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_integer">Integer Type</a></li>
44 <li><a href="#t_array">Array Type</a></li>
45 <li><a href="#t_function">Function Type</a></li>
46 <li><a href="#t_pointer">Pointer Type</a></li>
47 <li><a href="#t_struct">Structure Type</a></li>
48 <li><a href="#t_pstruct">Packed Structure Type</a></li>
49 <li><a href="#t_vector">Vector Type</a></li>
50 <li><a href="#t_opaque">Opaque Type</a></li>
55 <li><a href="#constants">Constants</a>
57 <li><a href="#simpleconstants">Simple Constants</a>
58 <li><a href="#aggregateconstants">Aggregate Constants</a>
59 <li><a href="#globalconstants">Global Variable and Function Addresses</a>
60 <li><a href="#undefvalues">Undefined Values</a>
61 <li><a href="#constantexprs">Constant Expressions</a>
64 <li><a href="#othervalues">Other Values</a>
66 <li><a href="#inlineasm">Inline Assembler Expressions</a>
69 <li><a href="#instref">Instruction Reference</a>
71 <li><a href="#terminators">Terminator Instructions</a>
73 <li><a href="#i_ret">'<tt>ret</tt>' Instruction</a></li>
74 <li><a href="#i_br">'<tt>br</tt>' Instruction</a></li>
75 <li><a href="#i_switch">'<tt>switch</tt>' Instruction</a></li>
76 <li><a href="#i_invoke">'<tt>invoke</tt>' Instruction</a></li>
77 <li><a href="#i_unwind">'<tt>unwind</tt>' Instruction</a></li>
78 <li><a href="#i_unreachable">'<tt>unreachable</tt>' Instruction</a></li>
81 <li><a href="#binaryops">Binary Operations</a>
83 <li><a href="#i_add">'<tt>add</tt>' Instruction</a></li>
84 <li><a href="#i_sub">'<tt>sub</tt>' Instruction</a></li>
85 <li><a href="#i_mul">'<tt>mul</tt>' Instruction</a></li>
86 <li><a href="#i_udiv">'<tt>udiv</tt>' Instruction</a></li>
87 <li><a href="#i_sdiv">'<tt>sdiv</tt>' Instruction</a></li>
88 <li><a href="#i_fdiv">'<tt>fdiv</tt>' Instruction</a></li>
89 <li><a href="#i_urem">'<tt>urem</tt>' Instruction</a></li>
90 <li><a href="#i_srem">'<tt>srem</tt>' Instruction</a></li>
91 <li><a href="#i_frem">'<tt>frem</tt>' Instruction</a></li>
94 <li><a href="#bitwiseops">Bitwise Binary Operations</a>
96 <li><a href="#i_shl">'<tt>shl</tt>' Instruction</a></li>
97 <li><a href="#i_lshr">'<tt>lshr</tt>' Instruction</a></li>
98 <li><a href="#i_ashr">'<tt>ashr</tt>' Instruction</a></li>
99 <li><a href="#i_and">'<tt>and</tt>' Instruction</a></li>
100 <li><a href="#i_or">'<tt>or</tt>' Instruction</a></li>
101 <li><a href="#i_xor">'<tt>xor</tt>' Instruction</a></li>
104 <li><a href="#vectorops">Vector Operations</a>
106 <li><a href="#i_extractelement">'<tt>extractelement</tt>' Instruction</a></li>
107 <li><a href="#i_insertelement">'<tt>insertelement</tt>' Instruction</a></li>
108 <li><a href="#i_shufflevector">'<tt>shufflevector</tt>' Instruction</a></li>
111 <li><a href="#memoryops">Memory Access and Addressing Operations</a>
113 <li><a href="#i_malloc">'<tt>malloc</tt>' Instruction</a></li>
114 <li><a href="#i_free">'<tt>free</tt>' Instruction</a></li>
115 <li><a href="#i_alloca">'<tt>alloca</tt>' Instruction</a></li>
116 <li><a href="#i_load">'<tt>load</tt>' Instruction</a></li>
117 <li><a href="#i_store">'<tt>store</tt>' Instruction</a></li>
118 <li><a href="#i_getelementptr">'<tt>getelementptr</tt>' Instruction</a></li>
121 <li><a href="#convertops">Conversion Operations</a>
123 <li><a href="#i_trunc">'<tt>trunc .. to</tt>' Instruction</a></li>
124 <li><a href="#i_zext">'<tt>zext .. to</tt>' Instruction</a></li>
125 <li><a href="#i_sext">'<tt>sext .. to</tt>' Instruction</a></li>
126 <li><a href="#i_fptrunc">'<tt>fptrunc .. to</tt>' Instruction</a></li>
127 <li><a href="#i_fpext">'<tt>fpext .. to</tt>' Instruction</a></li>
128 <li><a href="#i_fptoui">'<tt>fptoui .. to</tt>' Instruction</a></li>
129 <li><a href="#i_fptosi">'<tt>fptosi .. to</tt>' Instruction</a></li>
130 <li><a href="#i_uitofp">'<tt>uitofp .. to</tt>' Instruction</a></li>
131 <li><a href="#i_sitofp">'<tt>sitofp .. to</tt>' Instruction</a></li>
132 <li><a href="#i_ptrtoint">'<tt>ptrtoint .. to</tt>' Instruction</a></li>
133 <li><a href="#i_inttoptr">'<tt>inttoptr .. to</tt>' Instruction</a></li>
134 <li><a href="#i_bitcast">'<tt>bitcast .. to</tt>' Instruction</a></li>
136 <li><a href="#otherops">Other Operations</a>
138 <li><a href="#i_icmp">'<tt>icmp</tt>' Instruction</a></li>
139 <li><a href="#i_fcmp">'<tt>fcmp</tt>' Instruction</a></li>
140 <li><a href="#i_phi">'<tt>phi</tt>' Instruction</a></li>
141 <li><a href="#i_select">'<tt>select</tt>' Instruction</a></li>
142 <li><a href="#i_call">'<tt>call</tt>' Instruction</a></li>
143 <li><a href="#i_va_arg">'<tt>va_arg</tt>' Instruction</a></li>
148 <li><a href="#intrinsics">Intrinsic Functions</a>
150 <li><a href="#int_varargs">Variable Argument Handling Intrinsics</a>
152 <li><a href="#int_va_start">'<tt>llvm.va_start</tt>' Intrinsic</a></li>
153 <li><a href="#int_va_end">'<tt>llvm.va_end</tt>' Intrinsic</a></li>
154 <li><a href="#int_va_copy">'<tt>llvm.va_copy</tt>' Intrinsic</a></li>
157 <li><a href="#int_gc">Accurate Garbage Collection Intrinsics</a>
159 <li><a href="#int_gcroot">'<tt>llvm.gcroot</tt>' Intrinsic</a></li>
160 <li><a href="#int_gcread">'<tt>llvm.gcread</tt>' Intrinsic</a></li>
161 <li><a href="#int_gcwrite">'<tt>llvm.gcwrite</tt>' Intrinsic</a></li>
164 <li><a href="#int_codegen">Code Generator Intrinsics</a>
166 <li><a href="#int_returnaddress">'<tt>llvm.returnaddress</tt>' Intrinsic</a></li>
167 <li><a href="#int_frameaddress">'<tt>llvm.frameaddress</tt>' Intrinsic</a></li>
168 <li><a href="#int_stacksave">'<tt>llvm.stacksave</tt>' Intrinsic</a></li>
169 <li><a href="#int_stackrestore">'<tt>llvm.stackrestore</tt>' Intrinsic</a></li>
170 <li><a href="#int_prefetch">'<tt>llvm.prefetch</tt>' Intrinsic</a></li>
171 <li><a href="#int_pcmarker">'<tt>llvm.pcmarker</tt>' Intrinsic</a></li>
172 <li><a href="#int_readcyclecounter"><tt>llvm.readcyclecounter</tt>' Intrinsic</a></li>
175 <li><a href="#int_libc">Standard C Library Intrinsics</a>
177 <li><a href="#int_memcpy">'<tt>llvm.memcpy.*</tt>' Intrinsic</a></li>
178 <li><a href="#int_memmove">'<tt>llvm.memmove.*</tt>' Intrinsic</a></li>
179 <li><a href="#int_memset">'<tt>llvm.memset.*</tt>' Intrinsic</a></li>
180 <li><a href="#int_sqrt">'<tt>llvm.sqrt.*</tt>' Intrinsic</a></li>
181 <li><a href="#int_powi">'<tt>llvm.powi.*</tt>' Intrinsic</a></li>
182 <li><a href="#int_sin">'<tt>llvm.sin.*</tt>' Intrinsic</a></li>
183 <li><a href="#int_cos">'<tt>llvm.cos.*</tt>' Intrinsic</a></li>
184 <li><a href="#int_pow">'<tt>llvm.pow.*</tt>' Intrinsic</a></li>
187 <li><a href="#int_manip">Bit Manipulation Intrinsics</a>
189 <li><a href="#int_bswap">'<tt>llvm.bswap.*</tt>' Intrinsics</a></li>
190 <li><a href="#int_ctpop">'<tt>llvm.ctpop.*</tt>' Intrinsic </a></li>
191 <li><a href="#int_ctlz">'<tt>llvm.ctlz.*</tt>' Intrinsic </a></li>
192 <li><a href="#int_cttz">'<tt>llvm.cttz.*</tt>' Intrinsic </a></li>
193 <li><a href="#int_part_select">'<tt>llvm.part.select.*</tt>' Intrinsic </a></li>
194 <li><a href="#int_part_set">'<tt>llvm.part.set.*</tt>' Intrinsic </a></li>
197 <li><a href="#int_debugger">Debugger intrinsics</a></li>
198 <li><a href="#int_eh">Exception Handling intrinsics</a></li>
199 <li><a href="#int_trampoline">Trampoline Intrinsic</a>
201 <li><a href="#int_it">'<tt>llvm.init.trampoline</tt>' Intrinsic</a></li>
204 <li><a href="#int_general">General intrinsics</a>
206 <li><a href="#int_var_annotation">
207 <tt>llvm.var.annotation</tt>' Intrinsic</a></li>
210 <li><a href="#int_annotation">
211 <tt>llvm.annotation.*</tt>' Intrinsic</a></li>
218 <div class="doc_author">
219 <p>Written by <a href="mailto:sabre@nondot.org">Chris Lattner</a>
220 and <a href="mailto:vadve@cs.uiuc.edu">Vikram Adve</a></p>
223 <!-- *********************************************************************** -->
224 <div class="doc_section"> <a name="abstract">Abstract </a></div>
225 <!-- *********************************************************************** -->
227 <div class="doc_text">
228 <p>This document is a reference manual for the LLVM assembly language.
229 LLVM is an SSA based representation that provides type safety,
230 low-level operations, flexibility, and the capability of representing
231 'all' high-level languages cleanly. It is the common code
232 representation used throughout all phases of the LLVM compilation
236 <!-- *********************************************************************** -->
237 <div class="doc_section"> <a name="introduction">Introduction</a> </div>
238 <!-- *********************************************************************** -->
240 <div class="doc_text">
242 <p>The LLVM code representation is designed to be used in three
243 different forms: as an in-memory compiler IR, as an on-disk bitcode
244 representation (suitable for fast loading by a Just-In-Time compiler),
245 and as a human readable assembly language representation. This allows
246 LLVM to provide a powerful intermediate representation for efficient
247 compiler transformations and analysis, while providing a natural means
248 to debug and visualize the transformations. The three different forms
249 of LLVM are all equivalent. This document describes the human readable
250 representation and notation.</p>
252 <p>The LLVM representation aims to be light-weight and low-level
253 while being expressive, typed, and extensible at the same time. It
254 aims to be a "universal IR" of sorts, by being at a low enough level
255 that high-level ideas may be cleanly mapped to it (similar to how
256 microprocessors are "universal IR's", allowing many source languages to
257 be mapped to them). By providing type information, LLVM can be used as
258 the target of optimizations: for example, through pointer analysis, it
259 can be proven that a C automatic variable is never accessed outside of
260 the current function... allowing it to be promoted to a simple SSA
261 value instead of a memory location.</p>
265 <!-- _______________________________________________________________________ -->
266 <div class="doc_subsubsection"> <a name="wellformed">Well-Formedness</a> </div>
268 <div class="doc_text">
270 <p>It is important to note that this document describes 'well formed'
271 LLVM assembly language. There is a difference between what the parser
272 accepts and what is considered 'well formed'. For example, the
273 following instruction is syntactically okay, but not well formed:</p>
275 <div class="doc_code">
277 %x = <a href="#i_add">add</a> i32 1, %x
281 <p>...because the definition of <tt>%x</tt> does not dominate all of
282 its uses. The LLVM infrastructure provides a verification pass that may
283 be used to verify that an LLVM module is well formed. This pass is
284 automatically run by the parser after parsing input assembly and by
285 the optimizer before it outputs bitcode. The violations pointed out
286 by the verifier pass indicate bugs in transformation passes or input to
290 <!-- Describe the typesetting conventions here. -->
292 <!-- *********************************************************************** -->
293 <div class="doc_section"> <a name="identifiers">Identifiers</a> </div>
294 <!-- *********************************************************************** -->
296 <div class="doc_text">
298 <p>LLVM identifiers come in two basic types: global and local. Global
299 identifiers (functions, global variables) begin with the @ character. Local
300 identifiers (register names, types) begin with the % character. Additionally,
301 there are three different formats for identifiers, for different purposes:
304 <li>Named values are represented as a string of characters with their prefix.
305 For example, %foo, @DivisionByZero, %a.really.long.identifier. The actual
306 regular expression used is '<tt>[%@][a-zA-Z$._][a-zA-Z$._0-9]*</tt>'.
307 Identifiers which require other characters in their names can be surrounded
308 with quotes. In this way, anything except a <tt>"</tt> character can
309 be used in a named value.</li>
311 <li>Unnamed values are represented as an unsigned numeric value with their
312 prefix. For example, %12, @2, %44.</li>
314 <li>Constants, which are described in a <a href="#constants">section about
315 constants</a>, below.</li>
318 <p>LLVM requires that values start with a prefix for two reasons: Compilers
319 don't need to worry about name clashes with reserved words, and the set of
320 reserved words may be expanded in the future without penalty. Additionally,
321 unnamed identifiers allow a compiler to quickly come up with a temporary
322 variable without having to avoid symbol table conflicts.</p>
324 <p>Reserved words in LLVM are very similar to reserved words in other
325 languages. There are keywords for different opcodes
326 ('<tt><a href="#i_add">add</a></tt>',
327 '<tt><a href="#i_bitcast">bitcast</a></tt>',
328 '<tt><a href="#i_ret">ret</a></tt>', etc...), for primitive type names ('<tt><a
329 href="#t_void">void</a></tt>', '<tt><a href="#t_primitive">i32</a></tt>', etc...),
330 and others. These reserved words cannot conflict with variable names, because
331 none of them start with a prefix character ('%' or '@').</p>
333 <p>Here is an example of LLVM code to multiply the integer variable
334 '<tt>%X</tt>' by 8:</p>
338 <div class="doc_code">
340 %result = <a href="#i_mul">mul</a> i32 %X, 8
344 <p>After strength reduction:</p>
346 <div class="doc_code">
348 %result = <a href="#i_shl">shl</a> i32 %X, i8 3
352 <p>And the hard way:</p>
354 <div class="doc_code">
356 <a href="#i_add">add</a> i32 %X, %X <i>; yields {i32}:%0</i>
357 <a href="#i_add">add</a> i32 %0, %0 <i>; yields {i32}:%1</i>
358 %result = <a href="#i_add">add</a> i32 %1, %1
362 <p>This last way of multiplying <tt>%X</tt> by 8 illustrates several
363 important lexical features of LLVM:</p>
367 <li>Comments are delimited with a '<tt>;</tt>' and go until the end of
370 <li>Unnamed temporaries are created when the result of a computation is not
371 assigned to a named value.</li>
373 <li>Unnamed temporaries are numbered sequentially</li>
377 <p>...and it also shows a convention that we follow in this document. When
378 demonstrating instructions, we will follow an instruction with a comment that
379 defines the type and name of value produced. Comments are shown in italic
384 <!-- *********************************************************************** -->
385 <div class="doc_section"> <a name="highlevel">High Level Structure</a> </div>
386 <!-- *********************************************************************** -->
388 <!-- ======================================================================= -->
389 <div class="doc_subsection"> <a name="modulestructure">Module Structure</a>
392 <div class="doc_text">
394 <p>LLVM programs are composed of "Module"s, each of which is a
395 translation unit of the input programs. Each module consists of
396 functions, global variables, and symbol table entries. Modules may be
397 combined together with the LLVM linker, which merges function (and
398 global variable) definitions, resolves forward declarations, and merges
399 symbol table entries. Here is an example of the "hello world" module:</p>
401 <div class="doc_code">
402 <pre><i>; Declare the string constant as a global constant...</i>
403 <a href="#identifiers">@.LC0</a> = <a href="#linkage_internal">internal</a> <a
404 href="#globalvars">constant</a> <a href="#t_array">[13 x i8]</a> c"hello world\0A\00" <i>; [13 x i8]*</i>
406 <i>; External declaration of the puts function</i>
407 <a href="#functionstructure">declare</a> i32 @puts(i8 *) <i>; i32(i8 *)* </i>
409 <i>; Definition of main function</i>
410 define i32 @main() { <i>; i32()* </i>
411 <i>; Convert [13x i8 ]* to i8 *...</i>
413 href="#i_getelementptr">getelementptr</a> [13 x i8 ]* @.LC0, i64 0, i64 0 <i>; i8 *</i>
415 <i>; Call puts function to write out the string to stdout...</i>
417 href="#i_call">call</a> i32 @puts(i8 * %cast210) <i>; i32</i>
419 href="#i_ret">ret</a> i32 0<br>}<br>
423 <p>This example is made up of a <a href="#globalvars">global variable</a>
424 named "<tt>.LC0</tt>", an external declaration of the "<tt>puts</tt>"
425 function, and a <a href="#functionstructure">function definition</a>
426 for "<tt>main</tt>".</p>
428 <p>In general, a module is made up of a list of global values,
429 where both functions and global variables are global values. Global values are
430 represented by a pointer to a memory location (in this case, a pointer to an
431 array of char, and a pointer to a function), and have one of the following <a
432 href="#linkage">linkage types</a>.</p>
436 <!-- ======================================================================= -->
437 <div class="doc_subsection">
438 <a name="linkage">Linkage Types</a>
441 <div class="doc_text">
444 All Global Variables and Functions have one of the following types of linkage:
449 <dt><tt><b><a name="linkage_internal">internal</a></b></tt> </dt>
451 <dd>Global values with internal linkage are only directly accessible by
452 objects in the current module. In particular, linking code into a module with
453 an internal global value may cause the internal to be renamed as necessary to
454 avoid collisions. Because the symbol is internal to the module, all
455 references can be updated. This corresponds to the notion of the
456 '<tt>static</tt>' keyword in C.
459 <dt><tt><b><a name="linkage_linkonce">linkonce</a></b></tt>: </dt>
461 <dd>Globals with "<tt>linkonce</tt>" linkage are merged with other globals of
462 the same name when linkage occurs. This is typically used to implement
463 inline functions, templates, or other code which must be generated in each
464 translation unit that uses it. Unreferenced <tt>linkonce</tt> globals are
465 allowed to be discarded.
468 <dt><tt><b><a name="linkage_weak">weak</a></b></tt>: </dt>
470 <dd>"<tt>weak</tt>" linkage is exactly the same as <tt>linkonce</tt> linkage,
471 except that unreferenced <tt>weak</tt> globals may not be discarded. This is
472 used for globals that may be emitted in multiple translation units, but that
473 are not guaranteed to be emitted into every translation unit that uses them.
474 One example of this are common globals in C, such as "<tt>int X;</tt>" at
478 <dt><tt><b><a name="linkage_appending">appending</a></b></tt>: </dt>
480 <dd>"<tt>appending</tt>" linkage may only be applied to global variables of
481 pointer to array type. When two global variables with appending linkage are
482 linked together, the two global arrays are appended together. This is the
483 LLVM, typesafe, equivalent of having the system linker append together
484 "sections" with identical names when .o files are linked.
487 <dt><tt><b><a name="linkage_externweak">extern_weak</a></b></tt>: </dt>
488 <dd>The semantics of this linkage follow the ELF model: the symbol is weak
489 until linked, if not linked, the symbol becomes null instead of being an
493 <dt><tt><b><a name="linkage_external">externally visible</a></b></tt>:</dt>
495 <dd>If none of the above identifiers are used, the global is externally
496 visible, meaning that it participates in linkage and can be used to resolve
497 external symbol references.
502 The next two types of linkage are targeted for Microsoft Windows platform
503 only. They are designed to support importing (exporting) symbols from (to)
508 <dt><tt><b><a name="linkage_dllimport">dllimport</a></b></tt>: </dt>
510 <dd>"<tt>dllimport</tt>" linkage causes the compiler to reference a function
511 or variable via a global pointer to a pointer that is set up by the DLL
512 exporting the symbol. On Microsoft Windows targets, the pointer name is
513 formed by combining <code>_imp__</code> and the function or variable name.
516 <dt><tt><b><a name="linkage_dllexport">dllexport</a></b></tt>: </dt>
518 <dd>"<tt>dllexport</tt>" linkage causes the compiler to provide a global
519 pointer to a pointer in a DLL, so that it can be referenced with the
520 <tt>dllimport</tt> attribute. On Microsoft Windows targets, the pointer
521 name is formed by combining <code>_imp__</code> and the function or variable
527 <p><a name="linkage_external"></a>For example, since the "<tt>.LC0</tt>"
528 variable is defined to be internal, if another module defined a "<tt>.LC0</tt>"
529 variable and was linked with this one, one of the two would be renamed,
530 preventing a collision. Since "<tt>main</tt>" and "<tt>puts</tt>" are
531 external (i.e., lacking any linkage declarations), they are accessible
532 outside of the current module.</p>
533 <p>It is illegal for a function <i>declaration</i>
534 to have any linkage type other than "externally visible", <tt>dllimport</tt>,
535 or <tt>extern_weak</tt>.</p>
536 <p>Aliases can have only <tt>external</tt>, <tt>internal</tt> and <tt>weak</tt>
540 <!-- ======================================================================= -->
541 <div class="doc_subsection">
542 <a name="callingconv">Calling Conventions</a>
545 <div class="doc_text">
547 <p>LLVM <a href="#functionstructure">functions</a>, <a href="#i_call">calls</a>
548 and <a href="#i_invoke">invokes</a> can all have an optional calling convention
549 specified for the call. The calling convention of any pair of dynamic
550 caller/callee must match, or the behavior of the program is undefined. The
551 following calling conventions are supported by LLVM, and more may be added in
555 <dt><b>"<tt>ccc</tt>" - The C calling convention</b>:</dt>
557 <dd>This calling convention (the default if no other calling convention is
558 specified) matches the target C calling conventions. This calling convention
559 supports varargs function calls and tolerates some mismatch in the declared
560 prototype and implemented declaration of the function (as does normal C).
563 <dt><b>"<tt>fastcc</tt>" - The fast calling convention</b>:</dt>
565 <dd>This calling convention attempts to make calls as fast as possible
566 (e.g. by passing things in registers). This calling convention allows the
567 target to use whatever tricks it wants to produce fast code for the target,
568 without having to conform to an externally specified ABI. Implementations of
569 this convention should allow arbitrary tail call optimization to be supported.
570 This calling convention does not support varargs and requires the prototype of
571 all callees to exactly match the prototype of the function definition.
574 <dt><b>"<tt>coldcc</tt>" - The cold calling convention</b>:</dt>
576 <dd>This calling convention attempts to make code in the caller as efficient
577 as possible under the assumption that the call is not commonly executed. As
578 such, these calls often preserve all registers so that the call does not break
579 any live ranges in the caller side. This calling convention does not support
580 varargs and requires the prototype of all callees to exactly match the
581 prototype of the function definition.
584 <dt><b>"<tt>cc <<em>n</em>></tt>" - Numbered convention</b>:</dt>
586 <dd>Any calling convention may be specified by number, allowing
587 target-specific calling conventions to be used. Target specific calling
588 conventions start at 64.
592 <p>More calling conventions can be added/defined on an as-needed basis, to
593 support pascal conventions or any other well-known target-independent
598 <!-- ======================================================================= -->
599 <div class="doc_subsection">
600 <a name="visibility">Visibility Styles</a>
603 <div class="doc_text">
606 All Global Variables and Functions have one of the following visibility styles:
610 <dt><b>"<tt>default</tt>" - Default style</b>:</dt>
612 <dd>On ELF, default visibility means that the declaration is visible to other
613 modules and, in shared libraries, means that the declared entity may be
614 overridden. On Darwin, default visibility means that the declaration is
615 visible to other modules. Default visibility corresponds to "external
616 linkage" in the language.
619 <dt><b>"<tt>hidden</tt>" - Hidden style</b>:</dt>
621 <dd>Two declarations of an object with hidden visibility refer to the same
622 object if they are in the same shared object. Usually, hidden visibility
623 indicates that the symbol will not be placed into the dynamic symbol table,
624 so no other module (executable or shared library) can reference it
628 <dt><b>"<tt>protected</tt>" - Protected style</b>:</dt>
630 <dd>On ELF, protected visibility indicates that the symbol will be placed in
631 the dynamic symbol table, but that references within the defining module will
632 bind to the local symbol. That is, the symbol cannot be overridden by another
639 <!-- ======================================================================= -->
640 <div class="doc_subsection">
641 <a name="globalvars">Global Variables</a>
644 <div class="doc_text">
646 <p>Global variables define regions of memory allocated at compilation time
647 instead of run-time. Global variables may optionally be initialized, may have
648 an explicit section to be placed in, and may have an optional explicit alignment
649 specified. A variable may be defined as "thread_local", which means that it
650 will not be shared by threads (each thread will have a separated copy of the
651 variable). A variable may be defined as a global "constant," which indicates
652 that the contents of the variable will <b>never</b> be modified (enabling better
653 optimization, allowing the global data to be placed in the read-only section of
654 an executable, etc). Note that variables that need runtime initialization
655 cannot be marked "constant" as there is a store to the variable.</p>
658 LLVM explicitly allows <em>declarations</em> of global variables to be marked
659 constant, even if the final definition of the global is not. This capability
660 can be used to enable slightly better optimization of the program, but requires
661 the language definition to guarantee that optimizations based on the
662 'constantness' are valid for the translation units that do not include the
666 <p>As SSA values, global variables define pointer values that are in
667 scope (i.e. they dominate) all basic blocks in the program. Global
668 variables always define a pointer to their "content" type because they
669 describe a region of memory, and all memory objects in LLVM are
670 accessed through pointers.</p>
672 <p>A global variable may be declared to reside in a target-specifc numbered
673 address space. For targets that support them, address spaces may affect how
674 optimizations are performed and/or what target instructions are used to access
675 the variable. The default address space is zero. The address space qualifier
676 must precede any other attributes.</p>
678 <p>LLVM allows an explicit section to be specified for globals. If the target
679 supports it, it will emit globals to the section specified.</p>
681 <p>An explicit alignment may be specified for a global. If not present, or if
682 the alignment is set to zero, the alignment of the global is set by the target
683 to whatever it feels convenient. If an explicit alignment is specified, the
684 global is forced to have at least that much alignment. All alignments must be
687 <p>For example, the following defines a global in a numbered address space with
688 an initializer, section, and alignment:</p>
690 <div class="doc_code">
692 @G = constant float 1.0 addrspace(5), section "foo", align 4
699 <!-- ======================================================================= -->
700 <div class="doc_subsection">
701 <a name="functionstructure">Functions</a>
704 <div class="doc_text">
706 <p>LLVM function definitions consist of the "<tt>define</tt>" keyord,
707 an optional <a href="#linkage">linkage type</a>, an optional
708 <a href="#visibility">visibility style</a>, an optional
709 <a href="#callingconv">calling convention</a>, a return type, an optional
710 <a href="#paramattrs">parameter attribute</a> for the return type, a function
711 name, a (possibly empty) argument list (each with optional
712 <a href="#paramattrs">parameter attributes</a>), an optional section, an
713 optional alignment, an optional <a href="#gc">garbage collector name</a>, an
714 opening curly brace, a list of basic blocks, and a closing curly brace.
716 LLVM function declarations consist of the "<tt>declare</tt>" keyword, an
717 optional <a href="#linkage">linkage type</a>, an optional
718 <a href="#visibility">visibility style</a>, an optional
719 <a href="#callingconv">calling convention</a>, a return type, an optional
720 <a href="#paramattrs">parameter attribute</a> for the return type, a function
721 name, a possibly empty list of arguments, an optional alignment, and an optional
722 <a href="#gc">garbage collector name</a>.</p>
724 <p>A function definition contains a list of basic blocks, forming the CFG for
725 the function. Each basic block may optionally start with a label (giving the
726 basic block a symbol table entry), contains a list of instructions, and ends
727 with a <a href="#terminators">terminator</a> instruction (such as a branch or
728 function return).</p>
730 <p>The first basic block in a function is special in two ways: it is immediately
731 executed on entrance to the function, and it is not allowed to have predecessor
732 basic blocks (i.e. there can not be any branches to the entry block of a
733 function). Because the block can have no predecessors, it also cannot have any
734 <a href="#i_phi">PHI nodes</a>.</p>
736 <p>LLVM allows an explicit section to be specified for functions. If the target
737 supports it, it will emit functions to the section specified.</p>
739 <p>An explicit alignment may be specified for a function. If not present, or if
740 the alignment is set to zero, the alignment of the function is set by the target
741 to whatever it feels convenient. If an explicit alignment is specified, the
742 function is forced to have at least that much alignment. All alignments must be
748 <!-- ======================================================================= -->
749 <div class="doc_subsection">
750 <a name="aliasstructure">Aliases</a>
752 <div class="doc_text">
753 <p>Aliases act as "second name" for the aliasee value (which can be either
754 function or global variable or bitcast of global value). Aliases may have an
755 optional <a href="#linkage">linkage type</a>, and an
756 optional <a href="#visibility">visibility style</a>.</p>
760 <div class="doc_code">
762 @<Name> = [Linkage] [Visibility] alias <AliaseeTy> @<Aliasee>
770 <!-- ======================================================================= -->
771 <div class="doc_subsection"><a name="paramattrs">Parameter Attributes</a></div>
772 <div class="doc_text">
773 <p>The return type and each parameter of a function type may have a set of
774 <i>parameter attributes</i> associated with them. Parameter attributes are
775 used to communicate additional information about the result or parameters of
776 a function. Parameter attributes are considered to be part of the function,
777 not of the function type, so functions with different parameter attributes
778 can have the same function type.</p>
780 <p>Parameter attributes are simple keywords that follow the type specified. If
781 multiple parameter attributes are needed, they are space separated. For
784 <div class="doc_code">
786 declare i32 @printf(i8* noalias , ...) nounwind
787 declare i32 @atoi(i8*) nounwind readonly
791 <p>Note that any attributes for the function result (<tt>nounwind</tt>,
792 <tt>readonly</tt>) come immediately after the argument list.</p>
794 <p>Currently, only the following parameter attributes are defined:</p>
796 <dt><tt>zeroext</tt></dt>
797 <dd>This indicates that the parameter should be zero extended just before
798 a call to this function.</dd>
799 <dt><tt>signext</tt></dt>
800 <dd>This indicates that the parameter should be sign extended just before
801 a call to this function.</dd>
802 <dt><tt>inreg</tt></dt>
803 <dd>This indicates that the parameter should be placed in register (if
804 possible) during assembling function call. Support for this attribute is
806 <dt><tt>sret</tt></dt>
807 <dd>This indicates that the parameter specifies the address of a structure
808 that is the return value of the function in the source program.</dd>
809 <dt><tt>noalias</tt></dt>
810 <dd>This indicates that the parameter not alias any other object or any
811 other "noalias" objects during the function call.
812 <dt><tt>noreturn</tt></dt>
813 <dd>This function attribute indicates that the function never returns. This
814 indicates to LLVM that every call to this function should be treated as if
815 an <tt>unreachable</tt> instruction immediately followed the call.</dd>
816 <dt><tt>nounwind</tt></dt>
817 <dd>This function attribute indicates that the function type does not use
818 the unwind instruction and does not allow stack unwinding to propagate
820 <dt><tt>nest</tt></dt>
821 <dd>This indicates that the parameter can be excised using the
822 <a href="#int_trampoline">trampoline intrinsics</a>.</dd>
823 <dt><tt>readonly</tt></dt>
824 <dd>This function attribute indicates that the function has no side-effects
825 except for producing a return value or throwing an exception. The value
826 returned must only depend on the function arguments and/or global variables.
827 It may use values obtained by dereferencing pointers.</dd>
828 <dt><tt>readnone</tt></dt>
829 <dd>A <tt>readnone</tt> function has the same restrictions as a <tt>readonly</tt>
830 function, but in addition it is not allowed to dereference any pointer arguments
836 <!-- ======================================================================= -->
837 <div class="doc_subsection">
838 <a name="gc">Garbage Collector Names</a>
841 <div class="doc_text">
842 <p>Each function may specify a garbage collector name, which is simply a
845 <div class="doc_code"><pre
846 >define void @f() gc "name" { ...</pre></div>
848 <p>The compiler declares the supported values of <i>name</i>. Specifying a
849 collector which will cause the compiler to alter its output in order to support
850 the named garbage collection algorithm.</p>
853 <!-- ======================================================================= -->
854 <div class="doc_subsection">
855 <a name="moduleasm">Module-Level Inline Assembly</a>
858 <div class="doc_text">
860 Modules may contain "module-level inline asm" blocks, which corresponds to the
861 GCC "file scope inline asm" blocks. These blocks are internally concatenated by
862 LLVM and treated as a single unit, but may be separated in the .ll file if
863 desired. The syntax is very simple:
866 <div class="doc_code">
868 module asm "inline asm code goes here"
869 module asm "more can go here"
873 <p>The strings can contain any character by escaping non-printable characters.
874 The escape sequence used is simply "\xx" where "xx" is the two digit hex code
879 The inline asm code is simply printed to the machine code .s file when
880 assembly code is generated.
884 <!-- ======================================================================= -->
885 <div class="doc_subsection">
886 <a name="datalayout">Data Layout</a>
889 <div class="doc_text">
890 <p>A module may specify a target specific data layout string that specifies how
891 data is to be laid out in memory. The syntax for the data layout is simply:</p>
892 <pre> target datalayout = "<i>layout specification</i>"</pre>
893 <p>The <i>layout specification</i> consists of a list of specifications
894 separated by the minus sign character ('-'). Each specification starts with a
895 letter and may include other information after the letter to define some
896 aspect of the data layout. The specifications accepted are as follows: </p>
899 <dd>Specifies that the target lays out data in big-endian form. That is, the
900 bits with the most significance have the lowest address location.</dd>
902 <dd>Specifies that hte target lays out data in little-endian form. That is,
903 the bits with the least significance have the lowest address location.</dd>
904 <dt><tt>p:<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
905 <dd>This specifies the <i>size</i> of a pointer and its <i>abi</i> and
906 <i>preferred</i> alignments. All sizes are in bits. Specifying the <i>pref</i>
907 alignment is optional. If omitted, the preceding <tt>:</tt> should be omitted
909 <dt><tt>i<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
910 <dd>This specifies the alignment for an integer type of a given bit
911 <i>size</i>. The value of <i>size</i> must be in the range [1,2^23).</dd>
912 <dt><tt>v<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
913 <dd>This specifies the alignment for a vector type of a given bit
915 <dt><tt>f<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
916 <dd>This specifies the alignment for a floating point type of a given bit
917 <i>size</i>. The value of <i>size</i> must be either 32 (float) or 64
919 <dt><tt>a<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
920 <dd>This specifies the alignment for an aggregate type of a given bit
923 <p>When constructing the data layout for a given target, LLVM starts with a
924 default set of specifications which are then (possibly) overriden by the
925 specifications in the <tt>datalayout</tt> keyword. The default specifications
926 are given in this list:</p>
928 <li><tt>E</tt> - big endian</li>
929 <li><tt>p:32:64:64</tt> - 32-bit pointers with 64-bit alignment</li>
930 <li><tt>i1:8:8</tt> - i1 is 8-bit (byte) aligned</li>
931 <li><tt>i8:8:8</tt> - i8 is 8-bit (byte) aligned</li>
932 <li><tt>i16:16:16</tt> - i16 is 16-bit aligned</li>
933 <li><tt>i32:32:32</tt> - i32 is 32-bit aligned</li>
934 <li><tt>i64:32:64</tt> - i64 has abi alignment of 32-bits but preferred
935 alignment of 64-bits</li>
936 <li><tt>f32:32:32</tt> - float is 32-bit aligned</li>
937 <li><tt>f64:64:64</tt> - double is 64-bit aligned</li>
938 <li><tt>v64:64:64</tt> - 64-bit vector is 64-bit aligned</li>
939 <li><tt>v128:128:128</tt> - 128-bit vector is 128-bit aligned</li>
940 <li><tt>a0:0:1</tt> - aggregates are 8-bit aligned</li>
942 <p>When llvm is determining the alignment for a given type, it uses the
945 <li>If the type sought is an exact match for one of the specifications, that
946 specification is used.</li>
947 <li>If no match is found, and the type sought is an integer type, then the
948 smallest integer type that is larger than the bitwidth of the sought type is
949 used. If none of the specifications are larger than the bitwidth then the the
950 largest integer type is used. For example, given the default specifications
951 above, the i7 type will use the alignment of i8 (next largest) while both
952 i65 and i256 will use the alignment of i64 (largest specified).</li>
953 <li>If no match is found, and the type sought is a vector type, then the
954 largest vector type that is smaller than the sought vector type will be used
955 as a fall back. This happens because <128 x double> can be implemented in
956 terms of 64 <2 x double>, for example.</li>
960 <!-- *********************************************************************** -->
961 <div class="doc_section"> <a name="typesystem">Type System</a> </div>
962 <!-- *********************************************************************** -->
964 <div class="doc_text">
966 <p>The LLVM type system is one of the most important features of the
967 intermediate representation. Being typed enables a number of
968 optimizations to be performed on the IR directly, without having to do
969 extra analyses on the side before the transformation. A strong type
970 system makes it easier to read the generated code and enables novel
971 analyses and transformations that are not feasible to perform on normal
972 three address code representations.</p>
976 <!-- ======================================================================= -->
977 <div class="doc_subsection"> <a name="t_primitive">Primitive Types</a> </div>
978 <div class="doc_text">
979 <p>The primitive types are the fundamental building blocks of the LLVM
980 system. The current set of primitive types is as follows:</p>
982 <table class="layout">
987 <tr><th>Type</th><th>Description</th></tr>
988 <tr><td><tt><a name="t_void">void</a></tt></td><td>No value</td></tr>
989 <tr><td><tt>label</tt></td><td>Branch destination</td></tr>
996 <tr><th>Type</th><th>Description</th></tr>
997 <tr><td><tt>float</tt></td><td>32-bit floating point value</td></tr>
998 <tr><td><tt>double</tt></td><td>64-bit floating point value</td></tr>
1006 <!-- _______________________________________________________________________ -->
1007 <div class="doc_subsubsection"> <a name="t_classifications">Type
1008 Classifications</a> </div>
1009 <div class="doc_text">
1010 <p>These different primitive types fall into a few useful
1011 classifications:</p>
1013 <table border="1" cellspacing="0" cellpadding="4">
1015 <tr><th>Classification</th><th>Types</th></tr>
1017 <td><a name="t_integer">integer</a></td>
1018 <td><tt>i1, i2, i3, ... i8, ... i16, ... i32, ... i64, ... </tt></td>
1021 <td><a name="t_floating">floating point</a></td>
1022 <td><tt>float, double</tt></td>
1025 <td><a name="t_firstclass">first class</a></td>
1026 <td><tt>i1, ..., float, double, <br/>
1027 <a href="#t_pointer">pointer</a>,<a href="#t_vector">vector</a></tt>
1033 <p>The <a href="#t_firstclass">first class</a> types are perhaps the
1034 most important. Values of these types are the only ones which can be
1035 produced by instructions, passed as arguments, or used as operands to
1036 instructions. This means that all structures and arrays must be
1037 manipulated either by pointer or by component.</p>
1040 <!-- ======================================================================= -->
1041 <div class="doc_subsection"> <a name="t_derived">Derived Types</a> </div>
1043 <div class="doc_text">
1045 <p>The real power in LLVM comes from the derived types in the system.
1046 This is what allows a programmer to represent arrays, functions,
1047 pointers, and other useful types. Note that these derived types may be
1048 recursive: For example, it is possible to have a two dimensional array.</p>
1052 <!-- _______________________________________________________________________ -->
1053 <div class="doc_subsubsection"> <a name="t_integer">Integer Type</a> </div>
1055 <div class="doc_text">
1058 <p>The integer type is a very simple derived type that simply specifies an
1059 arbitrary bit width for the integer type desired. Any bit width from 1 bit to
1060 2^23-1 (about 8 million) can be specified.</p>
1068 <p>The number of bits the integer will occupy is specified by the <tt>N</tt>
1072 <table class="layout">
1075 <td><tt>i1</tt></td>
1076 <td>a single-bit integer.</td>
1078 <td><tt>i32</tt></td>
1079 <td>a 32-bit integer.</td>
1081 <td><tt>i1942652</tt></td>
1082 <td>a really big integer of over 1 million bits.</td>
1088 <!-- _______________________________________________________________________ -->
1089 <div class="doc_subsubsection"> <a name="t_array">Array Type</a> </div>
1091 <div class="doc_text">
1095 <p>The array type is a very simple derived type that arranges elements
1096 sequentially in memory. The array type requires a size (number of
1097 elements) and an underlying data type.</p>
1102 [<# elements> x <elementtype>]
1105 <p>The number of elements is a constant integer value; elementtype may
1106 be any type with a size.</p>
1109 <table class="layout">
1111 <td class="left"><tt>[40 x i32]</tt></td>
1112 <td class="left">Array of 40 32-bit integer values.</td>
1115 <td class="left"><tt>[41 x i32]</tt></td>
1116 <td class="left">Array of 41 32-bit integer values.</td>
1119 <td class="left"><tt>[4 x i8]</tt></td>
1120 <td class="left">Array of 4 8-bit integer values.</td>
1123 <p>Here are some examples of multidimensional arrays:</p>
1124 <table class="layout">
1126 <td class="left"><tt>[3 x [4 x i32]]</tt></td>
1127 <td class="left">3x4 array of 32-bit integer values.</td>
1130 <td class="left"><tt>[12 x [10 x float]]</tt></td>
1131 <td class="left">12x10 array of single precision floating point values.</td>
1134 <td class="left"><tt>[2 x [3 x [4 x i16]]]</tt></td>
1135 <td class="left">2x3x4 array of 16-bit integer values.</td>
1139 <p>Note that 'variable sized arrays' can be implemented in LLVM with a zero
1140 length array. Normally, accesses past the end of an array are undefined in
1141 LLVM (e.g. it is illegal to access the 5th element of a 3 element array).
1142 As a special case, however, zero length arrays are recognized to be variable
1143 length. This allows implementation of 'pascal style arrays' with the LLVM
1144 type "{ i32, [0 x float]}", for example.</p>
1148 <!-- _______________________________________________________________________ -->
1149 <div class="doc_subsubsection"> <a name="t_function">Function Type</a> </div>
1150 <div class="doc_text">
1152 <p>The function type can be thought of as a function signature. It
1153 consists of a return type and a list of formal parameter types.
1154 Function types are usually used to build virtual function tables
1155 (which are structures of pointers to functions), for indirect function
1156 calls, and when defining a function.</p>
1158 The return type of a function type cannot be an aggregate type.
1161 <pre> <returntype> (<parameter list>)<br></pre>
1162 <p>...where '<tt><parameter list></tt>' is a comma-separated list of type
1163 specifiers. Optionally, the parameter list may include a type <tt>...</tt>,
1164 which indicates that the function takes a variable number of arguments.
1165 Variable argument functions can access their arguments with the <a
1166 href="#int_varargs">variable argument handling intrinsic</a> functions.</p>
1168 <table class="layout">
1170 <td class="left"><tt>i32 (i32)</tt></td>
1171 <td class="left">function taking an <tt>i32</tt>, returning an <tt>i32</tt>
1173 </tr><tr class="layout">
1174 <td class="left"><tt>float (i16 signext, i32 *) *
1176 <td class="left"><a href="#t_pointer">Pointer</a> to a function that takes
1177 an <tt>i16</tt> that should be sign extended and a
1178 <a href="#t_pointer">pointer</a> to <tt>i32</tt>, returning
1181 </tr><tr class="layout">
1182 <td class="left"><tt>i32 (i8*, ...)</tt></td>
1183 <td class="left">A vararg function that takes at least one
1184 <a href="#t_pointer">pointer</a> to <tt>i8 </tt> (char in C),
1185 which returns an integer. This is the signature for <tt>printf</tt> in
1192 <!-- _______________________________________________________________________ -->
1193 <div class="doc_subsubsection"> <a name="t_struct">Structure Type</a> </div>
1194 <div class="doc_text">
1196 <p>The structure type is used to represent a collection of data members
1197 together in memory. The packing of the field types is defined to match
1198 the ABI of the underlying processor. The elements of a structure may
1199 be any type that has a size.</p>
1200 <p>Structures are accessed using '<tt><a href="#i_load">load</a></tt>
1201 and '<tt><a href="#i_store">store</a></tt>' by getting a pointer to a
1202 field with the '<tt><a href="#i_getelementptr">getelementptr</a></tt>'
1205 <pre> { <type list> }<br></pre>
1207 <table class="layout">
1209 <td class="left"><tt>{ i32, i32, i32 }</tt></td>
1210 <td class="left">A triple of three <tt>i32</tt> values</td>
1211 </tr><tr class="layout">
1212 <td class="left"><tt>{ float, i32 (i32) * }</tt></td>
1213 <td class="left">A pair, where the first element is a <tt>float</tt> and the
1214 second element is a <a href="#t_pointer">pointer</a> to a
1215 <a href="#t_function">function</a> that takes an <tt>i32</tt>, returning
1216 an <tt>i32</tt>.</td>
1221 <!-- _______________________________________________________________________ -->
1222 <div class="doc_subsubsection"> <a name="t_pstruct">Packed Structure Type</a>
1224 <div class="doc_text">
1226 <p>The packed structure type is used to represent a collection of data members
1227 together in memory. There is no padding between fields. Further, the alignment
1228 of a packed structure is 1 byte. The elements of a packed structure may
1229 be any type that has a size.</p>
1230 <p>Structures are accessed using '<tt><a href="#i_load">load</a></tt>
1231 and '<tt><a href="#i_store">store</a></tt>' by getting a pointer to a
1232 field with the '<tt><a href="#i_getelementptr">getelementptr</a></tt>'
1235 <pre> < { <type list> } > <br></pre>
1237 <table class="layout">
1239 <td class="left"><tt>< { i32, i32, i32 } ></tt></td>
1240 <td class="left">A triple of three <tt>i32</tt> values</td>
1241 </tr><tr class="layout">
1242 <td class="left"><tt>< { float, i32 (i32)* } ></tt></td>
1243 <td class="left">A pair, where the first element is a <tt>float</tt> and the
1244 second element is a <a href="#t_pointer">pointer</a> to a
1245 <a href="#t_function">function</a> that takes an <tt>i32</tt>, returning
1246 an <tt>i32</tt>.</td>
1251 <!-- _______________________________________________________________________ -->
1252 <div class="doc_subsubsection"> <a name="t_pointer">Pointer Type</a> </div>
1253 <div class="doc_text">
1255 <p>As in many languages, the pointer type represents a pointer or
1256 reference to another object, which must live in memory. Pointer types may have
1257 an optional address space attribute defining the target-specific numbered
1258 address space where the pointed-to object resides. The default address space is
1261 <pre> <type> *<br></pre>
1263 <table class="layout">
1265 <td class="left"><tt>[4x i32]*</tt></td>
1266 <td class="left">A <a href="#t_pointer">pointer</a> to <a
1267 href="#t_array">array</a> of four <tt>i32</tt> values.</td>
1270 <td class="left"><tt>i32 (i32 *) *</tt></td>
1271 <td class="left"> A <a href="#t_pointer">pointer</a> to a <a
1272 href="#t_function">function</a> that takes an <tt>i32*</tt>, returning an
1276 <td class="left"><tt>i32 addrspace(5)*</tt></td>
1277 <td class="left">A <a href="#t_pointer">pointer</a> to an <tt>i32</tt> value
1278 that resides in address space #5.</td>
1283 <!-- _______________________________________________________________________ -->
1284 <div class="doc_subsubsection"> <a name="t_vector">Vector Type</a> </div>
1285 <div class="doc_text">
1289 <p>A vector type is a simple derived type that represents a vector
1290 of elements. Vector types are used when multiple primitive data
1291 are operated in parallel using a single instruction (SIMD).
1292 A vector type requires a size (number of
1293 elements) and an underlying primitive data type. Vectors must have a power
1294 of two length (1, 2, 4, 8, 16 ...). Vector types are
1295 considered <a href="#t_firstclass">first class</a>.</p>
1300 < <# elements> x <elementtype> >
1303 <p>The number of elements is a constant integer value; elementtype may
1304 be any integer or floating point type.</p>
1308 <table class="layout">
1310 <td class="left"><tt><4 x i32></tt></td>
1311 <td class="left">Vector of 4 32-bit integer values.</td>
1314 <td class="left"><tt><8 x float></tt></td>
1315 <td class="left">Vector of 8 32-bit floating-point values.</td>
1318 <td class="left"><tt><2 x i64></tt></td>
1319 <td class="left">Vector of 2 64-bit integer values.</td>
1324 <!-- _______________________________________________________________________ -->
1325 <div class="doc_subsubsection"> <a name="t_opaque">Opaque Type</a> </div>
1326 <div class="doc_text">
1330 <p>Opaque types are used to represent unknown types in the system. This
1331 corresponds (for example) to the C notion of a forward declared structure type.
1332 In LLVM, opaque types can eventually be resolved to any type (not just a
1333 structure type).</p>
1343 <table class="layout">
1345 <td class="left"><tt>opaque</tt></td>
1346 <td class="left">An opaque type.</td>
1352 <!-- *********************************************************************** -->
1353 <div class="doc_section"> <a name="constants">Constants</a> </div>
1354 <!-- *********************************************************************** -->
1356 <div class="doc_text">
1358 <p>LLVM has several different basic types of constants. This section describes
1359 them all and their syntax.</p>
1363 <!-- ======================================================================= -->
1364 <div class="doc_subsection"><a name="simpleconstants">Simple Constants</a></div>
1366 <div class="doc_text">
1369 <dt><b>Boolean constants</b></dt>
1371 <dd>The two strings '<tt>true</tt>' and '<tt>false</tt>' are both valid
1372 constants of the <tt><a href="#t_primitive">i1</a></tt> type.
1375 <dt><b>Integer constants</b></dt>
1377 <dd>Standard integers (such as '4') are constants of the <a
1378 href="#t_integer">integer</a> type. Negative numbers may be used with
1382 <dt><b>Floating point constants</b></dt>
1384 <dd>Floating point constants use standard decimal notation (e.g. 123.421),
1385 exponential notation (e.g. 1.23421e+2), or a more precise hexadecimal
1386 notation (see below). Floating point constants must have a <a
1387 href="#t_floating">floating point</a> type. </dd>
1389 <dt><b>Null pointer constants</b></dt>
1391 <dd>The identifier '<tt>null</tt>' is recognized as a null pointer constant
1392 and must be of <a href="#t_pointer">pointer type</a>.</dd>
1396 <p>The one non-intuitive notation for constants is the optional hexadecimal form
1397 of floating point constants. For example, the form '<tt>double
1398 0x432ff973cafa8000</tt>' is equivalent to (but harder to read than) '<tt>double
1399 4.5e+15</tt>'. The only time hexadecimal floating point constants are required
1400 (and the only time that they are generated by the disassembler) is when a
1401 floating point constant must be emitted but it cannot be represented as a
1402 decimal floating point number. For example, NaN's, infinities, and other
1403 special values are represented in their IEEE hexadecimal format so that
1404 assembly and disassembly do not cause any bits to change in the constants.</p>
1408 <!-- ======================================================================= -->
1409 <div class="doc_subsection"><a name="aggregateconstants">Aggregate Constants</a>
1412 <div class="doc_text">
1413 <p>Aggregate constants arise from aggregation of simple constants
1414 and smaller aggregate constants.</p>
1417 <dt><b>Structure constants</b></dt>
1419 <dd>Structure constants are represented with notation similar to structure
1420 type definitions (a comma separated list of elements, surrounded by braces
1421 (<tt>{}</tt>)). For example: "<tt>{ i32 4, float 17.0, i32* @G }</tt>",
1422 where "<tt>@G</tt>" is declared as "<tt>@G = external global i32</tt>". Structure constants
1423 must have <a href="#t_struct">structure type</a>, and the number and
1424 types of elements must match those specified by the type.
1427 <dt><b>Array constants</b></dt>
1429 <dd>Array constants are represented with notation similar to array type
1430 definitions (a comma separated list of elements, surrounded by square brackets
1431 (<tt>[]</tt>)). For example: "<tt>[ i32 42, i32 11, i32 74 ]</tt>". Array
1432 constants must have <a href="#t_array">array type</a>, and the number and
1433 types of elements must match those specified by the type.
1436 <dt><b>Vector constants</b></dt>
1438 <dd>Vector constants are represented with notation similar to vector type
1439 definitions (a comma separated list of elements, surrounded by
1440 less-than/greater-than's (<tt><></tt>)). For example: "<tt>< i32 42,
1441 i32 11, i32 74, i32 100 ></tt>". Vector constants must have <a
1442 href="#t_vector">vector type</a>, and the number and types of elements must
1443 match those specified by the type.
1446 <dt><b>Zero initialization</b></dt>
1448 <dd>The string '<tt>zeroinitializer</tt>' can be used to zero initialize a
1449 value to zero of <em>any</em> type, including scalar and aggregate types.
1450 This is often used to avoid having to print large zero initializers (e.g. for
1451 large arrays) and is always exactly equivalent to using explicit zero
1458 <!-- ======================================================================= -->
1459 <div class="doc_subsection">
1460 <a name="globalconstants">Global Variable and Function Addresses</a>
1463 <div class="doc_text">
1465 <p>The addresses of <a href="#globalvars">global variables</a> and <a
1466 href="#functionstructure">functions</a> are always implicitly valid (link-time)
1467 constants. These constants are explicitly referenced when the <a
1468 href="#identifiers">identifier for the global</a> is used and always have <a
1469 href="#t_pointer">pointer</a> type. For example, the following is a legal LLVM
1472 <div class="doc_code">
1476 @Z = global [2 x i32*] [ i32* @X, i32* @Y ]
1482 <!-- ======================================================================= -->
1483 <div class="doc_subsection"><a name="undefvalues">Undefined Values</a></div>
1484 <div class="doc_text">
1485 <p>The string '<tt>undef</tt>' is recognized as a type-less constant that has
1486 no specific value. Undefined values may be of any type and be used anywhere
1487 a constant is permitted.</p>
1489 <p>Undefined values indicate to the compiler that the program is well defined
1490 no matter what value is used, giving the compiler more freedom to optimize.
1494 <!-- ======================================================================= -->
1495 <div class="doc_subsection"><a name="constantexprs">Constant Expressions</a>
1498 <div class="doc_text">
1500 <p>Constant expressions are used to allow expressions involving other constants
1501 to be used as constants. Constant expressions may be of any <a
1502 href="#t_firstclass">first class</a> type and may involve any LLVM operation
1503 that does not have side effects (e.g. load and call are not supported). The
1504 following is the syntax for constant expressions:</p>
1507 <dt><b><tt>trunc ( CST to TYPE )</tt></b></dt>
1508 <dd>Truncate a constant to another type. The bit size of CST must be larger
1509 than the bit size of TYPE. Both types must be integers.</dd>
1511 <dt><b><tt>zext ( CST to TYPE )</tt></b></dt>
1512 <dd>Zero extend a constant to another type. The bit size of CST must be
1513 smaller or equal to the bit size of TYPE. Both types must be integers.</dd>
1515 <dt><b><tt>sext ( CST to TYPE )</tt></b></dt>
1516 <dd>Sign extend a constant to another type. The bit size of CST must be
1517 smaller or equal to the bit size of TYPE. Both types must be integers.</dd>
1519 <dt><b><tt>fptrunc ( CST to TYPE )</tt></b></dt>
1520 <dd>Truncate a floating point constant to another floating point type. The
1521 size of CST must be larger than the size of TYPE. Both types must be
1522 floating point.</dd>
1524 <dt><b><tt>fpext ( CST to TYPE )</tt></b></dt>
1525 <dd>Floating point extend a constant to another type. The size of CST must be
1526 smaller or equal to the size of TYPE. Both types must be floating point.</dd>
1528 <dt><b><tt>fptoui ( CST to TYPE )</tt></b></dt>
1529 <dd>Convert a floating point constant to the corresponding unsigned integer
1530 constant. TYPE must be a scalar or vector integer type. CST must be of scalar
1531 or vector floating point type. Both CST and TYPE must be scalars, or vectors
1532 of the same number of elements. If the value won't fit in the integer type,
1533 the results are undefined.</dd>
1535 <dt><b><tt>fptosi ( CST to TYPE )</tt></b></dt>
1536 <dd>Convert a floating point constant to the corresponding signed integer
1537 constant. TYPE must be a scalar or vector integer type. CST must be of scalar
1538 or vector floating point type. Both CST and TYPE must be scalars, or vectors
1539 of the same number of elements. If the value won't fit in the integer type,
1540 the results are undefined.</dd>
1542 <dt><b><tt>uitofp ( CST to TYPE )</tt></b></dt>
1543 <dd>Convert an unsigned integer constant to the corresponding floating point
1544 constant. TYPE must be a scalar or vector floating point type. CST must be of
1545 scalar or vector integer type. Both CST and TYPE must be scalars, or vectors
1546 of the same number of elements. If the value won't fit in the floating point
1547 type, the results are undefined.</dd>
1549 <dt><b><tt>sitofp ( CST to TYPE )</tt></b></dt>
1550 <dd>Convert a signed integer constant to the corresponding floating point
1551 constant. TYPE must be a scalar or vector floating point type. CST must be of
1552 scalar or vector integer type. Both CST and TYPE must be scalars, or vectors
1553 of the same number of elements. If the value won't fit in the floating point
1554 type, the results are undefined.</dd>
1556 <dt><b><tt>ptrtoint ( CST to TYPE )</tt></b></dt>
1557 <dd>Convert a pointer typed constant to the corresponding integer constant
1558 TYPE must be an integer type. CST must be of pointer type. The CST value is
1559 zero extended, truncated, or unchanged to make it fit in TYPE.</dd>
1561 <dt><b><tt>inttoptr ( CST to TYPE )</tt></b></dt>
1562 <dd>Convert a integer constant to a pointer constant. TYPE must be a
1563 pointer type. CST must be of integer type. The CST value is zero extended,
1564 truncated, or unchanged to make it fit in a pointer size. This one is
1565 <i>really</i> dangerous!</dd>
1567 <dt><b><tt>bitcast ( CST to TYPE )</tt></b></dt>
1568 <dd>Convert a constant, CST, to another TYPE. The size of CST and TYPE must be
1569 identical (same number of bits). The conversion is done as if the CST value
1570 was stored to memory and read back as TYPE. In other words, no bits change
1571 with this operator, just the type. This can be used for conversion of
1572 vector types to any other type, as long as they have the same bit width. For
1573 pointers it is only valid to cast to another pointer type.
1576 <dt><b><tt>getelementptr ( CSTPTR, IDX0, IDX1, ... )</tt></b></dt>
1578 <dd>Perform the <a href="#i_getelementptr">getelementptr operation</a> on
1579 constants. As with the <a href="#i_getelementptr">getelementptr</a>
1580 instruction, the index list may have zero or more indexes, which are required
1581 to make sense for the type of "CSTPTR".</dd>
1583 <dt><b><tt>select ( COND, VAL1, VAL2 )</tt></b></dt>
1585 <dd>Perform the <a href="#i_select">select operation</a> on
1588 <dt><b><tt>icmp COND ( VAL1, VAL2 )</tt></b></dt>
1589 <dd>Performs the <a href="#i_icmp">icmp operation</a> on constants.</dd>
1591 <dt><b><tt>fcmp COND ( VAL1, VAL2 )</tt></b></dt>
1592 <dd>Performs the <a href="#i_fcmp">fcmp operation</a> on constants.</dd>
1594 <dt><b><tt>extractelement ( VAL, IDX )</tt></b></dt>
1596 <dd>Perform the <a href="#i_extractelement">extractelement
1597 operation</a> on constants.
1599 <dt><b><tt>insertelement ( VAL, ELT, IDX )</tt></b></dt>
1601 <dd>Perform the <a href="#i_insertelement">insertelement
1602 operation</a> on constants.</dd>
1605 <dt><b><tt>shufflevector ( VEC1, VEC2, IDXMASK )</tt></b></dt>
1607 <dd>Perform the <a href="#i_shufflevector">shufflevector
1608 operation</a> on constants.</dd>
1610 <dt><b><tt>OPCODE ( LHS, RHS )</tt></b></dt>
1612 <dd>Perform the specified operation of the LHS and RHS constants. OPCODE may
1613 be any of the <a href="#binaryops">binary</a> or <a href="#bitwiseops">bitwise
1614 binary</a> operations. The constraints on operands are the same as those for
1615 the corresponding instruction (e.g. no bitwise operations on floating point
1616 values are allowed).</dd>
1620 <!-- *********************************************************************** -->
1621 <div class="doc_section"> <a name="othervalues">Other Values</a> </div>
1622 <!-- *********************************************************************** -->
1624 <!-- ======================================================================= -->
1625 <div class="doc_subsection">
1626 <a name="inlineasm">Inline Assembler Expressions</a>
1629 <div class="doc_text">
1632 LLVM supports inline assembler expressions (as opposed to <a href="#moduleasm">
1633 Module-Level Inline Assembly</a>) through the use of a special value. This
1634 value represents the inline assembler as a string (containing the instructions
1635 to emit), a list of operand constraints (stored as a string), and a flag that
1636 indicates whether or not the inline asm expression has side effects. An example
1637 inline assembler expression is:
1640 <div class="doc_code">
1642 i32 (i32) asm "bswap $0", "=r,r"
1647 Inline assembler expressions may <b>only</b> be used as the callee operand of
1648 a <a href="#i_call"><tt>call</tt> instruction</a>. Thus, typically we have:
1651 <div class="doc_code">
1653 %X = call i32 asm "<a href="#int_bswap">bswap</a> $0", "=r,r"(i32 %Y)
1658 Inline asms with side effects not visible in the constraint list must be marked
1659 as having side effects. This is done through the use of the
1660 '<tt>sideeffect</tt>' keyword, like so:
1663 <div class="doc_code">
1665 call void asm sideeffect "eieio", ""()
1669 <p>TODO: The format of the asm and constraints string still need to be
1670 documented here. Constraints on what can be done (e.g. duplication, moving, etc
1671 need to be documented).
1676 <!-- *********************************************************************** -->
1677 <div class="doc_section"> <a name="instref">Instruction Reference</a> </div>
1678 <!-- *********************************************************************** -->
1680 <div class="doc_text">
1682 <p>The LLVM instruction set consists of several different
1683 classifications of instructions: <a href="#terminators">terminator
1684 instructions</a>, <a href="#binaryops">binary instructions</a>,
1685 <a href="#bitwiseops">bitwise binary instructions</a>, <a
1686 href="#memoryops">memory instructions</a>, and <a href="#otherops">other
1687 instructions</a>.</p>
1691 <!-- ======================================================================= -->
1692 <div class="doc_subsection"> <a name="terminators">Terminator
1693 Instructions</a> </div>
1695 <div class="doc_text">
1697 <p>As mentioned <a href="#functionstructure">previously</a>, every
1698 basic block in a program ends with a "Terminator" instruction, which
1699 indicates which block should be executed after the current block is
1700 finished. These terminator instructions typically yield a '<tt>void</tt>'
1701 value: they produce control flow, not values (the one exception being
1702 the '<a href="#i_invoke"><tt>invoke</tt></a>' instruction).</p>
1703 <p>There are six different terminator instructions: the '<a
1704 href="#i_ret"><tt>ret</tt></a>' instruction, the '<a href="#i_br"><tt>br</tt></a>'
1705 instruction, the '<a href="#i_switch"><tt>switch</tt></a>' instruction,
1706 the '<a href="#i_invoke"><tt>invoke</tt></a>' instruction, the '<a
1707 href="#i_unwind"><tt>unwind</tt></a>' instruction, and the '<a
1708 href="#i_unreachable"><tt>unreachable</tt></a>' instruction.</p>
1712 <!-- _______________________________________________________________________ -->
1713 <div class="doc_subsubsection"> <a name="i_ret">'<tt>ret</tt>'
1714 Instruction</a> </div>
1715 <div class="doc_text">
1717 <pre> ret <type> <value> <i>; Return a value from a non-void function</i>
1718 ret void <i>; Return from void function</i>
1721 <p>The '<tt>ret</tt>' instruction is used to return control flow (and a
1722 value) from a function back to the caller.</p>
1723 <p>There are two forms of the '<tt>ret</tt>' instruction: one that
1724 returns a value and then causes control flow, and one that just causes
1725 control flow to occur.</p>
1727 <p>The '<tt>ret</tt>' instruction may return any '<a
1728 href="#t_firstclass">first class</a>' type. Notice that a function is
1729 not <a href="#wellformed">well formed</a> if there exists a '<tt>ret</tt>'
1730 instruction inside of the function that returns a value that does not
1731 match the return type of the function.</p>
1733 <p>When the '<tt>ret</tt>' instruction is executed, control flow
1734 returns back to the calling function's context. If the caller is a "<a
1735 href="#i_call"><tt>call</tt></a>" instruction, execution continues at
1736 the instruction after the call. If the caller was an "<a
1737 href="#i_invoke"><tt>invoke</tt></a>" instruction, execution continues
1738 at the beginning of the "normal" destination block. If the instruction
1739 returns a value, that value shall set the call or invoke instruction's
1742 <pre> ret i32 5 <i>; Return an integer value of 5</i>
1743 ret void <i>; Return from a void function</i>
1746 <!-- _______________________________________________________________________ -->
1747 <div class="doc_subsubsection"> <a name="i_br">'<tt>br</tt>' Instruction</a> </div>
1748 <div class="doc_text">
1750 <pre> br i1 <cond>, label <iftrue>, label <iffalse><br> br label <dest> <i>; Unconditional branch</i>
1753 <p>The '<tt>br</tt>' instruction is used to cause control flow to
1754 transfer to a different basic block in the current function. There are
1755 two forms of this instruction, corresponding to a conditional branch
1756 and an unconditional branch.</p>
1758 <p>The conditional branch form of the '<tt>br</tt>' instruction takes a
1759 single '<tt>i1</tt>' value and two '<tt>label</tt>' values. The
1760 unconditional form of the '<tt>br</tt>' instruction takes a single
1761 '<tt>label</tt>' value as a target.</p>
1763 <p>Upon execution of a conditional '<tt>br</tt>' instruction, the '<tt>i1</tt>'
1764 argument is evaluated. If the value is <tt>true</tt>, control flows
1765 to the '<tt>iftrue</tt>' <tt>label</tt> argument. If "cond" is <tt>false</tt>,
1766 control flows to the '<tt>iffalse</tt>' <tt>label</tt> argument.</p>
1768 <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
1769 href="#i_ret">ret</a> i32 1<br>IfUnequal:<br> <a href="#i_ret">ret</a> i32 0<br></pre>
1771 <!-- _______________________________________________________________________ -->
1772 <div class="doc_subsubsection">
1773 <a name="i_switch">'<tt>switch</tt>' Instruction</a>
1776 <div class="doc_text">
1780 switch <intty> <value>, label <defaultdest> [ <intty> <val>, label <dest> ... ]
1785 <p>The '<tt>switch</tt>' instruction is used to transfer control flow to one of
1786 several different places. It is a generalization of the '<tt>br</tt>'
1787 instruction, allowing a branch to occur to one of many possible
1793 <p>The '<tt>switch</tt>' instruction uses three parameters: an integer
1794 comparison value '<tt>value</tt>', a default '<tt>label</tt>' destination, and
1795 an array of pairs of comparison value constants and '<tt>label</tt>'s. The
1796 table is not allowed to contain duplicate constant entries.</p>
1800 <p>The <tt>switch</tt> instruction specifies a table of values and
1801 destinations. When the '<tt>switch</tt>' instruction is executed, this
1802 table is searched for the given value. If the value is found, control flow is
1803 transfered to the corresponding destination; otherwise, control flow is
1804 transfered to the default destination.</p>
1806 <h5>Implementation:</h5>
1808 <p>Depending on properties of the target machine and the particular
1809 <tt>switch</tt> instruction, this instruction may be code generated in different
1810 ways. For example, it could be generated as a series of chained conditional
1811 branches or with a lookup table.</p>
1816 <i>; Emulate a conditional br instruction</i>
1817 %Val = <a href="#i_zext">zext</a> i1 %value to i32
1818 switch i32 %Val, label %truedest [i32 0, label %falsedest ]
1820 <i>; Emulate an unconditional br instruction</i>
1821 switch i32 0, label %dest [ ]
1823 <i>; Implement a jump table:</i>
1824 switch i32 %val, label %otherwise [ i32 0, label %onzero
1826 i32 2, label %ontwo ]
1830 <!-- _______________________________________________________________________ -->
1831 <div class="doc_subsubsection">
1832 <a name="i_invoke">'<tt>invoke</tt>' Instruction</a>
1835 <div class="doc_text">
1840 <result> = invoke [<a href="#callingconv">cconv</a>] <ptr to function ty> %<function ptr val>(<function args>)
1841 to label <normal label> unwind label <exception label>
1846 <p>The '<tt>invoke</tt>' instruction causes control to transfer to a specified
1847 function, with the possibility of control flow transfer to either the
1848 '<tt>normal</tt>' label or the
1849 '<tt>exception</tt>' label. If the callee function returns with the
1850 "<tt><a href="#i_ret">ret</a></tt>" instruction, control flow will return to the
1851 "normal" label. If the callee (or any indirect callees) returns with the "<a
1852 href="#i_unwind"><tt>unwind</tt></a>" instruction, control is interrupted and
1853 continued at the dynamically nearest "exception" label.</p>
1857 <p>This instruction requires several arguments:</p>
1861 The optional "cconv" marker indicates which <a href="#callingconv">calling
1862 convention</a> the call should use. If none is specified, the call defaults
1863 to using C calling conventions.
1865 <li>'<tt>ptr to function ty</tt>': shall be the signature of the pointer to
1866 function value being invoked. In most cases, this is a direct function
1867 invocation, but indirect <tt>invoke</tt>s are just as possible, branching off
1868 an arbitrary pointer to function value.
1871 <li>'<tt>function ptr val</tt>': An LLVM value containing a pointer to a
1872 function to be invoked. </li>
1874 <li>'<tt>function args</tt>': argument list whose types match the function
1875 signature argument types. If the function signature indicates the function
1876 accepts a variable number of arguments, the extra arguments can be
1879 <li>'<tt>normal label</tt>': the label reached when the called function
1880 executes a '<tt><a href="#i_ret">ret</a></tt>' instruction. </li>
1882 <li>'<tt>exception label</tt>': the label reached when a callee returns with
1883 the <a href="#i_unwind"><tt>unwind</tt></a> instruction. </li>
1889 <p>This instruction is designed to operate as a standard '<tt><a
1890 href="#i_call">call</a></tt>' instruction in most regards. The primary
1891 difference is that it establishes an association with a label, which is used by
1892 the runtime library to unwind the stack.</p>
1894 <p>This instruction is used in languages with destructors to ensure that proper
1895 cleanup is performed in the case of either a <tt>longjmp</tt> or a thrown
1896 exception. Additionally, this is important for implementation of
1897 '<tt>catch</tt>' clauses in high-level languages that support them.</p>
1901 %retval = invoke i32 %Test(i32 15) to label %Continue
1902 unwind label %TestCleanup <i>; {i32}:retval set</i>
1903 %retval = invoke <a href="#callingconv">coldcc</a> i32 %Test(i32 15) to label %Continue
1904 unwind label %TestCleanup <i>; {i32}:retval set</i>
1909 <!-- _______________________________________________________________________ -->
1911 <div class="doc_subsubsection"> <a name="i_unwind">'<tt>unwind</tt>'
1912 Instruction</a> </div>
1914 <div class="doc_text">
1923 <p>The '<tt>unwind</tt>' instruction unwinds the stack, continuing control flow
1924 at the first callee in the dynamic call stack which used an <a
1925 href="#i_invoke"><tt>invoke</tt></a> instruction to perform the call. This is
1926 primarily used to implement exception handling.</p>
1930 <p>The '<tt>unwind</tt>' intrinsic causes execution of the current function to
1931 immediately halt. The dynamic call stack is then searched for the first <a
1932 href="#i_invoke"><tt>invoke</tt></a> instruction on the call stack. Once found,
1933 execution continues at the "exceptional" destination block specified by the
1934 <tt>invoke</tt> instruction. If there is no <tt>invoke</tt> instruction in the
1935 dynamic call chain, undefined behavior results.</p>
1938 <!-- _______________________________________________________________________ -->
1940 <div class="doc_subsubsection"> <a name="i_unreachable">'<tt>unreachable</tt>'
1941 Instruction</a> </div>
1943 <div class="doc_text">
1952 <p>The '<tt>unreachable</tt>' instruction has no defined semantics. This
1953 instruction is used to inform the optimizer that a particular portion of the
1954 code is not reachable. This can be used to indicate that the code after a
1955 no-return function cannot be reached, and other facts.</p>
1959 <p>The '<tt>unreachable</tt>' instruction has no defined semantics.</p>
1964 <!-- ======================================================================= -->
1965 <div class="doc_subsection"> <a name="binaryops">Binary Operations</a> </div>
1966 <div class="doc_text">
1967 <p>Binary operators are used to do most of the computation in a
1968 program. They require two operands, execute an operation on them, and
1969 produce a single value. The operands might represent
1970 multiple data, as is the case with the <a href="#t_vector">vector</a> data type.
1971 The result value of a binary operator is not
1972 necessarily the same type as its operands.</p>
1973 <p>There are several different binary operators:</p>
1975 <!-- _______________________________________________________________________ -->
1976 <div class="doc_subsubsection"> <a name="i_add">'<tt>add</tt>'
1977 Instruction</a> </div>
1978 <div class="doc_text">
1980 <pre> <result> = add <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1983 <p>The '<tt>add</tt>' instruction returns the sum of its two operands.</p>
1985 <p>The two arguments to the '<tt>add</tt>' instruction must be either <a
1986 href="#t_integer">integer</a> or <a href="#t_floating">floating point</a> values.
1987 This instruction can also take <a href="#t_vector">vector</a> versions of the values.
1988 Both arguments must have identical types.</p>
1990 <p>The value produced is the integer or floating point sum of the two
1993 <pre> <result> = add i32 4, %var <i>; yields {i32}:result = 4 + %var</i>
1996 <!-- _______________________________________________________________________ -->
1997 <div class="doc_subsubsection"> <a name="i_sub">'<tt>sub</tt>'
1998 Instruction</a> </div>
1999 <div class="doc_text">
2001 <pre> <result> = sub <ty> <var1>, <var2> <i>; yields {ty}:result</i>
2004 <p>The '<tt>sub</tt>' instruction returns the difference of its two
2006 <p>Note that the '<tt>sub</tt>' instruction is used to represent the '<tt>neg</tt>'
2007 instruction present in most other intermediate representations.</p>
2009 <p>The two arguments to the '<tt>sub</tt>' instruction must be either <a
2010 href="#t_integer">integer</a> or <a href="#t_floating">floating point</a>
2012 This instruction can also take <a href="#t_vector">vector</a> versions of the values.
2013 Both arguments must have identical types.</p>
2015 <p>The value produced is the integer or floating point difference of
2016 the two operands.</p>
2019 <result> = sub i32 4, %var <i>; yields {i32}:result = 4 - %var</i>
2020 <result> = sub i32 0, %val <i>; yields {i32}:result = -%var</i>
2023 <!-- _______________________________________________________________________ -->
2024 <div class="doc_subsubsection"> <a name="i_mul">'<tt>mul</tt>'
2025 Instruction</a> </div>
2026 <div class="doc_text">
2028 <pre> <result> = mul <ty> <var1>, <var2> <i>; yields {ty}:result</i>
2031 <p>The '<tt>mul</tt>' instruction returns the product of its two
2034 <p>The two arguments to the '<tt>mul</tt>' instruction must be either <a
2035 href="#t_integer">integer</a> or <a href="#t_floating">floating point</a>
2037 This instruction can also take <a href="#t_vector">vector</a> versions of the values.
2038 Both arguments must have identical types.</p>
2040 <p>The value produced is the integer or floating point product of the
2042 <p>Because the operands are the same width, the result of an integer
2043 multiplication is the same whether the operands should be deemed unsigned or
2046 <pre> <result> = mul i32 4, %var <i>; yields {i32}:result = 4 * %var</i>
2049 <!-- _______________________________________________________________________ -->
2050 <div class="doc_subsubsection"> <a name="i_udiv">'<tt>udiv</tt>' Instruction
2052 <div class="doc_text">
2054 <pre> <result> = udiv <ty> <var1>, <var2> <i>; yields {ty}:result</i>
2057 <p>The '<tt>udiv</tt>' instruction returns the quotient of its two
2060 <p>The two arguments to the '<tt>udiv</tt>' instruction must be
2061 <a href="#t_integer">integer</a> values. Both arguments must have identical
2062 types. This instruction can also take <a href="#t_vector">vector</a> versions
2063 of the values in which case the elements must be integers.</p>
2065 <p>The value produced is the unsigned integer quotient of the two operands. This
2066 instruction always performs an unsigned division operation, regardless of
2067 whether the arguments are unsigned or not.</p>
2069 <pre> <result> = udiv i32 4, %var <i>; yields {i32}:result = 4 / %var</i>
2072 <!-- _______________________________________________________________________ -->
2073 <div class="doc_subsubsection"> <a name="i_sdiv">'<tt>sdiv</tt>' Instruction
2075 <div class="doc_text">
2077 <pre> <result> = sdiv <ty> <var1>, <var2> <i>; yields {ty}:result</i>
2080 <p>The '<tt>sdiv</tt>' instruction returns the quotient of its two
2083 <p>The two arguments to the '<tt>sdiv</tt>' instruction must be
2084 <a href="#t_integer">integer</a> values. Both arguments must have identical
2085 types. This instruction can also take <a href="#t_vector">vector</a> versions
2086 of the values in which case the elements must be integers.</p>
2088 <p>The value produced is the signed integer quotient of the two operands. This
2089 instruction always performs a signed division operation, regardless of whether
2090 the arguments are signed or not.</p>
2092 <pre> <result> = sdiv i32 4, %var <i>; yields {i32}:result = 4 / %var</i>
2095 <!-- _______________________________________________________________________ -->
2096 <div class="doc_subsubsection"> <a name="i_fdiv">'<tt>fdiv</tt>'
2097 Instruction</a> </div>
2098 <div class="doc_text">
2100 <pre> <result> = fdiv <ty> <var1>, <var2> <i>; yields {ty}:result</i>
2103 <p>The '<tt>fdiv</tt>' instruction returns the quotient of its two
2106 <p>The two arguments to the '<tt>fdiv</tt>' instruction must be
2107 <a href="#t_floating">floating point</a> values. Both arguments must have
2108 identical types. This instruction can also take <a href="#t_vector">vector</a>
2109 versions of floating point values.</p>
2111 <p>The value produced is the floating point quotient of the two operands.</p>
2113 <pre> <result> = fdiv float 4.0, %var <i>; yields {float}:result = 4.0 / %var</i>
2116 <!-- _______________________________________________________________________ -->
2117 <div class="doc_subsubsection"> <a name="i_urem">'<tt>urem</tt>' Instruction</a>
2119 <div class="doc_text">
2121 <pre> <result> = urem <ty> <var1>, <var2> <i>; yields {ty}:result</i>
2124 <p>The '<tt>urem</tt>' instruction returns the remainder from the
2125 unsigned division of its two arguments.</p>
2127 <p>The two arguments to the '<tt>urem</tt>' instruction must be
2128 <a href="#t_integer">integer</a> values. Both arguments must have identical
2129 types. This instruction can also take <a href="#t_vector">vector</a> versions
2130 of the values in which case the elements must be integers.</p>
2132 <p>This instruction returns the unsigned integer <i>remainder</i> of a division.
2133 This instruction always performs an unsigned division to get the remainder,
2134 regardless of whether the arguments are unsigned or not.</p>
2136 <pre> <result> = urem i32 4, %var <i>; yields {i32}:result = 4 % %var</i>
2140 <!-- _______________________________________________________________________ -->
2141 <div class="doc_subsubsection"> <a name="i_srem">'<tt>srem</tt>'
2142 Instruction</a> </div>
2143 <div class="doc_text">
2145 <pre> <result> = srem <ty> <var1>, <var2> <i>; yields {ty}:result</i>
2148 <p>The '<tt>srem</tt>' instruction returns the remainder from the
2149 signed division of its two operands. This instruction can also take
2150 <a href="#t_vector">vector</a> versions of the values in which case
2151 the elements must be integers.</p>
2154 <p>The two arguments to the '<tt>srem</tt>' instruction must be
2155 <a href="#t_integer">integer</a> values. Both arguments must have identical
2158 <p>This instruction returns the <i>remainder</i> of a division (where the result
2159 has the same sign as the dividend, <tt>var1</tt>), not the <i>modulo</i>
2160 operator (where the result has the same sign as the divisor, <tt>var2</tt>) of
2161 a value. For more information about the difference, see <a
2162 href="http://mathforum.org/dr.math/problems/anne.4.28.99.html">The
2163 Math Forum</a>. For a table of how this is implemented in various languages,
2164 please see <a href="http://en.wikipedia.org/wiki/Modulo_operation">
2165 Wikipedia: modulo operation</a>.</p>
2167 <pre> <result> = srem i32 4, %var <i>; yields {i32}:result = 4 % %var</i>
2171 <!-- _______________________________________________________________________ -->
2172 <div class="doc_subsubsection"> <a name="i_frem">'<tt>frem</tt>'
2173 Instruction</a> </div>
2174 <div class="doc_text">
2176 <pre> <result> = frem <ty> <var1>, <var2> <i>; yields {ty}:result</i>
2179 <p>The '<tt>frem</tt>' instruction returns the remainder from the
2180 division of its two operands.</p>
2182 <p>The two arguments to the '<tt>frem</tt>' instruction must be
2183 <a href="#t_floating">floating point</a> values. Both arguments must have
2184 identical types. This instruction can also take <a href="#t_vector">vector</a>
2185 versions of floating point values.</p>
2187 <p>This instruction returns the <i>remainder</i> of a division.</p>
2189 <pre> <result> = frem float 4.0, %var <i>; yields {float}:result = 4.0 % %var</i>
2193 <!-- ======================================================================= -->
2194 <div class="doc_subsection"> <a name="bitwiseops">Bitwise Binary
2195 Operations</a> </div>
2196 <div class="doc_text">
2197 <p>Bitwise binary operators are used to do various forms of
2198 bit-twiddling in a program. They are generally very efficient
2199 instructions and can commonly be strength reduced from other
2200 instructions. They require two operands, execute an operation on them,
2201 and produce a single value. The resulting value of the bitwise binary
2202 operators is always the same type as its first operand.</p>
2205 <!-- _______________________________________________________________________ -->
2206 <div class="doc_subsubsection"> <a name="i_shl">'<tt>shl</tt>'
2207 Instruction</a> </div>
2208 <div class="doc_text">
2210 <pre> <result> = shl <ty> <var1>, <var2> <i>; yields {ty}:result</i>
2215 <p>The '<tt>shl</tt>' instruction returns the first operand shifted to
2216 the left a specified number of bits.</p>
2220 <p>Both arguments to the '<tt>shl</tt>' instruction must be the same <a
2221 href="#t_integer">integer</a> type.</p>
2225 <p>The value produced is <tt>var1</tt> * 2<sup><tt>var2</tt></sup>. If
2226 <tt>var2</tt> is (statically or dynamically) equal to or larger than the number
2227 of bits in <tt>var1</tt>, the result is undefined.</p>
2229 <h5>Example:</h5><pre>
2230 <result> = shl i32 4, %var <i>; yields {i32}: 4 << %var</i>
2231 <result> = shl i32 4, 2 <i>; yields {i32}: 16</i>
2232 <result> = shl i32 1, 10 <i>; yields {i32}: 1024</i>
2233 <result> = shl i32 1, 32 <i>; undefined</i>
2236 <!-- _______________________________________________________________________ -->
2237 <div class="doc_subsubsection"> <a name="i_lshr">'<tt>lshr</tt>'
2238 Instruction</a> </div>
2239 <div class="doc_text">
2241 <pre> <result> = lshr <ty> <var1>, <var2> <i>; yields {ty}:result</i>
2245 <p>The '<tt>lshr</tt>' instruction (logical shift right) returns the first
2246 operand shifted to the right a specified number of bits with zero fill.</p>
2249 <p>Both arguments to the '<tt>lshr</tt>' instruction must be the same
2250 <a href="#t_integer">integer</a> type.</p>
2254 <p>This instruction always performs a logical shift right operation. The most
2255 significant bits of the result will be filled with zero bits after the
2256 shift. If <tt>var2</tt> is (statically or dynamically) equal to or larger than
2257 the number of bits in <tt>var1</tt>, the result is undefined.</p>
2261 <result> = lshr i32 4, 1 <i>; yields {i32}:result = 2</i>
2262 <result> = lshr i32 4, 2 <i>; yields {i32}:result = 1</i>
2263 <result> = lshr i8 4, 3 <i>; yields {i8}:result = 0</i>
2264 <result> = lshr i8 -2, 1 <i>; yields {i8}:result = 0x7FFFFFFF </i>
2265 <result> = lshr i32 1, 32 <i>; undefined</i>
2269 <!-- _______________________________________________________________________ -->
2270 <div class="doc_subsubsection"> <a name="i_ashr">'<tt>ashr</tt>'
2271 Instruction</a> </div>
2272 <div class="doc_text">
2275 <pre> <result> = ashr <ty> <var1>, <var2> <i>; yields {ty}:result</i>
2279 <p>The '<tt>ashr</tt>' instruction (arithmetic shift right) returns the first
2280 operand shifted to the right a specified number of bits with sign extension.</p>
2283 <p>Both arguments to the '<tt>ashr</tt>' instruction must be the same
2284 <a href="#t_integer">integer</a> type.</p>
2287 <p>This instruction always performs an arithmetic shift right operation,
2288 The most significant bits of the result will be filled with the sign bit
2289 of <tt>var1</tt>. If <tt>var2</tt> is (statically or dynamically) equal to or
2290 larger than the number of bits in <tt>var1</tt>, the result is undefined.
2295 <result> = ashr i32 4, 1 <i>; yields {i32}:result = 2</i>
2296 <result> = ashr i32 4, 2 <i>; yields {i32}:result = 1</i>
2297 <result> = ashr i8 4, 3 <i>; yields {i8}:result = 0</i>
2298 <result> = ashr i8 -2, 1 <i>; yields {i8}:result = -1</i>
2299 <result> = ashr i32 1, 32 <i>; undefined</i>
2303 <!-- _______________________________________________________________________ -->
2304 <div class="doc_subsubsection"> <a name="i_and">'<tt>and</tt>'
2305 Instruction</a> </div>
2306 <div class="doc_text">
2308 <pre> <result> = and <ty> <var1>, <var2> <i>; yields {ty}:result</i>
2311 <p>The '<tt>and</tt>' instruction returns the bitwise logical and of
2312 its two operands.</p>
2314 <p>The two arguments to the '<tt>and</tt>' instruction must be <a
2315 href="#t_integer">integer</a> values. Both arguments must have
2316 identical types.</p>
2318 <p>The truth table used for the '<tt>and</tt>' instruction is:</p>
2320 <div style="align: center">
2321 <table border="1" cellspacing="0" cellpadding="4">
2352 <pre> <result> = and i32 4, %var <i>; yields {i32}:result = 4 & %var</i>
2353 <result> = and i32 15, 40 <i>; yields {i32}:result = 8</i>
2354 <result> = and i32 4, 8 <i>; yields {i32}:result = 0</i>
2357 <!-- _______________________________________________________________________ -->
2358 <div class="doc_subsubsection"> <a name="i_or">'<tt>or</tt>' Instruction</a> </div>
2359 <div class="doc_text">
2361 <pre> <result> = or <ty> <var1>, <var2> <i>; yields {ty}:result</i>
2364 <p>The '<tt>or</tt>' instruction returns the bitwise logical inclusive
2365 or of its two operands.</p>
2367 <p>The two arguments to the '<tt>or</tt>' instruction must be <a
2368 href="#t_integer">integer</a> values. Both arguments must have
2369 identical types.</p>
2371 <p>The truth table used for the '<tt>or</tt>' instruction is:</p>
2373 <div style="align: center">
2374 <table border="1" cellspacing="0" cellpadding="4">
2405 <pre> <result> = or i32 4, %var <i>; yields {i32}:result = 4 | %var</i>
2406 <result> = or i32 15, 40 <i>; yields {i32}:result = 47</i>
2407 <result> = or i32 4, 8 <i>; yields {i32}:result = 12</i>
2410 <!-- _______________________________________________________________________ -->
2411 <div class="doc_subsubsection"> <a name="i_xor">'<tt>xor</tt>'
2412 Instruction</a> </div>
2413 <div class="doc_text">
2415 <pre> <result> = xor <ty> <var1>, <var2> <i>; yields {ty}:result</i>
2418 <p>The '<tt>xor</tt>' instruction returns the bitwise logical exclusive
2419 or of its two operands. The <tt>xor</tt> is used to implement the
2420 "one's complement" operation, which is the "~" operator in C.</p>
2422 <p>The two arguments to the '<tt>xor</tt>' instruction must be <a
2423 href="#t_integer">integer</a> values. Both arguments must have
2424 identical types.</p>
2426 <p>The truth table used for the '<tt>xor</tt>' instruction is:</p>
2428 <div style="align: center">
2429 <table border="1" cellspacing="0" cellpadding="4">
2461 <pre> <result> = xor i32 4, %var <i>; yields {i32}:result = 4 ^ %var</i>
2462 <result> = xor i32 15, 40 <i>; yields {i32}:result = 39</i>
2463 <result> = xor i32 4, 8 <i>; yields {i32}:result = 12</i>
2464 <result> = xor i32 %V, -1 <i>; yields {i32}:result = ~%V</i>
2468 <!-- ======================================================================= -->
2469 <div class="doc_subsection">
2470 <a name="vectorops">Vector Operations</a>
2473 <div class="doc_text">
2475 <p>LLVM supports several instructions to represent vector operations in a
2476 target-independent manner. These instructions cover the element-access and
2477 vector-specific operations needed to process vectors effectively. While LLVM
2478 does directly support these vector operations, many sophisticated algorithms
2479 will want to use target-specific intrinsics to take full advantage of a specific
2484 <!-- _______________________________________________________________________ -->
2485 <div class="doc_subsubsection">
2486 <a name="i_extractelement">'<tt>extractelement</tt>' Instruction</a>
2489 <div class="doc_text">
2494 <result> = extractelement <n x <ty>> <val>, i32 <idx> <i>; yields <ty></i>
2500 The '<tt>extractelement</tt>' instruction extracts a single scalar
2501 element from a vector at a specified index.
2508 The first operand of an '<tt>extractelement</tt>' instruction is a
2509 value of <a href="#t_vector">vector</a> type. The second operand is
2510 an index indicating the position from which to extract the element.
2511 The index may be a variable.</p>
2516 The result is a scalar of the same type as the element type of
2517 <tt>val</tt>. Its value is the value at position <tt>idx</tt> of
2518 <tt>val</tt>. If <tt>idx</tt> exceeds the length of <tt>val</tt>, the
2519 results are undefined.
2525 %result = extractelement <4 x i32> %vec, i32 0 <i>; yields i32</i>
2530 <!-- _______________________________________________________________________ -->
2531 <div class="doc_subsubsection">
2532 <a name="i_insertelement">'<tt>insertelement</tt>' Instruction</a>
2535 <div class="doc_text">
2540 <result> = insertelement <n x <ty>> <val>, <ty> <elt>, i32 <idx> <i>; yields <n x <ty>></i>
2546 The '<tt>insertelement</tt>' instruction inserts a scalar
2547 element into a vector at a specified index.
2554 The first operand of an '<tt>insertelement</tt>' instruction is a
2555 value of <a href="#t_vector">vector</a> type. The second operand is a
2556 scalar value whose type must equal the element type of the first
2557 operand. The third operand is an index indicating the position at
2558 which to insert the value. The index may be a variable.</p>
2563 The result is a vector of the same type as <tt>val</tt>. Its
2564 element values are those of <tt>val</tt> except at position
2565 <tt>idx</tt>, where it gets the value <tt>elt</tt>. If <tt>idx</tt>
2566 exceeds the length of <tt>val</tt>, the results are undefined.
2572 %result = insertelement <4 x i32> %vec, i32 1, i32 0 <i>; yields <4 x i32></i>
2576 <!-- _______________________________________________________________________ -->
2577 <div class="doc_subsubsection">
2578 <a name="i_shufflevector">'<tt>shufflevector</tt>' Instruction</a>
2581 <div class="doc_text">
2586 <result> = shufflevector <n x <ty>> <v1>, <n x <ty>> <v2>, <n x i32> <mask> <i>; yields <n x <ty>></i>
2592 The '<tt>shufflevector</tt>' instruction constructs a permutation of elements
2593 from two input vectors, returning a vector of the same type.
2599 The first two operands of a '<tt>shufflevector</tt>' instruction are vectors
2600 with types that match each other and types that match the result of the
2601 instruction. The third argument is a shuffle mask, which has the same number
2602 of elements as the other vector type, but whose element type is always 'i32'.
2606 The shuffle mask operand is required to be a constant vector with either
2607 constant integer or undef values.
2613 The elements of the two input vectors are numbered from left to right across
2614 both of the vectors. The shuffle mask operand specifies, for each element of
2615 the result vector, which element of the two input registers the result element
2616 gets. The element selector may be undef (meaning "don't care") and the second
2617 operand may be undef if performing a shuffle from only one vector.
2623 %result = shufflevector <4 x i32> %v1, <4 x i32> %v2,
2624 <4 x i32> <i32 0, i32 4, i32 1, i32 5> <i>; yields <4 x i32></i>
2625 %result = shufflevector <4 x i32> %v1, <4 x i32> undef,
2626 <4 x i32> <i32 0, i32 1, i32 2, i32 3> <i>; yields <4 x i32></i> - Identity shuffle.
2631 <!-- ======================================================================= -->
2632 <div class="doc_subsection">
2633 <a name="memoryops">Memory Access and Addressing Operations</a>
2636 <div class="doc_text">
2638 <p>A key design point of an SSA-based representation is how it
2639 represents memory. In LLVM, no memory locations are in SSA form, which
2640 makes things very simple. This section describes how to read, write,
2641 allocate, and free memory in LLVM.</p>
2645 <!-- _______________________________________________________________________ -->
2646 <div class="doc_subsubsection">
2647 <a name="i_malloc">'<tt>malloc</tt>' Instruction</a>
2650 <div class="doc_text">
2655 <result> = malloc <type>[, i32 <NumElements>][, align <alignment>] <i>; yields {type*}:result</i>
2660 <p>The '<tt>malloc</tt>' instruction allocates memory from the system
2661 heap and returns a pointer to it. The object is always allocated in the generic
2662 address space (address space zero).</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. The object is always allocated in the generic address
2750 space (address space zero).</p>
2754 <p>The '<tt>alloca</tt>' instruction allocates <tt>sizeof(<type>)*NumElements</tt>
2755 bytes of memory on the runtime stack, returning a pointer of the
2756 appropriate type to the program. If "NumElements" is specified, it is the
2757 number of elements allocated. If an alignment is specified, the value result
2758 of the allocation is guaranteed to be aligned to at least that boundary. If
2759 not specified, or if zero, the target can choose to align the allocation on any
2760 convenient boundary.</p>
2762 <p>'<tt>type</tt>' may be any sized type.</p>
2766 <p>Memory is allocated; a pointer is returned. '<tt>alloca</tt>'d
2767 memory is automatically released when the function returns. The '<tt>alloca</tt>'
2768 instruction is commonly used to represent automatic variables that must
2769 have an address available. When the function returns (either with the <tt><a
2770 href="#i_ret">ret</a></tt> or <tt><a href="#i_unwind">unwind</a></tt>
2771 instructions), the memory is reclaimed.</p>
2776 %ptr = alloca i32 <i>; yields {i32*}:ptr</i>
2777 %ptr = alloca i32, i32 4 <i>; yields {i32*}:ptr</i>
2778 %ptr = alloca i32, i32 4, align 1024 <i>; yields {i32*}:ptr</i>
2779 %ptr = alloca i32, align 1024 <i>; yields {i32*}:ptr</i>
2783 <!-- _______________________________________________________________________ -->
2784 <div class="doc_subsubsection"> <a name="i_load">'<tt>load</tt>'
2785 Instruction</a> </div>
2786 <div class="doc_text">
2788 <pre> <result> = load <ty>* <pointer>[, align <alignment>]<br> <result> = volatile load <ty>* <pointer>[, align <alignment>]<br></pre>
2790 <p>The '<tt>load</tt>' instruction is used to read from memory.</p>
2792 <p>The argument to the '<tt>load</tt>' instruction specifies the memory
2793 address from which to load. The pointer must point to a <a
2794 href="#t_firstclass">first class</a> type. If the <tt>load</tt> is
2795 marked as <tt>volatile</tt>, then the optimizer is not allowed to modify
2796 the number or order of execution of this <tt>load</tt> with other
2797 volatile <tt>load</tt> and <tt><a href="#i_store">store</a></tt>
2800 <p>The location of memory pointed to is loaded.</p>
2802 <pre> %ptr = <a href="#i_alloca">alloca</a> i32 <i>; yields {i32*}:ptr</i>
2804 href="#i_store">store</a> i32 3, i32* %ptr <i>; yields {void}</i>
2805 %val = load i32* %ptr <i>; yields {i32}:val = i32 3</i>
2808 <!-- _______________________________________________________________________ -->
2809 <div class="doc_subsubsection"> <a name="i_store">'<tt>store</tt>'
2810 Instruction</a> </div>
2811 <div class="doc_text">
2813 <pre> store <ty> <value>, <ty>* <pointer>[, align <alignment>] <i>; yields {void}</i>
2814 volatile store <ty> <value>, <ty>* <pointer>[, align <alignment>] <i>; yields {void}</i>
2817 <p>The '<tt>store</tt>' instruction is used to write to memory.</p>
2819 <p>There are two arguments to the '<tt>store</tt>' instruction: a value
2820 to store and an address at which to store it. The type of the '<tt><pointer></tt>'
2821 operand must be a pointer to the type of the '<tt><value></tt>'
2822 operand. If the <tt>store</tt> is marked as <tt>volatile</tt>, then the
2823 optimizer is not allowed to modify the number or order of execution of
2824 this <tt>store</tt> with other volatile <tt>load</tt> and <tt><a
2825 href="#i_store">store</a></tt> instructions.</p>
2827 <p>The contents of memory are updated to contain '<tt><value></tt>'
2828 at the location specified by the '<tt><pointer></tt>' operand.</p>
2830 <pre> %ptr = <a href="#i_alloca">alloca</a> i32 <i>; yields {i32*}:ptr</i>
2831 store i32 3, i32* %ptr <i>; yields {void}</i>
2832 %val = <a href="#i_load">load</a> i32* %ptr <i>; yields {i32}:val = i32 3</i>
2836 <!-- _______________________________________________________________________ -->
2837 <div class="doc_subsubsection">
2838 <a name="i_getelementptr">'<tt>getelementptr</tt>' Instruction</a>
2841 <div class="doc_text">
2844 <result> = getelementptr <ty>* <ptrval>{, <ty> <idx>}*
2850 The '<tt>getelementptr</tt>' instruction is used to get the address of a
2851 subelement of an aggregate data structure.</p>
2855 <p>This instruction takes a list of integer operands that indicate what
2856 elements of the aggregate object to index to. The actual types of the arguments
2857 provided depend on the type of the first pointer argument. The
2858 '<tt>getelementptr</tt>' instruction is used to index down through the type
2859 levels of a structure or to a specific index in an array. When indexing into a
2860 structure, only <tt>i32</tt> integer constants are allowed. When indexing
2861 into an array or pointer, only integers of 32 or 64 bits are allowed, and will
2862 be sign extended to 64-bit values.</p>
2864 <p>For example, let's consider a C code fragment and how it gets
2865 compiled to LLVM:</p>
2867 <div class="doc_code">
2880 int *foo(struct ST *s) {
2881 return &s[1].Z.B[5][13];
2886 <p>The LLVM code generated by the GCC frontend is:</p>
2888 <div class="doc_code">
2890 %RT = type { i8 , [10 x [20 x i32]], i8 }
2891 %ST = type { i32, double, %RT }
2893 define i32* %foo(%ST* %s) {
2895 %reg = getelementptr %ST* %s, i32 1, i32 2, i32 1, i32 5, i32 13
2903 <p>The index types specified for the '<tt>getelementptr</tt>' instruction depend
2904 on the pointer type that is being indexed into. <a href="#t_pointer">Pointer</a>
2905 and <a href="#t_array">array</a> types can use a 32-bit or 64-bit
2906 <a href="#t_integer">integer</a> type but the value will always be sign extended
2907 to 64-bits. <a href="#t_struct">Structure</a> types require <tt>i32</tt>
2908 <b>constants</b>.</p>
2910 <p>In the example above, the first index is indexing into the '<tt>%ST*</tt>'
2911 type, which is a pointer, yielding a '<tt>%ST</tt>' = '<tt>{ i32, double, %RT
2912 }</tt>' type, a structure. The second index indexes into the third element of
2913 the structure, yielding a '<tt>%RT</tt>' = '<tt>{ i8 , [10 x [20 x i32]],
2914 i8 }</tt>' type, another structure. The third index indexes into the second
2915 element of the structure, yielding a '<tt>[10 x [20 x i32]]</tt>' type, an
2916 array. The two dimensions of the array are subscripted into, yielding an
2917 '<tt>i32</tt>' type. The '<tt>getelementptr</tt>' instruction returns a pointer
2918 to this element, thus computing a value of '<tt>i32*</tt>' type.</p>
2920 <p>Note that it is perfectly legal to index partially through a
2921 structure, returning a pointer to an inner element. Because of this,
2922 the LLVM code for the given testcase is equivalent to:</p>
2925 define i32* %foo(%ST* %s) {
2926 %t1 = getelementptr %ST* %s, i32 1 <i>; yields %ST*:%t1</i>
2927 %t2 = getelementptr %ST* %t1, i32 0, i32 2 <i>; yields %RT*:%t2</i>
2928 %t3 = getelementptr %RT* %t2, i32 0, i32 1 <i>; yields [10 x [20 x i32]]*:%t3</i>
2929 %t4 = getelementptr [10 x [20 x i32]]* %t3, i32 0, i32 5 <i>; yields [20 x i32]*:%t4</i>
2930 %t5 = getelementptr [20 x i32]* %t4, i32 0, i32 13 <i>; yields i32*:%t5</i>
2935 <p>Note that it is undefined to access an array out of bounds: array and
2936 pointer indexes must always be within the defined bounds of the array type.
2937 The one exception for this rules is zero length arrays. These arrays are
2938 defined to be accessible as variable length arrays, which requires access
2939 beyond the zero'th element.</p>
2941 <p>The getelementptr instruction is often confusing. For some more insight
2942 into how it works, see <a href="GetElementPtr.html">the getelementptr
2948 <i>; yields [12 x i8]*:aptr</i>
2949 %aptr = getelementptr {i32, [12 x i8]}* %sptr, i64 0, i32 1
2953 <!-- ======================================================================= -->
2954 <div class="doc_subsection"> <a name="convertops">Conversion Operations</a>
2956 <div class="doc_text">
2957 <p>The instructions in this category are the conversion instructions (casting)
2958 which all take a single operand and a type. They perform various bit conversions
2962 <!-- _______________________________________________________________________ -->
2963 <div class="doc_subsubsection">
2964 <a name="i_trunc">'<tt>trunc .. to</tt>' Instruction</a>
2966 <div class="doc_text">
2970 <result> = trunc <ty> <value> to <ty2> <i>; yields ty2</i>
2975 The '<tt>trunc</tt>' instruction truncates its operand to the type <tt>ty2</tt>.
2980 The '<tt>trunc</tt>' instruction takes a <tt>value</tt> to trunc, which must
2981 be an <a href="#t_integer">integer</a> type, and a type that specifies the size
2982 and type of the result, which must be an <a href="#t_integer">integer</a>
2983 type. The bit size of <tt>value</tt> must be larger than the bit size of
2984 <tt>ty2</tt>. Equal sized types are not allowed.</p>
2988 The '<tt>trunc</tt>' instruction truncates the high order bits in <tt>value</tt>
2989 and converts the remaining bits to <tt>ty2</tt>. Since the source size must be
2990 larger than the destination size, <tt>trunc</tt> cannot be a <i>no-op cast</i>.
2991 It will always truncate bits.</p>
2995 %X = trunc i32 257 to i8 <i>; yields i8:1</i>
2996 %Y = trunc i32 123 to i1 <i>; yields i1:true</i>
2997 %Y = trunc i32 122 to i1 <i>; yields i1:false</i>
3001 <!-- _______________________________________________________________________ -->
3002 <div class="doc_subsubsection">
3003 <a name="i_zext">'<tt>zext .. to</tt>' Instruction</a>
3005 <div class="doc_text">
3009 <result> = zext <ty> <value> to <ty2> <i>; yields ty2</i>
3013 <p>The '<tt>zext</tt>' instruction zero extends its operand to type
3018 <p>The '<tt>zext</tt>' instruction takes a value to cast, which must be of
3019 <a href="#t_integer">integer</a> type, and a type to cast it to, which must
3020 also be of <a href="#t_integer">integer</a> type. The bit size of the
3021 <tt>value</tt> must be smaller than the bit size of the destination type,
3025 <p>The <tt>zext</tt> fills the high order bits of the <tt>value</tt> with zero
3026 bits until it reaches the size of the destination type, <tt>ty2</tt>.</p>
3028 <p>When zero extending from i1, the result will always be either 0 or 1.</p>
3032 %X = zext i32 257 to i64 <i>; yields i64:257</i>
3033 %Y = zext i1 true to i32 <i>; yields i32:1</i>
3037 <!-- _______________________________________________________________________ -->
3038 <div class="doc_subsubsection">
3039 <a name="i_sext">'<tt>sext .. to</tt>' Instruction</a>
3041 <div class="doc_text">
3045 <result> = sext <ty> <value> to <ty2> <i>; yields ty2</i>
3049 <p>The '<tt>sext</tt>' sign extends <tt>value</tt> to the type <tt>ty2</tt>.</p>
3053 The '<tt>sext</tt>' instruction takes a value to cast, which must be of
3054 <a href="#t_integer">integer</a> type, and a type to cast it to, which must
3055 also be of <a href="#t_integer">integer</a> type. The bit size of the
3056 <tt>value</tt> must be smaller than the bit size of the destination type,
3061 The '<tt>sext</tt>' instruction performs a sign extension by copying the sign
3062 bit (highest order bit) of the <tt>value</tt> until it reaches the bit size of
3063 the type <tt>ty2</tt>.</p>
3065 <p>When sign extending from i1, the extension always results in -1 or 0.</p>
3069 %X = sext i8 -1 to i16 <i>; yields i16 :65535</i>
3070 %Y = sext i1 true to i32 <i>; yields i32:-1</i>
3074 <!-- _______________________________________________________________________ -->
3075 <div class="doc_subsubsection">
3076 <a name="i_fptrunc">'<tt>fptrunc .. to</tt>' Instruction</a>
3079 <div class="doc_text">
3084 <result> = fptrunc <ty> <value> to <ty2> <i>; yields ty2</i>
3088 <p>The '<tt>fptrunc</tt>' instruction truncates <tt>value</tt> to type
3093 <p>The '<tt>fptrunc</tt>' instruction takes a <a href="#t_floating">floating
3094 point</a> value to cast and a <a href="#t_floating">floating point</a> type to
3095 cast it to. The size of <tt>value</tt> must be larger than the size of
3096 <tt>ty2</tt>. This implies that <tt>fptrunc</tt> cannot be used to make a
3097 <i>no-op cast</i>.</p>
3100 <p> The '<tt>fptrunc</tt>' instruction truncates a <tt>value</tt> from a larger
3101 <a href="#t_floating">floating point</a> type to a smaller
3102 <a href="#t_floating">floating point</a> type. If the value cannot fit within
3103 the destination type, <tt>ty2</tt>, then the results are undefined.</p>
3107 %X = fptrunc double 123.0 to float <i>; yields float:123.0</i>
3108 %Y = fptrunc double 1.0E+300 to float <i>; yields undefined</i>
3112 <!-- _______________________________________________________________________ -->
3113 <div class="doc_subsubsection">
3114 <a name="i_fpext">'<tt>fpext .. to</tt>' Instruction</a>
3116 <div class="doc_text">
3120 <result> = fpext <ty> <value> to <ty2> <i>; yields ty2</i>
3124 <p>The '<tt>fpext</tt>' extends a floating point <tt>value</tt> to a larger
3125 floating point value.</p>
3128 <p>The '<tt>fpext</tt>' instruction takes a
3129 <a href="#t_floating">floating point</a> <tt>value</tt> to cast,
3130 and a <a href="#t_floating">floating point</a> type to cast it to. The source
3131 type must be smaller than the destination type.</p>
3134 <p>The '<tt>fpext</tt>' instruction extends the <tt>value</tt> from a smaller
3135 <a href="#t_floating">floating point</a> type to a larger
3136 <a href="#t_floating">floating point</a> type. The <tt>fpext</tt> cannot be
3137 used to make a <i>no-op cast</i> because it always changes bits. Use
3138 <tt>bitcast</tt> to make a <i>no-op cast</i> for a floating point cast.</p>
3142 %X = fpext float 3.1415 to double <i>; yields double:3.1415</i>
3143 %Y = fpext float 1.0 to float <i>; yields float:1.0 (no-op)</i>
3147 <!-- _______________________________________________________________________ -->
3148 <div class="doc_subsubsection">
3149 <a name="i_fptoui">'<tt>fptoui .. to</tt>' Instruction</a>
3151 <div class="doc_text">
3155 <result> = fptoui <ty> <value> to <ty2> <i>; yields ty2</i>
3159 <p>The '<tt>fptoui</tt>' converts a floating point <tt>value</tt> to its
3160 unsigned integer equivalent of type <tt>ty2</tt>.
3164 <p>The '<tt>fptoui</tt>' instruction takes a value to cast, which must be a
3165 scalar or vector <a href="#t_floating">floating point</a> value, and a type
3166 to cast it to <tt>ty2</tt>, which must be an <a href="#t_integer">integer</a>
3167 type. If <tt>ty</tt> is a vector floating point type, <tt>ty2</tt> must be a
3168 vector integer type with the same number of elements as <tt>ty</tt></p>
3171 <p> The '<tt>fptoui</tt>' instruction converts its
3172 <a href="#t_floating">floating point</a> operand into the nearest (rounding
3173 towards zero) unsigned integer value. If the value cannot fit in <tt>ty2</tt>,
3174 the results are undefined.</p>
3178 %X = fptoui double 123.0 to i32 <i>; yields i32:123</i>
3179 %Y = fptoui float 1.0E+300 to i1 <i>; yields undefined:1</i>
3180 %X = fptoui float 1.04E+17 to i8 <i>; yields undefined:1</i>
3184 <!-- _______________________________________________________________________ -->
3185 <div class="doc_subsubsection">
3186 <a name="i_fptosi">'<tt>fptosi .. to</tt>' Instruction</a>
3188 <div class="doc_text">
3192 <result> = fptosi <ty> <value> to <ty2> <i>; yields ty2</i>
3196 <p>The '<tt>fptosi</tt>' instruction converts
3197 <a href="#t_floating">floating point</a> <tt>value</tt> to type <tt>ty2</tt>.
3201 <p> The '<tt>fptosi</tt>' instruction takes a value to cast, which must be a
3202 scalar or vector <a href="#t_floating">floating point</a> value, and a type
3203 to cast it to <tt>ty2</tt>, which must be an <a href="#t_integer">integer</a>
3204 type. If <tt>ty</tt> is a vector floating point type, <tt>ty2</tt> must be a
3205 vector integer type with the same number of elements as <tt>ty</tt></p>
3208 <p>The '<tt>fptosi</tt>' instruction converts its
3209 <a href="#t_floating">floating point</a> operand into the nearest (rounding
3210 towards zero) signed integer value. If the value cannot fit in <tt>ty2</tt>,
3211 the results are undefined.</p>
3215 %X = fptosi double -123.0 to i32 <i>; yields i32:-123</i>
3216 %Y = fptosi float 1.0E-247 to i1 <i>; yields undefined:1</i>
3217 %X = fptosi float 1.04E+17 to i8 <i>; yields undefined:1</i>
3221 <!-- _______________________________________________________________________ -->
3222 <div class="doc_subsubsection">
3223 <a name="i_uitofp">'<tt>uitofp .. to</tt>' Instruction</a>
3225 <div class="doc_text">
3229 <result> = uitofp <ty> <value> to <ty2> <i>; yields ty2</i>
3233 <p>The '<tt>uitofp</tt>' instruction regards <tt>value</tt> as an unsigned
3234 integer and converts that value to the <tt>ty2</tt> type.</p>
3237 <p>The '<tt>uitofp</tt>' instruction takes a value to cast, which must be a
3238 scalar or vector <a href="#t_integer">integer</a> value, and a type to cast it
3239 to <tt>ty2</tt>, which must be an <a href="#t_floating">floating point</a>
3240 type. If <tt>ty</tt> is a vector integer type, <tt>ty2</tt> must be a vector
3241 floating point type with the same number of elements as <tt>ty</tt></p>
3244 <p>The '<tt>uitofp</tt>' instruction interprets its operand as an unsigned
3245 integer quantity and converts it to the corresponding floating point value. If
3246 the value cannot fit in the floating point value, the results are undefined.</p>
3250 %X = uitofp i32 257 to float <i>; yields float:257.0</i>
3251 %Y = uitofp i8 -1 to double <i>; yields double:255.0</i>
3255 <!-- _______________________________________________________________________ -->
3256 <div class="doc_subsubsection">
3257 <a name="i_sitofp">'<tt>sitofp .. to</tt>' Instruction</a>
3259 <div class="doc_text">
3263 <result> = sitofp <ty> <value> to <ty2> <i>; yields ty2</i>
3267 <p>The '<tt>sitofp</tt>' instruction regards <tt>value</tt> as a signed
3268 integer and converts that value to the <tt>ty2</tt> type.</p>
3271 <p>The '<tt>sitofp</tt>' instruction takes a value to cast, which must be a
3272 scalar or vector <a href="#t_integer">integer</a> value, and a type to cast it
3273 to <tt>ty2</tt>, which must be an <a href="#t_floating">floating point</a>
3274 type. If <tt>ty</tt> is a vector integer type, <tt>ty2</tt> must be a vector
3275 floating point type with the same number of elements as <tt>ty</tt></p>
3278 <p>The '<tt>sitofp</tt>' instruction interprets its operand as a signed
3279 integer quantity and converts it to the corresponding floating point value. If
3280 the value cannot fit in the floating point value, the results are undefined.</p>
3284 %X = sitofp i32 257 to float <i>; yields float:257.0</i>
3285 %Y = sitofp i8 -1 to double <i>; yields double:-1.0</i>
3289 <!-- _______________________________________________________________________ -->
3290 <div class="doc_subsubsection">
3291 <a name="i_ptrtoint">'<tt>ptrtoint .. to</tt>' Instruction</a>
3293 <div class="doc_text">
3297 <result> = ptrtoint <ty> <value> to <ty2> <i>; yields ty2</i>
3301 <p>The '<tt>ptrtoint</tt>' instruction converts the pointer <tt>value</tt> to
3302 the integer type <tt>ty2</tt>.</p>
3305 <p>The '<tt>ptrtoint</tt>' instruction takes a <tt>value</tt> to cast, which
3306 must be a <a href="#t_pointer">pointer</a> value, and a type to cast it to
3307 <tt>ty2</tt>, which must be an <a href="#t_integer">integer</a> type.
3310 <p>The '<tt>ptrtoint</tt>' instruction converts <tt>value</tt> to integer type
3311 <tt>ty2</tt> by interpreting the pointer value as an integer and either
3312 truncating or zero extending that value to the size of the integer type. If
3313 <tt>value</tt> is smaller than <tt>ty2</tt> then a zero extension is done. If
3314 <tt>value</tt> is larger than <tt>ty2</tt> then a truncation is done. If they
3315 are the same size, then nothing is done (<i>no-op cast</i>) other than a type
3320 %X = ptrtoint i32* %X to i8 <i>; yields truncation on 32-bit architecture</i>
3321 %Y = ptrtoint i32* %x to i64 <i>; yields zero extension on 32-bit architecture</i>
3325 <!-- _______________________________________________________________________ -->
3326 <div class="doc_subsubsection">
3327 <a name="i_inttoptr">'<tt>inttoptr .. to</tt>' Instruction</a>
3329 <div class="doc_text">
3333 <result> = inttoptr <ty> <value> to <ty2> <i>; yields ty2</i>
3337 <p>The '<tt>inttoptr</tt>' instruction converts an integer <tt>value</tt> to
3338 a pointer type, <tt>ty2</tt>.</p>
3341 <p>The '<tt>inttoptr</tt>' instruction takes an <a href="#t_integer">integer</a>
3342 value to cast, and a type to cast it to, which must be a
3343 <a href="#t_pointer">pointer</a> type.
3346 <p>The '<tt>inttoptr</tt>' instruction converts <tt>value</tt> to type
3347 <tt>ty2</tt> by applying either a zero extension or a truncation depending on
3348 the size of the integer <tt>value</tt>. If <tt>value</tt> is larger than the
3349 size of a pointer then a truncation is done. If <tt>value</tt> is smaller than
3350 the size of a pointer then a zero extension is done. If they are the same size,
3351 nothing is done (<i>no-op cast</i>).</p>
3355 %X = inttoptr i32 255 to i32* <i>; yields zero extension on 64-bit architecture</i>
3356 %X = inttoptr i32 255 to i32* <i>; yields no-op on 32-bit architecture</i>
3357 %Y = inttoptr i64 0 to i32* <i>; yields truncation on 32-bit architecture</i>
3361 <!-- _______________________________________________________________________ -->
3362 <div class="doc_subsubsection">
3363 <a name="i_bitcast">'<tt>bitcast .. to</tt>' Instruction</a>
3365 <div class="doc_text">
3369 <result> = bitcast <ty> <value> to <ty2> <i>; yields ty2</i>
3373 <p>The '<tt>bitcast</tt>' instruction converts <tt>value</tt> to type
3374 <tt>ty2</tt> without changing any bits.</p>
3377 <p>The '<tt>bitcast</tt>' instruction takes a value to cast, which must be
3378 a first class value, and a type to cast it to, which must also be a <a
3379 href="#t_firstclass">first class</a> type. The bit sizes of <tt>value</tt>
3380 and the destination type, <tt>ty2</tt>, must be identical. If the source
3381 type is a pointer, the destination type must also be a pointer.</p>
3384 <p>The '<tt>bitcast</tt>' instruction converts <tt>value</tt> to type
3385 <tt>ty2</tt>. It is always a <i>no-op cast</i> because no bits change with
3386 this conversion. The conversion is done as if the <tt>value</tt> had been
3387 stored to memory and read back as type <tt>ty2</tt>. Pointer types may only be
3388 converted to other pointer types with this instruction. To convert pointers to
3389 other types, use the <a href="#i_inttoptr">inttoptr</a> or
3390 <a href="#i_ptrtoint">ptrtoint</a> instructions first.</p>
3394 %X = bitcast i8 255 to i8 <i>; yields i8 :-1</i>
3395 %Y = bitcast i32* %x to sint* <i>; yields sint*:%x</i>
3396 %Z = bitcast <2xint> %V to i64; <i>; yields i64: %V</i>
3400 <!-- ======================================================================= -->
3401 <div class="doc_subsection"> <a name="otherops">Other Operations</a> </div>
3402 <div class="doc_text">
3403 <p>The instructions in this category are the "miscellaneous"
3404 instructions, which defy better classification.</p>
3407 <!-- _______________________________________________________________________ -->
3408 <div class="doc_subsubsection"><a name="i_icmp">'<tt>icmp</tt>' Instruction</a>
3410 <div class="doc_text">
3412 <pre> <result> = icmp <cond> <ty> <var1>, <var2> <i>; yields {i1}:result</i>
3415 <p>The '<tt>icmp</tt>' instruction returns a boolean value based on comparison
3416 of its two integer operands.</p>
3418 <p>The '<tt>icmp</tt>' instruction takes three operands. The first operand is
3419 the condition code indicating the kind of comparison to perform. It is not
3420 a value, just a keyword. The possible condition code are:
3422 <li><tt>eq</tt>: equal</li>
3423 <li><tt>ne</tt>: not equal </li>
3424 <li><tt>ugt</tt>: unsigned greater than</li>
3425 <li><tt>uge</tt>: unsigned greater or equal</li>
3426 <li><tt>ult</tt>: unsigned less than</li>
3427 <li><tt>ule</tt>: unsigned less or equal</li>
3428 <li><tt>sgt</tt>: signed greater than</li>
3429 <li><tt>sge</tt>: signed greater or equal</li>
3430 <li><tt>slt</tt>: signed less than</li>
3431 <li><tt>sle</tt>: signed less or equal</li>
3433 <p>The remaining two arguments must be <a href="#t_integer">integer</a> or
3434 <a href="#t_pointer">pointer</a> typed. They must also be identical types.</p>
3436 <p>The '<tt>icmp</tt>' compares <tt>var1</tt> and <tt>var2</tt> according to
3437 the condition code given as <tt>cond</tt>. The comparison performed always
3438 yields a <a href="#t_primitive">i1</a> result, as follows:
3440 <li><tt>eq</tt>: yields <tt>true</tt> if the operands are equal,
3441 <tt>false</tt> otherwise. No sign interpretation is necessary or performed.
3443 <li><tt>ne</tt>: yields <tt>true</tt> if the operands are unequal,
3444 <tt>false</tt> otherwise. No sign interpretation is necessary or performed.
3445 <li><tt>ugt</tt>: interprets the operands as unsigned values and yields
3446 <tt>true</tt> if <tt>var1</tt> is greater than <tt>var2</tt>.</li>
3447 <li><tt>uge</tt>: interprets the operands as unsigned values and yields
3448 <tt>true</tt> if <tt>var1</tt> is greater than or equal to <tt>var2</tt>.</li>
3449 <li><tt>ult</tt>: interprets the operands as unsigned values and yields
3450 <tt>true</tt> if <tt>var1</tt> is less than <tt>var2</tt>.</li>
3451 <li><tt>ule</tt>: interprets the operands as unsigned values and yields
3452 <tt>true</tt> if <tt>var1</tt> is less than or equal to <tt>var2</tt>.</li>
3453 <li><tt>sgt</tt>: interprets the operands as signed values and yields
3454 <tt>true</tt> if <tt>var1</tt> is greater than <tt>var2</tt>.</li>
3455 <li><tt>sge</tt>: interprets the operands as signed values and yields
3456 <tt>true</tt> if <tt>var1</tt> is greater than or equal to <tt>var2</tt>.</li>
3457 <li><tt>slt</tt>: interprets the operands as signed values and yields
3458 <tt>true</tt> if <tt>var1</tt> is less than <tt>var2</tt>.</li>
3459 <li><tt>sle</tt>: interprets the operands as signed values and yields
3460 <tt>true</tt> if <tt>var1</tt> is less than or equal to <tt>var2</tt>.</li>
3462 <p>If the operands are <a href="#t_pointer">pointer</a> typed, the pointer
3463 values are compared as if they were integers.</p>
3466 <pre> <result> = icmp eq i32 4, 5 <i>; yields: result=false</i>
3467 <result> = icmp ne float* %X, %X <i>; yields: result=false</i>
3468 <result> = icmp ult i16 4, 5 <i>; yields: result=true</i>
3469 <result> = icmp sgt i16 4, 5 <i>; yields: result=false</i>
3470 <result> = icmp ule i16 -4, 5 <i>; yields: result=false</i>
3471 <result> = icmp sge i16 4, 5 <i>; yields: result=false</i>
3475 <!-- _______________________________________________________________________ -->
3476 <div class="doc_subsubsection"><a name="i_fcmp">'<tt>fcmp</tt>' Instruction</a>
3478 <div class="doc_text">
3480 <pre> <result> = fcmp <cond> <ty> <var1>, <var2> <i>; yields {i1}:result</i>
3483 <p>The '<tt>fcmp</tt>' instruction returns a boolean value based on comparison
3484 of its floating point operands.</p>
3486 <p>The '<tt>fcmp</tt>' instruction takes three operands. The first operand is
3487 the condition code indicating the kind of comparison to perform. It is not
3488 a value, just a keyword. The possible condition code are:
3490 <li><tt>false</tt>: no comparison, always returns false</li>
3491 <li><tt>oeq</tt>: ordered and equal</li>
3492 <li><tt>ogt</tt>: ordered and greater than </li>
3493 <li><tt>oge</tt>: ordered and greater than or equal</li>
3494 <li><tt>olt</tt>: ordered and less than </li>
3495 <li><tt>ole</tt>: ordered and less than or equal</li>
3496 <li><tt>one</tt>: ordered and not equal</li>
3497 <li><tt>ord</tt>: ordered (no nans)</li>
3498 <li><tt>ueq</tt>: unordered or equal</li>
3499 <li><tt>ugt</tt>: unordered or greater than </li>
3500 <li><tt>uge</tt>: unordered or greater than or equal</li>
3501 <li><tt>ult</tt>: unordered or less than </li>
3502 <li><tt>ule</tt>: unordered or less than or equal</li>
3503 <li><tt>une</tt>: unordered or not equal</li>
3504 <li><tt>uno</tt>: unordered (either nans)</li>
3505 <li><tt>true</tt>: no comparison, always returns true</li>
3507 <p><i>Ordered</i> means that neither operand is a QNAN while
3508 <i>unordered</i> means that either operand may be a QNAN.</p>
3509 <p>The <tt>val1</tt> and <tt>val2</tt> arguments must be
3510 <a href="#t_floating">floating point</a> typed. They must have identical
3513 <p>The '<tt>fcmp</tt>' compares <tt>var1</tt> and <tt>var2</tt> according to
3514 the condition code given as <tt>cond</tt>. The comparison performed always
3515 yields a <a href="#t_primitive">i1</a> result, as follows:
3517 <li><tt>false</tt>: always yields <tt>false</tt>, regardless of operands.</li>
3518 <li><tt>oeq</tt>: yields <tt>true</tt> if both operands are not a QNAN and
3519 <tt>var1</tt> is equal to <tt>var2</tt>.</li>
3520 <li><tt>ogt</tt>: yields <tt>true</tt> if both operands are not a QNAN and
3521 <tt>var1</tt> is greather than <tt>var2</tt>.</li>
3522 <li><tt>oge</tt>: yields <tt>true</tt> if both operands are not a QNAN and
3523 <tt>var1</tt> is greater than or equal to <tt>var2</tt>.</li>
3524 <li><tt>olt</tt>: yields <tt>true</tt> if both operands are not a QNAN and
3525 <tt>var1</tt> is less than <tt>var2</tt>.</li>
3526 <li><tt>ole</tt>: yields <tt>true</tt> if both operands are not a QNAN and
3527 <tt>var1</tt> is less than or equal to <tt>var2</tt>.</li>
3528 <li><tt>one</tt>: yields <tt>true</tt> if both operands are not a QNAN and
3529 <tt>var1</tt> is not equal to <tt>var2</tt>.</li>
3530 <li><tt>ord</tt>: yields <tt>true</tt> if both operands are not a QNAN.</li>
3531 <li><tt>ueq</tt>: yields <tt>true</tt> if either operand is a QNAN or
3532 <tt>var1</tt> is equal to <tt>var2</tt>.</li>
3533 <li><tt>ugt</tt>: yields <tt>true</tt> if either operand is a QNAN or
3534 <tt>var1</tt> is greater than <tt>var2</tt>.</li>
3535 <li><tt>uge</tt>: yields <tt>true</tt> if either operand is a QNAN or
3536 <tt>var1</tt> is greater than or equal to <tt>var2</tt>.</li>
3537 <li><tt>ult</tt>: yields <tt>true</tt> if either operand is a QNAN or
3538 <tt>var1</tt> is less than <tt>var2</tt>.</li>
3539 <li><tt>ule</tt>: yields <tt>true</tt> if either operand is a QNAN or
3540 <tt>var1</tt> is less than or equal to <tt>var2</tt>.</li>
3541 <li><tt>une</tt>: yields <tt>true</tt> if either operand is a QNAN or
3542 <tt>var1</tt> is not equal to <tt>var2</tt>.</li>
3543 <li><tt>uno</tt>: yields <tt>true</tt> if either operand is a QNAN.</li>
3544 <li><tt>true</tt>: always yields <tt>true</tt>, regardless of operands.</li>
3548 <pre> <result> = fcmp oeq float 4.0, 5.0 <i>; yields: result=false</i>
3549 <result> = icmp one float 4.0, 5.0 <i>; yields: result=true</i>
3550 <result> = icmp olt float 4.0, 5.0 <i>; yields: result=true</i>
3551 <result> = icmp ueq double 1.0, 2.0 <i>; yields: result=false</i>
3555 <!-- _______________________________________________________________________ -->
3556 <div class="doc_subsubsection"> <a name="i_phi">'<tt>phi</tt>'
3557 Instruction</a> </div>
3558 <div class="doc_text">
3560 <pre> <result> = phi <ty> [ <val0>, <label0>], ...<br></pre>
3562 <p>The '<tt>phi</tt>' instruction is used to implement the φ node in
3563 the SSA graph representing the function.</p>
3565 <p>The type of the incoming values is specified with the first type
3566 field. After this, the '<tt>phi</tt>' instruction takes a list of pairs
3567 as arguments, with one pair for each predecessor basic block of the
3568 current block. Only values of <a href="#t_firstclass">first class</a>
3569 type may be used as the value arguments to the PHI node. Only labels
3570 may be used as the label arguments.</p>
3571 <p>There must be no non-phi instructions between the start of a basic
3572 block and the PHI instructions: i.e. PHI instructions must be first in
3575 <p>At runtime, the '<tt>phi</tt>' instruction logically takes on the value
3576 specified by the pair corresponding to the predecessor basic block that executed
3577 just prior to the current block.</p>
3579 <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>
3582 <!-- _______________________________________________________________________ -->
3583 <div class="doc_subsubsection">
3584 <a name="i_select">'<tt>select</tt>' Instruction</a>
3587 <div class="doc_text">
3592 <result> = select i1 <cond>, <ty> <val1>, <ty> <val2> <i>; yields ty</i>
3598 The '<tt>select</tt>' instruction is used to choose one value based on a
3599 condition, without branching.
3606 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.
3612 If the boolean condition evaluates to true, the instruction returns the first
3613 value argument; otherwise, it returns the second value argument.
3619 %X = select i1 true, i8 17, i8 42 <i>; yields i8:17</i>
3624 <!-- _______________________________________________________________________ -->
3625 <div class="doc_subsubsection">
3626 <a name="i_call">'<tt>call</tt>' Instruction</a>
3629 <div class="doc_text">
3633 <result> = [tail] call [<a href="#callingconv">cconv</a>] <ty> [<fnty>*] <fnptrval>(<param list>)
3638 <p>The '<tt>call</tt>' instruction represents a simple function call.</p>
3642 <p>This instruction requires several arguments:</p>
3646 <p>The optional "tail" marker indicates whether the callee function accesses
3647 any allocas or varargs in the caller. If the "tail" marker is present, the
3648 function call is eligible for tail call optimization. Note that calls may
3649 be marked "tail" even if they do not occur before a <a
3650 href="#i_ret"><tt>ret</tt></a> instruction.
3653 <p>The optional "cconv" marker indicates which <a href="#callingconv">calling
3654 convention</a> the call should use. If none is specified, the call defaults
3655 to using C calling conventions.
3658 <p>'<tt>ty</tt>': the type of the call instruction itself which is also
3659 the type of the return value. Functions that return no value are marked
3660 <tt><a href="#t_void">void</a></tt>.</p>
3663 <p>'<tt>fnty</tt>': shall be the signature of the pointer to function
3664 value being invoked. The argument types must match the types implied by
3665 this signature. This type can be omitted if the function is not varargs
3666 and if the function type does not return a pointer to a function.</p>
3669 <p>'<tt>fnptrval</tt>': An LLVM value containing a pointer to a function to
3670 be invoked. In most cases, this is a direct function invocation, but
3671 indirect <tt>call</tt>s are just as possible, calling an arbitrary pointer
3672 to function value.</p>
3675 <p>'<tt>function args</tt>': argument list whose types match the
3676 function signature argument types. All arguments must be of
3677 <a href="#t_firstclass">first class</a> type. If the function signature
3678 indicates the function accepts a variable number of arguments, the extra
3679 arguments can be specified.</p>
3685 <p>The '<tt>call</tt>' instruction is used to cause control flow to
3686 transfer to a specified function, with its incoming arguments bound to
3687 the specified values. Upon a '<tt><a href="#i_ret">ret</a></tt>'
3688 instruction in the called function, control flow continues with the
3689 instruction after the function call, and the return value of the
3690 function is bound to the result argument. This is a simpler case of
3691 the <a href="#i_invoke">invoke</a> instruction.</p>
3696 %retval = call i32 @test(i32 %argc)
3697 call i32 (i8 *, ...)* @printf(i8 * %msg, i32 12, i8 42);
3698 %X = tail call i32 @foo()
3699 %Y = tail call <a href="#callingconv">fastcc</a> i32 @foo()
3700 %Z = call void %foo(i8 97 signext)
3705 <!-- _______________________________________________________________________ -->
3706 <div class="doc_subsubsection">
3707 <a name="i_va_arg">'<tt>va_arg</tt>' Instruction</a>
3710 <div class="doc_text">
3715 <resultval> = va_arg <va_list*> <arglist>, <argty>
3720 <p>The '<tt>va_arg</tt>' instruction is used to access arguments passed through
3721 the "variable argument" area of a function call. It is used to implement the
3722 <tt>va_arg</tt> macro in C.</p>
3726 <p>This instruction takes a <tt>va_list*</tt> value and the type of
3727 the argument. It returns a value of the specified argument type and
3728 increments the <tt>va_list</tt> to point to the next argument. The
3729 actual type of <tt>va_list</tt> is target specific.</p>
3733 <p>The '<tt>va_arg</tt>' instruction loads an argument of the specified
3734 type from the specified <tt>va_list</tt> and causes the
3735 <tt>va_list</tt> to point to the next argument. For more information,
3736 see the variable argument handling <a href="#int_varargs">Intrinsic
3739 <p>It is legal for this instruction to be called in a function which does not
3740 take a variable number of arguments, for example, the <tt>vfprintf</tt>
3743 <p><tt>va_arg</tt> is an LLVM instruction instead of an <a
3744 href="#intrinsics">intrinsic function</a> because it takes a type as an
3749 <p>See the <a href="#int_varargs">variable argument processing</a> section.</p>
3753 <!-- *********************************************************************** -->
3754 <div class="doc_section"> <a name="intrinsics">Intrinsic Functions</a> </div>
3755 <!-- *********************************************************************** -->
3757 <div class="doc_text">
3759 <p>LLVM supports the notion of an "intrinsic function". These functions have
3760 well known names and semantics and are required to follow certain restrictions.
3761 Overall, these intrinsics represent an extension mechanism for the LLVM
3762 language that does not require changing all of the transformations in LLVM when
3763 adding to the language (or the bitcode reader/writer, the parser, etc...).</p>
3765 <p>Intrinsic function names must all start with an "<tt>llvm.</tt>" prefix. This
3766 prefix is reserved in LLVM for intrinsic names; thus, function names may not
3767 begin with this prefix. Intrinsic functions must always be external functions:
3768 you cannot define the body of intrinsic functions. Intrinsic functions may
3769 only be used in call or invoke instructions: it is illegal to take the address
3770 of an intrinsic function. Additionally, because intrinsic functions are part
3771 of the LLVM language, it is required if any are added that they be documented
3774 <p>Some intrinsic functions can be overloaded, i.e., the intrinsic represents
3775 a family of functions that perform the same operation but on different data
3776 types. Because LLVM can represent over 8 million different integer types,
3777 overloading is used commonly to allow an intrinsic function to operate on any
3778 integer type. One or more of the argument types or the result type can be
3779 overloaded to accept any integer type. Argument types may also be defined as
3780 exactly matching a previous argument's type or the result type. This allows an
3781 intrinsic function which accepts multiple arguments, but needs all of them to
3782 be of the same type, to only be overloaded with respect to a single argument or
3785 <p>Overloaded intrinsics will have the names of its overloaded argument types
3786 encoded into its function name, each preceded by a period. Only those types
3787 which are overloaded result in a name suffix. Arguments whose type is matched
3788 against another type do not. For example, the <tt>llvm.ctpop</tt> function can
3789 take an integer of any width and returns an integer of exactly the same integer
3790 width. This leads to a family of functions such as
3791 <tt>i8 @llvm.ctpop.i8(i8 %val)</tt> and <tt>i29 @llvm.ctpop.i29(i29 %val)</tt>.
3792 Only one type, the return type, is overloaded, and only one type suffix is
3793 required. Because the argument's type is matched against the return type, it
3794 does not require its own name suffix.</p>
3796 <p>To learn how to add an intrinsic function, please see the
3797 <a href="ExtendingLLVM.html">Extending LLVM Guide</a>.
3802 <!-- ======================================================================= -->
3803 <div class="doc_subsection">
3804 <a name="int_varargs">Variable Argument Handling Intrinsics</a>
3807 <div class="doc_text">
3809 <p>Variable argument support is defined in LLVM with the <a
3810 href="#i_va_arg"><tt>va_arg</tt></a> instruction and these three
3811 intrinsic functions. These functions are related to the similarly
3812 named macros defined in the <tt><stdarg.h></tt> header file.</p>
3814 <p>All of these functions operate on arguments that use a
3815 target-specific value type "<tt>va_list</tt>". The LLVM assembly
3816 language reference manual does not define what this type is, so all
3817 transformations should be prepared to handle these functions regardless of
3820 <p>This example shows how the <a href="#i_va_arg"><tt>va_arg</tt></a>
3821 instruction and the variable argument handling intrinsic functions are
3824 <div class="doc_code">
3826 define i32 @test(i32 %X, ...) {
3827 ; Initialize variable argument processing
3829 %ap2 = bitcast i8** %ap to i8*
3830 call void @llvm.va_start(i8* %ap2)
3832 ; Read a single integer argument
3833 %tmp = va_arg i8** %ap, i32
3835 ; Demonstrate usage of llvm.va_copy and llvm.va_end
3837 %aq2 = bitcast i8** %aq to i8*
3838 call void @llvm.va_copy(i8* %aq2, i8* %ap2)
3839 call void @llvm.va_end(i8* %aq2)
3841 ; Stop processing of arguments.
3842 call void @llvm.va_end(i8* %ap2)
3846 declare void @llvm.va_start(i8*)
3847 declare void @llvm.va_copy(i8*, i8*)
3848 declare void @llvm.va_end(i8*)
3854 <!-- _______________________________________________________________________ -->
3855 <div class="doc_subsubsection">
3856 <a name="int_va_start">'<tt>llvm.va_start</tt>' Intrinsic</a>
3860 <div class="doc_text">
3862 <pre> declare void %llvm.va_start(i8* <arglist>)<br></pre>
3864 <P>The '<tt>llvm.va_start</tt>' intrinsic initializes
3865 <tt>*<arglist></tt> for subsequent use by <tt><a
3866 href="#i_va_arg">va_arg</a></tt>.</p>
3870 <P>The argument is a pointer to a <tt>va_list</tt> element to initialize.</p>
3874 <P>The '<tt>llvm.va_start</tt>' intrinsic works just like the <tt>va_start</tt>
3875 macro available in C. In a target-dependent way, it initializes the
3876 <tt>va_list</tt> element to which the argument points, so that the next call to
3877 <tt>va_arg</tt> will produce the first variable argument passed to the function.
3878 Unlike the C <tt>va_start</tt> macro, this intrinsic does not need to know the
3879 last argument of the function as the compiler can figure that out.</p>
3883 <!-- _______________________________________________________________________ -->
3884 <div class="doc_subsubsection">
3885 <a name="int_va_end">'<tt>llvm.va_end</tt>' Intrinsic</a>
3888 <div class="doc_text">
3890 <pre> declare void @llvm.va_end(i8* <arglist>)<br></pre>
3893 <p>The '<tt>llvm.va_end</tt>' intrinsic destroys <tt>*<arglist></tt>,
3894 which has been initialized previously with <tt><a href="#int_va_start">llvm.va_start</a></tt>
3895 or <tt><a href="#i_va_copy">llvm.va_copy</a></tt>.</p>
3899 <p>The argument is a pointer to a <tt>va_list</tt> to destroy.</p>
3903 <p>The '<tt>llvm.va_end</tt>' intrinsic works just like the <tt>va_end</tt>
3904 macro available in C. In a target-dependent way, it destroys the
3905 <tt>va_list</tt> element to which the argument points. Calls to <a
3906 href="#int_va_start"><tt>llvm.va_start</tt></a> and <a href="#int_va_copy">
3907 <tt>llvm.va_copy</tt></a> must be matched exactly with calls to
3908 <tt>llvm.va_end</tt>.</p>
3912 <!-- _______________________________________________________________________ -->
3913 <div class="doc_subsubsection">
3914 <a name="int_va_copy">'<tt>llvm.va_copy</tt>' Intrinsic</a>
3917 <div class="doc_text">
3922 declare void @llvm.va_copy(i8* <destarglist>, i8* <srcarglist>)
3927 <p>The '<tt>llvm.va_copy</tt>' intrinsic copies the current argument position
3928 from the source argument list to the destination argument list.</p>
3932 <p>The first argument is a pointer to a <tt>va_list</tt> element to initialize.
3933 The second argument is a pointer to a <tt>va_list</tt> element to copy from.</p>
3938 <p>The '<tt>llvm.va_copy</tt>' intrinsic works just like the <tt>va_copy</tt>
3939 macro available in C. In a target-dependent way, it copies the source
3940 <tt>va_list</tt> element into the destination <tt>va_list</tt> element. This
3941 intrinsic is necessary because the <tt><a href="#int_va_start">
3942 llvm.va_start</a></tt> intrinsic may be arbitrarily complex and require, for
3943 example, memory allocation.</p>
3947 <!-- ======================================================================= -->
3948 <div class="doc_subsection">
3949 <a name="int_gc">Accurate Garbage Collection Intrinsics</a>
3952 <div class="doc_text">
3955 LLVM support for <a href="GarbageCollection.html">Accurate Garbage
3956 Collection</a> requires the implementation and generation of these intrinsics.
3957 These intrinsics allow identification of <a href="#int_gcroot">GC roots on the
3958 stack</a>, as well as garbage collector implementations that require <a
3959 href="#int_gcread">read</a> and <a href="#int_gcwrite">write</a> barriers.
3960 Front-ends for type-safe garbage collected languages should generate these
3961 intrinsics to make use of the LLVM garbage collectors. For more details, see <a
3962 href="GarbageCollection.html">Accurate Garbage Collection with LLVM</a>.
3965 <p>The garbage collection intrinsics only operate on objects in the generic
3966 address space (address space zero).</p>
3970 <!-- _______________________________________________________________________ -->
3971 <div class="doc_subsubsection">
3972 <a name="int_gcroot">'<tt>llvm.gcroot</tt>' Intrinsic</a>
3975 <div class="doc_text">
3980 declare void @llvm.gcroot(i8** %ptrloc, i8* %metadata)
3985 <p>The '<tt>llvm.gcroot</tt>' intrinsic declares the existence of a GC root to
3986 the code generator, and allows some metadata to be associated with it.</p>
3990 <p>The first argument specifies the address of a stack object that contains the
3991 root pointer. The second pointer (which must be either a constant or a global
3992 value address) contains the meta-data to be associated with the root.</p>
3996 <p>At runtime, a call to this intrinsics stores a null pointer into the "ptrloc"
3997 location. At compile-time, the code generator generates information to allow
3998 the runtime to find the pointer at GC safe points. The '<tt>llvm.gcroot</tt>'
3999 intrinsic may only be used in a function which <a href="#gc">specifies a GC
4005 <!-- _______________________________________________________________________ -->
4006 <div class="doc_subsubsection">
4007 <a name="int_gcread">'<tt>llvm.gcread</tt>' Intrinsic</a>
4010 <div class="doc_text">
4015 declare i8* @llvm.gcread(i8* %ObjPtr, i8** %Ptr)
4020 <p>The '<tt>llvm.gcread</tt>' intrinsic identifies reads of references from heap
4021 locations, allowing garbage collector implementations that require read
4026 <p>The second argument is the address to read from, which should be an address
4027 allocated from the garbage collector. The first object is a pointer to the
4028 start of the referenced object, if needed by the language runtime (otherwise
4033 <p>The '<tt>llvm.gcread</tt>' intrinsic has the same semantics as a load
4034 instruction, but may be replaced with substantially more complex code by the
4035 garbage collector runtime, as needed. The '<tt>llvm.gcread</tt>' intrinsic
4036 may only be used in a function which <a href="#gc">specifies a GC
4042 <!-- _______________________________________________________________________ -->
4043 <div class="doc_subsubsection">
4044 <a name="int_gcwrite">'<tt>llvm.gcwrite</tt>' Intrinsic</a>
4047 <div class="doc_text">
4052 declare void @llvm.gcwrite(i8* %P1, i8* %Obj, i8** %P2)
4057 <p>The '<tt>llvm.gcwrite</tt>' intrinsic identifies writes of references to heap
4058 locations, allowing garbage collector implementations that require write
4059 barriers (such as generational or reference counting collectors).</p>
4063 <p>The first argument is the reference to store, the second is the start of the
4064 object to store it to, and the third is the address of the field of Obj to
4065 store to. If the runtime does not require a pointer to the object, Obj may be
4070 <p>The '<tt>llvm.gcwrite</tt>' intrinsic has the same semantics as a store
4071 instruction, but may be replaced with substantially more complex code by the
4072 garbage collector runtime, as needed. The '<tt>llvm.gcwrite</tt>' intrinsic
4073 may only be used in a function which <a href="#gc">specifies a GC
4080 <!-- ======================================================================= -->
4081 <div class="doc_subsection">
4082 <a name="int_codegen">Code Generator Intrinsics</a>
4085 <div class="doc_text">
4087 These intrinsics are provided by LLVM to expose special features that may only
4088 be implemented with code generator support.
4093 <!-- _______________________________________________________________________ -->
4094 <div class="doc_subsubsection">
4095 <a name="int_returnaddress">'<tt>llvm.returnaddress</tt>' Intrinsic</a>
4098 <div class="doc_text">
4102 declare i8 *@llvm.returnaddress(i32 <level>)
4108 The '<tt>llvm.returnaddress</tt>' intrinsic attempts to compute a
4109 target-specific value indicating the return address of the current function
4110 or one of its callers.
4116 The argument to this intrinsic indicates which function to return the address
4117 for. Zero indicates the calling function, one indicates its caller, etc. The
4118 argument is <b>required</b> to be a constant integer value.
4124 The '<tt>llvm.returnaddress</tt>' intrinsic either returns a pointer indicating
4125 the return address of the specified call frame, or zero if it cannot be
4126 identified. The value returned by this intrinsic is likely to be incorrect or 0
4127 for arguments other than zero, so it should only be used for debugging purposes.
4131 Note that calling this intrinsic does not prevent function inlining or other
4132 aggressive transformations, so the value returned may not be that of the obvious
4133 source-language caller.
4138 <!-- _______________________________________________________________________ -->
4139 <div class="doc_subsubsection">
4140 <a name="int_frameaddress">'<tt>llvm.frameaddress</tt>' Intrinsic</a>
4143 <div class="doc_text">
4147 declare i8 *@llvm.frameaddress(i32 <level>)
4153 The '<tt>llvm.frameaddress</tt>' intrinsic attempts to return the
4154 target-specific frame pointer value for the specified stack frame.
4160 The argument to this intrinsic indicates which function to return the frame
4161 pointer for. Zero indicates the calling function, one indicates its caller,
4162 etc. The argument is <b>required</b> to be a constant integer value.
4168 The '<tt>llvm.frameaddress</tt>' intrinsic either returns a pointer indicating
4169 the frame address of the specified call frame, or zero if it cannot be
4170 identified. The value returned by this intrinsic is likely to be incorrect or 0
4171 for arguments other than zero, so it should only be used for debugging purposes.
4175 Note that calling this intrinsic does not prevent function inlining or other
4176 aggressive transformations, so the value returned may not be that of the obvious
4177 source-language caller.
4181 <!-- _______________________________________________________________________ -->
4182 <div class="doc_subsubsection">
4183 <a name="int_stacksave">'<tt>llvm.stacksave</tt>' Intrinsic</a>
4186 <div class="doc_text">
4190 declare i8 *@llvm.stacksave()
4196 The '<tt>llvm.stacksave</tt>' intrinsic is used to remember the current state of
4197 the function stack, for use with <a href="#int_stackrestore">
4198 <tt>llvm.stackrestore</tt></a>. This is useful for implementing language
4199 features like scoped automatic variable sized arrays in C99.
4205 This intrinsic returns a opaque pointer value that can be passed to <a
4206 href="#int_stackrestore"><tt>llvm.stackrestore</tt></a>. When an
4207 <tt>llvm.stackrestore</tt> intrinsic is executed with a value saved from
4208 <tt>llvm.stacksave</tt>, it effectively restores the state of the stack to the
4209 state it was in when the <tt>llvm.stacksave</tt> intrinsic executed. In
4210 practice, this pops any <a href="#i_alloca">alloca</a> blocks from the stack
4211 that were allocated after the <tt>llvm.stacksave</tt> was executed.
4216 <!-- _______________________________________________________________________ -->
4217 <div class="doc_subsubsection">
4218 <a name="int_stackrestore">'<tt>llvm.stackrestore</tt>' Intrinsic</a>
4221 <div class="doc_text">
4225 declare void @llvm.stackrestore(i8 * %ptr)
4231 The '<tt>llvm.stackrestore</tt>' intrinsic is used to restore the state of
4232 the function stack to the state it was in when the corresponding <a
4233 href="#int_stacksave"><tt>llvm.stacksave</tt></a> intrinsic executed. This is
4234 useful for implementing language features like scoped automatic variable sized
4241 See the description for <a href="#int_stacksave"><tt>llvm.stacksave</tt></a>.
4247 <!-- _______________________________________________________________________ -->
4248 <div class="doc_subsubsection">
4249 <a name="int_prefetch">'<tt>llvm.prefetch</tt>' Intrinsic</a>
4252 <div class="doc_text">
4256 declare void @llvm.prefetch(i8* <address>, i32 <rw>, i32 <locality>)
4263 The '<tt>llvm.prefetch</tt>' intrinsic is a hint to the code generator to insert
4264 a prefetch instruction if supported; otherwise, it is a noop. Prefetches have
4266 effect on the behavior of the program but can change its performance
4273 <tt>address</tt> is the address to be prefetched, <tt>rw</tt> is the specifier
4274 determining if the fetch should be for a read (0) or write (1), and
4275 <tt>locality</tt> is a temporal locality specifier ranging from (0) - no
4276 locality, to (3) - extremely local keep in cache. The <tt>rw</tt> and
4277 <tt>locality</tt> arguments must be constant integers.
4283 This intrinsic does not modify the behavior of the program. In particular,
4284 prefetches cannot trap and do not produce a value. On targets that support this
4285 intrinsic, the prefetch can provide hints to the processor cache for better
4291 <!-- _______________________________________________________________________ -->
4292 <div class="doc_subsubsection">
4293 <a name="int_pcmarker">'<tt>llvm.pcmarker</tt>' Intrinsic</a>
4296 <div class="doc_text">
4300 declare void @llvm.pcmarker(i32 <id>)
4307 The '<tt>llvm.pcmarker</tt>' intrinsic is a method to export a Program Counter
4309 code to simulators and other tools. The method is target specific, but it is
4310 expected that the marker will use exported symbols to transmit the PC of the marker.
4311 The marker makes no guarantees that it will remain with any specific instruction
4312 after optimizations. It is possible that the presence of a marker will inhibit
4313 optimizations. The intended use is to be inserted after optimizations to allow
4314 correlations of simulation runs.
4320 <tt>id</tt> is a numerical id identifying the marker.
4326 This intrinsic does not modify the behavior of the program. Backends that do not
4327 support this intrinisic may ignore it.
4332 <!-- _______________________________________________________________________ -->
4333 <div class="doc_subsubsection">
4334 <a name="int_readcyclecounter">'<tt>llvm.readcyclecounter</tt>' Intrinsic</a>
4337 <div class="doc_text">
4341 declare i64 @llvm.readcyclecounter( )
4348 The '<tt>llvm.readcyclecounter</tt>' intrinsic provides access to the cycle
4349 counter register (or similar low latency, high accuracy clocks) on those targets
4350 that support it. On X86, it should map to RDTSC. On Alpha, it should map to RPCC.
4351 As the backing counters overflow quickly (on the order of 9 seconds on alpha), this
4352 should only be used for small timings.
4358 When directly supported, reading the cycle counter should not modify any memory.
4359 Implementations are allowed to either return a application specific value or a
4360 system wide value. On backends without support, this is lowered to a constant 0.
4365 <!-- ======================================================================= -->
4366 <div class="doc_subsection">
4367 <a name="int_libc">Standard C Library Intrinsics</a>
4370 <div class="doc_text">
4372 LLVM provides intrinsics for a few important standard C library functions.
4373 These intrinsics allow source-language front-ends to pass information about the
4374 alignment of the pointer arguments to the code generator, providing opportunity
4375 for more efficient code generation.
4380 <!-- _______________________________________________________________________ -->
4381 <div class="doc_subsubsection">
4382 <a name="int_memcpy">'<tt>llvm.memcpy</tt>' Intrinsic</a>
4385 <div class="doc_text">
4389 declare void @llvm.memcpy.i32(i8 * <dest>, i8 * <src>,
4390 i32 <len>, i32 <align>)
4391 declare void @llvm.memcpy.i64(i8 * <dest>, i8 * <src>,
4392 i64 <len>, i32 <align>)
4398 The '<tt>llvm.memcpy.*</tt>' intrinsics copy a block of memory from the source
4399 location to the destination location.
4403 Note that, unlike the standard libc function, the <tt>llvm.memcpy.*</tt>
4404 intrinsics do not return a value, and takes an extra alignment argument.
4410 The first argument is a pointer to the destination, the second is a pointer to
4411 the source. The third argument is an integer argument
4412 specifying the number of bytes to copy, and the fourth argument is the alignment
4413 of the source and destination locations.
4417 If the call to this intrinisic has an alignment value that is not 0 or 1, then
4418 the caller guarantees that both the source and destination pointers are aligned
4425 The '<tt>llvm.memcpy.*</tt>' intrinsics copy a block of memory from the source
4426 location to the destination location, which are not allowed to overlap. It
4427 copies "len" bytes of memory over. If the argument is known to be aligned to
4428 some boundary, this can be specified as the fourth argument, otherwise it should
4434 <!-- _______________________________________________________________________ -->
4435 <div class="doc_subsubsection">
4436 <a name="int_memmove">'<tt>llvm.memmove</tt>' Intrinsic</a>
4439 <div class="doc_text">
4443 declare void @llvm.memmove.i32(i8 * <dest>, i8 * <src>,
4444 i32 <len>, i32 <align>)
4445 declare void @llvm.memmove.i64(i8 * <dest>, i8 * <src>,
4446 i64 <len>, i32 <align>)
4452 The '<tt>llvm.memmove.*</tt>' intrinsics move a block of memory from the source
4453 location to the destination location. It is similar to the
4454 '<tt>llvm.memcmp</tt>' intrinsic but allows the two memory locations to overlap.
4458 Note that, unlike the standard libc function, the <tt>llvm.memmove.*</tt>
4459 intrinsics do not return a value, and takes an extra alignment argument.
4465 The first argument is a pointer to the destination, the second is a pointer to
4466 the source. The third argument is an integer argument
4467 specifying the number of bytes to copy, and the fourth argument is the alignment
4468 of the source and destination locations.
4472 If the call to this intrinisic has an alignment value that is not 0 or 1, then
4473 the caller guarantees that the source and destination pointers are aligned to
4480 The '<tt>llvm.memmove.*</tt>' intrinsics copy a block of memory from the source
4481 location to the destination location, which may overlap. It
4482 copies "len" bytes of memory over. If the argument is known to be aligned to
4483 some boundary, this can be specified as the fourth argument, otherwise it should
4489 <!-- _______________________________________________________________________ -->
4490 <div class="doc_subsubsection">
4491 <a name="int_memset">'<tt>llvm.memset.*</tt>' Intrinsics</a>
4494 <div class="doc_text">
4498 declare void @llvm.memset.i32(i8 * <dest>, i8 <val>,
4499 i32 <len>, i32 <align>)
4500 declare void @llvm.memset.i64(i8 * <dest>, i8 <val>,
4501 i64 <len>, i32 <align>)
4507 The '<tt>llvm.memset.*</tt>' intrinsics fill a block of memory with a particular
4512 Note that, unlike the standard libc function, the <tt>llvm.memset</tt> intrinsic
4513 does not return a value, and takes an extra alignment argument.
4519 The first argument is a pointer to the destination to fill, the second is the
4520 byte value to fill it with, the third argument is an integer
4521 argument specifying the number of bytes to fill, and the fourth argument is the
4522 known alignment of destination location.
4526 If the call to this intrinisic has an alignment value that is not 0 or 1, then
4527 the caller guarantees that the destination pointer is aligned to that boundary.
4533 The '<tt>llvm.memset.*</tt>' intrinsics fill "len" bytes of memory starting at
4535 destination location. If the argument is known to be aligned to some boundary,
4536 this can be specified as the fourth argument, otherwise it should be set to 0 or
4542 <!-- _______________________________________________________________________ -->
4543 <div class="doc_subsubsection">
4544 <a name="int_sqrt">'<tt>llvm.sqrt.*</tt>' Intrinsic</a>
4547 <div class="doc_text">
4550 <p>This is an overloaded intrinsic. You can use <tt>llvm.sqrt</tt> on any
4551 floating point or vector of floating point type. Not all targets support all
4554 declare float @llvm.sqrt.f32(float %Val)
4555 declare double @llvm.sqrt.f64(double %Val)
4556 declare x86_fp80 @llvm.sqrt.f80(x86_fp80 %Val)
4557 declare fp128 @llvm.sqrt.f128(fp128 %Val)
4558 declare ppc_fp128 @llvm.sqrt.ppcf128(ppc_fp128 %Val)
4564 The '<tt>llvm.sqrt</tt>' intrinsics return the sqrt of the specified operand,
4565 returning the same value as the libm '<tt>sqrt</tt>' functions would. Unlike
4566 <tt>sqrt</tt> in libm, however, <tt>llvm.sqrt</tt> has undefined behavior for
4567 negative numbers (which allows for better optimization).
4573 The argument and return value are floating point numbers of the same type.
4579 This function returns the sqrt of the specified operand if it is a nonnegative
4580 floating point number.
4584 <!-- _______________________________________________________________________ -->
4585 <div class="doc_subsubsection">
4586 <a name="int_powi">'<tt>llvm.powi.*</tt>' Intrinsic</a>
4589 <div class="doc_text">
4592 <p>This is an overloaded intrinsic. You can use <tt>llvm.powi</tt> on any
4593 floating point or vector of floating point type. Not all targets support all
4596 declare float @llvm.powi.f32(float %Val, i32 %power)
4597 declare double @llvm.powi.f64(double %Val, i32 %power)
4598 declare x86_fp80 @llvm.powi.f80(x86_fp80 %Val, i32 %power)
4599 declare fp128 @llvm.powi.f128(fp128 %Val, i32 %power)
4600 declare ppc_fp128 @llvm.powi.ppcf128(ppc_fp128 %Val, i32 %power)
4606 The '<tt>llvm.powi.*</tt>' intrinsics return the first operand raised to the
4607 specified (positive or negative) power. The order of evaluation of
4608 multiplications is not defined. When a vector of floating point type is
4609 used, the second argument remains a scalar integer value.
4615 The second argument is an integer power, and the first is a value to raise to
4622 This function returns the first value raised to the second power with an
4623 unspecified sequence of rounding operations.</p>
4626 <!-- _______________________________________________________________________ -->
4627 <div class="doc_subsubsection">
4628 <a name="int_sin">'<tt>llvm.sin.*</tt>' Intrinsic</a>
4631 <div class="doc_text">
4634 <p>This is an overloaded intrinsic. You can use <tt>llvm.sin</tt> on any
4635 floating point or vector of floating point type. Not all targets support all
4638 declare float @llvm.sin.f32(float %Val)
4639 declare double @llvm.sin.f64(double %Val)
4640 declare x86_fp80 @llvm.sin.f80(x86_fp80 %Val)
4641 declare fp128 @llvm.sin.f128(fp128 %Val)
4642 declare ppc_fp128 @llvm.sin.ppcf128(ppc_fp128 %Val)
4648 The '<tt>llvm.sin.*</tt>' intrinsics return the sine of the operand.
4654 The argument and return value are floating point numbers of the same type.
4660 This function returns the sine of the specified operand, returning the
4661 same values as the libm <tt>sin</tt> functions would, and handles error
4662 conditions in the same way.</p>
4665 <!-- _______________________________________________________________________ -->
4666 <div class="doc_subsubsection">
4667 <a name="int_cos">'<tt>llvm.cos.*</tt>' Intrinsic</a>
4670 <div class="doc_text">
4673 <p>This is an overloaded intrinsic. You can use <tt>llvm.cos</tt> on any
4674 floating point or vector of floating point type. Not all targets support all
4677 declare float @llvm.cos.f32(float %Val)
4678 declare double @llvm.cos.f64(double %Val)
4679 declare x86_fp80 @llvm.cos.f80(x86_fp80 %Val)
4680 declare fp128 @llvm.cos.f128(fp128 %Val)
4681 declare ppc_fp128 @llvm.cos.ppcf128(ppc_fp128 %Val)
4687 The '<tt>llvm.cos.*</tt>' intrinsics return the cosine of the operand.
4693 The argument and return value are floating point numbers of the same type.
4699 This function returns the cosine of the specified operand, returning the
4700 same values as the libm <tt>cos</tt> functions would, and handles error
4701 conditions in the same way.</p>
4704 <!-- _______________________________________________________________________ -->
4705 <div class="doc_subsubsection">
4706 <a name="int_pow">'<tt>llvm.pow.*</tt>' Intrinsic</a>
4709 <div class="doc_text">
4712 <p>This is an overloaded intrinsic. You can use <tt>llvm.pow</tt> on any
4713 floating point or vector of floating point type. Not all targets support all
4716 declare float @llvm.pow.f32(float %Val, float %Power)
4717 declare double @llvm.pow.f64(double %Val, double %Power)
4718 declare x86_fp80 @llvm.pow.f80(x86_fp80 %Val, x86_fp80 %Power)
4719 declare fp128 @llvm.pow.f128(fp128 %Val, fp128 %Power)
4720 declare ppc_fp128 @llvm.pow.ppcf128(ppc_fp128 %Val, ppc_fp128 Power)
4726 The '<tt>llvm.pow.*</tt>' intrinsics return the first operand raised to the
4727 specified (positive or negative) power.
4733 The second argument is a floating point power, and the first is a value to
4734 raise to that power.
4740 This function returns the first value raised to the second power,
4742 same values as the libm <tt>pow</tt> functions would, and handles error
4743 conditions in the same way.</p>
4747 <!-- ======================================================================= -->
4748 <div class="doc_subsection">
4749 <a name="int_manip">Bit Manipulation Intrinsics</a>
4752 <div class="doc_text">
4754 LLVM provides intrinsics for a few important bit manipulation operations.
4755 These allow efficient code generation for some algorithms.
4760 <!-- _______________________________________________________________________ -->
4761 <div class="doc_subsubsection">
4762 <a name="int_bswap">'<tt>llvm.bswap.*</tt>' Intrinsics</a>
4765 <div class="doc_text">
4768 <p>This is an overloaded intrinsic function. You can use bswap on any integer
4769 type that is an even number of bytes (i.e. BitWidth % 16 == 0).
4771 declare i16 @llvm.bswap.i16(i16 <id>)
4772 declare i32 @llvm.bswap.i32(i32 <id>)
4773 declare i64 @llvm.bswap.i64(i64 <id>)
4779 The '<tt>llvm.bswap</tt>' family of intrinsics is used to byte swap integer
4780 values with an even number of bytes (positive multiple of 16 bits). These are
4781 useful for performing operations on data that is not in the target's native
4788 The <tt>llvm.bswap.i16</tt> intrinsic returns an i16 value that has the high
4789 and low byte of the input i16 swapped. Similarly, the <tt>llvm.bswap.i32</tt>
4790 intrinsic returns an i32 value that has the four bytes of the input i32
4791 swapped, so that if the input bytes are numbered 0, 1, 2, 3 then the returned
4792 i32 will have its bytes in 3, 2, 1, 0 order. The <tt>llvm.bswap.i48</tt>,
4793 <tt>llvm.bswap.i64</tt> and other intrinsics extend this concept to
4794 additional even-byte lengths (6 bytes, 8 bytes and more, respectively).
4799 <!-- _______________________________________________________________________ -->
4800 <div class="doc_subsubsection">
4801 <a name="int_ctpop">'<tt>llvm.ctpop.*</tt>' Intrinsic</a>
4804 <div class="doc_text">
4807 <p>This is an overloaded intrinsic. You can use llvm.ctpop on any integer bit
4808 width. Not all targets support all bit widths however.
4810 declare i8 @llvm.ctpop.i8 (i8 <src>)
4811 declare i16 @llvm.ctpop.i16(i16 <src>)
4812 declare i32 @llvm.ctpop.i32(i32 <src>)
4813 declare i64 @llvm.ctpop.i64(i64 <src>)
4814 declare i256 @llvm.ctpop.i256(i256 <src>)
4820 The '<tt>llvm.ctpop</tt>' family of intrinsics counts the number of bits set in a
4827 The only argument is the value to be counted. The argument may be of any
4828 integer type. The return type must match the argument type.
4834 The '<tt>llvm.ctpop</tt>' intrinsic counts the 1's in a variable.
4838 <!-- _______________________________________________________________________ -->
4839 <div class="doc_subsubsection">
4840 <a name="int_ctlz">'<tt>llvm.ctlz.*</tt>' Intrinsic</a>
4843 <div class="doc_text">
4846 <p>This is an overloaded intrinsic. You can use <tt>llvm.ctlz</tt> on any
4847 integer bit width. Not all targets support all bit widths however.
4849 declare i8 @llvm.ctlz.i8 (i8 <src>)
4850 declare i16 @llvm.ctlz.i16(i16 <src>)
4851 declare i32 @llvm.ctlz.i32(i32 <src>)
4852 declare i64 @llvm.ctlz.i64(i64 <src>)
4853 declare i256 @llvm.ctlz.i256(i256 <src>)
4859 The '<tt>llvm.ctlz</tt>' family of intrinsic functions counts the number of
4860 leading zeros in a variable.
4866 The only argument is the value to be counted. The argument may be of any
4867 integer type. The return type must match the argument type.
4873 The '<tt>llvm.ctlz</tt>' intrinsic counts the leading (most significant) zeros
4874 in a variable. If the src == 0 then the result is the size in bits of the type
4875 of src. For example, <tt>llvm.ctlz(i32 2) = 30</tt>.
4881 <!-- _______________________________________________________________________ -->
4882 <div class="doc_subsubsection">
4883 <a name="int_cttz">'<tt>llvm.cttz.*</tt>' Intrinsic</a>
4886 <div class="doc_text">
4889 <p>This is an overloaded intrinsic. You can use <tt>llvm.cttz</tt> on any
4890 integer bit width. Not all targets support all bit widths however.
4892 declare i8 @llvm.cttz.i8 (i8 <src>)
4893 declare i16 @llvm.cttz.i16(i16 <src>)
4894 declare i32 @llvm.cttz.i32(i32 <src>)
4895 declare i64 @llvm.cttz.i64(i64 <src>)
4896 declare i256 @llvm.cttz.i256(i256 <src>)
4902 The '<tt>llvm.cttz</tt>' family of intrinsic functions counts the number of
4909 The only argument is the value to be counted. The argument may be of any
4910 integer type. The return type must match the argument type.
4916 The '<tt>llvm.cttz</tt>' intrinsic counts the trailing (least significant) zeros
4917 in a variable. If the src == 0 then the result is the size in bits of the type
4918 of src. For example, <tt>llvm.cttz(2) = 1</tt>.
4922 <!-- _______________________________________________________________________ -->
4923 <div class="doc_subsubsection">
4924 <a name="int_part_select">'<tt>llvm.part.select.*</tt>' Intrinsic</a>
4927 <div class="doc_text">
4930 <p>This is an overloaded intrinsic. You can use <tt>llvm.part.select</tt>
4931 on any integer bit width.
4933 declare i17 @llvm.part.select.i17 (i17 %val, i32 %loBit, i32 %hiBit)
4934 declare i29 @llvm.part.select.i29 (i29 %val, i32 %loBit, i32 %hiBit)
4938 <p>The '<tt>llvm.part.select</tt>' family of intrinsic functions selects a
4939 range of bits from an integer value and returns them in the same bit width as
4940 the original value.</p>
4943 <p>The first argument, <tt>%val</tt> and the result may be integer types of
4944 any bit width but they must have the same bit width. The second and third
4945 arguments must be <tt>i32</tt> type since they specify only a bit index.</p>
4948 <p>The operation of the '<tt>llvm.part.select</tt>' intrinsic has two modes
4949 of operation: forwards and reverse. If <tt>%loBit</tt> is greater than
4950 <tt>%hiBits</tt> then the intrinsic operates in reverse mode. Otherwise it
4951 operates in forward mode.</p>
4952 <p>In forward mode, this intrinsic is the equivalent of shifting <tt>%val</tt>
4953 right by <tt>%loBit</tt> bits and then ANDing it with a mask with
4954 only the <tt>%hiBit - %loBit</tt> bits set, as follows:</p>
4956 <li>The <tt>%val</tt> is shifted right (LSHR) by the number of bits specified
4957 by <tt>%loBits</tt>. This normalizes the value to the low order bits.</li>
4958 <li>The <tt>%loBits</tt> value is subtracted from the <tt>%hiBits</tt> value
4959 to determine the number of bits to retain.</li>
4960 <li>A mask of the retained bits is created by shifting a -1 value.</li>
4961 <li>The mask is ANDed with <tt>%val</tt> to produce the result.
4963 <p>In reverse mode, a similar computation is made except that the bits are
4964 returned in the reverse order. So, for example, if <tt>X</tt> has the value
4965 <tt>i16 0x0ACF (101011001111)</tt> and we apply
4966 <tt>part.select(i16 X, 8, 3)</tt> to it, we get back the value
4967 <tt>i16 0x0026 (000000100110)</tt>.</p>
4970 <div class="doc_subsubsection">
4971 <a name="int_part_set">'<tt>llvm.part.set.*</tt>' Intrinsic</a>
4974 <div class="doc_text">
4977 <p>This is an overloaded intrinsic. You can use <tt>llvm.part.set</tt>
4978 on any integer bit width.
4980 declare i17 @llvm.part.set.i17.i9 (i17 %val, i9 %repl, i32 %lo, i32 %hi)
4981 declare i29 @llvm.part.set.i29.i9 (i29 %val, i9 %repl, i32 %lo, i32 %hi)
4985 <p>The '<tt>llvm.part.set</tt>' family of intrinsic functions replaces a range
4986 of bits in an integer value with another integer value. It returns the integer
4987 with the replaced bits.</p>
4990 <p>The first argument, <tt>%val</tt> and the result may be integer types of
4991 any bit width but they must have the same bit width. <tt>%val</tt> is the value
4992 whose bits will be replaced. The second argument, <tt>%repl</tt> may be an
4993 integer of any bit width. The third and fourth arguments must be <tt>i32</tt>
4994 type since they specify only a bit index.</p>
4997 <p>The operation of the '<tt>llvm.part.set</tt>' intrinsic has two modes
4998 of operation: forwards and reverse. If <tt>%lo</tt> is greater than
4999 <tt>%hi</tt> then the intrinsic operates in reverse mode. Otherwise it
5000 operates in forward mode.</p>
5001 <p>For both modes, the <tt>%repl</tt> value is prepared for use by either
5002 truncating it down to the size of the replacement area or zero extending it
5003 up to that size.</p>
5004 <p>In forward mode, the bits between <tt>%lo</tt> and <tt>%hi</tt> (inclusive)
5005 are replaced with corresponding bits from <tt>%repl</tt>. That is the 0th bit
5006 in <tt>%repl</tt> replaces the <tt>%lo</tt>th bit in <tt>%val</tt> and etc. up
5007 to the <tt>%hi</tt>th bit.
5008 <p>In reverse mode, a similar computation is made except that the bits are
5009 reversed. That is, the <tt>0</tt>th bit in <tt>%repl</tt> replaces the
5010 <tt>%hi</tt> bit in <tt>%val</tt> and etc. down to the <tt>%lo</tt>th bit.
5013 llvm.part.set(0xFFFF, 0, 4, 7) -> 0xFF0F
5014 llvm.part.set(0xFFFF, 0, 7, 4) -> 0xFF0F
5015 llvm.part.set(0xFFFF, 1, 7, 4) -> 0xFF8F
5016 llvm.part.set(0xFFFF, F, 8, 3) -> 0xFFE7
5017 llvm.part.set(0xFFFF, 0, 3, 8) -> 0xFE07
5021 <!-- ======================================================================= -->
5022 <div class="doc_subsection">
5023 <a name="int_debugger">Debugger Intrinsics</a>
5026 <div class="doc_text">
5028 The LLVM debugger intrinsics (which all start with <tt>llvm.dbg.</tt> prefix),
5029 are described in the <a
5030 href="SourceLevelDebugging.html#format_common_intrinsics">LLVM Source Level
5031 Debugging</a> document.
5036 <!-- ======================================================================= -->
5037 <div class="doc_subsection">
5038 <a name="int_eh">Exception Handling Intrinsics</a>
5041 <div class="doc_text">
5042 <p> The LLVM exception handling intrinsics (which all start with
5043 <tt>llvm.eh.</tt> prefix), are described in the <a
5044 href="ExceptionHandling.html#format_common_intrinsics">LLVM Exception
5045 Handling</a> document. </p>
5048 <!-- ======================================================================= -->
5049 <div class="doc_subsection">
5050 <a name="int_trampoline">Trampoline Intrinsic</a>
5053 <div class="doc_text">
5055 This intrinsic makes it possible to excise one parameter, marked with
5056 the <tt>nest</tt> attribute, from a function. The result is a callable
5057 function pointer lacking the nest parameter - the caller does not need
5058 to provide a value for it. Instead, the value to use is stored in
5059 advance in a "trampoline", a block of memory usually allocated
5060 on the stack, which also contains code to splice the nest value into the
5061 argument list. This is used to implement the GCC nested function address
5065 For example, if the function is
5066 <tt>i32 f(i8* nest %c, i32 %x, i32 %y)</tt> then the resulting function
5067 pointer has signature <tt>i32 (i32, i32)*</tt>. It can be created as follows:</p>
5069 %tramp = alloca [10 x i8], align 4 ; size and alignment only correct for X86
5070 %tramp1 = getelementptr [10 x i8]* %tramp, i32 0, i32 0
5071 %p = call i8* @llvm.init.trampoline( i8* %tramp1, i8* bitcast (i32 (i8* nest , i32, i32)* @f to i8*), i8* %nval )
5072 %fp = bitcast i8* %p to i32 (i32, i32)*
5074 <p>The call <tt>%val = call i32 %fp( i32 %x, i32 %y )</tt> is then equivalent
5075 to <tt>%val = call i32 %f( i8* %nval, i32 %x, i32 %y )</tt>.</p>
5078 <!-- _______________________________________________________________________ -->
5079 <div class="doc_subsubsection">
5080 <a name="int_it">'<tt>llvm.init.trampoline</tt>' Intrinsic</a>
5082 <div class="doc_text">
5085 declare i8* @llvm.init.trampoline(i8* <tramp>, i8* <func>, i8* <nval>)
5089 This fills the memory pointed to by <tt>tramp</tt> with code
5090 and returns a function pointer suitable for executing it.
5094 The <tt>llvm.init.trampoline</tt> intrinsic takes three arguments, all
5095 pointers. The <tt>tramp</tt> argument must point to a sufficiently large
5096 and sufficiently aligned block of memory; this memory is written to by the
5097 intrinsic. Note that the size and the alignment are target-specific - LLVM
5098 currently provides no portable way of determining them, so a front-end that
5099 generates this intrinsic needs to have some target-specific knowledge.
5100 The <tt>func</tt> argument must hold a function bitcast to an <tt>i8*</tt>.
5104 The block of memory pointed to by <tt>tramp</tt> is filled with target
5105 dependent code, turning it into a function. A pointer to this function is
5106 returned, but needs to be bitcast to an
5107 <a href="#int_trampoline">appropriate function pointer type</a>
5108 before being called. The new function's signature is the same as that of
5109 <tt>func</tt> with any arguments marked with the <tt>nest</tt> attribute
5110 removed. At most one such <tt>nest</tt> argument is allowed, and it must be
5111 of pointer type. Calling the new function is equivalent to calling
5112 <tt>func</tt> with the same argument list, but with <tt>nval</tt> used for the
5113 missing <tt>nest</tt> argument. If, after calling
5114 <tt>llvm.init.trampoline</tt>, the memory pointed to by <tt>tramp</tt> is
5115 modified, then the effect of any later call to the returned function pointer is
5120 <!-- ======================================================================= -->
5121 <div class="doc_subsection">
5122 <a name="int_general">General Intrinsics</a>
5125 <div class="doc_text">
5126 <p> This class of intrinsics is designed to be generic and has
5127 no specific purpose. </p>
5130 <!-- _______________________________________________________________________ -->
5131 <div class="doc_subsubsection">
5132 <a name="int_var_annotation">'<tt>llvm.var.annotation</tt>' Intrinsic</a>
5135 <div class="doc_text">
5139 declare void @llvm.var.annotation(i8* <val>, i8* <str>, i8* <str>, i32 <int> )
5145 The '<tt>llvm.var.annotation</tt>' intrinsic
5151 The first argument is a pointer to a value, the second is a pointer to a
5152 global string, the third is a pointer to a global string which is the source
5153 file name, and the last argument is the line number.
5159 This intrinsic allows annotation of local variables with arbitrary strings.
5160 This can be useful for special purpose optimizations that want to look for these
5161 annotations. These have no other defined use, they are ignored by code
5162 generation and optimization.
5165 <!-- _______________________________________________________________________ -->
5166 <div class="doc_subsubsection">
5167 <a name="int_annotation">'<tt>llvm.annotation.*</tt>' Intrinsic</a>
5170 <div class="doc_text">
5173 <p>This is an overloaded intrinsic. You can use '<tt>llvm.annotation</tt>' on
5174 any integer bit width.
5177 declare i8 @llvm.annotation.i8(i8 <val>, i8* <str>, i8* <str>, i32 <int> )
5178 declare i16 @llvm.annotation.i16(i16 <val>, i8* <str>, i8* <str>, i32 <int> )
5179 declare i32 @llvm.annotation.i32(i32 <val>, i8* <str>, i8* <str>, i32 <int> )
5180 declare i64 @llvm.annotation.i64(i64 <val>, i8* <str>, i8* <str>, i32 <int> )
5181 declare i256 @llvm.annotation.i256(i256 <val>, i8* <str>, i8* <str>, i32 <int> )
5187 The '<tt>llvm.annotation</tt>' intrinsic.
5193 The first argument is an integer value (result of some expression),
5194 the second is a pointer to a global string, the third is a pointer to a global
5195 string which is the source file name, and the last argument is the line number.
5196 It returns the value of the first argument.
5202 This intrinsic allows annotations to be put on arbitrary expressions
5203 with arbitrary strings. This can be useful for special purpose optimizations
5204 that want to look for these annotations. These have no other defined use, they
5205 are ignored by code generation and optimization.
5208 <!-- *********************************************************************** -->
5211 <a href="http://jigsaw.w3.org/css-validator/check/referer"><img
5212 src="http://jigsaw.w3.org/css-validator/images/vcss" alt="Valid CSS!"></a>
5213 <a href="http://validator.w3.org/check/referer"><img
5214 src="http://www.w3.org/Icons/valid-html401" alt="Valid HTML 4.01!" /></a>
5216 <a href="mailto:sabre@nondot.org">Chris Lattner</a><br>
5217 <a href="http://llvm.org">The LLVM Compiler Infrastructure</a><br>
5218 Last modified: $Date$