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="#gc">Garbage Collector Names</a></li>
30 <li><a href="#moduleasm">Module-Level Inline Assembly</a></li>
31 <li><a href="#datalayout">Data Layout</a></li>
34 <li><a href="#typesystem">Type System</a>
36 <li><a href="#t_primitive">Primitive Types</a>
38 <li><a href="#t_classifications">Type Classifications</a></li>
41 <li><a href="#t_derived">Derived Types</a>
43 <li><a href="#t_array">Array Type</a></li>
44 <li><a href="#t_function">Function Type</a></li>
45 <li><a href="#t_pointer">Pointer Type</a></li>
46 <li><a href="#t_struct">Structure Type</a></li>
47 <li><a href="#t_pstruct">Packed Structure Type</a></li>
48 <li><a href="#t_vector">Vector Type</a></li>
49 <li><a href="#t_opaque">Opaque Type</a></li>
54 <li><a href="#constants">Constants</a>
56 <li><a href="#simpleconstants">Simple Constants</a>
57 <li><a href="#aggregateconstants">Aggregate Constants</a>
58 <li><a href="#globalconstants">Global Variable and Function Addresses</a>
59 <li><a href="#undefvalues">Undefined Values</a>
60 <li><a href="#constantexprs">Constant Expressions</a>
63 <li><a href="#othervalues">Other Values</a>
65 <li><a href="#inlineasm">Inline Assembler Expressions</a>
68 <li><a href="#instref">Instruction Reference</a>
70 <li><a href="#terminators">Terminator Instructions</a>
72 <li><a href="#i_ret">'<tt>ret</tt>' Instruction</a></li>
73 <li><a href="#i_br">'<tt>br</tt>' Instruction</a></li>
74 <li><a href="#i_switch">'<tt>switch</tt>' Instruction</a></li>
75 <li><a href="#i_invoke">'<tt>invoke</tt>' Instruction</a></li>
76 <li><a href="#i_unwind">'<tt>unwind</tt>' Instruction</a></li>
77 <li><a href="#i_unreachable">'<tt>unreachable</tt>' Instruction</a></li>
80 <li><a href="#binaryops">Binary Operations</a>
82 <li><a href="#i_add">'<tt>add</tt>' Instruction</a></li>
83 <li><a href="#i_sub">'<tt>sub</tt>' Instruction</a></li>
84 <li><a href="#i_mul">'<tt>mul</tt>' Instruction</a></li>
85 <li><a href="#i_udiv">'<tt>udiv</tt>' Instruction</a></li>
86 <li><a href="#i_sdiv">'<tt>sdiv</tt>' Instruction</a></li>
87 <li><a href="#i_fdiv">'<tt>fdiv</tt>' Instruction</a></li>
88 <li><a href="#i_urem">'<tt>urem</tt>' Instruction</a></li>
89 <li><a href="#i_srem">'<tt>srem</tt>' Instruction</a></li>
90 <li><a href="#i_frem">'<tt>frem</tt>' Instruction</a></li>
93 <li><a href="#bitwiseops">Bitwise Binary Operations</a>
95 <li><a href="#i_shl">'<tt>shl</tt>' Instruction</a></li>
96 <li><a href="#i_lshr">'<tt>lshr</tt>' Instruction</a></li>
97 <li><a href="#i_ashr">'<tt>ashr</tt>' Instruction</a></li>
98 <li><a href="#i_and">'<tt>and</tt>' Instruction</a></li>
99 <li><a href="#i_or">'<tt>or</tt>' Instruction</a></li>
100 <li><a href="#i_xor">'<tt>xor</tt>' Instruction</a></li>
103 <li><a href="#vectorops">Vector Operations</a>
105 <li><a href="#i_extractelement">'<tt>extractelement</tt>' Instruction</a></li>
106 <li><a href="#i_insertelement">'<tt>insertelement</tt>' Instruction</a></li>
107 <li><a href="#i_shufflevector">'<tt>shufflevector</tt>' Instruction</a></li>
110 <li><a href="#memoryops">Memory Access and Addressing Operations</a>
112 <li><a href="#i_malloc">'<tt>malloc</tt>' Instruction</a></li>
113 <li><a href="#i_free">'<tt>free</tt>' Instruction</a></li>
114 <li><a href="#i_alloca">'<tt>alloca</tt>' Instruction</a></li>
115 <li><a href="#i_load">'<tt>load</tt>' Instruction</a></li>
116 <li><a href="#i_store">'<tt>store</tt>' Instruction</a></li>
117 <li><a href="#i_getelementptr">'<tt>getelementptr</tt>' Instruction</a></li>
120 <li><a href="#convertops">Conversion Operations</a>
122 <li><a href="#i_trunc">'<tt>trunc .. to</tt>' Instruction</a></li>
123 <li><a href="#i_zext">'<tt>zext .. to</tt>' Instruction</a></li>
124 <li><a href="#i_sext">'<tt>sext .. to</tt>' Instruction</a></li>
125 <li><a href="#i_fptrunc">'<tt>fptrunc .. to</tt>' Instruction</a></li>
126 <li><a href="#i_fpext">'<tt>fpext .. to</tt>' Instruction</a></li>
127 <li><a href="#i_fptoui">'<tt>fptoui .. to</tt>' Instruction</a></li>
128 <li><a href="#i_fptosi">'<tt>fptosi .. to</tt>' Instruction</a></li>
129 <li><a href="#i_uitofp">'<tt>uitofp .. to</tt>' Instruction</a></li>
130 <li><a href="#i_sitofp">'<tt>sitofp .. to</tt>' Instruction</a></li>
131 <li><a href="#i_ptrtoint">'<tt>ptrtoint .. to</tt>' Instruction</a></li>
132 <li><a href="#i_inttoptr">'<tt>inttoptr .. to</tt>' Instruction</a></li>
133 <li><a href="#i_bitcast">'<tt>bitcast .. to</tt>' Instruction</a></li>
135 <li><a href="#otherops">Other Operations</a>
137 <li><a href="#i_icmp">'<tt>icmp</tt>' Instruction</a></li>
138 <li><a href="#i_fcmp">'<tt>fcmp</tt>' Instruction</a></li>
139 <li><a href="#i_phi">'<tt>phi</tt>' Instruction</a></li>
140 <li><a href="#i_select">'<tt>select</tt>' Instruction</a></li>
141 <li><a href="#i_call">'<tt>call</tt>' Instruction</a></li>
142 <li><a href="#i_va_arg">'<tt>va_arg</tt>' Instruction</a></li>
147 <li><a href="#intrinsics">Intrinsic Functions</a>
149 <li><a href="#int_varargs">Variable Argument Handling Intrinsics</a>
151 <li><a href="#int_va_start">'<tt>llvm.va_start</tt>' Intrinsic</a></li>
152 <li><a href="#int_va_end">'<tt>llvm.va_end</tt>' Intrinsic</a></li>
153 <li><a href="#int_va_copy">'<tt>llvm.va_copy</tt>' Intrinsic</a></li>
156 <li><a href="#int_gc">Accurate Garbage Collection Intrinsics</a>
158 <li><a href="#int_gcroot">'<tt>llvm.gcroot</tt>' Intrinsic</a></li>
159 <li><a href="#int_gcread">'<tt>llvm.gcread</tt>' Intrinsic</a></li>
160 <li><a href="#int_gcwrite">'<tt>llvm.gcwrite</tt>' Intrinsic</a></li>
163 <li><a href="#int_codegen">Code Generator Intrinsics</a>
165 <li><a href="#int_returnaddress">'<tt>llvm.returnaddress</tt>' Intrinsic</a></li>
166 <li><a href="#int_frameaddress">'<tt>llvm.frameaddress</tt>' Intrinsic</a></li>
167 <li><a href="#int_stacksave">'<tt>llvm.stacksave</tt>' Intrinsic</a></li>
168 <li><a href="#int_stackrestore">'<tt>llvm.stackrestore</tt>' Intrinsic</a></li>
169 <li><a href="#int_prefetch">'<tt>llvm.prefetch</tt>' Intrinsic</a></li>
170 <li><a href="#int_pcmarker">'<tt>llvm.pcmarker</tt>' Intrinsic</a></li>
171 <li><a href="#int_readcyclecounter"><tt>llvm.readcyclecounter</tt>' Intrinsic</a></li>
174 <li><a href="#int_libc">Standard C Library Intrinsics</a>
176 <li><a href="#int_memcpy">'<tt>llvm.memcpy.*</tt>' Intrinsic</a></li>
177 <li><a href="#int_memmove">'<tt>llvm.memmove.*</tt>' Intrinsic</a></li>
178 <li><a href="#int_memset">'<tt>llvm.memset.*</tt>' Intrinsic</a></li>
179 <li><a href="#int_sqrt">'<tt>llvm.sqrt.*</tt>' Intrinsic</a></li>
180 <li><a href="#int_powi">'<tt>llvm.powi.*</tt>' Intrinsic</a></li>
181 <li><a href="#int_sin">'<tt>llvm.sin.*</tt>' Intrinsic</a></li>
182 <li><a href="#int_cos">'<tt>llvm.cos.*</tt>' Intrinsic</a></li>
183 <li><a href="#int_pow">'<tt>llvm.pow.*</tt>' Intrinsic</a></li>
186 <li><a href="#int_manip">Bit Manipulation Intrinsics</a>
188 <li><a href="#int_bswap">'<tt>llvm.bswap.*</tt>' Intrinsics</a></li>
189 <li><a href="#int_ctpop">'<tt>llvm.ctpop.*</tt>' Intrinsic </a></li>
190 <li><a href="#int_ctlz">'<tt>llvm.ctlz.*</tt>' Intrinsic </a></li>
191 <li><a href="#int_cttz">'<tt>llvm.cttz.*</tt>' Intrinsic </a></li>
192 <li><a href="#int_part_select">'<tt>llvm.part.select.*</tt>' Intrinsic </a></li>
193 <li><a href="#int_part_set">'<tt>llvm.part.set.*</tt>' Intrinsic </a></li>
196 <li><a href="#int_debugger">Debugger intrinsics</a></li>
197 <li><a href="#int_eh">Exception Handling intrinsics</a></li>
198 <li><a href="#int_trampoline">Trampoline Intrinsic</a>
200 <li><a href="#int_it">'<tt>llvm.init.trampoline</tt>' Intrinsic</a></li>
203 <li><a href="#int_general">General intrinsics</a>
205 <li><a href="#int_var_annotation">
206 <tt>llvm.var.annotation</tt>' Intrinsic</a></li>
209 <li><a href="#int_annotation">
210 <tt>llvm.annotation.*</tt>' Intrinsic</a></li>
217 <div class="doc_author">
218 <p>Written by <a href="mailto:sabre@nondot.org">Chris Lattner</a>
219 and <a href="mailto:vadve@cs.uiuc.edu">Vikram Adve</a></p>
222 <!-- *********************************************************************** -->
223 <div class="doc_section"> <a name="abstract">Abstract </a></div>
224 <!-- *********************************************************************** -->
226 <div class="doc_text">
227 <p>This document is a reference manual for the LLVM assembly language.
228 LLVM is an SSA based representation that provides type safety,
229 low-level operations, flexibility, and the capability of representing
230 'all' high-level languages cleanly. It is the common code
231 representation used throughout all phases of the LLVM compilation
235 <!-- *********************************************************************** -->
236 <div class="doc_section"> <a name="introduction">Introduction</a> </div>
237 <!-- *********************************************************************** -->
239 <div class="doc_text">
241 <p>The LLVM code representation is designed to be used in three
242 different forms: as an in-memory compiler IR, as an on-disk bitcode
243 representation (suitable for fast loading by a Just-In-Time compiler),
244 and as a human readable assembly language representation. This allows
245 LLVM to provide a powerful intermediate representation for efficient
246 compiler transformations and analysis, while providing a natural means
247 to debug and visualize the transformations. The three different forms
248 of LLVM are all equivalent. This document describes the human readable
249 representation and notation.</p>
251 <p>The LLVM representation aims to be light-weight and low-level
252 while being expressive, typed, and extensible at the same time. It
253 aims to be a "universal IR" of sorts, by being at a low enough level
254 that high-level ideas may be cleanly mapped to it (similar to how
255 microprocessors are "universal IR's", allowing many source languages to
256 be mapped to them). By providing type information, LLVM can be used as
257 the target of optimizations: for example, through pointer analysis, it
258 can be proven that a C automatic variable is never accessed outside of
259 the current function... allowing it to be promoted to a simple SSA
260 value instead of a memory location.</p>
264 <!-- _______________________________________________________________________ -->
265 <div class="doc_subsubsection"> <a name="wellformed">Well-Formedness</a> </div>
267 <div class="doc_text">
269 <p>It is important to note that this document describes 'well formed'
270 LLVM assembly language. There is a difference between what the parser
271 accepts and what is considered 'well formed'. For example, the
272 following instruction is syntactically okay, but not well formed:</p>
274 <div class="doc_code">
276 %x = <a href="#i_add">add</a> i32 1, %x
280 <p>...because the definition of <tt>%x</tt> does not dominate all of
281 its uses. The LLVM infrastructure provides a verification pass that may
282 be used to verify that an LLVM module is well formed. This pass is
283 automatically run by the parser after parsing input assembly and by
284 the optimizer before it outputs bitcode. The violations pointed out
285 by the verifier pass indicate bugs in transformation passes or input to
289 <!-- Describe the typesetting conventions here. -->
291 <!-- *********************************************************************** -->
292 <div class="doc_section"> <a name="identifiers">Identifiers</a> </div>
293 <!-- *********************************************************************** -->
295 <div class="doc_text">
297 <p>LLVM identifiers come in two basic types: global and local. Global
298 identifiers (functions, global variables) begin with the @ character. Local
299 identifiers (register names, types) begin with the % character. Additionally,
300 there are three different formats for identifiers, for different purposes:
303 <li>Named values are represented as a string of characters with their prefix.
304 For example, %foo, @DivisionByZero, %a.really.long.identifier. The actual
305 regular expression used is '<tt>[%@][a-zA-Z$._][a-zA-Z$._0-9]*</tt>'.
306 Identifiers which require other characters in their names can be surrounded
307 with quotes. In this way, anything except a <tt>"</tt> character can
308 be used in a named value.</li>
310 <li>Unnamed values are represented as an unsigned numeric value with their
311 prefix. For example, %12, @2, %44.</li>
313 <li>Constants, which are described in a <a href="#constants">section about
314 constants</a>, below.</li>
317 <p>LLVM requires that values start with a prefix for two reasons: Compilers
318 don't need to worry about name clashes with reserved words, and the set of
319 reserved words may be expanded in the future without penalty. Additionally,
320 unnamed identifiers allow a compiler to quickly come up with a temporary
321 variable without having to avoid symbol table conflicts.</p>
323 <p>Reserved words in LLVM are very similar to reserved words in other
324 languages. There are keywords for different opcodes
325 ('<tt><a href="#i_add">add</a></tt>',
326 '<tt><a href="#i_bitcast">bitcast</a></tt>',
327 '<tt><a href="#i_ret">ret</a></tt>', etc...), for primitive type names ('<tt><a
328 href="#t_void">void</a></tt>', '<tt><a href="#t_primitive">i32</a></tt>', etc...),
329 and others. These reserved words cannot conflict with variable names, because
330 none of them start with a prefix character ('%' or '@').</p>
332 <p>Here is an example of LLVM code to multiply the integer variable
333 '<tt>%X</tt>' by 8:</p>
337 <div class="doc_code">
339 %result = <a href="#i_mul">mul</a> i32 %X, 8
343 <p>After strength reduction:</p>
345 <div class="doc_code">
347 %result = <a href="#i_shl">shl</a> i32 %X, i8 3
351 <p>And the hard way:</p>
353 <div class="doc_code">
355 <a href="#i_add">add</a> i32 %X, %X <i>; yields {i32}:%0</i>
356 <a href="#i_add">add</a> i32 %0, %0 <i>; yields {i32}:%1</i>
357 %result = <a href="#i_add">add</a> i32 %1, %1
361 <p>This last way of multiplying <tt>%X</tt> by 8 illustrates several
362 important lexical features of LLVM:</p>
366 <li>Comments are delimited with a '<tt>;</tt>' and go until the end of
369 <li>Unnamed temporaries are created when the result of a computation is not
370 assigned to a named value.</li>
372 <li>Unnamed temporaries are numbered sequentially</li>
376 <p>...and it also shows a convention that we follow in this document. When
377 demonstrating instructions, we will follow an instruction with a comment that
378 defines the type and name of value produced. Comments are shown in italic
383 <!-- *********************************************************************** -->
384 <div class="doc_section"> <a name="highlevel">High Level Structure</a> </div>
385 <!-- *********************************************************************** -->
387 <!-- ======================================================================= -->
388 <div class="doc_subsection"> <a name="modulestructure">Module Structure</a>
391 <div class="doc_text">
393 <p>LLVM programs are composed of "Module"s, each of which is a
394 translation unit of the input programs. Each module consists of
395 functions, global variables, and symbol table entries. Modules may be
396 combined together with the LLVM linker, which merges function (and
397 global variable) definitions, resolves forward declarations, and merges
398 symbol table entries. Here is an example of the "hello world" module:</p>
400 <div class="doc_code">
401 <pre><i>; Declare the string constant as a global constant...</i>
402 <a href="#identifiers">@.LC0</a> = <a href="#linkage_internal">internal</a> <a
403 href="#globalvars">constant</a> <a href="#t_array">[13 x i8]</a> c"hello world\0A\00" <i>; [13 x i8]*</i>
405 <i>; External declaration of the puts function</i>
406 <a href="#functionstructure">declare</a> i32 @puts(i8 *) <i>; i32(i8 *)* </i>
408 <i>; Definition of main function</i>
409 define i32 @main() { <i>; i32()* </i>
410 <i>; Convert [13x i8 ]* to i8 *...</i>
412 href="#i_getelementptr">getelementptr</a> [13 x i8 ]* @.LC0, i64 0, i64 0 <i>; i8 *</i>
414 <i>; Call puts function to write out the string to stdout...</i>
416 href="#i_call">call</a> i32 @puts(i8 * %cast210) <i>; i32</i>
418 href="#i_ret">ret</a> i32 0<br>}<br>
422 <p>This example is made up of a <a href="#globalvars">global variable</a>
423 named "<tt>.LC0</tt>", an external declaration of the "<tt>puts</tt>"
424 function, and a <a href="#functionstructure">function definition</a>
425 for "<tt>main</tt>".</p>
427 <p>In general, a module is made up of a list of global values,
428 where both functions and global variables are global values. Global values are
429 represented by a pointer to a memory location (in this case, a pointer to an
430 array of char, and a pointer to a function), and have one of the following <a
431 href="#linkage">linkage types</a>.</p>
435 <!-- ======================================================================= -->
436 <div class="doc_subsection">
437 <a name="linkage">Linkage Types</a>
440 <div class="doc_text">
443 All Global Variables and Functions have one of the following types of linkage:
448 <dt><tt><b><a name="linkage_internal">internal</a></b></tt> </dt>
450 <dd>Global values with internal linkage are only directly accessible by
451 objects in the current module. In particular, linking code into a module with
452 an internal global value may cause the internal to be renamed as necessary to
453 avoid collisions. Because the symbol is internal to the module, all
454 references can be updated. This corresponds to the notion of the
455 '<tt>static</tt>' keyword in C.
458 <dt><tt><b><a name="linkage_linkonce">linkonce</a></b></tt>: </dt>
460 <dd>Globals with "<tt>linkonce</tt>" linkage are merged with other globals of
461 the same name when linkage occurs. This is typically used to implement
462 inline functions, templates, or other code which must be generated in each
463 translation unit that uses it. Unreferenced <tt>linkonce</tt> globals are
464 allowed to be discarded.
467 <dt><tt><b><a name="linkage_weak">weak</a></b></tt>: </dt>
469 <dd>"<tt>weak</tt>" linkage is exactly the same as <tt>linkonce</tt> linkage,
470 except that unreferenced <tt>weak</tt> globals may not be discarded. This is
471 used for globals that may be emitted in multiple translation units, but that
472 are not guaranteed to be emitted into every translation unit that uses them.
473 One example of this are common globals in C, such as "<tt>int X;</tt>" at
477 <dt><tt><b><a name="linkage_appending">appending</a></b></tt>: </dt>
479 <dd>"<tt>appending</tt>" linkage may only be applied to global variables of
480 pointer to array type. When two global variables with appending linkage are
481 linked together, the two global arrays are appended together. This is the
482 LLVM, typesafe, equivalent of having the system linker append together
483 "sections" with identical names when .o files are linked.
486 <dt><tt><b><a name="linkage_externweak">extern_weak</a></b></tt>: </dt>
487 <dd>The semantics of this linkage follow the ELF model: the symbol is weak
488 until linked, if not linked, the symbol becomes null instead of being an
492 <dt><tt><b><a name="linkage_external">externally visible</a></b></tt>:</dt>
494 <dd>If none of the above identifiers are used, the global is externally
495 visible, meaning that it participates in linkage and can be used to resolve
496 external symbol references.
501 The next two types of linkage are targeted for Microsoft Windows platform
502 only. They are designed to support importing (exporting) symbols from (to)
507 <dt><tt><b><a name="linkage_dllimport">dllimport</a></b></tt>: </dt>
509 <dd>"<tt>dllimport</tt>" linkage causes the compiler to reference a function
510 or variable via a global pointer to a pointer that is set up by the DLL
511 exporting the symbol. On Microsoft Windows targets, the pointer name is
512 formed by combining <code>_imp__</code> and the function or variable name.
515 <dt><tt><b><a name="linkage_dllexport">dllexport</a></b></tt>: </dt>
517 <dd>"<tt>dllexport</tt>" linkage causes the compiler to provide a global
518 pointer to a pointer in a DLL, so that it can be referenced with the
519 <tt>dllimport</tt> attribute. On Microsoft Windows targets, the pointer
520 name is formed by combining <code>_imp__</code> and the function or variable
526 <p><a name="linkage_external"></a>For example, since the "<tt>.LC0</tt>"
527 variable is defined to be internal, if another module defined a "<tt>.LC0</tt>"
528 variable and was linked with this one, one of the two would be renamed,
529 preventing a collision. Since "<tt>main</tt>" and "<tt>puts</tt>" are
530 external (i.e., lacking any linkage declarations), they are accessible
531 outside of the current module.</p>
532 <p>It is illegal for a function <i>declaration</i>
533 to have any linkage type other than "externally visible", <tt>dllimport</tt>,
534 or <tt>extern_weak</tt>.</p>
535 <p>Aliases can have only <tt>external</tt>, <tt>internal</tt> and <tt>weak</tt>
539 <!-- ======================================================================= -->
540 <div class="doc_subsection">
541 <a name="callingconv">Calling Conventions</a>
544 <div class="doc_text">
546 <p>LLVM <a href="#functionstructure">functions</a>, <a href="#i_call">calls</a>
547 and <a href="#i_invoke">invokes</a> can all have an optional calling convention
548 specified for the call. The calling convention of any pair of dynamic
549 caller/callee must match, or the behavior of the program is undefined. The
550 following calling conventions are supported by LLVM, and more may be added in
554 <dt><b>"<tt>ccc</tt>" - The C calling convention</b>:</dt>
556 <dd>This calling convention (the default if no other calling convention is
557 specified) matches the target C calling conventions. This calling convention
558 supports varargs function calls and tolerates some mismatch in the declared
559 prototype and implemented declaration of the function (as does normal C).
562 <dt><b>"<tt>fastcc</tt>" - The fast calling convention</b>:</dt>
564 <dd>This calling convention attempts to make calls as fast as possible
565 (e.g. by passing things in registers). This calling convention allows the
566 target to use whatever tricks it wants to produce fast code for the target,
567 without having to conform to an externally specified ABI. Implementations of
568 this convention should allow arbitrary tail call optimization to be supported.
569 This calling convention does not support varargs and requires the prototype of
570 all callees to exactly match the prototype of the function definition.
573 <dt><b>"<tt>coldcc</tt>" - The cold calling convention</b>:</dt>
575 <dd>This calling convention attempts to make code in the caller as efficient
576 as possible under the assumption that the call is not commonly executed. As
577 such, these calls often preserve all registers so that the call does not break
578 any live ranges in the caller side. This calling convention does not support
579 varargs and requires the prototype of all callees to exactly match the
580 prototype of the function definition.
583 <dt><b>"<tt>cc <<em>n</em>></tt>" - Numbered convention</b>:</dt>
585 <dd>Any calling convention may be specified by number, allowing
586 target-specific calling conventions to be used. Target specific calling
587 conventions start at 64.
591 <p>More calling conventions can be added/defined on an as-needed basis, to
592 support pascal conventions or any other well-known target-independent
597 <!-- ======================================================================= -->
598 <div class="doc_subsection">
599 <a name="visibility">Visibility Styles</a>
602 <div class="doc_text">
605 All Global Variables and Functions have one of the following visibility styles:
609 <dt><b>"<tt>default</tt>" - Default style</b>:</dt>
611 <dd>On ELF, default visibility means that the declaration is visible to other
612 modules and, in shared libraries, means that the declared entity may be
613 overridden. On Darwin, default visibility means that the declaration is
614 visible to other modules. Default visibility corresponds to "external
615 linkage" in the language.
618 <dt><b>"<tt>hidden</tt>" - Hidden style</b>:</dt>
620 <dd>Two declarations of an object with hidden visibility refer to the same
621 object if they are in the same shared object. Usually, hidden visibility
622 indicates that the symbol will not be placed into the dynamic symbol table,
623 so no other module (executable or shared library) can reference it
627 <dt><b>"<tt>protected</tt>" - Protected style</b>:</dt>
629 <dd>On ELF, protected visibility indicates that the symbol will be placed in
630 the dynamic symbol table, but that references within the defining module will
631 bind to the local symbol. That is, the symbol cannot be overridden by another
638 <!-- ======================================================================= -->
639 <div class="doc_subsection">
640 <a name="globalvars">Global Variables</a>
643 <div class="doc_text">
645 <p>Global variables define regions of memory allocated at compilation time
646 instead of run-time. Global variables may optionally be initialized, may have
647 an explicit section to be placed in, and may have an optional explicit alignment
648 specified. A variable may be defined as "thread_local", which means that it
649 will not be shared by threads (each thread will have a separated copy of the
650 variable). A variable may be defined as a global "constant," which indicates
651 that the contents of the variable will <b>never</b> be modified (enabling better
652 optimization, allowing the global data to be placed in the read-only section of
653 an executable, etc). Note that variables that need runtime initialization
654 cannot be marked "constant" as there is a store to the variable.</p>
657 LLVM explicitly allows <em>declarations</em> of global variables to be marked
658 constant, even if the final definition of the global is not. This capability
659 can be used to enable slightly better optimization of the program, but requires
660 the language definition to guarantee that optimizations based on the
661 'constantness' are valid for the translation units that do not include the
665 <p>As SSA values, global variables define pointer values that are in
666 scope (i.e. they dominate) all basic blocks in the program. Global
667 variables always define a pointer to their "content" type because they
668 describe a region of memory, and all memory objects in LLVM are
669 accessed through pointers.</p>
671 <p>A global variable may be declared to reside in a target-specifc numbered
672 address space. For targets that support them, address spaces may affect how
673 optimizations are performed and/or what target instructions are used to access
674 the variable. The default address space is zero. The address space qualifier
675 must precede any other attributes.</p>
677 <p>LLVM allows an explicit section to be specified for globals. If the target
678 supports it, it will emit globals to the section specified.</p>
680 <p>An explicit alignment may be specified for a global. If not present, or if
681 the alignment is set to zero, the alignment of the global is set by the target
682 to whatever it feels convenient. If an explicit alignment is specified, the
683 global is forced to have at least that much alignment. All alignments must be
686 <p>For example, the following defines a global in a numbered address space with
687 an initializer, section, and alignment:</p>
689 <div class="doc_code">
691 @G = constant float 1.0 addrspace(5), section "foo", align 4
698 <!-- ======================================================================= -->
699 <div class="doc_subsection">
700 <a name="functionstructure">Functions</a>
703 <div class="doc_text">
705 <p>LLVM function definitions consist of the "<tt>define</tt>" keyord,
706 an 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) argument list (each with optional
711 <a href="#paramattrs">parameter attributes</a>), an optional section, an
712 optional alignment, an optional <a href="#gc">garbage collector name</a>, an
713 opening curly brace, a list of basic blocks, and a closing curly brace.
715 LLVM function declarations consist of the "<tt>declare</tt>" keyword, an
716 optional <a href="#linkage">linkage type</a>, an optional
717 <a href="#visibility">visibility style</a>, an optional
718 <a href="#callingconv">calling convention</a>, a return type, an optional
719 <a href="#paramattrs">parameter attribute</a> for the return type, a function
720 name, a possibly empty list of arguments, an optional alignment, and an optional
721 <a href="#gc">garbage collector name</a>.</p>
723 <p>A function definition contains a list of basic blocks, forming the CFG for
724 the function. Each basic block may optionally start with a label (giving the
725 basic block a symbol table entry), contains a list of instructions, and ends
726 with a <a href="#terminators">terminator</a> instruction (such as a branch or
727 function return).</p>
729 <p>The first basic block in a function is special in two ways: it is immediately
730 executed on entrance to the function, and it is not allowed to have predecessor
731 basic blocks (i.e. there can not be any branches to the entry block of a
732 function). Because the block can have no predecessors, it also cannot have any
733 <a href="#i_phi">PHI nodes</a>.</p>
735 <p>LLVM allows an explicit section to be specified for functions. If the target
736 supports it, it will emit functions to the section specified.</p>
738 <p>An explicit alignment may be specified for a function. If not present, or if
739 the alignment is set to zero, the alignment of the function is set by the target
740 to whatever it feels convenient. If an explicit alignment is specified, the
741 function is forced to have at least that much alignment. All alignments must be
747 <!-- ======================================================================= -->
748 <div class="doc_subsection">
749 <a name="aliasstructure">Aliases</a>
751 <div class="doc_text">
752 <p>Aliases act as "second name" for the aliasee value (which can be either
753 function or global variable or bitcast of global value). Aliases may have an
754 optional <a href="#linkage">linkage type</a>, and an
755 optional <a href="#visibility">visibility style</a>.</p>
759 <div class="doc_code">
761 @<Name> = [Linkage] [Visibility] alias <AliaseeTy> @<Aliasee>
769 <!-- ======================================================================= -->
770 <div class="doc_subsection"><a name="paramattrs">Parameter Attributes</a></div>
771 <div class="doc_text">
772 <p>The return type and each parameter of a function type may have a set of
773 <i>parameter attributes</i> associated with them. Parameter attributes are
774 used to communicate additional information about the result or parameters of
775 a function. Parameter attributes are considered to be part of the function,
776 not of the function type, so functions with different parameter attributes
777 can have the same function type.</p>
779 <p>Parameter attributes are simple keywords that follow the type specified. If
780 multiple parameter attributes are needed, they are space separated. For
783 <div class="doc_code">
785 declare i32 @printf(i8* noalias , ...) nounwind
786 declare i32 @atoi(i8*) nounwind readonly
790 <p>Note that any attributes for the function result (<tt>nounwind</tt>,
791 <tt>readonly</tt>) come immediately after the argument list.</p>
793 <p>Currently, only the following parameter attributes are defined:</p>
795 <dt><tt>zeroext</tt></dt>
796 <dd>This indicates that the parameter should be zero extended just before
797 a call to this function.</dd>
798 <dt><tt>signext</tt></dt>
799 <dd>This indicates that the parameter should be sign extended just before
800 a call to this function.</dd>
801 <dt><tt>inreg</tt></dt>
802 <dd>This indicates that the parameter should be placed in register (if
803 possible) during assembling function call. Support for this attribute is
805 <dt><tt>sret</tt></dt>
806 <dd>This indicates that the parameter specifies the address of a structure
807 that is the return value of the function in the source program.</dd>
808 <dt><tt>noalias</tt></dt>
809 <dd>This indicates that the parameter not alias any other object or any
810 other "noalias" objects during the function call.
811 <dt><tt>noreturn</tt></dt>
812 <dd>This function attribute indicates that the function never returns. This
813 indicates to LLVM that every call to this function should be treated as if
814 an <tt>unreachable</tt> instruction immediately followed the call.</dd>
815 <dt><tt>nounwind</tt></dt>
816 <dd>This function attribute indicates that the function type does not use
817 the unwind instruction and does not allow stack unwinding to propagate
819 <dt><tt>nest</tt></dt>
820 <dd>This indicates that the parameter can be excised using the
821 <a href="#int_trampoline">trampoline intrinsics</a>.</dd>
822 <dt><tt>readonly</tt></dt>
823 <dd>This function attribute indicates that the function has no side-effects
824 except for producing a return value or throwing an exception. The value
825 returned must only depend on the function arguments and/or global variables.
826 It may use values obtained by dereferencing pointers.</dd>
827 <dt><tt>readnone</tt></dt>
828 <dd>A <tt>readnone</tt> function has the same restrictions as a <tt>readonly</tt>
829 function, but in addition it is not allowed to dereference any pointer arguments
835 <!-- ======================================================================= -->
836 <div class="doc_subsection">
837 <a name="gc">Garbage Collector Names</a>
840 <div class="doc_text">
841 <p>Each function may specify a garbage collector name, which is simply a
844 <div class="doc_code"><pre
845 >define void @f() gc "name" { ...</pre></div>
847 <p>The compiler declares the supported values of <i>name</i>. Specifying a
848 collector which will cause the compiler to alter its output in order to support
849 the named garbage collection algorithm.</p>
852 <!-- ======================================================================= -->
853 <div class="doc_subsection">
854 <a name="moduleasm">Module-Level Inline Assembly</a>
857 <div class="doc_text">
859 Modules may contain "module-level inline asm" blocks, which corresponds to the
860 GCC "file scope inline asm" blocks. These blocks are internally concatenated by
861 LLVM and treated as a single unit, but may be separated in the .ll file if
862 desired. The syntax is very simple:
865 <div class="doc_code">
867 module asm "inline asm code goes here"
868 module asm "more can go here"
872 <p>The strings can contain any character by escaping non-printable characters.
873 The escape sequence used is simply "\xx" where "xx" is the two digit hex code
878 The inline asm code is simply printed to the machine code .s file when
879 assembly code is generated.
883 <!-- ======================================================================= -->
884 <div class="doc_subsection">
885 <a name="datalayout">Data Layout</a>
888 <div class="doc_text">
889 <p>A module may specify a target specific data layout string that specifies how
890 data is to be laid out in memory. The syntax for the data layout is simply:</p>
891 <pre> target datalayout = "<i>layout specification</i>"</pre>
892 <p>The <i>layout specification</i> consists of a list of specifications
893 separated by the minus sign character ('-'). Each specification starts with a
894 letter and may include other information after the letter to define some
895 aspect of the data layout. The specifications accepted are as follows: </p>
898 <dd>Specifies that the target lays out data in big-endian form. That is, the
899 bits with the most significance have the lowest address location.</dd>
901 <dd>Specifies that hte target lays out data in little-endian form. That is,
902 the bits with the least significance have the lowest address location.</dd>
903 <dt><tt>p:<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
904 <dd>This specifies the <i>size</i> of a pointer and its <i>abi</i> and
905 <i>preferred</i> alignments. All sizes are in bits. Specifying the <i>pref</i>
906 alignment is optional. If omitted, the preceding <tt>:</tt> should be omitted
908 <dt><tt>i<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
909 <dd>This specifies the alignment for an integer type of a given bit
910 <i>size</i>. The value of <i>size</i> must be in the range [1,2^23).</dd>
911 <dt><tt>v<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
912 <dd>This specifies the alignment for a vector type of a given bit
914 <dt><tt>f<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
915 <dd>This specifies the alignment for a floating point type of a given bit
916 <i>size</i>. The value of <i>size</i> must be either 32 (float) or 64
918 <dt><tt>a<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
919 <dd>This specifies the alignment for an aggregate type of a given bit
922 <p>When constructing the data layout for a given target, LLVM starts with a
923 default set of specifications which are then (possibly) overriden by the
924 specifications in the <tt>datalayout</tt> keyword. The default specifications
925 are given in this list:</p>
927 <li><tt>E</tt> - big endian</li>
928 <li><tt>p:32:64:64</tt> - 32-bit pointers with 64-bit alignment</li>
929 <li><tt>i1:8:8</tt> - i1 is 8-bit (byte) aligned</li>
930 <li><tt>i8:8:8</tt> - i8 is 8-bit (byte) aligned</li>
931 <li><tt>i16:16:16</tt> - i16 is 16-bit aligned</li>
932 <li><tt>i32:32:32</tt> - i32 is 32-bit aligned</li>
933 <li><tt>i64:32:64</tt> - i64 has abi alignment of 32-bits but preferred
934 alignment of 64-bits</li>
935 <li><tt>f32:32:32</tt> - float is 32-bit aligned</li>
936 <li><tt>f64:64:64</tt> - double is 64-bit aligned</li>
937 <li><tt>v64:64:64</tt> - 64-bit vector is 64-bit aligned</li>
938 <li><tt>v128:128:128</tt> - 128-bit vector is 128-bit aligned</li>
939 <li><tt>a0:0:1</tt> - aggregates are 8-bit aligned</li>
941 <p>When llvm is determining the alignment for a given type, it uses the
944 <li>If the type sought is an exact match for one of the specifications, that
945 specification is used.</li>
946 <li>If no match is found, and the type sought is an integer type, then the
947 smallest integer type that is larger than the bitwidth of the sought type is
948 used. If none of the specifications are larger than the bitwidth then the the
949 largest integer type is used. For example, given the default specifications
950 above, the i7 type will use the alignment of i8 (next largest) while both
951 i65 and i256 will use the alignment of i64 (largest specified).</li>
952 <li>If no match is found, and the type sought is a vector type, then the
953 largest vector type that is smaller than the sought vector type will be used
954 as a fall back. This happens because <128 x double> can be implemented in
955 terms of 64 <2 x double>, for example.</li>
959 <!-- *********************************************************************** -->
960 <div class="doc_section"> <a name="typesystem">Type System</a> </div>
961 <!-- *********************************************************************** -->
963 <div class="doc_text">
965 <p>The LLVM type system is one of the most important features of the
966 intermediate representation. Being typed enables a number of
967 optimizations to be performed on the IR directly, without having to do
968 extra analyses on the side before the transformation. A strong type
969 system makes it easier to read the generated code and enables novel
970 analyses and transformations that are not feasible to perform on normal
971 three address code representations.</p>
975 <!-- ======================================================================= -->
976 <div class="doc_subsection"> <a name="t_primitive">Primitive Types</a> </div>
977 <div class="doc_text">
978 <p>The primitive types are the fundamental building blocks of the LLVM
979 system. The current set of primitive types is as follows:</p>
981 <table class="layout">
986 <tr><th>Type</th><th>Description</th></tr>
987 <tr><td><tt><a name="t_void">void</a></tt></td><td>No value</td></tr>
988 <tr><td><tt>label</tt></td><td>Branch destination</td></tr>
995 <tr><th>Type</th><th>Description</th></tr>
996 <tr><td><tt>float</tt></td><td>32-bit floating point value</td></tr>
997 <tr><td><tt>double</tt></td><td>64-bit floating point value</td></tr>
1005 <!-- _______________________________________________________________________ -->
1006 <div class="doc_subsubsection"> <a name="t_classifications">Type
1007 Classifications</a> </div>
1008 <div class="doc_text">
1009 <p>These different primitive types fall into a few useful
1010 classifications:</p>
1012 <table border="1" cellspacing="0" cellpadding="4">
1014 <tr><th>Classification</th><th>Types</th></tr>
1016 <td><a name="t_integer">integer</a></td>
1017 <td><tt>i1, i2, i3, ... i8, ... i16, ... i32, ... i64, ... </tt></td>
1020 <td><a name="t_floating">floating point</a></td>
1021 <td><tt>float, double</tt></td>
1024 <td><a name="t_firstclass">first class</a></td>
1025 <td><tt>i1, ..., float, double, <br/>
1026 <a href="#t_pointer">pointer</a>,<a href="#t_vector">vector</a></tt>
1032 <p>The <a href="#t_firstclass">first class</a> types are perhaps the
1033 most important. Values of these types are the only ones which can be
1034 produced by instructions, passed as arguments, or used as operands to
1035 instructions. This means that all structures and arrays must be
1036 manipulated either by pointer or by component.</p>
1039 <!-- ======================================================================= -->
1040 <div class="doc_subsection"> <a name="t_derived">Derived Types</a> </div>
1042 <div class="doc_text">
1044 <p>The real power in LLVM comes from the derived types in the system.
1045 This is what allows a programmer to represent arrays, functions,
1046 pointers, and other useful types. Note that these derived types may be
1047 recursive: For example, it is possible to have a two dimensional array.</p>
1051 <!-- _______________________________________________________________________ -->
1052 <div class="doc_subsubsection"> <a name="t_integer">Integer Type</a> </div>
1054 <div class="doc_text">
1057 <p>The integer type is a very simple derived type that simply specifies an
1058 arbitrary bit width for the integer type desired. Any bit width from 1 bit to
1059 2^23-1 (about 8 million) can be specified.</p>
1067 <p>The number of bits the integer will occupy is specified by the <tt>N</tt>
1071 <table class="layout">
1081 <tt>i1942652</tt><br/>
1084 A boolean integer of 1 bit<br/>
1085 A nibble sized integer of 4 bits.<br/>
1086 A byte sized integer of 8 bits.<br/>
1087 A half word sized integer of 16 bits.<br/>
1088 A word sized integer of 32 bits.<br/>
1089 An integer whose bit width is the answer. <br/>
1090 A double word sized integer of 64 bits.<br/>
1091 A really big integer of over 1 million bits.<br/>
1097 <!-- _______________________________________________________________________ -->
1098 <div class="doc_subsubsection"> <a name="t_array">Array Type</a> </div>
1100 <div class="doc_text">
1104 <p>The array type is a very simple derived type that arranges elements
1105 sequentially in memory. The array type requires a size (number of
1106 elements) and an underlying data type.</p>
1111 [<# elements> x <elementtype>]
1114 <p>The number of elements is a constant integer value; elementtype may
1115 be any type with a size.</p>
1118 <table class="layout">
1121 <tt>[40 x i32 ]</tt><br/>
1122 <tt>[41 x i32 ]</tt><br/>
1123 <tt>[40 x i8]</tt><br/>
1126 Array of 40 32-bit integer values.<br/>
1127 Array of 41 32-bit integer values.<br/>
1128 Array of 40 8-bit integer values.<br/>
1132 <p>Here are some examples of multidimensional arrays:</p>
1133 <table class="layout">
1136 <tt>[3 x [4 x i32]]</tt><br/>
1137 <tt>[12 x [10 x float]]</tt><br/>
1138 <tt>[2 x [3 x [4 x i16]]]</tt><br/>
1141 3x4 array of 32-bit integer values.<br/>
1142 12x10 array of single precision floating point values.<br/>
1143 2x3x4 array of 16-bit integer values.<br/>
1148 <p>Note that 'variable sized arrays' can be implemented in LLVM with a zero
1149 length array. Normally, accesses past the end of an array are undefined in
1150 LLVM (e.g. it is illegal to access the 5th element of a 3 element array).
1151 As a special case, however, zero length arrays are recognized to be variable
1152 length. This allows implementation of 'pascal style arrays' with the LLVM
1153 type "{ i32, [0 x float]}", for example.</p>
1157 <!-- _______________________________________________________________________ -->
1158 <div class="doc_subsubsection"> <a name="t_function">Function Type</a> </div>
1159 <div class="doc_text">
1161 <p>The function type can be thought of as a function signature. It
1162 consists of a return type and a list of formal parameter types.
1163 Function types are usually used to build virtual function tables
1164 (which are structures of pointers to functions), for indirect function
1165 calls, and when defining a function.</p>
1167 The return type of a function type cannot be an aggregate type.
1170 <pre> <returntype> (<parameter list>)<br></pre>
1171 <p>...where '<tt><parameter list></tt>' is a comma-separated list of type
1172 specifiers. Optionally, the parameter list may include a type <tt>...</tt>,
1173 which indicates that the function takes a variable number of arguments.
1174 Variable argument functions can access their arguments with the <a
1175 href="#int_varargs">variable argument handling intrinsic</a> functions.</p>
1177 <table class="layout">
1179 <td class="left"><tt>i32 (i32)</tt></td>
1180 <td class="left">function taking an <tt>i32</tt>, returning an <tt>i32</tt>
1182 </tr><tr class="layout">
1183 <td class="left"><tt>float (i16 signext, i32 *) *
1185 <td class="left"><a href="#t_pointer">Pointer</a> to a function that takes
1186 an <tt>i16</tt> that should be sign extended and a
1187 <a href="#t_pointer">pointer</a> to <tt>i32</tt>, returning
1190 </tr><tr class="layout">
1191 <td class="left"><tt>i32 (i8*, ...)</tt></td>
1192 <td class="left">A vararg function that takes at least one
1193 <a href="#t_pointer">pointer</a> to <tt>i8 </tt> (char in C),
1194 which returns an integer. This is the signature for <tt>printf</tt> in
1201 <!-- _______________________________________________________________________ -->
1202 <div class="doc_subsubsection"> <a name="t_struct">Structure Type</a> </div>
1203 <div class="doc_text">
1205 <p>The structure type is used to represent a collection of data members
1206 together in memory. The packing of the field types is defined to match
1207 the ABI of the underlying processor. The elements of a structure may
1208 be any type that has a size.</p>
1209 <p>Structures are accessed using '<tt><a href="#i_load">load</a></tt>
1210 and '<tt><a href="#i_store">store</a></tt>' by getting a pointer to a
1211 field with the '<tt><a href="#i_getelementptr">getelementptr</a></tt>'
1214 <pre> { <type list> }<br></pre>
1216 <table class="layout">
1218 <td class="left"><tt>{ i32, i32, i32 }</tt></td>
1219 <td class="left">A triple of three <tt>i32</tt> values</td>
1220 </tr><tr class="layout">
1221 <td class="left"><tt>{ float, i32 (i32) * }</tt></td>
1222 <td class="left">A pair, where the first element is a <tt>float</tt> and the
1223 second element is a <a href="#t_pointer">pointer</a> to a
1224 <a href="#t_function">function</a> that takes an <tt>i32</tt>, returning
1225 an <tt>i32</tt>.</td>
1230 <!-- _______________________________________________________________________ -->
1231 <div class="doc_subsubsection"> <a name="t_pstruct">Packed Structure Type</a>
1233 <div class="doc_text">
1235 <p>The packed structure type is used to represent a collection of data members
1236 together in memory. There is no padding between fields. Further, the alignment
1237 of a packed structure is 1 byte. The elements of a packed structure may
1238 be any type that has a size.</p>
1239 <p>Structures are accessed using '<tt><a href="#i_load">load</a></tt>
1240 and '<tt><a href="#i_store">store</a></tt>' by getting a pointer to a
1241 field with the '<tt><a href="#i_getelementptr">getelementptr</a></tt>'
1244 <pre> < { <type list> } > <br></pre>
1246 <table class="layout">
1248 <td class="left"><tt>< { i32, i32, i32 } ></tt></td>
1249 <td class="left">A triple of three <tt>i32</tt> values</td>
1250 </tr><tr class="layout">
1251 <td class="left"><tt>< { float, i32 (i32) * } ></tt></td>
1252 <td class="left">A pair, where the first element is a <tt>float</tt> and the
1253 second element is a <a href="#t_pointer">pointer</a> to a
1254 <a href="#t_function">function</a> that takes an <tt>i32</tt>, returning
1255 an <tt>i32</tt>.</td>
1260 <!-- _______________________________________________________________________ -->
1261 <div class="doc_subsubsection"> <a name="t_pointer">Pointer Type</a> </div>
1262 <div class="doc_text">
1264 <p>As in many languages, the pointer type represents a pointer or
1265 reference to another object, which must live in memory. Pointer types may have
1266 an optional address space attribute defining the target-specific numbered
1267 address space where the pointed-to object resides. The default address space is
1270 <pre> <type> *<br></pre>
1272 <table class="layout">
1275 <tt>[4x i32]*</tt><br/>
1276 <tt>i32 (i32 *) *</tt><br/>
1277 <tt>i32 addrspace(5)*</tt><br/>
1280 A <a href="#t_pointer">pointer</a> to <a href="#t_array">array</a> of
1281 four <tt>i32</tt> values<br/>
1282 A <a href="#t_pointer">pointer</a> to a <a
1283 href="#t_function">function</a> that takes an <tt>i32*</tt>, returning an
1285 A <a href="#t_pointer">pointer</a> to an <tt>i32</tt> value that resides
1286 in address space 5.<br/>
1292 <!-- _______________________________________________________________________ -->
1293 <div class="doc_subsubsection"> <a name="t_vector">Vector Type</a> </div>
1294 <div class="doc_text">
1298 <p>A vector type is a simple derived type that represents a vector
1299 of elements. Vector types are used when multiple primitive data
1300 are operated in parallel using a single instruction (SIMD).
1301 A vector type requires a size (number of
1302 elements) and an underlying primitive data type. Vectors must have a power
1303 of two length (1, 2, 4, 8, 16 ...). Vector types are
1304 considered <a href="#t_firstclass">first class</a>.</p>
1309 < <# elements> x <elementtype> >
1312 <p>The number of elements is a constant integer value; elementtype may
1313 be any integer or floating point type.</p>
1317 <table class="layout">
1320 <tt><4 x i32></tt><br/>
1321 <tt><8 x float></tt><br/>
1322 <tt><2 x i64></tt><br/>
1325 Vector of 4 32-bit integer values.<br/>
1326 Vector of 8 floating-point values.<br/>
1327 Vector of 2 64-bit integer values.<br/>
1333 <!-- _______________________________________________________________________ -->
1334 <div class="doc_subsubsection"> <a name="t_opaque">Opaque Type</a> </div>
1335 <div class="doc_text">
1339 <p>Opaque types are used to represent unknown types in the system. This
1340 corresponds (for example) to the C notion of a forward declared structure type.
1341 In LLVM, opaque types can eventually be resolved to any type (not just a
1342 structure type).</p>
1352 <table class="layout">
1358 An opaque type.<br/>
1365 <!-- *********************************************************************** -->
1366 <div class="doc_section"> <a name="constants">Constants</a> </div>
1367 <!-- *********************************************************************** -->
1369 <div class="doc_text">
1371 <p>LLVM has several different basic types of constants. This section describes
1372 them all and their syntax.</p>
1376 <!-- ======================================================================= -->
1377 <div class="doc_subsection"><a name="simpleconstants">Simple Constants</a></div>
1379 <div class="doc_text">
1382 <dt><b>Boolean constants</b></dt>
1384 <dd>The two strings '<tt>true</tt>' and '<tt>false</tt>' are both valid
1385 constants of the <tt><a href="#t_primitive">i1</a></tt> type.
1388 <dt><b>Integer constants</b></dt>
1390 <dd>Standard integers (such as '4') are constants of the <a
1391 href="#t_integer">integer</a> type. Negative numbers may be used with
1395 <dt><b>Floating point constants</b></dt>
1397 <dd>Floating point constants use standard decimal notation (e.g. 123.421),
1398 exponential notation (e.g. 1.23421e+2), or a more precise hexadecimal
1399 notation (see below). Floating point constants must have a <a
1400 href="#t_floating">floating point</a> type. </dd>
1402 <dt><b>Null pointer constants</b></dt>
1404 <dd>The identifier '<tt>null</tt>' is recognized as a null pointer constant
1405 and must be of <a href="#t_pointer">pointer type</a>.</dd>
1409 <p>The one non-intuitive notation for constants is the optional hexadecimal form
1410 of floating point constants. For example, the form '<tt>double
1411 0x432ff973cafa8000</tt>' is equivalent to (but harder to read than) '<tt>double
1412 4.5e+15</tt>'. The only time hexadecimal floating point constants are required
1413 (and the only time that they are generated by the disassembler) is when a
1414 floating point constant must be emitted but it cannot be represented as a
1415 decimal floating point number. For example, NaN's, infinities, and other
1416 special values are represented in their IEEE hexadecimal format so that
1417 assembly and disassembly do not cause any bits to change in the constants.</p>
1421 <!-- ======================================================================= -->
1422 <div class="doc_subsection"><a name="aggregateconstants">Aggregate Constants</a>
1425 <div class="doc_text">
1426 <p>Aggregate constants arise from aggregation of simple constants
1427 and smaller aggregate constants.</p>
1430 <dt><b>Structure constants</b></dt>
1432 <dd>Structure constants are represented with notation similar to structure
1433 type definitions (a comma separated list of elements, surrounded by braces
1434 (<tt>{}</tt>)). For example: "<tt>{ i32 4, float 17.0, i32* %G }</tt>",
1435 where "<tt>%G</tt>" is declared as "<tt>@G = external global i32</tt>". Structure constants
1436 must have <a href="#t_struct">structure type</a>, and the number and
1437 types of elements must match those specified by the type.
1440 <dt><b>Array constants</b></dt>
1442 <dd>Array constants are represented with notation similar to array type
1443 definitions (a comma separated list of elements, surrounded by square brackets
1444 (<tt>[]</tt>)). For example: "<tt>[ i32 42, i32 11, i32 74 ]</tt>". Array
1445 constants must have <a href="#t_array">array type</a>, and the number and
1446 types of elements must match those specified by the type.
1449 <dt><b>Vector constants</b></dt>
1451 <dd>Vector constants are represented with notation similar to vector type
1452 definitions (a comma separated list of elements, surrounded by
1453 less-than/greater-than's (<tt><></tt>)). For example: "<tt>< i32 42,
1454 i32 11, i32 74, i32 100 ></tt>". Vector constants must have <a
1455 href="#t_vector">vector type</a>, and the number and types of elements must
1456 match those specified by the type.
1459 <dt><b>Zero initialization</b></dt>
1461 <dd>The string '<tt>zeroinitializer</tt>' can be used to zero initialize a
1462 value to zero of <em>any</em> type, including scalar and aggregate types.
1463 This is often used to avoid having to print large zero initializers (e.g. for
1464 large arrays) and is always exactly equivalent to using explicit zero
1471 <!-- ======================================================================= -->
1472 <div class="doc_subsection">
1473 <a name="globalconstants">Global Variable and Function Addresses</a>
1476 <div class="doc_text">
1478 <p>The addresses of <a href="#globalvars">global variables</a> and <a
1479 href="#functionstructure">functions</a> are always implicitly valid (link-time)
1480 constants. These constants are explicitly referenced when the <a
1481 href="#identifiers">identifier for the global</a> is used and always have <a
1482 href="#t_pointer">pointer</a> type. For example, the following is a legal LLVM
1485 <div class="doc_code">
1489 @Z = global [2 x i32*] [ i32* @X, i32* @Y ]
1495 <!-- ======================================================================= -->
1496 <div class="doc_subsection"><a name="undefvalues">Undefined Values</a></div>
1497 <div class="doc_text">
1498 <p>The string '<tt>undef</tt>' is recognized as a type-less constant that has
1499 no specific value. Undefined values may be of any type and be used anywhere
1500 a constant is permitted.</p>
1502 <p>Undefined values indicate to the compiler that the program is well defined
1503 no matter what value is used, giving the compiler more freedom to optimize.
1507 <!-- ======================================================================= -->
1508 <div class="doc_subsection"><a name="constantexprs">Constant Expressions</a>
1511 <div class="doc_text">
1513 <p>Constant expressions are used to allow expressions involving other constants
1514 to be used as constants. Constant expressions may be of any <a
1515 href="#t_firstclass">first class</a> type and may involve any LLVM operation
1516 that does not have side effects (e.g. load and call are not supported). The
1517 following is the syntax for constant expressions:</p>
1520 <dt><b><tt>trunc ( CST to TYPE )</tt></b></dt>
1521 <dd>Truncate a constant to another type. The bit size of CST must be larger
1522 than the bit size of TYPE. Both types must be integers.</dd>
1524 <dt><b><tt>zext ( CST to TYPE )</tt></b></dt>
1525 <dd>Zero extend a constant to another type. The bit size of CST must be
1526 smaller or equal to the bit size of TYPE. Both types must be integers.</dd>
1528 <dt><b><tt>sext ( CST to TYPE )</tt></b></dt>
1529 <dd>Sign extend a constant to another type. The bit size of CST must be
1530 smaller or equal to the bit size of TYPE. Both types must be integers.</dd>
1532 <dt><b><tt>fptrunc ( CST to TYPE )</tt></b></dt>
1533 <dd>Truncate a floating point constant to another floating point type. The
1534 size of CST must be larger than the size of TYPE. Both types must be
1535 floating point.</dd>
1537 <dt><b><tt>fpext ( CST to TYPE )</tt></b></dt>
1538 <dd>Floating point extend a constant to another type. The size of CST must be
1539 smaller or equal to the size of TYPE. Both types must be floating point.</dd>
1541 <dt><b><tt>fptoui ( CST to TYPE )</tt></b></dt>
1542 <dd>Convert a floating point constant to the corresponding unsigned integer
1543 constant. TYPE must be a scalar or vector integer type. CST must be of scalar
1544 or vector floating point type. Both CST and TYPE must be scalars, or vectors
1545 of the same number of elements. If the value won't fit in the integer type,
1546 the results are undefined.</dd>
1548 <dt><b><tt>fptosi ( CST to TYPE )</tt></b></dt>
1549 <dd>Convert a floating point constant to the corresponding signed integer
1550 constant. TYPE must be a scalar or vector integer type. CST must be of scalar
1551 or vector floating point type. Both CST and TYPE must be scalars, or vectors
1552 of the same number of elements. If the value won't fit in the integer type,
1553 the results are undefined.</dd>
1555 <dt><b><tt>uitofp ( CST to TYPE )</tt></b></dt>
1556 <dd>Convert an unsigned integer constant to the corresponding floating point
1557 constant. TYPE must be a scalar or vector floating point type. CST must be of
1558 scalar or vector integer type. Both CST and TYPE must be scalars, or vectors
1559 of the same number of elements. If the value won't fit in the floating point
1560 type, the results are undefined.</dd>
1562 <dt><b><tt>sitofp ( CST to TYPE )</tt></b></dt>
1563 <dd>Convert a signed integer constant to the corresponding floating point
1564 constant. TYPE must be a scalar or vector floating point type. CST must be of
1565 scalar or vector integer type. Both CST and TYPE must be scalars, or vectors
1566 of the same number of elements. If the value won't fit in the floating point
1567 type, the results are undefined.</dd>
1569 <dt><b><tt>ptrtoint ( CST to TYPE )</tt></b></dt>
1570 <dd>Convert a pointer typed constant to the corresponding integer constant
1571 TYPE must be an integer type. CST must be of pointer type. The CST value is
1572 zero extended, truncated, or unchanged to make it fit in TYPE.</dd>
1574 <dt><b><tt>inttoptr ( CST to TYPE )</tt></b></dt>
1575 <dd>Convert a integer constant to a pointer constant. TYPE must be a
1576 pointer type. CST must be of integer type. The CST value is zero extended,
1577 truncated, or unchanged to make it fit in a pointer size. This one is
1578 <i>really</i> dangerous!</dd>
1580 <dt><b><tt>bitcast ( CST to TYPE )</tt></b></dt>
1581 <dd>Convert a constant, CST, to another TYPE. The size of CST and TYPE must be
1582 identical (same number of bits). The conversion is done as if the CST value
1583 was stored to memory and read back as TYPE. In other words, no bits change
1584 with this operator, just the type. This can be used for conversion of
1585 vector types to any other type, as long as they have the same bit width. For
1586 pointers it is only valid to cast to another pointer type.
1589 <dt><b><tt>getelementptr ( CSTPTR, IDX0, IDX1, ... )</tt></b></dt>
1591 <dd>Perform the <a href="#i_getelementptr">getelementptr operation</a> on
1592 constants. As with the <a href="#i_getelementptr">getelementptr</a>
1593 instruction, the index list may have zero or more indexes, which are required
1594 to make sense for the type of "CSTPTR".</dd>
1596 <dt><b><tt>select ( COND, VAL1, VAL2 )</tt></b></dt>
1598 <dd>Perform the <a href="#i_select">select operation</a> on
1601 <dt><b><tt>icmp COND ( VAL1, VAL2 )</tt></b></dt>
1602 <dd>Performs the <a href="#i_icmp">icmp operation</a> on constants.</dd>
1604 <dt><b><tt>fcmp COND ( VAL1, VAL2 )</tt></b></dt>
1605 <dd>Performs the <a href="#i_fcmp">fcmp operation</a> on constants.</dd>
1607 <dt><b><tt>extractelement ( VAL, IDX )</tt></b></dt>
1609 <dd>Perform the <a href="#i_extractelement">extractelement
1610 operation</a> on constants.
1612 <dt><b><tt>insertelement ( VAL, ELT, IDX )</tt></b></dt>
1614 <dd>Perform the <a href="#i_insertelement">insertelement
1615 operation</a> on constants.</dd>
1618 <dt><b><tt>shufflevector ( VEC1, VEC2, IDXMASK )</tt></b></dt>
1620 <dd>Perform the <a href="#i_shufflevector">shufflevector
1621 operation</a> on constants.</dd>
1623 <dt><b><tt>OPCODE ( LHS, RHS )</tt></b></dt>
1625 <dd>Perform the specified operation of the LHS and RHS constants. OPCODE may
1626 be any of the <a href="#binaryops">binary</a> or <a href="#bitwiseops">bitwise
1627 binary</a> operations. The constraints on operands are the same as those for
1628 the corresponding instruction (e.g. no bitwise operations on floating point
1629 values are allowed).</dd>
1633 <!-- *********************************************************************** -->
1634 <div class="doc_section"> <a name="othervalues">Other Values</a> </div>
1635 <!-- *********************************************************************** -->
1637 <!-- ======================================================================= -->
1638 <div class="doc_subsection">
1639 <a name="inlineasm">Inline Assembler Expressions</a>
1642 <div class="doc_text">
1645 LLVM supports inline assembler expressions (as opposed to <a href="#moduleasm">
1646 Module-Level Inline Assembly</a>) through the use of a special value. This
1647 value represents the inline assembler as a string (containing the instructions
1648 to emit), a list of operand constraints (stored as a string), and a flag that
1649 indicates whether or not the inline asm expression has side effects. An example
1650 inline assembler expression is:
1653 <div class="doc_code">
1655 i32 (i32) asm "bswap $0", "=r,r"
1660 Inline assembler expressions may <b>only</b> be used as the callee operand of
1661 a <a href="#i_call"><tt>call</tt> instruction</a>. Thus, typically we have:
1664 <div class="doc_code">
1666 %X = call i32 asm "<a href="#int_bswap">bswap</a> $0", "=r,r"(i32 %Y)
1671 Inline asms with side effects not visible in the constraint list must be marked
1672 as having side effects. This is done through the use of the
1673 '<tt>sideeffect</tt>' keyword, like so:
1676 <div class="doc_code">
1678 call void asm sideeffect "eieio", ""()
1682 <p>TODO: The format of the asm and constraints string still need to be
1683 documented here. Constraints on what can be done (e.g. duplication, moving, etc
1684 need to be documented).
1689 <!-- *********************************************************************** -->
1690 <div class="doc_section"> <a name="instref">Instruction Reference</a> </div>
1691 <!-- *********************************************************************** -->
1693 <div class="doc_text">
1695 <p>The LLVM instruction set consists of several different
1696 classifications of instructions: <a href="#terminators">terminator
1697 instructions</a>, <a href="#binaryops">binary instructions</a>,
1698 <a href="#bitwiseops">bitwise binary instructions</a>, <a
1699 href="#memoryops">memory instructions</a>, and <a href="#otherops">other
1700 instructions</a>.</p>
1704 <!-- ======================================================================= -->
1705 <div class="doc_subsection"> <a name="terminators">Terminator
1706 Instructions</a> </div>
1708 <div class="doc_text">
1710 <p>As mentioned <a href="#functionstructure">previously</a>, every
1711 basic block in a program ends with a "Terminator" instruction, which
1712 indicates which block should be executed after the current block is
1713 finished. These terminator instructions typically yield a '<tt>void</tt>'
1714 value: they produce control flow, not values (the one exception being
1715 the '<a href="#i_invoke"><tt>invoke</tt></a>' instruction).</p>
1716 <p>There are six different terminator instructions: the '<a
1717 href="#i_ret"><tt>ret</tt></a>' instruction, the '<a href="#i_br"><tt>br</tt></a>'
1718 instruction, the '<a href="#i_switch"><tt>switch</tt></a>' instruction,
1719 the '<a href="#i_invoke"><tt>invoke</tt></a>' instruction, the '<a
1720 href="#i_unwind"><tt>unwind</tt></a>' instruction, and the '<a
1721 href="#i_unreachable"><tt>unreachable</tt></a>' instruction.</p>
1725 <!-- _______________________________________________________________________ -->
1726 <div class="doc_subsubsection"> <a name="i_ret">'<tt>ret</tt>'
1727 Instruction</a> </div>
1728 <div class="doc_text">
1730 <pre> ret <type> <value> <i>; Return a value from a non-void function</i>
1731 ret void <i>; Return from void function</i>
1734 <p>The '<tt>ret</tt>' instruction is used to return control flow (and a
1735 value) from a function back to the caller.</p>
1736 <p>There are two forms of the '<tt>ret</tt>' instruction: one that
1737 returns a value and then causes control flow, and one that just causes
1738 control flow to occur.</p>
1740 <p>The '<tt>ret</tt>' instruction may return any '<a
1741 href="#t_firstclass">first class</a>' type. Notice that a function is
1742 not <a href="#wellformed">well formed</a> if there exists a '<tt>ret</tt>'
1743 instruction inside of the function that returns a value that does not
1744 match the return type of the function.</p>
1746 <p>When the '<tt>ret</tt>' instruction is executed, control flow
1747 returns back to the calling function's context. If the caller is a "<a
1748 href="#i_call"><tt>call</tt></a>" instruction, execution continues at
1749 the instruction after the call. If the caller was an "<a
1750 href="#i_invoke"><tt>invoke</tt></a>" instruction, execution continues
1751 at the beginning of the "normal" destination block. If the instruction
1752 returns a value, that value shall set the call or invoke instruction's
1755 <pre> ret i32 5 <i>; Return an integer value of 5</i>
1756 ret void <i>; Return from a void function</i>
1759 <!-- _______________________________________________________________________ -->
1760 <div class="doc_subsubsection"> <a name="i_br">'<tt>br</tt>' Instruction</a> </div>
1761 <div class="doc_text">
1763 <pre> br i1 <cond>, label <iftrue>, label <iffalse><br> br label <dest> <i>; Unconditional branch</i>
1766 <p>The '<tt>br</tt>' instruction is used to cause control flow to
1767 transfer to a different basic block in the current function. There are
1768 two forms of this instruction, corresponding to a conditional branch
1769 and an unconditional branch.</p>
1771 <p>The conditional branch form of the '<tt>br</tt>' instruction takes a
1772 single '<tt>i1</tt>' value and two '<tt>label</tt>' values. The
1773 unconditional form of the '<tt>br</tt>' instruction takes a single
1774 '<tt>label</tt>' value as a target.</p>
1776 <p>Upon execution of a conditional '<tt>br</tt>' instruction, the '<tt>i1</tt>'
1777 argument is evaluated. If the value is <tt>true</tt>, control flows
1778 to the '<tt>iftrue</tt>' <tt>label</tt> argument. If "cond" is <tt>false</tt>,
1779 control flows to the '<tt>iffalse</tt>' <tt>label</tt> argument.</p>
1781 <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
1782 href="#i_ret">ret</a> i32 1<br>IfUnequal:<br> <a href="#i_ret">ret</a> i32 0<br></pre>
1784 <!-- _______________________________________________________________________ -->
1785 <div class="doc_subsubsection">
1786 <a name="i_switch">'<tt>switch</tt>' Instruction</a>
1789 <div class="doc_text">
1793 switch <intty> <value>, label <defaultdest> [ <intty> <val>, label <dest> ... ]
1798 <p>The '<tt>switch</tt>' instruction is used to transfer control flow to one of
1799 several different places. It is a generalization of the '<tt>br</tt>'
1800 instruction, allowing a branch to occur to one of many possible
1806 <p>The '<tt>switch</tt>' instruction uses three parameters: an integer
1807 comparison value '<tt>value</tt>', a default '<tt>label</tt>' destination, and
1808 an array of pairs of comparison value constants and '<tt>label</tt>'s. The
1809 table is not allowed to contain duplicate constant entries.</p>
1813 <p>The <tt>switch</tt> instruction specifies a table of values and
1814 destinations. When the '<tt>switch</tt>' instruction is executed, this
1815 table is searched for the given value. If the value is found, control flow is
1816 transfered to the corresponding destination; otherwise, control flow is
1817 transfered to the default destination.</p>
1819 <h5>Implementation:</h5>
1821 <p>Depending on properties of the target machine and the particular
1822 <tt>switch</tt> instruction, this instruction may be code generated in different
1823 ways. For example, it could be generated as a series of chained conditional
1824 branches or with a lookup table.</p>
1829 <i>; Emulate a conditional br instruction</i>
1830 %Val = <a href="#i_zext">zext</a> i1 %value to i32
1831 switch i32 %Val, label %truedest [i32 0, label %falsedest ]
1833 <i>; Emulate an unconditional br instruction</i>
1834 switch i32 0, label %dest [ ]
1836 <i>; Implement a jump table:</i>
1837 switch i32 %val, label %otherwise [ i32 0, label %onzero
1839 i32 2, label %ontwo ]
1843 <!-- _______________________________________________________________________ -->
1844 <div class="doc_subsubsection">
1845 <a name="i_invoke">'<tt>invoke</tt>' Instruction</a>
1848 <div class="doc_text">
1853 <result> = invoke [<a href="#callingconv">cconv</a>] <ptr to function ty> %<function ptr val>(<function args>)
1854 to label <normal label> unwind label <exception label>
1859 <p>The '<tt>invoke</tt>' instruction causes control to transfer to a specified
1860 function, with the possibility of control flow transfer to either the
1861 '<tt>normal</tt>' label or the
1862 '<tt>exception</tt>' label. If the callee function returns with the
1863 "<tt><a href="#i_ret">ret</a></tt>" instruction, control flow will return to the
1864 "normal" label. If the callee (or any indirect callees) returns with the "<a
1865 href="#i_unwind"><tt>unwind</tt></a>" instruction, control is interrupted and
1866 continued at the dynamically nearest "exception" label.</p>
1870 <p>This instruction requires several arguments:</p>
1874 The optional "cconv" marker indicates which <a href="#callingconv">calling
1875 convention</a> the call should use. If none is specified, the call defaults
1876 to using C calling conventions.
1878 <li>'<tt>ptr to function ty</tt>': shall be the signature of the pointer to
1879 function value being invoked. In most cases, this is a direct function
1880 invocation, but indirect <tt>invoke</tt>s are just as possible, branching off
1881 an arbitrary pointer to function value.
1884 <li>'<tt>function ptr val</tt>': An LLVM value containing a pointer to a
1885 function to be invoked. </li>
1887 <li>'<tt>function args</tt>': argument list whose types match the function
1888 signature argument types. If the function signature indicates the function
1889 accepts a variable number of arguments, the extra arguments can be
1892 <li>'<tt>normal label</tt>': the label reached when the called function
1893 executes a '<tt><a href="#i_ret">ret</a></tt>' instruction. </li>
1895 <li>'<tt>exception label</tt>': the label reached when a callee returns with
1896 the <a href="#i_unwind"><tt>unwind</tt></a> instruction. </li>
1902 <p>This instruction is designed to operate as a standard '<tt><a
1903 href="#i_call">call</a></tt>' instruction in most regards. The primary
1904 difference is that it establishes an association with a label, which is used by
1905 the runtime library to unwind the stack.</p>
1907 <p>This instruction is used in languages with destructors to ensure that proper
1908 cleanup is performed in the case of either a <tt>longjmp</tt> or a thrown
1909 exception. Additionally, this is important for implementation of
1910 '<tt>catch</tt>' clauses in high-level languages that support them.</p>
1914 %retval = invoke i32 %Test(i32 15) to label %Continue
1915 unwind label %TestCleanup <i>; {i32}:retval set</i>
1916 %retval = invoke <a href="#callingconv">coldcc</a> i32 %Test(i32 15) to label %Continue
1917 unwind label %TestCleanup <i>; {i32}:retval set</i>
1922 <!-- _______________________________________________________________________ -->
1924 <div class="doc_subsubsection"> <a name="i_unwind">'<tt>unwind</tt>'
1925 Instruction</a> </div>
1927 <div class="doc_text">
1936 <p>The '<tt>unwind</tt>' instruction unwinds the stack, continuing control flow
1937 at the first callee in the dynamic call stack which used an <a
1938 href="#i_invoke"><tt>invoke</tt></a> instruction to perform the call. This is
1939 primarily used to implement exception handling.</p>
1943 <p>The '<tt>unwind</tt>' intrinsic causes execution of the current function to
1944 immediately halt. The dynamic call stack is then searched for the first <a
1945 href="#i_invoke"><tt>invoke</tt></a> instruction on the call stack. Once found,
1946 execution continues at the "exceptional" destination block specified by the
1947 <tt>invoke</tt> instruction. If there is no <tt>invoke</tt> instruction in the
1948 dynamic call chain, undefined behavior results.</p>
1951 <!-- _______________________________________________________________________ -->
1953 <div class="doc_subsubsection"> <a name="i_unreachable">'<tt>unreachable</tt>'
1954 Instruction</a> </div>
1956 <div class="doc_text">
1965 <p>The '<tt>unreachable</tt>' instruction has no defined semantics. This
1966 instruction is used to inform the optimizer that a particular portion of the
1967 code is not reachable. This can be used to indicate that the code after a
1968 no-return function cannot be reached, and other facts.</p>
1972 <p>The '<tt>unreachable</tt>' instruction has no defined semantics.</p>
1977 <!-- ======================================================================= -->
1978 <div class="doc_subsection"> <a name="binaryops">Binary Operations</a> </div>
1979 <div class="doc_text">
1980 <p>Binary operators are used to do most of the computation in a
1981 program. They require two operands, execute an operation on them, and
1982 produce a single value. The operands might represent
1983 multiple data, as is the case with the <a href="#t_vector">vector</a> data type.
1984 The result value of a binary operator is not
1985 necessarily the same type as its operands.</p>
1986 <p>There are several different binary operators:</p>
1988 <!-- _______________________________________________________________________ -->
1989 <div class="doc_subsubsection"> <a name="i_add">'<tt>add</tt>'
1990 Instruction</a> </div>
1991 <div class="doc_text">
1993 <pre> <result> = add <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1996 <p>The '<tt>add</tt>' instruction returns the sum of its two operands.</p>
1998 <p>The two arguments to the '<tt>add</tt>' instruction must be either <a
1999 href="#t_integer">integer</a> or <a href="#t_floating">floating point</a> values.
2000 This instruction can also take <a href="#t_vector">vector</a> versions of the values.
2001 Both arguments must have identical types.</p>
2003 <p>The value produced is the integer or floating point sum of the two
2006 <pre> <result> = add i32 4, %var <i>; yields {i32}:result = 4 + %var</i>
2009 <!-- _______________________________________________________________________ -->
2010 <div class="doc_subsubsection"> <a name="i_sub">'<tt>sub</tt>'
2011 Instruction</a> </div>
2012 <div class="doc_text">
2014 <pre> <result> = sub <ty> <var1>, <var2> <i>; yields {ty}:result</i>
2017 <p>The '<tt>sub</tt>' instruction returns the difference of its two
2019 <p>Note that the '<tt>sub</tt>' instruction is used to represent the '<tt>neg</tt>'
2020 instruction present in most other intermediate representations.</p>
2022 <p>The two arguments to the '<tt>sub</tt>' instruction must be either <a
2023 href="#t_integer">integer</a> or <a href="#t_floating">floating point</a>
2025 This instruction can also take <a href="#t_vector">vector</a> versions of the values.
2026 Both arguments must have identical types.</p>
2028 <p>The value produced is the integer or floating point difference of
2029 the two operands.</p>
2032 <result> = sub i32 4, %var <i>; yields {i32}:result = 4 - %var</i>
2033 <result> = sub i32 0, %val <i>; yields {i32}:result = -%var</i>
2036 <!-- _______________________________________________________________________ -->
2037 <div class="doc_subsubsection"> <a name="i_mul">'<tt>mul</tt>'
2038 Instruction</a> </div>
2039 <div class="doc_text">
2041 <pre> <result> = mul <ty> <var1>, <var2> <i>; yields {ty}:result</i>
2044 <p>The '<tt>mul</tt>' instruction returns the product of its two
2047 <p>The two arguments to the '<tt>mul</tt>' instruction must be either <a
2048 href="#t_integer">integer</a> or <a href="#t_floating">floating point</a>
2050 This instruction can also take <a href="#t_vector">vector</a> versions of the values.
2051 Both arguments must have identical types.</p>
2053 <p>The value produced is the integer or floating point product of the
2055 <p>Because the operands are the same width, the result of an integer
2056 multiplication is the same whether the operands should be deemed unsigned or
2059 <pre> <result> = mul i32 4, %var <i>; yields {i32}:result = 4 * %var</i>
2062 <!-- _______________________________________________________________________ -->
2063 <div class="doc_subsubsection"> <a name="i_udiv">'<tt>udiv</tt>' Instruction
2065 <div class="doc_text">
2067 <pre> <result> = udiv <ty> <var1>, <var2> <i>; yields {ty}:result</i>
2070 <p>The '<tt>udiv</tt>' instruction returns the quotient of its two
2073 <p>The two arguments to the '<tt>udiv</tt>' instruction must be
2074 <a href="#t_integer">integer</a> values. Both arguments must have identical
2075 types. This instruction can also take <a href="#t_vector">vector</a> versions
2076 of the values in which case the elements must be integers.</p>
2078 <p>The value produced is the unsigned integer quotient of the two operands. This
2079 instruction always performs an unsigned division operation, regardless of
2080 whether the arguments are unsigned or not.</p>
2082 <pre> <result> = udiv i32 4, %var <i>; yields {i32}:result = 4 / %var</i>
2085 <!-- _______________________________________________________________________ -->
2086 <div class="doc_subsubsection"> <a name="i_sdiv">'<tt>sdiv</tt>' Instruction
2088 <div class="doc_text">
2090 <pre> <result> = sdiv <ty> <var1>, <var2> <i>; yields {ty}:result</i>
2093 <p>The '<tt>sdiv</tt>' instruction returns the quotient of its two
2096 <p>The two arguments to the '<tt>sdiv</tt>' instruction must be
2097 <a href="#t_integer">integer</a> values. Both arguments must have identical
2098 types. This instruction can also take <a href="#t_vector">vector</a> versions
2099 of the values in which case the elements must be integers.</p>
2101 <p>The value produced is the signed integer quotient of the two operands. This
2102 instruction always performs a signed division operation, regardless of whether
2103 the arguments are signed or not.</p>
2105 <pre> <result> = sdiv i32 4, %var <i>; yields {i32}:result = 4 / %var</i>
2108 <!-- _______________________________________________________________________ -->
2109 <div class="doc_subsubsection"> <a name="i_fdiv">'<tt>fdiv</tt>'
2110 Instruction</a> </div>
2111 <div class="doc_text">
2113 <pre> <result> = fdiv <ty> <var1>, <var2> <i>; yields {ty}:result</i>
2116 <p>The '<tt>fdiv</tt>' instruction returns the quotient of its two
2119 <p>The two arguments to the '<tt>fdiv</tt>' instruction must be
2120 <a href="#t_floating">floating point</a> values. Both arguments must have
2121 identical types. This instruction can also take <a href="#t_vector">vector</a>
2122 versions of floating point values.</p>
2124 <p>The value produced is the floating point quotient of the two operands.</p>
2126 <pre> <result> = fdiv float 4.0, %var <i>; yields {float}:result = 4.0 / %var</i>
2129 <!-- _______________________________________________________________________ -->
2130 <div class="doc_subsubsection"> <a name="i_urem">'<tt>urem</tt>' Instruction</a>
2132 <div class="doc_text">
2134 <pre> <result> = urem <ty> <var1>, <var2> <i>; yields {ty}:result</i>
2137 <p>The '<tt>urem</tt>' instruction returns the remainder from the
2138 unsigned division of its two arguments.</p>
2140 <p>The two arguments to the '<tt>urem</tt>' instruction must be
2141 <a href="#t_integer">integer</a> values. Both arguments must have identical
2142 types. This instruction can also take <a href="#t_vector">vector</a> versions
2143 of the values in which case the elements must be integers.</p>
2145 <p>This instruction returns the unsigned integer <i>remainder</i> of a division.
2146 This instruction always performs an unsigned division to get the remainder,
2147 regardless of whether the arguments are unsigned or not.</p>
2149 <pre> <result> = urem i32 4, %var <i>; yields {i32}:result = 4 % %var</i>
2153 <!-- _______________________________________________________________________ -->
2154 <div class="doc_subsubsection"> <a name="i_srem">'<tt>srem</tt>'
2155 Instruction</a> </div>
2156 <div class="doc_text">
2158 <pre> <result> = srem <ty> <var1>, <var2> <i>; yields {ty}:result</i>
2161 <p>The '<tt>srem</tt>' instruction returns the remainder from the
2162 signed division of its two operands. This instruction can also take
2163 <a href="#t_vector">vector</a> versions of the values in which case
2164 the elements must be integers.</p>
2167 <p>The two arguments to the '<tt>srem</tt>' instruction must be
2168 <a href="#t_integer">integer</a> values. Both arguments must have identical
2171 <p>This instruction returns the <i>remainder</i> of a division (where the result
2172 has the same sign as the dividend, <tt>var1</tt>), not the <i>modulo</i>
2173 operator (where the result has the same sign as the divisor, <tt>var2</tt>) of
2174 a value. For more information about the difference, see <a
2175 href="http://mathforum.org/dr.math/problems/anne.4.28.99.html">The
2176 Math Forum</a>. For a table of how this is implemented in various languages,
2177 please see <a href="http://en.wikipedia.org/wiki/Modulo_operation">
2178 Wikipedia: modulo operation</a>.</p>
2180 <pre> <result> = srem i32 4, %var <i>; yields {i32}:result = 4 % %var</i>
2184 <!-- _______________________________________________________________________ -->
2185 <div class="doc_subsubsection"> <a name="i_frem">'<tt>frem</tt>'
2186 Instruction</a> </div>
2187 <div class="doc_text">
2189 <pre> <result> = frem <ty> <var1>, <var2> <i>; yields {ty}:result</i>
2192 <p>The '<tt>frem</tt>' instruction returns the remainder from the
2193 division of its two operands.</p>
2195 <p>The two arguments to the '<tt>frem</tt>' instruction must be
2196 <a href="#t_floating">floating point</a> values. Both arguments must have
2197 identical types. This instruction can also take <a href="#t_vector">vector</a>
2198 versions of floating point values.</p>
2200 <p>This instruction returns the <i>remainder</i> of a division.</p>
2202 <pre> <result> = frem float 4.0, %var <i>; yields {float}:result = 4.0 % %var</i>
2206 <!-- ======================================================================= -->
2207 <div class="doc_subsection"> <a name="bitwiseops">Bitwise Binary
2208 Operations</a> </div>
2209 <div class="doc_text">
2210 <p>Bitwise binary operators are used to do various forms of
2211 bit-twiddling in a program. They are generally very efficient
2212 instructions and can commonly be strength reduced from other
2213 instructions. They require two operands, execute an operation on them,
2214 and produce a single value. The resulting value of the bitwise binary
2215 operators is always the same type as its first operand.</p>
2218 <!-- _______________________________________________________________________ -->
2219 <div class="doc_subsubsection"> <a name="i_shl">'<tt>shl</tt>'
2220 Instruction</a> </div>
2221 <div class="doc_text">
2223 <pre> <result> = shl <ty> <var1>, <var2> <i>; yields {ty}:result</i>
2228 <p>The '<tt>shl</tt>' instruction returns the first operand shifted to
2229 the left a specified number of bits.</p>
2233 <p>Both arguments to the '<tt>shl</tt>' instruction must be the same <a
2234 href="#t_integer">integer</a> type.</p>
2238 <p>The value produced is <tt>var1</tt> * 2<sup><tt>var2</tt></sup>. If
2239 <tt>var2</tt> is (statically or dynamically) equal to or larger than the number
2240 of bits in <tt>var1</tt>, the result is undefined.</p>
2242 <h5>Example:</h5><pre>
2243 <result> = shl i32 4, %var <i>; yields {i32}: 4 << %var</i>
2244 <result> = shl i32 4, 2 <i>; yields {i32}: 16</i>
2245 <result> = shl i32 1, 10 <i>; yields {i32}: 1024</i>
2246 <result> = shl i32 1, 32 <i>; undefined</i>
2249 <!-- _______________________________________________________________________ -->
2250 <div class="doc_subsubsection"> <a name="i_lshr">'<tt>lshr</tt>'
2251 Instruction</a> </div>
2252 <div class="doc_text">
2254 <pre> <result> = lshr <ty> <var1>, <var2> <i>; yields {ty}:result</i>
2258 <p>The '<tt>lshr</tt>' instruction (logical shift right) returns the first
2259 operand shifted to the right a specified number of bits with zero fill.</p>
2262 <p>Both arguments to the '<tt>lshr</tt>' instruction must be the same
2263 <a href="#t_integer">integer</a> type.</p>
2267 <p>This instruction always performs a logical shift right operation. The most
2268 significant bits of the result will be filled with zero bits after the
2269 shift. If <tt>var2</tt> is (statically or dynamically) equal to or larger than
2270 the number of bits in <tt>var1</tt>, the result is undefined.</p>
2274 <result> = lshr i32 4, 1 <i>; yields {i32}:result = 2</i>
2275 <result> = lshr i32 4, 2 <i>; yields {i32}:result = 1</i>
2276 <result> = lshr i8 4, 3 <i>; yields {i8}:result = 0</i>
2277 <result> = lshr i8 -2, 1 <i>; yields {i8}:result = 0x7FFFFFFF </i>
2278 <result> = lshr i32 1, 32 <i>; undefined</i>
2282 <!-- _______________________________________________________________________ -->
2283 <div class="doc_subsubsection"> <a name="i_ashr">'<tt>ashr</tt>'
2284 Instruction</a> </div>
2285 <div class="doc_text">
2288 <pre> <result> = ashr <ty> <var1>, <var2> <i>; yields {ty}:result</i>
2292 <p>The '<tt>ashr</tt>' instruction (arithmetic shift right) returns the first
2293 operand shifted to the right a specified number of bits with sign extension.</p>
2296 <p>Both arguments to the '<tt>ashr</tt>' instruction must be the same
2297 <a href="#t_integer">integer</a> type.</p>
2300 <p>This instruction always performs an arithmetic shift right operation,
2301 The most significant bits of the result will be filled with the sign bit
2302 of <tt>var1</tt>. If <tt>var2</tt> is (statically or dynamically) equal to or
2303 larger than the number of bits in <tt>var1</tt>, the result is undefined.
2308 <result> = ashr i32 4, 1 <i>; yields {i32}:result = 2</i>
2309 <result> = ashr i32 4, 2 <i>; yields {i32}:result = 1</i>
2310 <result> = ashr i8 4, 3 <i>; yields {i8}:result = 0</i>
2311 <result> = ashr i8 -2, 1 <i>; yields {i8}:result = -1</i>
2312 <result> = ashr i32 1, 32 <i>; undefined</i>
2316 <!-- _______________________________________________________________________ -->
2317 <div class="doc_subsubsection"> <a name="i_and">'<tt>and</tt>'
2318 Instruction</a> </div>
2319 <div class="doc_text">
2321 <pre> <result> = and <ty> <var1>, <var2> <i>; yields {ty}:result</i>
2324 <p>The '<tt>and</tt>' instruction returns the bitwise logical and of
2325 its two operands.</p>
2327 <p>The two arguments to the '<tt>and</tt>' instruction must be <a
2328 href="#t_integer">integer</a> values. Both arguments must have
2329 identical types.</p>
2331 <p>The truth table used for the '<tt>and</tt>' instruction is:</p>
2333 <div style="align: center">
2334 <table border="1" cellspacing="0" cellpadding="4">
2365 <pre> <result> = and i32 4, %var <i>; yields {i32}:result = 4 & %var</i>
2366 <result> = and i32 15, 40 <i>; yields {i32}:result = 8</i>
2367 <result> = and i32 4, 8 <i>; yields {i32}:result = 0</i>
2370 <!-- _______________________________________________________________________ -->
2371 <div class="doc_subsubsection"> <a name="i_or">'<tt>or</tt>' Instruction</a> </div>
2372 <div class="doc_text">
2374 <pre> <result> = or <ty> <var1>, <var2> <i>; yields {ty}:result</i>
2377 <p>The '<tt>or</tt>' instruction returns the bitwise logical inclusive
2378 or of its two operands.</p>
2380 <p>The two arguments to the '<tt>or</tt>' instruction must be <a
2381 href="#t_integer">integer</a> values. Both arguments must have
2382 identical types.</p>
2384 <p>The truth table used for the '<tt>or</tt>' instruction is:</p>
2386 <div style="align: center">
2387 <table border="1" cellspacing="0" cellpadding="4">
2418 <pre> <result> = or i32 4, %var <i>; yields {i32}:result = 4 | %var</i>
2419 <result> = or i32 15, 40 <i>; yields {i32}:result = 47</i>
2420 <result> = or i32 4, 8 <i>; yields {i32}:result = 12</i>
2423 <!-- _______________________________________________________________________ -->
2424 <div class="doc_subsubsection"> <a name="i_xor">'<tt>xor</tt>'
2425 Instruction</a> </div>
2426 <div class="doc_text">
2428 <pre> <result> = xor <ty> <var1>, <var2> <i>; yields {ty}:result</i>
2431 <p>The '<tt>xor</tt>' instruction returns the bitwise logical exclusive
2432 or of its two operands. The <tt>xor</tt> is used to implement the
2433 "one's complement" operation, which is the "~" operator in C.</p>
2435 <p>The two arguments to the '<tt>xor</tt>' instruction must be <a
2436 href="#t_integer">integer</a> values. Both arguments must have
2437 identical types.</p>
2439 <p>The truth table used for the '<tt>xor</tt>' instruction is:</p>
2441 <div style="align: center">
2442 <table border="1" cellspacing="0" cellpadding="4">
2474 <pre> <result> = xor i32 4, %var <i>; yields {i32}:result = 4 ^ %var</i>
2475 <result> = xor i32 15, 40 <i>; yields {i32}:result = 39</i>
2476 <result> = xor i32 4, 8 <i>; yields {i32}:result = 12</i>
2477 <result> = xor i32 %V, -1 <i>; yields {i32}:result = ~%V</i>
2481 <!-- ======================================================================= -->
2482 <div class="doc_subsection">
2483 <a name="vectorops">Vector Operations</a>
2486 <div class="doc_text">
2488 <p>LLVM supports several instructions to represent vector operations in a
2489 target-independent manner. These instructions cover the element-access and
2490 vector-specific operations needed to process vectors effectively. While LLVM
2491 does directly support these vector operations, many sophisticated algorithms
2492 will want to use target-specific intrinsics to take full advantage of a specific
2497 <!-- _______________________________________________________________________ -->
2498 <div class="doc_subsubsection">
2499 <a name="i_extractelement">'<tt>extractelement</tt>' Instruction</a>
2502 <div class="doc_text">
2507 <result> = extractelement <n x <ty>> <val>, i32 <idx> <i>; yields <ty></i>
2513 The '<tt>extractelement</tt>' instruction extracts a single scalar
2514 element from a vector at a specified index.
2521 The first operand of an '<tt>extractelement</tt>' instruction is a
2522 value of <a href="#t_vector">vector</a> type. The second operand is
2523 an index indicating the position from which to extract the element.
2524 The index may be a variable.</p>
2529 The result is a scalar of the same type as the element type of
2530 <tt>val</tt>. Its value is the value at position <tt>idx</tt> of
2531 <tt>val</tt>. If <tt>idx</tt> exceeds the length of <tt>val</tt>, the
2532 results are undefined.
2538 %result = extractelement <4 x i32> %vec, i32 0 <i>; yields i32</i>
2543 <!-- _______________________________________________________________________ -->
2544 <div class="doc_subsubsection">
2545 <a name="i_insertelement">'<tt>insertelement</tt>' Instruction</a>
2548 <div class="doc_text">
2553 <result> = insertelement <n x <ty>> <val>, <ty> <elt>, i32 <idx> <i>; yields <n x <ty>></i>
2559 The '<tt>insertelement</tt>' instruction inserts a scalar
2560 element into a vector at a specified index.
2567 The first operand of an '<tt>insertelement</tt>' instruction is a
2568 value of <a href="#t_vector">vector</a> type. The second operand is a
2569 scalar value whose type must equal the element type of the first
2570 operand. The third operand is an index indicating the position at
2571 which to insert the value. The index may be a variable.</p>
2576 The result is a vector of the same type as <tt>val</tt>. Its
2577 element values are those of <tt>val</tt> except at position
2578 <tt>idx</tt>, where it gets the value <tt>elt</tt>. If <tt>idx</tt>
2579 exceeds the length of <tt>val</tt>, the results are undefined.
2585 %result = insertelement <4 x i32> %vec, i32 1, i32 0 <i>; yields <4 x i32></i>
2589 <!-- _______________________________________________________________________ -->
2590 <div class="doc_subsubsection">
2591 <a name="i_shufflevector">'<tt>shufflevector</tt>' Instruction</a>
2594 <div class="doc_text">
2599 <result> = shufflevector <n x <ty>> <v1>, <n x <ty>> <v2>, <n x i32> <mask> <i>; yields <n x <ty>></i>
2605 The '<tt>shufflevector</tt>' instruction constructs a permutation of elements
2606 from two input vectors, returning a vector of the same type.
2612 The first two operands of a '<tt>shufflevector</tt>' instruction are vectors
2613 with types that match each other and types that match the result of the
2614 instruction. The third argument is a shuffle mask, which has the same number
2615 of elements as the other vector type, but whose element type is always 'i32'.
2619 The shuffle mask operand is required to be a constant vector with either
2620 constant integer or undef values.
2626 The elements of the two input vectors are numbered from left to right across
2627 both of the vectors. The shuffle mask operand specifies, for each element of
2628 the result vector, which element of the two input registers the result element
2629 gets. The element selector may be undef (meaning "don't care") and the second
2630 operand may be undef if performing a shuffle from only one vector.
2636 %result = shufflevector <4 x i32> %v1, <4 x i32> %v2,
2637 <4 x i32> <i32 0, i32 4, i32 1, i32 5> <i>; yields <4 x i32></i>
2638 %result = shufflevector <4 x i32> %v1, <4 x i32> undef,
2639 <4 x i32> <i32 0, i32 1, i32 2, i32 3> <i>; yields <4 x i32></i> - Identity shuffle.
2644 <!-- ======================================================================= -->
2645 <div class="doc_subsection">
2646 <a name="memoryops">Memory Access and Addressing Operations</a>
2649 <div class="doc_text">
2651 <p>A key design point of an SSA-based representation is how it
2652 represents memory. In LLVM, no memory locations are in SSA form, which
2653 makes things very simple. This section describes how to read, write,
2654 allocate, and free memory in LLVM.</p>
2658 <!-- _______________________________________________________________________ -->
2659 <div class="doc_subsubsection">
2660 <a name="i_malloc">'<tt>malloc</tt>' Instruction</a>
2663 <div class="doc_text">
2668 <result> = malloc <type>[, i32 <NumElements>][, align <alignment>] <i>; yields {type*}:result</i>
2673 <p>The '<tt>malloc</tt>' instruction allocates memory from the system
2674 heap and returns a pointer to it.</p>
2678 <p>The '<tt>malloc</tt>' instruction allocates
2679 <tt>sizeof(<type>)*NumElements</tt>
2680 bytes of memory from the operating system and returns a pointer of the
2681 appropriate type to the program. If "NumElements" is specified, it is the
2682 number of elements allocated. If an alignment is specified, the value result
2683 of the allocation is guaranteed to be aligned to at least that boundary. If
2684 not specified, or if zero, the target can choose to align the allocation on any
2685 convenient boundary.</p>
2687 <p>'<tt>type</tt>' must be a sized type.</p>
2691 <p>Memory is allocated using the system "<tt>malloc</tt>" function, and
2692 a pointer is returned.</p>
2697 %array = malloc [4 x i8 ] <i>; yields {[%4 x i8]*}:array</i>
2699 %size = <a href="#i_add">add</a> i32 2, 2 <i>; yields {i32}:size = i32 4</i>
2700 %array1 = malloc i8, i32 4 <i>; yields {i8*}:array1</i>
2701 %array2 = malloc [12 x i8], i32 %size <i>; yields {[12 x i8]*}:array2</i>
2702 %array3 = malloc i32, i32 4, align 1024 <i>; yields {i32*}:array3</i>
2703 %array4 = malloc i32, align 1024 <i>; yields {i32*}:array4</i>
2707 <!-- _______________________________________________________________________ -->
2708 <div class="doc_subsubsection">
2709 <a name="i_free">'<tt>free</tt>' Instruction</a>
2712 <div class="doc_text">
2717 free <type> <value> <i>; yields {void}</i>
2722 <p>The '<tt>free</tt>' instruction returns memory back to the unused
2723 memory heap to be reallocated in the future.</p>
2727 <p>'<tt>value</tt>' shall be a pointer value that points to a value
2728 that was allocated with the '<tt><a href="#i_malloc">malloc</a></tt>'
2733 <p>Access to the memory pointed to by the pointer is no longer defined
2734 after this instruction executes.</p>
2739 %array = <a href="#i_malloc">malloc</a> [4 x i8] <i>; yields {[4 x i8]*}:array</i>
2740 free [4 x i8]* %array
2744 <!-- _______________________________________________________________________ -->
2745 <div class="doc_subsubsection">
2746 <a name="i_alloca">'<tt>alloca</tt>' Instruction</a>
2749 <div class="doc_text">
2754 <result> = alloca <type>[, i32 <NumElements>][, align <alignment>] <i>; yields {type*}:result</i>
2759 <p>The '<tt>alloca</tt>' instruction allocates memory on the stack frame of the
2760 currently executing function, to be automatically released when this function
2761 returns to its caller.</p>
2765 <p>The '<tt>alloca</tt>' instruction allocates <tt>sizeof(<type>)*NumElements</tt>
2766 bytes of memory on the runtime stack, returning a pointer of the
2767 appropriate type to the program. If "NumElements" is specified, it is the
2768 number of elements allocated. If an alignment is specified, the value result
2769 of the allocation is guaranteed to be aligned to at least that boundary. If
2770 not specified, or if zero, the target can choose to align the allocation on any
2771 convenient boundary.</p>
2773 <p>'<tt>type</tt>' may be any sized type.</p>
2777 <p>Memory is allocated; a pointer is returned. '<tt>alloca</tt>'d
2778 memory is automatically released when the function returns. The '<tt>alloca</tt>'
2779 instruction is commonly used to represent automatic variables that must
2780 have an address available. When the function returns (either with the <tt><a
2781 href="#i_ret">ret</a></tt> or <tt><a href="#i_unwind">unwind</a></tt>
2782 instructions), the memory is reclaimed.</p>
2787 %ptr = alloca i32 <i>; yields {i32*}:ptr</i>
2788 %ptr = alloca i32, i32 4 <i>; yields {i32*}:ptr</i>
2789 %ptr = alloca i32, i32 4, align 1024 <i>; yields {i32*}:ptr</i>
2790 %ptr = alloca i32, align 1024 <i>; yields {i32*}:ptr</i>
2794 <!-- _______________________________________________________________________ -->
2795 <div class="doc_subsubsection"> <a name="i_load">'<tt>load</tt>'
2796 Instruction</a> </div>
2797 <div class="doc_text">
2799 <pre> <result> = load <ty>* <pointer>[, align <alignment>]<br> <result> = volatile load <ty>* <pointer>[, align <alignment>]<br></pre>
2801 <p>The '<tt>load</tt>' instruction is used to read from memory.</p>
2803 <p>The argument to the '<tt>load</tt>' instruction specifies the memory
2804 address from which to load. The pointer must point to a <a
2805 href="#t_firstclass">first class</a> type. If the <tt>load</tt> is
2806 marked as <tt>volatile</tt>, then the optimizer is not allowed to modify
2807 the number or order of execution of this <tt>load</tt> with other
2808 volatile <tt>load</tt> and <tt><a href="#i_store">store</a></tt>
2811 <p>The location of memory pointed to is loaded.</p>
2813 <pre> %ptr = <a href="#i_alloca">alloca</a> i32 <i>; yields {i32*}:ptr</i>
2815 href="#i_store">store</a> i32 3, i32* %ptr <i>; yields {void}</i>
2816 %val = load i32* %ptr <i>; yields {i32}:val = i32 3</i>
2819 <!-- _______________________________________________________________________ -->
2820 <div class="doc_subsubsection"> <a name="i_store">'<tt>store</tt>'
2821 Instruction</a> </div>
2822 <div class="doc_text">
2824 <pre> store <ty> <value>, <ty>* <pointer>[, align <alignment>] <i>; yields {void}</i>
2825 volatile store <ty> <value>, <ty>* <pointer>[, align <alignment>] <i>; yields {void}</i>
2828 <p>The '<tt>store</tt>' instruction is used to write to memory.</p>
2830 <p>There are two arguments to the '<tt>store</tt>' instruction: a value
2831 to store and an address at which to store it. The type of the '<tt><pointer></tt>'
2832 operand must be a pointer to the type of the '<tt><value></tt>'
2833 operand. If the <tt>store</tt> is marked as <tt>volatile</tt>, then the
2834 optimizer is not allowed to modify the number or order of execution of
2835 this <tt>store</tt> with other volatile <tt>load</tt> and <tt><a
2836 href="#i_store">store</a></tt> instructions.</p>
2838 <p>The contents of memory are updated to contain '<tt><value></tt>'
2839 at the location specified by the '<tt><pointer></tt>' operand.</p>
2841 <pre> %ptr = <a href="#i_alloca">alloca</a> i32 <i>; yields {i32*}:ptr</i>
2842 store i32 3, i32* %ptr <i>; yields {void}</i>
2843 %val = <a href="#i_load">load</a> i32* %ptr <i>; yields {i32}:val = i32 3</i>
2847 <!-- _______________________________________________________________________ -->
2848 <div class="doc_subsubsection">
2849 <a name="i_getelementptr">'<tt>getelementptr</tt>' Instruction</a>
2852 <div class="doc_text">
2855 <result> = getelementptr <ty>* <ptrval>{, <ty> <idx>}*
2861 The '<tt>getelementptr</tt>' instruction is used to get the address of a
2862 subelement of an aggregate data structure.</p>
2866 <p>This instruction takes a list of integer operands that indicate what
2867 elements of the aggregate object to index to. The actual types of the arguments
2868 provided depend on the type of the first pointer argument. The
2869 '<tt>getelementptr</tt>' instruction is used to index down through the type
2870 levels of a structure or to a specific index in an array. When indexing into a
2871 structure, only <tt>i32</tt> integer constants are allowed. When indexing
2872 into an array or pointer, only integers of 32 or 64 bits are allowed, and will
2873 be sign extended to 64-bit values.</p>
2875 <p>For example, let's consider a C code fragment and how it gets
2876 compiled to LLVM:</p>
2878 <div class="doc_code">
2891 int *foo(struct ST *s) {
2892 return &s[1].Z.B[5][13];
2897 <p>The LLVM code generated by the GCC frontend is:</p>
2899 <div class="doc_code">
2901 %RT = type { i8 , [10 x [20 x i32]], i8 }
2902 %ST = type { i32, double, %RT }
2904 define i32* %foo(%ST* %s) {
2906 %reg = getelementptr %ST* %s, i32 1, i32 2, i32 1, i32 5, i32 13
2914 <p>The index types specified for the '<tt>getelementptr</tt>' instruction depend
2915 on the pointer type that is being indexed into. <a href="#t_pointer">Pointer</a>
2916 and <a href="#t_array">array</a> types can use a 32-bit or 64-bit
2917 <a href="#t_integer">integer</a> type but the value will always be sign extended
2918 to 64-bits. <a href="#t_struct">Structure</a> types require <tt>i32</tt>
2919 <b>constants</b>.</p>
2921 <p>In the example above, the first index is indexing into the '<tt>%ST*</tt>'
2922 type, which is a pointer, yielding a '<tt>%ST</tt>' = '<tt>{ i32, double, %RT
2923 }</tt>' type, a structure. The second index indexes into the third element of
2924 the structure, yielding a '<tt>%RT</tt>' = '<tt>{ i8 , [10 x [20 x i32]],
2925 i8 }</tt>' type, another structure. The third index indexes into the second
2926 element of the structure, yielding a '<tt>[10 x [20 x i32]]</tt>' type, an
2927 array. The two dimensions of the array are subscripted into, yielding an
2928 '<tt>i32</tt>' type. The '<tt>getelementptr</tt>' instruction returns a pointer
2929 to this element, thus computing a value of '<tt>i32*</tt>' type.</p>
2931 <p>Note that it is perfectly legal to index partially through a
2932 structure, returning a pointer to an inner element. Because of this,
2933 the LLVM code for the given testcase is equivalent to:</p>
2936 define i32* %foo(%ST* %s) {
2937 %t1 = getelementptr %ST* %s, i32 1 <i>; yields %ST*:%t1</i>
2938 %t2 = getelementptr %ST* %t1, i32 0, i32 2 <i>; yields %RT*:%t2</i>
2939 %t3 = getelementptr %RT* %t2, i32 0, i32 1 <i>; yields [10 x [20 x i32]]*:%t3</i>
2940 %t4 = getelementptr [10 x [20 x i32]]* %t3, i32 0, i32 5 <i>; yields [20 x i32]*:%t4</i>
2941 %t5 = getelementptr [20 x i32]* %t4, i32 0, i32 13 <i>; yields i32*:%t5</i>
2946 <p>Note that it is undefined to access an array out of bounds: array and
2947 pointer indexes must always be within the defined bounds of the array type.
2948 The one exception for this rules is zero length arrays. These arrays are
2949 defined to be accessible as variable length arrays, which requires access
2950 beyond the zero'th element.</p>
2952 <p>The getelementptr instruction is often confusing. For some more insight
2953 into how it works, see <a href="GetElementPtr.html">the getelementptr
2959 <i>; yields [12 x i8]*:aptr</i>
2960 %aptr = getelementptr {i32, [12 x i8]}* %sptr, i64 0, i32 1
2964 <!-- ======================================================================= -->
2965 <div class="doc_subsection"> <a name="convertops">Conversion Operations</a>
2967 <div class="doc_text">
2968 <p>The instructions in this category are the conversion instructions (casting)
2969 which all take a single operand and a type. They perform various bit conversions
2973 <!-- _______________________________________________________________________ -->
2974 <div class="doc_subsubsection">
2975 <a name="i_trunc">'<tt>trunc .. to</tt>' Instruction</a>
2977 <div class="doc_text">
2981 <result> = trunc <ty> <value> to <ty2> <i>; yields ty2</i>
2986 The '<tt>trunc</tt>' instruction truncates its operand to the type <tt>ty2</tt>.
2991 The '<tt>trunc</tt>' instruction takes a <tt>value</tt> to trunc, which must
2992 be an <a href="#t_integer">integer</a> type, and a type that specifies the size
2993 and type of the result, which must be an <a href="#t_integer">integer</a>
2994 type. The bit size of <tt>value</tt> must be larger than the bit size of
2995 <tt>ty2</tt>. Equal sized types are not allowed.</p>
2999 The '<tt>trunc</tt>' instruction truncates the high order bits in <tt>value</tt>
3000 and converts the remaining bits to <tt>ty2</tt>. Since the source size must be
3001 larger than the destination size, <tt>trunc</tt> cannot be a <i>no-op cast</i>.
3002 It will always truncate bits.</p>
3006 %X = trunc i32 257 to i8 <i>; yields i8:1</i>
3007 %Y = trunc i32 123 to i1 <i>; yields i1:true</i>
3008 %Y = trunc i32 122 to i1 <i>; yields i1:false</i>
3012 <!-- _______________________________________________________________________ -->
3013 <div class="doc_subsubsection">
3014 <a name="i_zext">'<tt>zext .. to</tt>' Instruction</a>
3016 <div class="doc_text">
3020 <result> = zext <ty> <value> to <ty2> <i>; yields ty2</i>
3024 <p>The '<tt>zext</tt>' instruction zero extends its operand to type
3029 <p>The '<tt>zext</tt>' instruction takes a value to cast, which must be of
3030 <a href="#t_integer">integer</a> type, and a type to cast it to, which must
3031 also be of <a href="#t_integer">integer</a> type. The bit size of the
3032 <tt>value</tt> must be smaller than the bit size of the destination type,
3036 <p>The <tt>zext</tt> fills the high order bits of the <tt>value</tt> with zero
3037 bits until it reaches the size of the destination type, <tt>ty2</tt>.</p>
3039 <p>When zero extending from i1, the result will always be either 0 or 1.</p>
3043 %X = zext i32 257 to i64 <i>; yields i64:257</i>
3044 %Y = zext i1 true to i32 <i>; yields i32:1</i>
3048 <!-- _______________________________________________________________________ -->
3049 <div class="doc_subsubsection">
3050 <a name="i_sext">'<tt>sext .. to</tt>' Instruction</a>
3052 <div class="doc_text">
3056 <result> = sext <ty> <value> to <ty2> <i>; yields ty2</i>
3060 <p>The '<tt>sext</tt>' sign extends <tt>value</tt> to the type <tt>ty2</tt>.</p>
3064 The '<tt>sext</tt>' instruction takes a value to cast, which must be of
3065 <a href="#t_integer">integer</a> type, and a type to cast it to, which must
3066 also be of <a href="#t_integer">integer</a> type. The bit size of the
3067 <tt>value</tt> must be smaller than the bit size of the destination type,
3072 The '<tt>sext</tt>' instruction performs a sign extension by copying the sign
3073 bit (highest order bit) of the <tt>value</tt> until it reaches the bit size of
3074 the type <tt>ty2</tt>.</p>
3076 <p>When sign extending from i1, the extension always results in -1 or 0.</p>
3080 %X = sext i8 -1 to i16 <i>; yields i16 :65535</i>
3081 %Y = sext i1 true to i32 <i>; yields i32:-1</i>
3085 <!-- _______________________________________________________________________ -->
3086 <div class="doc_subsubsection">
3087 <a name="i_fptrunc">'<tt>fptrunc .. to</tt>' Instruction</a>
3090 <div class="doc_text">
3095 <result> = fptrunc <ty> <value> to <ty2> <i>; yields ty2</i>
3099 <p>The '<tt>fptrunc</tt>' instruction truncates <tt>value</tt> to type
3104 <p>The '<tt>fptrunc</tt>' instruction takes a <a href="#t_floating">floating
3105 point</a> value to cast and a <a href="#t_floating">floating point</a> type to
3106 cast it to. The size of <tt>value</tt> must be larger than the size of
3107 <tt>ty2</tt>. This implies that <tt>fptrunc</tt> cannot be used to make a
3108 <i>no-op cast</i>.</p>
3111 <p> The '<tt>fptrunc</tt>' instruction truncates a <tt>value</tt> from a larger
3112 <a href="#t_floating">floating point</a> type to a smaller
3113 <a href="#t_floating">floating point</a> type. If the value cannot fit within
3114 the destination type, <tt>ty2</tt>, then the results are undefined.</p>
3118 %X = fptrunc double 123.0 to float <i>; yields float:123.0</i>
3119 %Y = fptrunc double 1.0E+300 to float <i>; yields undefined</i>
3123 <!-- _______________________________________________________________________ -->
3124 <div class="doc_subsubsection">
3125 <a name="i_fpext">'<tt>fpext .. to</tt>' Instruction</a>
3127 <div class="doc_text">
3131 <result> = fpext <ty> <value> to <ty2> <i>; yields ty2</i>
3135 <p>The '<tt>fpext</tt>' extends a floating point <tt>value</tt> to a larger
3136 floating point value.</p>
3139 <p>The '<tt>fpext</tt>' instruction takes a
3140 <a href="#t_floating">floating point</a> <tt>value</tt> to cast,
3141 and a <a href="#t_floating">floating point</a> type to cast it to. The source
3142 type must be smaller than the destination type.</p>
3145 <p>The '<tt>fpext</tt>' instruction extends the <tt>value</tt> from a smaller
3146 <a href="#t_floating">floating point</a> type to a larger
3147 <a href="#t_floating">floating point</a> type. The <tt>fpext</tt> cannot be
3148 used to make a <i>no-op cast</i> because it always changes bits. Use
3149 <tt>bitcast</tt> to make a <i>no-op cast</i> for a floating point cast.</p>
3153 %X = fpext float 3.1415 to double <i>; yields double:3.1415</i>
3154 %Y = fpext float 1.0 to float <i>; yields float:1.0 (no-op)</i>
3158 <!-- _______________________________________________________________________ -->
3159 <div class="doc_subsubsection">
3160 <a name="i_fptoui">'<tt>fptoui .. to</tt>' Instruction</a>
3162 <div class="doc_text">
3166 <result> = fptoui <ty> <value> to <ty2> <i>; yields ty2</i>
3170 <p>The '<tt>fptoui</tt>' converts a floating point <tt>value</tt> to its
3171 unsigned integer equivalent of type <tt>ty2</tt>.
3175 <p>The '<tt>fptoui</tt>' instruction takes a value to cast, which must be a
3176 scalar or vector <a href="#t_floating">floating point</a> value, and a type
3177 to cast it to <tt>ty2</tt>, which must be an <a href="#t_integer">integer</a>
3178 type. If <tt>ty</tt> is a vector floating point type, <tt>ty2</tt> must be a
3179 vector integer type with the same number of elements as <tt>ty</tt></p>
3182 <p> The '<tt>fptoui</tt>' instruction converts its
3183 <a href="#t_floating">floating point</a> operand into the nearest (rounding
3184 towards zero) unsigned integer value. If the value cannot fit in <tt>ty2</tt>,
3185 the results are undefined.</p>
3189 %X = fptoui double 123.0 to i32 <i>; yields i32:123</i>
3190 %Y = fptoui float 1.0E+300 to i1 <i>; yields undefined:1</i>
3191 %X = fptoui float 1.04E+17 to i8 <i>; yields undefined:1</i>
3195 <!-- _______________________________________________________________________ -->
3196 <div class="doc_subsubsection">
3197 <a name="i_fptosi">'<tt>fptosi .. to</tt>' Instruction</a>
3199 <div class="doc_text">
3203 <result> = fptosi <ty> <value> to <ty2> <i>; yields ty2</i>
3207 <p>The '<tt>fptosi</tt>' instruction converts
3208 <a href="#t_floating">floating point</a> <tt>value</tt> to type <tt>ty2</tt>.
3212 <p> The '<tt>fptosi</tt>' instruction takes a value to cast, which must be a
3213 scalar or vector <a href="#t_floating">floating point</a> value, and a type
3214 to cast it to <tt>ty2</tt>, which must be an <a href="#t_integer">integer</a>
3215 type. If <tt>ty</tt> is a vector floating point type, <tt>ty2</tt> must be a
3216 vector integer type with the same number of elements as <tt>ty</tt></p>
3219 <p>The '<tt>fptosi</tt>' instruction converts its
3220 <a href="#t_floating">floating point</a> operand into the nearest (rounding
3221 towards zero) signed integer value. If the value cannot fit in <tt>ty2</tt>,
3222 the results are undefined.</p>
3226 %X = fptosi double -123.0 to i32 <i>; yields i32:-123</i>
3227 %Y = fptosi float 1.0E-247 to i1 <i>; yields undefined:1</i>
3228 %X = fptosi float 1.04E+17 to i8 <i>; yields undefined:1</i>
3232 <!-- _______________________________________________________________________ -->
3233 <div class="doc_subsubsection">
3234 <a name="i_uitofp">'<tt>uitofp .. to</tt>' Instruction</a>
3236 <div class="doc_text">
3240 <result> = uitofp <ty> <value> to <ty2> <i>; yields ty2</i>
3244 <p>The '<tt>uitofp</tt>' instruction regards <tt>value</tt> as an unsigned
3245 integer and converts that value to the <tt>ty2</tt> type.</p>
3248 <p>The '<tt>uitofp</tt>' instruction takes a value to cast, which must be a
3249 scalar or vector <a href="#t_integer">integer</a> value, and a type to cast it
3250 to <tt>ty2</tt>, which must be an <a href="#t_floating">floating point</a>
3251 type. If <tt>ty</tt> is a vector integer type, <tt>ty2</tt> must be a vector
3252 floating point type with the same number of elements as <tt>ty</tt></p>
3255 <p>The '<tt>uitofp</tt>' instruction interprets its operand as an unsigned
3256 integer quantity and converts it to the corresponding floating point value. If
3257 the value cannot fit in the floating point value, the results are undefined.</p>
3261 %X = uitofp i32 257 to float <i>; yields float:257.0</i>
3262 %Y = uitofp i8 -1 to double <i>; yields double:255.0</i>
3266 <!-- _______________________________________________________________________ -->
3267 <div class="doc_subsubsection">
3268 <a name="i_sitofp">'<tt>sitofp .. to</tt>' Instruction</a>
3270 <div class="doc_text">
3274 <result> = sitofp <ty> <value> to <ty2> <i>; yields ty2</i>
3278 <p>The '<tt>sitofp</tt>' instruction regards <tt>value</tt> as a signed
3279 integer and converts that value to the <tt>ty2</tt> type.</p>
3282 <p>The '<tt>sitofp</tt>' instruction takes a value to cast, which must be a
3283 scalar or vector <a href="#t_integer">integer</a> value, and a type to cast it
3284 to <tt>ty2</tt>, which must be an <a href="#t_floating">floating point</a>
3285 type. If <tt>ty</tt> is a vector integer type, <tt>ty2</tt> must be a vector
3286 floating point type with the same number of elements as <tt>ty</tt></p>
3289 <p>The '<tt>sitofp</tt>' instruction interprets its operand as a signed
3290 integer quantity and converts it to the corresponding floating point value. If
3291 the value cannot fit in the floating point value, the results are undefined.</p>
3295 %X = sitofp i32 257 to float <i>; yields float:257.0</i>
3296 %Y = sitofp i8 -1 to double <i>; yields double:-1.0</i>
3300 <!-- _______________________________________________________________________ -->
3301 <div class="doc_subsubsection">
3302 <a name="i_ptrtoint">'<tt>ptrtoint .. to</tt>' Instruction</a>
3304 <div class="doc_text">
3308 <result> = ptrtoint <ty> <value> to <ty2> <i>; yields ty2</i>
3312 <p>The '<tt>ptrtoint</tt>' instruction converts the pointer <tt>value</tt> to
3313 the integer type <tt>ty2</tt>.</p>
3316 <p>The '<tt>ptrtoint</tt>' instruction takes a <tt>value</tt> to cast, which
3317 must be a <a href="#t_pointer">pointer</a> value, and a type to cast it to
3318 <tt>ty2</tt>, which must be an <a href="#t_integer">integer</a> type.
3321 <p>The '<tt>ptrtoint</tt>' instruction converts <tt>value</tt> to integer type
3322 <tt>ty2</tt> by interpreting the pointer value as an integer and either
3323 truncating or zero extending that value to the size of the integer type. If
3324 <tt>value</tt> is smaller than <tt>ty2</tt> then a zero extension is done. If
3325 <tt>value</tt> is larger than <tt>ty2</tt> then a truncation is done. If they
3326 are the same size, then nothing is done (<i>no-op cast</i>) other than a type
3331 %X = ptrtoint i32* %X to i8 <i>; yields truncation on 32-bit architecture</i>
3332 %Y = ptrtoint i32* %x to i64 <i>; yields zero extension on 32-bit architecture</i>
3336 <!-- _______________________________________________________________________ -->
3337 <div class="doc_subsubsection">
3338 <a name="i_inttoptr">'<tt>inttoptr .. to</tt>' Instruction</a>
3340 <div class="doc_text">
3344 <result> = inttoptr <ty> <value> to <ty2> <i>; yields ty2</i>
3348 <p>The '<tt>inttoptr</tt>' instruction converts an integer <tt>value</tt> to
3349 a pointer type, <tt>ty2</tt>.</p>
3352 <p>The '<tt>inttoptr</tt>' instruction takes an <a href="#t_integer">integer</a>
3353 value to cast, and a type to cast it to, which must be a
3354 <a href="#t_pointer">pointer</a> type.
3357 <p>The '<tt>inttoptr</tt>' instruction converts <tt>value</tt> to type
3358 <tt>ty2</tt> by applying either a zero extension or a truncation depending on
3359 the size of the integer <tt>value</tt>. If <tt>value</tt> is larger than the
3360 size of a pointer then a truncation is done. If <tt>value</tt> is smaller than
3361 the size of a pointer then a zero extension is done. If they are the same size,
3362 nothing is done (<i>no-op cast</i>).</p>
3366 %X = inttoptr i32 255 to i32* <i>; yields zero extension on 64-bit architecture</i>
3367 %X = inttoptr i32 255 to i32* <i>; yields no-op on 32-bit architecture</i>
3368 %Y = inttoptr i64 0 to i32* <i>; yields truncation on 32-bit architecture</i>
3372 <!-- _______________________________________________________________________ -->
3373 <div class="doc_subsubsection">
3374 <a name="i_bitcast">'<tt>bitcast .. to</tt>' Instruction</a>
3376 <div class="doc_text">
3380 <result> = bitcast <ty> <value> to <ty2> <i>; yields ty2</i>
3384 <p>The '<tt>bitcast</tt>' instruction converts <tt>value</tt> to type
3385 <tt>ty2</tt> without changing any bits.</p>
3388 <p>The '<tt>bitcast</tt>' instruction takes a value to cast, which must be
3389 a first class value, and a type to cast it to, which must also be a <a
3390 href="#t_firstclass">first class</a> type. The bit sizes of <tt>value</tt>
3391 and the destination type, <tt>ty2</tt>, must be identical. If the source
3392 type is a pointer, the destination type must also be a pointer.</p>
3395 <p>The '<tt>bitcast</tt>' instruction converts <tt>value</tt> to type
3396 <tt>ty2</tt>. It is always a <i>no-op cast</i> because no bits change with
3397 this conversion. The conversion is done as if the <tt>value</tt> had been
3398 stored to memory and read back as type <tt>ty2</tt>. Pointer types may only be
3399 converted to other pointer types with this instruction. To convert pointers to
3400 other types, use the <a href="#i_inttoptr">inttoptr</a> or
3401 <a href="#i_ptrtoint">ptrtoint</a> instructions first.</p>
3405 %X = bitcast i8 255 to i8 <i>; yields i8 :-1</i>
3406 %Y = bitcast i32* %x to sint* <i>; yields sint*:%x</i>
3407 %Z = bitcast <2xint> %V to i64; <i>; yields i64: %V</i>
3411 <!-- ======================================================================= -->
3412 <div class="doc_subsection"> <a name="otherops">Other Operations</a> </div>
3413 <div class="doc_text">
3414 <p>The instructions in this category are the "miscellaneous"
3415 instructions, which defy better classification.</p>
3418 <!-- _______________________________________________________________________ -->
3419 <div class="doc_subsubsection"><a name="i_icmp">'<tt>icmp</tt>' Instruction</a>
3421 <div class="doc_text">
3423 <pre> <result> = icmp <cond> <ty> <var1>, <var2> <i>; yields {i1}:result</i>
3426 <p>The '<tt>icmp</tt>' instruction returns a boolean value based on comparison
3427 of its two integer operands.</p>
3429 <p>The '<tt>icmp</tt>' instruction takes three operands. The first operand is
3430 the condition code indicating the kind of comparison to perform. It is not
3431 a value, just a keyword. The possible condition code are:
3433 <li><tt>eq</tt>: equal</li>
3434 <li><tt>ne</tt>: not equal </li>
3435 <li><tt>ugt</tt>: unsigned greater than</li>
3436 <li><tt>uge</tt>: unsigned greater or equal</li>
3437 <li><tt>ult</tt>: unsigned less than</li>
3438 <li><tt>ule</tt>: unsigned less or equal</li>
3439 <li><tt>sgt</tt>: signed greater than</li>
3440 <li><tt>sge</tt>: signed greater or equal</li>
3441 <li><tt>slt</tt>: signed less than</li>
3442 <li><tt>sle</tt>: signed less or equal</li>
3444 <p>The remaining two arguments must be <a href="#t_integer">integer</a> or
3445 <a href="#t_pointer">pointer</a> typed. They must also be identical types.</p>
3447 <p>The '<tt>icmp</tt>' compares <tt>var1</tt> and <tt>var2</tt> according to
3448 the condition code given as <tt>cond</tt>. The comparison performed always
3449 yields a <a href="#t_primitive">i1</a> result, as follows:
3451 <li><tt>eq</tt>: yields <tt>true</tt> if the operands are equal,
3452 <tt>false</tt> otherwise. No sign interpretation is necessary or performed.
3454 <li><tt>ne</tt>: yields <tt>true</tt> if the operands are unequal,
3455 <tt>false</tt> otherwise. No sign interpretation is necessary or performed.
3456 <li><tt>ugt</tt>: interprets the operands as unsigned values and yields
3457 <tt>true</tt> if <tt>var1</tt> is greater than <tt>var2</tt>.</li>
3458 <li><tt>uge</tt>: interprets the operands as unsigned values and yields
3459 <tt>true</tt> if <tt>var1</tt> is greater than or equal to <tt>var2</tt>.</li>
3460 <li><tt>ult</tt>: interprets the operands as unsigned values and yields
3461 <tt>true</tt> if <tt>var1</tt> is less than <tt>var2</tt>.</li>
3462 <li><tt>ule</tt>: interprets the operands as unsigned values and yields
3463 <tt>true</tt> if <tt>var1</tt> is less than or equal to <tt>var2</tt>.</li>
3464 <li><tt>sgt</tt>: interprets the operands as signed values and yields
3465 <tt>true</tt> if <tt>var1</tt> is greater than <tt>var2</tt>.</li>
3466 <li><tt>sge</tt>: interprets the operands as signed values and yields
3467 <tt>true</tt> if <tt>var1</tt> is greater than or equal to <tt>var2</tt>.</li>
3468 <li><tt>slt</tt>: interprets the operands as signed values and yields
3469 <tt>true</tt> if <tt>var1</tt> is less than <tt>var2</tt>.</li>
3470 <li><tt>sle</tt>: interprets the operands as signed values and yields
3471 <tt>true</tt> if <tt>var1</tt> is less than or equal to <tt>var2</tt>.</li>
3473 <p>If the operands are <a href="#t_pointer">pointer</a> typed, the pointer
3474 values are compared as if they were integers.</p>
3477 <pre> <result> = icmp eq i32 4, 5 <i>; yields: result=false</i>
3478 <result> = icmp ne float* %X, %X <i>; yields: result=false</i>
3479 <result> = icmp ult i16 4, 5 <i>; yields: result=true</i>
3480 <result> = icmp sgt i16 4, 5 <i>; yields: result=false</i>
3481 <result> = icmp ule i16 -4, 5 <i>; yields: result=false</i>
3482 <result> = icmp sge i16 4, 5 <i>; yields: result=false</i>
3486 <!-- _______________________________________________________________________ -->
3487 <div class="doc_subsubsection"><a name="i_fcmp">'<tt>fcmp</tt>' Instruction</a>
3489 <div class="doc_text">
3491 <pre> <result> = fcmp <cond> <ty> <var1>, <var2> <i>; yields {i1}:result</i>
3494 <p>The '<tt>fcmp</tt>' instruction returns a boolean value based on comparison
3495 of its floating point operands.</p>
3497 <p>The '<tt>fcmp</tt>' instruction takes three operands. The first operand is
3498 the condition code indicating the kind of comparison to perform. It is not
3499 a value, just a keyword. The possible condition code are:
3501 <li><tt>false</tt>: no comparison, always returns false</li>
3502 <li><tt>oeq</tt>: ordered and equal</li>
3503 <li><tt>ogt</tt>: ordered and greater than </li>
3504 <li><tt>oge</tt>: ordered and greater than or equal</li>
3505 <li><tt>olt</tt>: ordered and less than </li>
3506 <li><tt>ole</tt>: ordered and less than or equal</li>
3507 <li><tt>one</tt>: ordered and not equal</li>
3508 <li><tt>ord</tt>: ordered (no nans)</li>
3509 <li><tt>ueq</tt>: unordered or equal</li>
3510 <li><tt>ugt</tt>: unordered or greater than </li>
3511 <li><tt>uge</tt>: unordered or greater than or equal</li>
3512 <li><tt>ult</tt>: unordered or less than </li>
3513 <li><tt>ule</tt>: unordered or less than or equal</li>
3514 <li><tt>une</tt>: unordered or not equal</li>
3515 <li><tt>uno</tt>: unordered (either nans)</li>
3516 <li><tt>true</tt>: no comparison, always returns true</li>
3518 <p><i>Ordered</i> means that neither operand is a QNAN while
3519 <i>unordered</i> means that either operand may be a QNAN.</p>
3520 <p>The <tt>val1</tt> and <tt>val2</tt> arguments must be
3521 <a href="#t_floating">floating point</a> typed. They must have identical
3524 <p>The '<tt>fcmp</tt>' compares <tt>var1</tt> and <tt>var2</tt> according to
3525 the condition code given as <tt>cond</tt>. The comparison performed always
3526 yields a <a href="#t_primitive">i1</a> result, as follows:
3528 <li><tt>false</tt>: always yields <tt>false</tt>, regardless of operands.</li>
3529 <li><tt>oeq</tt>: yields <tt>true</tt> if both operands are not a QNAN and
3530 <tt>var1</tt> is equal to <tt>var2</tt>.</li>
3531 <li><tt>ogt</tt>: yields <tt>true</tt> if both operands are not a QNAN and
3532 <tt>var1</tt> is greather than <tt>var2</tt>.</li>
3533 <li><tt>oge</tt>: yields <tt>true</tt> if both operands are not a QNAN and
3534 <tt>var1</tt> is greater than or equal to <tt>var2</tt>.</li>
3535 <li><tt>olt</tt>: yields <tt>true</tt> if both operands are not a QNAN and
3536 <tt>var1</tt> is less than <tt>var2</tt>.</li>
3537 <li><tt>ole</tt>: yields <tt>true</tt> if both operands are not a QNAN and
3538 <tt>var1</tt> is less than or equal to <tt>var2</tt>.</li>
3539 <li><tt>one</tt>: yields <tt>true</tt> if both operands are not a QNAN and
3540 <tt>var1</tt> is not equal to <tt>var2</tt>.</li>
3541 <li><tt>ord</tt>: yields <tt>true</tt> if both operands are not a QNAN.</li>
3542 <li><tt>ueq</tt>: yields <tt>true</tt> if either operand is a QNAN or
3543 <tt>var1</tt> is equal to <tt>var2</tt>.</li>
3544 <li><tt>ugt</tt>: yields <tt>true</tt> if either operand is a QNAN or
3545 <tt>var1</tt> is greater than <tt>var2</tt>.</li>
3546 <li><tt>uge</tt>: yields <tt>true</tt> if either operand is a QNAN or
3547 <tt>var1</tt> is greater than or equal to <tt>var2</tt>.</li>
3548 <li><tt>ult</tt>: yields <tt>true</tt> if either operand is a QNAN or
3549 <tt>var1</tt> is less than <tt>var2</tt>.</li>
3550 <li><tt>ule</tt>: yields <tt>true</tt> if either operand is a QNAN or
3551 <tt>var1</tt> is less than or equal to <tt>var2</tt>.</li>
3552 <li><tt>une</tt>: yields <tt>true</tt> if either operand is a QNAN or
3553 <tt>var1</tt> is not equal to <tt>var2</tt>.</li>
3554 <li><tt>uno</tt>: yields <tt>true</tt> if either operand is a QNAN.</li>
3555 <li><tt>true</tt>: always yields <tt>true</tt>, regardless of operands.</li>
3559 <pre> <result> = fcmp oeq float 4.0, 5.0 <i>; yields: result=false</i>
3560 <result> = icmp one float 4.0, 5.0 <i>; yields: result=true</i>
3561 <result> = icmp olt float 4.0, 5.0 <i>; yields: result=true</i>
3562 <result> = icmp ueq double 1.0, 2.0 <i>; yields: result=false</i>
3566 <!-- _______________________________________________________________________ -->
3567 <div class="doc_subsubsection"> <a name="i_phi">'<tt>phi</tt>'
3568 Instruction</a> </div>
3569 <div class="doc_text">
3571 <pre> <result> = phi <ty> [ <val0>, <label0>], ...<br></pre>
3573 <p>The '<tt>phi</tt>' instruction is used to implement the φ node in
3574 the SSA graph representing the function.</p>
3576 <p>The type of the incoming values is specified with the first type
3577 field. After this, the '<tt>phi</tt>' instruction takes a list of pairs
3578 as arguments, with one pair for each predecessor basic block of the
3579 current block. Only values of <a href="#t_firstclass">first class</a>
3580 type may be used as the value arguments to the PHI node. Only labels
3581 may be used as the label arguments.</p>
3582 <p>There must be no non-phi instructions between the start of a basic
3583 block and the PHI instructions: i.e. PHI instructions must be first in
3586 <p>At runtime, the '<tt>phi</tt>' instruction logically takes on the value
3587 specified by the pair corresponding to the predecessor basic block that executed
3588 just prior to the current block.</p>
3590 <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>
3593 <!-- _______________________________________________________________________ -->
3594 <div class="doc_subsubsection">
3595 <a name="i_select">'<tt>select</tt>' Instruction</a>
3598 <div class="doc_text">
3603 <result> = select i1 <cond>, <ty> <val1>, <ty> <val2> <i>; yields ty</i>
3609 The '<tt>select</tt>' instruction is used to choose one value based on a
3610 condition, without branching.
3617 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.
3623 If the boolean condition evaluates to true, the instruction returns the first
3624 value argument; otherwise, it returns the second value argument.
3630 %X = select i1 true, i8 17, i8 42 <i>; yields i8:17</i>
3635 <!-- _______________________________________________________________________ -->
3636 <div class="doc_subsubsection">
3637 <a name="i_call">'<tt>call</tt>' Instruction</a>
3640 <div class="doc_text">
3644 <result> = [tail] call [<a href="#callingconv">cconv</a>] <ty> [<fnty>*] <fnptrval>(<param list>)
3649 <p>The '<tt>call</tt>' instruction represents a simple function call.</p>
3653 <p>This instruction requires several arguments:</p>
3657 <p>The optional "tail" marker indicates whether the callee function accesses
3658 any allocas or varargs in the caller. If the "tail" marker is present, the
3659 function call is eligible for tail call optimization. Note that calls may
3660 be marked "tail" even if they do not occur before a <a
3661 href="#i_ret"><tt>ret</tt></a> instruction.
3664 <p>The optional "cconv" marker indicates which <a href="#callingconv">calling
3665 convention</a> the call should use. If none is specified, the call defaults
3666 to using C calling conventions.
3669 <p>'<tt>ty</tt>': the type of the call instruction itself which is also
3670 the type of the return value. Functions that return no value are marked
3671 <tt><a href="#t_void">void</a></tt>.</p>
3674 <p>'<tt>fnty</tt>': shall be the signature of the pointer to function
3675 value being invoked. The argument types must match the types implied by
3676 this signature. This type can be omitted if the function is not varargs
3677 and if the function type does not return a pointer to a function.</p>
3680 <p>'<tt>fnptrval</tt>': An LLVM value containing a pointer to a function to
3681 be invoked. In most cases, this is a direct function invocation, but
3682 indirect <tt>call</tt>s are just as possible, calling an arbitrary pointer
3683 to function value.</p>
3686 <p>'<tt>function args</tt>': argument list whose types match the
3687 function signature argument types. All arguments must be of
3688 <a href="#t_firstclass">first class</a> type. If the function signature
3689 indicates the function accepts a variable number of arguments, the extra
3690 arguments can be specified.</p>
3696 <p>The '<tt>call</tt>' instruction is used to cause control flow to
3697 transfer to a specified function, with its incoming arguments bound to
3698 the specified values. Upon a '<tt><a href="#i_ret">ret</a></tt>'
3699 instruction in the called function, control flow continues with the
3700 instruction after the function call, and the return value of the
3701 function is bound to the result argument. This is a simpler case of
3702 the <a href="#i_invoke">invoke</a> instruction.</p>
3707 %retval = call i32 @test(i32 %argc)
3708 call i32 (i8 *, ...)* @printf(i8 * %msg, i32 12, i8 42);
3709 %X = tail call i32 @foo()
3710 %Y = tail call <a href="#callingconv">fastcc</a> i32 @foo()
3711 %Z = call void %foo(i8 97 signext)
3716 <!-- _______________________________________________________________________ -->
3717 <div class="doc_subsubsection">
3718 <a name="i_va_arg">'<tt>va_arg</tt>' Instruction</a>
3721 <div class="doc_text">
3726 <resultval> = va_arg <va_list*> <arglist>, <argty>
3731 <p>The '<tt>va_arg</tt>' instruction is used to access arguments passed through
3732 the "variable argument" area of a function call. It is used to implement the
3733 <tt>va_arg</tt> macro in C.</p>
3737 <p>This instruction takes a <tt>va_list*</tt> value and the type of
3738 the argument. It returns a value of the specified argument type and
3739 increments the <tt>va_list</tt> to point to the next argument. The
3740 actual type of <tt>va_list</tt> is target specific.</p>
3744 <p>The '<tt>va_arg</tt>' instruction loads an argument of the specified
3745 type from the specified <tt>va_list</tt> and causes the
3746 <tt>va_list</tt> to point to the next argument. For more information,
3747 see the variable argument handling <a href="#int_varargs">Intrinsic
3750 <p>It is legal for this instruction to be called in a function which does not
3751 take a variable number of arguments, for example, the <tt>vfprintf</tt>
3754 <p><tt>va_arg</tt> is an LLVM instruction instead of an <a
3755 href="#intrinsics">intrinsic function</a> because it takes a type as an
3760 <p>See the <a href="#int_varargs">variable argument processing</a> section.</p>
3764 <!-- *********************************************************************** -->
3765 <div class="doc_section"> <a name="intrinsics">Intrinsic Functions</a> </div>
3766 <!-- *********************************************************************** -->
3768 <div class="doc_text">
3770 <p>LLVM supports the notion of an "intrinsic function". These functions have
3771 well known names and semantics and are required to follow certain restrictions.
3772 Overall, these intrinsics represent an extension mechanism for the LLVM
3773 language that does not require changing all of the transformations in LLVM when
3774 adding to the language (or the bitcode reader/writer, the parser, etc...).</p>
3776 <p>Intrinsic function names must all start with an "<tt>llvm.</tt>" prefix. This
3777 prefix is reserved in LLVM for intrinsic names; thus, function names may not
3778 begin with this prefix. Intrinsic functions must always be external functions:
3779 you cannot define the body of intrinsic functions. Intrinsic functions may
3780 only be used in call or invoke instructions: it is illegal to take the address
3781 of an intrinsic function. Additionally, because intrinsic functions are part
3782 of the LLVM language, it is required if any are added that they be documented
3785 <p>Some intrinsic functions can be overloaded, i.e., the intrinsic represents
3786 a family of functions that perform the same operation but on different data
3787 types. Because LLVM can represent over 8 million different integer types,
3788 overloading is used commonly to allow an intrinsic function to operate on any
3789 integer type. One or more of the argument types or the result type can be
3790 overloaded to accept any integer type. Argument types may also be defined as
3791 exactly matching a previous argument's type or the result type. This allows an
3792 intrinsic function which accepts multiple arguments, but needs all of them to
3793 be of the same type, to only be overloaded with respect to a single argument or
3796 <p>Overloaded intrinsics will have the names of its overloaded argument types
3797 encoded into its function name, each preceded by a period. Only those types
3798 which are overloaded result in a name suffix. Arguments whose type is matched
3799 against another type do not. For example, the <tt>llvm.ctpop</tt> function can
3800 take an integer of any width and returns an integer of exactly the same integer
3801 width. This leads to a family of functions such as
3802 <tt>i8 @llvm.ctpop.i8(i8 %val)</tt> and <tt>i29 @llvm.ctpop.i29(i29 %val)</tt>.
3803 Only one type, the return type, is overloaded, and only one type suffix is
3804 required. Because the argument's type is matched against the return type, it
3805 does not require its own name suffix.</p>
3807 <p>To learn how to add an intrinsic function, please see the
3808 <a href="ExtendingLLVM.html">Extending LLVM Guide</a>.
3813 <!-- ======================================================================= -->
3814 <div class="doc_subsection">
3815 <a name="int_varargs">Variable Argument Handling Intrinsics</a>
3818 <div class="doc_text">
3820 <p>Variable argument support is defined in LLVM with the <a
3821 href="#i_va_arg"><tt>va_arg</tt></a> instruction and these three
3822 intrinsic functions. These functions are related to the similarly
3823 named macros defined in the <tt><stdarg.h></tt> header file.</p>
3825 <p>All of these functions operate on arguments that use a
3826 target-specific value type "<tt>va_list</tt>". The LLVM assembly
3827 language reference manual does not define what this type is, so all
3828 transformations should be prepared to handle these functions regardless of
3831 <p>This example shows how the <a href="#i_va_arg"><tt>va_arg</tt></a>
3832 instruction and the variable argument handling intrinsic functions are
3835 <div class="doc_code">
3837 define i32 @test(i32 %X, ...) {
3838 ; Initialize variable argument processing
3840 %ap2 = bitcast i8** %ap to i8*
3841 call void @llvm.va_start(i8* %ap2)
3843 ; Read a single integer argument
3844 %tmp = va_arg i8** %ap, i32
3846 ; Demonstrate usage of llvm.va_copy and llvm.va_end
3848 %aq2 = bitcast i8** %aq to i8*
3849 call void @llvm.va_copy(i8* %aq2, i8* %ap2)
3850 call void @llvm.va_end(i8* %aq2)
3852 ; Stop processing of arguments.
3853 call void @llvm.va_end(i8* %ap2)
3857 declare void @llvm.va_start(i8*)
3858 declare void @llvm.va_copy(i8*, i8*)
3859 declare void @llvm.va_end(i8*)
3865 <!-- _______________________________________________________________________ -->
3866 <div class="doc_subsubsection">
3867 <a name="int_va_start">'<tt>llvm.va_start</tt>' Intrinsic</a>
3871 <div class="doc_text">
3873 <pre> declare void %llvm.va_start(i8* <arglist>)<br></pre>
3875 <P>The '<tt>llvm.va_start</tt>' intrinsic initializes
3876 <tt>*<arglist></tt> for subsequent use by <tt><a
3877 href="#i_va_arg">va_arg</a></tt>.</p>
3881 <P>The argument is a pointer to a <tt>va_list</tt> element to initialize.</p>
3885 <P>The '<tt>llvm.va_start</tt>' intrinsic works just like the <tt>va_start</tt>
3886 macro available in C. In a target-dependent way, it initializes the
3887 <tt>va_list</tt> element to which the argument points, so that the next call to
3888 <tt>va_arg</tt> will produce the first variable argument passed to the function.
3889 Unlike the C <tt>va_start</tt> macro, this intrinsic does not need to know the
3890 last argument of the function as the compiler can figure that out.</p>
3894 <!-- _______________________________________________________________________ -->
3895 <div class="doc_subsubsection">
3896 <a name="int_va_end">'<tt>llvm.va_end</tt>' Intrinsic</a>
3899 <div class="doc_text">
3901 <pre> declare void @llvm.va_end(i8* <arglist>)<br></pre>
3904 <p>The '<tt>llvm.va_end</tt>' intrinsic destroys <tt>*<arglist></tt>,
3905 which has been initialized previously with <tt><a href="#int_va_start">llvm.va_start</a></tt>
3906 or <tt><a href="#i_va_copy">llvm.va_copy</a></tt>.</p>
3910 <p>The argument is a pointer to a <tt>va_list</tt> to destroy.</p>
3914 <p>The '<tt>llvm.va_end</tt>' intrinsic works just like the <tt>va_end</tt>
3915 macro available in C. In a target-dependent way, it destroys the
3916 <tt>va_list</tt> element to which the argument points. Calls to <a
3917 href="#int_va_start"><tt>llvm.va_start</tt></a> and <a href="#int_va_copy">
3918 <tt>llvm.va_copy</tt></a> must be matched exactly with calls to
3919 <tt>llvm.va_end</tt>.</p>
3923 <!-- _______________________________________________________________________ -->
3924 <div class="doc_subsubsection">
3925 <a name="int_va_copy">'<tt>llvm.va_copy</tt>' Intrinsic</a>
3928 <div class="doc_text">
3933 declare void @llvm.va_copy(i8* <destarglist>, i8* <srcarglist>)
3938 <p>The '<tt>llvm.va_copy</tt>' intrinsic copies the current argument position
3939 from the source argument list to the destination argument list.</p>
3943 <p>The first argument is a pointer to a <tt>va_list</tt> element to initialize.
3944 The second argument is a pointer to a <tt>va_list</tt> element to copy from.</p>
3949 <p>The '<tt>llvm.va_copy</tt>' intrinsic works just like the <tt>va_copy</tt>
3950 macro available in C. In a target-dependent way, it copies the source
3951 <tt>va_list</tt> element into the destination <tt>va_list</tt> element. This
3952 intrinsic is necessary because the <tt><a href="#int_va_start">
3953 llvm.va_start</a></tt> intrinsic may be arbitrarily complex and require, for
3954 example, memory allocation.</p>
3958 <!-- ======================================================================= -->
3959 <div class="doc_subsection">
3960 <a name="int_gc">Accurate Garbage Collection Intrinsics</a>
3963 <div class="doc_text">
3966 LLVM support for <a href="GarbageCollection.html">Accurate Garbage
3967 Collection</a> requires the implementation and generation of these intrinsics.
3968 These intrinsics allow identification of <a href="#int_gcroot">GC roots on the
3969 stack</a>, as well as garbage collector implementations that require <a
3970 href="#int_gcread">read</a> and <a href="#int_gcwrite">write</a> barriers.
3971 Front-ends for type-safe garbage collected languages should generate these
3972 intrinsics to make use of the LLVM garbage collectors. For more details, see <a
3973 href="GarbageCollection.html">Accurate Garbage Collection with LLVM</a>.
3977 <!-- _______________________________________________________________________ -->
3978 <div class="doc_subsubsection">
3979 <a name="int_gcroot">'<tt>llvm.gcroot</tt>' Intrinsic</a>
3982 <div class="doc_text">
3987 declare void @llvm.gcroot(i8** %ptrloc, i8* %metadata)
3992 <p>The '<tt>llvm.gcroot</tt>' intrinsic declares the existence of a GC root to
3993 the code generator, and allows some metadata to be associated with it.</p>
3997 <p>The first argument specifies the address of a stack object that contains the
3998 root pointer. The second pointer (which must be either a constant or a global
3999 value address) contains the meta-data to be associated with the root.</p>
4003 <p>At runtime, a call to this intrinsics stores a null pointer into the "ptrloc"
4004 location. At compile-time, the code generator generates information to allow
4005 the runtime to find the pointer at GC safe points.
4011 <!-- _______________________________________________________________________ -->
4012 <div class="doc_subsubsection">
4013 <a name="int_gcread">'<tt>llvm.gcread</tt>' Intrinsic</a>
4016 <div class="doc_text">
4021 declare i8* @llvm.gcread(i8* %ObjPtr, i8** %Ptr)
4026 <p>The '<tt>llvm.gcread</tt>' intrinsic identifies reads of references from heap
4027 locations, allowing garbage collector implementations that require read
4032 <p>The second argument is the address to read from, which should be an address
4033 allocated from the garbage collector. The first object is a pointer to the
4034 start of the referenced object, if needed by the language runtime (otherwise
4039 <p>The '<tt>llvm.gcread</tt>' intrinsic has the same semantics as a load
4040 instruction, but may be replaced with substantially more complex code by the
4041 garbage collector runtime, as needed.</p>
4046 <!-- _______________________________________________________________________ -->
4047 <div class="doc_subsubsection">
4048 <a name="int_gcwrite">'<tt>llvm.gcwrite</tt>' Intrinsic</a>
4051 <div class="doc_text">
4056 declare void @llvm.gcwrite(i8* %P1, i8* %Obj, i8** %P2)
4061 <p>The '<tt>llvm.gcwrite</tt>' intrinsic identifies writes of references to heap
4062 locations, allowing garbage collector implementations that require write
4063 barriers (such as generational or reference counting collectors).</p>
4067 <p>The first argument is the reference to store, the second is the start of the
4068 object to store it to, and the third is the address of the field of Obj to
4069 store to. If the runtime does not require a pointer to the object, Obj may be
4074 <p>The '<tt>llvm.gcwrite</tt>' intrinsic has the same semantics as a store
4075 instruction, but may be replaced with substantially more complex code by the
4076 garbage collector runtime, as needed.</p>
4082 <!-- ======================================================================= -->
4083 <div class="doc_subsection">
4084 <a name="int_codegen">Code Generator Intrinsics</a>
4087 <div class="doc_text">
4089 These intrinsics are provided by LLVM to expose special features that may only
4090 be implemented with code generator support.
4095 <!-- _______________________________________________________________________ -->
4096 <div class="doc_subsubsection">
4097 <a name="int_returnaddress">'<tt>llvm.returnaddress</tt>' Intrinsic</a>
4100 <div class="doc_text">
4104 declare i8 *@llvm.returnaddress(i32 <level>)
4110 The '<tt>llvm.returnaddress</tt>' intrinsic attempts to compute a
4111 target-specific value indicating the return address of the current function
4112 or one of its callers.
4118 The argument to this intrinsic indicates which function to return the address
4119 for. Zero indicates the calling function, one indicates its caller, etc. The
4120 argument is <b>required</b> to be a constant integer value.
4126 The '<tt>llvm.returnaddress</tt>' intrinsic either returns a pointer indicating
4127 the return address of the specified call frame, or zero if it cannot be
4128 identified. The value returned by this intrinsic is likely to be incorrect or 0
4129 for arguments other than zero, so it should only be used for debugging purposes.
4133 Note that calling this intrinsic does not prevent function inlining or other
4134 aggressive transformations, so the value returned may not be that of the obvious
4135 source-language caller.
4140 <!-- _______________________________________________________________________ -->
4141 <div class="doc_subsubsection">
4142 <a name="int_frameaddress">'<tt>llvm.frameaddress</tt>' Intrinsic</a>
4145 <div class="doc_text">
4149 declare i8 *@llvm.frameaddress(i32 <level>)
4155 The '<tt>llvm.frameaddress</tt>' intrinsic attempts to return the
4156 target-specific frame pointer value for the specified stack frame.
4162 The argument to this intrinsic indicates which function to return the frame
4163 pointer for. Zero indicates the calling function, one indicates its caller,
4164 etc. The argument is <b>required</b> to be a constant integer value.
4170 The '<tt>llvm.frameaddress</tt>' intrinsic either returns a pointer indicating
4171 the frame address of the specified call frame, or zero if it cannot be
4172 identified. The value returned by this intrinsic is likely to be incorrect or 0
4173 for arguments other than zero, so it should only be used for debugging purposes.
4177 Note that calling this intrinsic does not prevent function inlining or other
4178 aggressive transformations, so the value returned may not be that of the obvious
4179 source-language caller.
4183 <!-- _______________________________________________________________________ -->
4184 <div class="doc_subsubsection">
4185 <a name="int_stacksave">'<tt>llvm.stacksave</tt>' Intrinsic</a>
4188 <div class="doc_text">
4192 declare i8 *@llvm.stacksave()
4198 The '<tt>llvm.stacksave</tt>' intrinsic is used to remember the current state of
4199 the function stack, for use with <a href="#int_stackrestore">
4200 <tt>llvm.stackrestore</tt></a>. This is useful for implementing language
4201 features like scoped automatic variable sized arrays in C99.
4207 This intrinsic returns a opaque pointer value that can be passed to <a
4208 href="#int_stackrestore"><tt>llvm.stackrestore</tt></a>. When an
4209 <tt>llvm.stackrestore</tt> intrinsic is executed with a value saved from
4210 <tt>llvm.stacksave</tt>, it effectively restores the state of the stack to the
4211 state it was in when the <tt>llvm.stacksave</tt> intrinsic executed. In
4212 practice, this pops any <a href="#i_alloca">alloca</a> blocks from the stack
4213 that were allocated after the <tt>llvm.stacksave</tt> was executed.
4218 <!-- _______________________________________________________________________ -->
4219 <div class="doc_subsubsection">
4220 <a name="int_stackrestore">'<tt>llvm.stackrestore</tt>' Intrinsic</a>
4223 <div class="doc_text">
4227 declare void @llvm.stackrestore(i8 * %ptr)
4233 The '<tt>llvm.stackrestore</tt>' intrinsic is used to restore the state of
4234 the function stack to the state it was in when the corresponding <a
4235 href="#int_stacksave"><tt>llvm.stacksave</tt></a> intrinsic executed. This is
4236 useful for implementing language features like scoped automatic variable sized
4243 See the description for <a href="#int_stacksave"><tt>llvm.stacksave</tt></a>.
4249 <!-- _______________________________________________________________________ -->
4250 <div class="doc_subsubsection">
4251 <a name="int_prefetch">'<tt>llvm.prefetch</tt>' Intrinsic</a>
4254 <div class="doc_text">
4258 declare void @llvm.prefetch(i8* <address>, i32 <rw>, i32 <locality>)
4265 The '<tt>llvm.prefetch</tt>' intrinsic is a hint to the code generator to insert
4266 a prefetch instruction if supported; otherwise, it is a noop. Prefetches have
4268 effect on the behavior of the program but can change its performance
4275 <tt>address</tt> is the address to be prefetched, <tt>rw</tt> is the specifier
4276 determining if the fetch should be for a read (0) or write (1), and
4277 <tt>locality</tt> is a temporal locality specifier ranging from (0) - no
4278 locality, to (3) - extremely local keep in cache. The <tt>rw</tt> and
4279 <tt>locality</tt> arguments must be constant integers.
4285 This intrinsic does not modify the behavior of the program. In particular,
4286 prefetches cannot trap and do not produce a value. On targets that support this
4287 intrinsic, the prefetch can provide hints to the processor cache for better
4293 <!-- _______________________________________________________________________ -->
4294 <div class="doc_subsubsection">
4295 <a name="int_pcmarker">'<tt>llvm.pcmarker</tt>' Intrinsic</a>
4298 <div class="doc_text">
4302 declare void @llvm.pcmarker(i32 <id>)
4309 The '<tt>llvm.pcmarker</tt>' intrinsic is a method to export a Program Counter
4311 code to simulators and other tools. The method is target specific, but it is
4312 expected that the marker will use exported symbols to transmit the PC of the marker.
4313 The marker makes no guarantees that it will remain with any specific instruction
4314 after optimizations. It is possible that the presence of a marker will inhibit
4315 optimizations. The intended use is to be inserted after optimizations to allow
4316 correlations of simulation runs.
4322 <tt>id</tt> is a numerical id identifying the marker.
4328 This intrinsic does not modify the behavior of the program. Backends that do not
4329 support this intrinisic may ignore it.
4334 <!-- _______________________________________________________________________ -->
4335 <div class="doc_subsubsection">
4336 <a name="int_readcyclecounter">'<tt>llvm.readcyclecounter</tt>' Intrinsic</a>
4339 <div class="doc_text">
4343 declare i64 @llvm.readcyclecounter( )
4350 The '<tt>llvm.readcyclecounter</tt>' intrinsic provides access to the cycle
4351 counter register (or similar low latency, high accuracy clocks) on those targets
4352 that support it. On X86, it should map to RDTSC. On Alpha, it should map to RPCC.
4353 As the backing counters overflow quickly (on the order of 9 seconds on alpha), this
4354 should only be used for small timings.
4360 When directly supported, reading the cycle counter should not modify any memory.
4361 Implementations are allowed to either return a application specific value or a
4362 system wide value. On backends without support, this is lowered to a constant 0.
4367 <!-- ======================================================================= -->
4368 <div class="doc_subsection">
4369 <a name="int_libc">Standard C Library Intrinsics</a>
4372 <div class="doc_text">
4374 LLVM provides intrinsics for a few important standard C library functions.
4375 These intrinsics allow source-language front-ends to pass information about the
4376 alignment of the pointer arguments to the code generator, providing opportunity
4377 for more efficient code generation.
4382 <!-- _______________________________________________________________________ -->
4383 <div class="doc_subsubsection">
4384 <a name="int_memcpy">'<tt>llvm.memcpy</tt>' Intrinsic</a>
4387 <div class="doc_text">
4391 declare void @llvm.memcpy.i32(i8 * <dest>, i8 * <src>,
4392 i32 <len>, i32 <align>)
4393 declare void @llvm.memcpy.i64(i8 * <dest>, i8 * <src>,
4394 i64 <len>, i32 <align>)
4400 The '<tt>llvm.memcpy.*</tt>' intrinsics copy a block of memory from the source
4401 location to the destination location.
4405 Note that, unlike the standard libc function, the <tt>llvm.memcpy.*</tt>
4406 intrinsics do not return a value, and takes an extra alignment argument.
4412 The first argument is a pointer to the destination, the second is a pointer to
4413 the source. The third argument is an integer argument
4414 specifying the number of bytes to copy, and the fourth argument is the alignment
4415 of the source and destination locations.
4419 If the call to this intrinisic has an alignment value that is not 0 or 1, then
4420 the caller guarantees that both the source and destination pointers are aligned
4427 The '<tt>llvm.memcpy.*</tt>' intrinsics copy a block of memory from the source
4428 location to the destination location, which are not allowed to overlap. It
4429 copies "len" bytes of memory over. If the argument is known to be aligned to
4430 some boundary, this can be specified as the fourth argument, otherwise it should
4436 <!-- _______________________________________________________________________ -->
4437 <div class="doc_subsubsection">
4438 <a name="int_memmove">'<tt>llvm.memmove</tt>' Intrinsic</a>
4441 <div class="doc_text">
4445 declare void @llvm.memmove.i32(i8 * <dest>, i8 * <src>,
4446 i32 <len>, i32 <align>)
4447 declare void @llvm.memmove.i64(i8 * <dest>, i8 * <src>,
4448 i64 <len>, i32 <align>)
4454 The '<tt>llvm.memmove.*</tt>' intrinsics move a block of memory from the source
4455 location to the destination location. It is similar to the
4456 '<tt>llvm.memcmp</tt>' intrinsic but allows the two memory locations to overlap.
4460 Note that, unlike the standard libc function, the <tt>llvm.memmove.*</tt>
4461 intrinsics do not return a value, and takes an extra alignment argument.
4467 The first argument is a pointer to the destination, the second is a pointer to
4468 the source. The third argument is an integer argument
4469 specifying the number of bytes to copy, and the fourth argument is the alignment
4470 of the source and destination locations.
4474 If the call to this intrinisic has an alignment value that is not 0 or 1, then
4475 the caller guarantees that the source and destination pointers are aligned to
4482 The '<tt>llvm.memmove.*</tt>' intrinsics copy a block of memory from the source
4483 location to the destination location, which may overlap. It
4484 copies "len" bytes of memory over. If the argument is known to be aligned to
4485 some boundary, this can be specified as the fourth argument, otherwise it should
4491 <!-- _______________________________________________________________________ -->
4492 <div class="doc_subsubsection">
4493 <a name="int_memset">'<tt>llvm.memset.*</tt>' Intrinsics</a>
4496 <div class="doc_text">
4500 declare void @llvm.memset.i32(i8 * <dest>, i8 <val>,
4501 i32 <len>, i32 <align>)
4502 declare void @llvm.memset.i64(i8 * <dest>, i8 <val>,
4503 i64 <len>, i32 <align>)
4509 The '<tt>llvm.memset.*</tt>' intrinsics fill a block of memory with a particular
4514 Note that, unlike the standard libc function, the <tt>llvm.memset</tt> intrinsic
4515 does not return a value, and takes an extra alignment argument.
4521 The first argument is a pointer to the destination to fill, the second is the
4522 byte value to fill it with, the third argument is an integer
4523 argument specifying the number of bytes to fill, and the fourth argument is the
4524 known alignment of destination location.
4528 If the call to this intrinisic has an alignment value that is not 0 or 1, then
4529 the caller guarantees that the destination pointer is aligned to that boundary.
4535 The '<tt>llvm.memset.*</tt>' intrinsics fill "len" bytes of memory starting at
4537 destination location. If the argument is known to be aligned to some boundary,
4538 this can be specified as the fourth argument, otherwise it should be set to 0 or
4544 <!-- _______________________________________________________________________ -->
4545 <div class="doc_subsubsection">
4546 <a name="int_sqrt">'<tt>llvm.sqrt.*</tt>' Intrinsic</a>
4549 <div class="doc_text">
4552 <p>This is an overloaded intrinsic. You can use <tt>llvm.sqrt</tt> on any
4553 floating point or vector of floating point type. Not all targets support all
4556 declare float @llvm.sqrt.f32(float %Val)
4557 declare double @llvm.sqrt.f64(double %Val)
4558 declare x86_fp80 @llvm.sqrt.f80(x86_fp80 %Val)
4559 declare fp128 @llvm.sqrt.f128(fp128 %Val)
4560 declare ppc_fp128 @llvm.sqrt.ppcf128(ppc_fp128 %Val)
4566 The '<tt>llvm.sqrt</tt>' intrinsics return the sqrt of the specified operand,
4567 returning the same value as the libm '<tt>sqrt</tt>' functions would. Unlike
4568 <tt>sqrt</tt> in libm, however, <tt>llvm.sqrt</tt> has undefined behavior for
4569 negative numbers (which allows for better optimization).
4575 The argument and return value are floating point numbers of the same type.
4581 This function returns the sqrt of the specified operand if it is a nonnegative
4582 floating point number.
4586 <!-- _______________________________________________________________________ -->
4587 <div class="doc_subsubsection">
4588 <a name="int_powi">'<tt>llvm.powi.*</tt>' Intrinsic</a>
4591 <div class="doc_text">
4594 <p>This is an overloaded intrinsic. You can use <tt>llvm.powi</tt> on any
4595 floating point or vector of floating point type. Not all targets support all
4598 declare float @llvm.powi.f32(float %Val, i32 %power)
4599 declare double @llvm.powi.f64(double %Val, i32 %power)
4600 declare x86_fp80 @llvm.powi.f80(x86_fp80 %Val, i32 %power)
4601 declare fp128 @llvm.powi.f128(fp128 %Val, i32 %power)
4602 declare ppc_fp128 @llvm.powi.ppcf128(ppc_fp128 %Val, i32 %power)
4608 The '<tt>llvm.powi.*</tt>' intrinsics return the first operand raised to the
4609 specified (positive or negative) power. The order of evaluation of
4610 multiplications is not defined. When a vector of floating point type is
4611 used, the second argument remains a scalar integer value.
4617 The second argument is an integer power, and the first is a value to raise to
4624 This function returns the first value raised to the second power with an
4625 unspecified sequence of rounding operations.</p>
4628 <!-- _______________________________________________________________________ -->
4629 <div class="doc_subsubsection">
4630 <a name="int_sin">'<tt>llvm.sin.*</tt>' Intrinsic</a>
4633 <div class="doc_text">
4636 <p>This is an overloaded intrinsic. You can use <tt>llvm.sin</tt> on any
4637 floating point or vector of floating point type. Not all targets support all
4640 declare float @llvm.sin.f32(float %Val)
4641 declare double @llvm.sin.f64(double %Val)
4642 declare x86_fp80 @llvm.sin.f80(x86_fp80 %Val)
4643 declare fp128 @llvm.sin.f128(fp128 %Val)
4644 declare ppc_fp128 @llvm.sin.ppcf128(ppc_fp128 %Val)
4650 The '<tt>llvm.sin.*</tt>' intrinsics return the sine of the operand.
4656 The argument and return value are floating point numbers of the same type.
4662 This function returns the sine of the specified operand, returning the
4663 same values as the libm <tt>sin</tt> functions would, and handles error
4664 conditions in the same way.</p>
4667 <!-- _______________________________________________________________________ -->
4668 <div class="doc_subsubsection">
4669 <a name="int_cos">'<tt>llvm.cos.*</tt>' Intrinsic</a>
4672 <div class="doc_text">
4675 <p>This is an overloaded intrinsic. You can use <tt>llvm.cos</tt> on any
4676 floating point or vector of floating point type. Not all targets support all
4679 declare float @llvm.cos.f32(float %Val)
4680 declare double @llvm.cos.f64(double %Val)
4681 declare x86_fp80 @llvm.cos.f80(x86_fp80 %Val)
4682 declare fp128 @llvm.cos.f128(fp128 %Val)
4683 declare ppc_fp128 @llvm.cos.ppcf128(ppc_fp128 %Val)
4689 The '<tt>llvm.cos.*</tt>' intrinsics return the cosine of the operand.
4695 The argument and return value are floating point numbers of the same type.
4701 This function returns the cosine of the specified operand, returning the
4702 same values as the libm <tt>cos</tt> functions would, and handles error
4703 conditions in the same way.</p>
4706 <!-- _______________________________________________________________________ -->
4707 <div class="doc_subsubsection">
4708 <a name="int_pow">'<tt>llvm.pow.*</tt>' Intrinsic</a>
4711 <div class="doc_text">
4714 <p>This is an overloaded intrinsic. You can use <tt>llvm.pow</tt> on any
4715 floating point or vector of floating point type. Not all targets support all
4718 declare float @llvm.pow.f32(float %Val, float %Power)
4719 declare double @llvm.pow.f64(double %Val, double %Power)
4720 declare x86_fp80 @llvm.pow.f80(x86_fp80 %Val, x86_fp80 %Power)
4721 declare fp128 @llvm.pow.f128(fp128 %Val, fp128 %Power)
4722 declare ppc_fp128 @llvm.pow.ppcf128(ppc_fp128 %Val, ppc_fp128 Power)
4728 The '<tt>llvm.pow.*</tt>' intrinsics return the first operand raised to the
4729 specified (positive or negative) power.
4735 The second argument is a floating point power, and the first is a value to
4736 raise to that power.
4742 This function returns the first value raised to the second power,
4744 same values as the libm <tt>pow</tt> functions would, and handles error
4745 conditions in the same way.</p>
4749 <!-- ======================================================================= -->
4750 <div class="doc_subsection">
4751 <a name="int_manip">Bit Manipulation Intrinsics</a>
4754 <div class="doc_text">
4756 LLVM provides intrinsics for a few important bit manipulation operations.
4757 These allow efficient code generation for some algorithms.
4762 <!-- _______________________________________________________________________ -->
4763 <div class="doc_subsubsection">
4764 <a name="int_bswap">'<tt>llvm.bswap.*</tt>' Intrinsics</a>
4767 <div class="doc_text">
4770 <p>This is an overloaded intrinsic function. You can use bswap on any integer
4771 type that is an even number of bytes (i.e. BitWidth % 16 == 0).
4773 declare i16 @llvm.bswap.i16(i16 <id>)
4774 declare i32 @llvm.bswap.i32(i32 <id>)
4775 declare i64 @llvm.bswap.i64(i64 <id>)
4781 The '<tt>llvm.bswap</tt>' family of intrinsics is used to byte swap integer
4782 values with an even number of bytes (positive multiple of 16 bits). These are
4783 useful for performing operations on data that is not in the target's native
4790 The <tt>llvm.bswap.i16</tt> intrinsic returns an i16 value that has the high
4791 and low byte of the input i16 swapped. Similarly, the <tt>llvm.bswap.i32</tt>
4792 intrinsic returns an i32 value that has the four bytes of the input i32
4793 swapped, so that if the input bytes are numbered 0, 1, 2, 3 then the returned
4794 i32 will have its bytes in 3, 2, 1, 0 order. The <tt>llvm.bswap.i48</tt>,
4795 <tt>llvm.bswap.i64</tt> and other intrinsics extend this concept to
4796 additional even-byte lengths (6 bytes, 8 bytes and more, respectively).
4801 <!-- _______________________________________________________________________ -->
4802 <div class="doc_subsubsection">
4803 <a name="int_ctpop">'<tt>llvm.ctpop.*</tt>' Intrinsic</a>
4806 <div class="doc_text">
4809 <p>This is an overloaded intrinsic. You can use llvm.ctpop on any integer bit
4810 width. Not all targets support all bit widths however.
4812 declare i8 @llvm.ctpop.i8 (i8 <src>)
4813 declare i16 @llvm.ctpop.i16(i16 <src>)
4814 declare i32 @llvm.ctpop.i32(i32 <src>)
4815 declare i64 @llvm.ctpop.i64(i64 <src>)
4816 declare i256 @llvm.ctpop.i256(i256 <src>)
4822 The '<tt>llvm.ctpop</tt>' family of intrinsics counts the number of bits set in a
4829 The only argument is the value to be counted. The argument may be of any
4830 integer type. The return type must match the argument type.
4836 The '<tt>llvm.ctpop</tt>' intrinsic counts the 1's in a variable.
4840 <!-- _______________________________________________________________________ -->
4841 <div class="doc_subsubsection">
4842 <a name="int_ctlz">'<tt>llvm.ctlz.*</tt>' Intrinsic</a>
4845 <div class="doc_text">
4848 <p>This is an overloaded intrinsic. You can use <tt>llvm.ctlz</tt> on any
4849 integer bit width. Not all targets support all bit widths however.
4851 declare i8 @llvm.ctlz.i8 (i8 <src>)
4852 declare i16 @llvm.ctlz.i16(i16 <src>)
4853 declare i32 @llvm.ctlz.i32(i32 <src>)
4854 declare i64 @llvm.ctlz.i64(i64 <src>)
4855 declare i256 @llvm.ctlz.i256(i256 <src>)
4861 The '<tt>llvm.ctlz</tt>' family of intrinsic functions counts the number of
4862 leading zeros in a variable.
4868 The only argument is the value to be counted. The argument may be of any
4869 integer type. The return type must match the argument type.
4875 The '<tt>llvm.ctlz</tt>' intrinsic counts the leading (most significant) zeros
4876 in a variable. If the src == 0 then the result is the size in bits of the type
4877 of src. For example, <tt>llvm.ctlz(i32 2) = 30</tt>.
4883 <!-- _______________________________________________________________________ -->
4884 <div class="doc_subsubsection">
4885 <a name="int_cttz">'<tt>llvm.cttz.*</tt>' Intrinsic</a>
4888 <div class="doc_text">
4891 <p>This is an overloaded intrinsic. You can use <tt>llvm.cttz</tt> on any
4892 integer bit width. Not all targets support all bit widths however.
4894 declare i8 @llvm.cttz.i8 (i8 <src>)
4895 declare i16 @llvm.cttz.i16(i16 <src>)
4896 declare i32 @llvm.cttz.i32(i32 <src>)
4897 declare i64 @llvm.cttz.i64(i64 <src>)
4898 declare i256 @llvm.cttz.i256(i256 <src>)
4904 The '<tt>llvm.cttz</tt>' family of intrinsic functions counts the number of
4911 The only argument is the value to be counted. The argument may be of any
4912 integer type. The return type must match the argument type.
4918 The '<tt>llvm.cttz</tt>' intrinsic counts the trailing (least significant) zeros
4919 in a variable. If the src == 0 then the result is the size in bits of the type
4920 of src. For example, <tt>llvm.cttz(2) = 1</tt>.
4924 <!-- _______________________________________________________________________ -->
4925 <div class="doc_subsubsection">
4926 <a name="int_part_select">'<tt>llvm.part.select.*</tt>' Intrinsic</a>
4929 <div class="doc_text">
4932 <p>This is an overloaded intrinsic. You can use <tt>llvm.part.select</tt>
4933 on any integer bit width.
4935 declare i17 @llvm.part.select.i17 (i17 %val, i32 %loBit, i32 %hiBit)
4936 declare i29 @llvm.part.select.i29 (i29 %val, i32 %loBit, i32 %hiBit)
4940 <p>The '<tt>llvm.part.select</tt>' family of intrinsic functions selects a
4941 range of bits from an integer value and returns them in the same bit width as
4942 the original value.</p>
4945 <p>The first argument, <tt>%val</tt> and the result may be integer types of
4946 any bit width but they must have the same bit width. The second and third
4947 arguments must be <tt>i32</tt> type since they specify only a bit index.</p>
4950 <p>The operation of the '<tt>llvm.part.select</tt>' intrinsic has two modes
4951 of operation: forwards and reverse. If <tt>%loBit</tt> is greater than
4952 <tt>%hiBits</tt> then the intrinsic operates in reverse mode. Otherwise it
4953 operates in forward mode.</p>
4954 <p>In forward mode, this intrinsic is the equivalent of shifting <tt>%val</tt>
4955 right by <tt>%loBit</tt> bits and then ANDing it with a mask with
4956 only the <tt>%hiBit - %loBit</tt> bits set, as follows:</p>
4958 <li>The <tt>%val</tt> is shifted right (LSHR) by the number of bits specified
4959 by <tt>%loBits</tt>. This normalizes the value to the low order bits.</li>
4960 <li>The <tt>%loBits</tt> value is subtracted from the <tt>%hiBits</tt> value
4961 to determine the number of bits to retain.</li>
4962 <li>A mask of the retained bits is created by shifting a -1 value.</li>
4963 <li>The mask is ANDed with <tt>%val</tt> to produce the result.
4965 <p>In reverse mode, a similar computation is made except that the bits are
4966 returned in the reverse order. So, for example, if <tt>X</tt> has the value
4967 <tt>i16 0x0ACF (101011001111)</tt> and we apply
4968 <tt>part.select(i16 X, 8, 3)</tt> to it, we get back the value
4969 <tt>i16 0x0026 (000000100110)</tt>.</p>
4972 <div class="doc_subsubsection">
4973 <a name="int_part_set">'<tt>llvm.part.set.*</tt>' Intrinsic</a>
4976 <div class="doc_text">
4979 <p>This is an overloaded intrinsic. You can use <tt>llvm.part.set</tt>
4980 on any integer bit width.
4982 declare i17 @llvm.part.set.i17.i9 (i17 %val, i9 %repl, i32 %lo, i32 %hi)
4983 declare i29 @llvm.part.set.i29.i9 (i29 %val, i9 %repl, i32 %lo, i32 %hi)
4987 <p>The '<tt>llvm.part.set</tt>' family of intrinsic functions replaces a range
4988 of bits in an integer value with another integer value. It returns the integer
4989 with the replaced bits.</p>
4992 <p>The first argument, <tt>%val</tt> and the result may be integer types of
4993 any bit width but they must have the same bit width. <tt>%val</tt> is the value
4994 whose bits will be replaced. The second argument, <tt>%repl</tt> may be an
4995 integer of any bit width. The third and fourth arguments must be <tt>i32</tt>
4996 type since they specify only a bit index.</p>
4999 <p>The operation of the '<tt>llvm.part.set</tt>' intrinsic has two modes
5000 of operation: forwards and reverse. If <tt>%lo</tt> is greater than
5001 <tt>%hi</tt> then the intrinsic operates in reverse mode. Otherwise it
5002 operates in forward mode.</p>
5003 <p>For both modes, the <tt>%repl</tt> value is prepared for use by either
5004 truncating it down to the size of the replacement area or zero extending it
5005 up to that size.</p>
5006 <p>In forward mode, the bits between <tt>%lo</tt> and <tt>%hi</tt> (inclusive)
5007 are replaced with corresponding bits from <tt>%repl</tt>. That is the 0th bit
5008 in <tt>%repl</tt> replaces the <tt>%lo</tt>th bit in <tt>%val</tt> and etc. up
5009 to the <tt>%hi</tt>th bit.
5010 <p>In reverse mode, a similar computation is made except that the bits are
5011 reversed. That is, the <tt>0</tt>th bit in <tt>%repl</tt> replaces the
5012 <tt>%hi</tt> bit in <tt>%val</tt> and etc. down to the <tt>%lo</tt>th bit.
5015 llvm.part.set(0xFFFF, 0, 4, 7) -> 0xFF0F
5016 llvm.part.set(0xFFFF, 0, 7, 4) -> 0xFF0F
5017 llvm.part.set(0xFFFF, 1, 7, 4) -> 0xFF8F
5018 llvm.part.set(0xFFFF, F, 8, 3) -> 0xFFE7
5019 llvm.part.set(0xFFFF, 0, 3, 8) -> 0xFE07
5023 <!-- ======================================================================= -->
5024 <div class="doc_subsection">
5025 <a name="int_debugger">Debugger Intrinsics</a>
5028 <div class="doc_text">
5030 The LLVM debugger intrinsics (which all start with <tt>llvm.dbg.</tt> prefix),
5031 are described in the <a
5032 href="SourceLevelDebugging.html#format_common_intrinsics">LLVM Source Level
5033 Debugging</a> document.
5038 <!-- ======================================================================= -->
5039 <div class="doc_subsection">
5040 <a name="int_eh">Exception Handling Intrinsics</a>
5043 <div class="doc_text">
5044 <p> The LLVM exception handling intrinsics (which all start with
5045 <tt>llvm.eh.</tt> prefix), are described in the <a
5046 href="ExceptionHandling.html#format_common_intrinsics">LLVM Exception
5047 Handling</a> document. </p>
5050 <!-- ======================================================================= -->
5051 <div class="doc_subsection">
5052 <a name="int_trampoline">Trampoline Intrinsic</a>
5055 <div class="doc_text">
5057 This intrinsic makes it possible to excise one parameter, marked with
5058 the <tt>nest</tt> attribute, from a function. The result is a callable
5059 function pointer lacking the nest parameter - the caller does not need
5060 to provide a value for it. Instead, the value to use is stored in
5061 advance in a "trampoline", a block of memory usually allocated
5062 on the stack, which also contains code to splice the nest value into the
5063 argument list. This is used to implement the GCC nested function address
5067 For example, if the function is
5068 <tt>i32 f(i8* nest %c, i32 %x, i32 %y)</tt> then the resulting function
5069 pointer has signature <tt>i32 (i32, i32)*</tt>. It can be created as follows:</p>
5071 %tramp = alloca [10 x i8], align 4 ; size and alignment only correct for X86
5072 %tramp1 = getelementptr [10 x i8]* %tramp, i32 0, i32 0
5073 %p = call i8* @llvm.init.trampoline( i8* %tramp1, i8* bitcast (i32 (i8* nest , i32, i32)* @f to i8*), i8* %nval )
5074 %fp = bitcast i8* %p to i32 (i32, i32)*
5076 <p>The call <tt>%val = call i32 %fp( i32 %x, i32 %y )</tt> is then equivalent
5077 to <tt>%val = call i32 %f( i8* %nval, i32 %x, i32 %y )</tt>.</p>
5080 <!-- _______________________________________________________________________ -->
5081 <div class="doc_subsubsection">
5082 <a name="int_it">'<tt>llvm.init.trampoline</tt>' Intrinsic</a>
5084 <div class="doc_text">
5087 declare i8* @llvm.init.trampoline(i8* <tramp>, i8* <func>, i8* <nval>)
5091 This fills the memory pointed to by <tt>tramp</tt> with code
5092 and returns a function pointer suitable for executing it.
5096 The <tt>llvm.init.trampoline</tt> intrinsic takes three arguments, all
5097 pointers. The <tt>tramp</tt> argument must point to a sufficiently large
5098 and sufficiently aligned block of memory; this memory is written to by the
5099 intrinsic. Note that the size and the alignment are target-specific - LLVM
5100 currently provides no portable way of determining them, so a front-end that
5101 generates this intrinsic needs to have some target-specific knowledge.
5102 The <tt>func</tt> argument must hold a function bitcast to an <tt>i8*</tt>.
5106 The block of memory pointed to by <tt>tramp</tt> is filled with target
5107 dependent code, turning it into a function. A pointer to this function is
5108 returned, but needs to be bitcast to an
5109 <a href="#int_trampoline">appropriate function pointer type</a>
5110 before being called. The new function's signature is the same as that of
5111 <tt>func</tt> with any arguments marked with the <tt>nest</tt> attribute
5112 removed. At most one such <tt>nest</tt> argument is allowed, and it must be
5113 of pointer type. Calling the new function is equivalent to calling
5114 <tt>func</tt> with the same argument list, but with <tt>nval</tt> used for the
5115 missing <tt>nest</tt> argument. If, after calling
5116 <tt>llvm.init.trampoline</tt>, the memory pointed to by <tt>tramp</tt> is
5117 modified, then the effect of any later call to the returned function pointer is
5122 <!-- ======================================================================= -->
5123 <div class="doc_subsection">
5124 <a name="int_general">General Intrinsics</a>
5127 <div class="doc_text">
5128 <p> This class of intrinsics is designed to be generic and has
5129 no specific purpose. </p>
5132 <!-- _______________________________________________________________________ -->
5133 <div class="doc_subsubsection">
5134 <a name="int_var_annotation">'<tt>llvm.var.annotation</tt>' Intrinsic</a>
5137 <div class="doc_text">
5141 declare void @llvm.var.annotation(i8* <val>, i8* <str>, i8* <str>, i32 <int> )
5147 The '<tt>llvm.var.annotation</tt>' intrinsic
5153 The first argument is a pointer to a value, the second is a pointer to a
5154 global string, the third is a pointer to a global string which is the source
5155 file name, and the last argument is the line number.
5161 This intrinsic allows annotation of local variables with arbitrary strings.
5162 This can be useful for special purpose optimizations that want to look for these
5163 annotations. These have no other defined use, they are ignored by code
5164 generation and optimization.
5167 <!-- _______________________________________________________________________ -->
5168 <div class="doc_subsubsection">
5169 <a name="int_annotation">'<tt>llvm.annotation.*</tt>' Intrinsic</a>
5172 <div class="doc_text">
5175 <p>This is an overloaded intrinsic. You can use '<tt>llvm.annotation</tt>' on
5176 any integer bit width.
5179 declare i8 @llvm.annotation.i8(i8 <val>, i8* <str>, i8* <str>, i32 <int> )
5180 declare i16 @llvm.annotation.i16(i16 <val>, i8* <str>, i8* <str>, i32 <int> )
5181 declare i32 @llvm.annotation.i32(i32 <val>, i8* <str>, i8* <str>, i32 <int> )
5182 declare i64 @llvm.annotation.i64(i64 <val>, i8* <str>, i8* <str>, i32 <int> )
5183 declare i256 @llvm.annotation.i256(i256 <val>, i8* <str>, i8* <str>, i32 <int> )
5189 The '<tt>llvm.annotation</tt>' intrinsic.
5195 The first argument is an integer value (result of some expression),
5196 the second is a pointer to a global string, the third is a pointer to a global
5197 string which is the source file name, and the last argument is the line number.
5198 It returns the value of the first argument.
5204 This intrinsic allows annotations to be put on arbitrary expressions
5205 with arbitrary strings. This can be useful for special purpose optimizations
5206 that want to look for these annotations. These have no other defined use, they
5207 are ignored by code generation and optimization.
5210 <!-- *********************************************************************** -->
5213 <a href="http://jigsaw.w3.org/css-validator/check/referer"><img
5214 src="http://jigsaw.w3.org/css-validator/images/vcss" alt="Valid CSS!"></a>
5215 <a href="http://validator.w3.org/check/referer"><img
5216 src="http://www.w3.org/Icons/valid-html401" alt="Valid HTML 4.01!" /></a>
5218 <a href="mailto:sabre@nondot.org">Chris Lattner</a><br>
5219 <a href="http://llvm.org">The LLVM Compiler Infrastructure</a><br>
5220 Last modified: $Date$