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="#namedtypes">Named Types</a></li>
26 <li><a href="#globalvars">Global Variables</a></li>
27 <li><a href="#functionstructure">Functions</a></li>
28 <li><a href="#aliasstructure">Aliases</a></li>
29 <li><a href="#paramattrs">Parameter Attributes</a></li>
30 <li><a href="#fnattrs">Function Attributes</a></li>
31 <li><a href="#gc">Garbage Collector Names</a></li>
32 <li><a href="#moduleasm">Module-Level Inline Assembly</a></li>
33 <li><a href="#datalayout">Data Layout</a></li>
36 <li><a href="#typesystem">Type System</a>
38 <li><a href="#t_classifications">Type Classifications</a></li>
39 <li><a href="#t_primitive">Primitive Types</a>
41 <li><a href="#t_floating">Floating Point Types</a></li>
42 <li><a href="#t_void">Void Type</a></li>
43 <li><a href="#t_label">Label Type</a></li>
46 <li><a href="#t_derived">Derived Types</a>
48 <li><a href="#t_integer">Integer Type</a></li>
49 <li><a href="#t_array">Array Type</a></li>
50 <li><a href="#t_function">Function Type</a></li>
51 <li><a href="#t_pointer">Pointer Type</a></li>
52 <li><a href="#t_struct">Structure Type</a></li>
53 <li><a href="#t_pstruct">Packed Structure Type</a></li>
54 <li><a href="#t_vector">Vector Type</a></li>
55 <li><a href="#t_opaque">Opaque Type</a></li>
58 <li><a href="#t_uprefs">Type Up-references</a></li>
61 <li><a href="#constants">Constants</a>
63 <li><a href="#simpleconstants">Simple Constants</a></li>
64 <li><a href="#complexconstants">Complex Constants</a></li>
65 <li><a href="#globalconstants">Global Variable and Function Addresses</a></li>
66 <li><a href="#undefvalues">Undefined Values</a></li>
67 <li><a href="#constantexprs">Constant Expressions</a></li>
70 <li><a href="#othervalues">Other Values</a>
72 <li><a href="#inlineasm">Inline Assembler Expressions</a></li>
75 <li><a href="#instref">Instruction Reference</a>
77 <li><a href="#terminators">Terminator Instructions</a>
79 <li><a href="#i_ret">'<tt>ret</tt>' Instruction</a></li>
80 <li><a href="#i_br">'<tt>br</tt>' Instruction</a></li>
81 <li><a href="#i_switch">'<tt>switch</tt>' Instruction</a></li>
82 <li><a href="#i_invoke">'<tt>invoke</tt>' Instruction</a></li>
83 <li><a href="#i_unwind">'<tt>unwind</tt>' Instruction</a></li>
84 <li><a href="#i_unreachable">'<tt>unreachable</tt>' Instruction</a></li>
87 <li><a href="#binaryops">Binary Operations</a>
89 <li><a href="#i_add">'<tt>add</tt>' Instruction</a></li>
90 <li><a href="#i_sub">'<tt>sub</tt>' Instruction</a></li>
91 <li><a href="#i_mul">'<tt>mul</tt>' Instruction</a></li>
92 <li><a href="#i_udiv">'<tt>udiv</tt>' Instruction</a></li>
93 <li><a href="#i_sdiv">'<tt>sdiv</tt>' Instruction</a></li>
94 <li><a href="#i_fdiv">'<tt>fdiv</tt>' Instruction</a></li>
95 <li><a href="#i_urem">'<tt>urem</tt>' Instruction</a></li>
96 <li><a href="#i_srem">'<tt>srem</tt>' Instruction</a></li>
97 <li><a href="#i_frem">'<tt>frem</tt>' Instruction</a></li>
100 <li><a href="#bitwiseops">Bitwise Binary Operations</a>
102 <li><a href="#i_shl">'<tt>shl</tt>' Instruction</a></li>
103 <li><a href="#i_lshr">'<tt>lshr</tt>' Instruction</a></li>
104 <li><a href="#i_ashr">'<tt>ashr</tt>' Instruction</a></li>
105 <li><a href="#i_and">'<tt>and</tt>' Instruction</a></li>
106 <li><a href="#i_or">'<tt>or</tt>' Instruction</a></li>
107 <li><a href="#i_xor">'<tt>xor</tt>' Instruction</a></li>
110 <li><a href="#vectorops">Vector Operations</a>
112 <li><a href="#i_extractelement">'<tt>extractelement</tt>' Instruction</a></li>
113 <li><a href="#i_insertelement">'<tt>insertelement</tt>' Instruction</a></li>
114 <li><a href="#i_shufflevector">'<tt>shufflevector</tt>' Instruction</a></li>
117 <li><a href="#aggregateops">Aggregate Operations</a>
119 <li><a href="#i_extractvalue">'<tt>extractvalue</tt>' Instruction</a></li>
120 <li><a href="#i_insertvalue">'<tt>insertvalue</tt>' Instruction</a></li>
123 <li><a href="#memoryops">Memory Access and Addressing Operations</a>
125 <li><a href="#i_malloc">'<tt>malloc</tt>' Instruction</a></li>
126 <li><a href="#i_free">'<tt>free</tt>' Instruction</a></li>
127 <li><a href="#i_alloca">'<tt>alloca</tt>' Instruction</a></li>
128 <li><a href="#i_load">'<tt>load</tt>' Instruction</a></li>
129 <li><a href="#i_store">'<tt>store</tt>' Instruction</a></li>
130 <li><a href="#i_getelementptr">'<tt>getelementptr</tt>' Instruction</a></li>
133 <li><a href="#convertops">Conversion Operations</a>
135 <li><a href="#i_trunc">'<tt>trunc .. to</tt>' Instruction</a></li>
136 <li><a href="#i_zext">'<tt>zext .. to</tt>' Instruction</a></li>
137 <li><a href="#i_sext">'<tt>sext .. to</tt>' Instruction</a></li>
138 <li><a href="#i_fptrunc">'<tt>fptrunc .. to</tt>' Instruction</a></li>
139 <li><a href="#i_fpext">'<tt>fpext .. to</tt>' Instruction</a></li>
140 <li><a href="#i_fptoui">'<tt>fptoui .. to</tt>' Instruction</a></li>
141 <li><a href="#i_fptosi">'<tt>fptosi .. to</tt>' Instruction</a></li>
142 <li><a href="#i_uitofp">'<tt>uitofp .. to</tt>' Instruction</a></li>
143 <li><a href="#i_sitofp">'<tt>sitofp .. to</tt>' Instruction</a></li>
144 <li><a href="#i_ptrtoint">'<tt>ptrtoint .. to</tt>' Instruction</a></li>
145 <li><a href="#i_inttoptr">'<tt>inttoptr .. to</tt>' Instruction</a></li>
146 <li><a href="#i_bitcast">'<tt>bitcast .. to</tt>' Instruction</a></li>
149 <li><a href="#otherops">Other Operations</a>
151 <li><a href="#i_icmp">'<tt>icmp</tt>' Instruction</a></li>
152 <li><a href="#i_fcmp">'<tt>fcmp</tt>' Instruction</a></li>
153 <li><a href="#i_vicmp">'<tt>vicmp</tt>' Instruction</a></li>
154 <li><a href="#i_vfcmp">'<tt>vfcmp</tt>' Instruction</a></li>
155 <li><a href="#i_phi">'<tt>phi</tt>' Instruction</a></li>
156 <li><a href="#i_select">'<tt>select</tt>' Instruction</a></li>
157 <li><a href="#i_call">'<tt>call</tt>' Instruction</a></li>
158 <li><a href="#i_va_arg">'<tt>va_arg</tt>' Instruction</a></li>
163 <li><a href="#intrinsics">Intrinsic Functions</a>
165 <li><a href="#int_varargs">Variable Argument Handling Intrinsics</a>
167 <li><a href="#int_va_start">'<tt>llvm.va_start</tt>' Intrinsic</a></li>
168 <li><a href="#int_va_end">'<tt>llvm.va_end</tt>' Intrinsic</a></li>
169 <li><a href="#int_va_copy">'<tt>llvm.va_copy</tt>' Intrinsic</a></li>
172 <li><a href="#int_gc">Accurate Garbage Collection Intrinsics</a>
174 <li><a href="#int_gcroot">'<tt>llvm.gcroot</tt>' Intrinsic</a></li>
175 <li><a href="#int_gcread">'<tt>llvm.gcread</tt>' Intrinsic</a></li>
176 <li><a href="#int_gcwrite">'<tt>llvm.gcwrite</tt>' Intrinsic</a></li>
179 <li><a href="#int_codegen">Code Generator Intrinsics</a>
181 <li><a href="#int_returnaddress">'<tt>llvm.returnaddress</tt>' Intrinsic</a></li>
182 <li><a href="#int_frameaddress">'<tt>llvm.frameaddress</tt>' Intrinsic</a></li>
183 <li><a href="#int_stacksave">'<tt>llvm.stacksave</tt>' Intrinsic</a></li>
184 <li><a href="#int_stackrestore">'<tt>llvm.stackrestore</tt>' Intrinsic</a></li>
185 <li><a href="#int_prefetch">'<tt>llvm.prefetch</tt>' Intrinsic</a></li>
186 <li><a href="#int_pcmarker">'<tt>llvm.pcmarker</tt>' Intrinsic</a></li>
187 <li><a href="#int_readcyclecounter"><tt>llvm.readcyclecounter</tt>' Intrinsic</a></li>
190 <li><a href="#int_libc">Standard C Library Intrinsics</a>
192 <li><a href="#int_memcpy">'<tt>llvm.memcpy.*</tt>' Intrinsic</a></li>
193 <li><a href="#int_memmove">'<tt>llvm.memmove.*</tt>' Intrinsic</a></li>
194 <li><a href="#int_memset">'<tt>llvm.memset.*</tt>' Intrinsic</a></li>
195 <li><a href="#int_sqrt">'<tt>llvm.sqrt.*</tt>' Intrinsic</a></li>
196 <li><a href="#int_powi">'<tt>llvm.powi.*</tt>' Intrinsic</a></li>
197 <li><a href="#int_sin">'<tt>llvm.sin.*</tt>' Intrinsic</a></li>
198 <li><a href="#int_cos">'<tt>llvm.cos.*</tt>' Intrinsic</a></li>
199 <li><a href="#int_pow">'<tt>llvm.pow.*</tt>' Intrinsic</a></li>
202 <li><a href="#int_manip">Bit Manipulation Intrinsics</a>
204 <li><a href="#int_bswap">'<tt>llvm.bswap.*</tt>' Intrinsics</a></li>
205 <li><a href="#int_ctpop">'<tt>llvm.ctpop.*</tt>' Intrinsic </a></li>
206 <li><a href="#int_ctlz">'<tt>llvm.ctlz.*</tt>' Intrinsic </a></li>
207 <li><a href="#int_cttz">'<tt>llvm.cttz.*</tt>' Intrinsic </a></li>
208 <li><a href="#int_part_select">'<tt>llvm.part.select.*</tt>' Intrinsic </a></li>
209 <li><a href="#int_part_set">'<tt>llvm.part.set.*</tt>' Intrinsic </a></li>
212 <li><a href="#int_overflow">Arithmetic with Overflow Intrinsics</a>
214 <li><a href="#int_sadd_overflow">'<tt>llvm.sadd.with.overflow.*</tt> Intrinsics</a></li>
215 <li><a href="#int_uadd_overflow">'<tt>llvm.uadd.with.overflow.*</tt> Intrinsics</a></li>
216 <li><a href="#int_ssub_overflow">'<tt>llvm.ssub.with.overflow.*</tt> Intrinsics</a></li>
217 <li><a href="#int_usub_overflow">'<tt>llvm.usub.with.overflow.*</tt> Intrinsics</a></li>
218 <li><a href="#int_smul_overflow">'<tt>llvm.smul.with.overflow.*</tt> Intrinsics</a></li>
219 <li><a href="#int_umul_overflow">'<tt>llvm.umul.with.overflow.*</tt> Intrinsics</a></li>
222 <li><a href="#int_debugger">Debugger intrinsics</a></li>
223 <li><a href="#int_eh">Exception Handling intrinsics</a></li>
224 <li><a href="#int_trampoline">Trampoline Intrinsic</a>
226 <li><a href="#int_it">'<tt>llvm.init.trampoline</tt>' Intrinsic</a></li>
229 <li><a href="#int_atomics">Atomic intrinsics</a>
231 <li><a href="#int_memory_barrier"><tt>llvm.memory_barrier</tt></a></li>
232 <li><a href="#int_atomic_cmp_swap"><tt>llvm.atomic.cmp.swap</tt></a></li>
233 <li><a href="#int_atomic_swap"><tt>llvm.atomic.swap</tt></a></li>
234 <li><a href="#int_atomic_load_add"><tt>llvm.atomic.load.add</tt></a></li>
235 <li><a href="#int_atomic_load_sub"><tt>llvm.atomic.load.sub</tt></a></li>
236 <li><a href="#int_atomic_load_and"><tt>llvm.atomic.load.and</tt></a></li>
237 <li><a href="#int_atomic_load_nand"><tt>llvm.atomic.load.nand</tt></a></li>
238 <li><a href="#int_atomic_load_or"><tt>llvm.atomic.load.or</tt></a></li>
239 <li><a href="#int_atomic_load_xor"><tt>llvm.atomic.load.xor</tt></a></li>
240 <li><a href="#int_atomic_load_max"><tt>llvm.atomic.load.max</tt></a></li>
241 <li><a href="#int_atomic_load_min"><tt>llvm.atomic.load.min</tt></a></li>
242 <li><a href="#int_atomic_load_umax"><tt>llvm.atomic.load.umax</tt></a></li>
243 <li><a href="#int_atomic_load_umin"><tt>llvm.atomic.load.umin</tt></a></li>
246 <li><a href="#int_general">General intrinsics</a>
248 <li><a href="#int_var_annotation">
249 '<tt>llvm.var.annotation</tt>' Intrinsic</a></li>
250 <li><a href="#int_annotation">
251 '<tt>llvm.annotation.*</tt>' Intrinsic</a></li>
252 <li><a href="#int_trap">
253 '<tt>llvm.trap</tt>' Intrinsic</a></li>
254 <li><a href="#int_stackprotector">
255 '<tt>llvm.stackprotector</tt>' Intrinsic</a></li>
262 <div class="doc_author">
263 <p>Written by <a href="mailto:sabre@nondot.org">Chris Lattner</a>
264 and <a href="mailto:vadve@cs.uiuc.edu">Vikram Adve</a></p>
267 <!-- *********************************************************************** -->
268 <div class="doc_section"> <a name="abstract">Abstract </a></div>
269 <!-- *********************************************************************** -->
271 <div class="doc_text">
272 <p>This document is a reference manual for the LLVM assembly language.
273 LLVM is a Static Single Assignment (SSA) based representation that provides
274 type safety, low-level operations, flexibility, and the capability of
275 representing 'all' high-level languages cleanly. It is the common code
276 representation used throughout all phases of the LLVM compilation
280 <!-- *********************************************************************** -->
281 <div class="doc_section"> <a name="introduction">Introduction</a> </div>
282 <!-- *********************************************************************** -->
284 <div class="doc_text">
286 <p>The LLVM code representation is designed to be used in three
287 different forms: as an in-memory compiler IR, as an on-disk bitcode
288 representation (suitable for fast loading by a Just-In-Time compiler),
289 and as a human readable assembly language representation. This allows
290 LLVM to provide a powerful intermediate representation for efficient
291 compiler transformations and analysis, while providing a natural means
292 to debug and visualize the transformations. The three different forms
293 of LLVM are all equivalent. This document describes the human readable
294 representation and notation.</p>
296 <p>The LLVM representation aims to be light-weight and low-level
297 while being expressive, typed, and extensible at the same time. It
298 aims to be a "universal IR" of sorts, by being at a low enough level
299 that high-level ideas may be cleanly mapped to it (similar to how
300 microprocessors are "universal IR's", allowing many source languages to
301 be mapped to them). By providing type information, LLVM can be used as
302 the target of optimizations: for example, through pointer analysis, it
303 can be proven that a C automatic variable is never accessed outside of
304 the current function... allowing it to be promoted to a simple SSA
305 value instead of a memory location.</p>
309 <!-- _______________________________________________________________________ -->
310 <div class="doc_subsubsection"> <a name="wellformed">Well-Formedness</a> </div>
312 <div class="doc_text">
314 <p>It is important to note that this document describes 'well formed'
315 LLVM assembly language. There is a difference between what the parser
316 accepts and what is considered 'well formed'. For example, the
317 following instruction is syntactically okay, but not well formed:</p>
319 <div class="doc_code">
321 %x = <a href="#i_add">add</a> i32 1, %x
325 <p>...because the definition of <tt>%x</tt> does not dominate all of
326 its uses. The LLVM infrastructure provides a verification pass that may
327 be used to verify that an LLVM module is well formed. This pass is
328 automatically run by the parser after parsing input assembly and by
329 the optimizer before it outputs bitcode. The violations pointed out
330 by the verifier pass indicate bugs in transformation passes or input to
334 <!-- Describe the typesetting conventions here. -->
336 <!-- *********************************************************************** -->
337 <div class="doc_section"> <a name="identifiers">Identifiers</a> </div>
338 <!-- *********************************************************************** -->
340 <div class="doc_text">
342 <p>LLVM identifiers come in two basic types: global and local. Global
343 identifiers (functions, global variables) begin with the @ character. Local
344 identifiers (register names, types) begin with the % character. Additionally,
345 there are three different formats for identifiers, for different purposes:</p>
348 <li>Named values are represented as a string of characters with their prefix.
349 For example, %foo, @DivisionByZero, %a.really.long.identifier. The actual
350 regular expression used is '<tt>[%@][a-zA-Z$._][a-zA-Z$._0-9]*</tt>'.
351 Identifiers which require other characters in their names can be surrounded
352 with quotes. Special characters may be escaped using "\xx" where xx is the
353 ASCII code for the character in hexadecimal. In this way, any character can
354 be used in a name value, even quotes themselves.
356 <li>Unnamed values are represented as an unsigned numeric value with their
357 prefix. For example, %12, @2, %44.</li>
359 <li>Constants, which are described in a <a href="#constants">section about
360 constants</a>, below.</li>
363 <p>LLVM requires that values start with a prefix for two reasons: Compilers
364 don't need to worry about name clashes with reserved words, and the set of
365 reserved words may be expanded in the future without penalty. Additionally,
366 unnamed identifiers allow a compiler to quickly come up with a temporary
367 variable without having to avoid symbol table conflicts.</p>
369 <p>Reserved words in LLVM are very similar to reserved words in other
370 languages. There are keywords for different opcodes
371 ('<tt><a href="#i_add">add</a></tt>',
372 '<tt><a href="#i_bitcast">bitcast</a></tt>',
373 '<tt><a href="#i_ret">ret</a></tt>', etc...), for primitive type names ('<tt><a
374 href="#t_void">void</a></tt>', '<tt><a href="#t_primitive">i32</a></tt>', etc...),
375 and others. These reserved words cannot conflict with variable names, because
376 none of them start with a prefix character ('%' or '@').</p>
378 <p>Here is an example of LLVM code to multiply the integer variable
379 '<tt>%X</tt>' by 8:</p>
383 <div class="doc_code">
385 %result = <a href="#i_mul">mul</a> i32 %X, 8
389 <p>After strength reduction:</p>
391 <div class="doc_code">
393 %result = <a href="#i_shl">shl</a> i32 %X, i8 3
397 <p>And the hard way:</p>
399 <div class="doc_code">
401 <a href="#i_add">add</a> i32 %X, %X <i>; yields {i32}:%0</i>
402 <a href="#i_add">add</a> i32 %0, %0 <i>; yields {i32}:%1</i>
403 %result = <a href="#i_add">add</a> i32 %1, %1
407 <p>This last way of multiplying <tt>%X</tt> by 8 illustrates several
408 important lexical features of LLVM:</p>
412 <li>Comments are delimited with a '<tt>;</tt>' and go until the end of
415 <li>Unnamed temporaries are created when the result of a computation is not
416 assigned to a named value.</li>
418 <li>Unnamed temporaries are numbered sequentially</li>
422 <p>...and it also shows a convention that we follow in this document. When
423 demonstrating instructions, we will follow an instruction with a comment that
424 defines the type and name of value produced. Comments are shown in italic
429 <!-- *********************************************************************** -->
430 <div class="doc_section"> <a name="highlevel">High Level Structure</a> </div>
431 <!-- *********************************************************************** -->
433 <!-- ======================================================================= -->
434 <div class="doc_subsection"> <a name="modulestructure">Module Structure</a>
437 <div class="doc_text">
439 <p>LLVM programs are composed of "Module"s, each of which is a
440 translation unit of the input programs. Each module consists of
441 functions, global variables, and symbol table entries. Modules may be
442 combined together with the LLVM linker, which merges function (and
443 global variable) definitions, resolves forward declarations, and merges
444 symbol table entries. Here is an example of the "hello world" module:</p>
446 <div class="doc_code">
447 <pre><i>; Declare the string constant as a global constant...</i>
448 <a href="#identifiers">@.LC0</a> = <a href="#linkage_internal">internal</a> <a
449 href="#globalvars">constant</a> <a href="#t_array">[13 x i8]</a> c"hello world\0A\00" <i>; [13 x i8]*</i>
451 <i>; External declaration of the puts function</i>
452 <a href="#functionstructure">declare</a> i32 @puts(i8 *) <i>; i32(i8 *)* </i>
454 <i>; Definition of main function</i>
455 define i32 @main() { <i>; i32()* </i>
456 <i>; Convert [13 x i8]* to i8 *...</i>
458 href="#i_getelementptr">getelementptr</a> [13 x i8]* @.LC0, i64 0, i64 0 <i>; i8 *</i>
460 <i>; Call puts function to write out the string to stdout...</i>
462 href="#i_call">call</a> i32 @puts(i8 * %cast210) <i>; i32</i>
464 href="#i_ret">ret</a> i32 0<br>}<br>
468 <p>This example is made up of a <a href="#globalvars">global variable</a>
469 named "<tt>.LC0</tt>", an external declaration of the "<tt>puts</tt>"
470 function, and a <a href="#functionstructure">function definition</a>
471 for "<tt>main</tt>".</p>
473 <p>In general, a module is made up of a list of global values,
474 where both functions and global variables are global values. Global values are
475 represented by a pointer to a memory location (in this case, a pointer to an
476 array of char, and a pointer to a function), and have one of the following <a
477 href="#linkage">linkage types</a>.</p>
481 <!-- ======================================================================= -->
482 <div class="doc_subsection">
483 <a name="linkage">Linkage Types</a>
486 <div class="doc_text">
489 All Global Variables and Functions have one of the following types of linkage:
494 <dt><tt><b><a name="linkage_private">private</a></b></tt>: </dt>
496 <dd>Global values with private linkage are only directly accessible by
497 objects in the current module. In particular, linking code into a module with
498 an private global value may cause the private to be renamed as necessary to
499 avoid collisions. Because the symbol is private to the module, all
500 references can be updated. This doesn't show up in any symbol table in the
504 <dt><tt><b><a name="linkage_internal">internal</a></b></tt>: </dt>
506 <dd> Similar to private, but the value shows as a local symbol (STB_LOCAL in
507 the case of ELF) in the object file. This corresponds to the notion of the
508 '<tt>static</tt>' keyword in C.
511 <dt><tt><b><a name="linkage_linkonce">linkonce</a></b></tt>: </dt>
513 <dd>Globals with "<tt>linkonce</tt>" linkage are merged with other globals of
514 the same name when linkage occurs. This is typically used to implement
515 inline functions, templates, or other code which must be generated in each
516 translation unit that uses it. Unreferenced <tt>linkonce</tt> globals are
517 allowed to be discarded.
520 <dt><tt><b><a name="linkage_common">common</a></b></tt>: </dt>
522 <dd>"<tt>common</tt>" linkage is exactly the same as <tt>linkonce</tt>
523 linkage, except that unreferenced <tt>common</tt> globals may not be
524 discarded. This is used for globals that may be emitted in multiple
525 translation units, but that are not guaranteed to be emitted into every
526 translation unit that uses them. One example of this is tentative
527 definitions in C, such as "<tt>int X;</tt>" at global scope.
530 <dt><tt><b><a name="linkage_weak">weak</a></b></tt>: </dt>
532 <dd>"<tt>weak</tt>" linkage is the same as <tt>common</tt> linkage, except
533 that some targets may choose to emit different assembly sequences for them
534 for target-dependent reasons. This is used for globals that are declared
535 "weak" in C source code.
538 <dt><tt><b><a name="linkage_appending">appending</a></b></tt>: </dt>
540 <dd>"<tt>appending</tt>" linkage may only be applied to global variables of
541 pointer to array type. When two global variables with appending linkage are
542 linked together, the two global arrays are appended together. This is the
543 LLVM, typesafe, equivalent of having the system linker append together
544 "sections" with identical names when .o files are linked.
547 <dt><tt><b><a name="linkage_externweak">extern_weak</a></b></tt>: </dt>
549 <dd>The semantics of this linkage follow the ELF object file model: the
550 symbol is weak until linked, if not linked, the symbol becomes null instead
551 of being an undefined reference.
554 <dt><tt><b><a name="linkage_linkonce">linkonce_odr</a></b></tt>: </dt>
555 <dt><tt><b><a name="linkage_weak">weak_odr</a></b></tt>: </dt>
556 <dd>Some languages allow inequivalent globals to be merged, such as two
557 functions with different semantics. Other languages, such as <tt>C++</tt>,
558 ensure that only equivalent globals are ever merged (the "one definition
559 rule" - <tt>odr</tt>). Such languages can use the <tt>linkonce_odr</tt>
560 and <tt>weak_odr</tt> linkage types to indicate that the global will only
561 be merged with equivalent globals. These linkage types are otherwise the
562 same as their non-<tt>odr</tt> versions.
565 <dt><tt><b><a name="linkage_external">externally visible</a></b></tt>:</dt>
567 <dd>If none of the above identifiers are used, the global is externally
568 visible, meaning that it participates in linkage and can be used to resolve
569 external symbol references.
574 The next two types of linkage are targeted for Microsoft Windows platform
575 only. They are designed to support importing (exporting) symbols from (to)
576 DLLs (Dynamic Link Libraries).
580 <dt><tt><b><a name="linkage_dllimport">dllimport</a></b></tt>: </dt>
582 <dd>"<tt>dllimport</tt>" linkage causes the compiler to reference a function
583 or variable via a global pointer to a pointer that is set up by the DLL
584 exporting the symbol. On Microsoft Windows targets, the pointer name is
585 formed by combining <code>__imp_</code> and the function or variable name.
588 <dt><tt><b><a name="linkage_dllexport">dllexport</a></b></tt>: </dt>
590 <dd>"<tt>dllexport</tt>" linkage causes the compiler to provide a global
591 pointer to a pointer in a DLL, so that it can be referenced with the
592 <tt>dllimport</tt> attribute. On Microsoft Windows targets, the pointer
593 name is formed by combining <code>__imp_</code> and the function or variable
599 <p>For example, since the "<tt>.LC0</tt>"
600 variable is defined to be internal, if another module defined a "<tt>.LC0</tt>"
601 variable and was linked with this one, one of the two would be renamed,
602 preventing a collision. Since "<tt>main</tt>" and "<tt>puts</tt>" are
603 external (i.e., lacking any linkage declarations), they are accessible
604 outside of the current module.</p>
605 <p>It is illegal for a function <i>declaration</i>
606 to have any linkage type other than "externally visible", <tt>dllimport</tt>
607 or <tt>extern_weak</tt>.</p>
608 <p>Aliases can have only <tt>external</tt>, <tt>internal</tt>, <tt>weak</tt>
609 or <tt>weak_odr</tt> linkages.</p>
612 <!-- ======================================================================= -->
613 <div class="doc_subsection">
614 <a name="callingconv">Calling Conventions</a>
617 <div class="doc_text">
619 <p>LLVM <a href="#functionstructure">functions</a>, <a href="#i_call">calls</a>
620 and <a href="#i_invoke">invokes</a> can all have an optional calling convention
621 specified for the call. The calling convention of any pair of dynamic
622 caller/callee must match, or the behavior of the program is undefined. The
623 following calling conventions are supported by LLVM, and more may be added in
627 <dt><b>"<tt>ccc</tt>" - The C calling convention</b>:</dt>
629 <dd>This calling convention (the default if no other calling convention is
630 specified) matches the target C calling conventions. This calling convention
631 supports varargs function calls and tolerates some mismatch in the declared
632 prototype and implemented declaration of the function (as does normal C).
635 <dt><b>"<tt>fastcc</tt>" - The fast calling convention</b>:</dt>
637 <dd>This calling convention attempts to make calls as fast as possible
638 (e.g. by passing things in registers). This calling convention allows the
639 target to use whatever tricks it wants to produce fast code for the target,
640 without having to conform to an externally specified ABI (Application Binary
641 Interface). Implementations of this convention should allow arbitrary
642 <a href="CodeGenerator.html#tailcallopt">tail call optimization</a> to be
643 supported. This calling convention does not support varargs and requires the
644 prototype of all callees to exactly match the prototype of the function
648 <dt><b>"<tt>coldcc</tt>" - The cold calling convention</b>:</dt>
650 <dd>This calling convention attempts to make code in the caller as efficient
651 as possible under the assumption that the call is not commonly executed. As
652 such, these calls often preserve all registers so that the call does not break
653 any live ranges in the caller side. This calling convention does not support
654 varargs and requires the prototype of all callees to exactly match the
655 prototype of the function definition.
658 <dt><b>"<tt>cc <<em>n</em>></tt>" - Numbered convention</b>:</dt>
660 <dd>Any calling convention may be specified by number, allowing
661 target-specific calling conventions to be used. Target specific calling
662 conventions start at 64.
666 <p>More calling conventions can be added/defined on an as-needed basis, to
667 support pascal conventions or any other well-known target-independent
672 <!-- ======================================================================= -->
673 <div class="doc_subsection">
674 <a name="visibility">Visibility Styles</a>
677 <div class="doc_text">
680 All Global Variables and Functions have one of the following visibility styles:
684 <dt><b>"<tt>default</tt>" - Default style</b>:</dt>
686 <dd>On targets that use the ELF object file format, default visibility means
687 that the declaration is visible to other
688 modules and, in shared libraries, means that the declared entity may be
689 overridden. On Darwin, default visibility means that the declaration is
690 visible to other modules. Default visibility corresponds to "external
691 linkage" in the language.
694 <dt><b>"<tt>hidden</tt>" - Hidden style</b>:</dt>
696 <dd>Two declarations of an object with hidden visibility refer to the same
697 object if they are in the same shared object. Usually, hidden visibility
698 indicates that the symbol will not be placed into the dynamic symbol table,
699 so no other module (executable or shared library) can reference it
703 <dt><b>"<tt>protected</tt>" - Protected style</b>:</dt>
705 <dd>On ELF, protected visibility indicates that the symbol will be placed in
706 the dynamic symbol table, but that references within the defining module will
707 bind to the local symbol. That is, the symbol cannot be overridden by another
714 <!-- ======================================================================= -->
715 <div class="doc_subsection">
716 <a name="namedtypes">Named Types</a>
719 <div class="doc_text">
721 <p>LLVM IR allows you to specify name aliases for certain types. This can make
722 it easier to read the IR and make the IR more condensed (particularly when
723 recursive types are involved). An example of a name specification is:
726 <div class="doc_code">
728 %mytype = type { %mytype*, i32 }
732 <p>You may give a name to any <a href="#typesystem">type</a> except "<a
733 href="t_void">void</a>". Type name aliases may be used anywhere a type is
734 expected with the syntax "%mytype".</p>
736 <p>Note that type names are aliases for the structural type that they indicate,
737 and that you can therefore specify multiple names for the same type. This often
738 leads to confusing behavior when dumping out a .ll file. Since LLVM IR uses
739 structural typing, the name is not part of the type. When printing out LLVM IR,
740 the printer will pick <em>one name</em> to render all types of a particular
741 shape. This means that if you have code where two different source types end up
742 having the same LLVM type, that the dumper will sometimes print the "wrong" or
743 unexpected type. This is an important design point and isn't going to
748 <!-- ======================================================================= -->
749 <div class="doc_subsection">
750 <a name="globalvars">Global Variables</a>
753 <div class="doc_text">
755 <p>Global variables define regions of memory allocated at compilation time
756 instead of run-time. Global variables may optionally be initialized, may have
757 an explicit section to be placed in, and may have an optional explicit alignment
758 specified. A variable may be defined as "thread_local", which means that it
759 will not be shared by threads (each thread will have a separated copy of the
760 variable). A variable may be defined as a global "constant," which indicates
761 that the contents of the variable will <b>never</b> be modified (enabling better
762 optimization, allowing the global data to be placed in the read-only section of
763 an executable, etc). Note that variables that need runtime initialization
764 cannot be marked "constant" as there is a store to the variable.</p>
767 LLVM explicitly allows <em>declarations</em> of global variables to be marked
768 constant, even if the final definition of the global is not. This capability
769 can be used to enable slightly better optimization of the program, but requires
770 the language definition to guarantee that optimizations based on the
771 'constantness' are valid for the translation units that do not include the
775 <p>As SSA values, global variables define pointer values that are in
776 scope (i.e. they dominate) all basic blocks in the program. Global
777 variables always define a pointer to their "content" type because they
778 describe a region of memory, and all memory objects in LLVM are
779 accessed through pointers.</p>
781 <p>A global variable may be declared to reside in a target-specifc numbered
782 address space. For targets that support them, address spaces may affect how
783 optimizations are performed and/or what target instructions are used to access
784 the variable. The default address space is zero. The address space qualifier
785 must precede any other attributes.</p>
787 <p>LLVM allows an explicit section to be specified for globals. If the target
788 supports it, it will emit globals to the section specified.</p>
790 <p>An explicit alignment may be specified for a global. If not present, or if
791 the alignment is set to zero, the alignment of the global is set by the target
792 to whatever it feels convenient. If an explicit alignment is specified, the
793 global is forced to have at least that much alignment. All alignments must be
796 <p>For example, the following defines a global in a numbered address space with
797 an initializer, section, and alignment:</p>
799 <div class="doc_code">
801 @G = addrspace(5) constant float 1.0, section "foo", align 4
808 <!-- ======================================================================= -->
809 <div class="doc_subsection">
810 <a name="functionstructure">Functions</a>
813 <div class="doc_text">
815 <p>LLVM function definitions consist of the "<tt>define</tt>" keyord,
816 an optional <a href="#linkage">linkage type</a>, an optional
817 <a href="#visibility">visibility style</a>, an optional
818 <a href="#callingconv">calling convention</a>, a return type, an optional
819 <a href="#paramattrs">parameter attribute</a> for the return type, a function
820 name, a (possibly empty) argument list (each with optional
821 <a href="#paramattrs">parameter attributes</a>), optional
822 <a href="#fnattrs">function attributes</a>, an optional section,
823 an optional alignment, an optional <a href="#gc">garbage collector name</a>,
824 an opening curly brace, a list of basic blocks, and a closing curly brace.
826 LLVM function declarations consist of the "<tt>declare</tt>" keyword, an
827 optional <a href="#linkage">linkage type</a>, an optional
828 <a href="#visibility">visibility style</a>, an optional
829 <a href="#callingconv">calling convention</a>, a return type, an optional
830 <a href="#paramattrs">parameter attribute</a> for the return type, a function
831 name, a possibly empty list of arguments, an optional alignment, and an optional
832 <a href="#gc">garbage collector name</a>.</p>
834 <p>A function definition contains a list of basic blocks, forming the CFG
835 (Control Flow Graph) for
836 the function. Each basic block may optionally start with a label (giving the
837 basic block a symbol table entry), contains a list of instructions, and ends
838 with a <a href="#terminators">terminator</a> instruction (such as a branch or
839 function return).</p>
841 <p>The first basic block in a function is special in two ways: it is immediately
842 executed on entrance to the function, and it is not allowed to have predecessor
843 basic blocks (i.e. there can not be any branches to the entry block of a
844 function). Because the block can have no predecessors, it also cannot have any
845 <a href="#i_phi">PHI nodes</a>.</p>
847 <p>LLVM allows an explicit section to be specified for functions. If the target
848 supports it, it will emit functions to the section specified.</p>
850 <p>An explicit alignment may be specified for a function. If not present, or if
851 the alignment is set to zero, the alignment of the function is set by the target
852 to whatever it feels convenient. If an explicit alignment is specified, the
853 function is forced to have at least that much alignment. All alignments must be
858 <div class="doc_code">
860 define [<a href="#linkage">linkage</a>] [<a href="#visibility">visibility</a>]
861 [<a href="#callingconv">cconv</a>] [<a href="#paramattrs">ret attrs</a>]
862 <ResultType> @<FunctionName> ([argument list])
863 [<a href="#fnattrs">fn Attrs</a>] [section "name"] [align N]
864 [<a href="#gc">gc</a>] { ... }
871 <!-- ======================================================================= -->
872 <div class="doc_subsection">
873 <a name="aliasstructure">Aliases</a>
875 <div class="doc_text">
876 <p>Aliases act as "second name" for the aliasee value (which can be either
877 function, global variable, another alias or bitcast of global value). Aliases
878 may have an optional <a href="#linkage">linkage type</a>, and an
879 optional <a href="#visibility">visibility style</a>.</p>
883 <div class="doc_code">
885 @<Name> = alias [Linkage] [Visibility] <AliaseeTy> @<Aliasee>
893 <!-- ======================================================================= -->
894 <div class="doc_subsection"><a name="paramattrs">Parameter Attributes</a></div>
895 <div class="doc_text">
896 <p>The return type and each parameter of a function type may have a set of
897 <i>parameter attributes</i> associated with them. Parameter attributes are
898 used to communicate additional information about the result or parameters of
899 a function. Parameter attributes are considered to be part of the function,
900 not of the function type, so functions with different parameter attributes
901 can have the same function type.</p>
903 <p>Parameter attributes are simple keywords that follow the type specified. If
904 multiple parameter attributes are needed, they are space separated. For
907 <div class="doc_code">
909 declare i32 @printf(i8* noalias nocapture, ...)
910 declare i32 @atoi(i8 zeroext)
911 declare signext i8 @returns_signed_char()
915 <p>Note that any attributes for the function result (<tt>nounwind</tt>,
916 <tt>readonly</tt>) come immediately after the argument list.</p>
918 <p>Currently, only the following parameter attributes are defined:</p>
920 <dt><tt>zeroext</tt></dt>
921 <dd>This indicates to the code generator that the parameter or return value
922 should be zero-extended to a 32-bit value by the caller (for a parameter)
923 or the callee (for a return value).</dd>
925 <dt><tt>signext</tt></dt>
926 <dd>This indicates to the code generator that the parameter or return value
927 should be sign-extended to a 32-bit value by the caller (for a parameter)
928 or the callee (for a return value).</dd>
930 <dt><tt>inreg</tt></dt>
931 <dd>This indicates that this parameter or return value should be treated
932 in a special target-dependent fashion during while emitting code for a
933 function call or return (usually, by putting it in a register as opposed
934 to memory, though some targets use it to distinguish between two different
935 kinds of registers). Use of this attribute is target-specific.</dd>
937 <dt><tt><a name="byval">byval</a></tt></dt>
938 <dd>This indicates that the pointer parameter should really be passed by
939 value to the function. The attribute implies that a hidden copy of the
940 pointee is made between the caller and the callee, so the callee is unable
941 to modify the value in the callee. This attribute is only valid on LLVM
942 pointer arguments. It is generally used to pass structs and arrays by
943 value, but is also valid on pointers to scalars. The copy is considered to
944 belong to the caller not the callee (for example,
945 <tt><a href="#readonly">readonly</a></tt> functions should not write to
946 <tt>byval</tt> parameters). This is not a valid attribute for return
947 values. The byval attribute also supports specifying an alignment with the
948 align attribute. This has a target-specific effect on the code generator
949 that usually indicates a desired alignment for the synthesized stack
952 <dt><tt>sret</tt></dt>
953 <dd>This indicates that the pointer parameter specifies the address of a
954 structure that is the return value of the function in the source program.
955 This pointer must be guaranteed by the caller to be valid: loads and stores
956 to the structure may be assumed by the callee to not to trap. This may only
957 be applied to the first parameter. This is not a valid attribute for
960 <dt><tt>noalias</tt></dt>
961 <dd>This indicates that the pointer does not alias any global or any other
962 parameter. The caller is responsible for ensuring that this is the
963 case. On a function return value, <tt>noalias</tt> additionally indicates
964 that the pointer does not alias any other pointers visible to the
965 caller. For further details, please see the discussion of the NoAlias
967 <a href="http://llvm.org/docs/AliasAnalysis.html#MustMayNo">alias
970 <dt><tt>nocapture</tt></dt>
971 <dd>This indicates that the callee does not make any copies of the pointer
972 that outlive the callee itself. This is not a valid attribute for return
975 <dt><tt>nest</tt></dt>
976 <dd>This indicates that the pointer parameter can be excised using the
977 <a href="#int_trampoline">trampoline intrinsics</a>. This is not a valid
978 attribute for return values.</dd>
983 <!-- ======================================================================= -->
984 <div class="doc_subsection">
985 <a name="gc">Garbage Collector Names</a>
988 <div class="doc_text">
989 <p>Each function may specify a garbage collector name, which is simply a
992 <div class="doc_code"><pre
993 >define void @f() gc "name" { ...</pre></div>
995 <p>The compiler declares the supported values of <i>name</i>. Specifying a
996 collector which will cause the compiler to alter its output in order to support
997 the named garbage collection algorithm.</p>
1000 <!-- ======================================================================= -->
1001 <div class="doc_subsection">
1002 <a name="fnattrs">Function Attributes</a>
1005 <div class="doc_text">
1007 <p>Function attributes are set to communicate additional information about
1008 a function. Function attributes are considered to be part of the function,
1009 not of the function type, so functions with different parameter attributes
1010 can have the same function type.</p>
1012 <p>Function attributes are simple keywords that follow the type specified. If
1013 multiple attributes are needed, they are space separated. For
1016 <div class="doc_code">
1018 define void @f() noinline { ... }
1019 define void @f() alwaysinline { ... }
1020 define void @f() alwaysinline optsize { ... }
1021 define void @f() optsize
1026 <dt><tt>alwaysinline</tt></dt>
1027 <dd>This attribute indicates that the inliner should attempt to inline this
1028 function into callers whenever possible, ignoring any active inlining size
1029 threshold for this caller.</dd>
1031 <dt><tt>noinline</tt></dt>
1032 <dd>This attribute indicates that the inliner should never inline this function
1033 in any situation. This attribute may not be used together with the
1034 <tt>alwaysinline</tt> attribute.</dd>
1036 <dt><tt>optsize</tt></dt>
1037 <dd>This attribute suggests that optimization passes and code generator passes
1038 make choices that keep the code size of this function low, and otherwise do
1039 optimizations specifically to reduce code size.</dd>
1041 <dt><tt>noreturn</tt></dt>
1042 <dd>This function attribute indicates that the function never returns normally.
1043 This produces undefined behavior at runtime if the function ever does
1044 dynamically return.</dd>
1046 <dt><tt>nounwind</tt></dt>
1047 <dd>This function attribute indicates that the function never returns with an
1048 unwind or exceptional control flow. If the function does unwind, its runtime
1049 behavior is undefined.</dd>
1051 <dt><tt>readnone</tt></dt>
1052 <dd>This attribute indicates that the function computes its result (or the
1053 exception it throws) based strictly on its arguments, without dereferencing any
1054 pointer arguments or otherwise accessing any mutable state (e.g. memory, control
1055 registers, etc) visible to caller functions. It does not write through any
1056 pointer arguments (including <tt><a href="#byval">byval</a></tt> arguments) and
1057 never changes any state visible to callers.</dd>
1059 <dt><tt><a name="readonly">readonly</a></tt></dt>
1060 <dd>This attribute indicates that the function does not write through any
1061 pointer arguments (including <tt><a href="#byval">byval</a></tt> arguments)
1062 or otherwise modify any state (e.g. memory, control registers, etc) visible to
1063 caller functions. It may dereference pointer arguments and read state that may
1064 be set in the caller. A readonly function always returns the same value (or
1065 throws the same exception) when called with the same set of arguments and global
1068 <dt><tt><a name="ssp">ssp</a></tt></dt>
1069 <dd>This attribute indicates that the function should emit a stack smashing
1070 protector. It is in the form of a "canary"—a random value placed on the
1071 stack before the local variables that's checked upon return from the function to
1072 see if it has been overwritten. A heuristic is used to determine if a function
1073 needs stack protectors or not.
1075 <p>If a function that has an <tt>ssp</tt> attribute is inlined into a function
1076 that doesn't have an <tt>ssp</tt> attribute, then the resulting function will
1077 have an <tt>ssp</tt> attribute.</p></dd>
1079 <dt><tt>sspreq</tt></dt>
1080 <dd>This attribute indicates that the function should <em>always</em> emit a
1081 stack smashing protector. This overrides the <tt><a href="#ssp">ssp</a></tt>
1084 <p>If a function that has an <tt>sspreq</tt> attribute is inlined into a
1085 function that doesn't have an <tt>sspreq</tt> attribute or which has
1086 an <tt>ssp</tt> attribute, then the resulting function will have
1087 an <tt>sspreq</tt> attribute.</p></dd>
1092 <!-- ======================================================================= -->
1093 <div class="doc_subsection">
1094 <a name="moduleasm">Module-Level Inline Assembly</a>
1097 <div class="doc_text">
1099 Modules may contain "module-level inline asm" blocks, which corresponds to the
1100 GCC "file scope inline asm" blocks. These blocks are internally concatenated by
1101 LLVM and treated as a single unit, but may be separated in the .ll file if
1102 desired. The syntax is very simple:
1105 <div class="doc_code">
1107 module asm "inline asm code goes here"
1108 module asm "more can go here"
1112 <p>The strings can contain any character by escaping non-printable characters.
1113 The escape sequence used is simply "\xx" where "xx" is the two digit hex code
1118 The inline asm code is simply printed to the machine code .s file when
1119 assembly code is generated.
1123 <!-- ======================================================================= -->
1124 <div class="doc_subsection">
1125 <a name="datalayout">Data Layout</a>
1128 <div class="doc_text">
1129 <p>A module may specify a target specific data layout string that specifies how
1130 data is to be laid out in memory. The syntax for the data layout is simply:</p>
1131 <pre> target datalayout = "<i>layout specification</i>"</pre>
1132 <p>The <i>layout specification</i> consists of a list of specifications
1133 separated by the minus sign character ('-'). Each specification starts with a
1134 letter and may include other information after the letter to define some
1135 aspect of the data layout. The specifications accepted are as follows: </p>
1138 <dd>Specifies that the target lays out data in big-endian form. That is, the
1139 bits with the most significance have the lowest address location.</dd>
1141 <dd>Specifies that the target lays out data in little-endian form. That is,
1142 the bits with the least significance have the lowest address location.</dd>
1143 <dt><tt>p:<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1144 <dd>This specifies the <i>size</i> of a pointer and its <i>abi</i> and
1145 <i>preferred</i> alignments. All sizes are in bits. Specifying the <i>pref</i>
1146 alignment is optional. If omitted, the preceding <tt>:</tt> should be omitted
1148 <dt><tt>i<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1149 <dd>This specifies the alignment for an integer type of a given bit
1150 <i>size</i>. The value of <i>size</i> must be in the range [1,2^23).</dd>
1151 <dt><tt>v<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1152 <dd>This specifies the alignment for a vector type of a given bit
1154 <dt><tt>f<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1155 <dd>This specifies the alignment for a floating point type of a given bit
1156 <i>size</i>. The value of <i>size</i> must be either 32 (float) or 64
1158 <dt><tt>a<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1159 <dd>This specifies the alignment for an aggregate type of a given bit
1162 <p>When constructing the data layout for a given target, LLVM starts with a
1163 default set of specifications which are then (possibly) overriden by the
1164 specifications in the <tt>datalayout</tt> keyword. The default specifications
1165 are given in this list:</p>
1167 <li><tt>E</tt> - big endian</li>
1168 <li><tt>p:32:64:64</tt> - 32-bit pointers with 64-bit alignment</li>
1169 <li><tt>i1:8:8</tt> - i1 is 8-bit (byte) aligned</li>
1170 <li><tt>i8:8:8</tt> - i8 is 8-bit (byte) aligned</li>
1171 <li><tt>i16:16:16</tt> - i16 is 16-bit aligned</li>
1172 <li><tt>i32:32:32</tt> - i32 is 32-bit aligned</li>
1173 <li><tt>i64:32:64</tt> - i64 has ABI alignment of 32-bits but preferred
1174 alignment of 64-bits</li>
1175 <li><tt>f32:32:32</tt> - float is 32-bit aligned</li>
1176 <li><tt>f64:64:64</tt> - double is 64-bit aligned</li>
1177 <li><tt>v64:64:64</tt> - 64-bit vector is 64-bit aligned</li>
1178 <li><tt>v128:128:128</tt> - 128-bit vector is 128-bit aligned</li>
1179 <li><tt>a0:0:1</tt> - aggregates are 8-bit aligned</li>
1181 <p>When LLVM is determining the alignment for a given type, it uses the
1182 following rules:</p>
1184 <li>If the type sought is an exact match for one of the specifications, that
1185 specification is used.</li>
1186 <li>If no match is found, and the type sought is an integer type, then the
1187 smallest integer type that is larger than the bitwidth of the sought type is
1188 used. If none of the specifications are larger than the bitwidth then the the
1189 largest integer type is used. For example, given the default specifications
1190 above, the i7 type will use the alignment of i8 (next largest) while both
1191 i65 and i256 will use the alignment of i64 (largest specified).</li>
1192 <li>If no match is found, and the type sought is a vector type, then the
1193 largest vector type that is smaller than the sought vector type will be used
1194 as a fall back. This happens because <128 x double> can be implemented
1195 in terms of 64 <2 x double>, for example.</li>
1199 <!-- *********************************************************************** -->
1200 <div class="doc_section"> <a name="typesystem">Type System</a> </div>
1201 <!-- *********************************************************************** -->
1203 <div class="doc_text">
1205 <p>The LLVM type system is one of the most important features of the
1206 intermediate representation. Being typed enables a number of
1207 optimizations to be performed on the intermediate representation directly,
1208 without having to do
1209 extra analyses on the side before the transformation. A strong type
1210 system makes it easier to read the generated code and enables novel
1211 analyses and transformations that are not feasible to perform on normal
1212 three address code representations.</p>
1216 <!-- ======================================================================= -->
1217 <div class="doc_subsection"> <a name="t_classifications">Type
1218 Classifications</a> </div>
1219 <div class="doc_text">
1220 <p>The types fall into a few useful
1221 classifications:</p>
1223 <table border="1" cellspacing="0" cellpadding="4">
1225 <tr><th>Classification</th><th>Types</th></tr>
1227 <td><a href="#t_integer">integer</a></td>
1228 <td><tt>i1, i2, i3, ... i8, ... i16, ... i32, ... i64, ... </tt></td>
1231 <td><a href="#t_floating">floating point</a></td>
1232 <td><tt>float, double, x86_fp80, fp128, ppc_fp128</tt></td>
1235 <td><a name="t_firstclass">first class</a></td>
1236 <td><a href="#t_integer">integer</a>,
1237 <a href="#t_floating">floating point</a>,
1238 <a href="#t_pointer">pointer</a>,
1239 <a href="#t_vector">vector</a>,
1240 <a href="#t_struct">structure</a>,
1241 <a href="#t_array">array</a>,
1242 <a href="#t_label">label</a>.
1246 <td><a href="#t_primitive">primitive</a></td>
1247 <td><a href="#t_label">label</a>,
1248 <a href="#t_void">void</a>,
1249 <a href="#t_floating">floating point</a>.</td>
1252 <td><a href="#t_derived">derived</a></td>
1253 <td><a href="#t_integer">integer</a>,
1254 <a href="#t_array">array</a>,
1255 <a href="#t_function">function</a>,
1256 <a href="#t_pointer">pointer</a>,
1257 <a href="#t_struct">structure</a>,
1258 <a href="#t_pstruct">packed structure</a>,
1259 <a href="#t_vector">vector</a>,
1260 <a href="#t_opaque">opaque</a>.
1266 <p>The <a href="#t_firstclass">first class</a> types are perhaps the
1267 most important. Values of these types are the only ones which can be
1268 produced by instructions, passed as arguments, or used as operands to
1272 <!-- ======================================================================= -->
1273 <div class="doc_subsection"> <a name="t_primitive">Primitive Types</a> </div>
1275 <div class="doc_text">
1276 <p>The primitive types are the fundamental building blocks of the LLVM
1281 <!-- _______________________________________________________________________ -->
1282 <div class="doc_subsubsection"> <a name="t_floating">Floating Point Types</a> </div>
1284 <div class="doc_text">
1287 <tr><th>Type</th><th>Description</th></tr>
1288 <tr><td><tt>float</tt></td><td>32-bit floating point value</td></tr>
1289 <tr><td><tt>double</tt></td><td>64-bit floating point value</td></tr>
1290 <tr><td><tt>fp128</tt></td><td>128-bit floating point value (112-bit mantissa)</td></tr>
1291 <tr><td><tt>x86_fp80</tt></td><td>80-bit floating point value (X87)</td></tr>
1292 <tr><td><tt>ppc_fp128</tt></td><td>128-bit floating point value (two 64-bits)</td></tr>
1297 <!-- _______________________________________________________________________ -->
1298 <div class="doc_subsubsection"> <a name="t_void">Void Type</a> </div>
1300 <div class="doc_text">
1302 <p>The void type does not represent any value and has no size.</p>
1311 <!-- _______________________________________________________________________ -->
1312 <div class="doc_subsubsection"> <a name="t_label">Label Type</a> </div>
1314 <div class="doc_text">
1316 <p>The label type represents code labels.</p>
1326 <!-- ======================================================================= -->
1327 <div class="doc_subsection"> <a name="t_derived">Derived Types</a> </div>
1329 <div class="doc_text">
1331 <p>The real power in LLVM comes from the derived types in the system.
1332 This is what allows a programmer to represent arrays, functions,
1333 pointers, and other useful types. Note that these derived types may be
1334 recursive: For example, it is possible to have a two dimensional array.</p>
1338 <!-- _______________________________________________________________________ -->
1339 <div class="doc_subsubsection"> <a name="t_integer">Integer Type</a> </div>
1341 <div class="doc_text">
1344 <p>The integer type is a very simple derived type that simply specifies an
1345 arbitrary bit width for the integer type desired. Any bit width from 1 bit to
1346 2^23-1 (about 8 million) can be specified.</p>
1354 <p>The number of bits the integer will occupy is specified by the <tt>N</tt>
1358 <table class="layout">
1361 <td><tt>i1</tt></td>
1362 <td>a single-bit integer.</td>
1364 <td><tt>i32</tt></td>
1365 <td>a 32-bit integer.</td>
1367 <td><tt>i1942652</tt></td>
1368 <td>a really big integer of over 1 million bits.</td>
1373 <p>Note that the code generator does not yet support large integer types
1374 to be used as function return types. The specific limit on how large a
1375 return type the code generator can currently handle is target-dependent;
1376 currently it's often 64 bits for 32-bit targets and 128 bits for 64-bit
1381 <!-- _______________________________________________________________________ -->
1382 <div class="doc_subsubsection"> <a name="t_array">Array Type</a> </div>
1384 <div class="doc_text">
1388 <p>The array type is a very simple derived type that arranges elements
1389 sequentially in memory. The array type requires a size (number of
1390 elements) and an underlying data type.</p>
1395 [<# elements> x <elementtype>]
1398 <p>The number of elements is a constant integer value; elementtype may
1399 be any type with a size.</p>
1402 <table class="layout">
1404 <td class="left"><tt>[40 x i32]</tt></td>
1405 <td class="left">Array of 40 32-bit integer values.</td>
1408 <td class="left"><tt>[41 x i32]</tt></td>
1409 <td class="left">Array of 41 32-bit integer values.</td>
1412 <td class="left"><tt>[4 x i8]</tt></td>
1413 <td class="left">Array of 4 8-bit integer values.</td>
1416 <p>Here are some examples of multidimensional arrays:</p>
1417 <table class="layout">
1419 <td class="left"><tt>[3 x [4 x i32]]</tt></td>
1420 <td class="left">3x4 array of 32-bit integer values.</td>
1423 <td class="left"><tt>[12 x [10 x float]]</tt></td>
1424 <td class="left">12x10 array of single precision floating point values.</td>
1427 <td class="left"><tt>[2 x [3 x [4 x i16]]]</tt></td>
1428 <td class="left">2x3x4 array of 16-bit integer values.</td>
1432 <p>Note that 'variable sized arrays' can be implemented in LLVM with a zero
1433 length array. Normally, accesses past the end of an array are undefined in
1434 LLVM (e.g. it is illegal to access the 5th element of a 3 element array).
1435 As a special case, however, zero length arrays are recognized to be variable
1436 length. This allows implementation of 'pascal style arrays' with the LLVM
1437 type "{ i32, [0 x float]}", for example.</p>
1439 <p>Note that the code generator does not yet support large aggregate types
1440 to be used as function return types. The specific limit on how large an
1441 aggregate return type the code generator can currently handle is
1442 target-dependent, and also dependent on the aggregate element types.</p>
1446 <!-- _______________________________________________________________________ -->
1447 <div class="doc_subsubsection"> <a name="t_function">Function Type</a> </div>
1448 <div class="doc_text">
1452 <p>The function type can be thought of as a function signature. It
1453 consists of a return type and a list of formal parameter types. The
1454 return type of a function type is a scalar type, a void type, or a struct type.
1455 If the return type is a struct type then all struct elements must be of first
1456 class types, and the struct must have at least one element.</p>
1461 <returntype list> (<parameter list>)
1464 <p>...where '<tt><parameter list></tt>' is a comma-separated list of type
1465 specifiers. Optionally, the parameter list may include a type <tt>...</tt>,
1466 which indicates that the function takes a variable number of arguments.
1467 Variable argument functions can access their arguments with the <a
1468 href="#int_varargs">variable argument handling intrinsic</a> functions.
1469 '<tt><returntype list></tt>' is a comma-separated list of
1470 <a href="#t_firstclass">first class</a> type specifiers.</p>
1473 <table class="layout">
1475 <td class="left"><tt>i32 (i32)</tt></td>
1476 <td class="left">function taking an <tt>i32</tt>, returning an <tt>i32</tt>
1478 </tr><tr class="layout">
1479 <td class="left"><tt>float (i16 signext, i32 *) *
1481 <td class="left"><a href="#t_pointer">Pointer</a> to a function that takes
1482 an <tt>i16</tt> that should be sign extended and a
1483 <a href="#t_pointer">pointer</a> to <tt>i32</tt>, returning
1486 </tr><tr class="layout">
1487 <td class="left"><tt>i32 (i8*, ...)</tt></td>
1488 <td class="left">A vararg function that takes at least one
1489 <a href="#t_pointer">pointer</a> to <tt>i8 </tt> (char in C),
1490 which returns an integer. This is the signature for <tt>printf</tt> in
1493 </tr><tr class="layout">
1494 <td class="left"><tt>{i32, i32} (i32)</tt></td>
1495 <td class="left">A function taking an <tt>i32</tt>, returning two
1496 <tt>i32</tt> values as an aggregate of type <tt>{ i32, i32 }</tt>
1502 <!-- _______________________________________________________________________ -->
1503 <div class="doc_subsubsection"> <a name="t_struct">Structure Type</a> </div>
1504 <div class="doc_text">
1506 <p>The structure type is used to represent a collection of data members
1507 together in memory. The packing of the field types is defined to match
1508 the ABI of the underlying processor. The elements of a structure may
1509 be any type that has a size.</p>
1510 <p>Structures are accessed using '<tt><a href="#i_load">load</a></tt>
1511 and '<tt><a href="#i_store">store</a></tt>' by getting a pointer to a
1512 field with the '<tt><a href="#i_getelementptr">getelementptr</a></tt>'
1515 <pre> { <type list> }<br></pre>
1517 <table class="layout">
1519 <td class="left"><tt>{ i32, i32, i32 }</tt></td>
1520 <td class="left">A triple of three <tt>i32</tt> values</td>
1521 </tr><tr class="layout">
1522 <td class="left"><tt>{ float, i32 (i32) * }</tt></td>
1523 <td class="left">A pair, where the first element is a <tt>float</tt> and the
1524 second element is a <a href="#t_pointer">pointer</a> to a
1525 <a href="#t_function">function</a> that takes an <tt>i32</tt>, returning
1526 an <tt>i32</tt>.</td>
1530 <p>Note that the code generator does not yet support large aggregate types
1531 to be used as function return types. The specific limit on how large an
1532 aggregate return type the code generator can currently handle is
1533 target-dependent, and also dependent on the aggregate element types.</p>
1537 <!-- _______________________________________________________________________ -->
1538 <div class="doc_subsubsection"> <a name="t_pstruct">Packed Structure Type</a>
1540 <div class="doc_text">
1542 <p>The packed structure type is used to represent a collection of data members
1543 together in memory. There is no padding between fields. Further, the alignment
1544 of a packed structure is 1 byte. The elements of a packed structure may
1545 be any type that has a size.</p>
1546 <p>Structures are accessed using '<tt><a href="#i_load">load</a></tt>
1547 and '<tt><a href="#i_store">store</a></tt>' by getting a pointer to a
1548 field with the '<tt><a href="#i_getelementptr">getelementptr</a></tt>'
1551 <pre> < { <type list> } > <br></pre>
1553 <table class="layout">
1555 <td class="left"><tt>< { i32, i32, i32 } ></tt></td>
1556 <td class="left">A triple of three <tt>i32</tt> values</td>
1557 </tr><tr class="layout">
1559 <tt>< { float, i32 (i32)* } ></tt></td>
1560 <td class="left">A pair, where the first element is a <tt>float</tt> and the
1561 second element is a <a href="#t_pointer">pointer</a> to a
1562 <a href="#t_function">function</a> that takes an <tt>i32</tt>, returning
1563 an <tt>i32</tt>.</td>
1568 <!-- _______________________________________________________________________ -->
1569 <div class="doc_subsubsection"> <a name="t_pointer">Pointer Type</a> </div>
1570 <div class="doc_text">
1572 <p>As in many languages, the pointer type represents a pointer or
1573 reference to another object, which must live in memory. Pointer types may have
1574 an optional address space attribute defining the target-specific numbered
1575 address space where the pointed-to object resides. The default address space is
1578 <p>Note that LLVM does not permit pointers to void (<tt>void*</tt>) nor does
1579 it permit pointers to labels (<tt>label*</tt>). Use <tt>i8*</tt> instead.</p>
1582 <pre> <type> *<br></pre>
1584 <table class="layout">
1586 <td class="left"><tt>[4 x i32]*</tt></td>
1587 <td class="left">A <a href="#t_pointer">pointer</a> to <a
1588 href="#t_array">array</a> of four <tt>i32</tt> values.</td>
1591 <td class="left"><tt>i32 (i32 *) *</tt></td>
1592 <td class="left"> A <a href="#t_pointer">pointer</a> to a <a
1593 href="#t_function">function</a> that takes an <tt>i32*</tt>, returning an
1597 <td class="left"><tt>i32 addrspace(5)*</tt></td>
1598 <td class="left">A <a href="#t_pointer">pointer</a> to an <tt>i32</tt> value
1599 that resides in address space #5.</td>
1604 <!-- _______________________________________________________________________ -->
1605 <div class="doc_subsubsection"> <a name="t_vector">Vector Type</a> </div>
1606 <div class="doc_text">
1610 <p>A vector type is a simple derived type that represents a vector
1611 of elements. Vector types are used when multiple primitive data
1612 are operated in parallel using a single instruction (SIMD).
1613 A vector type requires a size (number of
1614 elements) and an underlying primitive data type. Vectors must have a power
1615 of two length (1, 2, 4, 8, 16 ...). Vector types are
1616 considered <a href="#t_firstclass">first class</a>.</p>
1621 < <# elements> x <elementtype> >
1624 <p>The number of elements is a constant integer value; elementtype may
1625 be any integer or floating point type.</p>
1629 <table class="layout">
1631 <td class="left"><tt><4 x i32></tt></td>
1632 <td class="left">Vector of 4 32-bit integer values.</td>
1635 <td class="left"><tt><8 x float></tt></td>
1636 <td class="left">Vector of 8 32-bit floating-point values.</td>
1639 <td class="left"><tt><2 x i64></tt></td>
1640 <td class="left">Vector of 2 64-bit integer values.</td>
1644 <p>Note that the code generator does not yet support large vector types
1645 to be used as function return types. The specific limit on how large a
1646 vector return type codegen can currently handle is target-dependent;
1647 currently it's often a few times longer than a hardware vector register.</p>
1651 <!-- _______________________________________________________________________ -->
1652 <div class="doc_subsubsection"> <a name="t_opaque">Opaque Type</a> </div>
1653 <div class="doc_text">
1657 <p>Opaque types are used to represent unknown types in the system. This
1658 corresponds (for example) to the C notion of a forward declared structure type.
1659 In LLVM, opaque types can eventually be resolved to any type (not just a
1660 structure type).</p>
1670 <table class="layout">
1672 <td class="left"><tt>opaque</tt></td>
1673 <td class="left">An opaque type.</td>
1678 <!-- ======================================================================= -->
1679 <div class="doc_subsection">
1680 <a name="t_uprefs">Type Up-references</a>
1683 <div class="doc_text">
1686 An "up reference" allows you to refer to a lexically enclosing type without
1687 requiring it to have a name. For instance, a structure declaration may contain a
1688 pointer to any of the types it is lexically a member of. Example of up
1689 references (with their equivalent as named type declarations) include:</p>
1692 { \2 * } %x = type { %x* }
1693 { \2 }* %y = type { %y }*
1698 An up reference is needed by the asmprinter for printing out cyclic types when
1699 there is no declared name for a type in the cycle. Because the asmprinter does
1700 not want to print out an infinite type string, it needs a syntax to handle
1701 recursive types that have no names (all names are optional in llvm IR).
1710 The level is the count of the lexical type that is being referred to.
1715 <table class="layout">
1717 <td class="left"><tt>\1*</tt></td>
1718 <td class="left">Self-referential pointer.</td>
1721 <td class="left"><tt>{ { \3*, i8 }, i32 }</tt></td>
1722 <td class="left">Recursive structure where the upref refers to the out-most
1729 <!-- *********************************************************************** -->
1730 <div class="doc_section"> <a name="constants">Constants</a> </div>
1731 <!-- *********************************************************************** -->
1733 <div class="doc_text">
1735 <p>LLVM has several different basic types of constants. This section describes
1736 them all and their syntax.</p>
1740 <!-- ======================================================================= -->
1741 <div class="doc_subsection"><a name="simpleconstants">Simple Constants</a></div>
1743 <div class="doc_text">
1746 <dt><b>Boolean constants</b></dt>
1748 <dd>The two strings '<tt>true</tt>' and '<tt>false</tt>' are both valid
1749 constants of the <tt><a href="#t_primitive">i1</a></tt> type.
1752 <dt><b>Integer constants</b></dt>
1754 <dd>Standard integers (such as '4') are constants of the <a
1755 href="#t_integer">integer</a> type. Negative numbers may be used with
1759 <dt><b>Floating point constants</b></dt>
1761 <dd>Floating point constants use standard decimal notation (e.g. 123.421),
1762 exponential notation (e.g. 1.23421e+2), or a more precise hexadecimal
1763 notation (see below). The assembler requires the exact decimal value of
1764 a floating-point constant. For example, the assembler accepts 1.25 but
1765 rejects 1.3 because 1.3 is a repeating decimal in binary. Floating point
1766 constants must have a <a href="#t_floating">floating point</a> type. </dd>
1768 <dt><b>Null pointer constants</b></dt>
1770 <dd>The identifier '<tt>null</tt>' is recognized as a null pointer constant
1771 and must be of <a href="#t_pointer">pointer type</a>.</dd>
1775 <p>The one non-intuitive notation for constants is the hexadecimal form
1776 of floating point constants. For example, the form '<tt>double
1777 0x432ff973cafa8000</tt>' is equivalent to (but harder to read than) '<tt>double
1778 4.5e+15</tt>'. The only time hexadecimal floating point constants are required
1779 (and the only time that they are generated by the disassembler) is when a
1780 floating point constant must be emitted but it cannot be represented as a
1781 decimal floating point number in a reasonable number of digits. For example,
1782 NaN's, infinities, and other
1783 special values are represented in their IEEE hexadecimal format so that
1784 assembly and disassembly do not cause any bits to change in the constants.</p>
1785 <p>When using the hexadecimal form, constants of types float and double are
1786 represented using the 16-digit form shown above (which matches the IEEE754
1787 representation for double); float values must, however, be exactly representable
1788 as IEE754 single precision.
1789 Hexadecimal format is always used for long
1790 double, and there are three forms of long double. The 80-bit
1791 format used by x86 is represented as <tt>0xK</tt>
1792 followed by 20 hexadecimal digits.
1793 The 128-bit format used by PowerPC (two adjacent doubles) is represented
1794 by <tt>0xM</tt> followed by 32 hexadecimal digits. The IEEE 128-bit
1795 format is represented
1796 by <tt>0xL</tt> followed by 32 hexadecimal digits; no currently supported
1797 target uses this format. Long doubles will only work if they match
1798 the long double format on your target. All hexadecimal formats are big-endian
1799 (sign bit at the left).</p>
1802 <!-- ======================================================================= -->
1803 <div class="doc_subsection">
1804 <a name="aggregateconstants"> <!-- old anchor -->
1805 <a name="complexconstants">Complex Constants</a></a>
1808 <div class="doc_text">
1809 <p>Complex constants are a (potentially recursive) combination of simple
1810 constants and smaller complex constants.</p>
1813 <dt><b>Structure constants</b></dt>
1815 <dd>Structure constants are represented with notation similar to structure
1816 type definitions (a comma separated list of elements, surrounded by braces
1817 (<tt>{}</tt>)). For example: "<tt>{ i32 4, float 17.0, i32* @G }</tt>",
1818 where "<tt>@G</tt>" is declared as "<tt>@G = external global i32</tt>". Structure constants
1819 must have <a href="#t_struct">structure type</a>, and the number and
1820 types of elements must match those specified by the type.
1823 <dt><b>Array constants</b></dt>
1825 <dd>Array constants are represented with notation similar to array type
1826 definitions (a comma separated list of elements, surrounded by square brackets
1827 (<tt>[]</tt>)). For example: "<tt>[ i32 42, i32 11, i32 74 ]</tt>". Array
1828 constants must have <a href="#t_array">array type</a>, and the number and
1829 types of elements must match those specified by the type.
1832 <dt><b>Vector constants</b></dt>
1834 <dd>Vector constants are represented with notation similar to vector type
1835 definitions (a comma separated list of elements, surrounded by
1836 less-than/greater-than's (<tt><></tt>)). For example: "<tt>< i32 42,
1837 i32 11, i32 74, i32 100 ></tt>". Vector constants must have <a
1838 href="#t_vector">vector type</a>, and the number and types of elements must
1839 match those specified by the type.
1842 <dt><b>Zero initialization</b></dt>
1844 <dd>The string '<tt>zeroinitializer</tt>' can be used to zero initialize a
1845 value to zero of <em>any</em> type, including scalar and aggregate types.
1846 This is often used to avoid having to print large zero initializers (e.g. for
1847 large arrays) and is always exactly equivalent to using explicit zero
1854 <!-- ======================================================================= -->
1855 <div class="doc_subsection">
1856 <a name="globalconstants">Global Variable and Function Addresses</a>
1859 <div class="doc_text">
1861 <p>The addresses of <a href="#globalvars">global variables</a> and <a
1862 href="#functionstructure">functions</a> are always implicitly valid (link-time)
1863 constants. These constants are explicitly referenced when the <a
1864 href="#identifiers">identifier for the global</a> is used and always have <a
1865 href="#t_pointer">pointer</a> type. For example, the following is a legal LLVM
1868 <div class="doc_code">
1872 @Z = global [2 x i32*] [ i32* @X, i32* @Y ]
1878 <!-- ======================================================================= -->
1879 <div class="doc_subsection"><a name="undefvalues">Undefined Values</a></div>
1880 <div class="doc_text">
1881 <p>The string '<tt>undef</tt>' is recognized as a type-less constant that has
1882 no specific value. Undefined values may be of any type and be used anywhere
1883 a constant is permitted.</p>
1885 <p>Undefined values indicate to the compiler that the program is well defined
1886 no matter what value is used, giving the compiler more freedom to optimize.
1890 <!-- ======================================================================= -->
1891 <div class="doc_subsection"><a name="constantexprs">Constant Expressions</a>
1894 <div class="doc_text">
1896 <p>Constant expressions are used to allow expressions involving other constants
1897 to be used as constants. Constant expressions may be of any <a
1898 href="#t_firstclass">first class</a> type and may involve any LLVM operation
1899 that does not have side effects (e.g. load and call are not supported). The
1900 following is the syntax for constant expressions:</p>
1903 <dt><b><tt>trunc ( CST to TYPE )</tt></b></dt>
1904 <dd>Truncate a constant to another type. The bit size of CST must be larger
1905 than the bit size of TYPE. Both types must be integers.</dd>
1907 <dt><b><tt>zext ( CST to TYPE )</tt></b></dt>
1908 <dd>Zero extend a constant to another type. The bit size of CST must be
1909 smaller or equal to the bit size of TYPE. Both types must be integers.</dd>
1911 <dt><b><tt>sext ( CST to TYPE )</tt></b></dt>
1912 <dd>Sign extend a constant to another type. The bit size of CST must be
1913 smaller or equal to the bit size of TYPE. Both types must be integers.</dd>
1915 <dt><b><tt>fptrunc ( CST to TYPE )</tt></b></dt>
1916 <dd>Truncate a floating point constant to another floating point type. The
1917 size of CST must be larger than the size of TYPE. Both types must be
1918 floating point.</dd>
1920 <dt><b><tt>fpext ( CST to TYPE )</tt></b></dt>
1921 <dd>Floating point extend a constant to another type. The size of CST must be
1922 smaller or equal to the size of TYPE. Both types must be floating point.</dd>
1924 <dt><b><tt>fptoui ( CST to TYPE )</tt></b></dt>
1925 <dd>Convert a floating point constant to the corresponding unsigned integer
1926 constant. TYPE must be a scalar or vector integer type. CST must be of scalar
1927 or vector floating point type. Both CST and TYPE must be scalars, or vectors
1928 of the same number of elements. If the value won't fit in the integer type,
1929 the results are undefined.</dd>
1931 <dt><b><tt>fptosi ( CST to TYPE )</tt></b></dt>
1932 <dd>Convert a floating point constant to the corresponding signed integer
1933 constant. TYPE must be a scalar or vector integer type. CST must be of scalar
1934 or vector floating point type. Both CST and TYPE must be scalars, or vectors
1935 of the same number of elements. If the value won't fit in the integer type,
1936 the results are undefined.</dd>
1938 <dt><b><tt>uitofp ( CST to TYPE )</tt></b></dt>
1939 <dd>Convert an unsigned integer constant to the corresponding floating point
1940 constant. TYPE must be a scalar or vector floating point type. CST must be of
1941 scalar or vector integer type. Both CST and TYPE must be scalars, or vectors
1942 of the same number of elements. If the value won't fit in the floating point
1943 type, the results are undefined.</dd>
1945 <dt><b><tt>sitofp ( CST to TYPE )</tt></b></dt>
1946 <dd>Convert a signed integer constant to the corresponding floating point
1947 constant. TYPE must be a scalar or vector floating point type. CST must be of
1948 scalar or vector integer type. Both CST and TYPE must be scalars, or vectors
1949 of the same number of elements. If the value won't fit in the floating point
1950 type, the results are undefined.</dd>
1952 <dt><b><tt>ptrtoint ( CST to TYPE )</tt></b></dt>
1953 <dd>Convert a pointer typed constant to the corresponding integer constant
1954 TYPE must be an integer type. CST must be of pointer type. The CST value is
1955 zero extended, truncated, or unchanged to make it fit in TYPE.</dd>
1957 <dt><b><tt>inttoptr ( CST to TYPE )</tt></b></dt>
1958 <dd>Convert a integer constant to a pointer constant. TYPE must be a
1959 pointer type. CST must be of integer type. The CST value is zero extended,
1960 truncated, or unchanged to make it fit in a pointer size. This one is
1961 <i>really</i> dangerous!</dd>
1963 <dt><b><tt>bitcast ( CST to TYPE )</tt></b></dt>
1964 <dd>Convert a constant, CST, to another TYPE. The constraints of the operands
1965 are the same as those for the <a href="#i_bitcast">bitcast
1966 instruction</a>.</dd>
1968 <dt><b><tt>getelementptr ( CSTPTR, IDX0, IDX1, ... )</tt></b></dt>
1970 <dd>Perform the <a href="#i_getelementptr">getelementptr operation</a> on
1971 constants. As with the <a href="#i_getelementptr">getelementptr</a>
1972 instruction, the index list may have zero or more indexes, which are required
1973 to make sense for the type of "CSTPTR".</dd>
1975 <dt><b><tt>select ( COND, VAL1, VAL2 )</tt></b></dt>
1977 <dd>Perform the <a href="#i_select">select operation</a> on
1980 <dt><b><tt>icmp COND ( VAL1, VAL2 )</tt></b></dt>
1981 <dd>Performs the <a href="#i_icmp">icmp operation</a> on constants.</dd>
1983 <dt><b><tt>fcmp COND ( VAL1, VAL2 )</tt></b></dt>
1984 <dd>Performs the <a href="#i_fcmp">fcmp operation</a> on constants.</dd>
1986 <dt><b><tt>vicmp COND ( VAL1, VAL2 )</tt></b></dt>
1987 <dd>Performs the <a href="#i_vicmp">vicmp operation</a> on constants.</dd>
1989 <dt><b><tt>vfcmp COND ( VAL1, VAL2 )</tt></b></dt>
1990 <dd>Performs the <a href="#i_vfcmp">vfcmp operation</a> on constants.</dd>
1992 <dt><b><tt>extractelement ( VAL, IDX )</tt></b></dt>
1994 <dd>Perform the <a href="#i_extractelement">extractelement
1995 operation</a> on constants.</dd>
1997 <dt><b><tt>insertelement ( VAL, ELT, IDX )</tt></b></dt>
1999 <dd>Perform the <a href="#i_insertelement">insertelement
2000 operation</a> on constants.</dd>
2003 <dt><b><tt>shufflevector ( VEC1, VEC2, IDXMASK )</tt></b></dt>
2005 <dd>Perform the <a href="#i_shufflevector">shufflevector
2006 operation</a> on constants.</dd>
2008 <dt><b><tt>OPCODE ( LHS, RHS )</tt></b></dt>
2010 <dd>Perform the specified operation of the LHS and RHS constants. OPCODE may
2011 be any of the <a href="#binaryops">binary</a> or <a href="#bitwiseops">bitwise
2012 binary</a> operations. The constraints on operands are the same as those for
2013 the corresponding instruction (e.g. no bitwise operations on floating point
2014 values are allowed).</dd>
2018 <!-- *********************************************************************** -->
2019 <div class="doc_section"> <a name="othervalues">Other Values</a> </div>
2020 <!-- *********************************************************************** -->
2022 <!-- ======================================================================= -->
2023 <div class="doc_subsection">
2024 <a name="inlineasm">Inline Assembler Expressions</a>
2027 <div class="doc_text">
2030 LLVM supports inline assembler expressions (as opposed to <a href="#moduleasm">
2031 Module-Level Inline Assembly</a>) through the use of a special value. This
2032 value represents the inline assembler as a string (containing the instructions
2033 to emit), a list of operand constraints (stored as a string), and a flag that
2034 indicates whether or not the inline asm expression has side effects. An example
2035 inline assembler expression is:
2038 <div class="doc_code">
2040 i32 (i32) asm "bswap $0", "=r,r"
2045 Inline assembler expressions may <b>only</b> be used as the callee operand of
2046 a <a href="#i_call"><tt>call</tt> instruction</a>. Thus, typically we have:
2049 <div class="doc_code">
2051 %X = call i32 asm "<a href="#int_bswap">bswap</a> $0", "=r,r"(i32 %Y)
2056 Inline asms with side effects not visible in the constraint list must be marked
2057 as having side effects. This is done through the use of the
2058 '<tt>sideeffect</tt>' keyword, like so:
2061 <div class="doc_code">
2063 call void asm sideeffect "eieio", ""()
2067 <p>TODO: The format of the asm and constraints string still need to be
2068 documented here. Constraints on what can be done (e.g. duplication, moving, etc
2069 need to be documented). This is probably best done by reference to another
2070 document that covers inline asm from a holistic perspective.
2075 <!-- *********************************************************************** -->
2076 <div class="doc_section"> <a name="instref">Instruction Reference</a> </div>
2077 <!-- *********************************************************************** -->
2079 <div class="doc_text">
2081 <p>The LLVM instruction set consists of several different
2082 classifications of instructions: <a href="#terminators">terminator
2083 instructions</a>, <a href="#binaryops">binary instructions</a>,
2084 <a href="#bitwiseops">bitwise binary instructions</a>, <a
2085 href="#memoryops">memory instructions</a>, and <a href="#otherops">other
2086 instructions</a>.</p>
2090 <!-- ======================================================================= -->
2091 <div class="doc_subsection"> <a name="terminators">Terminator
2092 Instructions</a> </div>
2094 <div class="doc_text">
2096 <p>As mentioned <a href="#functionstructure">previously</a>, every
2097 basic block in a program ends with a "Terminator" instruction, which
2098 indicates which block should be executed after the current block is
2099 finished. These terminator instructions typically yield a '<tt>void</tt>'
2100 value: they produce control flow, not values (the one exception being
2101 the '<a href="#i_invoke"><tt>invoke</tt></a>' instruction).</p>
2102 <p>There are six different terminator instructions: the '<a
2103 href="#i_ret"><tt>ret</tt></a>' instruction, the '<a href="#i_br"><tt>br</tt></a>'
2104 instruction, the '<a href="#i_switch"><tt>switch</tt></a>' instruction,
2105 the '<a href="#i_invoke"><tt>invoke</tt></a>' instruction, the '<a
2106 href="#i_unwind"><tt>unwind</tt></a>' instruction, and the '<a
2107 href="#i_unreachable"><tt>unreachable</tt></a>' instruction.</p>
2111 <!-- _______________________________________________________________________ -->
2112 <div class="doc_subsubsection"> <a name="i_ret">'<tt>ret</tt>'
2113 Instruction</a> </div>
2114 <div class="doc_text">
2117 ret <type> <value> <i>; Return a value from a non-void function</i>
2118 ret void <i>; Return from void function</i>
2123 <p>The '<tt>ret</tt>' instruction is used to return control flow (and
2124 optionally a value) from a function back to the caller.</p>
2125 <p>There are two forms of the '<tt>ret</tt>' instruction: one that
2126 returns a value and then causes control flow, and one that just causes
2127 control flow to occur.</p>
2131 <p>The '<tt>ret</tt>' instruction optionally accepts a single argument,
2132 the return value. The type of the return value must be a
2133 '<a href="#t_firstclass">first class</a>' type.</p>
2135 <p>A function is not <a href="#wellformed">well formed</a> if
2136 it it has a non-void return type and contains a '<tt>ret</tt>'
2137 instruction with no return value or a return value with a type that
2138 does not match its type, or if it has a void return type and contains
2139 a '<tt>ret</tt>' instruction with a return value.</p>
2143 <p>When the '<tt>ret</tt>' instruction is executed, control flow
2144 returns back to the calling function's context. If the caller is a "<a
2145 href="#i_call"><tt>call</tt></a>" instruction, execution continues at
2146 the instruction after the call. If the caller was an "<a
2147 href="#i_invoke"><tt>invoke</tt></a>" instruction, execution continues
2148 at the beginning of the "normal" destination block. If the instruction
2149 returns a value, that value shall set the call or invoke instruction's
2155 ret i32 5 <i>; Return an integer value of 5</i>
2156 ret void <i>; Return from a void function</i>
2157 ret { i32, i8 } { i32 4, i8 2 } <i>; Return a struct of values 4 and 2</i>
2160 <p>Note that the code generator does not yet fully support large
2161 return values. The specific sizes that are currently supported are
2162 dependent on the target. For integers, on 32-bit targets the limit
2163 is often 64 bits, and on 64-bit targets the limit is often 128 bits.
2164 For aggregate types, the current limits are dependent on the element
2165 types; for example targets are often limited to 2 total integer
2166 elements and 2 total floating-point elements.</p>
2169 <!-- _______________________________________________________________________ -->
2170 <div class="doc_subsubsection"> <a name="i_br">'<tt>br</tt>' Instruction</a> </div>
2171 <div class="doc_text">
2173 <pre> br i1 <cond>, label <iftrue>, label <iffalse><br> br label <dest> <i>; Unconditional branch</i>
2176 <p>The '<tt>br</tt>' instruction is used to cause control flow to
2177 transfer to a different basic block in the current function. There are
2178 two forms of this instruction, corresponding to a conditional branch
2179 and an unconditional branch.</p>
2181 <p>The conditional branch form of the '<tt>br</tt>' instruction takes a
2182 single '<tt>i1</tt>' value and two '<tt>label</tt>' values. The
2183 unconditional form of the '<tt>br</tt>' instruction takes a single
2184 '<tt>label</tt>' value as a target.</p>
2186 <p>Upon execution of a conditional '<tt>br</tt>' instruction, the '<tt>i1</tt>'
2187 argument is evaluated. If the value is <tt>true</tt>, control flows
2188 to the '<tt>iftrue</tt>' <tt>label</tt> argument. If "cond" is <tt>false</tt>,
2189 control flows to the '<tt>iffalse</tt>' <tt>label</tt> argument.</p>
2191 <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
2192 href="#i_ret">ret</a> i32 1<br>IfUnequal:<br> <a href="#i_ret">ret</a> i32 0<br></pre>
2194 <!-- _______________________________________________________________________ -->
2195 <div class="doc_subsubsection">
2196 <a name="i_switch">'<tt>switch</tt>' Instruction</a>
2199 <div class="doc_text">
2203 switch <intty> <value>, label <defaultdest> [ <intty> <val>, label <dest> ... ]
2208 <p>The '<tt>switch</tt>' instruction is used to transfer control flow to one of
2209 several different places. It is a generalization of the '<tt>br</tt>'
2210 instruction, allowing a branch to occur to one of many possible
2216 <p>The '<tt>switch</tt>' instruction uses three parameters: an integer
2217 comparison value '<tt>value</tt>', a default '<tt>label</tt>' destination, and
2218 an array of pairs of comparison value constants and '<tt>label</tt>'s. The
2219 table is not allowed to contain duplicate constant entries.</p>
2223 <p>The <tt>switch</tt> instruction specifies a table of values and
2224 destinations. When the '<tt>switch</tt>' instruction is executed, this
2225 table is searched for the given value. If the value is found, control flow is
2226 transfered to the corresponding destination; otherwise, control flow is
2227 transfered to the default destination.</p>
2229 <h5>Implementation:</h5>
2231 <p>Depending on properties of the target machine and the particular
2232 <tt>switch</tt> instruction, this instruction may be code generated in different
2233 ways. For example, it could be generated as a series of chained conditional
2234 branches or with a lookup table.</p>
2239 <i>; Emulate a conditional br instruction</i>
2240 %Val = <a href="#i_zext">zext</a> i1 %value to i32
2241 switch i32 %Val, label %truedest [ i32 0, label %falsedest ]
2243 <i>; Emulate an unconditional br instruction</i>
2244 switch i32 0, label %dest [ ]
2246 <i>; Implement a jump table:</i>
2247 switch i32 %val, label %otherwise [ i32 0, label %onzero
2249 i32 2, label %ontwo ]
2253 <!-- _______________________________________________________________________ -->
2254 <div class="doc_subsubsection">
2255 <a name="i_invoke">'<tt>invoke</tt>' Instruction</a>
2258 <div class="doc_text">
2263 <result> = invoke [<a href="#callingconv">cconv</a>] [<a href="#paramattrs">ret attrs</a>] <ptr to function ty> <function ptr val>(<function args>) [<a href="#fnattrs">fn attrs</a>]
2264 to label <normal label> unwind label <exception label>
2269 <p>The '<tt>invoke</tt>' instruction causes control to transfer to a specified
2270 function, with the possibility of control flow transfer to either the
2271 '<tt>normal</tt>' label or the
2272 '<tt>exception</tt>' label. If the callee function returns with the
2273 "<tt><a href="#i_ret">ret</a></tt>" instruction, control flow will return to the
2274 "normal" label. If the callee (or any indirect callees) returns with the "<a
2275 href="#i_unwind"><tt>unwind</tt></a>" instruction, control is interrupted and
2276 continued at the dynamically nearest "exception" label.</p>
2280 <p>This instruction requires several arguments:</p>
2284 The optional "cconv" marker indicates which <a href="#callingconv">calling
2285 convention</a> the call should use. If none is specified, the call defaults
2286 to using C calling conventions.
2289 <li>The optional <a href="#paramattrs">Parameter Attributes</a> list for
2290 return values. Only '<tt>zeroext</tt>', '<tt>signext</tt>',
2291 and '<tt>inreg</tt>' attributes are valid here.</li>
2293 <li>'<tt>ptr to function ty</tt>': shall be the signature of the pointer to
2294 function value being invoked. In most cases, this is a direct function
2295 invocation, but indirect <tt>invoke</tt>s are just as possible, branching off
2296 an arbitrary pointer to function value.
2299 <li>'<tt>function ptr val</tt>': An LLVM value containing a pointer to a
2300 function to be invoked. </li>
2302 <li>'<tt>function args</tt>': argument list whose types match the function
2303 signature argument types. If the function signature indicates the function
2304 accepts a variable number of arguments, the extra arguments can be
2307 <li>'<tt>normal label</tt>': the label reached when the called function
2308 executes a '<tt><a href="#i_ret">ret</a></tt>' instruction. </li>
2310 <li>'<tt>exception label</tt>': the label reached when a callee returns with
2311 the <a href="#i_unwind"><tt>unwind</tt></a> instruction. </li>
2313 <li>The optional <a href="#fnattrs">function attributes</a> list. Only
2314 '<tt>noreturn</tt>', '<tt>nounwind</tt>', '<tt>readonly</tt>' and
2315 '<tt>readnone</tt>' attributes are valid here.</li>
2320 <p>This instruction is designed to operate as a standard '<tt><a
2321 href="#i_call">call</a></tt>' instruction in most regards. The primary
2322 difference is that it establishes an association with a label, which is used by
2323 the runtime library to unwind the stack.</p>
2325 <p>This instruction is used in languages with destructors to ensure that proper
2326 cleanup is performed in the case of either a <tt>longjmp</tt> or a thrown
2327 exception. Additionally, this is important for implementation of
2328 '<tt>catch</tt>' clauses in high-level languages that support them.</p>
2332 %retval = invoke i32 @Test(i32 15) to label %Continue
2333 unwind label %TestCleanup <i>; {i32}:retval set</i>
2334 %retval = invoke <a href="#callingconv">coldcc</a> i32 %Testfnptr(i32 15) to label %Continue
2335 unwind label %TestCleanup <i>; {i32}:retval set</i>
2340 <!-- _______________________________________________________________________ -->
2342 <div class="doc_subsubsection"> <a name="i_unwind">'<tt>unwind</tt>'
2343 Instruction</a> </div>
2345 <div class="doc_text">
2354 <p>The '<tt>unwind</tt>' instruction unwinds the stack, continuing control flow
2355 at the first callee in the dynamic call stack which used an <a
2356 href="#i_invoke"><tt>invoke</tt></a> instruction to perform the call. This is
2357 primarily used to implement exception handling.</p>
2361 <p>The '<tt>unwind</tt>' instruction causes execution of the current function to
2362 immediately halt. The dynamic call stack is then searched for the first <a
2363 href="#i_invoke"><tt>invoke</tt></a> instruction on the call stack. Once found,
2364 execution continues at the "exceptional" destination block specified by the
2365 <tt>invoke</tt> instruction. If there is no <tt>invoke</tt> instruction in the
2366 dynamic call chain, undefined behavior results.</p>
2369 <!-- _______________________________________________________________________ -->
2371 <div class="doc_subsubsection"> <a name="i_unreachable">'<tt>unreachable</tt>'
2372 Instruction</a> </div>
2374 <div class="doc_text">
2383 <p>The '<tt>unreachable</tt>' instruction has no defined semantics. This
2384 instruction is used to inform the optimizer that a particular portion of the
2385 code is not reachable. This can be used to indicate that the code after a
2386 no-return function cannot be reached, and other facts.</p>
2390 <p>The '<tt>unreachable</tt>' instruction has no defined semantics.</p>
2395 <!-- ======================================================================= -->
2396 <div class="doc_subsection"> <a name="binaryops">Binary Operations</a> </div>
2397 <div class="doc_text">
2398 <p>Binary operators are used to do most of the computation in a
2399 program. They require two operands of the same type, execute an operation on them, and
2400 produce a single value. The operands might represent
2401 multiple data, as is the case with the <a href="#t_vector">vector</a> data type.
2402 The result value has the same type as its operands.</p>
2403 <p>There are several different binary operators:</p>
2405 <!-- _______________________________________________________________________ -->
2406 <div class="doc_subsubsection">
2407 <a name="i_add">'<tt>add</tt>' Instruction</a>
2410 <div class="doc_text">
2415 <result> = add <ty> <op1>, <op2> <i>; yields {ty}:result</i>
2420 <p>The '<tt>add</tt>' instruction returns the sum of its two operands.</p>
2424 <p>The two arguments to the '<tt>add</tt>' instruction must be <a
2425 href="#t_integer">integer</a>, <a href="#t_floating">floating point</a>, or
2426 <a href="#t_vector">vector</a> values. Both arguments must have identical
2431 <p>The value produced is the integer or floating point sum of the two
2434 <p>If an integer sum has unsigned overflow, the result returned is the
2435 mathematical result modulo 2<sup>n</sup>, where n is the bit width of
2438 <p>Because LLVM integers use a two's complement representation, this
2439 instruction is appropriate for both signed and unsigned integers.</p>
2444 <result> = add i32 4, %var <i>; yields {i32}:result = 4 + %var</i>
2447 <!-- _______________________________________________________________________ -->
2448 <div class="doc_subsubsection">
2449 <a name="i_sub">'<tt>sub</tt>' Instruction</a>
2452 <div class="doc_text">
2457 <result> = sub <ty> <op1>, <op2> <i>; yields {ty}:result</i>
2462 <p>The '<tt>sub</tt>' instruction returns the difference of its two
2465 <p>Note that the '<tt>sub</tt>' instruction is used to represent the
2466 '<tt>neg</tt>' instruction present in most other intermediate
2467 representations.</p>
2471 <p>The two arguments to the '<tt>sub</tt>' instruction must be <a
2472 href="#t_integer">integer</a>, <a href="#t_floating">floating point</a>,
2473 or <a href="#t_vector">vector</a> values. Both arguments must have identical
2478 <p>The value produced is the integer or floating point difference of
2479 the two operands.</p>
2481 <p>If an integer difference has unsigned overflow, the result returned is the
2482 mathematical result modulo 2<sup>n</sup>, where n is the bit width of
2485 <p>Because LLVM integers use a two's complement representation, this
2486 instruction is appropriate for both signed and unsigned integers.</p>
2490 <result> = sub i32 4, %var <i>; yields {i32}:result = 4 - %var</i>
2491 <result> = sub i32 0, %val <i>; yields {i32}:result = -%var</i>
2495 <!-- _______________________________________________________________________ -->
2496 <div class="doc_subsubsection">
2497 <a name="i_mul">'<tt>mul</tt>' Instruction</a>
2500 <div class="doc_text">
2503 <pre> <result> = mul <ty> <op1>, <op2> <i>; yields {ty}:result</i>
2506 <p>The '<tt>mul</tt>' instruction returns the product of its two
2511 <p>The two arguments to the '<tt>mul</tt>' instruction must be <a
2512 href="#t_integer">integer</a>, <a href="#t_floating">floating point</a>,
2513 or <a href="#t_vector">vector</a> values. Both arguments must have identical
2518 <p>The value produced is the integer or floating point product of the
2521 <p>If the result of an integer multiplication has unsigned overflow,
2522 the result returned is the mathematical result modulo
2523 2<sup>n</sup>, where n is the bit width of the result.</p>
2524 <p>Because LLVM integers use a two's complement representation, and the
2525 result is the same width as the operands, this instruction returns the
2526 correct result for both signed and unsigned integers. If a full product
2527 (e.g. <tt>i32</tt>x<tt>i32</tt>-><tt>i64</tt>) is needed, the operands
2528 should be sign-extended or zero-extended as appropriate to the
2529 width of the full product.</p>
2531 <pre> <result> = mul i32 4, %var <i>; yields {i32}:result = 4 * %var</i>
2535 <!-- _______________________________________________________________________ -->
2536 <div class="doc_subsubsection"> <a name="i_udiv">'<tt>udiv</tt>' Instruction
2538 <div class="doc_text">
2540 <pre> <result> = udiv <ty> <op1>, <op2> <i>; yields {ty}:result</i>
2543 <p>The '<tt>udiv</tt>' instruction returns the quotient of its two
2548 <p>The two arguments to the '<tt>udiv</tt>' instruction must be
2549 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
2550 values. Both arguments must have identical types.</p>
2554 <p>The value produced is the unsigned integer quotient of the two operands.</p>
2555 <p>Note that unsigned integer division and signed integer division are distinct
2556 operations; for signed integer division, use '<tt>sdiv</tt>'.</p>
2557 <p>Division by zero leads to undefined behavior.</p>
2559 <pre> <result> = udiv i32 4, %var <i>; yields {i32}:result = 4 / %var</i>
2562 <!-- _______________________________________________________________________ -->
2563 <div class="doc_subsubsection"> <a name="i_sdiv">'<tt>sdiv</tt>' Instruction
2565 <div class="doc_text">
2568 <result> = sdiv <ty> <op1>, <op2> <i>; yields {ty}:result</i>
2573 <p>The '<tt>sdiv</tt>' instruction returns the quotient of its two
2578 <p>The two arguments to the '<tt>sdiv</tt>' instruction must be
2579 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
2580 values. Both arguments must have identical types.</p>
2583 <p>The value produced is the signed integer quotient of the two operands rounded towards zero.</p>
2584 <p>Note that signed integer division and unsigned integer division are distinct
2585 operations; for unsigned integer division, use '<tt>udiv</tt>'.</p>
2586 <p>Division by zero leads to undefined behavior. Overflow also leads to
2587 undefined behavior; this is a rare case, but can occur, for example,
2588 by doing a 32-bit division of -2147483648 by -1.</p>
2590 <pre> <result> = sdiv i32 4, %var <i>; yields {i32}:result = 4 / %var</i>
2593 <!-- _______________________________________________________________________ -->
2594 <div class="doc_subsubsection"> <a name="i_fdiv">'<tt>fdiv</tt>'
2595 Instruction</a> </div>
2596 <div class="doc_text">
2599 <result> = fdiv <ty> <op1>, <op2> <i>; yields {ty}:result</i>
2603 <p>The '<tt>fdiv</tt>' instruction returns the quotient of its two
2608 <p>The two arguments to the '<tt>fdiv</tt>' instruction must be
2609 <a href="#t_floating">floating point</a> or <a href="#t_vector">vector</a>
2610 of floating point values. Both arguments must have identical types.</p>
2614 <p>The value produced is the floating point quotient of the two operands.</p>
2619 <result> = fdiv float 4.0, %var <i>; yields {float}:result = 4.0 / %var</i>
2623 <!-- _______________________________________________________________________ -->
2624 <div class="doc_subsubsection"> <a name="i_urem">'<tt>urem</tt>' Instruction</a>
2626 <div class="doc_text">
2628 <pre> <result> = urem <ty> <op1>, <op2> <i>; yields {ty}:result</i>
2631 <p>The '<tt>urem</tt>' instruction returns the remainder from the
2632 unsigned division of its two arguments.</p>
2634 <p>The two arguments to the '<tt>urem</tt>' instruction must be
2635 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
2636 values. Both arguments must have identical types.</p>
2638 <p>This instruction returns the unsigned integer <i>remainder</i> of a division.
2639 This instruction always performs an unsigned division to get the remainder.</p>
2640 <p>Note that unsigned integer remainder and signed integer remainder are
2641 distinct operations; for signed integer remainder, use '<tt>srem</tt>'.</p>
2642 <p>Taking the remainder of a division by zero leads to undefined behavior.</p>
2644 <pre> <result> = urem i32 4, %var <i>; yields {i32}:result = 4 % %var</i>
2648 <!-- _______________________________________________________________________ -->
2649 <div class="doc_subsubsection">
2650 <a name="i_srem">'<tt>srem</tt>' Instruction</a>
2653 <div class="doc_text">
2658 <result> = srem <ty> <op1>, <op2> <i>; yields {ty}:result</i>
2663 <p>The '<tt>srem</tt>' instruction returns the remainder from the
2664 signed division of its two operands. This instruction can also take
2665 <a href="#t_vector">vector</a> versions of the values in which case
2666 the elements must be integers.</p>
2670 <p>The two arguments to the '<tt>srem</tt>' instruction must be
2671 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
2672 values. Both arguments must have identical types.</p>
2676 <p>This instruction returns the <i>remainder</i> of a division (where the result
2677 has the same sign as the dividend, <tt>op1</tt>), not the <i>modulo</i>
2678 operator (where the result has the same sign as the divisor, <tt>op2</tt>) of
2679 a value. For more information about the difference, see <a
2680 href="http://mathforum.org/dr.math/problems/anne.4.28.99.html">The
2681 Math Forum</a>. For a table of how this is implemented in various languages,
2682 please see <a href="http://en.wikipedia.org/wiki/Modulo_operation">
2683 Wikipedia: modulo operation</a>.</p>
2684 <p>Note that signed integer remainder and unsigned integer remainder are
2685 distinct operations; for unsigned integer remainder, use '<tt>urem</tt>'.</p>
2686 <p>Taking the remainder of a division by zero leads to undefined behavior.
2687 Overflow also leads to undefined behavior; this is a rare case, but can occur,
2688 for example, by taking the remainder of a 32-bit division of -2147483648 by -1.
2689 (The remainder doesn't actually overflow, but this rule lets srem be
2690 implemented using instructions that return both the result of the division
2691 and the remainder.)</p>
2693 <pre> <result> = srem i32 4, %var <i>; yields {i32}:result = 4 % %var</i>
2697 <!-- _______________________________________________________________________ -->
2698 <div class="doc_subsubsection">
2699 <a name="i_frem">'<tt>frem</tt>' Instruction</a> </div>
2701 <div class="doc_text">
2704 <pre> <result> = frem <ty> <op1>, <op2> <i>; yields {ty}:result</i>
2707 <p>The '<tt>frem</tt>' instruction returns the remainder from the
2708 division of its two operands.</p>
2710 <p>The two arguments to the '<tt>frem</tt>' instruction must be
2711 <a href="#t_floating">floating point</a> or <a href="#t_vector">vector</a>
2712 of floating point values. Both arguments must have identical types.</p>
2716 <p>This instruction returns the <i>remainder</i> of a division.
2717 The remainder has the same sign as the dividend.</p>
2722 <result> = frem float 4.0, %var <i>; yields {float}:result = 4.0 % %var</i>
2726 <!-- ======================================================================= -->
2727 <div class="doc_subsection"> <a name="bitwiseops">Bitwise Binary
2728 Operations</a> </div>
2729 <div class="doc_text">
2730 <p>Bitwise binary operators are used to do various forms of
2731 bit-twiddling in a program. They are generally very efficient
2732 instructions and can commonly be strength reduced from other
2733 instructions. They require two operands of the same type, execute an operation on them,
2734 and produce a single value. The resulting value is the same type as its operands.</p>
2737 <!-- _______________________________________________________________________ -->
2738 <div class="doc_subsubsection"> <a name="i_shl">'<tt>shl</tt>'
2739 Instruction</a> </div>
2740 <div class="doc_text">
2742 <pre> <result> = shl <ty> <op1>, <op2> <i>; yields {ty}:result</i>
2747 <p>The '<tt>shl</tt>' instruction returns the first operand shifted to
2748 the left a specified number of bits.</p>
2752 <p>Both arguments to the '<tt>shl</tt>' instruction must be the same <a
2753 href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
2754 type. '<tt>op2</tt>' is treated as an unsigned value.</p>
2758 <p>The value produced is <tt>op1</tt> * 2<sup><tt>op2</tt></sup> mod 2<sup>n</sup>,
2759 where n is the width of the result. If <tt>op2</tt> is (statically or dynamically) negative or
2760 equal to or larger than the number of bits in <tt>op1</tt>, the result is undefined.
2761 If the arguments are vectors, each vector element of <tt>op1</tt> is shifted by the
2762 corresponding shift amount in <tt>op2</tt>.</p>
2764 <h5>Example:</h5><pre>
2765 <result> = shl i32 4, %var <i>; yields {i32}: 4 << %var</i>
2766 <result> = shl i32 4, 2 <i>; yields {i32}: 16</i>
2767 <result> = shl i32 1, 10 <i>; yields {i32}: 1024</i>
2768 <result> = shl i32 1, 32 <i>; undefined</i>
2769 <result> = shl <2 x i32> < i32 1, i32 1>, < i32 1, i32 2> <i>; yields: result=<2 x i32> < i32 2, i32 4></i>
2772 <!-- _______________________________________________________________________ -->
2773 <div class="doc_subsubsection"> <a name="i_lshr">'<tt>lshr</tt>'
2774 Instruction</a> </div>
2775 <div class="doc_text">
2777 <pre> <result> = lshr <ty> <op1>, <op2> <i>; yields {ty}:result</i>
2781 <p>The '<tt>lshr</tt>' instruction (logical shift right) returns the first
2782 operand shifted to the right a specified number of bits with zero fill.</p>
2785 <p>Both arguments to the '<tt>lshr</tt>' instruction must be the same
2786 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
2787 type. '<tt>op2</tt>' is treated as an unsigned value.</p>
2791 <p>This instruction always performs a logical shift right operation. The most
2792 significant bits of the result will be filled with zero bits after the
2793 shift. If <tt>op2</tt> is (statically or dynamically) equal to or larger than
2794 the number of bits in <tt>op1</tt>, the result is undefined. If the arguments are
2795 vectors, each vector element of <tt>op1</tt> is shifted by the corresponding shift
2796 amount in <tt>op2</tt>.</p>
2800 <result> = lshr i32 4, 1 <i>; yields {i32}:result = 2</i>
2801 <result> = lshr i32 4, 2 <i>; yields {i32}:result = 1</i>
2802 <result> = lshr i8 4, 3 <i>; yields {i8}:result = 0</i>
2803 <result> = lshr i8 -2, 1 <i>; yields {i8}:result = 0x7FFFFFFF </i>
2804 <result> = lshr i32 1, 32 <i>; undefined</i>
2805 <result> = lshr <2 x i32> < i32 -2, i32 4>, < i32 1, i32 2> <i>; yields: result=<2 x i32> < i32 0x7FFFFFFF, i32 1></i>
2809 <!-- _______________________________________________________________________ -->
2810 <div class="doc_subsubsection"> <a name="i_ashr">'<tt>ashr</tt>'
2811 Instruction</a> </div>
2812 <div class="doc_text">
2815 <pre> <result> = ashr <ty> <op1>, <op2> <i>; yields {ty}:result</i>
2819 <p>The '<tt>ashr</tt>' instruction (arithmetic shift right) returns the first
2820 operand shifted to the right a specified number of bits with sign extension.</p>
2823 <p>Both arguments to the '<tt>ashr</tt>' instruction must be the same
2824 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
2825 type. '<tt>op2</tt>' is treated as an unsigned value.</p>
2828 <p>This instruction always performs an arithmetic shift right operation,
2829 The most significant bits of the result will be filled with the sign bit
2830 of <tt>op1</tt>. If <tt>op2</tt> is (statically or dynamically) equal to or
2831 larger than the number of bits in <tt>op1</tt>, the result is undefined. If the
2832 arguments are vectors, each vector element of <tt>op1</tt> is shifted by the
2833 corresponding shift amount in <tt>op2</tt>.</p>
2837 <result> = ashr i32 4, 1 <i>; yields {i32}:result = 2</i>
2838 <result> = ashr i32 4, 2 <i>; yields {i32}:result = 1</i>
2839 <result> = ashr i8 4, 3 <i>; yields {i8}:result = 0</i>
2840 <result> = ashr i8 -2, 1 <i>; yields {i8}:result = -1</i>
2841 <result> = ashr i32 1, 32 <i>; undefined</i>
2842 <result> = ashr <2 x i32> < i32 -2, i32 4>, < i32 1, i32 3> <i>; yields: result=<2 x i32> < i32 -1, i32 0></i>
2846 <!-- _______________________________________________________________________ -->
2847 <div class="doc_subsubsection"> <a name="i_and">'<tt>and</tt>'
2848 Instruction</a> </div>
2850 <div class="doc_text">
2855 <result> = and <ty> <op1>, <op2> <i>; yields {ty}:result</i>
2860 <p>The '<tt>and</tt>' instruction returns the bitwise logical and of
2861 its two operands.</p>
2865 <p>The two arguments to the '<tt>and</tt>' instruction must be
2866 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
2867 values. Both arguments must have identical types.</p>
2870 <p>The truth table used for the '<tt>and</tt>' instruction is:</p>
2873 <table border="1" cellspacing="0" cellpadding="4">
2905 <result> = and i32 4, %var <i>; yields {i32}:result = 4 & %var</i>
2906 <result> = and i32 15, 40 <i>; yields {i32}:result = 8</i>
2907 <result> = and i32 4, 8 <i>; yields {i32}:result = 0</i>
2910 <!-- _______________________________________________________________________ -->
2911 <div class="doc_subsubsection"> <a name="i_or">'<tt>or</tt>' Instruction</a> </div>
2912 <div class="doc_text">
2914 <pre> <result> = or <ty> <op1>, <op2> <i>; yields {ty}:result</i>
2917 <p>The '<tt>or</tt>' instruction returns the bitwise logical inclusive
2918 or of its two operands.</p>
2921 <p>The two arguments to the '<tt>or</tt>' instruction must be
2922 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
2923 values. Both arguments must have identical types.</p>
2925 <p>The truth table used for the '<tt>or</tt>' instruction is:</p>
2928 <table border="1" cellspacing="0" cellpadding="4">
2959 <pre> <result> = or i32 4, %var <i>; yields {i32}:result = 4 | %var</i>
2960 <result> = or i32 15, 40 <i>; yields {i32}:result = 47</i>
2961 <result> = or i32 4, 8 <i>; yields {i32}:result = 12</i>
2964 <!-- _______________________________________________________________________ -->
2965 <div class="doc_subsubsection"> <a name="i_xor">'<tt>xor</tt>'
2966 Instruction</a> </div>
2967 <div class="doc_text">
2969 <pre> <result> = xor <ty> <op1>, <op2> <i>; yields {ty}:result</i>
2972 <p>The '<tt>xor</tt>' instruction returns the bitwise logical exclusive
2973 or of its two operands. The <tt>xor</tt> is used to implement the
2974 "one's complement" operation, which is the "~" operator in C.</p>
2976 <p>The two arguments to the '<tt>xor</tt>' instruction must be
2977 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
2978 values. Both arguments must have identical types.</p>
2982 <p>The truth table used for the '<tt>xor</tt>' instruction is:</p>
2985 <table border="1" cellspacing="0" cellpadding="4">
3017 <pre> <result> = xor i32 4, %var <i>; yields {i32}:result = 4 ^ %var</i>
3018 <result> = xor i32 15, 40 <i>; yields {i32}:result = 39</i>
3019 <result> = xor i32 4, 8 <i>; yields {i32}:result = 12</i>
3020 <result> = xor i32 %V, -1 <i>; yields {i32}:result = ~%V</i>
3024 <!-- ======================================================================= -->
3025 <div class="doc_subsection">
3026 <a name="vectorops">Vector Operations</a>
3029 <div class="doc_text">
3031 <p>LLVM supports several instructions to represent vector operations in a
3032 target-independent manner. These instructions cover the element-access and
3033 vector-specific operations needed to process vectors effectively. While LLVM
3034 does directly support these vector operations, many sophisticated algorithms
3035 will want to use target-specific intrinsics to take full advantage of a specific
3040 <!-- _______________________________________________________________________ -->
3041 <div class="doc_subsubsection">
3042 <a name="i_extractelement">'<tt>extractelement</tt>' Instruction</a>
3045 <div class="doc_text">
3050 <result> = extractelement <n x <ty>> <val>, i32 <idx> <i>; yields <ty></i>
3056 The '<tt>extractelement</tt>' instruction extracts a single scalar
3057 element from a vector at a specified index.
3064 The first operand of an '<tt>extractelement</tt>' instruction is a
3065 value of <a href="#t_vector">vector</a> type. The second operand is
3066 an index indicating the position from which to extract the element.
3067 The index may be a variable.</p>
3072 The result is a scalar of the same type as the element type of
3073 <tt>val</tt>. Its value is the value at position <tt>idx</tt> of
3074 <tt>val</tt>. If <tt>idx</tt> exceeds the length of <tt>val</tt>, the
3075 results are undefined.
3081 %result = extractelement <4 x i32> %vec, i32 0 <i>; yields i32</i>
3086 <!-- _______________________________________________________________________ -->
3087 <div class="doc_subsubsection">
3088 <a name="i_insertelement">'<tt>insertelement</tt>' Instruction</a>
3091 <div class="doc_text">
3096 <result> = insertelement <n x <ty>> <val>, <ty> <elt>, i32 <idx> <i>; yields <n x <ty>></i>
3102 The '<tt>insertelement</tt>' instruction inserts a scalar
3103 element into a vector at a specified index.
3110 The first operand of an '<tt>insertelement</tt>' instruction is a
3111 value of <a href="#t_vector">vector</a> type. The second operand is a
3112 scalar value whose type must equal the element type of the first
3113 operand. The third operand is an index indicating the position at
3114 which to insert the value. The index may be a variable.</p>
3119 The result is a vector of the same type as <tt>val</tt>. Its
3120 element values are those of <tt>val</tt> except at position
3121 <tt>idx</tt>, where it gets the value <tt>elt</tt>. If <tt>idx</tt>
3122 exceeds the length of <tt>val</tt>, the results are undefined.
3128 %result = insertelement <4 x i32> %vec, i32 1, i32 0 <i>; yields <4 x i32></i>
3132 <!-- _______________________________________________________________________ -->
3133 <div class="doc_subsubsection">
3134 <a name="i_shufflevector">'<tt>shufflevector</tt>' Instruction</a>
3137 <div class="doc_text">
3142 <result> = shufflevector <n x <ty>> <v1>, <n x <ty>> <v2>, <m x i32> <mask> <i>; yields <m x <ty>></i>
3148 The '<tt>shufflevector</tt>' instruction constructs a permutation of elements
3149 from two input vectors, returning a vector with the same element type as
3150 the input and length that is the same as the shuffle mask.
3156 The first two operands of a '<tt>shufflevector</tt>' instruction are vectors
3157 with types that match each other. The third argument is a shuffle mask whose
3158 element type is always 'i32'. The result of the instruction is a vector whose
3159 length is the same as the shuffle mask and whose element type is the same as
3160 the element type of the first two operands.
3164 The shuffle mask operand is required to be a constant vector with either
3165 constant integer or undef values.
3171 The elements of the two input vectors are numbered from left to right across
3172 both of the vectors. The shuffle mask operand specifies, for each element of
3173 the result vector, which element of the two input vectors the result element
3174 gets. The element selector may be undef (meaning "don't care") and the second
3175 operand may be undef if performing a shuffle from only one vector.
3181 %result = shufflevector <4 x i32> %v1, <4 x i32> %v2,
3182 <4 x i32> <i32 0, i32 4, i32 1, i32 5> <i>; yields <4 x i32></i>
3183 %result = shufflevector <4 x i32> %v1, <4 x i32> undef,
3184 <4 x i32> <i32 0, i32 1, i32 2, i32 3> <i>; yields <4 x i32></i> - Identity shuffle.
3185 %result = shufflevector <8 x i32> %v1, <8 x i32> undef,
3186 <4 x i32> <i32 0, i32 1, i32 2, i32 3> <i>; yields <4 x i32></i>
3187 %result = shufflevector <4 x i32> %v1, <4 x i32> %v2,
3188 <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7 > <i>; yields <8 x i32></i>
3193 <!-- ======================================================================= -->
3194 <div class="doc_subsection">
3195 <a name="aggregateops">Aggregate Operations</a>
3198 <div class="doc_text">
3200 <p>LLVM supports several instructions for working with aggregate values.
3205 <!-- _______________________________________________________________________ -->
3206 <div class="doc_subsubsection">
3207 <a name="i_extractvalue">'<tt>extractvalue</tt>' Instruction</a>
3210 <div class="doc_text">
3215 <result> = extractvalue <aggregate type> <val>, <idx>{, <idx>}*
3221 The '<tt>extractvalue</tt>' instruction extracts the value of a struct field
3222 or array element from an aggregate value.
3229 The first operand of an '<tt>extractvalue</tt>' instruction is a
3230 value of <a href="#t_struct">struct</a> or <a href="#t_array">array</a>
3231 type. The operands are constant indices to specify which value to extract
3232 in a similar manner as indices in a
3233 '<tt><a href="#i_getelementptr">getelementptr</a></tt>' instruction.
3239 The result is the value at the position in the aggregate specified by
3246 %result = extractvalue {i32, float} %agg, 0 <i>; yields i32</i>
3251 <!-- _______________________________________________________________________ -->
3252 <div class="doc_subsubsection">
3253 <a name="i_insertvalue">'<tt>insertvalue</tt>' Instruction</a>
3256 <div class="doc_text">
3261 <result> = insertvalue <aggregate type> <val>, <ty> <val>, <idx> <i>; yields <n x <ty>></i>
3267 The '<tt>insertvalue</tt>' instruction inserts a value
3268 into a struct field or array element in an aggregate.
3275 The first operand of an '<tt>insertvalue</tt>' instruction is a
3276 value of <a href="#t_struct">struct</a> or <a href="#t_array">array</a> type.
3277 The second operand is a first-class value to insert.
3278 The following operands are constant indices
3279 indicating the position at which to insert the value in a similar manner as
3281 '<tt><a href="#i_getelementptr">getelementptr</a></tt>' instruction.
3282 The value to insert must have the same type as the value identified
3289 The result is an aggregate of the same type as <tt>val</tt>. Its
3290 value is that of <tt>val</tt> except that the value at the position
3291 specified by the indices is that of <tt>elt</tt>.
3297 %result = insertvalue {i32, float} %agg, i32 1, 0 <i>; yields {i32, float}</i>
3302 <!-- ======================================================================= -->
3303 <div class="doc_subsection">
3304 <a name="memoryops">Memory Access and Addressing Operations</a>
3307 <div class="doc_text">
3309 <p>A key design point of an SSA-based representation is how it
3310 represents memory. In LLVM, no memory locations are in SSA form, which
3311 makes things very simple. This section describes how to read, write,
3312 allocate, and free memory in LLVM.</p>
3316 <!-- _______________________________________________________________________ -->
3317 <div class="doc_subsubsection">
3318 <a name="i_malloc">'<tt>malloc</tt>' Instruction</a>
3321 <div class="doc_text">
3326 <result> = malloc <type>[, i32 <NumElements>][, align <alignment>] <i>; yields {type*}:result</i>
3331 <p>The '<tt>malloc</tt>' instruction allocates memory from the system
3332 heap and returns a pointer to it. The object is always allocated in the generic
3333 address space (address space zero).</p>
3337 <p>The '<tt>malloc</tt>' instruction allocates
3338 <tt>sizeof(<type>)*NumElements</tt>
3339 bytes of memory from the operating system and returns a pointer of the
3340 appropriate type to the program. If "NumElements" is specified, it is the
3341 number of elements allocated, otherwise "NumElements" is defaulted to be one.
3342 If a constant alignment is specified, the value result of the allocation is guaranteed to
3343 be aligned to at least that boundary. If not specified, or if zero, the target can
3344 choose to align the allocation on any convenient boundary.</p>
3346 <p>'<tt>type</tt>' must be a sized type.</p>
3350 <p>Memory is allocated using the system "<tt>malloc</tt>" function, and
3351 a pointer is returned. The result of a zero byte allocation is undefined. The
3352 result is null if there is insufficient memory available.</p>
3357 %array = malloc [4 x i8] <i>; yields {[%4 x i8]*}:array</i>
3359 %size = <a href="#i_add">add</a> i32 2, 2 <i>; yields {i32}:size = i32 4</i>
3360 %array1 = malloc i8, i32 4 <i>; yields {i8*}:array1</i>
3361 %array2 = malloc [12 x i8], i32 %size <i>; yields {[12 x i8]*}:array2</i>
3362 %array3 = malloc i32, i32 4, align 1024 <i>; yields {i32*}:array3</i>
3363 %array4 = malloc i32, align 1024 <i>; yields {i32*}:array4</i>
3366 <p>Note that the code generator does not yet respect the
3367 alignment value.</p>
3371 <!-- _______________________________________________________________________ -->
3372 <div class="doc_subsubsection">
3373 <a name="i_free">'<tt>free</tt>' Instruction</a>
3376 <div class="doc_text">
3381 free <type> <value> <i>; yields {void}</i>
3386 <p>The '<tt>free</tt>' instruction returns memory back to the unused
3387 memory heap to be reallocated in the future.</p>
3391 <p>'<tt>value</tt>' shall be a pointer value that points to a value
3392 that was allocated with the '<tt><a href="#i_malloc">malloc</a></tt>'
3397 <p>Access to the memory pointed to by the pointer is no longer defined
3398 after this instruction executes. If the pointer is null, the operation
3404 %array = <a href="#i_malloc">malloc</a> [4 x i8] <i>; yields {[4 x i8]*}:array</i>
3405 free [4 x i8]* %array
3409 <!-- _______________________________________________________________________ -->
3410 <div class="doc_subsubsection">
3411 <a name="i_alloca">'<tt>alloca</tt>' Instruction</a>
3414 <div class="doc_text">
3419 <result> = alloca <type>[, i32 <NumElements>][, align <alignment>] <i>; yields {type*}:result</i>
3424 <p>The '<tt>alloca</tt>' instruction allocates memory on the stack frame of the
3425 currently executing function, to be automatically released when this function
3426 returns to its caller. The object is always allocated in the generic address
3427 space (address space zero).</p>
3431 <p>The '<tt>alloca</tt>' instruction allocates <tt>sizeof(<type>)*NumElements</tt>
3432 bytes of memory on the runtime stack, returning a pointer of the
3433 appropriate type to the program. If "NumElements" is specified, it is the
3434 number of elements allocated, otherwise "NumElements" is defaulted to be one.
3435 If a constant alignment is specified, the value result of the allocation is guaranteed
3436 to be aligned to at least that boundary. If not specified, or if zero, the target
3437 can choose to align the allocation on any convenient boundary.</p>
3439 <p>'<tt>type</tt>' may be any sized type.</p>
3443 <p>Memory is allocated; a pointer is returned. The operation is undefiend if
3444 there is insufficient stack space for the allocation. '<tt>alloca</tt>'d
3445 memory is automatically released when the function returns. The '<tt>alloca</tt>'
3446 instruction is commonly used to represent automatic variables that must
3447 have an address available. When the function returns (either with the <tt><a
3448 href="#i_ret">ret</a></tt> or <tt><a href="#i_unwind">unwind</a></tt>
3449 instructions), the memory is reclaimed. Allocating zero bytes
3450 is legal, but the result is undefined.</p>
3455 %ptr = alloca i32 <i>; yields {i32*}:ptr</i>
3456 %ptr = alloca i32, i32 4 <i>; yields {i32*}:ptr</i>
3457 %ptr = alloca i32, i32 4, align 1024 <i>; yields {i32*}:ptr</i>
3458 %ptr = alloca i32, align 1024 <i>; yields {i32*}:ptr</i>
3462 <!-- _______________________________________________________________________ -->
3463 <div class="doc_subsubsection"> <a name="i_load">'<tt>load</tt>'
3464 Instruction</a> </div>
3465 <div class="doc_text">
3467 <pre> <result> = load <ty>* <pointer>[, align <alignment>]<br> <result> = volatile load <ty>* <pointer>[, align <alignment>]<br></pre>
3469 <p>The '<tt>load</tt>' instruction is used to read from memory.</p>
3471 <p>The argument to the '<tt>load</tt>' instruction specifies the memory
3472 address from which to load. The pointer must point to a <a
3473 href="#t_firstclass">first class</a> type. If the <tt>load</tt> is
3474 marked as <tt>volatile</tt>, then the optimizer is not allowed to modify
3475 the number or order of execution of this <tt>load</tt> with other
3476 volatile <tt>load</tt> and <tt><a href="#i_store">store</a></tt>
3479 The optional constant "align" argument specifies the alignment of the operation
3480 (that is, the alignment of the memory address). A value of 0 or an
3481 omitted "align" argument means that the operation has the preferential
3482 alignment for the target. It is the responsibility of the code emitter
3483 to ensure that the alignment information is correct. Overestimating
3484 the alignment results in an undefined behavior. Underestimating the
3485 alignment may produce less efficient code. An alignment of 1 is always
3489 <p>The location of memory pointed to is loaded.</p>
3491 <pre> %ptr = <a href="#i_alloca">alloca</a> i32 <i>; yields {i32*}:ptr</i>
3493 href="#i_store">store</a> i32 3, i32* %ptr <i>; yields {void}</i>
3494 %val = load i32* %ptr <i>; yields {i32}:val = i32 3</i>
3497 <!-- _______________________________________________________________________ -->
3498 <div class="doc_subsubsection"> <a name="i_store">'<tt>store</tt>'
3499 Instruction</a> </div>
3500 <div class="doc_text">
3502 <pre> store <ty> <value>, <ty>* <pointer>[, align <alignment>] <i>; yields {void}</i>
3503 volatile store <ty> <value>, <ty>* <pointer>[, align <alignment>] <i>; yields {void}</i>
3506 <p>The '<tt>store</tt>' instruction is used to write to memory.</p>
3508 <p>There are two arguments to the '<tt>store</tt>' instruction: a value
3509 to store and an address at which to store it. The type of the '<tt><pointer></tt>'
3510 operand must be a pointer to the <a href="#t_firstclass">first class</a> type
3511 of the '<tt><value></tt>'
3512 operand. If the <tt>store</tt> is marked as <tt>volatile</tt>, then the
3513 optimizer is not allowed to modify the number or order of execution of
3514 this <tt>store</tt> with other volatile <tt>load</tt> and <tt><a
3515 href="#i_store">store</a></tt> instructions.</p>
3517 The optional constant "align" argument specifies the alignment of the operation
3518 (that is, the alignment of the memory address). A value of 0 or an
3519 omitted "align" argument means that the operation has the preferential
3520 alignment for the target. It is the responsibility of the code emitter
3521 to ensure that the alignment information is correct. Overestimating
3522 the alignment results in an undefined behavior. Underestimating the
3523 alignment may produce less efficient code. An alignment of 1 is always
3527 <p>The contents of memory are updated to contain '<tt><value></tt>'
3528 at the location specified by the '<tt><pointer></tt>' operand.</p>
3530 <pre> %ptr = <a href="#i_alloca">alloca</a> i32 <i>; yields {i32*}:ptr</i>
3531 store i32 3, i32* %ptr <i>; yields {void}</i>
3532 %val = <a href="#i_load">load</a> i32* %ptr <i>; yields {i32}:val = i32 3</i>
3536 <!-- _______________________________________________________________________ -->
3537 <div class="doc_subsubsection">
3538 <a name="i_getelementptr">'<tt>getelementptr</tt>' Instruction</a>
3541 <div class="doc_text">
3544 <result> = getelementptr <pty>* <ptrval>{, <ty> <idx>}*
3550 The '<tt>getelementptr</tt>' instruction is used to get the address of a
3551 subelement of an aggregate data structure. It performs address calculation only
3552 and does not access memory.</p>
3556 <p>The first argument is always a pointer, and forms the basis of the
3557 calculation. The remaining arguments are indices, that indicate which of the
3558 elements of the aggregate object are indexed. The interpretation of each index
3559 is dependent on the type being indexed into. The first index always indexes the
3560 pointer value given as the first argument, the second index indexes a value of
3561 the type pointed to (not necessarily the value directly pointed to, since the
3562 first index can be non-zero), etc. The first type indexed into must be a pointer
3563 value, subsequent types can be arrays, vectors and structs. Note that subsequent
3564 types being indexed into can never be pointers, since that would require loading
3565 the pointer before continuing calculation.</p>
3567 <p>The type of each index argument depends on the type it is indexing into.
3568 When indexing into a (packed) structure, only <tt>i32</tt> integer
3569 <b>constants</b> are allowed. When indexing into an array, pointer or vector,
3570 only integers of 32 or 64 bits are allowed (also non-constants). 32-bit values
3571 will be sign extended to 64-bits if required.</p>
3573 <p>For example, let's consider a C code fragment and how it gets
3574 compiled to LLVM:</p>
3576 <div class="doc_code">
3589 int *foo(struct ST *s) {
3590 return &s[1].Z.B[5][13];
3595 <p>The LLVM code generated by the GCC frontend is:</p>
3597 <div class="doc_code">
3599 %RT = <a href="#namedtypes">type</a> { i8 , [10 x [20 x i32]], i8 }
3600 %ST = <a href="#namedtypes">type</a> { i32, double, %RT }
3602 define i32* %foo(%ST* %s) {
3604 %reg = getelementptr %ST* %s, i32 1, i32 2, i32 1, i32 5, i32 13
3612 <p>In the example above, the first index is indexing into the '<tt>%ST*</tt>'
3613 type, which is a pointer, yielding a '<tt>%ST</tt>' = '<tt>{ i32, double, %RT
3614 }</tt>' type, a structure. The second index indexes into the third element of
3615 the structure, yielding a '<tt>%RT</tt>' = '<tt>{ i8 , [10 x [20 x i32]],
3616 i8 }</tt>' type, another structure. The third index indexes into the second
3617 element of the structure, yielding a '<tt>[10 x [20 x i32]]</tt>' type, an
3618 array. The two dimensions of the array are subscripted into, yielding an
3619 '<tt>i32</tt>' type. The '<tt>getelementptr</tt>' instruction returns a pointer
3620 to this element, thus computing a value of '<tt>i32*</tt>' type.</p>
3622 <p>Note that it is perfectly legal to index partially through a
3623 structure, returning a pointer to an inner element. Because of this,
3624 the LLVM code for the given testcase is equivalent to:</p>
3627 define i32* %foo(%ST* %s) {
3628 %t1 = getelementptr %ST* %s, i32 1 <i>; yields %ST*:%t1</i>
3629 %t2 = getelementptr %ST* %t1, i32 0, i32 2 <i>; yields %RT*:%t2</i>
3630 %t3 = getelementptr %RT* %t2, i32 0, i32 1 <i>; yields [10 x [20 x i32]]*:%t3</i>
3631 %t4 = getelementptr [10 x [20 x i32]]* %t3, i32 0, i32 5 <i>; yields [20 x i32]*:%t4</i>
3632 %t5 = getelementptr [20 x i32]* %t4, i32 0, i32 13 <i>; yields i32*:%t5</i>
3637 <p>Note that it is undefined to access an array out of bounds: array
3638 and pointer indexes must always be within the defined bounds of the
3639 array type when accessed with an instruction that dereferences the
3640 pointer (e.g. a load or store instruction). The one exception for
3641 this rule is zero length arrays. These arrays are defined to be
3642 accessible as variable length arrays, which requires access beyond the
3643 zero'th element.</p>
3645 <p>The getelementptr instruction is often confusing. For some more insight
3646 into how it works, see <a href="GetElementPtr.html">the getelementptr
3652 <i>; yields [12 x i8]*:aptr</i>
3653 %aptr = getelementptr {i32, [12 x i8]}* %saptr, i64 0, i32 1
3654 <i>; yields i8*:vptr</i>
3655 %vptr = getelementptr {i32, <2 x i8>}* %svptr, i64 0, i32 1, i32 1
3656 <i>; yields i8*:eptr</i>
3657 %eptr = getelementptr [12 x i8]* %aptr, i64 0, i32 1
3661 <!-- ======================================================================= -->
3662 <div class="doc_subsection"> <a name="convertops">Conversion Operations</a>
3664 <div class="doc_text">
3665 <p>The instructions in this category are the conversion instructions (casting)
3666 which all take a single operand and a type. They perform various bit conversions
3670 <!-- _______________________________________________________________________ -->
3671 <div class="doc_subsubsection">
3672 <a name="i_trunc">'<tt>trunc .. to</tt>' Instruction</a>
3674 <div class="doc_text">
3678 <result> = trunc <ty> <value> to <ty2> <i>; yields ty2</i>
3683 The '<tt>trunc</tt>' instruction truncates its operand to the type <tt>ty2</tt>.
3688 The '<tt>trunc</tt>' instruction takes a <tt>value</tt> to trunc, which must
3689 be an <a href="#t_integer">integer</a> type, and a type that specifies the size
3690 and type of the result, which must be an <a href="#t_integer">integer</a>
3691 type. The bit size of <tt>value</tt> must be larger than the bit size of
3692 <tt>ty2</tt>. Equal sized types are not allowed.</p>
3696 The '<tt>trunc</tt>' instruction truncates the high order bits in <tt>value</tt>
3697 and converts the remaining bits to <tt>ty2</tt>. Since the source size must be
3698 larger than the destination size, <tt>trunc</tt> cannot be a <i>no-op cast</i>.
3699 It will always truncate bits.</p>
3703 %X = trunc i32 257 to i8 <i>; yields i8:1</i>
3704 %Y = trunc i32 123 to i1 <i>; yields i1:true</i>
3705 %Y = trunc i32 122 to i1 <i>; yields i1:false</i>
3709 <!-- _______________________________________________________________________ -->
3710 <div class="doc_subsubsection">
3711 <a name="i_zext">'<tt>zext .. to</tt>' Instruction</a>
3713 <div class="doc_text">
3717 <result> = zext <ty> <value> to <ty2> <i>; yields ty2</i>
3721 <p>The '<tt>zext</tt>' instruction zero extends its operand to type
3726 <p>The '<tt>zext</tt>' instruction takes a value to cast, which must be of
3727 <a href="#t_integer">integer</a> type, and a type to cast it to, which must
3728 also be of <a href="#t_integer">integer</a> type. The bit size of the
3729 <tt>value</tt> must be smaller than the bit size of the destination type,
3733 <p>The <tt>zext</tt> fills the high order bits of the <tt>value</tt> with zero
3734 bits until it reaches the size of the destination type, <tt>ty2</tt>.</p>
3736 <p>When zero extending from i1, the result will always be either 0 or 1.</p>
3740 %X = zext i32 257 to i64 <i>; yields i64:257</i>
3741 %Y = zext i1 true to i32 <i>; yields i32:1</i>
3745 <!-- _______________________________________________________________________ -->
3746 <div class="doc_subsubsection">
3747 <a name="i_sext">'<tt>sext .. to</tt>' Instruction</a>
3749 <div class="doc_text">
3753 <result> = sext <ty> <value> to <ty2> <i>; yields ty2</i>
3757 <p>The '<tt>sext</tt>' sign extends <tt>value</tt> to the type <tt>ty2</tt>.</p>
3761 The '<tt>sext</tt>' instruction takes a value to cast, which must be of
3762 <a href="#t_integer">integer</a> type, and a type to cast it to, which must
3763 also be of <a href="#t_integer">integer</a> type. The bit size of the
3764 <tt>value</tt> must be smaller than the bit size of the destination type,
3769 The '<tt>sext</tt>' instruction performs a sign extension by copying the sign
3770 bit (highest order bit) of the <tt>value</tt> until it reaches the bit size of
3771 the type <tt>ty2</tt>.</p>
3773 <p>When sign extending from i1, the extension always results in -1 or 0.</p>
3777 %X = sext i8 -1 to i16 <i>; yields i16 :65535</i>
3778 %Y = sext i1 true to i32 <i>; yields i32:-1</i>
3782 <!-- _______________________________________________________________________ -->
3783 <div class="doc_subsubsection">
3784 <a name="i_fptrunc">'<tt>fptrunc .. to</tt>' Instruction</a>
3787 <div class="doc_text">
3792 <result> = fptrunc <ty> <value> to <ty2> <i>; yields ty2</i>
3796 <p>The '<tt>fptrunc</tt>' instruction truncates <tt>value</tt> to type
3801 <p>The '<tt>fptrunc</tt>' instruction takes a <a href="#t_floating">floating
3802 point</a> value to cast and a <a href="#t_floating">floating point</a> type to
3803 cast it to. The size of <tt>value</tt> must be larger than the size of
3804 <tt>ty2</tt>. This implies that <tt>fptrunc</tt> cannot be used to make a
3805 <i>no-op cast</i>.</p>
3808 <p> The '<tt>fptrunc</tt>' instruction truncates a <tt>value</tt> from a larger
3809 <a href="#t_floating">floating point</a> type to a smaller
3810 <a href="#t_floating">floating point</a> type. If the value cannot fit within
3811 the destination type, <tt>ty2</tt>, then the results are undefined.</p>
3815 %X = fptrunc double 123.0 to float <i>; yields float:123.0</i>
3816 %Y = fptrunc double 1.0E+300 to float <i>; yields undefined</i>
3820 <!-- _______________________________________________________________________ -->
3821 <div class="doc_subsubsection">
3822 <a name="i_fpext">'<tt>fpext .. to</tt>' Instruction</a>
3824 <div class="doc_text">
3828 <result> = fpext <ty> <value> to <ty2> <i>; yields ty2</i>
3832 <p>The '<tt>fpext</tt>' extends a floating point <tt>value</tt> to a larger
3833 floating point value.</p>
3836 <p>The '<tt>fpext</tt>' instruction takes a
3837 <a href="#t_floating">floating point</a> <tt>value</tt> to cast,
3838 and a <a href="#t_floating">floating point</a> type to cast it to. The source
3839 type must be smaller than the destination type.</p>
3842 <p>The '<tt>fpext</tt>' instruction extends the <tt>value</tt> from a smaller
3843 <a href="#t_floating">floating point</a> type to a larger
3844 <a href="#t_floating">floating point</a> type. The <tt>fpext</tt> cannot be
3845 used to make a <i>no-op cast</i> because it always changes bits. Use
3846 <tt>bitcast</tt> to make a <i>no-op cast</i> for a floating point cast.</p>
3850 %X = fpext float 3.1415 to double <i>; yields double:3.1415</i>
3851 %Y = fpext float 1.0 to float <i>; yields float:1.0 (no-op)</i>
3855 <!-- _______________________________________________________________________ -->
3856 <div class="doc_subsubsection">
3857 <a name="i_fptoui">'<tt>fptoui .. to</tt>' Instruction</a>
3859 <div class="doc_text">
3863 <result> = fptoui <ty> <value> to <ty2> <i>; yields ty2</i>
3867 <p>The '<tt>fptoui</tt>' converts a floating point <tt>value</tt> to its
3868 unsigned integer equivalent of type <tt>ty2</tt>.
3872 <p>The '<tt>fptoui</tt>' instruction takes a value to cast, which must be a
3873 scalar or vector <a href="#t_floating">floating point</a> value, and a type
3874 to cast it to <tt>ty2</tt>, which must be an <a href="#t_integer">integer</a>
3875 type. If <tt>ty</tt> is a vector floating point type, <tt>ty2</tt> must be a
3876 vector integer type with the same number of elements as <tt>ty</tt></p>
3879 <p> The '<tt>fptoui</tt>' instruction converts its
3880 <a href="#t_floating">floating point</a> operand into the nearest (rounding
3881 towards zero) unsigned integer value. If the value cannot fit in <tt>ty2</tt>,
3882 the results are undefined.</p>
3886 %X = fptoui double 123.0 to i32 <i>; yields i32:123</i>
3887 %Y = fptoui float 1.0E+300 to i1 <i>; yields undefined:1</i>
3888 %X = fptoui float 1.04E+17 to i8 <i>; yields undefined:1</i>
3892 <!-- _______________________________________________________________________ -->
3893 <div class="doc_subsubsection">
3894 <a name="i_fptosi">'<tt>fptosi .. to</tt>' Instruction</a>
3896 <div class="doc_text">
3900 <result> = fptosi <ty> <value> to <ty2> <i>; yields ty2</i>
3904 <p>The '<tt>fptosi</tt>' instruction converts
3905 <a href="#t_floating">floating point</a> <tt>value</tt> to type <tt>ty2</tt>.
3909 <p> The '<tt>fptosi</tt>' instruction takes a value to cast, which must be a
3910 scalar or vector <a href="#t_floating">floating point</a> value, and a type
3911 to cast it to <tt>ty2</tt>, which must be an <a href="#t_integer">integer</a>
3912 type. If <tt>ty</tt> is a vector floating point type, <tt>ty2</tt> must be a
3913 vector integer type with the same number of elements as <tt>ty</tt></p>
3916 <p>The '<tt>fptosi</tt>' instruction converts its
3917 <a href="#t_floating">floating point</a> operand into the nearest (rounding
3918 towards zero) signed integer value. If the value cannot fit in <tt>ty2</tt>,
3919 the results are undefined.</p>
3923 %X = fptosi double -123.0 to i32 <i>; yields i32:-123</i>
3924 %Y = fptosi float 1.0E-247 to i1 <i>; yields undefined:1</i>
3925 %X = fptosi float 1.04E+17 to i8 <i>; yields undefined:1</i>
3929 <!-- _______________________________________________________________________ -->
3930 <div class="doc_subsubsection">
3931 <a name="i_uitofp">'<tt>uitofp .. to</tt>' Instruction</a>
3933 <div class="doc_text">
3937 <result> = uitofp <ty> <value> to <ty2> <i>; yields ty2</i>
3941 <p>The '<tt>uitofp</tt>' instruction regards <tt>value</tt> as an unsigned
3942 integer and converts that value to the <tt>ty2</tt> type.</p>
3945 <p>The '<tt>uitofp</tt>' instruction takes a value to cast, which must be a
3946 scalar or vector <a href="#t_integer">integer</a> value, and a type to cast it
3947 to <tt>ty2</tt>, which must be an <a href="#t_floating">floating point</a>
3948 type. If <tt>ty</tt> is a vector integer type, <tt>ty2</tt> must be a vector
3949 floating point type with the same number of elements as <tt>ty</tt></p>
3952 <p>The '<tt>uitofp</tt>' instruction interprets its operand as an unsigned
3953 integer quantity and converts it to the corresponding floating point value. If
3954 the value cannot fit in the floating point value, the results are undefined.</p>
3958 %X = uitofp i32 257 to float <i>; yields float:257.0</i>
3959 %Y = uitofp i8 -1 to double <i>; yields double:255.0</i>
3963 <!-- _______________________________________________________________________ -->
3964 <div class="doc_subsubsection">
3965 <a name="i_sitofp">'<tt>sitofp .. to</tt>' Instruction</a>
3967 <div class="doc_text">
3971 <result> = sitofp <ty> <value> to <ty2> <i>; yields ty2</i>
3975 <p>The '<tt>sitofp</tt>' instruction regards <tt>value</tt> as a signed
3976 integer and converts that value to the <tt>ty2</tt> type.</p>
3979 <p>The '<tt>sitofp</tt>' instruction takes a value to cast, which must be a
3980 scalar or vector <a href="#t_integer">integer</a> value, and a type to cast it
3981 to <tt>ty2</tt>, which must be an <a href="#t_floating">floating point</a>
3982 type. If <tt>ty</tt> is a vector integer type, <tt>ty2</tt> must be a vector
3983 floating point type with the same number of elements as <tt>ty</tt></p>
3986 <p>The '<tt>sitofp</tt>' instruction interprets its operand as a signed
3987 integer quantity and converts it to the corresponding floating point value. If
3988 the value cannot fit in the floating point value, the results are undefined.</p>
3992 %X = sitofp i32 257 to float <i>; yields float:257.0</i>
3993 %Y = sitofp i8 -1 to double <i>; yields double:-1.0</i>
3997 <!-- _______________________________________________________________________ -->
3998 <div class="doc_subsubsection">
3999 <a name="i_ptrtoint">'<tt>ptrtoint .. to</tt>' Instruction</a>
4001 <div class="doc_text">
4005 <result> = ptrtoint <ty> <value> to <ty2> <i>; yields ty2</i>
4009 <p>The '<tt>ptrtoint</tt>' instruction converts the pointer <tt>value</tt> to
4010 the integer type <tt>ty2</tt>.</p>
4013 <p>The '<tt>ptrtoint</tt>' instruction takes a <tt>value</tt> to cast, which
4014 must be a <a href="#t_pointer">pointer</a> value, and a type to cast it to
4015 <tt>ty2</tt>, which must be an <a href="#t_integer">integer</a> type.</p>
4018 <p>The '<tt>ptrtoint</tt>' instruction converts <tt>value</tt> to integer type
4019 <tt>ty2</tt> by interpreting the pointer value as an integer and either
4020 truncating or zero extending that value to the size of the integer type. If
4021 <tt>value</tt> is smaller than <tt>ty2</tt> then a zero extension is done. If
4022 <tt>value</tt> is larger than <tt>ty2</tt> then a truncation is done. If they
4023 are the same size, then nothing is done (<i>no-op cast</i>) other than a type
4028 %X = ptrtoint i32* %X to i8 <i>; yields truncation on 32-bit architecture</i>
4029 %Y = ptrtoint i32* %x to i64 <i>; yields zero extension on 32-bit architecture</i>
4033 <!-- _______________________________________________________________________ -->
4034 <div class="doc_subsubsection">
4035 <a name="i_inttoptr">'<tt>inttoptr .. to</tt>' Instruction</a>
4037 <div class="doc_text">
4041 <result> = inttoptr <ty> <value> to <ty2> <i>; yields ty2</i>
4045 <p>The '<tt>inttoptr</tt>' instruction converts an integer <tt>value</tt> to
4046 a pointer type, <tt>ty2</tt>.</p>
4049 <p>The '<tt>inttoptr</tt>' instruction takes an <a href="#t_integer">integer</a>
4050 value to cast, and a type to cast it to, which must be a
4051 <a href="#t_pointer">pointer</a> type.</p>
4054 <p>The '<tt>inttoptr</tt>' instruction converts <tt>value</tt> to type
4055 <tt>ty2</tt> by applying either a zero extension or a truncation depending on
4056 the size of the integer <tt>value</tt>. If <tt>value</tt> is larger than the
4057 size of a pointer then a truncation is done. If <tt>value</tt> is smaller than
4058 the size of a pointer then a zero extension is done. If they are the same size,
4059 nothing is done (<i>no-op cast</i>).</p>
4063 %X = inttoptr i32 255 to i32* <i>; yields zero extension on 64-bit architecture</i>
4064 %X = inttoptr i32 255 to i32* <i>; yields no-op on 32-bit architecture</i>
4065 %Y = inttoptr i64 0 to i32* <i>; yields truncation on 32-bit architecture</i>
4069 <!-- _______________________________________________________________________ -->
4070 <div class="doc_subsubsection">
4071 <a name="i_bitcast">'<tt>bitcast .. to</tt>' Instruction</a>
4073 <div class="doc_text">
4077 <result> = bitcast <ty> <value> to <ty2> <i>; yields ty2</i>
4082 <p>The '<tt>bitcast</tt>' instruction converts <tt>value</tt> to type
4083 <tt>ty2</tt> without changing any bits.</p>
4087 <p>The '<tt>bitcast</tt>' instruction takes a value to cast, which must be
4088 a non-aggregate first class value, and a type to cast it to, which must also be
4089 a non-aggregate <a href="#t_firstclass">first class</a> type. The bit sizes of
4091 and the destination type, <tt>ty2</tt>, must be identical. If the source
4092 type is a pointer, the destination type must also be a pointer. This
4093 instruction supports bitwise conversion of vectors to integers and to vectors
4094 of other types (as long as they have the same size).</p>
4097 <p>The '<tt>bitcast</tt>' instruction converts <tt>value</tt> to type
4098 <tt>ty2</tt>. It is always a <i>no-op cast</i> because no bits change with
4099 this conversion. The conversion is done as if the <tt>value</tt> had been
4100 stored to memory and read back as type <tt>ty2</tt>. Pointer types may only be
4101 converted to other pointer types with this instruction. To convert pointers to
4102 other types, use the <a href="#i_inttoptr">inttoptr</a> or
4103 <a href="#i_ptrtoint">ptrtoint</a> instructions first.</p>
4107 %X = bitcast i8 255 to i8 <i>; yields i8 :-1</i>
4108 %Y = bitcast i32* %x to sint* <i>; yields sint*:%x</i>
4109 %Z = bitcast <2 x int> %V to i64; <i>; yields i64: %V</i>
4113 <!-- ======================================================================= -->
4114 <div class="doc_subsection"> <a name="otherops">Other Operations</a> </div>
4115 <div class="doc_text">
4116 <p>The instructions in this category are the "miscellaneous"
4117 instructions, which defy better classification.</p>
4120 <!-- _______________________________________________________________________ -->
4121 <div class="doc_subsubsection"><a name="i_icmp">'<tt>icmp</tt>' Instruction</a>
4123 <div class="doc_text">
4125 <pre> <result> = icmp <cond> <ty> <op1>, <op2> <i>; yields {i1} or {<N x i1>}:result</i>
4128 <p>The '<tt>icmp</tt>' instruction returns a boolean value or
4129 a vector of boolean values based on comparison
4130 of its two integer, integer vector, or pointer operands.</p>
4132 <p>The '<tt>icmp</tt>' instruction takes three operands. The first operand is
4133 the condition code indicating the kind of comparison to perform. It is not
4134 a value, just a keyword. The possible condition code are:
4137 <li><tt>eq</tt>: equal</li>
4138 <li><tt>ne</tt>: not equal </li>
4139 <li><tt>ugt</tt>: unsigned greater than</li>
4140 <li><tt>uge</tt>: unsigned greater or equal</li>
4141 <li><tt>ult</tt>: unsigned less than</li>
4142 <li><tt>ule</tt>: unsigned less or equal</li>
4143 <li><tt>sgt</tt>: signed greater than</li>
4144 <li><tt>sge</tt>: signed greater or equal</li>
4145 <li><tt>slt</tt>: signed less than</li>
4146 <li><tt>sle</tt>: signed less or equal</li>
4148 <p>The remaining two arguments must be <a href="#t_integer">integer</a> or
4149 <a href="#t_pointer">pointer</a>
4150 or integer <a href="#t_vector">vector</a> typed.
4151 They must also be identical types.</p>
4153 <p>The '<tt>icmp</tt>' compares <tt>op1</tt> and <tt>op2</tt> according to
4154 the condition code given as <tt>cond</tt>. The comparison performed always
4155 yields either an <a href="#t_primitive"><tt>i1</tt></a> or vector of <tt>i1</tt> result, as follows:
4158 <li><tt>eq</tt>: yields <tt>true</tt> if the operands are equal,
4159 <tt>false</tt> otherwise. No sign interpretation is necessary or performed.
4161 <li><tt>ne</tt>: yields <tt>true</tt> if the operands are unequal,
4162 <tt>false</tt> otherwise. No sign interpretation is necessary or performed.</li>
4163 <li><tt>ugt</tt>: interprets the operands as unsigned values and yields
4164 <tt>true</tt> if <tt>op1</tt> is greater than <tt>op2</tt>.</li>
4165 <li><tt>uge</tt>: interprets the operands as unsigned values and yields
4166 <tt>true</tt> if <tt>op1</tt> is greater than or equal to <tt>op2</tt>.</li>
4167 <li><tt>ult</tt>: interprets the operands as unsigned values and yields
4168 <tt>true</tt> if <tt>op1</tt> is less than <tt>op2</tt>.</li>
4169 <li><tt>ule</tt>: interprets the operands as unsigned values and yields
4170 <tt>true</tt> if <tt>op1</tt> is less than or equal to <tt>op2</tt>.</li>
4171 <li><tt>sgt</tt>: interprets the operands as signed values and yields
4172 <tt>true</tt> if <tt>op1</tt> is greater than <tt>op2</tt>.</li>
4173 <li><tt>sge</tt>: interprets the operands as signed values and yields
4174 <tt>true</tt> if <tt>op1</tt> is greater than or equal to <tt>op2</tt>.</li>
4175 <li><tt>slt</tt>: interprets the operands as signed values and yields
4176 <tt>true</tt> if <tt>op1</tt> is less than <tt>op2</tt>.</li>
4177 <li><tt>sle</tt>: interprets the operands as signed values and yields
4178 <tt>true</tt> if <tt>op1</tt> is less than or equal to <tt>op2</tt>.</li>
4180 <p>If the operands are <a href="#t_pointer">pointer</a> typed, the pointer
4181 values are compared as if they were integers.</p>
4182 <p>If the operands are integer vectors, then they are compared
4183 element by element. The result is an <tt>i1</tt> vector with
4184 the same number of elements as the values being compared.
4185 Otherwise, the result is an <tt>i1</tt>.
4189 <pre> <result> = icmp eq i32 4, 5 <i>; yields: result=false</i>
4190 <result> = icmp ne float* %X, %X <i>; yields: result=false</i>
4191 <result> = icmp ult i16 4, 5 <i>; yields: result=true</i>
4192 <result> = icmp sgt i16 4, 5 <i>; yields: result=false</i>
4193 <result> = icmp ule i16 -4, 5 <i>; yields: result=false</i>
4194 <result> = icmp sge i16 4, 5 <i>; yields: result=false</i>
4197 <p>Note that the code generator does not yet support vector types with
4198 the <tt>icmp</tt> instruction.</p>
4202 <!-- _______________________________________________________________________ -->
4203 <div class="doc_subsubsection"><a name="i_fcmp">'<tt>fcmp</tt>' Instruction</a>
4205 <div class="doc_text">
4207 <pre> <result> = fcmp <cond> <ty> <op1>, <op2> <i>; yields {i1} or {<N x i1>}:result</i>
4210 <p>The '<tt>fcmp</tt>' instruction returns a boolean value
4211 or vector of boolean values based on comparison
4212 of its operands.</p>
4214 If the operands are floating point scalars, then the result
4215 type is a boolean (<a href="#t_primitive"><tt>i1</tt></a>).
4217 <p>If the operands are floating point vectors, then the result type
4218 is a vector of boolean with the same number of elements as the
4219 operands being compared.</p>
4221 <p>The '<tt>fcmp</tt>' instruction takes three operands. The first operand is
4222 the condition code indicating the kind of comparison to perform. It is not
4223 a value, just a keyword. The possible condition code are:</p>
4225 <li><tt>false</tt>: no comparison, always returns false</li>
4226 <li><tt>oeq</tt>: ordered and equal</li>
4227 <li><tt>ogt</tt>: ordered and greater than </li>
4228 <li><tt>oge</tt>: ordered and greater than or equal</li>
4229 <li><tt>olt</tt>: ordered and less than </li>
4230 <li><tt>ole</tt>: ordered and less than or equal</li>
4231 <li><tt>one</tt>: ordered and not equal</li>
4232 <li><tt>ord</tt>: ordered (no nans)</li>
4233 <li><tt>ueq</tt>: unordered or equal</li>
4234 <li><tt>ugt</tt>: unordered or greater than </li>
4235 <li><tt>uge</tt>: unordered or greater than or equal</li>
4236 <li><tt>ult</tt>: unordered or less than </li>
4237 <li><tt>ule</tt>: unordered or less than or equal</li>
4238 <li><tt>une</tt>: unordered or not equal</li>
4239 <li><tt>uno</tt>: unordered (either nans)</li>
4240 <li><tt>true</tt>: no comparison, always returns true</li>
4242 <p><i>Ordered</i> means that neither operand is a QNAN while
4243 <i>unordered</i> means that either operand may be a QNAN.</p>
4244 <p>Each of <tt>val1</tt> and <tt>val2</tt> arguments must be
4245 either a <a href="#t_floating">floating point</a> type
4246 or a <a href="#t_vector">vector</a> of floating point type.
4247 They must have identical types.</p>
4249 <p>The '<tt>fcmp</tt>' instruction compares <tt>op1</tt> and <tt>op2</tt>
4250 according to the condition code given as <tt>cond</tt>.
4251 If the operands are vectors, then the vectors are compared
4253 Each comparison performed
4254 always yields an <a href="#t_primitive">i1</a> result, as follows:</p>
4256 <li><tt>false</tt>: always yields <tt>false</tt>, regardless of operands.</li>
4257 <li><tt>oeq</tt>: yields <tt>true</tt> if both operands are not a QNAN and
4258 <tt>op1</tt> is equal to <tt>op2</tt>.</li>
4259 <li><tt>ogt</tt>: yields <tt>true</tt> if both operands are not a QNAN and
4260 <tt>op1</tt> is greather than <tt>op2</tt>.</li>
4261 <li><tt>oge</tt>: yields <tt>true</tt> if both operands are not a QNAN and
4262 <tt>op1</tt> is greater than or equal to <tt>op2</tt>.</li>
4263 <li><tt>olt</tt>: yields <tt>true</tt> if both operands are not a QNAN and
4264 <tt>op1</tt> is less than <tt>op2</tt>.</li>
4265 <li><tt>ole</tt>: yields <tt>true</tt> if both operands are not a QNAN and
4266 <tt>op1</tt> is less than or equal to <tt>op2</tt>.</li>
4267 <li><tt>one</tt>: yields <tt>true</tt> if both operands are not a QNAN and
4268 <tt>op1</tt> is not equal to <tt>op2</tt>.</li>
4269 <li><tt>ord</tt>: yields <tt>true</tt> if both operands are not a QNAN.</li>
4270 <li><tt>ueq</tt>: yields <tt>true</tt> if either operand is a QNAN or
4271 <tt>op1</tt> is equal to <tt>op2</tt>.</li>
4272 <li><tt>ugt</tt>: yields <tt>true</tt> if either operand is a QNAN or
4273 <tt>op1</tt> is greater than <tt>op2</tt>.</li>
4274 <li><tt>uge</tt>: yields <tt>true</tt> if either operand is a QNAN or
4275 <tt>op1</tt> is greater than or equal to <tt>op2</tt>.</li>
4276 <li><tt>ult</tt>: yields <tt>true</tt> if either operand is a QNAN or
4277 <tt>op1</tt> is less than <tt>op2</tt>.</li>
4278 <li><tt>ule</tt>: yields <tt>true</tt> if either operand is a QNAN or
4279 <tt>op1</tt> is less than or equal to <tt>op2</tt>.</li>
4280 <li><tt>une</tt>: yields <tt>true</tt> if either operand is a QNAN or
4281 <tt>op1</tt> is not equal to <tt>op2</tt>.</li>
4282 <li><tt>uno</tt>: yields <tt>true</tt> if either operand is a QNAN.</li>
4283 <li><tt>true</tt>: always yields <tt>true</tt>, regardless of operands.</li>
4287 <pre> <result> = fcmp oeq float 4.0, 5.0 <i>; yields: result=false</i>
4288 <result> = fcmp one float 4.0, 5.0 <i>; yields: result=true</i>
4289 <result> = fcmp olt float 4.0, 5.0 <i>; yields: result=true</i>
4290 <result> = fcmp ueq double 1.0, 2.0 <i>; yields: result=false</i>
4293 <p>Note that the code generator does not yet support vector types with
4294 the <tt>fcmp</tt> instruction.</p>
4298 <!-- _______________________________________________________________________ -->
4299 <div class="doc_subsubsection">
4300 <a name="i_vicmp">'<tt>vicmp</tt>' Instruction</a>
4302 <div class="doc_text">
4304 <pre> <result> = vicmp <cond> <ty> <op1>, <op2> <i>; yields {ty}:result</i>
4307 <p>The '<tt>vicmp</tt>' instruction returns an integer vector value based on
4308 element-wise comparison of its two integer vector operands.</p>
4310 <p>The '<tt>vicmp</tt>' instruction takes three operands. The first operand is
4311 the condition code indicating the kind of comparison to perform. It is not
4312 a value, just a keyword. The possible condition code are:</p>
4314 <li><tt>eq</tt>: equal</li>
4315 <li><tt>ne</tt>: not equal </li>
4316 <li><tt>ugt</tt>: unsigned greater than</li>
4317 <li><tt>uge</tt>: unsigned greater or equal</li>
4318 <li><tt>ult</tt>: unsigned less than</li>
4319 <li><tt>ule</tt>: unsigned less or equal</li>
4320 <li><tt>sgt</tt>: signed greater than</li>
4321 <li><tt>sge</tt>: signed greater or equal</li>
4322 <li><tt>slt</tt>: signed less than</li>
4323 <li><tt>sle</tt>: signed less or equal</li>
4325 <p>The remaining two arguments must be <a href="#t_vector">vector</a> or
4326 <a href="#t_integer">integer</a> typed. They must also be identical types.</p>
4328 <p>The '<tt>vicmp</tt>' instruction compares <tt>op1</tt> and <tt>op2</tt>
4329 according to the condition code given as <tt>cond</tt>. The comparison yields a
4330 <a href="#t_vector">vector</a> of <a href="#t_integer">integer</a> result, of
4331 identical type as the values being compared. The most significant bit in each
4332 element is 1 if the element-wise comparison evaluates to true, and is 0
4333 otherwise. All other bits of the result are undefined. The condition codes
4334 are evaluated identically to the <a href="#i_icmp">'<tt>icmp</tt>'
4335 instruction</a>.</p>
4339 <result> = vicmp eq <2 x i32> < i32 4, i32 0>, < i32 5, i32 0> <i>; yields: result=<2 x i32> < i32 0, i32 -1 ></i>
4340 <result> = vicmp ult <2 x i8 > < i8 1, i8 2>, < i8 2, i8 2 > <i>; yields: result=<2 x i8> < i8 -1, i8 0 ></i>
4344 <!-- _______________________________________________________________________ -->
4345 <div class="doc_subsubsection">
4346 <a name="i_vfcmp">'<tt>vfcmp</tt>' Instruction</a>
4348 <div class="doc_text">
4350 <pre> <result> = vfcmp <cond> <ty> <op1>, <op2></pre>
4352 <p>The '<tt>vfcmp</tt>' instruction returns an integer vector value based on
4353 element-wise comparison of its two floating point vector operands. The output
4354 elements have the same width as the input elements.</p>
4356 <p>The '<tt>vfcmp</tt>' instruction takes three operands. The first operand is
4357 the condition code indicating the kind of comparison to perform. It is not
4358 a value, just a keyword. The possible condition code are:</p>
4360 <li><tt>false</tt>: no comparison, always returns false</li>
4361 <li><tt>oeq</tt>: ordered and equal</li>
4362 <li><tt>ogt</tt>: ordered and greater than </li>
4363 <li><tt>oge</tt>: ordered and greater than or equal</li>
4364 <li><tt>olt</tt>: ordered and less than </li>
4365 <li><tt>ole</tt>: ordered and less than or equal</li>
4366 <li><tt>one</tt>: ordered and not equal</li>
4367 <li><tt>ord</tt>: ordered (no nans)</li>
4368 <li><tt>ueq</tt>: unordered or equal</li>
4369 <li><tt>ugt</tt>: unordered or greater than </li>
4370 <li><tt>uge</tt>: unordered or greater than or equal</li>
4371 <li><tt>ult</tt>: unordered or less than </li>
4372 <li><tt>ule</tt>: unordered or less than or equal</li>
4373 <li><tt>une</tt>: unordered or not equal</li>
4374 <li><tt>uno</tt>: unordered (either nans)</li>
4375 <li><tt>true</tt>: no comparison, always returns true</li>
4377 <p>The remaining two arguments must be <a href="#t_vector">vector</a> of
4378 <a href="#t_floating">floating point</a> typed. They must also be identical
4381 <p>The '<tt>vfcmp</tt>' instruction compares <tt>op1</tt> and <tt>op2</tt>
4382 according to the condition code given as <tt>cond</tt>. The comparison yields a
4383 <a href="#t_vector">vector</a> of <a href="#t_integer">integer</a> result, with
4384 an identical number of elements as the values being compared, and each element
4385 having identical with to the width of the floating point elements. The most
4386 significant bit in each element is 1 if the element-wise comparison evaluates to
4387 true, and is 0 otherwise. All other bits of the result are undefined. The
4388 condition codes are evaluated identically to the
4389 <a href="#i_fcmp">'<tt>fcmp</tt>' instruction</a>.</p>
4393 <i>; yields: result=<2 x i32> < i32 0, i32 -1 ></i>
4394 <result> = vfcmp oeq <2 x float> < float 4, float 0 >, < float 5, float 0 >
4396 <i>; yields: result=<2 x i64> < i64 -1, i64 0 ></i>
4397 <result> = vfcmp ult <2 x double> < double 1, double 2 >, < double 2, double 2>
4401 <!-- _______________________________________________________________________ -->
4402 <div class="doc_subsubsection">
4403 <a name="i_phi">'<tt>phi</tt>' Instruction</a>
4406 <div class="doc_text">
4410 <pre> <result> = phi <ty> [ <val0>, <label0>], ...<br></pre>
4412 <p>The '<tt>phi</tt>' instruction is used to implement the φ node in
4413 the SSA graph representing the function.</p>
4416 <p>The type of the incoming values is specified with the first type
4417 field. After this, the '<tt>phi</tt>' instruction takes a list of pairs
4418 as arguments, with one pair for each predecessor basic block of the
4419 current block. Only values of <a href="#t_firstclass">first class</a>
4420 type may be used as the value arguments to the PHI node. Only labels
4421 may be used as the label arguments.</p>
4423 <p>There must be no non-phi instructions between the start of a basic
4424 block and the PHI instructions: i.e. PHI instructions must be first in
4429 <p>At runtime, the '<tt>phi</tt>' instruction logically takes on the value
4430 specified by the pair corresponding to the predecessor basic block that executed
4431 just prior to the current block.</p>
4435 Loop: ; Infinite loop that counts from 0 on up...
4436 %indvar = phi i32 [ 0, %LoopHeader ], [ %nextindvar, %Loop ]
4437 %nextindvar = add i32 %indvar, 1
4442 <!-- _______________________________________________________________________ -->
4443 <div class="doc_subsubsection">
4444 <a name="i_select">'<tt>select</tt>' Instruction</a>
4447 <div class="doc_text">
4452 <result> = select <i>selty</i> <cond>, <ty> <val1>, <ty> <val2> <i>; yields ty</i>
4454 <i>selty</i> is either i1 or {<N x i1>}
4460 The '<tt>select</tt>' instruction is used to choose one value based on a
4461 condition, without branching.
4468 The '<tt>select</tt>' instruction requires an 'i1' value or
4469 a vector of 'i1' values indicating the
4470 condition, and two values of the same <a href="#t_firstclass">first class</a>
4471 type. If the val1/val2 are vectors and
4472 the condition is a scalar, then entire vectors are selected, not
4473 individual elements.
4479 If the condition is an i1 and it evaluates to 1, the instruction returns the first
4480 value argument; otherwise, it returns the second value argument.
4483 If the condition is a vector of i1, then the value arguments must
4484 be vectors of the same size, and the selection is done element
4491 %X = select i1 true, i8 17, i8 42 <i>; yields i8:17</i>
4494 <p>Note that the code generator does not yet support conditions
4495 with vector type.</p>
4500 <!-- _______________________________________________________________________ -->
4501 <div class="doc_subsubsection">
4502 <a name="i_call">'<tt>call</tt>' Instruction</a>
4505 <div class="doc_text">
4509 <result> = [tail] call [<a href="#callingconv">cconv</a>] [<a href="#paramattrs">ret attrs</a>] <ty> [<fnty>*] <fnptrval>(<function args>) [<a href="#fnattrs">fn attrs</a>]
4514 <p>The '<tt>call</tt>' instruction represents a simple function call.</p>
4518 <p>This instruction requires several arguments:</p>
4522 <p>The optional "tail" marker indicates whether the callee function accesses
4523 any allocas or varargs in the caller. If the "tail" marker is present, the
4524 function call is eligible for tail call optimization. Note that calls may
4525 be marked "tail" even if they do not occur before a <a
4526 href="#i_ret"><tt>ret</tt></a> instruction.</p>
4529 <p>The optional "cconv" marker indicates which <a href="#callingconv">calling
4530 convention</a> the call should use. If none is specified, the call defaults
4531 to using C calling conventions.</p>
4535 <p>The optional <a href="#paramattrs">Parameter Attributes</a> list for
4536 return values. Only '<tt>zeroext</tt>', '<tt>signext</tt>',
4537 and '<tt>inreg</tt>' attributes are valid here.</p>
4541 <p>'<tt>ty</tt>': the type of the call instruction itself which is also
4542 the type of the return value. Functions that return no value are marked
4543 <tt><a href="#t_void">void</a></tt>.</p>
4546 <p>'<tt>fnty</tt>': shall be the signature of the pointer to function
4547 value being invoked. The argument types must match the types implied by
4548 this signature. This type can be omitted if the function is not varargs
4549 and if the function type does not return a pointer to a function.</p>
4552 <p>'<tt>fnptrval</tt>': An LLVM value containing a pointer to a function to
4553 be invoked. In most cases, this is a direct function invocation, but
4554 indirect <tt>call</tt>s are just as possible, calling an arbitrary pointer
4555 to function value.</p>
4558 <p>'<tt>function args</tt>': argument list whose types match the
4559 function signature argument types. All arguments must be of
4560 <a href="#t_firstclass">first class</a> type. If the function signature
4561 indicates the function accepts a variable number of arguments, the extra
4562 arguments can be specified.</p>
4565 <p>The optional <a href="#fnattrs">function attributes</a> list. Only
4566 '<tt>noreturn</tt>', '<tt>nounwind</tt>', '<tt>readonly</tt>' and
4567 '<tt>readnone</tt>' attributes are valid here.</p>
4573 <p>The '<tt>call</tt>' instruction is used to cause control flow to
4574 transfer to a specified function, with its incoming arguments bound to
4575 the specified values. Upon a '<tt><a href="#i_ret">ret</a></tt>'
4576 instruction in the called function, control flow continues with the
4577 instruction after the function call, and the return value of the
4578 function is bound to the result argument.</p>
4583 %retval = call i32 @test(i32 %argc)
4584 call i32 (i8 *, ...)* @printf(i8 * %msg, i32 12, i8 42) <i>; yields i32</i>
4585 %X = tail call i32 @foo() <i>; yields i32</i>
4586 %Y = tail call <a href="#callingconv">fastcc</a> i32 @foo() <i>; yields i32</i>
4587 call void %foo(i8 97 signext)
4589 %struct.A = type { i32, i8 }
4590 %r = call %struct.A @foo() <i>; yields { 32, i8 }</i>
4591 %gr = extractvalue %struct.A %r, 0 <i>; yields i32</i>
4592 %gr1 = extractvalue %struct.A %r, 1 <i>; yields i8</i>
4593 %Z = call void @foo() noreturn <i>; indicates that %foo never returns normally</i>
4594 %ZZ = call zeroext i32 @bar() <i>; Return value is %zero extended</i>
4599 <!-- _______________________________________________________________________ -->
4600 <div class="doc_subsubsection">
4601 <a name="i_va_arg">'<tt>va_arg</tt>' Instruction</a>
4604 <div class="doc_text">
4609 <resultval> = va_arg <va_list*> <arglist>, <argty>
4614 <p>The '<tt>va_arg</tt>' instruction is used to access arguments passed through
4615 the "variable argument" area of a function call. It is used to implement the
4616 <tt>va_arg</tt> macro in C.</p>
4620 <p>This instruction takes a <tt>va_list*</tt> value and the type of
4621 the argument. It returns a value of the specified argument type and
4622 increments the <tt>va_list</tt> to point to the next argument. The
4623 actual type of <tt>va_list</tt> is target specific.</p>
4627 <p>The '<tt>va_arg</tt>' instruction loads an argument of the specified
4628 type from the specified <tt>va_list</tt> and causes the
4629 <tt>va_list</tt> to point to the next argument. For more information,
4630 see the variable argument handling <a href="#int_varargs">Intrinsic
4633 <p>It is legal for this instruction to be called in a function which does not
4634 take a variable number of arguments, for example, the <tt>vfprintf</tt>
4637 <p><tt>va_arg</tt> is an LLVM instruction instead of an <a
4638 href="#intrinsics">intrinsic function</a> because it takes a type as an
4643 <p>See the <a href="#int_varargs">variable argument processing</a> section.</p>
4645 <p>Note that the code generator does not yet fully support va_arg
4646 on many targets. Also, it does not currently support va_arg with
4647 aggregate types on any target.</p>
4651 <!-- *********************************************************************** -->
4652 <div class="doc_section"> <a name="intrinsics">Intrinsic Functions</a> </div>
4653 <!-- *********************************************************************** -->
4655 <div class="doc_text">
4657 <p>LLVM supports the notion of an "intrinsic function". These functions have
4658 well known names and semantics and are required to follow certain restrictions.
4659 Overall, these intrinsics represent an extension mechanism for the LLVM
4660 language that does not require changing all of the transformations in LLVM when
4661 adding to the language (or the bitcode reader/writer, the parser, etc...).</p>
4663 <p>Intrinsic function names must all start with an "<tt>llvm.</tt>" prefix. This
4664 prefix is reserved in LLVM for intrinsic names; thus, function names may not
4665 begin with this prefix. Intrinsic functions must always be external functions:
4666 you cannot define the body of intrinsic functions. Intrinsic functions may
4667 only be used in call or invoke instructions: it is illegal to take the address
4668 of an intrinsic function. Additionally, because intrinsic functions are part
4669 of the LLVM language, it is required if any are added that they be documented
4672 <p>Some intrinsic functions can be overloaded, i.e., the intrinsic represents
4673 a family of functions that perform the same operation but on different data
4674 types. Because LLVM can represent over 8 million different integer types,
4675 overloading is used commonly to allow an intrinsic function to operate on any
4676 integer type. One or more of the argument types or the result type can be
4677 overloaded to accept any integer type. Argument types may also be defined as
4678 exactly matching a previous argument's type or the result type. This allows an
4679 intrinsic function which accepts multiple arguments, but needs all of them to
4680 be of the same type, to only be overloaded with respect to a single argument or
4683 <p>Overloaded intrinsics will have the names of its overloaded argument types
4684 encoded into its function name, each preceded by a period. Only those types
4685 which are overloaded result in a name suffix. Arguments whose type is matched
4686 against another type do not. For example, the <tt>llvm.ctpop</tt> function can
4687 take an integer of any width and returns an integer of exactly the same integer
4688 width. This leads to a family of functions such as
4689 <tt>i8 @llvm.ctpop.i8(i8 %val)</tt> and <tt>i29 @llvm.ctpop.i29(i29 %val)</tt>.
4690 Only one type, the return type, is overloaded, and only one type suffix is
4691 required. Because the argument's type is matched against the return type, it
4692 does not require its own name suffix.</p>
4694 <p>To learn how to add an intrinsic function, please see the
4695 <a href="ExtendingLLVM.html">Extending LLVM Guide</a>.
4700 <!-- ======================================================================= -->
4701 <div class="doc_subsection">
4702 <a name="int_varargs">Variable Argument Handling Intrinsics</a>
4705 <div class="doc_text">
4707 <p>Variable argument support is defined in LLVM with the <a
4708 href="#i_va_arg"><tt>va_arg</tt></a> instruction and these three
4709 intrinsic functions. These functions are related to the similarly
4710 named macros defined in the <tt><stdarg.h></tt> header file.</p>
4712 <p>All of these functions operate on arguments that use a
4713 target-specific value type "<tt>va_list</tt>". The LLVM assembly
4714 language reference manual does not define what this type is, so all
4715 transformations should be prepared to handle these functions regardless of
4718 <p>This example shows how the <a href="#i_va_arg"><tt>va_arg</tt></a>
4719 instruction and the variable argument handling intrinsic functions are
4722 <div class="doc_code">
4724 define i32 @test(i32 %X, ...) {
4725 ; Initialize variable argument processing
4727 %ap2 = bitcast i8** %ap to i8*
4728 call void @llvm.va_start(i8* %ap2)
4730 ; Read a single integer argument
4731 %tmp = va_arg i8** %ap, i32
4733 ; Demonstrate usage of llvm.va_copy and llvm.va_end
4735 %aq2 = bitcast i8** %aq to i8*
4736 call void @llvm.va_copy(i8* %aq2, i8* %ap2)
4737 call void @llvm.va_end(i8* %aq2)
4739 ; Stop processing of arguments.
4740 call void @llvm.va_end(i8* %ap2)
4744 declare void @llvm.va_start(i8*)
4745 declare void @llvm.va_copy(i8*, i8*)
4746 declare void @llvm.va_end(i8*)
4752 <!-- _______________________________________________________________________ -->
4753 <div class="doc_subsubsection">
4754 <a name="int_va_start">'<tt>llvm.va_start</tt>' Intrinsic</a>
4758 <div class="doc_text">
4760 <pre> declare void %llvm.va_start(i8* <arglist>)<br></pre>
4762 <p>The '<tt>llvm.va_start</tt>' intrinsic initializes
4763 <tt>*<arglist></tt> for subsequent use by <tt><a
4764 href="#i_va_arg">va_arg</a></tt>.</p>
4768 <p>The argument is a pointer to a <tt>va_list</tt> element to initialize.</p>
4772 <p>The '<tt>llvm.va_start</tt>' intrinsic works just like the <tt>va_start</tt>
4773 macro available in C. In a target-dependent way, it initializes the
4774 <tt>va_list</tt> element to which the argument points, so that the next call to
4775 <tt>va_arg</tt> will produce the first variable argument passed to the function.
4776 Unlike the C <tt>va_start</tt> macro, this intrinsic does not need to know the
4777 last argument of the function as the compiler can figure that out.</p>
4781 <!-- _______________________________________________________________________ -->
4782 <div class="doc_subsubsection">
4783 <a name="int_va_end">'<tt>llvm.va_end</tt>' Intrinsic</a>
4786 <div class="doc_text">
4788 <pre> declare void @llvm.va_end(i8* <arglist>)<br></pre>
4791 <p>The '<tt>llvm.va_end</tt>' intrinsic destroys <tt>*<arglist></tt>,
4792 which has been initialized previously with <tt><a href="#int_va_start">llvm.va_start</a></tt>
4793 or <tt><a href="#i_va_copy">llvm.va_copy</a></tt>.</p>
4797 <p>The argument is a pointer to a <tt>va_list</tt> to destroy.</p>
4801 <p>The '<tt>llvm.va_end</tt>' intrinsic works just like the <tt>va_end</tt>
4802 macro available in C. In a target-dependent way, it destroys the
4803 <tt>va_list</tt> element to which the argument points. Calls to <a
4804 href="#int_va_start"><tt>llvm.va_start</tt></a> and <a href="#int_va_copy">
4805 <tt>llvm.va_copy</tt></a> must be matched exactly with calls to
4806 <tt>llvm.va_end</tt>.</p>
4810 <!-- _______________________________________________________________________ -->
4811 <div class="doc_subsubsection">
4812 <a name="int_va_copy">'<tt>llvm.va_copy</tt>' Intrinsic</a>
4815 <div class="doc_text">
4820 declare void @llvm.va_copy(i8* <destarglist>, i8* <srcarglist>)
4825 <p>The '<tt>llvm.va_copy</tt>' intrinsic copies the current argument position
4826 from the source argument list to the destination argument list.</p>
4830 <p>The first argument is a pointer to a <tt>va_list</tt> element to initialize.
4831 The second argument is a pointer to a <tt>va_list</tt> element to copy from.</p>
4836 <p>The '<tt>llvm.va_copy</tt>' intrinsic works just like the <tt>va_copy</tt>
4837 macro available in C. In a target-dependent way, it copies the source
4838 <tt>va_list</tt> element into the destination <tt>va_list</tt> element. This
4839 intrinsic is necessary because the <tt><a href="#int_va_start">
4840 llvm.va_start</a></tt> intrinsic may be arbitrarily complex and require, for
4841 example, memory allocation.</p>
4845 <!-- ======================================================================= -->
4846 <div class="doc_subsection">
4847 <a name="int_gc">Accurate Garbage Collection Intrinsics</a>
4850 <div class="doc_text">
4853 LLVM support for <a href="GarbageCollection.html">Accurate Garbage
4854 Collection</a> (GC) requires the implementation and generation of these
4856 These intrinsics allow identification of <a href="#int_gcroot">GC roots on the
4857 stack</a>, as well as garbage collector implementations that require <a
4858 href="#int_gcread">read</a> and <a href="#int_gcwrite">write</a> barriers.
4859 Front-ends for type-safe garbage collected languages should generate these
4860 intrinsics to make use of the LLVM garbage collectors. For more details, see <a
4861 href="GarbageCollection.html">Accurate Garbage Collection with LLVM</a>.
4864 <p>The garbage collection intrinsics only operate on objects in the generic
4865 address space (address space zero).</p>
4869 <!-- _______________________________________________________________________ -->
4870 <div class="doc_subsubsection">
4871 <a name="int_gcroot">'<tt>llvm.gcroot</tt>' Intrinsic</a>
4874 <div class="doc_text">
4879 declare void @llvm.gcroot(i8** %ptrloc, i8* %metadata)
4884 <p>The '<tt>llvm.gcroot</tt>' intrinsic declares the existence of a GC root to
4885 the code generator, and allows some metadata to be associated with it.</p>
4889 <p>The first argument specifies the address of a stack object that contains the
4890 root pointer. The second pointer (which must be either a constant or a global
4891 value address) contains the meta-data to be associated with the root.</p>
4895 <p>At runtime, a call to this intrinsic stores a null pointer into the "ptrloc"
4896 location. At compile-time, the code generator generates information to allow
4897 the runtime to find the pointer at GC safe points. The '<tt>llvm.gcroot</tt>'
4898 intrinsic may only be used in a function which <a href="#gc">specifies a GC
4904 <!-- _______________________________________________________________________ -->
4905 <div class="doc_subsubsection">
4906 <a name="int_gcread">'<tt>llvm.gcread</tt>' Intrinsic</a>
4909 <div class="doc_text">
4914 declare i8* @llvm.gcread(i8* %ObjPtr, i8** %Ptr)
4919 <p>The '<tt>llvm.gcread</tt>' intrinsic identifies reads of references from heap
4920 locations, allowing garbage collector implementations that require read
4925 <p>The second argument is the address to read from, which should be an address
4926 allocated from the garbage collector. The first object is a pointer to the
4927 start of the referenced object, if needed by the language runtime (otherwise
4932 <p>The '<tt>llvm.gcread</tt>' intrinsic has the same semantics as a load
4933 instruction, but may be replaced with substantially more complex code by the
4934 garbage collector runtime, as needed. The '<tt>llvm.gcread</tt>' intrinsic
4935 may only be used in a function which <a href="#gc">specifies a GC
4941 <!-- _______________________________________________________________________ -->
4942 <div class="doc_subsubsection">
4943 <a name="int_gcwrite">'<tt>llvm.gcwrite</tt>' Intrinsic</a>
4946 <div class="doc_text">
4951 declare void @llvm.gcwrite(i8* %P1, i8* %Obj, i8** %P2)
4956 <p>The '<tt>llvm.gcwrite</tt>' intrinsic identifies writes of references to heap
4957 locations, allowing garbage collector implementations that require write
4958 barriers (such as generational or reference counting collectors).</p>
4962 <p>The first argument is the reference to store, the second is the start of the
4963 object to store it to, and the third is the address of the field of Obj to
4964 store to. If the runtime does not require a pointer to the object, Obj may be
4969 <p>The '<tt>llvm.gcwrite</tt>' intrinsic has the same semantics as a store
4970 instruction, but may be replaced with substantially more complex code by the
4971 garbage collector runtime, as needed. The '<tt>llvm.gcwrite</tt>' intrinsic
4972 may only be used in a function which <a href="#gc">specifies a GC
4979 <!-- ======================================================================= -->
4980 <div class="doc_subsection">
4981 <a name="int_codegen">Code Generator Intrinsics</a>
4984 <div class="doc_text">
4986 These intrinsics are provided by LLVM to expose special features that may only
4987 be implemented with code generator support.
4992 <!-- _______________________________________________________________________ -->
4993 <div class="doc_subsubsection">
4994 <a name="int_returnaddress">'<tt>llvm.returnaddress</tt>' Intrinsic</a>
4997 <div class="doc_text">
5001 declare i8 *@llvm.returnaddress(i32 <level>)
5007 The '<tt>llvm.returnaddress</tt>' intrinsic attempts to compute a
5008 target-specific value indicating the return address of the current function
5009 or one of its callers.
5015 The argument to this intrinsic indicates which function to return the address
5016 for. Zero indicates the calling function, one indicates its caller, etc. The
5017 argument is <b>required</b> to be a constant integer value.
5023 The '<tt>llvm.returnaddress</tt>' intrinsic either returns a pointer indicating
5024 the return address of the specified call frame, or zero if it cannot be
5025 identified. The value returned by this intrinsic is likely to be incorrect or 0
5026 for arguments other than zero, so it should only be used for debugging purposes.
5030 Note that calling this intrinsic does not prevent function inlining or other
5031 aggressive transformations, so the value returned may not be that of the obvious
5032 source-language caller.
5037 <!-- _______________________________________________________________________ -->
5038 <div class="doc_subsubsection">
5039 <a name="int_frameaddress">'<tt>llvm.frameaddress</tt>' Intrinsic</a>
5042 <div class="doc_text">
5046 declare i8 *@llvm.frameaddress(i32 <level>)
5052 The '<tt>llvm.frameaddress</tt>' intrinsic attempts to return the
5053 target-specific frame pointer value for the specified stack frame.
5059 The argument to this intrinsic indicates which function to return the frame
5060 pointer for. Zero indicates the calling function, one indicates its caller,
5061 etc. The argument is <b>required</b> to be a constant integer value.
5067 The '<tt>llvm.frameaddress</tt>' intrinsic either returns a pointer indicating
5068 the frame address of the specified call frame, or zero if it cannot be
5069 identified. The value returned by this intrinsic is likely to be incorrect or 0
5070 for arguments other than zero, so it should only be used for debugging purposes.
5074 Note that calling this intrinsic does not prevent function inlining or other
5075 aggressive transformations, so the value returned may not be that of the obvious
5076 source-language caller.
5080 <!-- _______________________________________________________________________ -->
5081 <div class="doc_subsubsection">
5082 <a name="int_stacksave">'<tt>llvm.stacksave</tt>' Intrinsic</a>
5085 <div class="doc_text">
5089 declare i8 *@llvm.stacksave()
5095 The '<tt>llvm.stacksave</tt>' intrinsic is used to remember the current state of
5096 the function stack, for use with <a href="#int_stackrestore">
5097 <tt>llvm.stackrestore</tt></a>. This is useful for implementing language
5098 features like scoped automatic variable sized arrays in C99.
5104 This intrinsic returns a opaque pointer value that can be passed to <a
5105 href="#int_stackrestore"><tt>llvm.stackrestore</tt></a>. When an
5106 <tt>llvm.stackrestore</tt> intrinsic is executed with a value saved from
5107 <tt>llvm.stacksave</tt>, it effectively restores the state of the stack to the
5108 state it was in when the <tt>llvm.stacksave</tt> intrinsic executed. In
5109 practice, this pops any <a href="#i_alloca">alloca</a> blocks from the stack
5110 that were allocated after the <tt>llvm.stacksave</tt> was executed.
5115 <!-- _______________________________________________________________________ -->
5116 <div class="doc_subsubsection">
5117 <a name="int_stackrestore">'<tt>llvm.stackrestore</tt>' Intrinsic</a>
5120 <div class="doc_text">
5124 declare void @llvm.stackrestore(i8 * %ptr)
5130 The '<tt>llvm.stackrestore</tt>' intrinsic is used to restore the state of
5131 the function stack to the state it was in when the corresponding <a
5132 href="#int_stacksave"><tt>llvm.stacksave</tt></a> intrinsic executed. This is
5133 useful for implementing language features like scoped automatic variable sized
5140 See the description for <a href="#int_stacksave"><tt>llvm.stacksave</tt></a>.
5146 <!-- _______________________________________________________________________ -->
5147 <div class="doc_subsubsection">
5148 <a name="int_prefetch">'<tt>llvm.prefetch</tt>' Intrinsic</a>
5151 <div class="doc_text">
5155 declare void @llvm.prefetch(i8* <address>, i32 <rw>, i32 <locality>)
5162 The '<tt>llvm.prefetch</tt>' intrinsic is a hint to the code generator to insert
5163 a prefetch instruction if supported; otherwise, it is a noop. Prefetches have
5165 effect on the behavior of the program but can change its performance
5172 <tt>address</tt> is the address to be prefetched, <tt>rw</tt> is the specifier
5173 determining if the fetch should be for a read (0) or write (1), and
5174 <tt>locality</tt> is a temporal locality specifier ranging from (0) - no
5175 locality, to (3) - extremely local keep in cache. The <tt>rw</tt> and
5176 <tt>locality</tt> arguments must be constant integers.
5182 This intrinsic does not modify the behavior of the program. In particular,
5183 prefetches cannot trap and do not produce a value. On targets that support this
5184 intrinsic, the prefetch can provide hints to the processor cache for better
5190 <!-- _______________________________________________________________________ -->
5191 <div class="doc_subsubsection">
5192 <a name="int_pcmarker">'<tt>llvm.pcmarker</tt>' Intrinsic</a>
5195 <div class="doc_text">
5199 declare void @llvm.pcmarker(i32 <id>)
5206 The '<tt>llvm.pcmarker</tt>' intrinsic is a method to export a Program Counter
5208 code to simulators and other tools. The method is target specific, but it is
5209 expected that the marker will use exported symbols to transmit the PC of the
5211 The marker makes no guarantees that it will remain with any specific instruction
5212 after optimizations. It is possible that the presence of a marker will inhibit
5213 optimizations. The intended use is to be inserted after optimizations to allow
5214 correlations of simulation runs.
5220 <tt>id</tt> is a numerical id identifying the marker.
5226 This intrinsic does not modify the behavior of the program. Backends that do not
5227 support this intrinisic may ignore it.
5232 <!-- _______________________________________________________________________ -->
5233 <div class="doc_subsubsection">
5234 <a name="int_readcyclecounter">'<tt>llvm.readcyclecounter</tt>' Intrinsic</a>
5237 <div class="doc_text">
5241 declare i64 @llvm.readcyclecounter( )
5248 The '<tt>llvm.readcyclecounter</tt>' intrinsic provides access to the cycle
5249 counter register (or similar low latency, high accuracy clocks) on those targets
5250 that support it. On X86, it should map to RDTSC. On Alpha, it should map to RPCC.
5251 As the backing counters overflow quickly (on the order of 9 seconds on alpha), this
5252 should only be used for small timings.
5258 When directly supported, reading the cycle counter should not modify any memory.
5259 Implementations are allowed to either return a application specific value or a
5260 system wide value. On backends without support, this is lowered to a constant 0.
5265 <!-- ======================================================================= -->
5266 <div class="doc_subsection">
5267 <a name="int_libc">Standard C Library Intrinsics</a>
5270 <div class="doc_text">
5272 LLVM provides intrinsics for a few important standard C library functions.
5273 These intrinsics allow source-language front-ends to pass information about the
5274 alignment of the pointer arguments to the code generator, providing opportunity
5275 for more efficient code generation.
5280 <!-- _______________________________________________________________________ -->
5281 <div class="doc_subsubsection">
5282 <a name="int_memcpy">'<tt>llvm.memcpy</tt>' Intrinsic</a>
5285 <div class="doc_text">
5288 <p>This is an overloaded intrinsic. You can use llvm.memcpy on any integer bit
5289 width. Not all targets support all bit widths however.</p>
5291 declare void @llvm.memcpy.i8(i8 * <dest>, i8 * <src>,
5292 i8 <len>, i32 <align>)
5293 declare void @llvm.memcpy.i16(i8 * <dest>, i8 * <src>,
5294 i16 <len>, i32 <align>)
5295 declare void @llvm.memcpy.i32(i8 * <dest>, i8 * <src>,
5296 i32 <len>, i32 <align>)
5297 declare void @llvm.memcpy.i64(i8 * <dest>, i8 * <src>,
5298 i64 <len>, i32 <align>)
5304 The '<tt>llvm.memcpy.*</tt>' intrinsics copy a block of memory from the source
5305 location to the destination location.
5309 Note that, unlike the standard libc function, the <tt>llvm.memcpy.*</tt>
5310 intrinsics do not return a value, and takes an extra alignment argument.
5316 The first argument is a pointer to the destination, the second is a pointer to
5317 the source. The third argument is an integer argument
5318 specifying the number of bytes to copy, and the fourth argument is the alignment
5319 of the source and destination locations.
5323 If the call to this intrinisic has an alignment value that is not 0 or 1, then
5324 the caller guarantees that both the source and destination pointers are aligned
5331 The '<tt>llvm.memcpy.*</tt>' intrinsics copy a block of memory from the source
5332 location to the destination location, which are not allowed to overlap. It
5333 copies "len" bytes of memory over. If the argument is known to be aligned to
5334 some boundary, this can be specified as the fourth argument, otherwise it should
5340 <!-- _______________________________________________________________________ -->
5341 <div class="doc_subsubsection">
5342 <a name="int_memmove">'<tt>llvm.memmove</tt>' Intrinsic</a>
5345 <div class="doc_text">
5348 <p>This is an overloaded intrinsic. You can use llvm.memmove on any integer bit
5349 width. Not all targets support all bit widths however.</p>
5351 declare void @llvm.memmove.i8(i8 * <dest>, i8 * <src>,
5352 i8 <len>, i32 <align>)
5353 declare void @llvm.memmove.i16(i8 * <dest>, i8 * <src>,
5354 i16 <len>, i32 <align>)
5355 declare void @llvm.memmove.i32(i8 * <dest>, i8 * <src>,
5356 i32 <len>, i32 <align>)
5357 declare void @llvm.memmove.i64(i8 * <dest>, i8 * <src>,
5358 i64 <len>, i32 <align>)
5364 The '<tt>llvm.memmove.*</tt>' intrinsics move a block of memory from the source
5365 location to the destination location. It is similar to the
5366 '<tt>llvm.memcpy</tt>' intrinsic but allows the two memory locations to overlap.
5370 Note that, unlike the standard libc function, the <tt>llvm.memmove.*</tt>
5371 intrinsics do not return a value, and takes an extra alignment argument.
5377 The first argument is a pointer to the destination, the second is a pointer to
5378 the source. The third argument is an integer argument
5379 specifying the number of bytes to copy, and the fourth argument is the alignment
5380 of the source and destination locations.
5384 If the call to this intrinisic has an alignment value that is not 0 or 1, then
5385 the caller guarantees that the source and destination pointers are aligned to
5392 The '<tt>llvm.memmove.*</tt>' intrinsics copy a block of memory from the source
5393 location to the destination location, which may overlap. It
5394 copies "len" bytes of memory over. If the argument is known to be aligned to
5395 some boundary, this can be specified as the fourth argument, otherwise it should
5401 <!-- _______________________________________________________________________ -->
5402 <div class="doc_subsubsection">
5403 <a name="int_memset">'<tt>llvm.memset.*</tt>' Intrinsics</a>
5406 <div class="doc_text">
5409 <p>This is an overloaded intrinsic. You can use llvm.memset on any integer bit
5410 width. Not all targets support all bit widths however.</p>
5412 declare void @llvm.memset.i8(i8 * <dest>, i8 <val>,
5413 i8 <len>, i32 <align>)
5414 declare void @llvm.memset.i16(i8 * <dest>, i8 <val>,
5415 i16 <len>, i32 <align>)
5416 declare void @llvm.memset.i32(i8 * <dest>, i8 <val>,
5417 i32 <len>, i32 <align>)
5418 declare void @llvm.memset.i64(i8 * <dest>, i8 <val>,
5419 i64 <len>, i32 <align>)
5425 The '<tt>llvm.memset.*</tt>' intrinsics fill a block of memory with a particular
5430 Note that, unlike the standard libc function, the <tt>llvm.memset</tt> intrinsic
5431 does not return a value, and takes an extra alignment argument.
5437 The first argument is a pointer to the destination to fill, the second is the
5438 byte value to fill it with, the third argument is an integer
5439 argument specifying the number of bytes to fill, and the fourth argument is the
5440 known alignment of destination location.
5444 If the call to this intrinisic has an alignment value that is not 0 or 1, then
5445 the caller guarantees that the destination pointer is aligned to that boundary.
5451 The '<tt>llvm.memset.*</tt>' intrinsics fill "len" bytes of memory starting at
5453 destination location. If the argument is known to be aligned to some boundary,
5454 this can be specified as the fourth argument, otherwise it should be set to 0 or
5460 <!-- _______________________________________________________________________ -->
5461 <div class="doc_subsubsection">
5462 <a name="int_sqrt">'<tt>llvm.sqrt.*</tt>' Intrinsic</a>
5465 <div class="doc_text">
5468 <p>This is an overloaded intrinsic. You can use <tt>llvm.sqrt</tt> on any
5469 floating point or vector of floating point type. Not all targets support all
5472 declare float @llvm.sqrt.f32(float %Val)
5473 declare double @llvm.sqrt.f64(double %Val)
5474 declare x86_fp80 @llvm.sqrt.f80(x86_fp80 %Val)
5475 declare fp128 @llvm.sqrt.f128(fp128 %Val)
5476 declare ppc_fp128 @llvm.sqrt.ppcf128(ppc_fp128 %Val)
5482 The '<tt>llvm.sqrt</tt>' intrinsics return the sqrt of the specified operand,
5483 returning the same value as the libm '<tt>sqrt</tt>' functions would. Unlike
5484 <tt>sqrt</tt> in libm, however, <tt>llvm.sqrt</tt> has undefined behavior for
5485 negative numbers other than -0.0 (which allows for better optimization, because
5486 there is no need to worry about errno being set). <tt>llvm.sqrt(-0.0)</tt> is
5487 defined to return -0.0 like IEEE sqrt.
5493 The argument and return value are floating point numbers of the same type.
5499 This function returns the sqrt of the specified operand if it is a nonnegative
5500 floating point number.
5504 <!-- _______________________________________________________________________ -->
5505 <div class="doc_subsubsection">
5506 <a name="int_powi">'<tt>llvm.powi.*</tt>' Intrinsic</a>
5509 <div class="doc_text">
5512 <p>This is an overloaded intrinsic. You can use <tt>llvm.powi</tt> on any
5513 floating point or vector of floating point type. Not all targets support all
5516 declare float @llvm.powi.f32(float %Val, i32 %power)
5517 declare double @llvm.powi.f64(double %Val, i32 %power)
5518 declare x86_fp80 @llvm.powi.f80(x86_fp80 %Val, i32 %power)
5519 declare fp128 @llvm.powi.f128(fp128 %Val, i32 %power)
5520 declare ppc_fp128 @llvm.powi.ppcf128(ppc_fp128 %Val, i32 %power)
5526 The '<tt>llvm.powi.*</tt>' intrinsics return the first operand raised to the
5527 specified (positive or negative) power. The order of evaluation of
5528 multiplications is not defined. When a vector of floating point type is
5529 used, the second argument remains a scalar integer value.
5535 The second argument is an integer power, and the first is a value to raise to
5542 This function returns the first value raised to the second power with an
5543 unspecified sequence of rounding operations.</p>
5546 <!-- _______________________________________________________________________ -->
5547 <div class="doc_subsubsection">
5548 <a name="int_sin">'<tt>llvm.sin.*</tt>' Intrinsic</a>
5551 <div class="doc_text">
5554 <p>This is an overloaded intrinsic. You can use <tt>llvm.sin</tt> on any
5555 floating point or vector of floating point type. Not all targets support all
5558 declare float @llvm.sin.f32(float %Val)
5559 declare double @llvm.sin.f64(double %Val)
5560 declare x86_fp80 @llvm.sin.f80(x86_fp80 %Val)
5561 declare fp128 @llvm.sin.f128(fp128 %Val)
5562 declare ppc_fp128 @llvm.sin.ppcf128(ppc_fp128 %Val)
5568 The '<tt>llvm.sin.*</tt>' intrinsics return the sine of the operand.
5574 The argument and return value are floating point numbers of the same type.
5580 This function returns the sine of the specified operand, returning the
5581 same values as the libm <tt>sin</tt> functions would, and handles error
5582 conditions in the same way.</p>
5585 <!-- _______________________________________________________________________ -->
5586 <div class="doc_subsubsection">
5587 <a name="int_cos">'<tt>llvm.cos.*</tt>' Intrinsic</a>
5590 <div class="doc_text">
5593 <p>This is an overloaded intrinsic. You can use <tt>llvm.cos</tt> on any
5594 floating point or vector of floating point type. Not all targets support all
5597 declare float @llvm.cos.f32(float %Val)
5598 declare double @llvm.cos.f64(double %Val)
5599 declare x86_fp80 @llvm.cos.f80(x86_fp80 %Val)
5600 declare fp128 @llvm.cos.f128(fp128 %Val)
5601 declare ppc_fp128 @llvm.cos.ppcf128(ppc_fp128 %Val)
5607 The '<tt>llvm.cos.*</tt>' intrinsics return the cosine of the operand.
5613 The argument and return value are floating point numbers of the same type.
5619 This function returns the cosine of the specified operand, returning the
5620 same values as the libm <tt>cos</tt> functions would, and handles error
5621 conditions in the same way.</p>
5624 <!-- _______________________________________________________________________ -->
5625 <div class="doc_subsubsection">
5626 <a name="int_pow">'<tt>llvm.pow.*</tt>' Intrinsic</a>
5629 <div class="doc_text">
5632 <p>This is an overloaded intrinsic. You can use <tt>llvm.pow</tt> on any
5633 floating point or vector of floating point type. Not all targets support all
5636 declare float @llvm.pow.f32(float %Val, float %Power)
5637 declare double @llvm.pow.f64(double %Val, double %Power)
5638 declare x86_fp80 @llvm.pow.f80(x86_fp80 %Val, x86_fp80 %Power)
5639 declare fp128 @llvm.pow.f128(fp128 %Val, fp128 %Power)
5640 declare ppc_fp128 @llvm.pow.ppcf128(ppc_fp128 %Val, ppc_fp128 Power)
5646 The '<tt>llvm.pow.*</tt>' intrinsics return the first operand raised to the
5647 specified (positive or negative) power.
5653 The second argument is a floating point power, and the first is a value to
5654 raise to that power.
5660 This function returns the first value raised to the second power,
5662 same values as the libm <tt>pow</tt> functions would, and handles error
5663 conditions in the same way.</p>
5667 <!-- ======================================================================= -->
5668 <div class="doc_subsection">
5669 <a name="int_manip">Bit Manipulation Intrinsics</a>
5672 <div class="doc_text">
5674 LLVM provides intrinsics for a few important bit manipulation operations.
5675 These allow efficient code generation for some algorithms.
5680 <!-- _______________________________________________________________________ -->
5681 <div class="doc_subsubsection">
5682 <a name="int_bswap">'<tt>llvm.bswap.*</tt>' Intrinsics</a>
5685 <div class="doc_text">
5688 <p>This is an overloaded intrinsic function. You can use bswap on any integer
5689 type that is an even number of bytes (i.e. BitWidth % 16 == 0).</p>
5691 declare i16 @llvm.bswap.i16(i16 <id>)
5692 declare i32 @llvm.bswap.i32(i32 <id>)
5693 declare i64 @llvm.bswap.i64(i64 <id>)
5699 The '<tt>llvm.bswap</tt>' family of intrinsics is used to byte swap integer
5700 values with an even number of bytes (positive multiple of 16 bits). These are
5701 useful for performing operations on data that is not in the target's native
5708 The <tt>llvm.bswap.i16</tt> intrinsic returns an i16 value that has the high
5709 and low byte of the input i16 swapped. Similarly, the <tt>llvm.bswap.i32</tt>
5710 intrinsic returns an i32 value that has the four bytes of the input i32
5711 swapped, so that if the input bytes are numbered 0, 1, 2, 3 then the returned
5712 i32 will have its bytes in 3, 2, 1, 0 order. The <tt>llvm.bswap.i48</tt>,
5713 <tt>llvm.bswap.i64</tt> and other intrinsics extend this concept to
5714 additional even-byte lengths (6 bytes, 8 bytes and more, respectively).
5719 <!-- _______________________________________________________________________ -->
5720 <div class="doc_subsubsection">
5721 <a name="int_ctpop">'<tt>llvm.ctpop.*</tt>' Intrinsic</a>
5724 <div class="doc_text">
5727 <p>This is an overloaded intrinsic. You can use llvm.ctpop on any integer bit
5728 width. Not all targets support all bit widths however.</p>
5730 declare i8 @llvm.ctpop.i8(i8 <src>)
5731 declare i16 @llvm.ctpop.i16(i16 <src>)
5732 declare i32 @llvm.ctpop.i32(i32 <src>)
5733 declare i64 @llvm.ctpop.i64(i64 <src>)
5734 declare i256 @llvm.ctpop.i256(i256 <src>)
5740 The '<tt>llvm.ctpop</tt>' family of intrinsics counts the number of bits set in a
5747 The only argument is the value to be counted. The argument may be of any
5748 integer type. The return type must match the argument type.
5754 The '<tt>llvm.ctpop</tt>' intrinsic counts the 1's in a variable.
5758 <!-- _______________________________________________________________________ -->
5759 <div class="doc_subsubsection">
5760 <a name="int_ctlz">'<tt>llvm.ctlz.*</tt>' Intrinsic</a>
5763 <div class="doc_text">
5766 <p>This is an overloaded intrinsic. You can use <tt>llvm.ctlz</tt> on any
5767 integer bit width. Not all targets support all bit widths however.</p>
5769 declare i8 @llvm.ctlz.i8 (i8 <src>)
5770 declare i16 @llvm.ctlz.i16(i16 <src>)
5771 declare i32 @llvm.ctlz.i32(i32 <src>)
5772 declare i64 @llvm.ctlz.i64(i64 <src>)
5773 declare i256 @llvm.ctlz.i256(i256 <src>)
5779 The '<tt>llvm.ctlz</tt>' family of intrinsic functions counts the number of
5780 leading zeros in a variable.
5786 The only argument is the value to be counted. The argument may be of any
5787 integer type. The return type must match the argument type.
5793 The '<tt>llvm.ctlz</tt>' intrinsic counts the leading (most significant) zeros
5794 in a variable. If the src == 0 then the result is the size in bits of the type
5795 of src. For example, <tt>llvm.ctlz(i32 2) = 30</tt>.
5801 <!-- _______________________________________________________________________ -->
5802 <div class="doc_subsubsection">
5803 <a name="int_cttz">'<tt>llvm.cttz.*</tt>' Intrinsic</a>
5806 <div class="doc_text">
5809 <p>This is an overloaded intrinsic. You can use <tt>llvm.cttz</tt> on any
5810 integer bit width. Not all targets support all bit widths however.</p>
5812 declare i8 @llvm.cttz.i8 (i8 <src>)
5813 declare i16 @llvm.cttz.i16(i16 <src>)
5814 declare i32 @llvm.cttz.i32(i32 <src>)
5815 declare i64 @llvm.cttz.i64(i64 <src>)
5816 declare i256 @llvm.cttz.i256(i256 <src>)
5822 The '<tt>llvm.cttz</tt>' family of intrinsic functions counts the number of
5829 The only argument is the value to be counted. The argument may be of any
5830 integer type. The return type must match the argument type.
5836 The '<tt>llvm.cttz</tt>' intrinsic counts the trailing (least significant) zeros
5837 in a variable. If the src == 0 then the result is the size in bits of the type
5838 of src. For example, <tt>llvm.cttz(2) = 1</tt>.
5842 <!-- _______________________________________________________________________ -->
5843 <div class="doc_subsubsection">
5844 <a name="int_part_select">'<tt>llvm.part.select.*</tt>' Intrinsic</a>
5847 <div class="doc_text">
5850 <p>This is an overloaded intrinsic. You can use <tt>llvm.part.select</tt>
5851 on any integer bit width.</p>
5853 declare i17 @llvm.part.select.i17 (i17 %val, i32 %loBit, i32 %hiBit)
5854 declare i29 @llvm.part.select.i29 (i29 %val, i32 %loBit, i32 %hiBit)
5858 <p>The '<tt>llvm.part.select</tt>' family of intrinsic functions selects a
5859 range of bits from an integer value and returns them in the same bit width as
5860 the original value.</p>
5863 <p>The first argument, <tt>%val</tt> and the result may be integer types of
5864 any bit width but they must have the same bit width. The second and third
5865 arguments must be <tt>i32</tt> type since they specify only a bit index.</p>
5868 <p>The operation of the '<tt>llvm.part.select</tt>' intrinsic has two modes
5869 of operation: forwards and reverse. If <tt>%loBit</tt> is greater than
5870 <tt>%hiBits</tt> then the intrinsic operates in reverse mode. Otherwise it
5871 operates in forward mode.</p>
5872 <p>In forward mode, this intrinsic is the equivalent of shifting <tt>%val</tt>
5873 right by <tt>%loBit</tt> bits and then ANDing it with a mask with
5874 only the <tt>%hiBit - %loBit</tt> bits set, as follows:</p>
5876 <li>The <tt>%val</tt> is shifted right (LSHR) by the number of bits specified
5877 by <tt>%loBits</tt>. This normalizes the value to the low order bits.</li>
5878 <li>The <tt>%loBits</tt> value is subtracted from the <tt>%hiBits</tt> value
5879 to determine the number of bits to retain.</li>
5880 <li>A mask of the retained bits is created by shifting a -1 value.</li>
5881 <li>The mask is ANDed with <tt>%val</tt> to produce the result.</li>
5883 <p>In reverse mode, a similar computation is made except that the bits are
5884 returned in the reverse order. So, for example, if <tt>X</tt> has the value
5885 <tt>i16 0x0ACF (101011001111)</tt> and we apply
5886 <tt>part.select(i16 X, 8, 3)</tt> to it, we get back the value
5887 <tt>i16 0x0026 (000000100110)</tt>.</p>
5890 <div class="doc_subsubsection">
5891 <a name="int_part_set">'<tt>llvm.part.set.*</tt>' Intrinsic</a>
5894 <div class="doc_text">
5897 <p>This is an overloaded intrinsic. You can use <tt>llvm.part.set</tt>
5898 on any integer bit width.</p>
5900 declare i17 @llvm.part.set.i17.i9 (i17 %val, i9 %repl, i32 %lo, i32 %hi)
5901 declare i29 @llvm.part.set.i29.i9 (i29 %val, i9 %repl, i32 %lo, i32 %hi)
5905 <p>The '<tt>llvm.part.set</tt>' family of intrinsic functions replaces a range
5906 of bits in an integer value with another integer value. It returns the integer
5907 with the replaced bits.</p>
5910 <p>The first argument, <tt>%val</tt>, and the result may be integer types of
5911 any bit width, but they must have the same bit width. <tt>%val</tt> is the value
5912 whose bits will be replaced. The second argument, <tt>%repl</tt> may be an
5913 integer of any bit width. The third and fourth arguments must be <tt>i32</tt>
5914 type since they specify only a bit index.</p>
5917 <p>The operation of the '<tt>llvm.part.set</tt>' intrinsic has two modes
5918 of operation: forwards and reverse. If <tt>%lo</tt> is greater than
5919 <tt>%hi</tt> then the intrinsic operates in reverse mode. Otherwise it
5920 operates in forward mode.</p>
5922 <p>For both modes, the <tt>%repl</tt> value is prepared for use by either
5923 truncating it down to the size of the replacement area or zero extending it
5924 up to that size.</p>
5926 <p>In forward mode, the bits between <tt>%lo</tt> and <tt>%hi</tt> (inclusive)
5927 are replaced with corresponding bits from <tt>%repl</tt>. That is the 0th bit
5928 in <tt>%repl</tt> replaces the <tt>%lo</tt>th bit in <tt>%val</tt> and etc. up
5929 to the <tt>%hi</tt>th bit.</p>
5931 <p>In reverse mode, a similar computation is made except that the bits are
5932 reversed. That is, the <tt>0</tt>th bit in <tt>%repl</tt> replaces the
5933 <tt>%hi</tt> bit in <tt>%val</tt> and etc. down to the <tt>%lo</tt>th bit.</p>
5938 llvm.part.set(0xFFFF, 0, 4, 7) -> 0xFF0F
5939 llvm.part.set(0xFFFF, 0, 7, 4) -> 0xFF0F
5940 llvm.part.set(0xFFFF, 1, 7, 4) -> 0xFF8F
5941 llvm.part.set(0xFFFF, F, 8, 3) -> 0xFFE7
5942 llvm.part.set(0xFFFF, 0, 3, 8) -> 0xFE07
5947 <!-- ======================================================================= -->
5948 <div class="doc_subsection">
5949 <a name="int_overflow">Arithmetic with Overflow Intrinsics</a>
5952 <div class="doc_text">
5954 LLVM provides intrinsics for some arithmetic with overflow operations.
5959 <!-- _______________________________________________________________________ -->
5960 <div class="doc_subsubsection">
5961 <a name="int_sadd_overflow">'<tt>llvm.sadd.with.overflow.*</tt>' Intrinsics</a>
5964 <div class="doc_text">
5968 <p>This is an overloaded intrinsic. You can use <tt>llvm.sadd.with.overflow</tt>
5969 on any integer bit width.</p>
5972 declare {i16, i1} @llvm.sadd.with.overflow.i16(i16 %a, i16 %b)
5973 declare {i32, i1} @llvm.sadd.with.overflow.i32(i32 %a, i32 %b)
5974 declare {i64, i1} @llvm.sadd.with.overflow.i64(i64 %a, i64 %b)
5979 <p>The '<tt>llvm.sadd.with.overflow</tt>' family of intrinsic functions perform
5980 a signed addition of the two arguments, and indicate whether an overflow
5981 occurred during the signed summation.</p>
5985 <p>The arguments (%a and %b) and the first element of the result structure may
5986 be of integer types of any bit width, but they must have the same bit width. The
5987 second element of the result structure must be of type <tt>i1</tt>. <tt>%a</tt>
5988 and <tt>%b</tt> are the two values that will undergo signed addition.</p>
5992 <p>The '<tt>llvm.sadd.with.overflow</tt>' family of intrinsic functions perform
5993 a signed addition of the two variables. They return a structure — the
5994 first element of which is the signed summation, and the second element of which
5995 is a bit specifying if the signed summation resulted in an overflow.</p>
5999 %res = call {i32, i1} @llvm.sadd.with.overflow.i32(i32 %a, i32 %b)
6000 %sum = extractvalue {i32, i1} %res, 0
6001 %obit = extractvalue {i32, i1} %res, 1
6002 br i1 %obit, label %overflow, label %normal
6007 <!-- _______________________________________________________________________ -->
6008 <div class="doc_subsubsection">
6009 <a name="int_uadd_overflow">'<tt>llvm.uadd.with.overflow.*</tt>' Intrinsics</a>
6012 <div class="doc_text">
6016 <p>This is an overloaded intrinsic. You can use <tt>llvm.uadd.with.overflow</tt>
6017 on any integer bit width.</p>
6020 declare {i16, i1} @llvm.uadd.with.overflow.i16(i16 %a, i16 %b)
6021 declare {i32, i1} @llvm.uadd.with.overflow.i32(i32 %a, i32 %b)
6022 declare {i64, i1} @llvm.uadd.with.overflow.i64(i64 %a, i64 %b)
6027 <p>The '<tt>llvm.uadd.with.overflow</tt>' family of intrinsic functions perform
6028 an unsigned addition of the two arguments, and indicate whether a carry occurred
6029 during the unsigned summation.</p>
6033 <p>The arguments (%a and %b) and the first element of the result structure may
6034 be of integer types of any bit width, but they must have the same bit width. The
6035 second element of the result structure must be of type <tt>i1</tt>. <tt>%a</tt>
6036 and <tt>%b</tt> are the two values that will undergo unsigned addition.</p>
6040 <p>The '<tt>llvm.uadd.with.overflow</tt>' family of intrinsic functions perform
6041 an unsigned addition of the two arguments. They return a structure — the
6042 first element of which is the sum, and the second element of which is a bit
6043 specifying if the unsigned summation resulted in a carry.</p>
6047 %res = call {i32, i1} @llvm.uadd.with.overflow.i32(i32 %a, i32 %b)
6048 %sum = extractvalue {i32, i1} %res, 0
6049 %obit = extractvalue {i32, i1} %res, 1
6050 br i1 %obit, label %carry, label %normal
6055 <!-- _______________________________________________________________________ -->
6056 <div class="doc_subsubsection">
6057 <a name="int_ssub_overflow">'<tt>llvm.ssub.with.overflow.*</tt>' Intrinsics</a>
6060 <div class="doc_text">
6064 <p>This is an overloaded intrinsic. You can use <tt>llvm.ssub.with.overflow</tt>
6065 on any integer bit width.</p>
6068 declare {i16, i1} @llvm.ssub.with.overflow.i16(i16 %a, i16 %b)
6069 declare {i32, i1} @llvm.ssub.with.overflow.i32(i32 %a, i32 %b)
6070 declare {i64, i1} @llvm.ssub.with.overflow.i64(i64 %a, i64 %b)
6075 <p>The '<tt>llvm.ssub.with.overflow</tt>' family of intrinsic functions perform
6076 a signed subtraction of the two arguments, and indicate whether an overflow
6077 occurred during the signed subtraction.</p>
6081 <p>The arguments (%a and %b) and the first element of the result structure may
6082 be of integer types of any bit width, but they must have the same bit width. The
6083 second element of the result structure must be of type <tt>i1</tt>. <tt>%a</tt>
6084 and <tt>%b</tt> are the two values that will undergo signed subtraction.</p>
6088 <p>The '<tt>llvm.ssub.with.overflow</tt>' family of intrinsic functions perform
6089 a signed subtraction of the two arguments. They return a structure — the
6090 first element of which is the subtraction, and the second element of which is a bit
6091 specifying if the signed subtraction resulted in an overflow.</p>
6095 %res = call {i32, i1} @llvm.ssub.with.overflow.i32(i32 %a, i32 %b)
6096 %sum = extractvalue {i32, i1} %res, 0
6097 %obit = extractvalue {i32, i1} %res, 1
6098 br i1 %obit, label %overflow, label %normal
6103 <!-- _______________________________________________________________________ -->
6104 <div class="doc_subsubsection">
6105 <a name="int_usub_overflow">'<tt>llvm.usub.with.overflow.*</tt>' Intrinsics</a>
6108 <div class="doc_text">
6112 <p>This is an overloaded intrinsic. You can use <tt>llvm.usub.with.overflow</tt>
6113 on any integer bit width.</p>
6116 declare {i16, i1} @llvm.usub.with.overflow.i16(i16 %a, i16 %b)
6117 declare {i32, i1} @llvm.usub.with.overflow.i32(i32 %a, i32 %b)
6118 declare {i64, i1} @llvm.usub.with.overflow.i64(i64 %a, i64 %b)
6123 <p>The '<tt>llvm.usub.with.overflow</tt>' family of intrinsic functions perform
6124 an unsigned subtraction of the two arguments, and indicate whether an overflow
6125 occurred during the unsigned subtraction.</p>
6129 <p>The arguments (%a and %b) and the first element of the result structure may
6130 be of integer types of any bit width, but they must have the same bit width. The
6131 second element of the result structure must be of type <tt>i1</tt>. <tt>%a</tt>
6132 and <tt>%b</tt> are the two values that will undergo unsigned subtraction.</p>
6136 <p>The '<tt>llvm.usub.with.overflow</tt>' family of intrinsic functions perform
6137 an unsigned subtraction of the two arguments. They return a structure — the
6138 first element of which is the subtraction, and the second element of which is a bit
6139 specifying if the unsigned subtraction resulted in an overflow.</p>
6143 %res = call {i32, i1} @llvm.usub.with.overflow.i32(i32 %a, i32 %b)
6144 %sum = extractvalue {i32, i1} %res, 0
6145 %obit = extractvalue {i32, i1} %res, 1
6146 br i1 %obit, label %overflow, label %normal
6151 <!-- _______________________________________________________________________ -->
6152 <div class="doc_subsubsection">
6153 <a name="int_smul_overflow">'<tt>llvm.smul.with.overflow.*</tt>' Intrinsics</a>
6156 <div class="doc_text">
6160 <p>This is an overloaded intrinsic. You can use <tt>llvm.smul.with.overflow</tt>
6161 on any integer bit width.</p>
6164 declare {i16, i1} @llvm.smul.with.overflow.i16(i16 %a, i16 %b)
6165 declare {i32, i1} @llvm.smul.with.overflow.i32(i32 %a, i32 %b)
6166 declare {i64, i1} @llvm.smul.with.overflow.i64(i64 %a, i64 %b)
6171 <p>The '<tt>llvm.smul.with.overflow</tt>' family of intrinsic functions perform
6172 a signed multiplication of the two arguments, and indicate whether an overflow
6173 occurred during the signed multiplication.</p>
6177 <p>The arguments (%a and %b) and the first element of the result structure may
6178 be of integer types of any bit width, but they must have the same bit width. The
6179 second element of the result structure must be of type <tt>i1</tt>. <tt>%a</tt>
6180 and <tt>%b</tt> are the two values that will undergo signed multiplication.</p>
6184 <p>The '<tt>llvm.smul.with.overflow</tt>' family of intrinsic functions perform
6185 a signed multiplication of the two arguments. They return a structure —
6186 the first element of which is the multiplication, and the second element of
6187 which is a bit specifying if the signed multiplication resulted in an
6192 %res = call {i32, i1} @llvm.smul.with.overflow.i32(i32 %a, i32 %b)
6193 %sum = extractvalue {i32, i1} %res, 0
6194 %obit = extractvalue {i32, i1} %res, 1
6195 br i1 %obit, label %overflow, label %normal
6200 <!-- _______________________________________________________________________ -->
6201 <div class="doc_subsubsection">
6202 <a name="int_umul_overflow">'<tt>llvm.umul.with.overflow.*</tt>' Intrinsics</a>
6205 <div class="doc_text">
6209 <p>This is an overloaded intrinsic. You can use <tt>llvm.umul.with.overflow</tt>
6210 on any integer bit width.</p>
6213 declare {i16, i1} @llvm.umul.with.overflow.i16(i16 %a, i16 %b)
6214 declare {i32, i1} @llvm.umul.with.overflow.i32(i32 %a, i32 %b)
6215 declare {i64, i1} @llvm.umul.with.overflow.i64(i64 %a, i64 %b)
6220 <p><i><b>Warning:</b> '<tt>llvm.umul.with.overflow</tt>' is badly broken. It is
6221 actively being fixed, but it should not currently be used!</i></p>
6223 <p>The '<tt>llvm.umul.with.overflow</tt>' family of intrinsic functions perform
6224 a unsigned multiplication of the two arguments, and indicate whether an overflow
6225 occurred during the unsigned multiplication.</p>
6229 <p>The arguments (%a and %b) and the first element of the result structure may
6230 be of integer types of any bit width, but they must have the same bit width. The
6231 second element of the result structure must be of type <tt>i1</tt>. <tt>%a</tt>
6232 and <tt>%b</tt> are the two values that will undergo unsigned
6237 <p>The '<tt>llvm.umul.with.overflow</tt>' family of intrinsic functions perform
6238 an unsigned multiplication of the two arguments. They return a structure —
6239 the first element of which is the multiplication, and the second element of
6240 which is a bit specifying if the unsigned multiplication resulted in an
6245 %res = call {i32, i1} @llvm.umul.with.overflow.i32(i32 %a, i32 %b)
6246 %sum = extractvalue {i32, i1} %res, 0
6247 %obit = extractvalue {i32, i1} %res, 1
6248 br i1 %obit, label %overflow, label %normal
6253 <!-- ======================================================================= -->
6254 <div class="doc_subsection">
6255 <a name="int_debugger">Debugger Intrinsics</a>
6258 <div class="doc_text">
6260 The LLVM debugger intrinsics (which all start with <tt>llvm.dbg.</tt> prefix),
6261 are described in the <a
6262 href="SourceLevelDebugging.html#format_common_intrinsics">LLVM Source Level
6263 Debugging</a> document.
6268 <!-- ======================================================================= -->
6269 <div class="doc_subsection">
6270 <a name="int_eh">Exception Handling Intrinsics</a>
6273 <div class="doc_text">
6274 <p> The LLVM exception handling intrinsics (which all start with
6275 <tt>llvm.eh.</tt> prefix), are described in the <a
6276 href="ExceptionHandling.html#format_common_intrinsics">LLVM Exception
6277 Handling</a> document. </p>
6280 <!-- ======================================================================= -->
6281 <div class="doc_subsection">
6282 <a name="int_trampoline">Trampoline Intrinsic</a>
6285 <div class="doc_text">
6287 This intrinsic makes it possible to excise one parameter, marked with
6288 the <tt>nest</tt> attribute, from a function. The result is a callable
6289 function pointer lacking the nest parameter - the caller does not need
6290 to provide a value for it. Instead, the value to use is stored in
6291 advance in a "trampoline", a block of memory usually allocated
6292 on the stack, which also contains code to splice the nest value into the
6293 argument list. This is used to implement the GCC nested function address
6297 For example, if the function is
6298 <tt>i32 f(i8* nest %c, i32 %x, i32 %y)</tt> then the resulting function
6299 pointer has signature <tt>i32 (i32, i32)*</tt>. It can be created as follows:</p>
6301 %tramp = alloca [10 x i8], align 4 ; size and alignment only correct for X86
6302 %tramp1 = getelementptr [10 x i8]* %tramp, i32 0, i32 0
6303 %p = call i8* @llvm.init.trampoline( i8* %tramp1, i8* bitcast (i32 (i8* nest , i32, i32)* @f to i8*), i8* %nval )
6304 %fp = bitcast i8* %p to i32 (i32, i32)*
6306 <p>The call <tt>%val = call i32 %fp( i32 %x, i32 %y )</tt> is then equivalent
6307 to <tt>%val = call i32 %f( i8* %nval, i32 %x, i32 %y )</tt>.</p>
6310 <!-- _______________________________________________________________________ -->
6311 <div class="doc_subsubsection">
6312 <a name="int_it">'<tt>llvm.init.trampoline</tt>' Intrinsic</a>
6314 <div class="doc_text">
6317 declare i8* @llvm.init.trampoline(i8* <tramp>, i8* <func>, i8* <nval>)
6321 This fills the memory pointed to by <tt>tramp</tt> with code
6322 and returns a function pointer suitable for executing it.
6326 The <tt>llvm.init.trampoline</tt> intrinsic takes three arguments, all
6327 pointers. The <tt>tramp</tt> argument must point to a sufficiently large
6328 and sufficiently aligned block of memory; this memory is written to by the
6329 intrinsic. Note that the size and the alignment are target-specific - LLVM
6330 currently provides no portable way of determining them, so a front-end that
6331 generates this intrinsic needs to have some target-specific knowledge.
6332 The <tt>func</tt> argument must hold a function bitcast to an <tt>i8*</tt>.
6336 The block of memory pointed to by <tt>tramp</tt> is filled with target
6337 dependent code, turning it into a function. A pointer to this function is
6338 returned, but needs to be bitcast to an
6339 <a href="#int_trampoline">appropriate function pointer type</a>
6340 before being called. The new function's signature is the same as that of
6341 <tt>func</tt> with any arguments marked with the <tt>nest</tt> attribute
6342 removed. At most one such <tt>nest</tt> argument is allowed, and it must be
6343 of pointer type. Calling the new function is equivalent to calling
6344 <tt>func</tt> with the same argument list, but with <tt>nval</tt> used for the
6345 missing <tt>nest</tt> argument. If, after calling
6346 <tt>llvm.init.trampoline</tt>, the memory pointed to by <tt>tramp</tt> is
6347 modified, then the effect of any later call to the returned function pointer is
6352 <!-- ======================================================================= -->
6353 <div class="doc_subsection">
6354 <a name="int_atomics">Atomic Operations and Synchronization Intrinsics</a>
6357 <div class="doc_text">
6359 These intrinsic functions expand the "universal IR" of LLVM to represent
6360 hardware constructs for atomic operations and memory synchronization. This
6361 provides an interface to the hardware, not an interface to the programmer. It
6362 is aimed at a low enough level to allow any programming models or APIs
6363 (Application Programming Interfaces) which
6364 need atomic behaviors to map cleanly onto it. It is also modeled primarily on
6365 hardware behavior. Just as hardware provides a "universal IR" for source
6366 languages, it also provides a starting point for developing a "universal"
6367 atomic operation and synchronization IR.
6370 These do <em>not</em> form an API such as high-level threading libraries,
6371 software transaction memory systems, atomic primitives, and intrinsic
6372 functions as found in BSD, GNU libc, atomic_ops, APR, and other system and
6373 application libraries. The hardware interface provided by LLVM should allow
6374 a clean implementation of all of these APIs and parallel programming models.
6375 No one model or paradigm should be selected above others unless the hardware
6376 itself ubiquitously does so.
6381 <!-- _______________________________________________________________________ -->
6382 <div class="doc_subsubsection">
6383 <a name="int_memory_barrier">'<tt>llvm.memory.barrier</tt>' Intrinsic</a>
6385 <div class="doc_text">
6388 declare void @llvm.memory.barrier( i1 <ll>, i1 <ls>, i1 <sl>, i1 <ss>,
6394 The <tt>llvm.memory.barrier</tt> intrinsic guarantees ordering between
6395 specific pairs of memory access types.
6399 The <tt>llvm.memory.barrier</tt> intrinsic requires five boolean arguments.
6400 The first four arguments enables a specific barrier as listed below. The fith
6401 argument specifies that the barrier applies to io or device or uncached memory.
6405 <li><tt>ll</tt>: load-load barrier</li>
6406 <li><tt>ls</tt>: load-store barrier</li>
6407 <li><tt>sl</tt>: store-load barrier</li>
6408 <li><tt>ss</tt>: store-store barrier</li>
6409 <li><tt>device</tt>: barrier applies to device and uncached memory also.</li>
6413 This intrinsic causes the system to enforce some ordering constraints upon
6414 the loads and stores of the program. This barrier does not indicate
6415 <em>when</em> any events will occur, it only enforces an <em>order</em> in
6416 which they occur. For any of the specified pairs of load and store operations
6417 (f.ex. load-load, or store-load), all of the first operations preceding the
6418 barrier will complete before any of the second operations succeeding the
6419 barrier begin. Specifically the semantics for each pairing is as follows:
6422 <li><tt>ll</tt>: All loads before the barrier must complete before any load
6423 after the barrier begins.</li>
6425 <li><tt>ls</tt>: All loads before the barrier must complete before any
6426 store after the barrier begins.</li>
6427 <li><tt>ss</tt>: All stores before the barrier must complete before any
6428 store after the barrier begins.</li>
6429 <li><tt>sl</tt>: All stores before the barrier must complete before any
6430 load after the barrier begins.</li>
6433 These semantics are applied with a logical "and" behavior when more than one
6434 is enabled in a single memory barrier intrinsic.
6437 Backends may implement stronger barriers than those requested when they do not
6438 support as fine grained a barrier as requested. Some architectures do not
6439 need all types of barriers and on such architectures, these become noops.
6446 %result1 = load i32* %ptr <i>; yields {i32}:result1 = 4</i>
6447 call void @llvm.memory.barrier( i1 false, i1 true, i1 false, i1 false )
6448 <i>; guarantee the above finishes</i>
6449 store i32 8, %ptr <i>; before this begins</i>
6453 <!-- _______________________________________________________________________ -->
6454 <div class="doc_subsubsection">
6455 <a name="int_atomic_cmp_swap">'<tt>llvm.atomic.cmp.swap.*</tt>' Intrinsic</a>
6457 <div class="doc_text">
6460 This is an overloaded intrinsic. You can use <tt>llvm.atomic.cmp.swap</tt> on
6461 any integer bit width and for different address spaces. Not all targets
6462 support all bit widths however.</p>
6465 declare i8 @llvm.atomic.cmp.swap.i8.p0i8( i8* <ptr>, i8 <cmp>, i8 <val> )
6466 declare i16 @llvm.atomic.cmp.swap.i16.p0i16( i16* <ptr>, i16 <cmp>, i16 <val> )
6467 declare i32 @llvm.atomic.cmp.swap.i32.p0i32( i32* <ptr>, i32 <cmp>, i32 <val> )
6468 declare i64 @llvm.atomic.cmp.swap.i64.p0i64( i64* <ptr>, i64 <cmp>, i64 <val> )
6473 This loads a value in memory and compares it to a given value. If they are
6474 equal, it stores a new value into the memory.
6478 The <tt>llvm.atomic.cmp.swap</tt> intrinsic takes three arguments. The result as
6479 well as both <tt>cmp</tt> and <tt>val</tt> must be integer values with the
6480 same bit width. The <tt>ptr</tt> argument must be a pointer to a value of
6481 this integer type. While any bit width integer may be used, targets may only
6482 lower representations they support in hardware.
6487 This entire intrinsic must be executed atomically. It first loads the value
6488 in memory pointed to by <tt>ptr</tt> and compares it with the value
6489 <tt>cmp</tt>. If they are equal, <tt>val</tt> is stored into the memory. The
6490 loaded value is yielded in all cases. This provides the equivalent of an
6491 atomic compare-and-swap operation within the SSA framework.
6499 %val1 = add i32 4, 4
6500 %result1 = call i32 @llvm.atomic.cmp.swap.i32.p0i32( i32* %ptr, i32 4, %val1 )
6501 <i>; yields {i32}:result1 = 4</i>
6502 %stored1 = icmp eq i32 %result1, 4 <i>; yields {i1}:stored1 = true</i>
6503 %memval1 = load i32* %ptr <i>; yields {i32}:memval1 = 8</i>
6505 %val2 = add i32 1, 1
6506 %result2 = call i32 @llvm.atomic.cmp.swap.i32.p0i32( i32* %ptr, i32 5, %val2 )
6507 <i>; yields {i32}:result2 = 8</i>
6508 %stored2 = icmp eq i32 %result2, 5 <i>; yields {i1}:stored2 = false</i>
6510 %memval2 = load i32* %ptr <i>; yields {i32}:memval2 = 8</i>
6514 <!-- _______________________________________________________________________ -->
6515 <div class="doc_subsubsection">
6516 <a name="int_atomic_swap">'<tt>llvm.atomic.swap.*</tt>' Intrinsic</a>
6518 <div class="doc_text">
6522 This is an overloaded intrinsic. You can use <tt>llvm.atomic.swap</tt> on any
6523 integer bit width. Not all targets support all bit widths however.</p>
6525 declare i8 @llvm.atomic.swap.i8.p0i8( i8* <ptr>, i8 <val> )
6526 declare i16 @llvm.atomic.swap.i16.p0i16( i16* <ptr>, i16 <val> )
6527 declare i32 @llvm.atomic.swap.i32.p0i32( i32* <ptr>, i32 <val> )
6528 declare i64 @llvm.atomic.swap.i64.p0i64( i64* <ptr>, i64 <val> )
6533 This intrinsic loads the value stored in memory at <tt>ptr</tt> and yields
6534 the value from memory. It then stores the value in <tt>val</tt> in the memory
6540 The <tt>llvm.atomic.swap</tt> intrinsic takes two arguments. Both the
6541 <tt>val</tt> argument and the result must be integers of the same bit width.
6542 The first argument, <tt>ptr</tt>, must be a pointer to a value of this
6543 integer type. The targets may only lower integer representations they
6548 This intrinsic loads the value pointed to by <tt>ptr</tt>, yields it, and
6549 stores <tt>val</tt> back into <tt>ptr</tt> atomically. This provides the
6550 equivalent of an atomic swap operation within the SSA framework.
6558 %val1 = add i32 4, 4
6559 %result1 = call i32 @llvm.atomic.swap.i32.p0i32( i32* %ptr, i32 %val1 )
6560 <i>; yields {i32}:result1 = 4</i>
6561 %stored1 = icmp eq i32 %result1, 4 <i>; yields {i1}:stored1 = true</i>
6562 %memval1 = load i32* %ptr <i>; yields {i32}:memval1 = 8</i>
6564 %val2 = add i32 1, 1
6565 %result2 = call i32 @llvm.atomic.swap.i32.p0i32( i32* %ptr, i32 %val2 )
6566 <i>; yields {i32}:result2 = 8</i>
6568 %stored2 = icmp eq i32 %result2, 8 <i>; yields {i1}:stored2 = true</i>
6569 %memval2 = load i32* %ptr <i>; yields {i32}:memval2 = 2</i>
6573 <!-- _______________________________________________________________________ -->
6574 <div class="doc_subsubsection">
6575 <a name="int_atomic_load_add">'<tt>llvm.atomic.load.add.*</tt>' Intrinsic</a>
6578 <div class="doc_text">
6581 This is an overloaded intrinsic. You can use <tt>llvm.atomic.load.add</tt> on any
6582 integer bit width. Not all targets support all bit widths however.</p>
6584 declare i8 @llvm.atomic.load.add.i8..p0i8( i8* <ptr>, i8 <delta> )
6585 declare i16 @llvm.atomic.load.add.i16..p0i16( i16* <ptr>, i16 <delta> )
6586 declare i32 @llvm.atomic.load.add.i32..p0i32( i32* <ptr>, i32 <delta> )
6587 declare i64 @llvm.atomic.load.add.i64..p0i64( i64* <ptr>, i64 <delta> )
6592 This intrinsic adds <tt>delta</tt> to the value stored in memory at
6593 <tt>ptr</tt>. It yields the original value at <tt>ptr</tt>.
6598 The intrinsic takes two arguments, the first a pointer to an integer value
6599 and the second an integer value. The result is also an integer value. These
6600 integer types can have any bit width, but they must all have the same bit
6601 width. The targets may only lower integer representations they support.
6605 This intrinsic does a series of operations atomically. It first loads the
6606 value stored at <tt>ptr</tt>. It then adds <tt>delta</tt>, stores the result
6607 to <tt>ptr</tt>. It yields the original value stored at <tt>ptr</tt>.
6614 %result1 = call i32 @llvm.atomic.load.add.i32.p0i32( i32* %ptr, i32 4 )
6615 <i>; yields {i32}:result1 = 4</i>
6616 %result2 = call i32 @llvm.atomic.load.add.i32.p0i32( i32* %ptr, i32 2 )
6617 <i>; yields {i32}:result2 = 8</i>
6618 %result3 = call i32 @llvm.atomic.load.add.i32.p0i32( i32* %ptr, i32 5 )
6619 <i>; yields {i32}:result3 = 10</i>
6620 %memval1 = load i32* %ptr <i>; yields {i32}:memval1 = 15</i>
6624 <!-- _______________________________________________________________________ -->
6625 <div class="doc_subsubsection">
6626 <a name="int_atomic_load_sub">'<tt>llvm.atomic.load.sub.*</tt>' Intrinsic</a>
6629 <div class="doc_text">
6632 This is an overloaded intrinsic. You can use <tt>llvm.atomic.load.sub</tt> on
6633 any integer bit width and for different address spaces. Not all targets
6634 support all bit widths however.</p>
6636 declare i8 @llvm.atomic.load.sub.i8.p0i32( i8* <ptr>, i8 <delta> )
6637 declare i16 @llvm.atomic.load.sub.i16.p0i32( i16* <ptr>, i16 <delta> )
6638 declare i32 @llvm.atomic.load.sub.i32.p0i32( i32* <ptr>, i32 <delta> )
6639 declare i64 @llvm.atomic.load.sub.i64.p0i32( i64* <ptr>, i64 <delta> )
6644 This intrinsic subtracts <tt>delta</tt> to the value stored in memory at
6645 <tt>ptr</tt>. It yields the original value at <tt>ptr</tt>.
6650 The intrinsic takes two arguments, the first a pointer to an integer value
6651 and the second an integer value. The result is also an integer value. These
6652 integer types can have any bit width, but they must all have the same bit
6653 width. The targets may only lower integer representations they support.
6657 This intrinsic does a series of operations atomically. It first loads the
6658 value stored at <tt>ptr</tt>. It then subtracts <tt>delta</tt>, stores the
6659 result to <tt>ptr</tt>. It yields the original value stored at <tt>ptr</tt>.
6666 %result1 = call i32 @llvm.atomic.load.sub.i32.p0i32( i32* %ptr, i32 4 )
6667 <i>; yields {i32}:result1 = 8</i>
6668 %result2 = call i32 @llvm.atomic.load.sub.i32.p0i32( i32* %ptr, i32 2 )
6669 <i>; yields {i32}:result2 = 4</i>
6670 %result3 = call i32 @llvm.atomic.load.sub.i32.p0i32( i32* %ptr, i32 5 )
6671 <i>; yields {i32}:result3 = 2</i>
6672 %memval1 = load i32* %ptr <i>; yields {i32}:memval1 = -3</i>
6676 <!-- _______________________________________________________________________ -->
6677 <div class="doc_subsubsection">
6678 <a name="int_atomic_load_and">'<tt>llvm.atomic.load.and.*</tt>' Intrinsic</a><br>
6679 <a name="int_atomic_load_nand">'<tt>llvm.atomic.load.nand.*</tt>' Intrinsic</a><br>
6680 <a name="int_atomic_load_or">'<tt>llvm.atomic.load.or.*</tt>' Intrinsic</a><br>
6681 <a name="int_atomic_load_xor">'<tt>llvm.atomic.load.xor.*</tt>' Intrinsic</a><br>
6684 <div class="doc_text">
6687 These are overloaded intrinsics. You can use <tt>llvm.atomic.load_and</tt>,
6688 <tt>llvm.atomic.load_nand</tt>, <tt>llvm.atomic.load_or</tt>, and
6689 <tt>llvm.atomic.load_xor</tt> on any integer bit width and for different
6690 address spaces. Not all targets support all bit widths however.</p>
6692 declare i8 @llvm.atomic.load.and.i8.p0i8( i8* <ptr>, i8 <delta> )
6693 declare i16 @llvm.atomic.load.and.i16.p0i16( i16* <ptr>, i16 <delta> )
6694 declare i32 @llvm.atomic.load.and.i32.p0i32( i32* <ptr>, i32 <delta> )
6695 declare i64 @llvm.atomic.load.and.i64.p0i64( i64* <ptr>, i64 <delta> )
6700 declare i8 @llvm.atomic.load.or.i8.p0i8( i8* <ptr>, i8 <delta> )
6701 declare i16 @llvm.atomic.load.or.i16.p0i16( i16* <ptr>, i16 <delta> )
6702 declare i32 @llvm.atomic.load.or.i32.p0i32( i32* <ptr>, i32 <delta> )
6703 declare i64 @llvm.atomic.load.or.i64.p0i64( i64* <ptr>, i64 <delta> )
6708 declare i8 @llvm.atomic.load.nand.i8.p0i32( i8* <ptr>, i8 <delta> )
6709 declare i16 @llvm.atomic.load.nand.i16.p0i32( i16* <ptr>, i16 <delta> )
6710 declare i32 @llvm.atomic.load.nand.i32.p0i32( i32* <ptr>, i32 <delta> )
6711 declare i64 @llvm.atomic.load.nand.i64.p0i32( i64* <ptr>, i64 <delta> )
6716 declare i8 @llvm.atomic.load.xor.i8.p0i32( i8* <ptr>, i8 <delta> )
6717 declare i16 @llvm.atomic.load.xor.i16.p0i32( i16* <ptr>, i16 <delta> )
6718 declare i32 @llvm.atomic.load.xor.i32.p0i32( i32* <ptr>, i32 <delta> )
6719 declare i64 @llvm.atomic.load.xor.i64.p0i32( i64* <ptr>, i64 <delta> )
6724 These intrinsics bitwise the operation (and, nand, or, xor) <tt>delta</tt> to
6725 the value stored in memory at <tt>ptr</tt>. It yields the original value
6731 These intrinsics take two arguments, the first a pointer to an integer value
6732 and the second an integer value. The result is also an integer value. These
6733 integer types can have any bit width, but they must all have the same bit
6734 width. The targets may only lower integer representations they support.
6738 These intrinsics does a series of operations atomically. They first load the
6739 value stored at <tt>ptr</tt>. They then do the bitwise operation
6740 <tt>delta</tt>, store the result to <tt>ptr</tt>. They yield the original
6741 value stored at <tt>ptr</tt>.
6747 store i32 0x0F0F, %ptr
6748 %result0 = call i32 @llvm.atomic.load.nand.i32.p0i32( i32* %ptr, i32 0xFF )
6749 <i>; yields {i32}:result0 = 0x0F0F</i>
6750 %result1 = call i32 @llvm.atomic.load.and.i32.p0i32( i32* %ptr, i32 0xFF )
6751 <i>; yields {i32}:result1 = 0xFFFFFFF0</i>
6752 %result2 = call i32 @llvm.atomic.load.or.i32.p0i32( i32* %ptr, i32 0F )
6753 <i>; yields {i32}:result2 = 0xF0</i>
6754 %result3 = call i32 @llvm.atomic.load.xor.i32.p0i32( i32* %ptr, i32 0F )
6755 <i>; yields {i32}:result3 = FF</i>
6756 %memval1 = load i32* %ptr <i>; yields {i32}:memval1 = F0</i>
6761 <!-- _______________________________________________________________________ -->
6762 <div class="doc_subsubsection">
6763 <a name="int_atomic_load_max">'<tt>llvm.atomic.load.max.*</tt>' Intrinsic</a><br>
6764 <a name="int_atomic_load_min">'<tt>llvm.atomic.load.min.*</tt>' Intrinsic</a><br>
6765 <a name="int_atomic_load_umax">'<tt>llvm.atomic.load.umax.*</tt>' Intrinsic</a><br>
6766 <a name="int_atomic_load_umin">'<tt>llvm.atomic.load.umin.*</tt>' Intrinsic</a><br>
6769 <div class="doc_text">
6772 These are overloaded intrinsics. You can use <tt>llvm.atomic.load_max</tt>,
6773 <tt>llvm.atomic.load_min</tt>, <tt>llvm.atomic.load_umax</tt>, and
6774 <tt>llvm.atomic.load_umin</tt> on any integer bit width and for different
6775 address spaces. Not all targets
6776 support all bit widths however.</p>
6778 declare i8 @llvm.atomic.load.max.i8.p0i8( i8* <ptr>, i8 <delta> )
6779 declare i16 @llvm.atomic.load.max.i16.p0i16( i16* <ptr>, i16 <delta> )
6780 declare i32 @llvm.atomic.load.max.i32.p0i32( i32* <ptr>, i32 <delta> )
6781 declare i64 @llvm.atomic.load.max.i64.p0i64( i64* <ptr>, i64 <delta> )
6786 declare i8 @llvm.atomic.load.min.i8.p0i8( i8* <ptr>, i8 <delta> )
6787 declare i16 @llvm.atomic.load.min.i16.p0i16( i16* <ptr>, i16 <delta> )
6788 declare i32 @llvm.atomic.load.min.i32..p0i32( i32* <ptr>, i32 <delta> )
6789 declare i64 @llvm.atomic.load.min.i64..p0i64( i64* <ptr>, i64 <delta> )
6794 declare i8 @llvm.atomic.load.umax.i8.p0i8( i8* <ptr>, i8 <delta> )
6795 declare i16 @llvm.atomic.load.umax.i16.p0i16( i16* <ptr>, i16 <delta> )
6796 declare i32 @llvm.atomic.load.umax.i32.p0i32( i32* <ptr>, i32 <delta> )
6797 declare i64 @llvm.atomic.load.umax.i64.p0i64( i64* <ptr>, i64 <delta> )
6802 declare i8 @llvm.atomic.load.umin.i8..p0i8( i8* <ptr>, i8 <delta> )
6803 declare i16 @llvm.atomic.load.umin.i16.p0i16( i16* <ptr>, i16 <delta> )
6804 declare i32 @llvm.atomic.load.umin.i32..p0i32( i32* <ptr>, i32 <delta> )
6805 declare i64 @llvm.atomic.load.umin.i64..p0i64( i64* <ptr>, i64 <delta> )
6810 These intrinsics takes the signed or unsigned minimum or maximum of
6811 <tt>delta</tt> and the value stored in memory at <tt>ptr</tt>. It yields the
6812 original value at <tt>ptr</tt>.
6817 These intrinsics take two arguments, the first a pointer to an integer value
6818 and the second an integer value. The result is also an integer value. These
6819 integer types can have any bit width, but they must all have the same bit
6820 width. The targets may only lower integer representations they support.
6824 These intrinsics does a series of operations atomically. They first load the
6825 value stored at <tt>ptr</tt>. They then do the signed or unsigned min or max
6826 <tt>delta</tt> and the value, store the result to <tt>ptr</tt>. They yield
6827 the original value stored at <tt>ptr</tt>.
6834 %result0 = call i32 @llvm.atomic.load.min.i32.p0i32( i32* %ptr, i32 -2 )
6835 <i>; yields {i32}:result0 = 7</i>
6836 %result1 = call i32 @llvm.atomic.load.max.i32.p0i32( i32* %ptr, i32 8 )
6837 <i>; yields {i32}:result1 = -2</i>
6838 %result2 = call i32 @llvm.atomic.load.umin.i32.p0i32( i32* %ptr, i32 10 )
6839 <i>; yields {i32}:result2 = 8</i>
6840 %result3 = call i32 @llvm.atomic.load.umax.i32.p0i32( i32* %ptr, i32 30 )
6841 <i>; yields {i32}:result3 = 8</i>
6842 %memval1 = load i32* %ptr <i>; yields {i32}:memval1 = 30</i>
6846 <!-- ======================================================================= -->
6847 <div class="doc_subsection">
6848 <a name="int_general">General Intrinsics</a>
6851 <div class="doc_text">
6852 <p> This class of intrinsics is designed to be generic and has
6853 no specific purpose. </p>
6856 <!-- _______________________________________________________________________ -->
6857 <div class="doc_subsubsection">
6858 <a name="int_var_annotation">'<tt>llvm.var.annotation</tt>' Intrinsic</a>
6861 <div class="doc_text">
6865 declare void @llvm.var.annotation(i8* <val>, i8* <str>, i8* <str>, i32 <int> )
6871 The '<tt>llvm.var.annotation</tt>' intrinsic
6877 The first argument is a pointer to a value, the second is a pointer to a
6878 global string, the third is a pointer to a global string which is the source
6879 file name, and the last argument is the line number.
6885 This intrinsic allows annotation of local variables with arbitrary strings.
6886 This can be useful for special purpose optimizations that want to look for these
6887 annotations. These have no other defined use, they are ignored by code
6888 generation and optimization.
6892 <!-- _______________________________________________________________________ -->
6893 <div class="doc_subsubsection">
6894 <a name="int_annotation">'<tt>llvm.annotation.*</tt>' Intrinsic</a>
6897 <div class="doc_text">
6900 <p>This is an overloaded intrinsic. You can use '<tt>llvm.annotation</tt>' on
6901 any integer bit width.
6904 declare i8 @llvm.annotation.i8(i8 <val>, i8* <str>, i8* <str>, i32 <int> )
6905 declare i16 @llvm.annotation.i16(i16 <val>, i8* <str>, i8* <str>, i32 <int> )
6906 declare i32 @llvm.annotation.i32(i32 <val>, i8* <str>, i8* <str>, i32 <int> )
6907 declare i64 @llvm.annotation.i64(i64 <val>, i8* <str>, i8* <str>, i32 <int> )
6908 declare i256 @llvm.annotation.i256(i256 <val>, i8* <str>, i8* <str>, i32 <int> )
6914 The '<tt>llvm.annotation</tt>' intrinsic.
6920 The first argument is an integer value (result of some expression),
6921 the second is a pointer to a global string, the third is a pointer to a global
6922 string which is the source file name, and the last argument is the line number.
6923 It returns the value of the first argument.
6929 This intrinsic allows annotations to be put on arbitrary expressions
6930 with arbitrary strings. This can be useful for special purpose optimizations
6931 that want to look for these annotations. These have no other defined use, they
6932 are ignored by code generation and optimization.
6936 <!-- _______________________________________________________________________ -->
6937 <div class="doc_subsubsection">
6938 <a name="int_trap">'<tt>llvm.trap</tt>' Intrinsic</a>
6941 <div class="doc_text">
6945 declare void @llvm.trap()
6951 The '<tt>llvm.trap</tt>' intrinsic
6963 This intrinsics is lowered to the target dependent trap instruction. If the
6964 target does not have a trap instruction, this intrinsic will be lowered to the
6965 call of the abort() function.
6969 <!-- _______________________________________________________________________ -->
6970 <div class="doc_subsubsection">
6971 <a name="int_stackprotector">'<tt>llvm.stackprotector</tt>' Intrinsic</a>
6973 <div class="doc_text">
6976 declare void @llvm.stackprotector( i8* <guard>, i8** <slot> )
6981 The <tt>llvm.stackprotector</tt> intrinsic takes the <tt>guard</tt> and stores
6982 it onto the stack at <tt>slot</tt>. The stack slot is adjusted to ensure that
6983 it is placed on the stack before local variables.
6987 The <tt>llvm.stackprotector</tt> intrinsic requires two pointer arguments. The
6988 first argument is the value loaded from the stack guard
6989 <tt>@__stack_chk_guard</tt>. The second variable is an <tt>alloca</tt> that
6990 has enough space to hold the value of the guard.
6994 This intrinsic causes the prologue/epilogue inserter to force the position of
6995 the <tt>AllocaInst</tt> stack slot to be before local variables on the
6996 stack. This is to ensure that if a local variable on the stack is overwritten,
6997 it will destroy the value of the guard. When the function exits, the guard on
6998 the stack is checked against the original guard. If they're different, then
6999 the program aborts by calling the <tt>__stack_chk_fail()</tt> function.
7003 <!-- *********************************************************************** -->
7006 <a href="http://jigsaw.w3.org/css-validator/check/referer"><img
7007 src="http://jigsaw.w3.org/css-validator/images/vcss-blue" alt="Valid CSS"></a>
7008 <a href="http://validator.w3.org/check/referer"><img
7009 src="http://www.w3.org/Icons/valid-html401-blue" alt="Valid HTML 4.01"></a>
7011 <a href="mailto:sabre@nondot.org">Chris Lattner</a><br>
7012 <a href="http://llvm.org">The LLVM Compiler Infrastructure</a><br>
7013 Last modified: $Date$