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>LLVM allows an explicit section to be specified for globals. If the target
672 supports it, it will emit globals to the section specified.</p>
674 <p>An explicit alignment may be specified for a global. If not present, or if
675 the alignment is set to zero, the alignment of the global is set by the target
676 to whatever it feels convenient. If an explicit alignment is specified, the
677 global is forced to have at least that much alignment. All alignments must be
680 <p>For example, the following defines a global with an initializer, section,
683 <div class="doc_code">
685 @G = constant float 1.0, section "foo", align 4
692 <!-- ======================================================================= -->
693 <div class="doc_subsection">
694 <a name="functionstructure">Functions</a>
697 <div class="doc_text">
699 <p>LLVM function definitions consist of the "<tt>define</tt>" keyord,
700 an optional <a href="#linkage">linkage type</a>, an optional
701 <a href="#visibility">visibility style</a>, an optional
702 <a href="#callingconv">calling convention</a>, a return type, an optional
703 <a href="#paramattrs">parameter attribute</a> for the return type, a function
704 name, a (possibly empty) argument list (each with optional
705 <a href="#paramattrs">parameter attributes</a>), an optional section, an
706 optional alignment, an optional <a href="#gc">garbage collector name</a>, an
707 opening curly brace, a list of basic blocks, and a closing curly brace.
709 LLVM function declarations consist of the "<tt>declare</tt>" keyword, an
710 optional <a href="#linkage">linkage type</a>, an optional
711 <a href="#visibility">visibility style</a>, an optional
712 <a href="#callingconv">calling convention</a>, a return type, an optional
713 <a href="#paramattrs">parameter attribute</a> for the return type, a function
714 name, a possibly empty list of arguments, an optional alignment, and an optional
715 <a href="#gc">garbage collector name</a>.</p>
717 <p>A function definition contains a list of basic blocks, forming the CFG for
718 the function. Each basic block may optionally start with a label (giving the
719 basic block a symbol table entry), contains a list of instructions, and ends
720 with a <a href="#terminators">terminator</a> instruction (such as a branch or
721 function return).</p>
723 <p>The first basic block in a function is special in two ways: it is immediately
724 executed on entrance to the function, and it is not allowed to have predecessor
725 basic blocks (i.e. there can not be any branches to the entry block of a
726 function). Because the block can have no predecessors, it also cannot have any
727 <a href="#i_phi">PHI nodes</a>.</p>
729 <p>LLVM allows an explicit section to be specified for functions. If the target
730 supports it, it will emit functions to the section specified.</p>
732 <p>An explicit alignment may be specified for a function. If not present, or if
733 the alignment is set to zero, the alignment of the function is set by the target
734 to whatever it feels convenient. If an explicit alignment is specified, the
735 function is forced to have at least that much alignment. All alignments must be
741 <!-- ======================================================================= -->
742 <div class="doc_subsection">
743 <a name="aliasstructure">Aliases</a>
745 <div class="doc_text">
746 <p>Aliases act as "second name" for the aliasee value (which can be either
747 function or global variable or bitcast of global value). Aliases may have an
748 optional <a href="#linkage">linkage type</a>, and an
749 optional <a href="#visibility">visibility style</a>.</p>
753 <div class="doc_code">
755 @<Name> = [Linkage] [Visibility] alias <AliaseeTy> @<Aliasee>
763 <!-- ======================================================================= -->
764 <div class="doc_subsection"><a name="paramattrs">Parameter Attributes</a></div>
765 <div class="doc_text">
766 <p>The return type and each parameter of a function type may have a set of
767 <i>parameter attributes</i> associated with them. Parameter attributes are
768 used to communicate additional information about the result or parameters of
769 a function. Parameter attributes are considered to be part of the function,
770 not of the function type, so functions with different parameter attributes
771 can have the same function type.</p>
773 <p>Parameter attributes are simple keywords that follow the type specified. If
774 multiple parameter attributes are needed, they are space separated. For
777 <div class="doc_code">
779 declare i32 @printf(i8* noalias , ...) nounwind
780 declare i32 @atoi(i8*) nounwind readonly
784 <p>Note that any attributes for the function result (<tt>nounwind</tt>,
785 <tt>readonly</tt>) come immediately after the argument list.</p>
787 <p>Currently, only the following parameter attributes are defined:</p>
789 <dt><tt>zeroext</tt></dt>
790 <dd>This indicates that the parameter should be zero extended just before
791 a call to this function.</dd>
792 <dt><tt>signext</tt></dt>
793 <dd>This indicates that the parameter should be sign extended just before
794 a call to this function.</dd>
795 <dt><tt>inreg</tt></dt>
796 <dd>This indicates that the parameter should be placed in register (if
797 possible) during assembling function call. Support for this attribute is
799 <dt><tt>sret</tt></dt>
800 <dd>This indicates that the parameter specifies the address of a structure
801 that is the return value of the function in the source program.</dd>
802 <dt><tt>noalias</tt></dt>
803 <dd>This indicates that the parameter not alias any other object or any
804 other "noalias" objects during the function call.
805 <dt><tt>noreturn</tt></dt>
806 <dd>This function attribute indicates that the function never returns. This
807 indicates to LLVM that every call to this function should be treated as if
808 an <tt>unreachable</tt> instruction immediately followed the call.</dd>
809 <dt><tt>nounwind</tt></dt>
810 <dd>This function attribute indicates that the function type does not use
811 the unwind instruction and does not allow stack unwinding to propagate
813 <dt><tt>nest</tt></dt>
814 <dd>This indicates that the parameter can be excised using the
815 <a href="#int_trampoline">trampoline intrinsics</a>.</dd>
816 <dt><tt>readonly</tt></dt>
817 <dd>This function attribute indicates that the function has no side-effects
818 except for producing a return value or throwing an exception. The value
819 returned must only depend on the function arguments and/or global variables.
820 It may use values obtained by dereferencing pointers.</dd>
821 <dt><tt>readnone</tt></dt>
822 <dd>A <tt>readnone</tt> function has the same restrictions as a <tt>readonly</tt>
823 function, but in addition it is not allowed to dereference any pointer arguments
829 <!-- ======================================================================= -->
830 <div class="doc_subsection">
831 <a name="gc">Garbage Collector Names</a>
834 <div class="doc_text">
835 <p>Each function may specify a garbage collector name, which is simply a
838 <div class="doc_code"><pre
839 >define void @f() gc "name" { ...</pre></div>
841 <p>The compiler declares the supported values of <i>name</i>. Specifying a
842 collector which will cause the compiler to alter its output in order to support
843 the named garbage collection algorithm.</p>
846 <!-- ======================================================================= -->
847 <div class="doc_subsection">
848 <a name="moduleasm">Module-Level Inline Assembly</a>
851 <div class="doc_text">
853 Modules may contain "module-level inline asm" blocks, which corresponds to the
854 GCC "file scope inline asm" blocks. These blocks are internally concatenated by
855 LLVM and treated as a single unit, but may be separated in the .ll file if
856 desired. The syntax is very simple:
859 <div class="doc_code">
861 module asm "inline asm code goes here"
862 module asm "more can go here"
866 <p>The strings can contain any character by escaping non-printable characters.
867 The escape sequence used is simply "\xx" where "xx" is the two digit hex code
872 The inline asm code is simply printed to the machine code .s file when
873 assembly code is generated.
877 <!-- ======================================================================= -->
878 <div class="doc_subsection">
879 <a name="datalayout">Data Layout</a>
882 <div class="doc_text">
883 <p>A module may specify a target specific data layout string that specifies how
884 data is to be laid out in memory. The syntax for the data layout is simply:</p>
885 <pre> target datalayout = "<i>layout specification</i>"</pre>
886 <p>The <i>layout specification</i> consists of a list of specifications
887 separated by the minus sign character ('-'). Each specification starts with a
888 letter and may include other information after the letter to define some
889 aspect of the data layout. The specifications accepted are as follows: </p>
892 <dd>Specifies that the target lays out data in big-endian form. That is, the
893 bits with the most significance have the lowest address location.</dd>
895 <dd>Specifies that hte target lays out data in little-endian form. That is,
896 the bits with the least significance have the lowest address location.</dd>
897 <dt><tt>p:<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
898 <dd>This specifies the <i>size</i> of a pointer and its <i>abi</i> and
899 <i>preferred</i> alignments. All sizes are in bits. Specifying the <i>pref</i>
900 alignment is optional. If omitted, the preceding <tt>:</tt> should be omitted
902 <dt><tt>i<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
903 <dd>This specifies the alignment for an integer type of a given bit
904 <i>size</i>. The value of <i>size</i> must be in the range [1,2^23).</dd>
905 <dt><tt>v<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
906 <dd>This specifies the alignment for a vector type of a given bit
908 <dt><tt>f<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
909 <dd>This specifies the alignment for a floating point type of a given bit
910 <i>size</i>. The value of <i>size</i> must be either 32 (float) or 64
912 <dt><tt>a<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
913 <dd>This specifies the alignment for an aggregate type of a given bit
916 <p>When constructing the data layout for a given target, LLVM starts with a
917 default set of specifications which are then (possibly) overriden by the
918 specifications in the <tt>datalayout</tt> keyword. The default specifications
919 are given in this list:</p>
921 <li><tt>E</tt> - big endian</li>
922 <li><tt>p:32:64:64</tt> - 32-bit pointers with 64-bit alignment</li>
923 <li><tt>i1:8:8</tt> - i1 is 8-bit (byte) aligned</li>
924 <li><tt>i8:8:8</tt> - i8 is 8-bit (byte) aligned</li>
925 <li><tt>i16:16:16</tt> - i16 is 16-bit aligned</li>
926 <li><tt>i32:32:32</tt> - i32 is 32-bit aligned</li>
927 <li><tt>i64:32:64</tt> - i64 has abi alignment of 32-bits but preferred
928 alignment of 64-bits</li>
929 <li><tt>f32:32:32</tt> - float is 32-bit aligned</li>
930 <li><tt>f64:64:64</tt> - double is 64-bit aligned</li>
931 <li><tt>v64:64:64</tt> - 64-bit vector is 64-bit aligned</li>
932 <li><tt>v128:128:128</tt> - 128-bit vector is 128-bit aligned</li>
933 <li><tt>a0:0:1</tt> - aggregates are 8-bit aligned</li>
935 <p>When llvm is determining the alignment for a given type, it uses the
938 <li>If the type sought is an exact match for one of the specifications, that
939 specification is used.</li>
940 <li>If no match is found, and the type sought is an integer type, then the
941 smallest integer type that is larger than the bitwidth of the sought type is
942 used. If none of the specifications are larger than the bitwidth then the the
943 largest integer type is used. For example, given the default specifications
944 above, the i7 type will use the alignment of i8 (next largest) while both
945 i65 and i256 will use the alignment of i64 (largest specified).</li>
946 <li>If no match is found, and the type sought is a vector type, then the
947 largest vector type that is smaller than the sought vector type will be used
948 as a fall back. This happens because <128 x double> can be implemented in
949 terms of 64 <2 x double>, for example.</li>
953 <!-- *********************************************************************** -->
954 <div class="doc_section"> <a name="typesystem">Type System</a> </div>
955 <!-- *********************************************************************** -->
957 <div class="doc_text">
959 <p>The LLVM type system is one of the most important features of the
960 intermediate representation. Being typed enables a number of
961 optimizations to be performed on the IR directly, without having to do
962 extra analyses on the side before the transformation. A strong type
963 system makes it easier to read the generated code and enables novel
964 analyses and transformations that are not feasible to perform on normal
965 three address code representations.</p>
969 <!-- ======================================================================= -->
970 <div class="doc_subsection"> <a name="t_primitive">Primitive Types</a> </div>
971 <div class="doc_text">
972 <p>The primitive types are the fundamental building blocks of the LLVM
973 system. The current set of primitive types is as follows:</p>
975 <table class="layout">
980 <tr><th>Type</th><th>Description</th></tr>
981 <tr><td><tt><a name="t_void">void</a></tt></td><td>No value</td></tr>
982 <tr><td><tt>label</tt></td><td>Branch destination</td></tr>
989 <tr><th>Type</th><th>Description</th></tr>
990 <tr><td><tt>float</tt></td><td>32-bit floating point value</td></tr>
991 <tr><td><tt>double</tt></td><td>64-bit floating point value</td></tr>
999 <!-- _______________________________________________________________________ -->
1000 <div class="doc_subsubsection"> <a name="t_classifications">Type
1001 Classifications</a> </div>
1002 <div class="doc_text">
1003 <p>These different primitive types fall into a few useful
1004 classifications:</p>
1006 <table border="1" cellspacing="0" cellpadding="4">
1008 <tr><th>Classification</th><th>Types</th></tr>
1010 <td><a name="t_integer">integer</a></td>
1011 <td><tt>i1, i2, i3, ... i8, ... i16, ... i32, ... i64, ... </tt></td>
1014 <td><a name="t_floating">floating point</a></td>
1015 <td><tt>float, double</tt></td>
1018 <td><a name="t_firstclass">first class</a></td>
1019 <td><tt>i1, ..., float, double, <br/>
1020 <a href="#t_pointer">pointer</a>,<a href="#t_vector">vector</a></tt>
1026 <p>The <a href="#t_firstclass">first class</a> types are perhaps the
1027 most important. Values of these types are the only ones which can be
1028 produced by instructions, passed as arguments, or used as operands to
1029 instructions. This means that all structures and arrays must be
1030 manipulated either by pointer or by component.</p>
1033 <!-- ======================================================================= -->
1034 <div class="doc_subsection"> <a name="t_derived">Derived Types</a> </div>
1036 <div class="doc_text">
1038 <p>The real power in LLVM comes from the derived types in the system.
1039 This is what allows a programmer to represent arrays, functions,
1040 pointers, and other useful types. Note that these derived types may be
1041 recursive: For example, it is possible to have a two dimensional array.</p>
1045 <!-- _______________________________________________________________________ -->
1046 <div class="doc_subsubsection"> <a name="t_integer">Integer Type</a> </div>
1048 <div class="doc_text">
1051 <p>The integer type is a very simple derived type that simply specifies an
1052 arbitrary bit width for the integer type desired. Any bit width from 1 bit to
1053 2^23-1 (about 8 million) can be specified.</p>
1061 <p>The number of bits the integer will occupy is specified by the <tt>N</tt>
1065 <table class="layout">
1075 <tt>i1942652</tt><br/>
1078 A boolean integer of 1 bit<br/>
1079 A nibble sized integer of 4 bits.<br/>
1080 A byte sized integer of 8 bits.<br/>
1081 A half word sized integer of 16 bits.<br/>
1082 A word sized integer of 32 bits.<br/>
1083 An integer whose bit width is the answer. <br/>
1084 A double word sized integer of 64 bits.<br/>
1085 A really big integer of over 1 million bits.<br/>
1091 <!-- _______________________________________________________________________ -->
1092 <div class="doc_subsubsection"> <a name="t_array">Array Type</a> </div>
1094 <div class="doc_text">
1098 <p>The array type is a very simple derived type that arranges elements
1099 sequentially in memory. The array type requires a size (number of
1100 elements) and an underlying data type.</p>
1105 [<# elements> x <elementtype>]
1108 <p>The number of elements is a constant integer value; elementtype may
1109 be any type with a size.</p>
1112 <table class="layout">
1115 <tt>[40 x i32 ]</tt><br/>
1116 <tt>[41 x i32 ]</tt><br/>
1117 <tt>[40 x i8]</tt><br/>
1120 Array of 40 32-bit integer values.<br/>
1121 Array of 41 32-bit integer values.<br/>
1122 Array of 40 8-bit integer values.<br/>
1126 <p>Here are some examples of multidimensional arrays:</p>
1127 <table class="layout">
1130 <tt>[3 x [4 x i32]]</tt><br/>
1131 <tt>[12 x [10 x float]]</tt><br/>
1132 <tt>[2 x [3 x [4 x i16]]]</tt><br/>
1135 3x4 array of 32-bit integer values.<br/>
1136 12x10 array of single precision floating point values.<br/>
1137 2x3x4 array of 16-bit integer values.<br/>
1142 <p>Note that 'variable sized arrays' can be implemented in LLVM with a zero
1143 length array. Normally, accesses past the end of an array are undefined in
1144 LLVM (e.g. it is illegal to access the 5th element of a 3 element array).
1145 As a special case, however, zero length arrays are recognized to be variable
1146 length. This allows implementation of 'pascal style arrays' with the LLVM
1147 type "{ i32, [0 x float]}", for example.</p>
1151 <!-- _______________________________________________________________________ -->
1152 <div class="doc_subsubsection"> <a name="t_function">Function Type</a> </div>
1153 <div class="doc_text">
1155 <p>The function type can be thought of as a function signature. It
1156 consists of a return type and a list of formal parameter types.
1157 Function types are usually used to build virtual function tables
1158 (which are structures of pointers to functions), for indirect function
1159 calls, and when defining a function.</p>
1161 The return type of a function type cannot be an aggregate type.
1164 <pre> <returntype> (<parameter list>)<br></pre>
1165 <p>...where '<tt><parameter list></tt>' is a comma-separated list of type
1166 specifiers. Optionally, the parameter list may include a type <tt>...</tt>,
1167 which indicates that the function takes a variable number of arguments.
1168 Variable argument functions can access their arguments with the <a
1169 href="#int_varargs">variable argument handling intrinsic</a> functions.</p>
1171 <table class="layout">
1173 <td class="left"><tt>i32 (i32)</tt></td>
1174 <td class="left">function taking an <tt>i32</tt>, returning an <tt>i32</tt>
1176 </tr><tr class="layout">
1177 <td class="left"><tt>float (i16 signext, i32 *) *
1179 <td class="left"><a href="#t_pointer">Pointer</a> to a function that takes
1180 an <tt>i16</tt> that should be sign extended and a
1181 <a href="#t_pointer">pointer</a> to <tt>i32</tt>, returning
1184 </tr><tr class="layout">
1185 <td class="left"><tt>i32 (i8*, ...)</tt></td>
1186 <td class="left">A vararg function that takes at least one
1187 <a href="#t_pointer">pointer</a> to <tt>i8 </tt> (char in C),
1188 which returns an integer. This is the signature for <tt>printf</tt> in
1195 <!-- _______________________________________________________________________ -->
1196 <div class="doc_subsubsection"> <a name="t_struct">Structure Type</a> </div>
1197 <div class="doc_text">
1199 <p>The structure type is used to represent a collection of data members
1200 together in memory. The packing of the field types is defined to match
1201 the ABI of the underlying processor. The elements of a structure may
1202 be any type that has a size.</p>
1203 <p>Structures are accessed using '<tt><a href="#i_load">load</a></tt>
1204 and '<tt><a href="#i_store">store</a></tt>' by getting a pointer to a
1205 field with the '<tt><a href="#i_getelementptr">getelementptr</a></tt>'
1208 <pre> { <type list> }<br></pre>
1210 <table class="layout">
1212 <td class="left"><tt>{ i32, i32, i32 }</tt></td>
1213 <td class="left">A triple of three <tt>i32</tt> values</td>
1214 </tr><tr class="layout">
1215 <td class="left"><tt>{ float, i32 (i32) * }</tt></td>
1216 <td class="left">A pair, where the first element is a <tt>float</tt> and the
1217 second element is a <a href="#t_pointer">pointer</a> to a
1218 <a href="#t_function">function</a> that takes an <tt>i32</tt>, returning
1219 an <tt>i32</tt>.</td>
1224 <!-- _______________________________________________________________________ -->
1225 <div class="doc_subsubsection"> <a name="t_pstruct">Packed Structure Type</a>
1227 <div class="doc_text">
1229 <p>The packed structure type is used to represent a collection of data members
1230 together in memory. There is no padding between fields. Further, the alignment
1231 of a packed structure is 1 byte. The elements of a packed structure may
1232 be any type that has a size.</p>
1233 <p>Structures are accessed using '<tt><a href="#i_load">load</a></tt>
1234 and '<tt><a href="#i_store">store</a></tt>' by getting a pointer to a
1235 field with the '<tt><a href="#i_getelementptr">getelementptr</a></tt>'
1238 <pre> < { <type list> } > <br></pre>
1240 <table class="layout">
1242 <td class="left"><tt>< { i32, i32, i32 } ></tt></td>
1243 <td class="left">A triple of three <tt>i32</tt> values</td>
1244 </tr><tr class="layout">
1245 <td class="left"><tt>< { float, i32 (i32) * } ></tt></td>
1246 <td class="left">A pair, where the first element is a <tt>float</tt> and the
1247 second element is a <a href="#t_pointer">pointer</a> to a
1248 <a href="#t_function">function</a> that takes an <tt>i32</tt>, returning
1249 an <tt>i32</tt>.</td>
1254 <!-- _______________________________________________________________________ -->
1255 <div class="doc_subsubsection"> <a name="t_pointer">Pointer Type</a> </div>
1256 <div class="doc_text">
1258 <p>As in many languages, the pointer type represents a pointer or
1259 reference to another object, which must live in memory.</p>
1261 <pre> <type> *<br></pre>
1263 <table class="layout">
1266 <tt>[4x i32]*</tt><br/>
1267 <tt>i32 (i32 *) *</tt><br/>
1270 A <a href="#t_pointer">pointer</a> to <a href="#t_array">array</a> of
1271 four <tt>i32</tt> values<br/>
1272 A <a href="#t_pointer">pointer</a> to a <a
1273 href="#t_function">function</a> that takes an <tt>i32*</tt>, returning an
1280 <!-- _______________________________________________________________________ -->
1281 <div class="doc_subsubsection"> <a name="t_vector">Vector Type</a> </div>
1282 <div class="doc_text">
1286 <p>A vector type is a simple derived type that represents a vector
1287 of elements. Vector types are used when multiple primitive data
1288 are operated in parallel using a single instruction (SIMD).
1289 A vector type requires a size (number of
1290 elements) and an underlying primitive data type. Vectors must have a power
1291 of two length (1, 2, 4, 8, 16 ...). Vector types are
1292 considered <a href="#t_firstclass">first class</a>.</p>
1297 < <# elements> x <elementtype> >
1300 <p>The number of elements is a constant integer value; elementtype may
1301 be any integer or floating point type.</p>
1305 <table class="layout">
1308 <tt><4 x i32></tt><br/>
1309 <tt><8 x float></tt><br/>
1310 <tt><2 x i64></tt><br/>
1313 Vector of 4 32-bit integer values.<br/>
1314 Vector of 8 floating-point values.<br/>
1315 Vector of 2 64-bit integer values.<br/>
1321 <!-- _______________________________________________________________________ -->
1322 <div class="doc_subsubsection"> <a name="t_opaque">Opaque Type</a> </div>
1323 <div class="doc_text">
1327 <p>Opaque types are used to represent unknown types in the system. This
1328 corresponds (for example) to the C notion of a forward declared structure type.
1329 In LLVM, opaque types can eventually be resolved to any type (not just a
1330 structure type).</p>
1340 <table class="layout">
1346 An opaque type.<br/>
1353 <!-- *********************************************************************** -->
1354 <div class="doc_section"> <a name="constants">Constants</a> </div>
1355 <!-- *********************************************************************** -->
1357 <div class="doc_text">
1359 <p>LLVM has several different basic types of constants. This section describes
1360 them all and their syntax.</p>
1364 <!-- ======================================================================= -->
1365 <div class="doc_subsection"><a name="simpleconstants">Simple Constants</a></div>
1367 <div class="doc_text">
1370 <dt><b>Boolean constants</b></dt>
1372 <dd>The two strings '<tt>true</tt>' and '<tt>false</tt>' are both valid
1373 constants of the <tt><a href="#t_primitive">i1</a></tt> type.
1376 <dt><b>Integer constants</b></dt>
1378 <dd>Standard integers (such as '4') are constants of the <a
1379 href="#t_integer">integer</a> type. Negative numbers may be used with
1383 <dt><b>Floating point constants</b></dt>
1385 <dd>Floating point constants use standard decimal notation (e.g. 123.421),
1386 exponential notation (e.g. 1.23421e+2), or a more precise hexadecimal
1387 notation (see below). Floating point constants must have a <a
1388 href="#t_floating">floating point</a> type. </dd>
1390 <dt><b>Null pointer constants</b></dt>
1392 <dd>The identifier '<tt>null</tt>' is recognized as a null pointer constant
1393 and must be of <a href="#t_pointer">pointer type</a>.</dd>
1397 <p>The one non-intuitive notation for constants is the optional hexadecimal form
1398 of floating point constants. For example, the form '<tt>double
1399 0x432ff973cafa8000</tt>' is equivalent to (but harder to read than) '<tt>double
1400 4.5e+15</tt>'. The only time hexadecimal floating point constants are required
1401 (and the only time that they are generated by the disassembler) is when a
1402 floating point constant must be emitted but it cannot be represented as a
1403 decimal floating point number. For example, NaN's, infinities, and other
1404 special values are represented in their IEEE hexadecimal format so that
1405 assembly and disassembly do not cause any bits to change in the constants.</p>
1409 <!-- ======================================================================= -->
1410 <div class="doc_subsection"><a name="aggregateconstants">Aggregate Constants</a>
1413 <div class="doc_text">
1414 <p>Aggregate constants arise from aggregation of simple constants
1415 and smaller aggregate constants.</p>
1418 <dt><b>Structure constants</b></dt>
1420 <dd>Structure constants are represented with notation similar to structure
1421 type definitions (a comma separated list of elements, surrounded by braces
1422 (<tt>{}</tt>)). For example: "<tt>{ i32 4, float 17.0, i32* %G }</tt>",
1423 where "<tt>%G</tt>" is declared as "<tt>@G = external global i32</tt>". Structure constants
1424 must have <a href="#t_struct">structure type</a>, and the number and
1425 types of elements must match those specified by the type.
1428 <dt><b>Array constants</b></dt>
1430 <dd>Array constants are represented with notation similar to array type
1431 definitions (a comma separated list of elements, surrounded by square brackets
1432 (<tt>[]</tt>)). For example: "<tt>[ i32 42, i32 11, i32 74 ]</tt>". Array
1433 constants must have <a href="#t_array">array type</a>, and the number and
1434 types of elements must match those specified by the type.
1437 <dt><b>Vector constants</b></dt>
1439 <dd>Vector constants are represented with notation similar to vector type
1440 definitions (a comma separated list of elements, surrounded by
1441 less-than/greater-than's (<tt><></tt>)). For example: "<tt>< i32 42,
1442 i32 11, i32 74, i32 100 ></tt>". Vector constants must have <a
1443 href="#t_vector">vector type</a>, and the number and types of elements must
1444 match those specified by the type.
1447 <dt><b>Zero initialization</b></dt>
1449 <dd>The string '<tt>zeroinitializer</tt>' can be used to zero initialize a
1450 value to zero of <em>any</em> type, including scalar and aggregate types.
1451 This is often used to avoid having to print large zero initializers (e.g. for
1452 large arrays) and is always exactly equivalent to using explicit zero
1459 <!-- ======================================================================= -->
1460 <div class="doc_subsection">
1461 <a name="globalconstants">Global Variable and Function Addresses</a>
1464 <div class="doc_text">
1466 <p>The addresses of <a href="#globalvars">global variables</a> and <a
1467 href="#functionstructure">functions</a> are always implicitly valid (link-time)
1468 constants. These constants are explicitly referenced when the <a
1469 href="#identifiers">identifier for the global</a> is used and always have <a
1470 href="#t_pointer">pointer</a> type. For example, the following is a legal LLVM
1473 <div class="doc_code">
1477 @Z = global [2 x i32*] [ i32* @X, i32* @Y ]
1483 <!-- ======================================================================= -->
1484 <div class="doc_subsection"><a name="undefvalues">Undefined Values</a></div>
1485 <div class="doc_text">
1486 <p>The string '<tt>undef</tt>' is recognized as a type-less constant that has
1487 no specific value. Undefined values may be of any type and be used anywhere
1488 a constant is permitted.</p>
1490 <p>Undefined values indicate to the compiler that the program is well defined
1491 no matter what value is used, giving the compiler more freedom to optimize.
1495 <!-- ======================================================================= -->
1496 <div class="doc_subsection"><a name="constantexprs">Constant Expressions</a>
1499 <div class="doc_text">
1501 <p>Constant expressions are used to allow expressions involving other constants
1502 to be used as constants. Constant expressions may be of any <a
1503 href="#t_firstclass">first class</a> type and may involve any LLVM operation
1504 that does not have side effects (e.g. load and call are not supported). The
1505 following is the syntax for constant expressions:</p>
1508 <dt><b><tt>trunc ( CST to TYPE )</tt></b></dt>
1509 <dd>Truncate a constant to another type. The bit size of CST must be larger
1510 than the bit size of TYPE. Both types must be integers.</dd>
1512 <dt><b><tt>zext ( CST to TYPE )</tt></b></dt>
1513 <dd>Zero extend a constant to another type. The bit size of CST must be
1514 smaller or equal to the bit size of TYPE. Both types must be integers.</dd>
1516 <dt><b><tt>sext ( CST to TYPE )</tt></b></dt>
1517 <dd>Sign extend a constant to another type. The bit size of CST must be
1518 smaller or equal to the bit size of TYPE. Both types must be integers.</dd>
1520 <dt><b><tt>fptrunc ( CST to TYPE )</tt></b></dt>
1521 <dd>Truncate a floating point constant to another floating point type. The
1522 size of CST must be larger than the size of TYPE. Both types must be
1523 floating point.</dd>
1525 <dt><b><tt>fpext ( CST to TYPE )</tt></b></dt>
1526 <dd>Floating point extend a constant to another type. The size of CST must be
1527 smaller or equal to the size of TYPE. Both types must be floating point.</dd>
1529 <dt><b><tt>fptoui ( CST to TYPE )</tt></b></dt>
1530 <dd>Convert a floating point constant to the corresponding unsigned integer
1531 constant. TYPE must be a scalar or vector integer type. CST must be of scalar
1532 or vector floating point type. Both CST and TYPE must be scalars, or vectors
1533 of the same number of elements. If the value won't fit in the integer type,
1534 the results are undefined.</dd>
1536 <dt><b><tt>fptosi ( CST to TYPE )</tt></b></dt>
1537 <dd>Convert a floating point constant to the corresponding signed integer
1538 constant. TYPE must be a scalar or vector integer type. CST must be of scalar
1539 or vector floating point type. Both CST and TYPE must be scalars, or vectors
1540 of the same number of elements. If the value won't fit in the integer type,
1541 the results are undefined.</dd>
1543 <dt><b><tt>uitofp ( CST to TYPE )</tt></b></dt>
1544 <dd>Convert an unsigned integer constant to the corresponding floating point
1545 constant. TYPE must be a scalar or vector floating point type. CST must be of
1546 scalar or vector integer type. Both CST and TYPE must be scalars, or vectors
1547 of the same number of elements. If the value won't fit in the floating point
1548 type, the results are undefined.</dd>
1550 <dt><b><tt>sitofp ( CST to TYPE )</tt></b></dt>
1551 <dd>Convert a signed integer constant to the corresponding floating point
1552 constant. TYPE must be a scalar or vector floating point type. CST must be of
1553 scalar or vector integer type. Both CST and TYPE must be scalars, or vectors
1554 of the same number of elements. If the value won't fit in the floating point
1555 type, the results are undefined.</dd>
1557 <dt><b><tt>ptrtoint ( CST to TYPE )</tt></b></dt>
1558 <dd>Convert a pointer typed constant to the corresponding integer constant
1559 TYPE must be an integer type. CST must be of pointer type. The CST value is
1560 zero extended, truncated, or unchanged to make it fit in TYPE.</dd>
1562 <dt><b><tt>inttoptr ( CST to TYPE )</tt></b></dt>
1563 <dd>Convert a integer constant to a pointer constant. TYPE must be a
1564 pointer type. CST must be of integer type. The CST value is zero extended,
1565 truncated, or unchanged to make it fit in a pointer size. This one is
1566 <i>really</i> dangerous!</dd>
1568 <dt><b><tt>bitcast ( CST to TYPE )</tt></b></dt>
1569 <dd>Convert a constant, CST, to another TYPE. The size of CST and TYPE must be
1570 identical (same number of bits). The conversion is done as if the CST value
1571 was stored to memory and read back as TYPE. In other words, no bits change
1572 with this operator, just the type. This can be used for conversion of
1573 vector types to any other type, as long as they have the same bit width. For
1574 pointers it is only valid to cast to another pointer type.
1577 <dt><b><tt>getelementptr ( CSTPTR, IDX0, IDX1, ... )</tt></b></dt>
1579 <dd>Perform the <a href="#i_getelementptr">getelementptr operation</a> on
1580 constants. As with the <a href="#i_getelementptr">getelementptr</a>
1581 instruction, the index list may have zero or more indexes, which are required
1582 to make sense for the type of "CSTPTR".</dd>
1584 <dt><b><tt>select ( COND, VAL1, VAL2 )</tt></b></dt>
1586 <dd>Perform the <a href="#i_select">select operation</a> on
1589 <dt><b><tt>icmp COND ( VAL1, VAL2 )</tt></b></dt>
1590 <dd>Performs the <a href="#i_icmp">icmp operation</a> on constants.</dd>
1592 <dt><b><tt>fcmp COND ( VAL1, VAL2 )</tt></b></dt>
1593 <dd>Performs the <a href="#i_fcmp">fcmp operation</a> on constants.</dd>
1595 <dt><b><tt>extractelement ( VAL, IDX )</tt></b></dt>
1597 <dd>Perform the <a href="#i_extractelement">extractelement
1598 operation</a> on constants.
1600 <dt><b><tt>insertelement ( VAL, ELT, IDX )</tt></b></dt>
1602 <dd>Perform the <a href="#i_insertelement">insertelement
1603 operation</a> on constants.</dd>
1606 <dt><b><tt>shufflevector ( VEC1, VEC2, IDXMASK )</tt></b></dt>
1608 <dd>Perform the <a href="#i_shufflevector">shufflevector
1609 operation</a> on constants.</dd>
1611 <dt><b><tt>OPCODE ( LHS, RHS )</tt></b></dt>
1613 <dd>Perform the specified operation of the LHS and RHS constants. OPCODE may
1614 be any of the <a href="#binaryops">binary</a> or <a href="#bitwiseops">bitwise
1615 binary</a> operations. The constraints on operands are the same as those for
1616 the corresponding instruction (e.g. no bitwise operations on floating point
1617 values are allowed).</dd>
1621 <!-- *********************************************************************** -->
1622 <div class="doc_section"> <a name="othervalues">Other Values</a> </div>
1623 <!-- *********************************************************************** -->
1625 <!-- ======================================================================= -->
1626 <div class="doc_subsection">
1627 <a name="inlineasm">Inline Assembler Expressions</a>
1630 <div class="doc_text">
1633 LLVM supports inline assembler expressions (as opposed to <a href="#moduleasm">
1634 Module-Level Inline Assembly</a>) through the use of a special value. This
1635 value represents the inline assembler as a string (containing the instructions
1636 to emit), a list of operand constraints (stored as a string), and a flag that
1637 indicates whether or not the inline asm expression has side effects. An example
1638 inline assembler expression is:
1641 <div class="doc_code">
1643 i32 (i32) asm "bswap $0", "=r,r"
1648 Inline assembler expressions may <b>only</b> be used as the callee operand of
1649 a <a href="#i_call"><tt>call</tt> instruction</a>. Thus, typically we have:
1652 <div class="doc_code">
1654 %X = call i32 asm "<a href="#int_bswap">bswap</a> $0", "=r,r"(i32 %Y)
1659 Inline asms with side effects not visible in the constraint list must be marked
1660 as having side effects. This is done through the use of the
1661 '<tt>sideeffect</tt>' keyword, like so:
1664 <div class="doc_code">
1666 call void asm sideeffect "eieio", ""()
1670 <p>TODO: The format of the asm and constraints string still need to be
1671 documented here. Constraints on what can be done (e.g. duplication, moving, etc
1672 need to be documented).
1677 <!-- *********************************************************************** -->
1678 <div class="doc_section"> <a name="instref">Instruction Reference</a> </div>
1679 <!-- *********************************************************************** -->
1681 <div class="doc_text">
1683 <p>The LLVM instruction set consists of several different
1684 classifications of instructions: <a href="#terminators">terminator
1685 instructions</a>, <a href="#binaryops">binary instructions</a>,
1686 <a href="#bitwiseops">bitwise binary instructions</a>, <a
1687 href="#memoryops">memory instructions</a>, and <a href="#otherops">other
1688 instructions</a>.</p>
1692 <!-- ======================================================================= -->
1693 <div class="doc_subsection"> <a name="terminators">Terminator
1694 Instructions</a> </div>
1696 <div class="doc_text">
1698 <p>As mentioned <a href="#functionstructure">previously</a>, every
1699 basic block in a program ends with a "Terminator" instruction, which
1700 indicates which block should be executed after the current block is
1701 finished. These terminator instructions typically yield a '<tt>void</tt>'
1702 value: they produce control flow, not values (the one exception being
1703 the '<a href="#i_invoke"><tt>invoke</tt></a>' instruction).</p>
1704 <p>There are six different terminator instructions: the '<a
1705 href="#i_ret"><tt>ret</tt></a>' instruction, the '<a href="#i_br"><tt>br</tt></a>'
1706 instruction, the '<a href="#i_switch"><tt>switch</tt></a>' instruction,
1707 the '<a href="#i_invoke"><tt>invoke</tt></a>' instruction, the '<a
1708 href="#i_unwind"><tt>unwind</tt></a>' instruction, and the '<a
1709 href="#i_unreachable"><tt>unreachable</tt></a>' instruction.</p>
1713 <!-- _______________________________________________________________________ -->
1714 <div class="doc_subsubsection"> <a name="i_ret">'<tt>ret</tt>'
1715 Instruction</a> </div>
1716 <div class="doc_text">
1718 <pre> ret <type> <value> <i>; Return a value from a non-void function</i>
1719 ret void <i>; Return from void function</i>
1722 <p>The '<tt>ret</tt>' instruction is used to return control flow (and a
1723 value) from a function back to the caller.</p>
1724 <p>There are two forms of the '<tt>ret</tt>' instruction: one that
1725 returns a value and then causes control flow, and one that just causes
1726 control flow to occur.</p>
1728 <p>The '<tt>ret</tt>' instruction may return any '<a
1729 href="#t_firstclass">first class</a>' type. Notice that a function is
1730 not <a href="#wellformed">well formed</a> if there exists a '<tt>ret</tt>'
1731 instruction inside of the function that returns a value that does not
1732 match the return type of the function.</p>
1734 <p>When the '<tt>ret</tt>' instruction is executed, control flow
1735 returns back to the calling function's context. If the caller is a "<a
1736 href="#i_call"><tt>call</tt></a>" instruction, execution continues at
1737 the instruction after the call. If the caller was an "<a
1738 href="#i_invoke"><tt>invoke</tt></a>" instruction, execution continues
1739 at the beginning of the "normal" destination block. If the instruction
1740 returns a value, that value shall set the call or invoke instruction's
1743 <pre> ret i32 5 <i>; Return an integer value of 5</i>
1744 ret void <i>; Return from a void function</i>
1747 <!-- _______________________________________________________________________ -->
1748 <div class="doc_subsubsection"> <a name="i_br">'<tt>br</tt>' Instruction</a> </div>
1749 <div class="doc_text">
1751 <pre> br i1 <cond>, label <iftrue>, label <iffalse><br> br label <dest> <i>; Unconditional branch</i>
1754 <p>The '<tt>br</tt>' instruction is used to cause control flow to
1755 transfer to a different basic block in the current function. There are
1756 two forms of this instruction, corresponding to a conditional branch
1757 and an unconditional branch.</p>
1759 <p>The conditional branch form of the '<tt>br</tt>' instruction takes a
1760 single '<tt>i1</tt>' value and two '<tt>label</tt>' values. The
1761 unconditional form of the '<tt>br</tt>' instruction takes a single
1762 '<tt>label</tt>' value as a target.</p>
1764 <p>Upon execution of a conditional '<tt>br</tt>' instruction, the '<tt>i1</tt>'
1765 argument is evaluated. If the value is <tt>true</tt>, control flows
1766 to the '<tt>iftrue</tt>' <tt>label</tt> argument. If "cond" is <tt>false</tt>,
1767 control flows to the '<tt>iffalse</tt>' <tt>label</tt> argument.</p>
1769 <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
1770 href="#i_ret">ret</a> i32 1<br>IfUnequal:<br> <a href="#i_ret">ret</a> i32 0<br></pre>
1772 <!-- _______________________________________________________________________ -->
1773 <div class="doc_subsubsection">
1774 <a name="i_switch">'<tt>switch</tt>' Instruction</a>
1777 <div class="doc_text">
1781 switch <intty> <value>, label <defaultdest> [ <intty> <val>, label <dest> ... ]
1786 <p>The '<tt>switch</tt>' instruction is used to transfer control flow to one of
1787 several different places. It is a generalization of the '<tt>br</tt>'
1788 instruction, allowing a branch to occur to one of many possible
1794 <p>The '<tt>switch</tt>' instruction uses three parameters: an integer
1795 comparison value '<tt>value</tt>', a default '<tt>label</tt>' destination, and
1796 an array of pairs of comparison value constants and '<tt>label</tt>'s. The
1797 table is not allowed to contain duplicate constant entries.</p>
1801 <p>The <tt>switch</tt> instruction specifies a table of values and
1802 destinations. When the '<tt>switch</tt>' instruction is executed, this
1803 table is searched for the given value. If the value is found, control flow is
1804 transfered to the corresponding destination; otherwise, control flow is
1805 transfered to the default destination.</p>
1807 <h5>Implementation:</h5>
1809 <p>Depending on properties of the target machine and the particular
1810 <tt>switch</tt> instruction, this instruction may be code generated in different
1811 ways. For example, it could be generated as a series of chained conditional
1812 branches or with a lookup table.</p>
1817 <i>; Emulate a conditional br instruction</i>
1818 %Val = <a href="#i_zext">zext</a> i1 %value to i32
1819 switch i32 %Val, label %truedest [i32 0, label %falsedest ]
1821 <i>; Emulate an unconditional br instruction</i>
1822 switch i32 0, label %dest [ ]
1824 <i>; Implement a jump table:</i>
1825 switch i32 %val, label %otherwise [ i32 0, label %onzero
1827 i32 2, label %ontwo ]
1831 <!-- _______________________________________________________________________ -->
1832 <div class="doc_subsubsection">
1833 <a name="i_invoke">'<tt>invoke</tt>' Instruction</a>
1836 <div class="doc_text">
1841 <result> = invoke [<a href="#callingconv">cconv</a>] <ptr to function ty> %<function ptr val>(<function args>)
1842 to label <normal label> unwind label <exception label>
1847 <p>The '<tt>invoke</tt>' instruction causes control to transfer to a specified
1848 function, with the possibility of control flow transfer to either the
1849 '<tt>normal</tt>' label or the
1850 '<tt>exception</tt>' label. If the callee function returns with the
1851 "<tt><a href="#i_ret">ret</a></tt>" instruction, control flow will return to the
1852 "normal" label. If the callee (or any indirect callees) returns with the "<a
1853 href="#i_unwind"><tt>unwind</tt></a>" instruction, control is interrupted and
1854 continued at the dynamically nearest "exception" label.</p>
1858 <p>This instruction requires several arguments:</p>
1862 The optional "cconv" marker indicates which <a href="#callingconv">calling
1863 convention</a> the call should use. If none is specified, the call defaults
1864 to using C calling conventions.
1866 <li>'<tt>ptr to function ty</tt>': shall be the signature of the pointer to
1867 function value being invoked. In most cases, this is a direct function
1868 invocation, but indirect <tt>invoke</tt>s are just as possible, branching off
1869 an arbitrary pointer to function value.
1872 <li>'<tt>function ptr val</tt>': An LLVM value containing a pointer to a
1873 function to be invoked. </li>
1875 <li>'<tt>function args</tt>': argument list whose types match the function
1876 signature argument types. If the function signature indicates the function
1877 accepts a variable number of arguments, the extra arguments can be
1880 <li>'<tt>normal label</tt>': the label reached when the called function
1881 executes a '<tt><a href="#i_ret">ret</a></tt>' instruction. </li>
1883 <li>'<tt>exception label</tt>': the label reached when a callee returns with
1884 the <a href="#i_unwind"><tt>unwind</tt></a> instruction. </li>
1890 <p>This instruction is designed to operate as a standard '<tt><a
1891 href="#i_call">call</a></tt>' instruction in most regards. The primary
1892 difference is that it establishes an association with a label, which is used by
1893 the runtime library to unwind the stack.</p>
1895 <p>This instruction is used in languages with destructors to ensure that proper
1896 cleanup is performed in the case of either a <tt>longjmp</tt> or a thrown
1897 exception. Additionally, this is important for implementation of
1898 '<tt>catch</tt>' clauses in high-level languages that support them.</p>
1902 %retval = invoke i32 %Test(i32 15) to label %Continue
1903 unwind label %TestCleanup <i>; {i32}:retval set</i>
1904 %retval = invoke <a href="#callingconv">coldcc</a> i32 %Test(i32 15) to label %Continue
1905 unwind label %TestCleanup <i>; {i32}:retval set</i>
1910 <!-- _______________________________________________________________________ -->
1912 <div class="doc_subsubsection"> <a name="i_unwind">'<tt>unwind</tt>'
1913 Instruction</a> </div>
1915 <div class="doc_text">
1924 <p>The '<tt>unwind</tt>' instruction unwinds the stack, continuing control flow
1925 at the first callee in the dynamic call stack which used an <a
1926 href="#i_invoke"><tt>invoke</tt></a> instruction to perform the call. This is
1927 primarily used to implement exception handling.</p>
1931 <p>The '<tt>unwind</tt>' intrinsic causes execution of the current function to
1932 immediately halt. The dynamic call stack is then searched for the first <a
1933 href="#i_invoke"><tt>invoke</tt></a> instruction on the call stack. Once found,
1934 execution continues at the "exceptional" destination block specified by the
1935 <tt>invoke</tt> instruction. If there is no <tt>invoke</tt> instruction in the
1936 dynamic call chain, undefined behavior results.</p>
1939 <!-- _______________________________________________________________________ -->
1941 <div class="doc_subsubsection"> <a name="i_unreachable">'<tt>unreachable</tt>'
1942 Instruction</a> </div>
1944 <div class="doc_text">
1953 <p>The '<tt>unreachable</tt>' instruction has no defined semantics. This
1954 instruction is used to inform the optimizer that a particular portion of the
1955 code is not reachable. This can be used to indicate that the code after a
1956 no-return function cannot be reached, and other facts.</p>
1960 <p>The '<tt>unreachable</tt>' instruction has no defined semantics.</p>
1965 <!-- ======================================================================= -->
1966 <div class="doc_subsection"> <a name="binaryops">Binary Operations</a> </div>
1967 <div class="doc_text">
1968 <p>Binary operators are used to do most of the computation in a
1969 program. They require two operands, execute an operation on them, and
1970 produce a single value. The operands might represent
1971 multiple data, as is the case with the <a href="#t_vector">vector</a> data type.
1972 The result value of a binary operator is not
1973 necessarily the same type as its operands.</p>
1974 <p>There are several different binary operators:</p>
1976 <!-- _______________________________________________________________________ -->
1977 <div class="doc_subsubsection"> <a name="i_add">'<tt>add</tt>'
1978 Instruction</a> </div>
1979 <div class="doc_text">
1981 <pre> <result> = add <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1984 <p>The '<tt>add</tt>' instruction returns the sum of its two operands.</p>
1986 <p>The two arguments to the '<tt>add</tt>' instruction must be either <a
1987 href="#t_integer">integer</a> or <a href="#t_floating">floating point</a> values.
1988 This instruction can also take <a href="#t_vector">vector</a> versions of the values.
1989 Both arguments must have identical types.</p>
1991 <p>The value produced is the integer or floating point sum of the two
1994 <pre> <result> = add i32 4, %var <i>; yields {i32}:result = 4 + %var</i>
1997 <!-- _______________________________________________________________________ -->
1998 <div class="doc_subsubsection"> <a name="i_sub">'<tt>sub</tt>'
1999 Instruction</a> </div>
2000 <div class="doc_text">
2002 <pre> <result> = sub <ty> <var1>, <var2> <i>; yields {ty}:result</i>
2005 <p>The '<tt>sub</tt>' instruction returns the difference of its two
2007 <p>Note that the '<tt>sub</tt>' instruction is used to represent the '<tt>neg</tt>'
2008 instruction present in most other intermediate representations.</p>
2010 <p>The two arguments to the '<tt>sub</tt>' instruction must be either <a
2011 href="#t_integer">integer</a> or <a href="#t_floating">floating point</a>
2013 This instruction can also take <a href="#t_vector">vector</a> versions of the values.
2014 Both arguments must have identical types.</p>
2016 <p>The value produced is the integer or floating point difference of
2017 the two operands.</p>
2020 <result> = sub i32 4, %var <i>; yields {i32}:result = 4 - %var</i>
2021 <result> = sub i32 0, %val <i>; yields {i32}:result = -%var</i>
2024 <!-- _______________________________________________________________________ -->
2025 <div class="doc_subsubsection"> <a name="i_mul">'<tt>mul</tt>'
2026 Instruction</a> </div>
2027 <div class="doc_text">
2029 <pre> <result> = mul <ty> <var1>, <var2> <i>; yields {ty}:result</i>
2032 <p>The '<tt>mul</tt>' instruction returns the product of its two
2035 <p>The two arguments to the '<tt>mul</tt>' instruction must be either <a
2036 href="#t_integer">integer</a> or <a href="#t_floating">floating point</a>
2038 This instruction can also take <a href="#t_vector">vector</a> versions of the values.
2039 Both arguments must have identical types.</p>
2041 <p>The value produced is the integer or floating point product of the
2043 <p>Because the operands are the same width, the result of an integer
2044 multiplication is the same whether the operands should be deemed unsigned or
2047 <pre> <result> = mul i32 4, %var <i>; yields {i32}:result = 4 * %var</i>
2050 <!-- _______________________________________________________________________ -->
2051 <div class="doc_subsubsection"> <a name="i_udiv">'<tt>udiv</tt>' Instruction
2053 <div class="doc_text">
2055 <pre> <result> = udiv <ty> <var1>, <var2> <i>; yields {ty}:result</i>
2058 <p>The '<tt>udiv</tt>' instruction returns the quotient of its two
2061 <p>The two arguments to the '<tt>udiv</tt>' instruction must be
2062 <a href="#t_integer">integer</a> values. Both arguments must have identical
2063 types. This instruction can also take <a href="#t_vector">vector</a> versions
2064 of the values in which case the elements must be integers.</p>
2066 <p>The value produced is the unsigned integer quotient of the two operands. This
2067 instruction always performs an unsigned division operation, regardless of
2068 whether the arguments are unsigned or not.</p>
2070 <pre> <result> = udiv i32 4, %var <i>; yields {i32}:result = 4 / %var</i>
2073 <!-- _______________________________________________________________________ -->
2074 <div class="doc_subsubsection"> <a name="i_sdiv">'<tt>sdiv</tt>' Instruction
2076 <div class="doc_text">
2078 <pre> <result> = sdiv <ty> <var1>, <var2> <i>; yields {ty}:result</i>
2081 <p>The '<tt>sdiv</tt>' instruction returns the quotient of its two
2084 <p>The two arguments to the '<tt>sdiv</tt>' instruction must be
2085 <a href="#t_integer">integer</a> values. Both arguments must have identical
2086 types. This instruction can also take <a href="#t_vector">vector</a> versions
2087 of the values in which case the elements must be integers.</p>
2089 <p>The value produced is the signed integer quotient of the two operands. This
2090 instruction always performs a signed division operation, regardless of whether
2091 the arguments are signed or not.</p>
2093 <pre> <result> = sdiv i32 4, %var <i>; yields {i32}:result = 4 / %var</i>
2096 <!-- _______________________________________________________________________ -->
2097 <div class="doc_subsubsection"> <a name="i_fdiv">'<tt>fdiv</tt>'
2098 Instruction</a> </div>
2099 <div class="doc_text">
2101 <pre> <result> = fdiv <ty> <var1>, <var2> <i>; yields {ty}:result</i>
2104 <p>The '<tt>fdiv</tt>' instruction returns the quotient of its two
2107 <p>The two arguments to the '<tt>fdiv</tt>' instruction must be
2108 <a href="#t_floating">floating point</a> values. Both arguments must have
2109 identical types. This instruction can also take <a href="#t_vector">vector</a>
2110 versions of floating point values.</p>
2112 <p>The value produced is the floating point quotient of the two operands.</p>
2114 <pre> <result> = fdiv float 4.0, %var <i>; yields {float}:result = 4.0 / %var</i>
2117 <!-- _______________________________________________________________________ -->
2118 <div class="doc_subsubsection"> <a name="i_urem">'<tt>urem</tt>' Instruction</a>
2120 <div class="doc_text">
2122 <pre> <result> = urem <ty> <var1>, <var2> <i>; yields {ty}:result</i>
2125 <p>The '<tt>urem</tt>' instruction returns the remainder from the
2126 unsigned division of its two arguments.</p>
2128 <p>The two arguments to the '<tt>urem</tt>' instruction must be
2129 <a href="#t_integer">integer</a> values. Both arguments must have identical
2130 types. This instruction can also take <a href="#t_vector">vector</a> versions
2131 of the values in which case the elements must be integers.</p>
2133 <p>This instruction returns the unsigned integer <i>remainder</i> of a division.
2134 This instruction always performs an unsigned division to get the remainder,
2135 regardless of whether the arguments are unsigned or not.</p>
2137 <pre> <result> = urem i32 4, %var <i>; yields {i32}:result = 4 % %var</i>
2141 <!-- _______________________________________________________________________ -->
2142 <div class="doc_subsubsection"> <a name="i_srem">'<tt>srem</tt>'
2143 Instruction</a> </div>
2144 <div class="doc_text">
2146 <pre> <result> = srem <ty> <var1>, <var2> <i>; yields {ty}:result</i>
2149 <p>The '<tt>srem</tt>' instruction returns the remainder from the
2150 signed division of its two operands. This instruction can also take
2151 <a href="#t_vector">vector</a> versions of the values in which case
2152 the elements must be integers.</p>
2155 <p>The two arguments to the '<tt>srem</tt>' instruction must be
2156 <a href="#t_integer">integer</a> values. Both arguments must have identical
2159 <p>This instruction returns the <i>remainder</i> of a division (where the result
2160 has the same sign as the dividend, <tt>var1</tt>), not the <i>modulo</i>
2161 operator (where the result has the same sign as the divisor, <tt>var2</tt>) of
2162 a value. For more information about the difference, see <a
2163 href="http://mathforum.org/dr.math/problems/anne.4.28.99.html">The
2164 Math Forum</a>. For a table of how this is implemented in various languages,
2165 please see <a href="http://en.wikipedia.org/wiki/Modulo_operation">
2166 Wikipedia: modulo operation</a>.</p>
2168 <pre> <result> = srem i32 4, %var <i>; yields {i32}:result = 4 % %var</i>
2172 <!-- _______________________________________________________________________ -->
2173 <div class="doc_subsubsection"> <a name="i_frem">'<tt>frem</tt>'
2174 Instruction</a> </div>
2175 <div class="doc_text">
2177 <pre> <result> = frem <ty> <var1>, <var2> <i>; yields {ty}:result</i>
2180 <p>The '<tt>frem</tt>' instruction returns the remainder from the
2181 division of its two operands.</p>
2183 <p>The two arguments to the '<tt>frem</tt>' instruction must be
2184 <a href="#t_floating">floating point</a> values. Both arguments must have
2185 identical types. This instruction can also take <a href="#t_vector">vector</a>
2186 versions of floating point values.</p>
2188 <p>This instruction returns the <i>remainder</i> of a division.</p>
2190 <pre> <result> = frem float 4.0, %var <i>; yields {float}:result = 4.0 % %var</i>
2194 <!-- ======================================================================= -->
2195 <div class="doc_subsection"> <a name="bitwiseops">Bitwise Binary
2196 Operations</a> </div>
2197 <div class="doc_text">
2198 <p>Bitwise binary operators are used to do various forms of
2199 bit-twiddling in a program. They are generally very efficient
2200 instructions and can commonly be strength reduced from other
2201 instructions. They require two operands, execute an operation on them,
2202 and produce a single value. The resulting value of the bitwise binary
2203 operators is always the same type as its first operand.</p>
2206 <!-- _______________________________________________________________________ -->
2207 <div class="doc_subsubsection"> <a name="i_shl">'<tt>shl</tt>'
2208 Instruction</a> </div>
2209 <div class="doc_text">
2211 <pre> <result> = shl <ty> <var1>, <var2> <i>; yields {ty}:result</i>
2216 <p>The '<tt>shl</tt>' instruction returns the first operand shifted to
2217 the left a specified number of bits.</p>
2221 <p>Both arguments to the '<tt>shl</tt>' instruction must be the same <a
2222 href="#t_integer">integer</a> type.</p>
2226 <p>The value produced is <tt>var1</tt> * 2<sup><tt>var2</tt></sup>. If
2227 <tt>var2</tt> is (statically or dynamically) equal to or larger than the number
2228 of bits in <tt>var1</tt>, the result is undefined.</p>
2230 <h5>Example:</h5><pre>
2231 <result> = shl i32 4, %var <i>; yields {i32}: 4 << %var</i>
2232 <result> = shl i32 4, 2 <i>; yields {i32}: 16</i>
2233 <result> = shl i32 1, 10 <i>; yields {i32}: 1024</i>
2234 <result> = shl i32 1, 32 <i>; undefined</i>
2237 <!-- _______________________________________________________________________ -->
2238 <div class="doc_subsubsection"> <a name="i_lshr">'<tt>lshr</tt>'
2239 Instruction</a> </div>
2240 <div class="doc_text">
2242 <pre> <result> = lshr <ty> <var1>, <var2> <i>; yields {ty}:result</i>
2246 <p>The '<tt>lshr</tt>' instruction (logical shift right) returns the first
2247 operand shifted to the right a specified number of bits with zero fill.</p>
2250 <p>Both arguments to the '<tt>lshr</tt>' instruction must be the same
2251 <a href="#t_integer">integer</a> type.</p>
2255 <p>This instruction always performs a logical shift right operation. The most
2256 significant bits of the result will be filled with zero bits after the
2257 shift. If <tt>var2</tt> is (statically or dynamically) equal to or larger than
2258 the number of bits in <tt>var1</tt>, the result is undefined.</p>
2262 <result> = lshr i32 4, 1 <i>; yields {i32}:result = 2</i>
2263 <result> = lshr i32 4, 2 <i>; yields {i32}:result = 1</i>
2264 <result> = lshr i8 4, 3 <i>; yields {i8}:result = 0</i>
2265 <result> = lshr i8 -2, 1 <i>; yields {i8}:result = 0x7FFFFFFF </i>
2266 <result> = lshr i32 1, 32 <i>; undefined</i>
2270 <!-- _______________________________________________________________________ -->
2271 <div class="doc_subsubsection"> <a name="i_ashr">'<tt>ashr</tt>'
2272 Instruction</a> </div>
2273 <div class="doc_text">
2276 <pre> <result> = ashr <ty> <var1>, <var2> <i>; yields {ty}:result</i>
2280 <p>The '<tt>ashr</tt>' instruction (arithmetic shift right) returns the first
2281 operand shifted to the right a specified number of bits with sign extension.</p>
2284 <p>Both arguments to the '<tt>ashr</tt>' instruction must be the same
2285 <a href="#t_integer">integer</a> type.</p>
2288 <p>This instruction always performs an arithmetic shift right operation,
2289 The most significant bits of the result will be filled with the sign bit
2290 of <tt>var1</tt>. If <tt>var2</tt> is (statically or dynamically) equal to or
2291 larger than the number of bits in <tt>var1</tt>, the result is undefined.
2296 <result> = ashr i32 4, 1 <i>; yields {i32}:result = 2</i>
2297 <result> = ashr i32 4, 2 <i>; yields {i32}:result = 1</i>
2298 <result> = ashr i8 4, 3 <i>; yields {i8}:result = 0</i>
2299 <result> = ashr i8 -2, 1 <i>; yields {i8}:result = -1</i>
2300 <result> = ashr i32 1, 32 <i>; undefined</i>
2304 <!-- _______________________________________________________________________ -->
2305 <div class="doc_subsubsection"> <a name="i_and">'<tt>and</tt>'
2306 Instruction</a> </div>
2307 <div class="doc_text">
2309 <pre> <result> = and <ty> <var1>, <var2> <i>; yields {ty}:result</i>
2312 <p>The '<tt>and</tt>' instruction returns the bitwise logical and of
2313 its two operands.</p>
2315 <p>The two arguments to the '<tt>and</tt>' instruction must be <a
2316 href="#t_integer">integer</a> values. Both arguments must have
2317 identical types.</p>
2319 <p>The truth table used for the '<tt>and</tt>' instruction is:</p>
2321 <div style="align: center">
2322 <table border="1" cellspacing="0" cellpadding="4">
2353 <pre> <result> = and i32 4, %var <i>; yields {i32}:result = 4 & %var</i>
2354 <result> = and i32 15, 40 <i>; yields {i32}:result = 8</i>
2355 <result> = and i32 4, 8 <i>; yields {i32}:result = 0</i>
2358 <!-- _______________________________________________________________________ -->
2359 <div class="doc_subsubsection"> <a name="i_or">'<tt>or</tt>' Instruction</a> </div>
2360 <div class="doc_text">
2362 <pre> <result> = or <ty> <var1>, <var2> <i>; yields {ty}:result</i>
2365 <p>The '<tt>or</tt>' instruction returns the bitwise logical inclusive
2366 or of its two operands.</p>
2368 <p>The two arguments to the '<tt>or</tt>' instruction must be <a
2369 href="#t_integer">integer</a> values. Both arguments must have
2370 identical types.</p>
2372 <p>The truth table used for the '<tt>or</tt>' instruction is:</p>
2374 <div style="align: center">
2375 <table border="1" cellspacing="0" cellpadding="4">
2406 <pre> <result> = or i32 4, %var <i>; yields {i32}:result = 4 | %var</i>
2407 <result> = or i32 15, 40 <i>; yields {i32}:result = 47</i>
2408 <result> = or i32 4, 8 <i>; yields {i32}:result = 12</i>
2411 <!-- _______________________________________________________________________ -->
2412 <div class="doc_subsubsection"> <a name="i_xor">'<tt>xor</tt>'
2413 Instruction</a> </div>
2414 <div class="doc_text">
2416 <pre> <result> = xor <ty> <var1>, <var2> <i>; yields {ty}:result</i>
2419 <p>The '<tt>xor</tt>' instruction returns the bitwise logical exclusive
2420 or of its two operands. The <tt>xor</tt> is used to implement the
2421 "one's complement" operation, which is the "~" operator in C.</p>
2423 <p>The two arguments to the '<tt>xor</tt>' instruction must be <a
2424 href="#t_integer">integer</a> values. Both arguments must have
2425 identical types.</p>
2427 <p>The truth table used for the '<tt>xor</tt>' instruction is:</p>
2429 <div style="align: center">
2430 <table border="1" cellspacing="0" cellpadding="4">
2462 <pre> <result> = xor i32 4, %var <i>; yields {i32}:result = 4 ^ %var</i>
2463 <result> = xor i32 15, 40 <i>; yields {i32}:result = 39</i>
2464 <result> = xor i32 4, 8 <i>; yields {i32}:result = 12</i>
2465 <result> = xor i32 %V, -1 <i>; yields {i32}:result = ~%V</i>
2469 <!-- ======================================================================= -->
2470 <div class="doc_subsection">
2471 <a name="vectorops">Vector Operations</a>
2474 <div class="doc_text">
2476 <p>LLVM supports several instructions to represent vector operations in a
2477 target-independent manner. These instructions cover the element-access and
2478 vector-specific operations needed to process vectors effectively. While LLVM
2479 does directly support these vector operations, many sophisticated algorithms
2480 will want to use target-specific intrinsics to take full advantage of a specific
2485 <!-- _______________________________________________________________________ -->
2486 <div class="doc_subsubsection">
2487 <a name="i_extractelement">'<tt>extractelement</tt>' Instruction</a>
2490 <div class="doc_text">
2495 <result> = extractelement <n x <ty>> <val>, i32 <idx> <i>; yields <ty></i>
2501 The '<tt>extractelement</tt>' instruction extracts a single scalar
2502 element from a vector at a specified index.
2509 The first operand of an '<tt>extractelement</tt>' instruction is a
2510 value of <a href="#t_vector">vector</a> type. The second operand is
2511 an index indicating the position from which to extract the element.
2512 The index may be a variable.</p>
2517 The result is a scalar of the same type as the element type of
2518 <tt>val</tt>. Its value is the value at position <tt>idx</tt> of
2519 <tt>val</tt>. If <tt>idx</tt> exceeds the length of <tt>val</tt>, the
2520 results are undefined.
2526 %result = extractelement <4 x i32> %vec, i32 0 <i>; yields i32</i>
2531 <!-- _______________________________________________________________________ -->
2532 <div class="doc_subsubsection">
2533 <a name="i_insertelement">'<tt>insertelement</tt>' Instruction</a>
2536 <div class="doc_text">
2541 <result> = insertelement <n x <ty>> <val>, <ty> <elt>, i32 <idx> <i>; yields <n x <ty>></i>
2547 The '<tt>insertelement</tt>' instruction inserts a scalar
2548 element into a vector at a specified index.
2555 The first operand of an '<tt>insertelement</tt>' instruction is a
2556 value of <a href="#t_vector">vector</a> type. The second operand is a
2557 scalar value whose type must equal the element type of the first
2558 operand. The third operand is an index indicating the position at
2559 which to insert the value. The index may be a variable.</p>
2564 The result is a vector of the same type as <tt>val</tt>. Its
2565 element values are those of <tt>val</tt> except at position
2566 <tt>idx</tt>, where it gets the value <tt>elt</tt>. If <tt>idx</tt>
2567 exceeds the length of <tt>val</tt>, the results are undefined.
2573 %result = insertelement <4 x i32> %vec, i32 1, i32 0 <i>; yields <4 x i32></i>
2577 <!-- _______________________________________________________________________ -->
2578 <div class="doc_subsubsection">
2579 <a name="i_shufflevector">'<tt>shufflevector</tt>' Instruction</a>
2582 <div class="doc_text">
2587 <result> = shufflevector <n x <ty>> <v1>, <n x <ty>> <v2>, <n x i32> <mask> <i>; yields <n x <ty>></i>
2593 The '<tt>shufflevector</tt>' instruction constructs a permutation of elements
2594 from two input vectors, returning a vector of the same type.
2600 The first two operands of a '<tt>shufflevector</tt>' instruction are vectors
2601 with types that match each other and types that match the result of the
2602 instruction. The third argument is a shuffle mask, which has the same number
2603 of elements as the other vector type, but whose element type is always 'i32'.
2607 The shuffle mask operand is required to be a constant vector with either
2608 constant integer or undef values.
2614 The elements of the two input vectors are numbered from left to right across
2615 both of the vectors. The shuffle mask operand specifies, for each element of
2616 the result vector, which element of the two input registers the result element
2617 gets. The element selector may be undef (meaning "don't care") and the second
2618 operand may be undef if performing a shuffle from only one vector.
2624 %result = shufflevector <4 x i32> %v1, <4 x i32> %v2,
2625 <4 x i32> <i32 0, i32 4, i32 1, i32 5> <i>; yields <4 x i32></i>
2626 %result = shufflevector <4 x i32> %v1, <4 x i32> undef,
2627 <4 x i32> <i32 0, i32 1, i32 2, i32 3> <i>; yields <4 x i32></i> - Identity shuffle.
2632 <!-- ======================================================================= -->
2633 <div class="doc_subsection">
2634 <a name="memoryops">Memory Access and Addressing Operations</a>
2637 <div class="doc_text">
2639 <p>A key design point of an SSA-based representation is how it
2640 represents memory. In LLVM, no memory locations are in SSA form, which
2641 makes things very simple. This section describes how to read, write,
2642 allocate, and free memory in LLVM.</p>
2646 <!-- _______________________________________________________________________ -->
2647 <div class="doc_subsubsection">
2648 <a name="i_malloc">'<tt>malloc</tt>' Instruction</a>
2651 <div class="doc_text">
2656 <result> = malloc <type>[, i32 <NumElements>][, align <alignment>] <i>; yields {type*}:result</i>
2661 <p>The '<tt>malloc</tt>' instruction allocates memory from the system
2662 heap and returns a pointer to it.</p>
2666 <p>The '<tt>malloc</tt>' instruction allocates
2667 <tt>sizeof(<type>)*NumElements</tt>
2668 bytes of memory from the operating system and returns a pointer of the
2669 appropriate type to the program. If "NumElements" is specified, it is the
2670 number of elements allocated. If an alignment is specified, the value result
2671 of the allocation is guaranteed to be aligned to at least that boundary. If
2672 not specified, or if zero, the target can choose to align the allocation on any
2673 convenient boundary.</p>
2675 <p>'<tt>type</tt>' must be a sized type.</p>
2679 <p>Memory is allocated using the system "<tt>malloc</tt>" function, and
2680 a pointer is returned.</p>
2685 %array = malloc [4 x i8 ] <i>; yields {[%4 x i8]*}:array</i>
2687 %size = <a href="#i_add">add</a> i32 2, 2 <i>; yields {i32}:size = i32 4</i>
2688 %array1 = malloc i8, i32 4 <i>; yields {i8*}:array1</i>
2689 %array2 = malloc [12 x i8], i32 %size <i>; yields {[12 x i8]*}:array2</i>
2690 %array3 = malloc i32, i32 4, align 1024 <i>; yields {i32*}:array3</i>
2691 %array4 = malloc i32, align 1024 <i>; yields {i32*}:array4</i>
2695 <!-- _______________________________________________________________________ -->
2696 <div class="doc_subsubsection">
2697 <a name="i_free">'<tt>free</tt>' Instruction</a>
2700 <div class="doc_text">
2705 free <type> <value> <i>; yields {void}</i>
2710 <p>The '<tt>free</tt>' instruction returns memory back to the unused
2711 memory heap to be reallocated in the future.</p>
2715 <p>'<tt>value</tt>' shall be a pointer value that points to a value
2716 that was allocated with the '<tt><a href="#i_malloc">malloc</a></tt>'
2721 <p>Access to the memory pointed to by the pointer is no longer defined
2722 after this instruction executes.</p>
2727 %array = <a href="#i_malloc">malloc</a> [4 x i8] <i>; yields {[4 x i8]*}:array</i>
2728 free [4 x i8]* %array
2732 <!-- _______________________________________________________________________ -->
2733 <div class="doc_subsubsection">
2734 <a name="i_alloca">'<tt>alloca</tt>' Instruction</a>
2737 <div class="doc_text">
2742 <result> = alloca <type>[, i32 <NumElements>][, align <alignment>] <i>; yields {type*}:result</i>
2747 <p>The '<tt>alloca</tt>' instruction allocates memory on the stack frame of the
2748 currently executing function, to be automatically released when this function
2749 returns to its caller.</p>
2753 <p>The '<tt>alloca</tt>' instruction allocates <tt>sizeof(<type>)*NumElements</tt>
2754 bytes of memory on the runtime stack, returning a pointer of the
2755 appropriate type to the program. If "NumElements" is specified, it is the
2756 number of elements allocated. If an alignment is specified, the value result
2757 of the allocation is guaranteed to be aligned to at least that boundary. If
2758 not specified, or if zero, the target can choose to align the allocation on any
2759 convenient boundary.</p>
2761 <p>'<tt>type</tt>' may be any sized type.</p>
2765 <p>Memory is allocated; a pointer is returned. '<tt>alloca</tt>'d
2766 memory is automatically released when the function returns. The '<tt>alloca</tt>'
2767 instruction is commonly used to represent automatic variables that must
2768 have an address available. When the function returns (either with the <tt><a
2769 href="#i_ret">ret</a></tt> or <tt><a href="#i_unwind">unwind</a></tt>
2770 instructions), the memory is reclaimed.</p>
2775 %ptr = alloca i32 <i>; yields {i32*}:ptr</i>
2776 %ptr = alloca i32, i32 4 <i>; yields {i32*}:ptr</i>
2777 %ptr = alloca i32, i32 4, align 1024 <i>; yields {i32*}:ptr</i>
2778 %ptr = alloca i32, align 1024 <i>; yields {i32*}:ptr</i>
2782 <!-- _______________________________________________________________________ -->
2783 <div class="doc_subsubsection"> <a name="i_load">'<tt>load</tt>'
2784 Instruction</a> </div>
2785 <div class="doc_text">
2787 <pre> <result> = load <ty>* <pointer>[, align <alignment>]<br> <result> = volatile load <ty>* <pointer>[, align <alignment>]<br></pre>
2789 <p>The '<tt>load</tt>' instruction is used to read from memory.</p>
2791 <p>The argument to the '<tt>load</tt>' instruction specifies the memory
2792 address from which to load. The pointer must point to a <a
2793 href="#t_firstclass">first class</a> type. If the <tt>load</tt> is
2794 marked as <tt>volatile</tt>, then the optimizer is not allowed to modify
2795 the number or order of execution of this <tt>load</tt> with other
2796 volatile <tt>load</tt> and <tt><a href="#i_store">store</a></tt>
2799 <p>The location of memory pointed to is loaded.</p>
2801 <pre> %ptr = <a href="#i_alloca">alloca</a> i32 <i>; yields {i32*}:ptr</i>
2803 href="#i_store">store</a> i32 3, i32* %ptr <i>; yields {void}</i>
2804 %val = load i32* %ptr <i>; yields {i32}:val = i32 3</i>
2807 <!-- _______________________________________________________________________ -->
2808 <div class="doc_subsubsection"> <a name="i_store">'<tt>store</tt>'
2809 Instruction</a> </div>
2810 <div class="doc_text">
2812 <pre> store <ty> <value>, <ty>* <pointer>[, align <alignment>] <i>; yields {void}</i>
2813 volatile store <ty> <value>, <ty>* <pointer>[, align <alignment>] <i>; yields {void}</i>
2816 <p>The '<tt>store</tt>' instruction is used to write to memory.</p>
2818 <p>There are two arguments to the '<tt>store</tt>' instruction: a value
2819 to store and an address at which to store it. The type of the '<tt><pointer></tt>'
2820 operand must be a pointer to the type of the '<tt><value></tt>'
2821 operand. If the <tt>store</tt> is marked as <tt>volatile</tt>, then the
2822 optimizer is not allowed to modify the number or order of execution of
2823 this <tt>store</tt> with other volatile <tt>load</tt> and <tt><a
2824 href="#i_store">store</a></tt> instructions.</p>
2826 <p>The contents of memory are updated to contain '<tt><value></tt>'
2827 at the location specified by the '<tt><pointer></tt>' operand.</p>
2829 <pre> %ptr = <a href="#i_alloca">alloca</a> i32 <i>; yields {i32*}:ptr</i>
2830 store i32 3, i32* %ptr <i>; yields {void}</i>
2831 %val = <a href="#i_load">load</a> i32* %ptr <i>; yields {i32}:val = i32 3</i>
2835 <!-- _______________________________________________________________________ -->
2836 <div class="doc_subsubsection">
2837 <a name="i_getelementptr">'<tt>getelementptr</tt>' Instruction</a>
2840 <div class="doc_text">
2843 <result> = getelementptr <ty>* <ptrval>{, <ty> <idx>}*
2849 The '<tt>getelementptr</tt>' instruction is used to get the address of a
2850 subelement of an aggregate data structure.</p>
2854 <p>This instruction takes a list of integer operands that indicate what
2855 elements of the aggregate object to index to. The actual types of the arguments
2856 provided depend on the type of the first pointer argument. The
2857 '<tt>getelementptr</tt>' instruction is used to index down through the type
2858 levels of a structure or to a specific index in an array. When indexing into a
2859 structure, only <tt>i32</tt> integer constants are allowed. When indexing
2860 into an array or pointer, only integers of 32 or 64 bits are allowed, and will
2861 be sign extended to 64-bit values.</p>
2863 <p>For example, let's consider a C code fragment and how it gets
2864 compiled to LLVM:</p>
2866 <div class="doc_code">
2879 int *foo(struct ST *s) {
2880 return &s[1].Z.B[5][13];
2885 <p>The LLVM code generated by the GCC frontend is:</p>
2887 <div class="doc_code">
2889 %RT = type { i8 , [10 x [20 x i32]], i8 }
2890 %ST = type { i32, double, %RT }
2892 define i32* %foo(%ST* %s) {
2894 %reg = getelementptr %ST* %s, i32 1, i32 2, i32 1, i32 5, i32 13
2902 <p>The index types specified for the '<tt>getelementptr</tt>' instruction depend
2903 on the pointer type that is being indexed into. <a href="#t_pointer">Pointer</a>
2904 and <a href="#t_array">array</a> types can use a 32-bit or 64-bit
2905 <a href="#t_integer">integer</a> type but the value will always be sign extended
2906 to 64-bits. <a href="#t_struct">Structure</a> types require <tt>i32</tt>
2907 <b>constants</b>.</p>
2909 <p>In the example above, the first index is indexing into the '<tt>%ST*</tt>'
2910 type, which is a pointer, yielding a '<tt>%ST</tt>' = '<tt>{ i32, double, %RT
2911 }</tt>' type, a structure. The second index indexes into the third element of
2912 the structure, yielding a '<tt>%RT</tt>' = '<tt>{ i8 , [10 x [20 x i32]],
2913 i8 }</tt>' type, another structure. The third index indexes into the second
2914 element of the structure, yielding a '<tt>[10 x [20 x i32]]</tt>' type, an
2915 array. The two dimensions of the array are subscripted into, yielding an
2916 '<tt>i32</tt>' type. The '<tt>getelementptr</tt>' instruction returns a pointer
2917 to this element, thus computing a value of '<tt>i32*</tt>' type.</p>
2919 <p>Note that it is perfectly legal to index partially through a
2920 structure, returning a pointer to an inner element. Because of this,
2921 the LLVM code for the given testcase is equivalent to:</p>
2924 define i32* %foo(%ST* %s) {
2925 %t1 = getelementptr %ST* %s, i32 1 <i>; yields %ST*:%t1</i>
2926 %t2 = getelementptr %ST* %t1, i32 0, i32 2 <i>; yields %RT*:%t2</i>
2927 %t3 = getelementptr %RT* %t2, i32 0, i32 1 <i>; yields [10 x [20 x i32]]*:%t3</i>
2928 %t4 = getelementptr [10 x [20 x i32]]* %t3, i32 0, i32 5 <i>; yields [20 x i32]*:%t4</i>
2929 %t5 = getelementptr [20 x i32]* %t4, i32 0, i32 13 <i>; yields i32*:%t5</i>
2934 <p>Note that it is undefined to access an array out of bounds: array and
2935 pointer indexes must always be within the defined bounds of the array type.
2936 The one exception for this rules is zero length arrays. These arrays are
2937 defined to be accessible as variable length arrays, which requires access
2938 beyond the zero'th element.</p>
2940 <p>The getelementptr instruction is often confusing. For some more insight
2941 into how it works, see <a href="GetElementPtr.html">the getelementptr
2947 <i>; yields [12 x i8]*:aptr</i>
2948 %aptr = getelementptr {i32, [12 x i8]}* %sptr, i64 0, i32 1
2952 <!-- ======================================================================= -->
2953 <div class="doc_subsection"> <a name="convertops">Conversion Operations</a>
2955 <div class="doc_text">
2956 <p>The instructions in this category are the conversion instructions (casting)
2957 which all take a single operand and a type. They perform various bit conversions
2961 <!-- _______________________________________________________________________ -->
2962 <div class="doc_subsubsection">
2963 <a name="i_trunc">'<tt>trunc .. to</tt>' Instruction</a>
2965 <div class="doc_text">
2969 <result> = trunc <ty> <value> to <ty2> <i>; yields ty2</i>
2974 The '<tt>trunc</tt>' instruction truncates its operand to the type <tt>ty2</tt>.
2979 The '<tt>trunc</tt>' instruction takes a <tt>value</tt> to trunc, which must
2980 be an <a href="#t_integer">integer</a> type, and a type that specifies the size
2981 and type of the result, which must be an <a href="#t_integer">integer</a>
2982 type. The bit size of <tt>value</tt> must be larger than the bit size of
2983 <tt>ty2</tt>. Equal sized types are not allowed.</p>
2987 The '<tt>trunc</tt>' instruction truncates the high order bits in <tt>value</tt>
2988 and converts the remaining bits to <tt>ty2</tt>. Since the source size must be
2989 larger than the destination size, <tt>trunc</tt> cannot be a <i>no-op cast</i>.
2990 It will always truncate bits.</p>
2994 %X = trunc i32 257 to i8 <i>; yields i8:1</i>
2995 %Y = trunc i32 123 to i1 <i>; yields i1:true</i>
2996 %Y = trunc i32 122 to i1 <i>; yields i1:false</i>
3000 <!-- _______________________________________________________________________ -->
3001 <div class="doc_subsubsection">
3002 <a name="i_zext">'<tt>zext .. to</tt>' Instruction</a>
3004 <div class="doc_text">
3008 <result> = zext <ty> <value> to <ty2> <i>; yields ty2</i>
3012 <p>The '<tt>zext</tt>' instruction zero extends its operand to type
3017 <p>The '<tt>zext</tt>' instruction takes a value to cast, which must be of
3018 <a href="#t_integer">integer</a> type, and a type to cast it to, which must
3019 also be of <a href="#t_integer">integer</a> type. The bit size of the
3020 <tt>value</tt> must be smaller than the bit size of the destination type,
3024 <p>The <tt>zext</tt> fills the high order bits of the <tt>value</tt> with zero
3025 bits until it reaches the size of the destination type, <tt>ty2</tt>.</p>
3027 <p>When zero extending from i1, the result will always be either 0 or 1.</p>
3031 %X = zext i32 257 to i64 <i>; yields i64:257</i>
3032 %Y = zext i1 true to i32 <i>; yields i32:1</i>
3036 <!-- _______________________________________________________________________ -->
3037 <div class="doc_subsubsection">
3038 <a name="i_sext">'<tt>sext .. to</tt>' Instruction</a>
3040 <div class="doc_text">
3044 <result> = sext <ty> <value> to <ty2> <i>; yields ty2</i>
3048 <p>The '<tt>sext</tt>' sign extends <tt>value</tt> to the type <tt>ty2</tt>.</p>
3052 The '<tt>sext</tt>' instruction takes a value to cast, which must be of
3053 <a href="#t_integer">integer</a> type, and a type to cast it to, which must
3054 also be of <a href="#t_integer">integer</a> type. The bit size of the
3055 <tt>value</tt> must be smaller than the bit size of the destination type,
3060 The '<tt>sext</tt>' instruction performs a sign extension by copying the sign
3061 bit (highest order bit) of the <tt>value</tt> until it reaches the bit size of
3062 the type <tt>ty2</tt>.</p>
3064 <p>When sign extending from i1, the extension always results in -1 or 0.</p>
3068 %X = sext i8 -1 to i16 <i>; yields i16 :65535</i>
3069 %Y = sext i1 true to i32 <i>; yields i32:-1</i>
3073 <!-- _______________________________________________________________________ -->
3074 <div class="doc_subsubsection">
3075 <a name="i_fptrunc">'<tt>fptrunc .. to</tt>' Instruction</a>
3078 <div class="doc_text">
3083 <result> = fptrunc <ty> <value> to <ty2> <i>; yields ty2</i>
3087 <p>The '<tt>fptrunc</tt>' instruction truncates <tt>value</tt> to type
3092 <p>The '<tt>fptrunc</tt>' instruction takes a <a href="#t_floating">floating
3093 point</a> value to cast and a <a href="#t_floating">floating point</a> type to
3094 cast it to. The size of <tt>value</tt> must be larger than the size of
3095 <tt>ty2</tt>. This implies that <tt>fptrunc</tt> cannot be used to make a
3096 <i>no-op cast</i>.</p>
3099 <p> The '<tt>fptrunc</tt>' instruction truncates a <tt>value</tt> from a larger
3100 <a href="#t_floating">floating point</a> type to a smaller
3101 <a href="#t_floating">floating point</a> type. If the value cannot fit within
3102 the destination type, <tt>ty2</tt>, then the results are undefined.</p>
3106 %X = fptrunc double 123.0 to float <i>; yields float:123.0</i>
3107 %Y = fptrunc double 1.0E+300 to float <i>; yields undefined</i>
3111 <!-- _______________________________________________________________________ -->
3112 <div class="doc_subsubsection">
3113 <a name="i_fpext">'<tt>fpext .. to</tt>' Instruction</a>
3115 <div class="doc_text">
3119 <result> = fpext <ty> <value> to <ty2> <i>; yields ty2</i>
3123 <p>The '<tt>fpext</tt>' extends a floating point <tt>value</tt> to a larger
3124 floating point value.</p>
3127 <p>The '<tt>fpext</tt>' instruction takes a
3128 <a href="#t_floating">floating point</a> <tt>value</tt> to cast,
3129 and a <a href="#t_floating">floating point</a> type to cast it to. The source
3130 type must be smaller than the destination type.</p>
3133 <p>The '<tt>fpext</tt>' instruction extends the <tt>value</tt> from a smaller
3134 <a href="#t_floating">floating point</a> type to a larger
3135 <a href="#t_floating">floating point</a> type. The <tt>fpext</tt> cannot be
3136 used to make a <i>no-op cast</i> because it always changes bits. Use
3137 <tt>bitcast</tt> to make a <i>no-op cast</i> for a floating point cast.</p>
3141 %X = fpext float 3.1415 to double <i>; yields double:3.1415</i>
3142 %Y = fpext float 1.0 to float <i>; yields float:1.0 (no-op)</i>
3146 <!-- _______________________________________________________________________ -->
3147 <div class="doc_subsubsection">
3148 <a name="i_fptoui">'<tt>fptoui .. to</tt>' Instruction</a>
3150 <div class="doc_text">
3154 <result> = fptoui <ty> <value> to <ty2> <i>; yields ty2</i>
3158 <p>The '<tt>fptoui</tt>' converts a floating point <tt>value</tt> to its
3159 unsigned integer equivalent of type <tt>ty2</tt>.
3163 <p>The '<tt>fptoui</tt>' instruction takes a value to cast, which must be a
3164 scalar or vector <a href="#t_floating">floating point</a> value, and a type
3165 to cast it to <tt>ty2</tt>, which must be an <a href="#t_integer">integer</a>
3166 type. If <tt>ty</tt> is a vector floating point type, <tt>ty2</tt> must be a
3167 vector integer type with the same number of elements as <tt>ty</tt></p>
3170 <p> The '<tt>fptoui</tt>' instruction converts its
3171 <a href="#t_floating">floating point</a> operand into the nearest (rounding
3172 towards zero) unsigned integer value. If the value cannot fit in <tt>ty2</tt>,
3173 the results are undefined.</p>
3177 %X = fptoui double 123.0 to i32 <i>; yields i32:123</i>
3178 %Y = fptoui float 1.0E+300 to i1 <i>; yields undefined:1</i>
3179 %X = fptoui float 1.04E+17 to i8 <i>; yields undefined:1</i>
3183 <!-- _______________________________________________________________________ -->
3184 <div class="doc_subsubsection">
3185 <a name="i_fptosi">'<tt>fptosi .. to</tt>' Instruction</a>
3187 <div class="doc_text">
3191 <result> = fptosi <ty> <value> to <ty2> <i>; yields ty2</i>
3195 <p>The '<tt>fptosi</tt>' instruction converts
3196 <a href="#t_floating">floating point</a> <tt>value</tt> to type <tt>ty2</tt>.
3200 <p> The '<tt>fptosi</tt>' instruction takes a value to cast, which must be a
3201 scalar or vector <a href="#t_floating">floating point</a> value, and a type
3202 to cast it to <tt>ty2</tt>, which must be an <a href="#t_integer">integer</a>
3203 type. If <tt>ty</tt> is a vector floating point type, <tt>ty2</tt> must be a
3204 vector integer type with the same number of elements as <tt>ty</tt></p>
3207 <p>The '<tt>fptosi</tt>' instruction converts its
3208 <a href="#t_floating">floating point</a> operand into the nearest (rounding
3209 towards zero) signed integer value. If the value cannot fit in <tt>ty2</tt>,
3210 the results are undefined.</p>
3214 %X = fptosi double -123.0 to i32 <i>; yields i32:-123</i>
3215 %Y = fptosi float 1.0E-247 to i1 <i>; yields undefined:1</i>
3216 %X = fptosi float 1.04E+17 to i8 <i>; yields undefined:1</i>
3220 <!-- _______________________________________________________________________ -->
3221 <div class="doc_subsubsection">
3222 <a name="i_uitofp">'<tt>uitofp .. to</tt>' Instruction</a>
3224 <div class="doc_text">
3228 <result> = uitofp <ty> <value> to <ty2> <i>; yields ty2</i>
3232 <p>The '<tt>uitofp</tt>' instruction regards <tt>value</tt> as an unsigned
3233 integer and converts that value to the <tt>ty2</tt> type.</p>
3236 <p>The '<tt>uitofp</tt>' instruction takes a value to cast, which must be a
3237 scalar or vector <a href="#t_integer">integer</a> value, and a type to cast it
3238 to <tt>ty2</tt>, which must be an <a href="#t_floating">floating point</a>
3239 type. If <tt>ty</tt> is a vector integer type, <tt>ty2</tt> must be a vector
3240 floating point type with the same number of elements as <tt>ty</tt></p>
3243 <p>The '<tt>uitofp</tt>' instruction interprets its operand as an unsigned
3244 integer quantity and converts it to the corresponding floating point value. If
3245 the value cannot fit in the floating point value, the results are undefined.</p>
3249 %X = uitofp i32 257 to float <i>; yields float:257.0</i>
3250 %Y = uitofp i8 -1 to double <i>; yields double:255.0</i>
3254 <!-- _______________________________________________________________________ -->
3255 <div class="doc_subsubsection">
3256 <a name="i_sitofp">'<tt>sitofp .. to</tt>' Instruction</a>
3258 <div class="doc_text">
3262 <result> = sitofp <ty> <value> to <ty2> <i>; yields ty2</i>
3266 <p>The '<tt>sitofp</tt>' instruction regards <tt>value</tt> as a signed
3267 integer and converts that value to the <tt>ty2</tt> type.</p>
3270 <p>The '<tt>sitofp</tt>' instruction takes a value to cast, which must be a
3271 scalar or vector <a href="#t_integer">integer</a> value, and a type to cast it
3272 to <tt>ty2</tt>, which must be an <a href="#t_floating">floating point</a>
3273 type. If <tt>ty</tt> is a vector integer type, <tt>ty2</tt> must be a vector
3274 floating point type with the same number of elements as <tt>ty</tt></p>
3277 <p>The '<tt>sitofp</tt>' instruction interprets its operand as a signed
3278 integer quantity and converts it to the corresponding floating point value. If
3279 the value cannot fit in the floating point value, the results are undefined.</p>
3283 %X = sitofp i32 257 to float <i>; yields float:257.0</i>
3284 %Y = sitofp i8 -1 to double <i>; yields double:-1.0</i>
3288 <!-- _______________________________________________________________________ -->
3289 <div class="doc_subsubsection">
3290 <a name="i_ptrtoint">'<tt>ptrtoint .. to</tt>' Instruction</a>
3292 <div class="doc_text">
3296 <result> = ptrtoint <ty> <value> to <ty2> <i>; yields ty2</i>
3300 <p>The '<tt>ptrtoint</tt>' instruction converts the pointer <tt>value</tt> to
3301 the integer type <tt>ty2</tt>.</p>
3304 <p>The '<tt>ptrtoint</tt>' instruction takes a <tt>value</tt> to cast, which
3305 must be a <a href="#t_pointer">pointer</a> value, and a type to cast it to
3306 <tt>ty2</tt>, which must be an <a href="#t_integer">integer</a> type.
3309 <p>The '<tt>ptrtoint</tt>' instruction converts <tt>value</tt> to integer type
3310 <tt>ty2</tt> by interpreting the pointer value as an integer and either
3311 truncating or zero extending that value to the size of the integer type. If
3312 <tt>value</tt> is smaller than <tt>ty2</tt> then a zero extension is done. If
3313 <tt>value</tt> is larger than <tt>ty2</tt> then a truncation is done. If they
3314 are the same size, then nothing is done (<i>no-op cast</i>) other than a type
3319 %X = ptrtoint i32* %X to i8 <i>; yields truncation on 32-bit architecture</i>
3320 %Y = ptrtoint i32* %x to i64 <i>; yields zero extension on 32-bit architecture</i>
3324 <!-- _______________________________________________________________________ -->
3325 <div class="doc_subsubsection">
3326 <a name="i_inttoptr">'<tt>inttoptr .. to</tt>' Instruction</a>
3328 <div class="doc_text">
3332 <result> = inttoptr <ty> <value> to <ty2> <i>; yields ty2</i>
3336 <p>The '<tt>inttoptr</tt>' instruction converts an integer <tt>value</tt> to
3337 a pointer type, <tt>ty2</tt>.</p>
3340 <p>The '<tt>inttoptr</tt>' instruction takes an <a href="#t_integer">integer</a>
3341 value to cast, and a type to cast it to, which must be a
3342 <a href="#t_pointer">pointer</a> type.
3345 <p>The '<tt>inttoptr</tt>' instruction converts <tt>value</tt> to type
3346 <tt>ty2</tt> by applying either a zero extension or a truncation depending on
3347 the size of the integer <tt>value</tt>. If <tt>value</tt> is larger than the
3348 size of a pointer then a truncation is done. If <tt>value</tt> is smaller than
3349 the size of a pointer then a zero extension is done. If they are the same size,
3350 nothing is done (<i>no-op cast</i>).</p>
3354 %X = inttoptr i32 255 to i32* <i>; yields zero extension on 64-bit architecture</i>
3355 %X = inttoptr i32 255 to i32* <i>; yields no-op on 32-bit architecture</i>
3356 %Y = inttoptr i64 0 to i32* <i>; yields truncation on 32-bit architecture</i>
3360 <!-- _______________________________________________________________________ -->
3361 <div class="doc_subsubsection">
3362 <a name="i_bitcast">'<tt>bitcast .. to</tt>' Instruction</a>
3364 <div class="doc_text">
3368 <result> = bitcast <ty> <value> to <ty2> <i>; yields ty2</i>
3372 <p>The '<tt>bitcast</tt>' instruction converts <tt>value</tt> to type
3373 <tt>ty2</tt> without changing any bits.</p>
3376 <p>The '<tt>bitcast</tt>' instruction takes a value to cast, which must be
3377 a first class value, and a type to cast it to, which must also be a <a
3378 href="#t_firstclass">first class</a> type. The bit sizes of <tt>value</tt>
3379 and the destination type, <tt>ty2</tt>, must be identical. If the source
3380 type is a pointer, the destination type must also be a pointer.</p>
3383 <p>The '<tt>bitcast</tt>' instruction converts <tt>value</tt> to type
3384 <tt>ty2</tt>. It is always a <i>no-op cast</i> because no bits change with
3385 this conversion. The conversion is done as if the <tt>value</tt> had been
3386 stored to memory and read back as type <tt>ty2</tt>. Pointer types may only be
3387 converted to other pointer types with this instruction. To convert pointers to
3388 other types, use the <a href="#i_inttoptr">inttoptr</a> or
3389 <a href="#i_ptrtoint">ptrtoint</a> instructions first.</p>
3393 %X = bitcast i8 255 to i8 <i>; yields i8 :-1</i>
3394 %Y = bitcast i32* %x to sint* <i>; yields sint*:%x</i>
3395 %Z = bitcast <2xint> %V to i64; <i>; yields i64: %V</i>
3399 <!-- ======================================================================= -->
3400 <div class="doc_subsection"> <a name="otherops">Other Operations</a> </div>
3401 <div class="doc_text">
3402 <p>The instructions in this category are the "miscellaneous"
3403 instructions, which defy better classification.</p>
3406 <!-- _______________________________________________________________________ -->
3407 <div class="doc_subsubsection"><a name="i_icmp">'<tt>icmp</tt>' Instruction</a>
3409 <div class="doc_text">
3411 <pre> <result> = icmp <cond> <ty> <var1>, <var2> <i>; yields {i1}:result</i>
3414 <p>The '<tt>icmp</tt>' instruction returns a boolean value based on comparison
3415 of its two integer operands.</p>
3417 <p>The '<tt>icmp</tt>' instruction takes three operands. The first operand is
3418 the condition code indicating the kind of comparison to perform. It is not
3419 a value, just a keyword. The possible condition code are:
3421 <li><tt>eq</tt>: equal</li>
3422 <li><tt>ne</tt>: not equal </li>
3423 <li><tt>ugt</tt>: unsigned greater than</li>
3424 <li><tt>uge</tt>: unsigned greater or equal</li>
3425 <li><tt>ult</tt>: unsigned less than</li>
3426 <li><tt>ule</tt>: unsigned less or equal</li>
3427 <li><tt>sgt</tt>: signed greater than</li>
3428 <li><tt>sge</tt>: signed greater or equal</li>
3429 <li><tt>slt</tt>: signed less than</li>
3430 <li><tt>sle</tt>: signed less or equal</li>
3432 <p>The remaining two arguments must be <a href="#t_integer">integer</a> or
3433 <a href="#t_pointer">pointer</a> typed. They must also be identical types.</p>
3435 <p>The '<tt>icmp</tt>' compares <tt>var1</tt> and <tt>var2</tt> according to
3436 the condition code given as <tt>cond</tt>. The comparison performed always
3437 yields a <a href="#t_primitive">i1</a> result, as follows:
3439 <li><tt>eq</tt>: yields <tt>true</tt> if the operands are equal,
3440 <tt>false</tt> otherwise. No sign interpretation is necessary or performed.
3442 <li><tt>ne</tt>: yields <tt>true</tt> if the operands are unequal,
3443 <tt>false</tt> otherwise. No sign interpretation is necessary or performed.
3444 <li><tt>ugt</tt>: interprets the operands as unsigned values and yields
3445 <tt>true</tt> if <tt>var1</tt> is greater than <tt>var2</tt>.</li>
3446 <li><tt>uge</tt>: interprets the operands as unsigned values and yields
3447 <tt>true</tt> if <tt>var1</tt> is greater than or equal to <tt>var2</tt>.</li>
3448 <li><tt>ult</tt>: interprets the operands as unsigned values and yields
3449 <tt>true</tt> if <tt>var1</tt> is less than <tt>var2</tt>.</li>
3450 <li><tt>ule</tt>: interprets the operands as unsigned values and yields
3451 <tt>true</tt> if <tt>var1</tt> is less than or equal to <tt>var2</tt>.</li>
3452 <li><tt>sgt</tt>: interprets the operands as signed values and yields
3453 <tt>true</tt> if <tt>var1</tt> is greater than <tt>var2</tt>.</li>
3454 <li><tt>sge</tt>: interprets the operands as signed values and yields
3455 <tt>true</tt> if <tt>var1</tt> is greater than or equal to <tt>var2</tt>.</li>
3456 <li><tt>slt</tt>: interprets the operands as signed values and yields
3457 <tt>true</tt> if <tt>var1</tt> is less than <tt>var2</tt>.</li>
3458 <li><tt>sle</tt>: interprets the operands as signed values and yields
3459 <tt>true</tt> if <tt>var1</tt> is less than or equal to <tt>var2</tt>.</li>
3461 <p>If the operands are <a href="#t_pointer">pointer</a> typed, the pointer
3462 values are compared as if they were integers.</p>
3465 <pre> <result> = icmp eq i32 4, 5 <i>; yields: result=false</i>
3466 <result> = icmp ne float* %X, %X <i>; yields: result=false</i>
3467 <result> = icmp ult i16 4, 5 <i>; yields: result=true</i>
3468 <result> = icmp sgt i16 4, 5 <i>; yields: result=false</i>
3469 <result> = icmp ule i16 -4, 5 <i>; yields: result=false</i>
3470 <result> = icmp sge i16 4, 5 <i>; yields: result=false</i>
3474 <!-- _______________________________________________________________________ -->
3475 <div class="doc_subsubsection"><a name="i_fcmp">'<tt>fcmp</tt>' Instruction</a>
3477 <div class="doc_text">
3479 <pre> <result> = fcmp <cond> <ty> <var1>, <var2> <i>; yields {i1}:result</i>
3482 <p>The '<tt>fcmp</tt>' instruction returns a boolean value based on comparison
3483 of its floating point operands.</p>
3485 <p>The '<tt>fcmp</tt>' instruction takes three operands. The first operand is
3486 the condition code indicating the kind of comparison to perform. It is not
3487 a value, just a keyword. The possible condition code are:
3489 <li><tt>false</tt>: no comparison, always returns false</li>
3490 <li><tt>oeq</tt>: ordered and equal</li>
3491 <li><tt>ogt</tt>: ordered and greater than </li>
3492 <li><tt>oge</tt>: ordered and greater than or equal</li>
3493 <li><tt>olt</tt>: ordered and less than </li>
3494 <li><tt>ole</tt>: ordered and less than or equal</li>
3495 <li><tt>one</tt>: ordered and not equal</li>
3496 <li><tt>ord</tt>: ordered (no nans)</li>
3497 <li><tt>ueq</tt>: unordered or equal</li>
3498 <li><tt>ugt</tt>: unordered or greater than </li>
3499 <li><tt>uge</tt>: unordered or greater than or equal</li>
3500 <li><tt>ult</tt>: unordered or less than </li>
3501 <li><tt>ule</tt>: unordered or less than or equal</li>
3502 <li><tt>une</tt>: unordered or not equal</li>
3503 <li><tt>uno</tt>: unordered (either nans)</li>
3504 <li><tt>true</tt>: no comparison, always returns true</li>
3506 <p><i>Ordered</i> means that neither operand is a QNAN while
3507 <i>unordered</i> means that either operand may be a QNAN.</p>
3508 <p>The <tt>val1</tt> and <tt>val2</tt> arguments must be
3509 <a href="#t_floating">floating point</a> typed. They must have identical
3512 <p>The '<tt>fcmp</tt>' compares <tt>var1</tt> and <tt>var2</tt> according to
3513 the condition code given as <tt>cond</tt>. The comparison performed always
3514 yields a <a href="#t_primitive">i1</a> result, as follows:
3516 <li><tt>false</tt>: always yields <tt>false</tt>, regardless of operands.</li>
3517 <li><tt>oeq</tt>: yields <tt>true</tt> if both operands are not a QNAN and
3518 <tt>var1</tt> is equal to <tt>var2</tt>.</li>
3519 <li><tt>ogt</tt>: yields <tt>true</tt> if both operands are not a QNAN and
3520 <tt>var1</tt> is greather than <tt>var2</tt>.</li>
3521 <li><tt>oge</tt>: yields <tt>true</tt> if both operands are not a QNAN and
3522 <tt>var1</tt> is greater than or equal to <tt>var2</tt>.</li>
3523 <li><tt>olt</tt>: yields <tt>true</tt> if both operands are not a QNAN and
3524 <tt>var1</tt> is less than <tt>var2</tt>.</li>
3525 <li><tt>ole</tt>: yields <tt>true</tt> if both operands are not a QNAN and
3526 <tt>var1</tt> is less than or equal to <tt>var2</tt>.</li>
3527 <li><tt>one</tt>: yields <tt>true</tt> if both operands are not a QNAN and
3528 <tt>var1</tt> is not equal to <tt>var2</tt>.</li>
3529 <li><tt>ord</tt>: yields <tt>true</tt> if both operands are not a QNAN.</li>
3530 <li><tt>ueq</tt>: yields <tt>true</tt> if either operand is a QNAN or
3531 <tt>var1</tt> is equal to <tt>var2</tt>.</li>
3532 <li><tt>ugt</tt>: yields <tt>true</tt> if either operand is a QNAN or
3533 <tt>var1</tt> is greater than <tt>var2</tt>.</li>
3534 <li><tt>uge</tt>: yields <tt>true</tt> if either operand is a QNAN or
3535 <tt>var1</tt> is greater than or equal to <tt>var2</tt>.</li>
3536 <li><tt>ult</tt>: yields <tt>true</tt> if either operand is a QNAN or
3537 <tt>var1</tt> is less than <tt>var2</tt>.</li>
3538 <li><tt>ule</tt>: yields <tt>true</tt> if either operand is a QNAN or
3539 <tt>var1</tt> is less than or equal to <tt>var2</tt>.</li>
3540 <li><tt>une</tt>: yields <tt>true</tt> if either operand is a QNAN or
3541 <tt>var1</tt> is not equal to <tt>var2</tt>.</li>
3542 <li><tt>uno</tt>: yields <tt>true</tt> if either operand is a QNAN.</li>
3543 <li><tt>true</tt>: always yields <tt>true</tt>, regardless of operands.</li>
3547 <pre> <result> = fcmp oeq float 4.0, 5.0 <i>; yields: result=false</i>
3548 <result> = icmp one float 4.0, 5.0 <i>; yields: result=true</i>
3549 <result> = icmp olt float 4.0, 5.0 <i>; yields: result=true</i>
3550 <result> = icmp ueq double 1.0, 2.0 <i>; yields: result=false</i>
3554 <!-- _______________________________________________________________________ -->
3555 <div class="doc_subsubsection"> <a name="i_phi">'<tt>phi</tt>'
3556 Instruction</a> </div>
3557 <div class="doc_text">
3559 <pre> <result> = phi <ty> [ <val0>, <label0>], ...<br></pre>
3561 <p>The '<tt>phi</tt>' instruction is used to implement the φ node in
3562 the SSA graph representing the function.</p>
3564 <p>The type of the incoming values is specified with the first type
3565 field. After this, the '<tt>phi</tt>' instruction takes a list of pairs
3566 as arguments, with one pair for each predecessor basic block of the
3567 current block. Only values of <a href="#t_firstclass">first class</a>
3568 type may be used as the value arguments to the PHI node. Only labels
3569 may be used as the label arguments.</p>
3570 <p>There must be no non-phi instructions between the start of a basic
3571 block and the PHI instructions: i.e. PHI instructions must be first in
3574 <p>At runtime, the '<tt>phi</tt>' instruction logically takes on the value
3575 specified by the pair corresponding to the predecessor basic block that executed
3576 just prior to the current block.</p>
3578 <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>
3581 <!-- _______________________________________________________________________ -->
3582 <div class="doc_subsubsection">
3583 <a name="i_select">'<tt>select</tt>' Instruction</a>
3586 <div class="doc_text">
3591 <result> = select i1 <cond>, <ty> <val1>, <ty> <val2> <i>; yields ty</i>
3597 The '<tt>select</tt>' instruction is used to choose one value based on a
3598 condition, without branching.
3605 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.
3611 If the boolean condition evaluates to true, the instruction returns the first
3612 value argument; otherwise, it returns the second value argument.
3618 %X = select i1 true, i8 17, i8 42 <i>; yields i8:17</i>
3623 <!-- _______________________________________________________________________ -->
3624 <div class="doc_subsubsection">
3625 <a name="i_call">'<tt>call</tt>' Instruction</a>
3628 <div class="doc_text">
3632 <result> = [tail] call [<a href="#callingconv">cconv</a>] <ty> [<fnty>*] <fnptrval>(<param list>)
3637 <p>The '<tt>call</tt>' instruction represents a simple function call.</p>
3641 <p>This instruction requires several arguments:</p>
3645 <p>The optional "tail" marker indicates whether the callee function accesses
3646 any allocas or varargs in the caller. If the "tail" marker is present, the
3647 function call is eligible for tail call optimization. Note that calls may
3648 be marked "tail" even if they do not occur before a <a
3649 href="#i_ret"><tt>ret</tt></a> instruction.
3652 <p>The optional "cconv" marker indicates which <a href="#callingconv">calling
3653 convention</a> the call should use. If none is specified, the call defaults
3654 to using C calling conventions.
3657 <p>'<tt>ty</tt>': the type of the call instruction itself which is also
3658 the type of the return value. Functions that return no value are marked
3659 <tt><a href="#t_void">void</a></tt>.</p>
3662 <p>'<tt>fnty</tt>': shall be the signature of the pointer to function
3663 value being invoked. The argument types must match the types implied by
3664 this signature. This type can be omitted if the function is not varargs
3665 and if the function type does not return a pointer to a function.</p>
3668 <p>'<tt>fnptrval</tt>': An LLVM value containing a pointer to a function to
3669 be invoked. In most cases, this is a direct function invocation, but
3670 indirect <tt>call</tt>s are just as possible, calling an arbitrary pointer
3671 to function value.</p>
3674 <p>'<tt>function args</tt>': argument list whose types match the
3675 function signature argument types. All arguments must be of
3676 <a href="#t_firstclass">first class</a> type. If the function signature
3677 indicates the function accepts a variable number of arguments, the extra
3678 arguments can be specified.</p>
3684 <p>The '<tt>call</tt>' instruction is used to cause control flow to
3685 transfer to a specified function, with its incoming arguments bound to
3686 the specified values. Upon a '<tt><a href="#i_ret">ret</a></tt>'
3687 instruction in the called function, control flow continues with the
3688 instruction after the function call, and the return value of the
3689 function is bound to the result argument. This is a simpler case of
3690 the <a href="#i_invoke">invoke</a> instruction.</p>
3695 %retval = call i32 @test(i32 %argc)
3696 call i32 (i8 *, ...)* @printf(i8 * %msg, i32 12, i8 42);
3697 %X = tail call i32 @foo()
3698 %Y = tail call <a href="#callingconv">fastcc</a> i32 @foo()
3699 %Z = call void %foo(i8 97 signext)
3704 <!-- _______________________________________________________________________ -->
3705 <div class="doc_subsubsection">
3706 <a name="i_va_arg">'<tt>va_arg</tt>' Instruction</a>
3709 <div class="doc_text">
3714 <resultval> = va_arg <va_list*> <arglist>, <argty>
3719 <p>The '<tt>va_arg</tt>' instruction is used to access arguments passed through
3720 the "variable argument" area of a function call. It is used to implement the
3721 <tt>va_arg</tt> macro in C.</p>
3725 <p>This instruction takes a <tt>va_list*</tt> value and the type of
3726 the argument. It returns a value of the specified argument type and
3727 increments the <tt>va_list</tt> to point to the next argument. The
3728 actual type of <tt>va_list</tt> is target specific.</p>
3732 <p>The '<tt>va_arg</tt>' instruction loads an argument of the specified
3733 type from the specified <tt>va_list</tt> and causes the
3734 <tt>va_list</tt> to point to the next argument. For more information,
3735 see the variable argument handling <a href="#int_varargs">Intrinsic
3738 <p>It is legal for this instruction to be called in a function which does not
3739 take a variable number of arguments, for example, the <tt>vfprintf</tt>
3742 <p><tt>va_arg</tt> is an LLVM instruction instead of an <a
3743 href="#intrinsics">intrinsic function</a> because it takes a type as an
3748 <p>See the <a href="#int_varargs">variable argument processing</a> section.</p>
3752 <!-- *********************************************************************** -->
3753 <div class="doc_section"> <a name="intrinsics">Intrinsic Functions</a> </div>
3754 <!-- *********************************************************************** -->
3756 <div class="doc_text">
3758 <p>LLVM supports the notion of an "intrinsic function". These functions have
3759 well known names and semantics and are required to follow certain restrictions.
3760 Overall, these intrinsics represent an extension mechanism for the LLVM
3761 language that does not require changing all of the transformations in LLVM when
3762 adding to the language (or the bitcode reader/writer, the parser, etc...).</p>
3764 <p>Intrinsic function names must all start with an "<tt>llvm.</tt>" prefix. This
3765 prefix is reserved in LLVM for intrinsic names; thus, function names may not
3766 begin with this prefix. Intrinsic functions must always be external functions:
3767 you cannot define the body of intrinsic functions. Intrinsic functions may
3768 only be used in call or invoke instructions: it is illegal to take the address
3769 of an intrinsic function. Additionally, because intrinsic functions are part
3770 of the LLVM language, it is required if any are added that they be documented
3773 <p>Some intrinsic functions can be overloaded, i.e., the intrinsic represents
3774 a family of functions that perform the same operation but on different data
3775 types. Because LLVM can represent over 8 million different integer types,
3776 overloading is used commonly to allow an intrinsic function to operate on any
3777 integer type. One or more of the argument types or the result type can be
3778 overloaded to accept any integer type. Argument types may also be defined as
3779 exactly matching a previous argument's type or the result type. This allows an
3780 intrinsic function which accepts multiple arguments, but needs all of them to
3781 be of the same type, to only be overloaded with respect to a single argument or
3784 <p>Overloaded intrinsics will have the names of its overloaded argument types
3785 encoded into its function name, each preceded by a period. Only those types
3786 which are overloaded result in a name suffix. Arguments whose type is matched
3787 against another type do not. For example, the <tt>llvm.ctpop</tt> function can
3788 take an integer of any width and returns an integer of exactly the same integer
3789 width. This leads to a family of functions such as
3790 <tt>i8 @llvm.ctpop.i8(i8 %val)</tt> and <tt>i29 @llvm.ctpop.i29(i29 %val)</tt>.
3791 Only one type, the return type, is overloaded, and only one type suffix is
3792 required. Because the argument's type is matched against the return type, it
3793 does not require its own name suffix.</p>
3795 <p>To learn how to add an intrinsic function, please see the
3796 <a href="ExtendingLLVM.html">Extending LLVM Guide</a>.
3801 <!-- ======================================================================= -->
3802 <div class="doc_subsection">
3803 <a name="int_varargs">Variable Argument Handling Intrinsics</a>
3806 <div class="doc_text">
3808 <p>Variable argument support is defined in LLVM with the <a
3809 href="#i_va_arg"><tt>va_arg</tt></a> instruction and these three
3810 intrinsic functions. These functions are related to the similarly
3811 named macros defined in the <tt><stdarg.h></tt> header file.</p>
3813 <p>All of these functions operate on arguments that use a
3814 target-specific value type "<tt>va_list</tt>". The LLVM assembly
3815 language reference manual does not define what this type is, so all
3816 transformations should be prepared to handle these functions regardless of
3819 <p>This example shows how the <a href="#i_va_arg"><tt>va_arg</tt></a>
3820 instruction and the variable argument handling intrinsic functions are
3823 <div class="doc_code">
3825 define i32 @test(i32 %X, ...) {
3826 ; Initialize variable argument processing
3828 %ap2 = bitcast i8** %ap to i8*
3829 call void @llvm.va_start(i8* %ap2)
3831 ; Read a single integer argument
3832 %tmp = va_arg i8** %ap, i32
3834 ; Demonstrate usage of llvm.va_copy and llvm.va_end
3836 %aq2 = bitcast i8** %aq to i8*
3837 call void @llvm.va_copy(i8* %aq2, i8* %ap2)
3838 call void @llvm.va_end(i8* %aq2)
3840 ; Stop processing of arguments.
3841 call void @llvm.va_end(i8* %ap2)
3845 declare void @llvm.va_start(i8*)
3846 declare void @llvm.va_copy(i8*, i8*)
3847 declare void @llvm.va_end(i8*)
3853 <!-- _______________________________________________________________________ -->
3854 <div class="doc_subsubsection">
3855 <a name="int_va_start">'<tt>llvm.va_start</tt>' Intrinsic</a>
3859 <div class="doc_text">
3861 <pre> declare void %llvm.va_start(i8* <arglist>)<br></pre>
3863 <P>The '<tt>llvm.va_start</tt>' intrinsic initializes
3864 <tt>*<arglist></tt> for subsequent use by <tt><a
3865 href="#i_va_arg">va_arg</a></tt>.</p>
3869 <P>The argument is a pointer to a <tt>va_list</tt> element to initialize.</p>
3873 <P>The '<tt>llvm.va_start</tt>' intrinsic works just like the <tt>va_start</tt>
3874 macro available in C. In a target-dependent way, it initializes the
3875 <tt>va_list</tt> element to which the argument points, so that the next call to
3876 <tt>va_arg</tt> will produce the first variable argument passed to the function.
3877 Unlike the C <tt>va_start</tt> macro, this intrinsic does not need to know the
3878 last argument of the function as the compiler can figure that out.</p>
3882 <!-- _______________________________________________________________________ -->
3883 <div class="doc_subsubsection">
3884 <a name="int_va_end">'<tt>llvm.va_end</tt>' Intrinsic</a>
3887 <div class="doc_text">
3889 <pre> declare void @llvm.va_end(i8* <arglist>)<br></pre>
3892 <p>The '<tt>llvm.va_end</tt>' intrinsic destroys <tt>*<arglist></tt>,
3893 which has been initialized previously with <tt><a href="#int_va_start">llvm.va_start</a></tt>
3894 or <tt><a href="#i_va_copy">llvm.va_copy</a></tt>.</p>
3898 <p>The argument is a pointer to a <tt>va_list</tt> to destroy.</p>
3902 <p>The '<tt>llvm.va_end</tt>' intrinsic works just like the <tt>va_end</tt>
3903 macro available in C. In a target-dependent way, it destroys the
3904 <tt>va_list</tt> element to which the argument points. Calls to <a
3905 href="#int_va_start"><tt>llvm.va_start</tt></a> and <a href="#int_va_copy">
3906 <tt>llvm.va_copy</tt></a> must be matched exactly with calls to
3907 <tt>llvm.va_end</tt>.</p>
3911 <!-- _______________________________________________________________________ -->
3912 <div class="doc_subsubsection">
3913 <a name="int_va_copy">'<tt>llvm.va_copy</tt>' Intrinsic</a>
3916 <div class="doc_text">
3921 declare void @llvm.va_copy(i8* <destarglist>, i8* <srcarglist>)
3926 <p>The '<tt>llvm.va_copy</tt>' intrinsic copies the current argument position
3927 from the source argument list to the destination argument list.</p>
3931 <p>The first argument is a pointer to a <tt>va_list</tt> element to initialize.
3932 The second argument is a pointer to a <tt>va_list</tt> element to copy from.</p>
3937 <p>The '<tt>llvm.va_copy</tt>' intrinsic works just like the <tt>va_copy</tt>
3938 macro available in C. In a target-dependent way, it copies the source
3939 <tt>va_list</tt> element into the destination <tt>va_list</tt> element. This
3940 intrinsic is necessary because the <tt><a href="#int_va_start">
3941 llvm.va_start</a></tt> intrinsic may be arbitrarily complex and require, for
3942 example, memory allocation.</p>
3946 <!-- ======================================================================= -->
3947 <div class="doc_subsection">
3948 <a name="int_gc">Accurate Garbage Collection Intrinsics</a>
3951 <div class="doc_text">
3954 LLVM support for <a href="GarbageCollection.html">Accurate Garbage
3955 Collection</a> requires the implementation and generation of these intrinsics.
3956 These intrinsics allow identification of <a href="#int_gcroot">GC roots on the
3957 stack</a>, as well as garbage collector implementations that require <a
3958 href="#int_gcread">read</a> and <a href="#int_gcwrite">write</a> barriers.
3959 Front-ends for type-safe garbage collected languages should generate these
3960 intrinsics to make use of the LLVM garbage collectors. For more details, see <a
3961 href="GarbageCollection.html">Accurate Garbage Collection with LLVM</a>.
3965 <!-- _______________________________________________________________________ -->
3966 <div class="doc_subsubsection">
3967 <a name="int_gcroot">'<tt>llvm.gcroot</tt>' Intrinsic</a>
3970 <div class="doc_text">
3975 declare void @llvm.gcroot(i8** %ptrloc, i8* %metadata)
3980 <p>The '<tt>llvm.gcroot</tt>' intrinsic declares the existence of a GC root to
3981 the code generator, and allows some metadata to be associated with it.</p>
3985 <p>The first argument specifies the address of a stack object that contains the
3986 root pointer. The second pointer (which must be either a constant or a global
3987 value address) contains the meta-data to be associated with the root.</p>
3991 <p>At runtime, a call to this intrinsics stores a null pointer into the "ptrloc"
3992 location. At compile-time, the code generator generates information to allow
3993 the runtime to find the pointer at GC safe points.
3999 <!-- _______________________________________________________________________ -->
4000 <div class="doc_subsubsection">
4001 <a name="int_gcread">'<tt>llvm.gcread</tt>' Intrinsic</a>
4004 <div class="doc_text">
4009 declare i8* @llvm.gcread(i8* %ObjPtr, i8** %Ptr)
4014 <p>The '<tt>llvm.gcread</tt>' intrinsic identifies reads of references from heap
4015 locations, allowing garbage collector implementations that require read
4020 <p>The second argument is the address to read from, which should be an address
4021 allocated from the garbage collector. The first object is a pointer to the
4022 start of the referenced object, if needed by the language runtime (otherwise
4027 <p>The '<tt>llvm.gcread</tt>' intrinsic has the same semantics as a load
4028 instruction, but may be replaced with substantially more complex code by the
4029 garbage collector runtime, as needed.</p>
4034 <!-- _______________________________________________________________________ -->
4035 <div class="doc_subsubsection">
4036 <a name="int_gcwrite">'<tt>llvm.gcwrite</tt>' Intrinsic</a>
4039 <div class="doc_text">
4044 declare void @llvm.gcwrite(i8* %P1, i8* %Obj, i8** %P2)
4049 <p>The '<tt>llvm.gcwrite</tt>' intrinsic identifies writes of references to heap
4050 locations, allowing garbage collector implementations that require write
4051 barriers (such as generational or reference counting collectors).</p>
4055 <p>The first argument is the reference to store, the second is the start of the
4056 object to store it to, and the third is the address of the field of Obj to
4057 store to. If the runtime does not require a pointer to the object, Obj may be
4062 <p>The '<tt>llvm.gcwrite</tt>' intrinsic has the same semantics as a store
4063 instruction, but may be replaced with substantially more complex code by the
4064 garbage collector runtime, as needed.</p>
4070 <!-- ======================================================================= -->
4071 <div class="doc_subsection">
4072 <a name="int_codegen">Code Generator Intrinsics</a>
4075 <div class="doc_text">
4077 These intrinsics are provided by LLVM to expose special features that may only
4078 be implemented with code generator support.
4083 <!-- _______________________________________________________________________ -->
4084 <div class="doc_subsubsection">
4085 <a name="int_returnaddress">'<tt>llvm.returnaddress</tt>' Intrinsic</a>
4088 <div class="doc_text">
4092 declare i8 *@llvm.returnaddress(i32 <level>)
4098 The '<tt>llvm.returnaddress</tt>' intrinsic attempts to compute a
4099 target-specific value indicating the return address of the current function
4100 or one of its callers.
4106 The argument to this intrinsic indicates which function to return the address
4107 for. Zero indicates the calling function, one indicates its caller, etc. The
4108 argument is <b>required</b> to be a constant integer value.
4114 The '<tt>llvm.returnaddress</tt>' intrinsic either returns a pointer indicating
4115 the return address of the specified call frame, or zero if it cannot be
4116 identified. The value returned by this intrinsic is likely to be incorrect or 0
4117 for arguments other than zero, so it should only be used for debugging purposes.
4121 Note that calling this intrinsic does not prevent function inlining or other
4122 aggressive transformations, so the value returned may not be that of the obvious
4123 source-language caller.
4128 <!-- _______________________________________________________________________ -->
4129 <div class="doc_subsubsection">
4130 <a name="int_frameaddress">'<tt>llvm.frameaddress</tt>' Intrinsic</a>
4133 <div class="doc_text">
4137 declare i8 *@llvm.frameaddress(i32 <level>)
4143 The '<tt>llvm.frameaddress</tt>' intrinsic attempts to return the
4144 target-specific frame pointer value for the specified stack frame.
4150 The argument to this intrinsic indicates which function to return the frame
4151 pointer for. Zero indicates the calling function, one indicates its caller,
4152 etc. The argument is <b>required</b> to be a constant integer value.
4158 The '<tt>llvm.frameaddress</tt>' intrinsic either returns a pointer indicating
4159 the frame address of the specified call frame, or zero if it cannot be
4160 identified. The value returned by this intrinsic is likely to be incorrect or 0
4161 for arguments other than zero, so it should only be used for debugging purposes.
4165 Note that calling this intrinsic does not prevent function inlining or other
4166 aggressive transformations, so the value returned may not be that of the obvious
4167 source-language caller.
4171 <!-- _______________________________________________________________________ -->
4172 <div class="doc_subsubsection">
4173 <a name="int_stacksave">'<tt>llvm.stacksave</tt>' Intrinsic</a>
4176 <div class="doc_text">
4180 declare i8 *@llvm.stacksave()
4186 The '<tt>llvm.stacksave</tt>' intrinsic is used to remember the current state of
4187 the function stack, for use with <a href="#int_stackrestore">
4188 <tt>llvm.stackrestore</tt></a>. This is useful for implementing language
4189 features like scoped automatic variable sized arrays in C99.
4195 This intrinsic returns a opaque pointer value that can be passed to <a
4196 href="#int_stackrestore"><tt>llvm.stackrestore</tt></a>. When an
4197 <tt>llvm.stackrestore</tt> intrinsic is executed with a value saved from
4198 <tt>llvm.stacksave</tt>, it effectively restores the state of the stack to the
4199 state it was in when the <tt>llvm.stacksave</tt> intrinsic executed. In
4200 practice, this pops any <a href="#i_alloca">alloca</a> blocks from the stack
4201 that were allocated after the <tt>llvm.stacksave</tt> was executed.
4206 <!-- _______________________________________________________________________ -->
4207 <div class="doc_subsubsection">
4208 <a name="int_stackrestore">'<tt>llvm.stackrestore</tt>' Intrinsic</a>
4211 <div class="doc_text">
4215 declare void @llvm.stackrestore(i8 * %ptr)
4221 The '<tt>llvm.stackrestore</tt>' intrinsic is used to restore the state of
4222 the function stack to the state it was in when the corresponding <a
4223 href="#int_stacksave"><tt>llvm.stacksave</tt></a> intrinsic executed. This is
4224 useful for implementing language features like scoped automatic variable sized
4231 See the description for <a href="#int_stacksave"><tt>llvm.stacksave</tt></a>.
4237 <!-- _______________________________________________________________________ -->
4238 <div class="doc_subsubsection">
4239 <a name="int_prefetch">'<tt>llvm.prefetch</tt>' Intrinsic</a>
4242 <div class="doc_text">
4246 declare void @llvm.prefetch(i8* <address>, i32 <rw>, i32 <locality>)
4253 The '<tt>llvm.prefetch</tt>' intrinsic is a hint to the code generator to insert
4254 a prefetch instruction if supported; otherwise, it is a noop. Prefetches have
4256 effect on the behavior of the program but can change its performance
4263 <tt>address</tt> is the address to be prefetched, <tt>rw</tt> is the specifier
4264 determining if the fetch should be for a read (0) or write (1), and
4265 <tt>locality</tt> is a temporal locality specifier ranging from (0) - no
4266 locality, to (3) - extremely local keep in cache. The <tt>rw</tt> and
4267 <tt>locality</tt> arguments must be constant integers.
4273 This intrinsic does not modify the behavior of the program. In particular,
4274 prefetches cannot trap and do not produce a value. On targets that support this
4275 intrinsic, the prefetch can provide hints to the processor cache for better
4281 <!-- _______________________________________________________________________ -->
4282 <div class="doc_subsubsection">
4283 <a name="int_pcmarker">'<tt>llvm.pcmarker</tt>' Intrinsic</a>
4286 <div class="doc_text">
4290 declare void @llvm.pcmarker(i32 <id>)
4297 The '<tt>llvm.pcmarker</tt>' intrinsic is a method to export a Program Counter
4299 code to simulators and other tools. The method is target specific, but it is
4300 expected that the marker will use exported symbols to transmit the PC of the marker.
4301 The marker makes no guarantees that it will remain with any specific instruction
4302 after optimizations. It is possible that the presence of a marker will inhibit
4303 optimizations. The intended use is to be inserted after optimizations to allow
4304 correlations of simulation runs.
4310 <tt>id</tt> is a numerical id identifying the marker.
4316 This intrinsic does not modify the behavior of the program. Backends that do not
4317 support this intrinisic may ignore it.
4322 <!-- _______________________________________________________________________ -->
4323 <div class="doc_subsubsection">
4324 <a name="int_readcyclecounter">'<tt>llvm.readcyclecounter</tt>' Intrinsic</a>
4327 <div class="doc_text">
4331 declare i64 @llvm.readcyclecounter( )
4338 The '<tt>llvm.readcyclecounter</tt>' intrinsic provides access to the cycle
4339 counter register (or similar low latency, high accuracy clocks) on those targets
4340 that support it. On X86, it should map to RDTSC. On Alpha, it should map to RPCC.
4341 As the backing counters overflow quickly (on the order of 9 seconds on alpha), this
4342 should only be used for small timings.
4348 When directly supported, reading the cycle counter should not modify any memory.
4349 Implementations are allowed to either return a application specific value or a
4350 system wide value. On backends without support, this is lowered to a constant 0.
4355 <!-- ======================================================================= -->
4356 <div class="doc_subsection">
4357 <a name="int_libc">Standard C Library Intrinsics</a>
4360 <div class="doc_text">
4362 LLVM provides intrinsics for a few important standard C library functions.
4363 These intrinsics allow source-language front-ends to pass information about the
4364 alignment of the pointer arguments to the code generator, providing opportunity
4365 for more efficient code generation.
4370 <!-- _______________________________________________________________________ -->
4371 <div class="doc_subsubsection">
4372 <a name="int_memcpy">'<tt>llvm.memcpy</tt>' Intrinsic</a>
4375 <div class="doc_text">
4379 declare void @llvm.memcpy.i32(i8 * <dest>, i8 * <src>,
4380 i32 <len>, i32 <align>)
4381 declare void @llvm.memcpy.i64(i8 * <dest>, i8 * <src>,
4382 i64 <len>, i32 <align>)
4388 The '<tt>llvm.memcpy.*</tt>' intrinsics copy a block of memory from the source
4389 location to the destination location.
4393 Note that, unlike the standard libc function, the <tt>llvm.memcpy.*</tt>
4394 intrinsics do not return a value, and takes an extra alignment argument.
4400 The first argument is a pointer to the destination, the second is a pointer to
4401 the source. The third argument is an integer argument
4402 specifying the number of bytes to copy, and the fourth argument is the alignment
4403 of the source and destination locations.
4407 If the call to this intrinisic has an alignment value that is not 0 or 1, then
4408 the caller guarantees that both the source and destination pointers are aligned
4415 The '<tt>llvm.memcpy.*</tt>' intrinsics copy a block of memory from the source
4416 location to the destination location, which are not allowed to overlap. It
4417 copies "len" bytes of memory over. If the argument is known to be aligned to
4418 some boundary, this can be specified as the fourth argument, otherwise it should
4424 <!-- _______________________________________________________________________ -->
4425 <div class="doc_subsubsection">
4426 <a name="int_memmove">'<tt>llvm.memmove</tt>' Intrinsic</a>
4429 <div class="doc_text">
4433 declare void @llvm.memmove.i32(i8 * <dest>, i8 * <src>,
4434 i32 <len>, i32 <align>)
4435 declare void @llvm.memmove.i64(i8 * <dest>, i8 * <src>,
4436 i64 <len>, i32 <align>)
4442 The '<tt>llvm.memmove.*</tt>' intrinsics move a block of memory from the source
4443 location to the destination location. It is similar to the
4444 '<tt>llvm.memcmp</tt>' intrinsic but allows the two memory locations to overlap.
4448 Note that, unlike the standard libc function, the <tt>llvm.memmove.*</tt>
4449 intrinsics do not return a value, and takes an extra alignment argument.
4455 The first argument is a pointer to the destination, the second is a pointer to
4456 the source. The third argument is an integer argument
4457 specifying the number of bytes to copy, and the fourth argument is the alignment
4458 of the source and destination locations.
4462 If the call to this intrinisic has an alignment value that is not 0 or 1, then
4463 the caller guarantees that the source and destination pointers are aligned to
4470 The '<tt>llvm.memmove.*</tt>' intrinsics copy a block of memory from the source
4471 location to the destination location, which may overlap. It
4472 copies "len" bytes of memory over. If the argument is known to be aligned to
4473 some boundary, this can be specified as the fourth argument, otherwise it should
4479 <!-- _______________________________________________________________________ -->
4480 <div class="doc_subsubsection">
4481 <a name="int_memset">'<tt>llvm.memset.*</tt>' Intrinsics</a>
4484 <div class="doc_text">
4488 declare void @llvm.memset.i32(i8 * <dest>, i8 <val>,
4489 i32 <len>, i32 <align>)
4490 declare void @llvm.memset.i64(i8 * <dest>, i8 <val>,
4491 i64 <len>, i32 <align>)
4497 The '<tt>llvm.memset.*</tt>' intrinsics fill a block of memory with a particular
4502 Note that, unlike the standard libc function, the <tt>llvm.memset</tt> intrinsic
4503 does not return a value, and takes an extra alignment argument.
4509 The first argument is a pointer to the destination to fill, the second is the
4510 byte value to fill it with, the third argument is an integer
4511 argument specifying the number of bytes to fill, and the fourth argument is the
4512 known alignment of destination location.
4516 If the call to this intrinisic has an alignment value that is not 0 or 1, then
4517 the caller guarantees that the destination pointer is aligned to that boundary.
4523 The '<tt>llvm.memset.*</tt>' intrinsics fill "len" bytes of memory starting at
4525 destination location. If the argument is known to be aligned to some boundary,
4526 this can be specified as the fourth argument, otherwise it should be set to 0 or
4532 <!-- _______________________________________________________________________ -->
4533 <div class="doc_subsubsection">
4534 <a name="int_sqrt">'<tt>llvm.sqrt.*</tt>' Intrinsic</a>
4537 <div class="doc_text">
4540 <p>This is an overloaded intrinsic. You can use <tt>llvm.sqrt</tt> on any
4541 floating point or vector of floating point type. Not all targets support all
4544 declare float @llvm.sqrt.f32(float %Val)
4545 declare double @llvm.sqrt.f64(double %Val)
4546 declare x86_fp80 @llvm.sqrt.f80(x86_fp80 %Val)
4547 declare fp128 @llvm.sqrt.f128(fp128 %Val)
4548 declare ppc_fp128 @llvm.sqrt.ppcf128(ppc_fp128 %Val)
4554 The '<tt>llvm.sqrt</tt>' intrinsics return the sqrt of the specified operand,
4555 returning the same value as the libm '<tt>sqrt</tt>' functions would. Unlike
4556 <tt>sqrt</tt> in libm, however, <tt>llvm.sqrt</tt> has undefined behavior for
4557 negative numbers (which allows for better optimization).
4563 The argument and return value are floating point numbers of the same type.
4569 This function returns the sqrt of the specified operand if it is a nonnegative
4570 floating point number.
4574 <!-- _______________________________________________________________________ -->
4575 <div class="doc_subsubsection">
4576 <a name="int_powi">'<tt>llvm.powi.*</tt>' Intrinsic</a>
4579 <div class="doc_text">
4582 <p>This is an overloaded intrinsic. You can use <tt>llvm.powi</tt> on any
4583 floating point or vector of floating point type. Not all targets support all
4586 declare float @llvm.powi.f32(float %Val, i32 %power)
4587 declare double @llvm.powi.f64(double %Val, i32 %power)
4588 declare x86_fp80 @llvm.powi.f80(x86_fp80 %Val, i32 %power)
4589 declare fp128 @llvm.powi.f128(fp128 %Val, i32 %power)
4590 declare ppc_fp128 @llvm.powi.ppcf128(ppc_fp128 %Val, i32 %power)
4596 The '<tt>llvm.powi.*</tt>' intrinsics return the first operand raised to the
4597 specified (positive or negative) power. The order of evaluation of
4598 multiplications is not defined. When a vector of floating point type is
4599 used, the second argument remains a scalar integer value.
4605 The second argument is an integer power, and the first is a value to raise to
4612 This function returns the first value raised to the second power with an
4613 unspecified sequence of rounding operations.</p>
4616 <!-- _______________________________________________________________________ -->
4617 <div class="doc_subsubsection">
4618 <a name="int_sin">'<tt>llvm.sin.*</tt>' Intrinsic</a>
4621 <div class="doc_text">
4624 <p>This is an overloaded intrinsic. You can use <tt>llvm.sin</tt> on any
4625 floating point or vector of floating point type. Not all targets support all
4628 declare float @llvm.sin.f32(float %Val)
4629 declare double @llvm.sin.f64(double %Val)
4630 declare x86_fp80 @llvm.sin.f80(x86_fp80 %Val)
4631 declare fp128 @llvm.sin.f128(fp128 %Val)
4632 declare ppc_fp128 @llvm.sin.ppcf128(ppc_fp128 %Val)
4638 The '<tt>llvm.sin.*</tt>' intrinsics return the sine of the operand.
4644 The argument and return value are floating point numbers of the same type.
4650 This function returns the sine of the specified operand, returning the
4651 same values as the libm <tt>sin</tt> functions would, and handles error
4652 conditions in the same way.</p>
4655 <!-- _______________________________________________________________________ -->
4656 <div class="doc_subsubsection">
4657 <a name="int_cos">'<tt>llvm.cos.*</tt>' Intrinsic</a>
4660 <div class="doc_text">
4663 <p>This is an overloaded intrinsic. You can use <tt>llvm.cos</tt> on any
4664 floating point or vector of floating point type. Not all targets support all
4667 declare float @llvm.cos.f32(float %Val)
4668 declare double @llvm.cos.f64(double %Val)
4669 declare x86_fp80 @llvm.cos.f80(x86_fp80 %Val)
4670 declare fp128 @llvm.cos.f128(fp128 %Val)
4671 declare ppc_fp128 @llvm.cos.ppcf128(ppc_fp128 %Val)
4677 The '<tt>llvm.cos.*</tt>' intrinsics return the cosine of the operand.
4683 The argument and return value are floating point numbers of the same type.
4689 This function returns the cosine of the specified operand, returning the
4690 same values as the libm <tt>cos</tt> functions would, and handles error
4691 conditions in the same way.</p>
4694 <!-- _______________________________________________________________________ -->
4695 <div class="doc_subsubsection">
4696 <a name="int_pow">'<tt>llvm.pow.*</tt>' Intrinsic</a>
4699 <div class="doc_text">
4702 <p>This is an overloaded intrinsic. You can use <tt>llvm.pow</tt> on any
4703 floating point or vector of floating point type. Not all targets support all
4706 declare float @llvm.pow.f32(float %Val, float %Power)
4707 declare double @llvm.pow.f64(double %Val, double %Power)
4708 declare x86_fp80 @llvm.pow.f80(x86_fp80 %Val, x86_fp80 %Power)
4709 declare fp128 @llvm.pow.f128(fp128 %Val, fp128 %Power)
4710 declare ppc_fp128 @llvm.pow.ppcf128(ppc_fp128 %Val, ppc_fp128 Power)
4716 The '<tt>llvm.pow.*</tt>' intrinsics return the first operand raised to the
4717 specified (positive or negative) power.
4723 The second argument is a floating point power, and the first is a value to
4724 raise to that power.
4730 This function returns the first value raised to the second power,
4732 same values as the libm <tt>pow</tt> functions would, and handles error
4733 conditions in the same way.</p>
4737 <!-- ======================================================================= -->
4738 <div class="doc_subsection">
4739 <a name="int_manip">Bit Manipulation Intrinsics</a>
4742 <div class="doc_text">
4744 LLVM provides intrinsics for a few important bit manipulation operations.
4745 These allow efficient code generation for some algorithms.
4750 <!-- _______________________________________________________________________ -->
4751 <div class="doc_subsubsection">
4752 <a name="int_bswap">'<tt>llvm.bswap.*</tt>' Intrinsics</a>
4755 <div class="doc_text">
4758 <p>This is an overloaded intrinsic function. You can use bswap on any integer
4759 type that is an even number of bytes (i.e. BitWidth % 16 == 0).
4761 declare i16 @llvm.bswap.i16(i16 <id>)
4762 declare i32 @llvm.bswap.i32(i32 <id>)
4763 declare i64 @llvm.bswap.i64(i64 <id>)
4769 The '<tt>llvm.bswap</tt>' family of intrinsics is used to byte swap integer
4770 values with an even number of bytes (positive multiple of 16 bits). These are
4771 useful for performing operations on data that is not in the target's native
4778 The <tt>llvm.bswap.i16</tt> intrinsic returns an i16 value that has the high
4779 and low byte of the input i16 swapped. Similarly, the <tt>llvm.bswap.i32</tt>
4780 intrinsic returns an i32 value that has the four bytes of the input i32
4781 swapped, so that if the input bytes are numbered 0, 1, 2, 3 then the returned
4782 i32 will have its bytes in 3, 2, 1, 0 order. The <tt>llvm.bswap.i48</tt>,
4783 <tt>llvm.bswap.i64</tt> and other intrinsics extend this concept to
4784 additional even-byte lengths (6 bytes, 8 bytes and more, respectively).
4789 <!-- _______________________________________________________________________ -->
4790 <div class="doc_subsubsection">
4791 <a name="int_ctpop">'<tt>llvm.ctpop.*</tt>' Intrinsic</a>
4794 <div class="doc_text">
4797 <p>This is an overloaded intrinsic. You can use llvm.ctpop on any integer bit
4798 width. Not all targets support all bit widths however.
4800 declare i8 @llvm.ctpop.i8 (i8 <src>)
4801 declare i16 @llvm.ctpop.i16(i16 <src>)
4802 declare i32 @llvm.ctpop.i32(i32 <src>)
4803 declare i64 @llvm.ctpop.i64(i64 <src>)
4804 declare i256 @llvm.ctpop.i256(i256 <src>)
4810 The '<tt>llvm.ctpop</tt>' family of intrinsics counts the number of bits set in a
4817 The only argument is the value to be counted. The argument may be of any
4818 integer type. The return type must match the argument type.
4824 The '<tt>llvm.ctpop</tt>' intrinsic counts the 1's in a variable.
4828 <!-- _______________________________________________________________________ -->
4829 <div class="doc_subsubsection">
4830 <a name="int_ctlz">'<tt>llvm.ctlz.*</tt>' Intrinsic</a>
4833 <div class="doc_text">
4836 <p>This is an overloaded intrinsic. You can use <tt>llvm.ctlz</tt> on any
4837 integer bit width. Not all targets support all bit widths however.
4839 declare i8 @llvm.ctlz.i8 (i8 <src>)
4840 declare i16 @llvm.ctlz.i16(i16 <src>)
4841 declare i32 @llvm.ctlz.i32(i32 <src>)
4842 declare i64 @llvm.ctlz.i64(i64 <src>)
4843 declare i256 @llvm.ctlz.i256(i256 <src>)
4849 The '<tt>llvm.ctlz</tt>' family of intrinsic functions counts the number of
4850 leading zeros in a variable.
4856 The only argument is the value to be counted. The argument may be of any
4857 integer type. The return type must match the argument type.
4863 The '<tt>llvm.ctlz</tt>' intrinsic counts the leading (most significant) zeros
4864 in a variable. If the src == 0 then the result is the size in bits of the type
4865 of src. For example, <tt>llvm.ctlz(i32 2) = 30</tt>.
4871 <!-- _______________________________________________________________________ -->
4872 <div class="doc_subsubsection">
4873 <a name="int_cttz">'<tt>llvm.cttz.*</tt>' Intrinsic</a>
4876 <div class="doc_text">
4879 <p>This is an overloaded intrinsic. You can use <tt>llvm.cttz</tt> on any
4880 integer bit width. Not all targets support all bit widths however.
4882 declare i8 @llvm.cttz.i8 (i8 <src>)
4883 declare i16 @llvm.cttz.i16(i16 <src>)
4884 declare i32 @llvm.cttz.i32(i32 <src>)
4885 declare i64 @llvm.cttz.i64(i64 <src>)
4886 declare i256 @llvm.cttz.i256(i256 <src>)
4892 The '<tt>llvm.cttz</tt>' family of intrinsic functions counts the number of
4899 The only argument is the value to be counted. The argument may be of any
4900 integer type. The return type must match the argument type.
4906 The '<tt>llvm.cttz</tt>' intrinsic counts the trailing (least significant) zeros
4907 in a variable. If the src == 0 then the result is the size in bits of the type
4908 of src. For example, <tt>llvm.cttz(2) = 1</tt>.
4912 <!-- _______________________________________________________________________ -->
4913 <div class="doc_subsubsection">
4914 <a name="int_part_select">'<tt>llvm.part.select.*</tt>' Intrinsic</a>
4917 <div class="doc_text">
4920 <p>This is an overloaded intrinsic. You can use <tt>llvm.part.select</tt>
4921 on any integer bit width.
4923 declare i17 @llvm.part.select.i17 (i17 %val, i32 %loBit, i32 %hiBit)
4924 declare i29 @llvm.part.select.i29 (i29 %val, i32 %loBit, i32 %hiBit)
4928 <p>The '<tt>llvm.part.select</tt>' family of intrinsic functions selects a
4929 range of bits from an integer value and returns them in the same bit width as
4930 the original value.</p>
4933 <p>The first argument, <tt>%val</tt> and the result may be integer types of
4934 any bit width but they must have the same bit width. The second and third
4935 arguments must be <tt>i32</tt> type since they specify only a bit index.</p>
4938 <p>The operation of the '<tt>llvm.part.select</tt>' intrinsic has two modes
4939 of operation: forwards and reverse. If <tt>%loBit</tt> is greater than
4940 <tt>%hiBits</tt> then the intrinsic operates in reverse mode. Otherwise it
4941 operates in forward mode.</p>
4942 <p>In forward mode, this intrinsic is the equivalent of shifting <tt>%val</tt>
4943 right by <tt>%loBit</tt> bits and then ANDing it with a mask with
4944 only the <tt>%hiBit - %loBit</tt> bits set, as follows:</p>
4946 <li>The <tt>%val</tt> is shifted right (LSHR) by the number of bits specified
4947 by <tt>%loBits</tt>. This normalizes the value to the low order bits.</li>
4948 <li>The <tt>%loBits</tt> value is subtracted from the <tt>%hiBits</tt> value
4949 to determine the number of bits to retain.</li>
4950 <li>A mask of the retained bits is created by shifting a -1 value.</li>
4951 <li>The mask is ANDed with <tt>%val</tt> to produce the result.
4953 <p>In reverse mode, a similar computation is made except that the bits are
4954 returned in the reverse order. So, for example, if <tt>X</tt> has the value
4955 <tt>i16 0x0ACF (101011001111)</tt> and we apply
4956 <tt>part.select(i16 X, 8, 3)</tt> to it, we get back the value
4957 <tt>i16 0x0026 (000000100110)</tt>.</p>
4960 <div class="doc_subsubsection">
4961 <a name="int_part_set">'<tt>llvm.part.set.*</tt>' Intrinsic</a>
4964 <div class="doc_text">
4967 <p>This is an overloaded intrinsic. You can use <tt>llvm.part.set</tt>
4968 on any integer bit width.
4970 declare i17 @llvm.part.set.i17.i9 (i17 %val, i9 %repl, i32 %lo, i32 %hi)
4971 declare i29 @llvm.part.set.i29.i9 (i29 %val, i9 %repl, i32 %lo, i32 %hi)
4975 <p>The '<tt>llvm.part.set</tt>' family of intrinsic functions replaces a range
4976 of bits in an integer value with another integer value. It returns the integer
4977 with the replaced bits.</p>
4980 <p>The first argument, <tt>%val</tt> and the result may be integer types of
4981 any bit width but they must have the same bit width. <tt>%val</tt> is the value
4982 whose bits will be replaced. The second argument, <tt>%repl</tt> may be an
4983 integer of any bit width. The third and fourth arguments must be <tt>i32</tt>
4984 type since they specify only a bit index.</p>
4987 <p>The operation of the '<tt>llvm.part.set</tt>' intrinsic has two modes
4988 of operation: forwards and reverse. If <tt>%lo</tt> is greater than
4989 <tt>%hi</tt> then the intrinsic operates in reverse mode. Otherwise it
4990 operates in forward mode.</p>
4991 <p>For both modes, the <tt>%repl</tt> value is prepared for use by either
4992 truncating it down to the size of the replacement area or zero extending it
4993 up to that size.</p>
4994 <p>In forward mode, the bits between <tt>%lo</tt> and <tt>%hi</tt> (inclusive)
4995 are replaced with corresponding bits from <tt>%repl</tt>. That is the 0th bit
4996 in <tt>%repl</tt> replaces the <tt>%lo</tt>th bit in <tt>%val</tt> and etc. up
4997 to the <tt>%hi</tt>th bit.
4998 <p>In reverse mode, a similar computation is made except that the bits are
4999 reversed. That is, the <tt>0</tt>th bit in <tt>%repl</tt> replaces the
5000 <tt>%hi</tt> bit in <tt>%val</tt> and etc. down to the <tt>%lo</tt>th bit.
5003 llvm.part.set(0xFFFF, 0, 4, 7) -> 0xFF0F
5004 llvm.part.set(0xFFFF, 0, 7, 4) -> 0xFF0F
5005 llvm.part.set(0xFFFF, 1, 7, 4) -> 0xFF8F
5006 llvm.part.set(0xFFFF, F, 8, 3) -> 0xFFE7
5007 llvm.part.set(0xFFFF, 0, 3, 8) -> 0xFE07
5011 <!-- ======================================================================= -->
5012 <div class="doc_subsection">
5013 <a name="int_debugger">Debugger Intrinsics</a>
5016 <div class="doc_text">
5018 The LLVM debugger intrinsics (which all start with <tt>llvm.dbg.</tt> prefix),
5019 are described in the <a
5020 href="SourceLevelDebugging.html#format_common_intrinsics">LLVM Source Level
5021 Debugging</a> document.
5026 <!-- ======================================================================= -->
5027 <div class="doc_subsection">
5028 <a name="int_eh">Exception Handling Intrinsics</a>
5031 <div class="doc_text">
5032 <p> The LLVM exception handling intrinsics (which all start with
5033 <tt>llvm.eh.</tt> prefix), are described in the <a
5034 href="ExceptionHandling.html#format_common_intrinsics">LLVM Exception
5035 Handling</a> document. </p>
5038 <!-- ======================================================================= -->
5039 <div class="doc_subsection">
5040 <a name="int_trampoline">Trampoline Intrinsic</a>
5043 <div class="doc_text">
5045 This intrinsic makes it possible to excise one parameter, marked with
5046 the <tt>nest</tt> attribute, from a function. The result is a callable
5047 function pointer lacking the nest parameter - the caller does not need
5048 to provide a value for it. Instead, the value to use is stored in
5049 advance in a "trampoline", a block of memory usually allocated
5050 on the stack, which also contains code to splice the nest value into the
5051 argument list. This is used to implement the GCC nested function address
5055 For example, if the function is
5056 <tt>i32 f(i8* nest %c, i32 %x, i32 %y)</tt> then the resulting function
5057 pointer has signature <tt>i32 (i32, i32)*</tt>. It can be created as follows:</p>
5059 %tramp = alloca [10 x i8], align 4 ; size and alignment only correct for X86
5060 %tramp1 = getelementptr [10 x i8]* %tramp, i32 0, i32 0
5061 %p = call i8* @llvm.init.trampoline( i8* %tramp1, i8* bitcast (i32 (i8* nest , i32, i32)* @f to i8*), i8* %nval )
5062 %fp = bitcast i8* %p to i32 (i32, i32)*
5064 <p>The call <tt>%val = call i32 %fp( i32 %x, i32 %y )</tt> is then equivalent
5065 to <tt>%val = call i32 %f( i8* %nval, i32 %x, i32 %y )</tt>.</p>
5068 <!-- _______________________________________________________________________ -->
5069 <div class="doc_subsubsection">
5070 <a name="int_it">'<tt>llvm.init.trampoline</tt>' Intrinsic</a>
5072 <div class="doc_text">
5075 declare i8* @llvm.init.trampoline(i8* <tramp>, i8* <func>, i8* <nval>)
5079 This fills the memory pointed to by <tt>tramp</tt> with code
5080 and returns a function pointer suitable for executing it.
5084 The <tt>llvm.init.trampoline</tt> intrinsic takes three arguments, all
5085 pointers. The <tt>tramp</tt> argument must point to a sufficiently large
5086 and sufficiently aligned block of memory; this memory is written to by the
5087 intrinsic. Note that the size and the alignment are target-specific - LLVM
5088 currently provides no portable way of determining them, so a front-end that
5089 generates this intrinsic needs to have some target-specific knowledge.
5090 The <tt>func</tt> argument must hold a function bitcast to an <tt>i8*</tt>.
5094 The block of memory pointed to by <tt>tramp</tt> is filled with target
5095 dependent code, turning it into a function. A pointer to this function is
5096 returned, but needs to be bitcast to an
5097 <a href="#int_trampoline">appropriate function pointer type</a>
5098 before being called. The new function's signature is the same as that of
5099 <tt>func</tt> with any arguments marked with the <tt>nest</tt> attribute
5100 removed. At most one such <tt>nest</tt> argument is allowed, and it must be
5101 of pointer type. Calling the new function is equivalent to calling
5102 <tt>func</tt> with the same argument list, but with <tt>nval</tt> used for the
5103 missing <tt>nest</tt> argument. If, after calling
5104 <tt>llvm.init.trampoline</tt>, the memory pointed to by <tt>tramp</tt> is
5105 modified, then the effect of any later call to the returned function pointer is
5110 <!-- ======================================================================= -->
5111 <div class="doc_subsection">
5112 <a name="int_general">General Intrinsics</a>
5115 <div class="doc_text">
5116 <p> This class of intrinsics is designed to be generic and has
5117 no specific purpose. </p>
5120 <!-- _______________________________________________________________________ -->
5121 <div class="doc_subsubsection">
5122 <a name="int_var_annotation">'<tt>llvm.var.annotation</tt>' Intrinsic</a>
5125 <div class="doc_text">
5129 declare void @llvm.var.annotation(i8* <val>, i8* <str>, i8* <str>, i32 <int> )
5135 The '<tt>llvm.var.annotation</tt>' intrinsic
5141 The first argument is a pointer to a value, the second is a pointer to a
5142 global string, the third is a pointer to a global string which is the source
5143 file name, and the last argument is the line number.
5149 This intrinsic allows annotation of local variables with arbitrary strings.
5150 This can be useful for special purpose optimizations that want to look for these
5151 annotations. These have no other defined use, they are ignored by code
5152 generation and optimization.
5155 <!-- _______________________________________________________________________ -->
5156 <div class="doc_subsubsection">
5157 <a name="int_annotation">'<tt>llvm.annotation.*</tt>' Intrinsic</a>
5160 <div class="doc_text">
5163 <p>This is an overloaded intrinsic. You can use '<tt>llvm.annotation</tt>' on
5164 any integer bit width.
5167 declare i8 @llvm.annotation.i8(i8 <val>, i8* <str>, i8* <str>, i32 <int> )
5168 declare i16 @llvm.annotation.i16(i16 <val>, i8* <str>, i8* <str>, i32 <int> )
5169 declare i32 @llvm.annotation.i32(i32 <val>, i8* <str>, i8* <str>, i32 <int> )
5170 declare i64 @llvm.annotation.i64(i64 <val>, i8* <str>, i8* <str>, i32 <int> )
5171 declare i256 @llvm.annotation.i256(i256 <val>, i8* <str>, i8* <str>, i32 <int> )
5177 The '<tt>llvm.annotation</tt>' intrinsic.
5183 The first argument is an integer value (result of some expression),
5184 the second is a pointer to a global string, the third is a pointer to a global
5185 string which is the source file name, and the last argument is the line number.
5186 It returns the value of the first argument.
5192 This intrinsic allows annotations to be put on arbitrary expressions
5193 with arbitrary strings. This can be useful for special purpose optimizations
5194 that want to look for these annotations. These have no other defined use, they
5195 are ignored by code generation and optimization.
5198 <!-- *********************************************************************** -->
5201 <a href="http://jigsaw.w3.org/css-validator/check/referer"><img
5202 src="http://jigsaw.w3.org/css-validator/images/vcss" alt="Valid CSS!"></a>
5203 <a href="http://validator.w3.org/check/referer"><img
5204 src="http://www.w3.org/Icons/valid-html401" alt="Valid HTML 4.01!" /></a>
5206 <a href="mailto:sabre@nondot.org">Chris Lattner</a><br>
5207 <a href="http://llvm.org">The LLVM Compiler Infrastructure</a><br>
5208 Last modified: $Date$