt>' Instruction</a>
3282 <div class="doc_text">
3286 <result> = inttoptr <ty> <value> to <ty2> <i>; yields ty2</i>
3290 <p>The '<tt>inttoptr</tt>' instruction converts an integer <tt>value</tt> to
3291 a pointer type, <tt>ty2</tt>.</p>
3294 <p>The '<tt>inttoptr</tt>' instruction takes an <a href="#t_integer">integer</a>
3295 value to cast, and a type to cast it to, which must be a
3296 <a href="#t_pointer">pointer</a> type.
3299 <p>The '<tt>inttoptr</tt>' instruction converts <tt>value</tt> to type
3300 <tt>ty2</tt> by applying either a zero extension or a truncation depending on
3301 the size of the integer <tt>value</tt>. If <tt>value</tt> is larger than the
3302 size of a pointer then a truncation is done. If <tt>value</tt> is smaller than
3303 the size of a pointer then a zero extension is done. If they are the same size,
3304 nothing is done (<i>no-op cast</i>).</p>
3308 %X = inttoptr i32 255 to i32* <i>; yields zero extension on 64-bit architecture</i>
3309 %X = inttoptr i32 255 to i32* <i>; yields no-op on 32-bit architecture</i>
3310 %Y = inttoptr i64 0 to i32* <i>; yields truncation on 32-bit architecture</i>
3314 <!-- _______________________________________________________________________ -->
3315 <div class="doc_subsubsection">
3316 <a name="i_bitcast">'<tt>bitcast .. to</tt>' Instruction</a>
3318 <div class="doc_text">
3322 <result> = bitcast <ty> <value> to <ty2> <i>; yields ty2</i>
3326 <p>The '<tt>bitcast</tt>' instruction converts <tt>value</tt> to type
3327 <tt>ty2</tt> without changing any bits.</p>
3330 <p>The '<tt>bitcast</tt>' instruction takes a value to cast, which must be
3331 a first class value, and a type to cast it to, which must also be a <a
3332 href="#t_firstclass">first class</a> type. The bit sizes of <tt>value</tt>
3333 and the destination type, <tt>ty2</tt>, must be identical. If the source
3334 type is a pointer, the destination type must also be a pointer.</p>
3337 <p>The '<tt>bitcast</tt>' instruction converts <tt>value</tt> to type
3338 <tt>ty2</tt>. It is always a <i>no-op cast</i> because no bits change with
3339 this conversion. The conversion is done as if the <tt>value</tt> had been
3340 stored to memory and read back as type <tt>ty2</tt>. Pointer types may only be
3341 converted to other pointer types with this instruction. To convert pointers to
3342 other types, use the <a href="#i_inttoptr">inttoptr</a> or
3343 <a href="#i_ptrtoint">ptrtoint</a> instructions first.</p>
3347 %X = bitcast i8 255 to i8 <i>; yields i8 :-1</i>
3348 %Y = bitcast i32* %x to sint* <i>; yields sint*:%x</i>
3349 %Z = bitcast <2xint> %V to i64; <i>; yields i64: %V</i>
3353 <!-- ======================================================================= -->
3354 <div class="doc_subsection"> <a name="otherops">Other Operations</a> </div>
3355 <div class="doc_text">
3356 <p>The instructions in this category are the "miscellaneous"
3357 instructions, which defy better classification.</p>
3360 <!-- _______________________________________________________________________ -->
3361 <div class="doc_subsubsection"><a name="i_icmp">'<tt>icmp</tt>' Instruction</a>
3363 <div class="doc_text">
3365 <pre> <result> = icmp <cond> <ty> <var1>, <var2> <i>; yields {i1}:result</i>
3368 <p>The '<tt>icmp</tt>' instruction returns a boolean value based on comparison
3369 of its two integer operands.</p>
3371 <p>The '<tt>icmp</tt>' instruction takes three operands. The first operand is
3372 the condition code indicating the kind of comparison to perform. It is not
3373 a value, just a keyword. The possible condition code are:
3375 <li><tt>eq</tt>: equal</li>
3376 <li><tt>ne</tt>: not equal </li>
3377 <li><tt>ugt</tt>: unsigned greater than</li>
3378 <li><tt>uge</tt>: unsigned greater or equal</li>
3379 <li><tt>ult</tt>: unsigned less than</li>
3380 <li><tt>ule</tt>: unsigned less or equal</li>
3381 <li><tt>sgt</tt>: signed greater than</li>
3382 <li><tt>sge</tt>: signed greater or equal</li>
3383 <li><tt>slt</tt>: signed less than</li>
3384 <li><tt>sle</tt>: signed less or equal</li>
3386 <p>The remaining two arguments must be <a href="#t_integer">integer</a> or
3387 <a href="#t_pointer">pointer</a> typed. They must also be identical types.</p>
3389 <p>The '<tt>icmp</tt>' compares <tt>var1</tt> and <tt>var2</tt> according to
3390 the condition code given as <tt>cond</tt>. The comparison performed always
3391 yields a <a href="#t_primitive">i1</a> result, as follows:
3393 <li><tt>eq</tt>: yields <tt>true</tt> if the operands are equal,
3394 <tt>false</tt> otherwise. No sign interpretation is necessary or performed.
3396 <li><tt>ne</tt>: yields <tt>true</tt> if the operands are unequal,
3397 <tt>false</tt> otherwise. No sign interpretation is necessary or performed.
3398 <li><tt>ugt</tt>: interprets the operands as unsigned values and yields
3399 <tt>true</tt> if <tt>var1</tt> is greater than <tt>var2</tt>.</li>
3400 <li><tt>uge</tt>: interprets the operands as unsigned values and yields
3401 <tt>true</tt> if <tt>var1</tt> is greater than or equal to <tt>var2</tt>.</li>
3402 <li><tt>ult</tt>: interprets the operands as unsigned values and yields
3403 <tt>true</tt> if <tt>var1</tt> is less than <tt>var2</tt>.</li>
3404 <li><tt>ule</tt>: interprets the operands as unsigned values and yields
3405 <tt>true</tt> if <tt>var1</tt> is less than or equal to <tt>var2</tt>.</li>
3406 <li><tt>sgt</tt>: interprets the operands as signed values and yields
3407 <tt>true</tt> if <tt>var1</tt> is greater than <tt>var2</tt>.</li>
3408 <li><tt>sge</tt>: interprets the operands as signed values and yields
3409 <tt>true</tt> if <tt>var1</tt> is greater than or equal to <tt>var2</tt>.</li>
3410 <li><tt>slt</tt>: interprets the operands as signed values and yields
3411 <tt>true</tt> if <tt>var1</tt> is less than <tt>var2</tt>.</li>
3412 <li><tt>sle</tt>: interprets the operands as signed values and yields
3413 <tt>true</tt> if <tt>var1</tt> is less than or equal to <tt>var2</tt>.</li>
3415 <p>If the operands are <a href="#t_pointer">pointer</a> typed, the pointer
3416 values are compared as if they were integers.</p>
3419 <pre> <result> = icmp eq i32 4, 5 <i>; yields: result=false</i>
3420 <result> = icmp ne float* %X, %X <i>; yields: result=false</i>
3421 <result> = icmp ult i16 4, 5 <i>; yields: result=true</i>
3422 <result> = icmp sgt i16 4, 5 <i>; yields: result=false</i>
3423 <result> = icmp ule i16 -4, 5 <i>; yields: result=false</i>
3424 <result> = icmp sge i16 4, 5 <i>; yields: result=false</i>
3428 <!-- _______________________________________________________________________ -->
3429 <div class="doc_subsubsection"><a name="i_fcmp">'<tt>fcmp</tt>' Instruction</a>
3431 <div class="doc_text">
3433 <pre> <result> = fcmp <cond> <ty> <var1>, <var2> <i>; yields {i1}:result</i>
3436 <p>The '<tt>fcmp</tt>' instruction returns a boolean value based on comparison
3437 of its floating point operands.</p>
3439 <p>The '<tt>fcmp</tt>' instruction takes three operands. The first operand is
3440 the condition code indicating the kind of comparison to perform. It is not
3441 a value, just a keyword. The possible condition code are:
3443 <li><tt>false</tt>: no comparison, always returns false</li>
3444 <li><tt>oeq</tt>: ordered and equal</li>
3445 <li><tt>ogt</tt>: ordered and greater than </li>
3446 <li><tt>oge</tt>: ordered and greater than or equal</li>
3447 <li><tt>olt</tt>: ordered and less than </li>
3448 <li><tt>ole</tt>: ordered and less than or equal</li>
3449 <li><tt>one</tt>: ordered and not equal</li>
3450 <li><tt>ord</tt>: ordered (no nans)</li>
3451 <li><tt>ueq</tt>: unordered or equal</li>
3452 <li><tt>ugt</tt>: unordered or greater than </li>
3453 <li><tt>uge</tt>: unordered or greater than or equal</li>
3454 <li><tt>ult</tt>: unordered or less than </li>
3455 <li><tt>ule</tt>: unordered or less than or equal</li>
3456 <li><tt>une</tt>: unordered or not equal</li>
3457 <li><tt>uno</tt>: unordered (either nans)</li>
3458 <li><tt>true</tt>: no comparison, always returns true</li>
3460 <p><i>Ordered</i> means that neither operand is a QNAN while
3461 <i>unordered</i> means that either operand may be a QNAN.</p>
3462 <p>The <tt>val1</tt> and <tt>val2</tt> arguments must be
3463 <a href="#t_floating">floating point</a> typed. They must have identical
3466 <p>The '<tt>fcmp</tt>' compares <tt>var1</tt> and <tt>var2</tt> according to
3467 the condition code given as <tt>cond</tt>. The comparison performed always
3468 yields a <a href="#t_primitive">i1</a> result, as follows:
3470 <li><tt>false</tt>: always yields <tt>false</tt>, regardless of operands.</li>
3471 <li><tt>oeq</tt>: yields <tt>true</tt> if both operands are not a QNAN and
3472 <tt>var1</tt> is equal to <tt>var2</tt>.</li>
3473 <li><tt>ogt</tt>: yields <tt>true</tt> if both operands are not a QNAN and
3474 <tt>var1</tt> is greather than <tt>var2</tt>.</li>
3475 <li><tt>oge</tt>: yields <tt>true</tt> if both operands are not a QNAN and
3476 <tt>var1</tt> is greater than or equal to <tt>var2</tt>.</li>
3477 <li><tt>olt</tt>: yields <tt>true</tt> if both operands are not a QNAN and
3478 <tt>var1</tt> is less than <tt>var2</tt>.</li>
3479 <li><tt>ole</tt>: yields <tt>true</tt> if both operands are not a QNAN and
3480 <tt>var1</tt> is less than or equal to <tt>var2</tt>.</li>
3481 <li><tt>one</tt>: yields <tt>true</tt> if both operands are not a QNAN and
3482 <tt>var1</tt> is not equal to <tt>var2</tt>.</li>
3483 <li><tt>ord</tt>: yields <tt>true</tt> if both operands are not a QNAN.</li>
3484 <li><tt>ueq</tt>: yields <tt>true</tt> if either operand is a QNAN or
3485 <tt>var1</tt> is equal to <tt>var2</tt>.</li>
3486 <li><tt>ugt</tt>: yields <tt>true</tt> if either operand is a QNAN or
3487 <tt>var1</tt> is greater than <tt>var2</tt>.</li>
3488 <li><tt>uge</tt>: yields <tt>true</tt> if either operand is a QNAN or
3489 <tt>var1</tt> is greater than or equal to <tt>var2</tt>.</li>
3490 <li><tt>ult</tt>: yields <tt>true</tt> if either operand is a QNAN or
3491 <tt>var1</tt> is less than <tt>var2</tt>.</li>
3492 <li><tt>ule</tt>: yields <tt>true</tt> if either operand is a QNAN or
3493 <tt>var1</tt> is less than or equal to <tt>var2</tt>.</li>
3494 <li><tt>une</tt>: yields <tt>true</tt> if either operand is a QNAN or
3495 <tt>var1</tt> is not equal to <tt>var2</tt>.</li>
3496 <li><tt>uno</tt>: yields <tt>true</tt> if either operand is a QNAN.</li>
3497 <li><tt>true</tt>: always yields <tt>true</tt>, regardless of operands.</li>
3501 <pre> <result> = fcmp oeq float 4.0, 5.0 <i>; yields: result=false</i>
3502 <result> = icmp one float 4.0, 5.0 <i>; yields: result=true</i>
3503 <result> = icmp olt float 4.0, 5.0 <i>; yields: result=true</i>
3504 <result> = icmp ueq double 1.0, 2.0 <i>; yields: result=false</i>
3508 <!-- _______________________________________________________________________ -->
3509 <div class="doc_subsubsection"> <a name="i_phi">'<tt>phi</tt>'
3510 Instruction</a> </div>
3511 <div class="doc_text">
3513 <pre> <result> = phi <ty> [ <val0>, <label0>], ...<br></pre>
3515 <p>The '<tt>phi</tt>' instruction is used to implement the φ node in
3516 the SSA graph representing the function.</p>
3518 <p>The type of the incoming values is specified with the first type
3519 field. After this, the '<tt>phi</tt>' instruction takes a list of pairs
3520 as arguments, with one pair for each predecessor basic block of the
3521 current block. Only values of <a href="#t_firstclass">first class</a>
3522 type may be used as the value arguments to the PHI node. Only labels
3523 may be used as the label arguments.</p>
3524 <p>There must be no non-phi instructions between the start of a basic
3525 block and the PHI instructions: i.e. PHI instructions must be first in
3528 <p>At runtime, the '<tt>phi</tt>' instruction logically takes on the value
3529 specified by the pair corresponding to the predecessor basic block that executed
3530 just prior to the current block.</p>
3532 <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>
3535 <!-- _______________________________________________________________________ -->
3536 <div class="doc_subsubsection">
3537 <a name="i_select">'<tt>select</tt>' Instruction</a>
3540 <div class="doc_text">
3545 <result> = select i1 <cond>, <ty> <val1>, <ty> <val2> <i>; yields ty</i>
3551 The '<tt>select</tt>' instruction is used to choose one value based on a
3552 condition, without branching.
3559 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.
3565 If the boolean condition evaluates to true, the instruction returns the first
3566 value argument; otherwise, it returns the second value argument.
3572 %X = select i1 true, i8 17, i8 42 <i>; yields i8:17</i>
3577 <!-- _______________________________________________________________________ -->
3578 <div class="doc_subsubsection">
3579 <a name="i_call">'<tt>call</tt>' Instruction</a>
3582 <div class="doc_text">
3586 <result> = [tail] call [<a href="#callingconv">cconv</a>] <ty> [<fnty>*] <fnptrval>(<param list>)
3591 <p>The '<tt>call</tt>' instruction represents a simple function call.</p>
3595 <p>This instruction requires several arguments:</p>
3599 <p>The optional "tail" marker indicates whether the callee function accesses
3600 any allocas or varargs in the caller. If the "tail" marker is present, the
3601 function call is eligible for tail call optimization. Note that calls may
3602 be marked "tail" even if they do not occur before a <a
3603 href="#i_ret"><tt>ret</tt></a> instruction.
3606 <p>The optional "cconv" marker indicates which <a href="#callingconv">calling
3607 convention</a> the call should use. If none is specified, the call defaults
3608 to using C calling conventions.
3611 <p>'<tt>ty</tt>': the type of the call instruction itself which is also
3612 the type of the return value. Functions that return no value are marked
3613 <tt><a href="#t_void">void</a></tt>.</p>
3616 <p>'<tt>fnty</tt>': shall be the signature of the pointer to function
3617 value being invoked. The argument types must match the types implied by
3618 this signature. This type can be omitted if the function is not varargs
3619 and if the function type does not return a pointer to a function.</p>
3622 <p>'<tt>fnptrval</tt>': An LLVM value containing a pointer to a function to
3623 be invoked. In most cases, this is a direct function invocation, but
3624 indirect <tt>call</tt>s are just as possible, calling an arbitrary pointer
3625 to function value.</p>
3628 <p>'<tt>function args</tt>': argument list whose types match the
3629 function signature argument types. All arguments must be of
3630 <a href="#t_firstclass">first class</a> type. If the function signature
3631 indicates the function accepts a variable number of arguments, the extra
3632 arguments can be specified.</p>
3638 <p>The '<tt>call</tt>' instruction is used to cause control flow to
3639 transfer to a specified function, with its incoming arguments bound to
3640 the specified values. Upon a '<tt><a href="#i_ret">ret</a></tt>'
3641 instruction in the called function, control flow continues with the
3642 instruction after the function call, and the return value of the
3643 function is bound to the result argument. This is a simpler case of
3644 the <a href="#i_invoke">invoke</a> instruction.</p>
3649 %retval = call i32 @test(i32 %argc)
3650 call i32 (i8 *, ...)* @printf(i8 * %msg, i32 12, i8 42);
3651 %X = tail call i32 @foo()
3652 %Y = tail call <a href="#callingconv">fastcc</a> i32 @foo()
3653 %Z = call void %foo(i8 97 signext)
3658 <!-- _______________________________________________________________________ -->
3659 <div class="doc_subsubsection">
3660 <a name="i_va_arg">'<tt>va_arg</tt>' Instruction</a>
3663 <div class="doc_text">
3668 <resultval> = va_arg <va_list*> <arglist>, <argty>
3673 <p>The '<tt>va_arg</tt>' instruction is used to access arguments passed through
3674 the "variable argument" area of a function call. It is used to implement the
3675 <tt>va_arg</tt> macro in C.</p>
3679 <p>This instruction takes a <tt>va_list*</tt> value and the type of
3680 the argument. It returns a value of the specified argument type and
3681 increments the <tt>va_list</tt> to point to the next argument. The
3682 actual type of <tt>va_list</tt> is target specific.</p>
3686 <p>The '<tt>va_arg</tt>' instruction loads an argument of the specified
3687 type from the specified <tt>va_list</tt> and causes the
3688 <tt>va_list</tt> to point to the next argument. For more information,
3689 see the variable argument handling <a href="#int_varargs">Intrinsic
3692 <p>It is legal for this instruction to be called in a function which does not
3693 take a variable number of arguments, for example, the <tt>vfprintf</tt>
3696 <p><tt>va_arg</tt> is an LLVM instruction instead of an <a
3697 href="#intrinsics">intrinsic function</a> because it takes a type as an
3702 <p>See the <a href="#int_varargs">variable argument processing</a> section.</p>
3706 <!-- *********************************************************************** -->
3707 <div class="doc_section"> <a name="intrinsics">Intrinsic Functions</a> </div>
3708 <!-- *********************************************************************** -->
3710 <div class="doc_text">
3712 <p>LLVM supports the notion of an "intrinsic function". These functions have
3713 well known names and semantics and are required to follow certain restrictions.
3714 Overall, these intrinsics represent an extension mechanism for the LLVM
3715 language that does not require changing all of the transformations in LLVM when
3716 adding to the language (or the bitcode reader/writer, the parser, etc...).</p>
3718 <p>Intrinsic function names must all start with an "<tt>llvm.</tt>" prefix. This
3719 prefix is reserved in LLVM for intrinsic names; thus, function names may not
3720 begin with this prefix. Intrinsic functions must always be external functions:
3721 you cannot define the body of intrinsic functions. Intrinsic functions may
3722 only be used in call or invoke instructions: it is illegal to take the address
3723 of an intrinsic function. Additionally, because intrinsic functions are part
3724 of the LLVM language, it is required if any are added that they be documented
3727 <p>Some intrinsic functions can be overloaded, i.e., the intrinsic represents
3728 a family of functions that perform the same operation but on different data
3729 types. Because LLVM can represent over 8 million different integer types,
3730 overloading is used commonly to allow an intrinsic function to operate on any
3731 integer type. One or more of the argument types or the result type can be
3732 overloaded to accept any integer type. Argument types may also be defined as
3733 exactly matching a previous argument's type or the result type. This allows an
3734 intrinsic function which accepts multiple arguments, but needs all of them to
3735 be of the same type, to only be overloaded with respect to a single argument or
3738 <p>Overloaded intrinsics will have the names of its overloaded argument types
3739 encoded into its function name, each preceded by a period. Only those types
3740 which are overloaded result in a name suffix. Arguments whose type is matched
3741 against another type do not. For example, the <tt>llvm.ctpop</tt> function can
3742 take an integer of any width and returns an integer of exactly the same integer
3743 width. This leads to a family of functions such as
3744 <tt>i8 @llvm.ctpop.i8(i8 %val)</tt> and <tt>i29 @llvm.ctpop.i29(i29 %val)</tt>.
3745 Only one type, the return type, is overloaded, and only one type suffix is
3746 required. Because the argument's type is matched against the return type, it
3747 does not require its own name suffix.</p>
3749 <p>To learn how to add an intrinsic function, please see the
3750 <a href="ExtendingLLVM.html">Extending LLVM Guide</a>.
3755 <!-- ======================================================================= -->
3756 <div class="doc_subsection">
3757 <a name="int_varargs">Variable Argument Handling Intrinsics</a>
3760 <div class="doc_text">
3762 <p>Variable argument support is defined in LLVM with the <a
3763 href="#i_va_arg"><tt>va_arg</tt></a> instruction and these three
3764 intrinsic functions. These functions are related to the similarly
3765 named macros defined in the <tt><stdarg.h></tt> header file.</p>
3767 <p>All of these functions operate on arguments that use a
3768 target-specific value type "<tt>va_list</tt>". The LLVM assembly
3769 language reference manual does not define what this type is, so all
3770 transformations should be prepared to handle these functions regardless of
3773 <p>This example shows how the <a href="#i_va_arg"><tt>va_arg</tt></a>
3774 instruction and the variable argument handling intrinsic functions are
3777 <div class="doc_code">
3779 define i32 @test(i32 %X, ...) {
3780 ; Initialize variable argument processing
3782 %ap2 = bitcast i8** %ap to i8*
3783 call void @llvm.va_start(i8* %ap2)
3785 ; Read a single integer argument
3786 %tmp = va_arg i8** %ap, i32
3788 ; Demonstrate usage of llvm.va_copy and llvm.va_end
3790 %aq2 = bitcast i8** %aq to i8*
3791 call void @llvm.va_copy(i8* %aq2, i8* %ap2)
3792 call void @llvm.va_end(i8* %aq2)
3794 ; Stop processing of arguments.
3795 call void @llvm.va_end(i8* %ap2)
3799 declare void @llvm.va_start(i8*)
3800 declare void @llvm.va_copy(i8*, i8*)
3801 declare void @llvm.va_end(i8*)
3807 <!-- _______________________________________________________________________ -->
3808 <div class="doc_subsubsection">
3809 <a name="int_va_start">'<tt>llvm.va_start</tt>' Intrinsic</a>
3813 <div class="doc_text">
3815 <pre> declare void %llvm.va_start(i8* <arglist>)<br></pre>
3817 <P>The '<tt>llvm.va_start</tt>' intrinsic initializes
3818 <tt>*<arglist></tt> for subsequent use by <tt><a
3819 href="#i_va_arg">va_arg</a></tt>.</p>
3823 <P>The argument is a pointer to a <tt>va_list</tt> element to initialize.</p>
3827 <P>The '<tt>llvm.va_start</tt>' intrinsic works just like the <tt>va_start</tt>
3828 macro available in C. In a target-dependent way, it initializes the
3829 <tt>va_list</tt> element to which the argument points, so that the next call to
3830 <tt>va_arg</tt> will produce the first variable argument passed to the function.
3831 Unlike the C <tt>va_start</tt> macro, this intrinsic does not need to know the
3832 last argument of the function as the compiler can figure that out.</p>
3836 <!-- _______________________________________________________________________ -->
3837 <div class="doc_subsubsection">
3838 <a name="int_va_end">'<tt>llvm.va_end</tt>' Intrinsic</a>
3841 <div class="doc_text">
3843 <pre> declare void @llvm.va_end(i8* <arglist>)<br></pre>
3846 <p>The '<tt>llvm.va_end</tt>' intrinsic destroys <tt>*<arglist></tt>,
3847 which has been initialized previously with <tt><a href="#int_va_start">llvm.va_start</a></tt>
3848 or <tt><a href="#i_va_copy">llvm.va_copy</a></tt>.</p>
3852 <p>The argument is a pointer to a <tt>va_list</tt> to destroy.</p>
3856 <p>The '<tt>llvm.va_end</tt>' intrinsic works just like the <tt>va_end</tt>
3857 macro available in C. In a target-dependent way, it destroys the
3858 <tt>va_list</tt> element to which the argument points. Calls to <a
3859 href="#int_va_start"><tt>llvm.va_start</tt></a> and <a href="#int_va_copy">
3860 <tt>llvm.va_copy</tt></a> must be matched exactly with calls to
3861 <tt>llvm.va_end</tt>.</p>
3865 <!-- _______________________________________________________________________ -->
3866 <div class="doc_subsubsection">
3867 <a name="int_va_copy">'<tt>llvm.va_copy</tt>' Intrinsic</a>
3870 <div class="doc_text">
3875 declare void @llvm.va_copy(i8* <destarglist>, i8* <srcarglist>)
3880 <p>The '<tt>llvm.va_copy</tt>' intrinsic copies the current argument position
3881 from the source argument list to the destination argument list.</p>
3885 <p>The first argument is a pointer to a <tt>va_list</tt> element to initialize.
3886 The second argument is a pointer to a <tt>va_list</tt> element to copy from.</p>
3891 <p>The '<tt>llvm.va_copy</tt>' intrinsic works just like the <tt>va_copy</tt>
3892 macro available in C. In a target-dependent way, it copies the source
3893 <tt>va_list</tt> element into the destination <tt>va_list</tt> element. This
3894 intrinsic is necessary because the <tt><a href="#int_va_start">
3895 llvm.va_start</a></tt> intrinsic may be arbitrarily complex and require, for
3896 example, memory allocation.</p>
3900 <!-- ======================================================================= -->
3901 <div class="doc_subsection">
3902 <a name="int_gc">Accurate Garbage Collection Intrinsics</a>
3905 <div class="doc_text">
3908 LLVM support for <a href="GarbageCollection.html">Accurate Garbage
3909 Collection</a> requires the implementation and generation of these intrinsics.
3910 These intrinsics allow identification of <a href="#int_gcroot">GC roots on the
3911 stack</a>, as well as garbage collector implementations that require <a
3912 href="#int_gcread">read</a> and <a href="#int_gcwrite">write</a> barriers.
3913 Front-ends for type-safe garbage collected languages should generate these
3914 intrinsics to make use of the LLVM garbage collectors. For more details, see <a
3915 href="GarbageCollection.html">Accurate Garbage Collection with LLVM</a>.
3919 <!-- _______________________________________________________________________ -->
3920 <div class="doc_subsubsection">
3921 <a name="int_gcroot">'<tt>llvm.gcroot</tt>' Intrinsic</a>
3924 <div class="doc_text">
3929 declare void @llvm.gcroot(i8** %ptrloc, i8* %metadata)
3934 <p>The '<tt>llvm.gcroot</tt>' intrinsic declares the existence of a GC root to
3935 the code generator, and allows some metadata to be associated with it.</p>
3939 <p>The first argument specifies the address of a stack object that contains the
3940 root pointer. The second pointer (which must be either a constant or a global
3941 value address) contains the meta-data to be associated with the root.</p>
3945 <p>At runtime, a call to this intrinsics stores a null pointer into the "ptrloc"
3946 location. At compile-time, the code generator generates information to allow
3947 the runtime to find the pointer at GC safe points.
3953 <!-- _______________________________________________________________________ -->
3954 <div class="doc_subsubsection">
3955 <a name="int_gcread">'<tt>llvm.gcread</tt>' Intrinsic</a>
3958 <div class="doc_text">
3963 declare i8* @llvm.gcread(i8* %ObjPtr, i8** %Ptr)
3968 <p>The '<tt>llvm.gcread</tt>' intrinsic identifies reads of references from heap
3969 locations, allowing garbage collector implementations that require read
3974 <p>The second argument is the address to read from, which should be an address
3975 allocated from the garbage collector. The first object is a pointer to the
3976 start of the referenced object, if needed by the language runtime (otherwise
3981 <p>The '<tt>llvm.gcread</tt>' intrinsic has the same semantics as a load
3982 instruction, but may be replaced with substantially more complex code by the
3983 garbage collector runtime, as needed.</p>
3988 <!-- _______________________________________________________________________ -->
3989 <div class="doc_subsubsection">
3990 <a name="int_gcwrite">'<tt>llvm.gcwrite</tt>' Intrinsic</a>
3993 <div class="doc_text">
3998 declare void @llvm.gcwrite(i8* %P1, i8* %Obj, i8** %P2)
4003 <p>The '<tt>llvm.gcwrite</tt>' intrinsic identifies writes of references to heap
4004 locations, allowing garbage collector implementations that require write
4005 barriers (such as generational or reference counting collectors).</p>
4009 <p>The first argument is the reference to store, the second is the start of the
4010 object to store it to, and the third is the address of the field of Obj to
4011 store to. If the runtime does not require a pointer to the object, Obj may be
4016 <p>The '<tt>llvm.gcwrite</tt>' intrinsic has the same semantics as a store
4017 instruction, but may be replaced with substantially more complex code by the
4018 garbage collector runtime, as needed.</p>
4024 <!-- ======================================================================= -->
4025 <div class="doc_subsection">
4026 <a name="int_codegen">Code Generator Intrinsics</a>
4029 <div class="doc_text">
4031 These intrinsics are provided by LLVM to expose special features that may only
4032 be implemented with code generator support.
4037 <!-- _______________________________________________________________________ -->
4038 <div class="doc_subsubsection">
4039 <a name="int_returnaddress">'<tt>llvm.returnaddress</tt>' Intrinsic</a>
4042 <div class="doc_text">
4046 declare i8 *@llvm.returnaddress(i32 <level>)
4052 The '<tt>llvm.returnaddress</tt>' intrinsic attempts to compute a
4053 target-specific value indicating the return address of the current function
4054 or one of its callers.
4060 The argument to this intrinsic indicates which function to return the address
4061 for. Zero indicates the calling function, one indicates its caller, etc. The
4062 argument is <b>required</b> to be a constant integer value.
4068 The '<tt>llvm.returnaddress</tt>' intrinsic either returns a pointer indicating
4069 the return address of the specified call frame, or zero if it cannot be
4070 identified. The value returned by this intrinsic is likely to be incorrect or 0
4071 for arguments other than zero, so it should only be used for debugging purposes.
4075 Note that calling this intrinsic does not prevent function inlining or other
4076 aggressive transformations, so the value returned may not be that of the obvious
4077 source-language caller.
4082 <!-- _______________________________________________________________________ -->
4083 <div class="doc_subsubsection">
4084 <a name="int_frameaddress">'<tt>llvm.frameaddress</tt>' Intrinsic</a>
4087 <div class="doc_text">
4091 declare i8 *@llvm.frameaddress(i32 <level>)
4097 The '<tt>llvm.frameaddress</tt>' intrinsic attempts to return the
4098 target-specific frame pointer value for the specified stack frame.
4104 The argument to this intrinsic indicates which function to return the frame
4105 pointer for. Zero indicates the calling function, one indicates its caller,
4106 etc. The argument is <b>required</b> to be a constant integer value.
4112 The '<tt>llvm.frameaddress</tt>' intrinsic either returns a pointer indicating
4113 the frame address of the specified call frame, or zero if it cannot be
4114 identified. The value returned by this intrinsic is likely to be incorrect or 0
4115 for arguments other than zero, so it should only be used for debugging purposes.
4119 Note that calling this intrinsic does not prevent function inlining or other
4120 aggressive transformations, so the value returned may not be that of the obvious
4121 source-language caller.
4125 <!-- _______________________________________________________________________ -->
4126 <div class="doc_subsubsection">
4127 <a name="int_stacksave">'<tt>llvm.stacksave</tt>' Intrinsic</a>
4130 <div class="doc_text">
4134 declare i8 *@llvm.stacksave()
4140 The '<tt>llvm.stacksave</tt>' intrinsic is used to remember the current state of
4141 the function stack, for use with <a href="#int_stackrestore">
4142 <tt>llvm.stackrestore</tt></a>. This is useful for implementing language
4143 features like scoped automatic variable sized arrays in C99.
4149 This intrinsic returns a opaque pointer value that can be passed to <a
4150 href="#int_stackrestore"><tt>llvm.stackrestore</tt></a>. When an
4151 <tt>llvm.stackrestore</tt> intrinsic is executed with a value saved from
4152 <tt>llvm.stacksave</tt>, it effectively restores the state of the stack to the
4153 state it was in when the <tt>llvm.stacksave</tt> intrinsic executed. In
4154 practice, this pops any <a href="#i_alloca">alloca</a> blocks from the stack
4155 that were allocated after the <tt>llvm.stacksave</tt> was executed.
4160 <!-- _______________________________________________________________________ -->
4161 <div class="doc_subsubsection">
4162 <a name="int_stackrestore">'<tt>llvm.stackrestore</tt>' Intrinsic</a>
4165 <div class="doc_text">
4169 declare void @llvm.stackrestore(i8 * %ptr)
4175 The '<tt>llvm.stackrestore</tt>' intrinsic is used to restore the state of
4176 the function stack to the state it was in when the corresponding <a
4177 href="#int_stacksave"><tt>llvm.stacksave</tt></a> intrinsic executed. This is
4178 useful for implementing language features like scoped automatic variable sized
4185 See the description for <a href="#int_stacksave"><tt>llvm.stacksave</tt></a>.
4191 <!-- _______________________________________________________________________ -->
4192 <div class="doc_subsubsection">
4193 <a name="int_prefetch">'<tt>llvm.prefetch</tt>' Intrinsic</a>
4196 <div class="doc_text">
4200 declare void @llvm.prefetch(i8* <address>, i32 <rw>, i32 <locality>)
4207 The '<tt>llvm.prefetch</tt>' intrinsic is a hint to the code generator to insert
4208 a prefetch instruction if supported; otherwise, it is a noop. Prefetches have
4210 effect on the behavior of the program but can change its performance
4217 <tt>address</tt> is the address to be prefetched, <tt>rw</tt> is the specifier
4218 determining if the fetch should be for a read (0) or write (1), and
4219 <tt>locality</tt> is a temporal locality specifier ranging from (0) - no
4220 locality, to (3) - extremely local keep in cache. The <tt>rw</tt> and
4221 <tt>locality</tt> arguments must be constant integers.
4227 This intrinsic does not modify the behavior of the program. In particular,
4228 prefetches cannot trap and do not produce a value. On targets that support this
4229 intrinsic, the prefetch can provide hints to the processor cache for better
4235 <!-- _______________________________________________________________________ -->
4236 <div class="doc_subsubsection">
4237 <a name="int_pcmarker">'<tt>llvm.pcmarker</tt>' Intrinsic</a>
4240 <div class="doc_text">
4244 declare void @llvm.pcmarker(i32 <id>)
4251 The '<tt>llvm.pcmarker</tt>' intrinsic is a method to export a Program Counter
4253 code to simulators and other tools. The method is target specific, but it is
4254 expected that the marker will use exported symbols to transmit the PC of the marker.
4255 The marker makes no guarantees that it will remain with any specific instruction
4256 after optimizations. It is possible that the presence of a marker will inhibit
4257 optimizations. The intended use is to be inserted after optimizations to allow
4258 correlations of simulation runs.
4264 <tt>id</tt> is a numerical id identifying the marker.
4270 This intrinsic does not modify the behavior of the program. Backends that do not
4271 support this intrinisic may ignore it.
4276 <!-- _______________________________________________________________________ -->
4277 <div class="doc_subsubsection">
4278 <a name="int_readcyclecounter">'<tt>llvm.readcyclecounter</tt>' Intrinsic</a>
4281 <div class="doc_text">
4285 declare i64 @llvm.readcyclecounter( )
4292 The '<tt>llvm.readcyclecounter</tt>' intrinsic provides access to the cycle
4293 counter register (or similar low latency, high accuracy clocks) on those targets
4294 that support it. On X86, it should map to RDTSC. On Alpha, it should map to RPCC.
4295 As the backing counters overflow quickly (on the order of 9 seconds on alpha), this
4296 should only be used for small timings.
4302 When directly supported, reading the cycle counter should not modify any memory.
4303 Implementations are allowed to either return a application specific value or a
4304 system wide value. On backends without support, this is lowered to a constant 0.
4309 <!-- ======================================================================= -->
4310 <div class="doc_subsection">
4311 <a name="int_libc">Standard C Library Intrinsics</a>
4314 <div class="doc_text">
4316 LLVM provides intrinsics for a few important standard C library functions.
4317 These intrinsics allow source-language front-ends to pass information about the
4318 alignment of the pointer arguments to the code generator, providing opportunity
4319 for more efficient code generation.
4324 <!-- _______________________________________________________________________ -->
4325 <div class="doc_subsubsection">
4326 <a name="int_memcpy">'<tt>llvm.memcpy</tt>' Intrinsic</a>
4329 <div class="doc_text">
4333 declare void @llvm.memcpy.i32(i8 * <dest>, i8 * <src>,
4334 i32 <len>, i32 <align>)
4335 declare void @llvm.memcpy.i64(i8 * <dest>, i8 * <src>,
4336 i64 <len>, i32 <align>)
4342 The '<tt>llvm.memcpy.*</tt>' intrinsics copy a block of memory from the source
4343 location to the destination location.
4347 Note that, unlike the standard libc function, the <tt>llvm.memcpy.*</tt>
4348 intrinsics do not return a value, and takes an extra alignment argument.
4354 The first argument is a pointer to the destination, the second is a pointer to
4355 the source. The third argument is an integer argument
4356 specifying the number of bytes to copy, and the fourth argument is the alignment
4357 of the source and destination locations.
4361 If the call to this intrinisic has an alignment value that is not 0 or 1, then
4362 the caller guarantees that both the source and destination pointers are aligned
4369 The '<tt>llvm.memcpy.*</tt>' intrinsics copy a block of memory from the source
4370 location to the destination location, which are not allowed to overlap. It
4371 copies "len" bytes of memory over. If the argument is known to be aligned to
4372 some boundary, this can be specified as the fourth argument, otherwise it should
4378 <!-- _______________________________________________________________________ -->
4379 <div class="doc_subsubsection">
4380 <a name="int_memmove">'<tt>llvm.memmove</tt>' Intrinsic</a>
4383 <div class="doc_text">
4387 declare void @llvm.memmove.i32(i8 * <dest>, i8 * <src>,
4388 i32 <len>, i32 <align>)
4389 declare void @llvm.memmove.i64(i8 * <dest>, i8 * <src>,
4390 i64 <len>, i32 <align>)
4396 The '<tt>llvm.memmove.*</tt>' intrinsics move a block of memory from the source
4397 location to the destination location. It is similar to the
4398 '<tt>llvm.memcmp</tt>' intrinsic but allows the two memory locations to overlap.
4402 Note that, unlike the standard libc function, the <tt>llvm.memmove.*</tt>
4403 intrinsics do not return a value, and takes an extra alignment argument.
4409 The first argument is a pointer to the destination, the second is a pointer to
4410 the source. The third argument is an integer argument
4411 specifying the number of bytes to copy, and the fourth argument is the alignment
4412 of the source and destination locations.
4416 If the call to this intrinisic has an alignment value that is not 0 or 1, then
4417 the caller guarantees that the source and destination pointers are aligned to
4424 The '<tt>llvm.memmove.*</tt>' intrinsics copy a block of memory from the source
4425 location to the destination location, which may overlap. It
4426 copies "len" bytes of memory over. If the argument is known to be aligned to
4427 some boundary, this can be specified as the fourth argument, otherwise it should
4433 <!-- _______________________________________________________________________ -->
4434 <div class="doc_subsubsection">
4435 <a name="int_memset">'<tt>llvm.memset.*</tt>' Intrinsics</a>
4438 <div class="doc_text">
4442 declare void @llvm.memset.i32(i8 * <dest>, i8 <val>,
4443 i32 <len>, i32 <align>)
4444 declare void @llvm.memset.i64(i8 * <dest>, i8 <val>,
4445 i64 <len>, i32 <align>)
4451 The '<tt>llvm.memset.*</tt>' intrinsics fill a block of memory with a particular
4456 Note that, unlike the standard libc function, the <tt>llvm.memset</tt> intrinsic
4457 does not return a value, and takes an extra alignment argument.
4463 The first argument is a pointer to the destination to fill, the second is the
4464 byte value to fill it with, the third argument is an integer
4465 argument specifying the number of bytes to fill, and the fourth argument is the
4466 known alignment of destination location.
4470 If the call to this intrinisic has an alignment value that is not 0 or 1, then
4471 the caller guarantees that the destination pointer is aligned to that boundary.
4477 The '<tt>llvm.memset.*</tt>' intrinsics fill "len" bytes of memory starting at
4479 destination location. If the argument is known to be aligned to some boundary,
4480 this can be specified as the fourth argument, otherwise it should be set to 0 or
4486 <!-- _______________________________________________________________________ -->
4487 <div class="doc_subsubsection">
4488 <a name="int_sqrt">'<tt>llvm.sqrt.*</tt>' Intrinsic</a>
4491 <div class="doc_text">
4494 <p>This is an overloaded intrinsic. You can use <tt>llvm.sqrt</tt> on any
4495 floating point type. Not all targets support all types however.
4497 declare float @llvm.sqrt.f32(float %Val)
4498 declare double @llvm.sqrt.f64(double %Val)
4499 declare x86_fp80 @llvm.sqrt.f80(x86_fp80 %Val)
4500 declare fp128 @llvm.sqrt.f128(fp128 %Val)
4501 declare ppc_fp128 @llvm.sqrt.ppcf128(ppc_fp128 %Val)
4507 The '<tt>llvm.sqrt</tt>' intrinsics return the sqrt of the specified operand,
4508 returning the same value as the libm '<tt>sqrt</tt>' function would. Unlike
4509 <tt>sqrt</tt> in libm, however, <tt>llvm.sqrt</tt> has undefined behavior for
4510 negative numbers (which allows for better optimization).
4516 The argument and return value are floating point numbers of the same type.
4522 This function returns the sqrt of the specified operand if it is a nonnegative
4523 floating point number.
4527 <!-- _______________________________________________________________________ -->
4528 <div class="doc_subsubsection">
4529 <a name="int_powi">'<tt>llvm.powi.*</tt>' Intrinsic</a>
4532 <div class="doc_text">
4535 <p>This is an overloaded intrinsic. You can use <tt>llvm.powi</tt> on any
4536 floating point type. Not all targets support all types however.
4538 declare float @llvm.powi.f32(float %Val, i32 %power)
4539 declare double @llvm.powi.f64(double %Val, i32 %power)
4540 declare x86_fp80 @llvm.powi.f80(x86_fp80 %Val, i32 %power)
4541 declare fp128 @llvm.powi.f128(fp128 %Val, i32 %power)
4542 declare ppc_fp128 @llvm.powi.ppcf128(ppc_fp128 %Val, i32 %power)
4548 The '<tt>llvm.powi.*</tt>' intrinsics return the first operand raised to the
4549 specified (positive or negative) power. The order of evaluation of
4550 multiplications is not defined.
4556 The second argument is an integer power, and the first is a value to raise to
4563 This function returns the first value raised to the second power with an
4564 unspecified sequence of rounding operations.</p>
4568 <!-- ======================================================================= -->
4569 <div class="doc_subsection">
4570 <a name="int_manip">Bit Manipulation Intrinsics</a>
4573 <div class="doc_text">
4575 LLVM provides intrinsics for a few important bit manipulation operations.
4576 These allow efficient code generation for some algorithms.
4581 <!-- _______________________________________________________________________ -->
4582 <div class="doc_subsubsection">
4583 <a name="int_bswap">'<tt>llvm.bswap.*</tt>' Intrinsics</a>
4586 <div class="doc_text">
4589 <p>This is an overloaded intrinsic function. You can use bswap on any integer
4590 type that is an even number of bytes (i.e. BitWidth % 16 == 0).
4592 declare i16 @llvm.bswap.i16(i16 <id>)
4593 declare i32 @llvm.bswap.i32(i32 <id>)
4594 declare i64 @llvm.bswap.i64(i64 <id>)
4600 The '<tt>llvm.bswap</tt>' family of intrinsics is used to byte swap integer
4601 values with an even number of bytes (positive multiple of 16 bits). These are
4602 useful for performing operations on data that is not in the target's native
4609 The <tt>llvm.bswap.i16</tt> intrinsic returns an i16 value that has the high
4610 and low byte of the input i16 swapped. Similarly, the <tt>llvm.bswap.i32</tt>
4611 intrinsic returns an i32 value that has the four bytes of the input i32
4612 swapped, so that if the input bytes are numbered 0, 1, 2, 3 then the returned
4613 i32 will have its bytes in 3, 2, 1, 0 order. The <tt>llvm.bswap.i48</tt>,
4614 <tt>llvm.bswap.i64</tt> and other intrinsics extend this concept to
4615 additional even-byte lengths (6 bytes, 8 bytes and more, respectively).
4620 <!-- _______________________________________________________________________ -->
4621 <div class="doc_subsubsection">
4622 <a name="int_ctpop">'<tt>llvm.ctpop.*</tt>' Intrinsic</a>
4625 <div class="doc_text">
4628 <p>This is an overloaded intrinsic. You can use llvm.ctpop on any integer bit
4629 width. Not all targets support all bit widths however.
4631 declare i8 @llvm.ctpop.i8 (i8 <src>)
4632 declare i16 @llvm.ctpop.i16(i16 <src>)
4633 declare i32 @llvm.ctpop.i32(i32 <src>)
4634 declare i64 @llvm.ctpop.i64(i64 <src>)
4635 declare i256 @llvm.ctpop.i256(i256 <src>)
4641 The '<tt>llvm.ctpop</tt>' family of intrinsics counts the number of bits set in a
4648 The only argument is the value to be counted. The argument may be of any
4649 integer type. The return type must match the argument type.
4655 The '<tt>llvm.ctpop</tt>' intrinsic counts the 1's in a variable.
4659 <!-- _______________________________________________________________________ -->
4660 <div class="doc_subsubsection">
4661 <a name="int_ctlz">'<tt>llvm.ctlz.*</tt>' Intrinsic</a>
4664 <div class="doc_text">
4667 <p>This is an overloaded intrinsic. You can use <tt>llvm.ctlz</tt> on any
4668 integer bit width. Not all targets support all bit widths however.
4670 declare i8 @llvm.ctlz.i8 (i8 <src>)
4671 declare i16 @llvm.ctlz.i16(i16 <src>)
4672 declare i32 @llvm.ctlz.i32(i32 <src>)
4673 declare i64 @llvm.ctlz.i64(i64 <src>)
4674 declare i256 @llvm.ctlz.i256(i256 <src>)
4680 The '<tt>llvm.ctlz</tt>' family of intrinsic functions counts the number of
4681 leading zeros in a variable.
4687 The only argument is the value to be counted. The argument may be of any
4688 integer type. The return type must match the argument type.
4694 The '<tt>llvm.ctlz</tt>' intrinsic counts the leading (most significant) zeros
4695 in a variable. If the src == 0 then the result is the size in bits of the type
4696 of src. For example, <tt>llvm.ctlz(i32 2) = 30</tt>.
4702 <!-- _______________________________________________________________________ -->
4703 <div class="doc_subsubsection">
4704 <a name="int_cttz">'<tt>llvm.cttz.*</tt>' Intrinsic</a>
4707 <div class="doc_text">
4710 <p>This is an overloaded intrinsic. You can use <tt>llvm.cttz</tt> on any
4711 integer bit width. Not all targets support all bit widths however.
4713 declare i8 @llvm.cttz.i8 (i8 <src>)
4714 declare i16 @llvm.cttz.i16(i16 <src>)
4715 declare i32 @llvm.cttz.i32(i32 <src>)
4716 declare i64 @llvm.cttz.i64(i64 <src>)
4717 declare i256 @llvm.cttz.i256(i256 <src>)
4723 The '<tt>llvm.cttz</tt>' family of intrinsic functions counts the number of
4730 The only argument is the value to be counted. The argument may be of any
4731 integer type. The return type must match the argument type.
4737 The '<tt>llvm.cttz</tt>' intrinsic counts the trailing (least significant) zeros
4738 in a variable. If the src == 0 then the result is the size in bits of the type
4739 of src. For example, <tt>llvm.cttz(2) = 1</tt>.
4743 <!-- _______________________________________________________________________ -->
4744 <div class="doc_subsubsection">
4745 <a name="int_part_select">'<tt>llvm.part.select.*</tt>' Intrinsic</a>
4748 <div class="doc_text">
4751 <p>This is an overloaded intrinsic. You can use <tt>llvm.part.select</tt>
4752 on any integer bit width.
4754 declare i17 @llvm.part.select.i17 (i17 %val, i32 %loBit, i32 %hiBit)
4755 declare i29 @llvm.part.select.i29 (i29 %val, i32 %loBit, i32 %hiBit)
4759 <p>The '<tt>llvm.part.select</tt>' family of intrinsic functions selects a
4760 range of bits from an integer value and returns them in the same bit width as
4761 the original value.</p>
4764 <p>The first argument, <tt>%val</tt> and the result may be integer types of
4765 any bit width but they must have the same bit width. The second and third
4766 arguments must be <tt>i32</tt> type since they specify only a bit index.</p>
4769 <p>The operation of the '<tt>llvm.part.select</tt>' intrinsic has two modes
4770 of operation: forwards and reverse. If <tt>%loBit</tt> is greater than
4771 <tt>%hiBits</tt> then the intrinsic operates in reverse mode. Otherwise it
4772 operates in forward mode.</p>
4773 <p>In forward mode, this intrinsic is the equivalent of shifting <tt>%val</tt>
4774 right by <tt>%loBit</tt> bits and then ANDing it with a mask with
4775 only the <tt>%hiBit - %loBit</tt> bits set, as follows:</p>
4777 <li>The <tt>%val</tt> is shifted right (LSHR) by the number of bits specified
4778 by <tt>%loBits</tt>. This normalizes the value to the low order bits.</li>
4779 <li>The <tt>%loBits</tt> value is subtracted from the <tt>%hiBits</tt> value
4780 to determine the number of bits to retain.</li>
4781 <li>A mask of the retained bits is created by shifting a -1 value.</li>
4782 <li>The mask is ANDed with <tt>%val</tt> to produce the result.
4784 <p>In reverse mode, a similar computation is made except that the bits are
4785 returned in the reverse order. So, for example, if <tt>X</tt> has the value
4786 <tt>i16 0x0ACF (101011001111)</tt> and we apply
4787 <tt>part.select(i16 X, 8, 3)</tt> to it, we get back the value
4788 <tt>i16 0x0026 (000000100110)</tt>.</p>
4791 <div class="doc_subsubsection">
4792 <a name="int_part_set">'<tt>llvm.part.set.*</tt>' Intrinsic</a>
4795 <div class="doc_text">
4798 <p>This is an overloaded intrinsic. You can use <tt>llvm.part.set</tt>
4799 on any integer bit width.
4801 declare i17 @llvm.part.set.i17.i9 (i17 %val, i9 %repl, i32 %lo, i32 %hi)
4802 declare i29 @llvm.part.set.i29.i9 (i29 %val, i9 %repl, i32 %lo, i32 %hi)
4806 <p>The '<tt>llvm.part.set</tt>' family of intrinsic functions replaces a range
4807 of bits in an integer value with another integer value. It returns the integer
4808 with the replaced bits.</p>
4811 <p>The first argument, <tt>%val</tt> and the result may be integer types of
4812 any bit width but they must have the same bit width. <tt>%val</tt> is the value
4813 whose bits will be replaced. The second argument, <tt>%repl</tt> may be an
4814 integer of any bit width. The third and fourth arguments must be <tt>i32</tt>
4815 type since they specify only a bit index.</p>
4818 <p>The operation of the '<tt>llvm.part.set</tt>' intrinsic has two modes
4819 of operation: forwards and reverse. If <tt>%lo</tt> is greater than
4820 <tt>%hi</tt> then the intrinsic operates in reverse mode. Otherwise it
4821 operates in forward mode.</p>
4822 <p>For both modes, the <tt>%repl</tt> value is prepared for use by either
4823 truncating it down to the size of the replacement area or zero extending it
4824 up to that size.</p>
4825 <p>In forward mode, the bits between <tt>%lo</tt> and <tt>%hi</tt> (inclusive)
4826 are replaced with corresponding bits from <tt>%repl</tt>. That is the 0th bit
4827 in <tt>%repl</tt> replaces the <tt>%lo</tt>th bit in <tt>%val</tt> and etc. up
4828 to the <tt>%hi</tt>th bit.
4829 <p>In reverse mode, a similar computation is made except that the bits are
4830 reversed. That is, the <tt>0</tt>th bit in <tt>%repl</tt> replaces the
4831 <tt>%hi</tt> bit in <tt>%val</tt> and etc. down to the <tt>%lo</tt>th bit.
4834 llvm.part.set(0xFFFF, 0, 4, 7) -> 0xFF0F
4835 llvm.part.set(0xFFFF, 0, 7, 4) -> 0xFF0F
4836 llvm.part.set(0xFFFF, 1, 7, 4) -> 0xFF8F
4837 llvm.part.set(0xFFFF, F, 8, 3) -> 0xFFE7
4838 llvm.part.set(0xFFFF, 0, 3, 8) -> 0xFE07
4842 <!-- ======================================================================= -->
4843 <div class="doc_subsection">
4844 <a name="int_debugger">Debugger Intrinsics</a>
4847 <div class="doc_text">
4849 The LLVM debugger intrinsics (which all start with <tt>llvm.dbg.</tt> prefix),
4850 are described in the <a
4851 href="SourceLevelDebugging.html#format_common_intrinsics">LLVM Source Level
4852 Debugging</a> document.
4857 <!-- ======================================================================= -->
4858 <div class="doc_subsection">
4859 <a name="int_eh">Exception Handling Intrinsics</a>
4862 <div class="doc_text">
4863 <p> The LLVM exception handling intrinsics (which all start with
4864 <tt>llvm.eh.</tt> prefix), are described in the <a
4865 href="ExceptionHandling.html#format_common_intrinsics">LLVM Exception
4866 Handling</a> document. </p>
4869 <!-- ======================================================================= -->
4870 <div class="doc_subsection">
4871 <a name="int_trampoline">Trampoline Intrinsic</a>
4874 <div class="doc_text">
4876 This intrinsic makes it possible to excise one parameter, marked with
4877 the <tt>nest</tt> attribute, from a function. The result is a callable
4878 function pointer lacking the nest parameter - the caller does not need
4879 to provide a value for it. Instead, the value to use is stored in
4880 advance in a "trampoline", a block of memory usually allocated
4881 on the stack, which also contains code to splice the nest value into the
4882 argument list. This is used to implement the GCC nested function address
4886 For example, if the function is
4887 <tt>i32 f(i8* nest %c, i32 %x, i32 %y)</tt> then the resulting function
4888 pointer has signature <tt>i32 (i32, i32)*</tt>. It can be created as follows:</p>
4890 %tramp = alloca [10 x i8], align 4 ; size and alignment only correct for X86
4891 %tramp1 = getelementptr [10 x i8]* %tramp, i32 0, i32 0
4892 %p = call i8* @llvm.init.trampoline( i8* %tramp1, i8* bitcast (i32 (i8* nest , i32, i32)* @f to i8*), i8* %nval )
4893 %fp = bitcast i8* %p to i32 (i32, i32)*
4895 <p>The call <tt>%val = call i32 %fp( i32 %x, i32 %y )</tt> is then equivalent
4896 to <tt>%val = call i32 %f( i8* %nval, i32 %x, i32 %y )</tt>.</p>
4899 <!-- _______________________________________________________________________ -->
4900 <div class="doc_subsubsection">
4901 <a name="int_it">'<tt>llvm.init.trampoline</tt>' Intrinsic</a>
4903 <div class="doc_text">
4906 declare i8* @llvm.init.trampoline(i8* <tramp>, i8* <func>, i8* <nval>)
4910 This fills the memory pointed to by <tt>tramp</tt> with code
4911 and returns a function pointer suitable for executing it.
4915 The <tt>llvm.init.trampoline</tt> intrinsic takes three arguments, all
4916 pointers. The <tt>tramp</tt> argument must point to a sufficiently large
4917 and sufficiently aligned block of memory; this memory is written to by the
4918 intrinsic. Note that the size and the alignment are target-specific - LLVM
4919 currently provides no portable way of determining them, so a front-end that
4920 generates this intrinsic needs to have some target-specific knowledge.
4921 The <tt>func</tt> argument must hold a function bitcast to an <tt>i8*</tt>.
4925 The block of memory pointed to by <tt>tramp</tt> is filled with target
4926 dependent code, turning it into a function. A pointer to this function is
4927 returned, but needs to be bitcast to an
4928 <a href="#int_trampoline">appropriate function pointer type</a>
4929 before being called. The new function's signature is the same as that of
4930 <tt>func</tt> with any arguments marked with the <tt>nest</tt> attribute
4931 removed. At most one such <tt>nest</tt> argument is allowed, and it must be
4932 of pointer type. Calling the new function is equivalent to calling
4933 <tt>func</tt> with the same argument list, but with <tt>nval</tt> used for the
4934 missing <tt>nest</tt> argument. If, after calling
4935 <tt>llvm.init.trampoline</tt>, the memory pointed to by <tt>tramp</tt> is
4936 modified, then the effect of any later call to the returned function pointer is
4941 <!-- ======================================================================= -->
4942 <div class="doc_subsection">
4943 <a name="int_general">General Intrinsics</a>
4946 <div class="doc_text">
4947 <p> This class of intrinsics is designed to be generic and has
4948 no specific purpose. </p>
4951 <!-- _______________________________________________________________________ -->
4952 <div class="doc_subsubsection">
4953 <a name="int_var_annotation">'<tt>llvm.var.annotation</tt>' Intrinsic</a>
4956 <div class="doc_text">
4960 declare void @llvm.var.annotation(i8* <val>, i8* <str>, i8* <str>, i32 <int> )
4966 The '<tt>llvm.var.annotation</tt>' intrinsic
4972 The first argument is a pointer to a value, the second is a pointer to a
4973 global string, the third is a pointer to a global string which is the source
4974 file name, and the last argument is the line number.
4980 This intrinsic allows annotation of local variables with arbitrary strings.
4981 This can be useful for special purpose optimizations that want to look for these
4982 annotations. These have no other defined use, they are ignored by code
4983 generation and optimization.
4986 <!-- _______________________________________________________________________ -->
4987 <div class="doc_subsubsection">
4988 <a name="int_annotation">'<tt>llvm.annotation.*</tt>' Intrinsic</a>
4991 <div class="doc_text">
4994 <p>This is an overloaded intrinsic. You can use '<tt>llvm.annotation</tt>' on
4995 any integer bit width.
4998 declare i8 @llvm.annotation.i8(i8 <val>, i8* <str>, i8* <str>, i32 <int> )
4999 declare i16 @llvm.annotation.i16(i16 <val>, i8* <str>, i8* <str>, i32 <int> )
5000 declare i32 @llvm.annotation.i32(i32 <val>, i8* <str>, i8* <str>, i32 <int> )
5001 declare i64 @llvm.annotation.i64(i64 <val>, i8* <str>, i8* <str>, i32 <int> )
5002 declare i256 @llvm.annotation.i256(i256 <val>, i8* <str>, i8* <str>, i32 <int> )
5008 The '<tt>llvm.annotation</tt>' intrinsic.
5014 The first argument is an integer value (result of some expression),
5015 the second is a pointer to a global string, the third is a pointer to a global
5016 string which is the source file name, and the last argument is the line number.
5017 It returns the value of the first argument.
5023 This intrinsic allows annotations to be put on arbitrary expressions
5024 with arbitrary strings. This can be useful for special purpose optimizations
5025 that want to look for these annotations. These have no other defined use, they
5026 are ignored by code generation and optimization.
5029 <!-- *********************************************************************** -->
5032 <a href="http://jigsaw.w3.org/css-validator/check/referer"><img
5033 src="http://jigsaw.w3.org/css-validator/images/vcss" alt="Valid CSS!"></a>
5034 <a href="http://validator.w3.org/check/referer"><img
5035 src="http://www.w3.org/Icons/valid-html401" alt="Valid HTML 4.01!" /></a>
5037 <a href="mailto:sabre@nondot.org">Chris Lattner</a><br>
5038 <a href="http://llvm.org">The LLVM Compiler Infrastructure</a><br>
5039 Last modified: $Date$