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="#aggregateconstants">Aggregate 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_debugger">Debugger intrinsics</a></li>
213 <li><a href="#int_eh">Exception Handling intrinsics</a></li>
214 <li><a href="#int_trampoline">Trampoline Intrinsic</a>
216 <li><a href="#int_it">'<tt>llvm.init.trampoline</tt>' Intrinsic</a></li>
219 <li><a href="#int_atomics">Atomic intrinsics</a>
221 <li><a href="#int_memory_barrier"><tt>llvm.memory_barrier</tt></a></li>
222 <li><a href="#int_atomic_cmp_swap"><tt>llvm.atomic.cmp.swap</tt></a></li>
223 <li><a href="#int_atomic_swap"><tt>llvm.atomic.swap</tt></a></li>
224 <li><a href="#int_atomic_load_add"><tt>llvm.atomic.load.add</tt></a></li>
225 <li><a href="#int_atomic_load_sub"><tt>llvm.atomic.load.sub</tt></a></li>
226 <li><a href="#int_atomic_load_and"><tt>llvm.atomic.load.and</tt></a></li>
227 <li><a href="#int_atomic_load_nand"><tt>llvm.atomic.load.nand</tt></a></li>
228 <li><a href="#int_atomic_load_or"><tt>llvm.atomic.load.or</tt></a></li>
229 <li><a href="#int_atomic_load_xor"><tt>llvm.atomic.load.xor</tt></a></li>
230 <li><a href="#int_atomic_load_max"><tt>llvm.atomic.load.max</tt></a></li>
231 <li><a href="#int_atomic_load_min"><tt>llvm.atomic.load.min</tt></a></li>
232 <li><a href="#int_atomic_load_umax"><tt>llvm.atomic.load.umax</tt></a></li>
233 <li><a href="#int_atomic_load_umin"><tt>llvm.atomic.load.umin</tt></a></li>
236 <li><a href="#int_general">General intrinsics</a>
238 <li><a href="#int_var_annotation">
239 '<tt>llvm.var.annotation</tt>' Intrinsic</a></li>
240 <li><a href="#int_annotation">
241 '<tt>llvm.annotation.*</tt>' Intrinsic</a></li>
242 <li><a href="#int_trap">
243 '<tt>llvm.trap</tt>' Intrinsic</a></li>
244 <li><a href="#int_stackprotector">
245 '<tt>llvm.stackprotector</tt>' Intrinsic</a></li>
252 <div class="doc_author">
253 <p>Written by <a href="mailto:sabre@nondot.org">Chris Lattner</a>
254 and <a href="mailto:vadve@cs.uiuc.edu">Vikram Adve</a></p>
257 <!-- *********************************************************************** -->
258 <div class="doc_section"> <a name="abstract">Abstract </a></div>
259 <!-- *********************************************************************** -->
261 <div class="doc_text">
262 <p>This document is a reference manual for the LLVM assembly language.
263 LLVM is a Static Single Assignment (SSA) based representation that provides
264 type safety, low-level operations, flexibility, and the capability of
265 representing 'all' high-level languages cleanly. It is the common code
266 representation used throughout all phases of the LLVM compilation
270 <!-- *********************************************************************** -->
271 <div class="doc_section"> <a name="introduction">Introduction</a> </div>
272 <!-- *********************************************************************** -->
274 <div class="doc_text">
276 <p>The LLVM code representation is designed to be used in three
277 different forms: as an in-memory compiler IR, as an on-disk bitcode
278 representation (suitable for fast loading by a Just-In-Time compiler),
279 and as a human readable assembly language representation. This allows
280 LLVM to provide a powerful intermediate representation for efficient
281 compiler transformations and analysis, while providing a natural means
282 to debug and visualize the transformations. The three different forms
283 of LLVM are all equivalent. This document describes the human readable
284 representation and notation.</p>
286 <p>The LLVM representation aims to be light-weight and low-level
287 while being expressive, typed, and extensible at the same time. It
288 aims to be a "universal IR" of sorts, by being at a low enough level
289 that high-level ideas may be cleanly mapped to it (similar to how
290 microprocessors are "universal IR's", allowing many source languages to
291 be mapped to them). By providing type information, LLVM can be used as
292 the target of optimizations: for example, through pointer analysis, it
293 can be proven that a C automatic variable is never accessed outside of
294 the current function... allowing it to be promoted to a simple SSA
295 value instead of a memory location.</p>
299 <!-- _______________________________________________________________________ -->
300 <div class="doc_subsubsection"> <a name="wellformed">Well-Formedness</a> </div>
302 <div class="doc_text">
304 <p>It is important to note that this document describes 'well formed'
305 LLVM assembly language. There is a difference between what the parser
306 accepts and what is considered 'well formed'. For example, the
307 following instruction is syntactically okay, but not well formed:</p>
309 <div class="doc_code">
311 %x = <a href="#i_add">add</a> i32 1, %x
315 <p>...because the definition of <tt>%x</tt> does not dominate all of
316 its uses. The LLVM infrastructure provides a verification pass that may
317 be used to verify that an LLVM module is well formed. This pass is
318 automatically run by the parser after parsing input assembly and by
319 the optimizer before it outputs bitcode. The violations pointed out
320 by the verifier pass indicate bugs in transformation passes or input to
324 <!-- Describe the typesetting conventions here. -->
326 <!-- *********************************************************************** -->
327 <div class="doc_section"> <a name="identifiers">Identifiers</a> </div>
328 <!-- *********************************************************************** -->
330 <div class="doc_text">
332 <p>LLVM identifiers come in two basic types: global and local. Global
333 identifiers (functions, global variables) begin with the @ character. Local
334 identifiers (register names, types) begin with the % character. Additionally,
335 there are three different formats for identifiers, for different purposes:</p>
338 <li>Named values are represented as a string of characters with their prefix.
339 For example, %foo, @DivisionByZero, %a.really.long.identifier. The actual
340 regular expression used is '<tt>[%@][a-zA-Z$._][a-zA-Z$._0-9]*</tt>'.
341 Identifiers which require other characters in their names can be surrounded
342 with quotes. Special characters may be escaped using "\xx" where xx is the
343 ASCII code for the character in hexadecimal. In this way, any character can
344 be used in a name value, even quotes themselves.
346 <li>Unnamed values are represented as an unsigned numeric value with their
347 prefix. For example, %12, @2, %44.</li>
349 <li>Constants, which are described in a <a href="#constants">section about
350 constants</a>, below.</li>
353 <p>LLVM requires that values start with a prefix for two reasons: Compilers
354 don't need to worry about name clashes with reserved words, and the set of
355 reserved words may be expanded in the future without penalty. Additionally,
356 unnamed identifiers allow a compiler to quickly come up with a temporary
357 variable without having to avoid symbol table conflicts.</p>
359 <p>Reserved words in LLVM are very similar to reserved words in other
360 languages. There are keywords for different opcodes
361 ('<tt><a href="#i_add">add</a></tt>',
362 '<tt><a href="#i_bitcast">bitcast</a></tt>',
363 '<tt><a href="#i_ret">ret</a></tt>', etc...), for primitive type names ('<tt><a
364 href="#t_void">void</a></tt>', '<tt><a href="#t_primitive">i32</a></tt>', etc...),
365 and others. These reserved words cannot conflict with variable names, because
366 none of them start with a prefix character ('%' or '@').</p>
368 <p>Here is an example of LLVM code to multiply the integer variable
369 '<tt>%X</tt>' by 8:</p>
373 <div class="doc_code">
375 %result = <a href="#i_mul">mul</a> i32 %X, 8
379 <p>After strength reduction:</p>
381 <div class="doc_code">
383 %result = <a href="#i_shl">shl</a> i32 %X, i8 3
387 <p>And the hard way:</p>
389 <div class="doc_code">
391 <a href="#i_add">add</a> i32 %X, %X <i>; yields {i32}:%0</i>
392 <a href="#i_add">add</a> i32 %0, %0 <i>; yields {i32}:%1</i>
393 %result = <a href="#i_add">add</a> i32 %1, %1
397 <p>This last way of multiplying <tt>%X</tt> by 8 illustrates several
398 important lexical features of LLVM:</p>
402 <li>Comments are delimited with a '<tt>;</tt>' and go until the end of
405 <li>Unnamed temporaries are created when the result of a computation is not
406 assigned to a named value.</li>
408 <li>Unnamed temporaries are numbered sequentially</li>
412 <p>...and it also shows a convention that we follow in this document. When
413 demonstrating instructions, we will follow an instruction with a comment that
414 defines the type and name of value produced. Comments are shown in italic
419 <!-- *********************************************************************** -->
420 <div class="doc_section"> <a name="highlevel">High Level Structure</a> </div>
421 <!-- *********************************************************************** -->
423 <!-- ======================================================================= -->
424 <div class="doc_subsection"> <a name="modulestructure">Module Structure</a>
427 <div class="doc_text">
429 <p>LLVM programs are composed of "Module"s, each of which is a
430 translation unit of the input programs. Each module consists of
431 functions, global variables, and symbol table entries. Modules may be
432 combined together with the LLVM linker, which merges function (and
433 global variable) definitions, resolves forward declarations, and merges
434 symbol table entries. Here is an example of the "hello world" module:</p>
436 <div class="doc_code">
437 <pre><i>; Declare the string constant as a global constant...</i>
438 <a href="#identifiers">@.LC0</a> = <a href="#linkage_internal">internal</a> <a
439 href="#globalvars">constant</a> <a href="#t_array">[13 x i8]</a> c"hello world\0A\00" <i>; [13 x i8]*</i>
441 <i>; External declaration of the puts function</i>
442 <a href="#functionstructure">declare</a> i32 @puts(i8 *) <i>; i32(i8 *)* </i>
444 <i>; Definition of main function</i>
445 define i32 @main() { <i>; i32()* </i>
446 <i>; Convert [13 x i8]* to i8 *...</i>
448 href="#i_getelementptr">getelementptr</a> [13 x i8]* @.LC0, i64 0, i64 0 <i>; i8 *</i>
450 <i>; Call puts function to write out the string to stdout...</i>
452 href="#i_call">call</a> i32 @puts(i8 * %cast210) <i>; i32</i>
454 href="#i_ret">ret</a> i32 0<br>}<br>
458 <p>This example is made up of a <a href="#globalvars">global variable</a>
459 named "<tt>.LC0</tt>", an external declaration of the "<tt>puts</tt>"
460 function, and a <a href="#functionstructure">function definition</a>
461 for "<tt>main</tt>".</p>
463 <p>In general, a module is made up of a list of global values,
464 where both functions and global variables are global values. Global values are
465 represented by a pointer to a memory location (in this case, a pointer to an
466 array of char, and a pointer to a function), and have one of the following <a
467 href="#linkage">linkage types</a>.</p>
471 <!-- ======================================================================= -->
472 <div class="doc_subsection">
473 <a name="linkage">Linkage Types</a>
476 <div class="doc_text">
479 All Global Variables and Functions have one of the following types of linkage:
484 <dt><tt><b><a name="linkage_private">private</a></b></tt>: </dt>
486 <dd>Global values with private linkage are only directly accessible by
487 objects in the current module. In particular, linking code into a module with
488 an private global value may cause the private to be renamed as necessary to
489 avoid collisions. Because the symbol is private to the module, all
490 references can be updated. This doesn't show up in any symbol table in the
494 <dt><tt><b><a name="linkage_internal">internal</a></b></tt>: </dt>
496 <dd> Similar to private, but the value shows as a local symbol (STB_LOCAL in
497 the case of ELF) in the object file. This corresponds to the notion of the
498 '<tt>static</tt>' keyword in C.
501 <dt><tt><b><a name="linkage_linkonce">linkonce</a></b></tt>: </dt>
503 <dd>Globals with "<tt>linkonce</tt>" linkage are merged with other globals of
504 the same name when linkage occurs. This is typically used to implement
505 inline functions, templates, or other code which must be generated in each
506 translation unit that uses it. Unreferenced <tt>linkonce</tt> globals are
507 allowed to be discarded.
510 <dt><tt><b><a name="linkage_common">common</a></b></tt>: </dt>
512 <dd>"<tt>common</tt>" linkage is exactly the same as <tt>linkonce</tt>
513 linkage, except that unreferenced <tt>common</tt> globals may not be
514 discarded. This is used for globals that may be emitted in multiple
515 translation units, but that are not guaranteed to be emitted into every
516 translation unit that uses them. One example of this is tentative
517 definitions in C, such as "<tt>int X;</tt>" at global scope.
520 <dt><tt><b><a name="linkage_weak">weak</a></b></tt>: </dt>
522 <dd>"<tt>weak</tt>" linkage is the same as <tt>common</tt> linkage, except
523 that some targets may choose to emit different assembly sequences for them
524 for target-dependent reasons. This is used for globals that are declared
525 "weak" in C source code.
528 <dt><tt><b><a name="linkage_appending">appending</a></b></tt>: </dt>
530 <dd>"<tt>appending</tt>" linkage may only be applied to global variables of
531 pointer to array type. When two global variables with appending linkage are
532 linked together, the two global arrays are appended together. This is the
533 LLVM, typesafe, equivalent of having the system linker append together
534 "sections" with identical names when .o files are linked.
537 <dt><tt><b><a name="linkage_externweak">extern_weak</a></b></tt>: </dt>
538 <dd>The semantics of this linkage follow the ELF object file model: the
539 symbol is weak until linked, if not linked, the symbol becomes null instead
540 of being an undefined reference.
543 <dt><tt><b><a name="linkage_external">externally visible</a></b></tt>:</dt>
545 <dd>If none of the above identifiers are used, the global is externally
546 visible, meaning that it participates in linkage and can be used to resolve
547 external symbol references.
552 The next two types of linkage are targeted for Microsoft Windows platform
553 only. They are designed to support importing (exporting) symbols from (to)
554 DLLs (Dynamic Link Libraries).
558 <dt><tt><b><a name="linkage_dllimport">dllimport</a></b></tt>: </dt>
560 <dd>"<tt>dllimport</tt>" linkage causes the compiler to reference a function
561 or variable via a global pointer to a pointer that is set up by the DLL
562 exporting the symbol. On Microsoft Windows targets, the pointer name is
563 formed by combining <code>__imp_</code> and the function or variable name.
566 <dt><tt><b><a name="linkage_dllexport">dllexport</a></b></tt>: </dt>
568 <dd>"<tt>dllexport</tt>" linkage causes the compiler to provide a global
569 pointer to a pointer in a DLL, so that it can be referenced with the
570 <tt>dllimport</tt> attribute. On Microsoft Windows targets, the pointer
571 name is formed by combining <code>__imp_</code> and the function or variable
577 <p>For example, since the "<tt>.LC0</tt>"
578 variable is defined to be internal, if another module defined a "<tt>.LC0</tt>"
579 variable and was linked with this one, one of the two would be renamed,
580 preventing a collision. Since "<tt>main</tt>" and "<tt>puts</tt>" are
581 external (i.e., lacking any linkage declarations), they are accessible
582 outside of the current module.</p>
583 <p>It is illegal for a function <i>declaration</i>
584 to have any linkage type other than "externally visible", <tt>dllimport</tt>,
585 or <tt>extern_weak</tt>.</p>
586 <p>Aliases can have only <tt>external</tt>, <tt>internal</tt> and <tt>weak</tt>
590 <!-- ======================================================================= -->
591 <div class="doc_subsection">
592 <a name="callingconv">Calling Conventions</a>
595 <div class="doc_text">
597 <p>LLVM <a href="#functionstructure">functions</a>, <a href="#i_call">calls</a>
598 and <a href="#i_invoke">invokes</a> can all have an optional calling convention
599 specified for the call. The calling convention of any pair of dynamic
600 caller/callee must match, or the behavior of the program is undefined. The
601 following calling conventions are supported by LLVM, and more may be added in
605 <dt><b>"<tt>ccc</tt>" - The C calling convention</b>:</dt>
607 <dd>This calling convention (the default if no other calling convention is
608 specified) matches the target C calling conventions. This calling convention
609 supports varargs function calls and tolerates some mismatch in the declared
610 prototype and implemented declaration of the function (as does normal C).
613 <dt><b>"<tt>fastcc</tt>" - The fast calling convention</b>:</dt>
615 <dd>This calling convention attempts to make calls as fast as possible
616 (e.g. by passing things in registers). This calling convention allows the
617 target to use whatever tricks it wants to produce fast code for the target,
618 without having to conform to an externally specified ABI (Application Binary
619 Interface). Implementations of this convention should allow arbitrary
620 <a href="CodeGenerator.html#tailcallopt">tail call optimization</a> to be
621 supported. This calling convention does not support varargs and requires the
622 prototype of all callees to exactly match the prototype of the function
626 <dt><b>"<tt>coldcc</tt>" - The cold calling convention</b>:</dt>
628 <dd>This calling convention attempts to make code in the caller as efficient
629 as possible under the assumption that the call is not commonly executed. As
630 such, these calls often preserve all registers so that the call does not break
631 any live ranges in the caller side. This calling convention does not support
632 varargs and requires the prototype of all callees to exactly match the
633 prototype of the function definition.
636 <dt><b>"<tt>cc <<em>n</em>></tt>" - Numbered convention</b>:</dt>
638 <dd>Any calling convention may be specified by number, allowing
639 target-specific calling conventions to be used. Target specific calling
640 conventions start at 64.
644 <p>More calling conventions can be added/defined on an as-needed basis, to
645 support pascal conventions or any other well-known target-independent
650 <!-- ======================================================================= -->
651 <div class="doc_subsection">
652 <a name="visibility">Visibility Styles</a>
655 <div class="doc_text">
658 All Global Variables and Functions have one of the following visibility styles:
662 <dt><b>"<tt>default</tt>" - Default style</b>:</dt>
664 <dd>On targets that use the ELF object file format, default visibility means
665 that the declaration is visible to other
666 modules and, in shared libraries, means that the declared entity may be
667 overridden. On Darwin, default visibility means that the declaration is
668 visible to other modules. Default visibility corresponds to "external
669 linkage" in the language.
672 <dt><b>"<tt>hidden</tt>" - Hidden style</b>:</dt>
674 <dd>Two declarations of an object with hidden visibility refer to the same
675 object if they are in the same shared object. Usually, hidden visibility
676 indicates that the symbol will not be placed into the dynamic symbol table,
677 so no other module (executable or shared library) can reference it
681 <dt><b>"<tt>protected</tt>" - Protected style</b>:</dt>
683 <dd>On ELF, protected visibility indicates that the symbol will be placed in
684 the dynamic symbol table, but that references within the defining module will
685 bind to the local symbol. That is, the symbol cannot be overridden by another
692 <!-- ======================================================================= -->
693 <div class="doc_subsection">
694 <a name="namedtypes">Named Types</a>
697 <div class="doc_text">
699 <p>LLVM IR allows you to specify name aliases for certain types. This can make
700 it easier to read the IR and make the IR more condensed (particularly when
701 recursive types are involved). An example of a name specification is:
704 <div class="doc_code">
706 %mytype = type { %mytype*, i32 }
710 <p>You may give a name to any <a href="#typesystem">type</a> except "<a
711 href="t_void">void</a>". Type name aliases may be used anywhere a type is
712 expected with the syntax "%mytype".</p>
714 <p>Note that type names are aliases for the structural type that they indicate,
715 and that you can therefore specify multiple names for the same type. This often
716 leads to confusing behavior when dumping out a .ll file. Since LLVM IR uses
717 structural typing, the name is not part of the type. When printing out LLVM IR,
718 the printer will pick <em>one name</em> to render all types of a particular
719 shape. This means that if you have code where two different source types end up
720 having the same LLVM type, that the dumper will sometimes print the "wrong" or
721 unexpected type. This is an important design point and isn't going to
726 <!-- ======================================================================= -->
727 <div class="doc_subsection">
728 <a name="globalvars">Global Variables</a>
731 <div class="doc_text">
733 <p>Global variables define regions of memory allocated at compilation time
734 instead of run-time. Global variables may optionally be initialized, may have
735 an explicit section to be placed in, and may have an optional explicit alignment
736 specified. A variable may be defined as "thread_local", which means that it
737 will not be shared by threads (each thread will have a separated copy of the
738 variable). A variable may be defined as a global "constant," which indicates
739 that the contents of the variable will <b>never</b> be modified (enabling better
740 optimization, allowing the global data to be placed in the read-only section of
741 an executable, etc). Note that variables that need runtime initialization
742 cannot be marked "constant" as there is a store to the variable.</p>
745 LLVM explicitly allows <em>declarations</em> of global variables to be marked
746 constant, even if the final definition of the global is not. This capability
747 can be used to enable slightly better optimization of the program, but requires
748 the language definition to guarantee that optimizations based on the
749 'constantness' are valid for the translation units that do not include the
753 <p>As SSA values, global variables define pointer values that are in
754 scope (i.e. they dominate) all basic blocks in the program. Global
755 variables always define a pointer to their "content" type because they
756 describe a region of memory, and all memory objects in LLVM are
757 accessed through pointers.</p>
759 <p>A global variable may be declared to reside in a target-specifc numbered
760 address space. For targets that support them, address spaces may affect how
761 optimizations are performed and/or what target instructions are used to access
762 the variable. The default address space is zero. The address space qualifier
763 must precede any other attributes.</p>
765 <p>LLVM allows an explicit section to be specified for globals. If the target
766 supports it, it will emit globals to the section specified.</p>
768 <p>An explicit alignment may be specified for a global. If not present, or if
769 the alignment is set to zero, the alignment of the global is set by the target
770 to whatever it feels convenient. If an explicit alignment is specified, the
771 global is forced to have at least that much alignment. All alignments must be
774 <p>For example, the following defines a global in a numbered address space with
775 an initializer, section, and alignment:</p>
777 <div class="doc_code">
779 @G = addrspace(5) constant float 1.0, section "foo", align 4
786 <!-- ======================================================================= -->
787 <div class="doc_subsection">
788 <a name="functionstructure">Functions</a>
791 <div class="doc_text">
793 <p>LLVM function definitions consist of the "<tt>define</tt>" keyord,
794 an optional <a href="#linkage">linkage type</a>, an optional
795 <a href="#visibility">visibility style</a>, an optional
796 <a href="#callingconv">calling convention</a>, a return type, an optional
797 <a href="#paramattrs">parameter attribute</a> for the return type, a function
798 name, a (possibly empty) argument list (each with optional
799 <a href="#paramattrs">parameter attributes</a>), optional
800 <a href="#fnattrs">function attributes</a>, an optional section,
801 an optional alignment, an optional <a href="#gc">garbage collector name</a>,
802 an opening curly brace, a list of basic blocks, and a closing curly brace.
804 LLVM function declarations consist of the "<tt>declare</tt>" keyword, an
805 optional <a href="#linkage">linkage type</a>, an optional
806 <a href="#visibility">visibility style</a>, an optional
807 <a href="#callingconv">calling convention</a>, a return type, an optional
808 <a href="#paramattrs">parameter attribute</a> for the return type, a function
809 name, a possibly empty list of arguments, an optional alignment, and an optional
810 <a href="#gc">garbage collector name</a>.</p>
812 <p>A function definition contains a list of basic blocks, forming the CFG
813 (Control Flow Graph) for
814 the function. Each basic block may optionally start with a label (giving the
815 basic block a symbol table entry), contains a list of instructions, and ends
816 with a <a href="#terminators">terminator</a> instruction (such as a branch or
817 function return).</p>
819 <p>The first basic block in a function is special in two ways: it is immediately
820 executed on entrance to the function, and it is not allowed to have predecessor
821 basic blocks (i.e. there can not be any branches to the entry block of a
822 function). Because the block can have no predecessors, it also cannot have any
823 <a href="#i_phi">PHI nodes</a>.</p>
825 <p>LLVM allows an explicit section to be specified for functions. If the target
826 supports it, it will emit functions to the section specified.</p>
828 <p>An explicit alignment may be specified for a function. If not present, or if
829 the alignment is set to zero, the alignment of the function is set by the target
830 to whatever it feels convenient. If an explicit alignment is specified, the
831 function is forced to have at least that much alignment. All alignments must be
836 <div class="doc_code">
838 define [<a href="#linkage">linkage</a>] [<a href="#visibility">visibility</a>]
839 [<a href="#callingconv">cconv</a>] [<a href="#paramattrs">ret attrs</a>]
840 <ResultType> @<FunctionName> ([argument list])
841 [<a href="#fnattrs">fn Attrs</a>] [section "name"] [align N]
842 [<a href="#gc">gc</a>] { ... }
849 <!-- ======================================================================= -->
850 <div class="doc_subsection">
851 <a name="aliasstructure">Aliases</a>
853 <div class="doc_text">
854 <p>Aliases act as "second name" for the aliasee value (which can be either
855 function, global variable, another alias or bitcast of global value). Aliases
856 may have an optional <a href="#linkage">linkage type</a>, and an
857 optional <a href="#visibility">visibility style</a>.</p>
861 <div class="doc_code">
863 @<Name> = alias [Linkage] [Visibility] <AliaseeTy> @<Aliasee>
871 <!-- ======================================================================= -->
872 <div class="doc_subsection"><a name="paramattrs">Parameter Attributes</a></div>
873 <div class="doc_text">
874 <p>The return type and each parameter of a function type may have a set of
875 <i>parameter attributes</i> associated with them. Parameter attributes are
876 used to communicate additional information about the result or parameters of
877 a function. Parameter attributes are considered to be part of the function,
878 not of the function type, so functions with different parameter attributes
879 can have the same function type.</p>
881 <p>Parameter attributes are simple keywords that follow the type specified. If
882 multiple parameter attributes are needed, they are space separated. For
885 <div class="doc_code">
887 declare i32 @printf(i8* noalias , ...)
888 declare i32 @atoi(i8 zeroext)
889 declare signext i8 @returns_signed_char()
893 <p>Note that any attributes for the function result (<tt>nounwind</tt>,
894 <tt>readonly</tt>) come immediately after the argument list.</p>
896 <p>Currently, only the following parameter attributes are defined:</p>
898 <dt><tt>zeroext</tt></dt>
899 <dd>This indicates to the code generator that the parameter or return value
900 should be zero-extended to a 32-bit value by the caller (for a parameter)
901 or the callee (for a return value).</dd>
903 <dt><tt>signext</tt></dt>
904 <dd>This indicates to the code generator that the parameter or return value
905 should be sign-extended to a 32-bit value by the caller (for a parameter)
906 or the callee (for a return value).</dd>
908 <dt><tt>inreg</tt></dt>
909 <dd>This indicates that this parameter or return value should be treated
910 in a special target-dependent fashion during while emitting code for a
911 function call or return (usually, by putting it in a register as opposed
912 to memory, though some targets use it to distinguish between two different
913 kinds of registers). Use of this attribute is target-specific.</dd>
915 <dt><tt><a name="byval">byval</a></tt></dt>
916 <dd>This indicates that the pointer parameter should really be passed by
917 value to the function. The attribute implies that a hidden copy of the
918 pointee is made between the caller and the callee, so the callee is unable
919 to modify the value in the callee. This attribute is only valid on LLVM
920 pointer arguments. It is generally used to pass structs and arrays by
921 value, but is also valid on pointers to scalars. The copy is considered to
922 belong to the caller not the callee (for example,
923 <tt><a href="#readonly">readonly</a></tt> functions should not write to
924 <tt>byval</tt> parameters). This is not a valid attribute for return
927 <dt><tt>sret</tt></dt>
928 <dd>This indicates that the pointer parameter specifies the address of a
929 structure that is the return value of the function in the source program.
930 This pointer must be guaranteed by the caller to be valid: loads and stores
931 to the structure may be assumed by the callee to not to trap. This may only
932 be applied to the first parameter. This is not a valid attribute for
935 <dt><tt>noalias</tt></dt>
936 <dd>This indicates that the pointer does not alias any global or any other
937 parameter. The caller is responsible for ensuring that this is the
938 case. On a function return value, <tt>noalias</tt> additionally indicates
939 that the pointer does not alias any other pointers visible to the
940 caller. For further details, please see the discussion of the NoAlias
942 <a href="http://llvm.org/docs/AliasAnalysis.html#MustMayNo">alias
945 <dt><tt>nocapture</tt></dt>
946 <dd>This indicates that the callee does not make any copies of the pointer
947 that outlive the callee itself. This is not a valid attribute for return
950 <dt><tt>nest</tt></dt>
951 <dd>This indicates that the pointer parameter can be excised using the
952 <a href="#int_trampoline">trampoline intrinsics</a>. This is not a valid
953 attribute for return values.</dd>
958 <!-- ======================================================================= -->
959 <div class="doc_subsection">
960 <a name="gc">Garbage Collector Names</a>
963 <div class="doc_text">
964 <p>Each function may specify a garbage collector name, which is simply a
967 <div class="doc_code"><pre
968 >define void @f() gc "name" { ...</pre></div>
970 <p>The compiler declares the supported values of <i>name</i>. Specifying a
971 collector which will cause the compiler to alter its output in order to support
972 the named garbage collection algorithm.</p>
975 <!-- ======================================================================= -->
976 <div class="doc_subsection">
977 <a name="fnattrs">Function Attributes</a>
980 <div class="doc_text">
982 <p>Function attributes are set to communicate additional information about
983 a function. Function attributes are considered to be part of the function,
984 not of the function type, so functions with different parameter attributes
985 can have the same function type.</p>
987 <p>Function attributes are simple keywords that follow the type specified. If
988 multiple attributes are needed, they are space separated. For
991 <div class="doc_code">
993 define void @f() noinline { ... }
994 define void @f() alwaysinline { ... }
995 define void @f() alwaysinline optsize { ... }
996 define void @f() optsize
1001 <dt><tt>alwaysinline</tt></dt>
1002 <dd>This attribute indicates that the inliner should attempt to inline this
1003 function into callers whenever possible, ignoring any active inlining size
1004 threshold for this caller.</dd>
1006 <dt><tt>noinline</tt></dt>
1007 <dd>This attribute indicates that the inliner should never inline this function
1008 in any situation. This attribute may not be used together with the
1009 <tt>alwaysinline</tt> attribute.</dd>
1011 <dt><tt>optsize</tt></dt>
1012 <dd>This attribute suggests that optimization passes and code generator passes
1013 make choices that keep the code size of this function low, and otherwise do
1014 optimizations specifically to reduce code size.</dd>
1016 <dt><tt>noreturn</tt></dt>
1017 <dd>This function attribute indicates that the function never returns normally.
1018 This produces undefined behavior at runtime if the function ever does
1019 dynamically return.</dd>
1021 <dt><tt>nounwind</tt></dt>
1022 <dd>This function attribute indicates that the function never returns with an
1023 unwind or exceptional control flow. If the function does unwind, its runtime
1024 behavior is undefined.</dd>
1026 <dt><tt>readnone</tt></dt>
1027 <dd>This attribute indicates that the function computes its result (or the
1028 exception it throws) based strictly on its arguments, without dereferencing any
1029 pointer arguments or otherwise accessing any mutable state (e.g. memory, control
1030 registers, etc) visible to caller functions. It does not write through any
1031 pointer arguments (including <tt><a href="#byval">byval</a></tt> arguments) and
1032 never changes any state visible to callers.</dd>
1034 <dt><tt><a name="readonly">readonly</a></tt></dt>
1035 <dd>This attribute indicates that the function does not write through any
1036 pointer arguments (including <tt><a href="#byval">byval</a></tt> arguments)
1037 or otherwise modify any state (e.g. memory, control registers, etc) visible to
1038 caller functions. It may dereference pointer arguments and read state that may
1039 be set in the caller. A readonly function always returns the same value (or
1040 throws the same exception) when called with the same set of arguments and global
1043 <dt><tt><a name="ssp">ssp</a></tt></dt>
1044 <dd>This attribute indicates that the function should emit a stack smashing
1045 protector. It is in the form of a "canary"—a random value placed on the
1046 stack before the local variables that's checked upon return from the function to
1047 see if it has been overwritten. A heuristic is used to determine if a function
1048 needs stack protectors or not.
1050 <p>If a function that has an <tt>ssp</tt> attribute is inlined into a function
1051 that doesn't have an <tt>ssp</tt> attribute, then the resulting function will
1052 have an <tt>ssp</tt> attribute.</p></dd>
1054 <dt><tt>sspreq</tt></dt>
1055 <dd>This attribute indicates that the function should <em>always</em> emit a
1056 stack smashing protector. This overrides the <tt><a href="#ssp">ssp</a></tt>
1059 <p>If a function that has an <tt>sspreq</tt> attribute is inlined into a
1060 function that doesn't have an <tt>sspreq</tt> attribute or which has
1061 an <tt>ssp</tt> attribute, then the resulting function will have
1062 an <tt>sspreq</tt> attribute.</p></dd>
1067 <!-- ======================================================================= -->
1068 <div class="doc_subsection">
1069 <a name="moduleasm">Module-Level Inline Assembly</a>
1072 <div class="doc_text">
1074 Modules may contain "module-level inline asm" blocks, which corresponds to the
1075 GCC "file scope inline asm" blocks. These blocks are internally concatenated by
1076 LLVM and treated as a single unit, but may be separated in the .ll file if
1077 desired. The syntax is very simple:
1080 <div class="doc_code">
1082 module asm "inline asm code goes here"
1083 module asm "more can go here"
1087 <p>The strings can contain any character by escaping non-printable characters.
1088 The escape sequence used is simply "\xx" where "xx" is the two digit hex code
1093 The inline asm code is simply printed to the machine code .s file when
1094 assembly code is generated.
1098 <!-- ======================================================================= -->
1099 <div class="doc_subsection">
1100 <a name="datalayout">Data Layout</a>
1103 <div class="doc_text">
1104 <p>A module may specify a target specific data layout string that specifies how
1105 data is to be laid out in memory. The syntax for the data layout is simply:</p>
1106 <pre> target datalayout = "<i>layout specification</i>"</pre>
1107 <p>The <i>layout specification</i> consists of a list of specifications
1108 separated by the minus sign character ('-'). Each specification starts with a
1109 letter and may include other information after the letter to define some
1110 aspect of the data layout. The specifications accepted are as follows: </p>
1113 <dd>Specifies that the target lays out data in big-endian form. That is, the
1114 bits with the most significance have the lowest address location.</dd>
1116 <dd>Specifies that the target lays out data in little-endian form. That is,
1117 the bits with the least significance have the lowest address location.</dd>
1118 <dt><tt>p:<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1119 <dd>This specifies the <i>size</i> of a pointer and its <i>abi</i> and
1120 <i>preferred</i> alignments. All sizes are in bits. Specifying the <i>pref</i>
1121 alignment is optional. If omitted, the preceding <tt>:</tt> should be omitted
1123 <dt><tt>i<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1124 <dd>This specifies the alignment for an integer type of a given bit
1125 <i>size</i>. The value of <i>size</i> must be in the range [1,2^23).</dd>
1126 <dt><tt>v<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1127 <dd>This specifies the alignment for a vector type of a given bit
1129 <dt><tt>f<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1130 <dd>This specifies the alignment for a floating point type of a given bit
1131 <i>size</i>. The value of <i>size</i> must be either 32 (float) or 64
1133 <dt><tt>a<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1134 <dd>This specifies the alignment for an aggregate type of a given bit
1137 <p>When constructing the data layout for a given target, LLVM starts with a
1138 default set of specifications which are then (possibly) overriden by the
1139 specifications in the <tt>datalayout</tt> keyword. The default specifications
1140 are given in this list:</p>
1142 <li><tt>E</tt> - big endian</li>
1143 <li><tt>p:32:64:64</tt> - 32-bit pointers with 64-bit alignment</li>
1144 <li><tt>i1:8:8</tt> - i1 is 8-bit (byte) aligned</li>
1145 <li><tt>i8:8:8</tt> - i8 is 8-bit (byte) aligned</li>
1146 <li><tt>i16:16:16</tt> - i16 is 16-bit aligned</li>
1147 <li><tt>i32:32:32</tt> - i32 is 32-bit aligned</li>
1148 <li><tt>i64:32:64</tt> - i64 has ABI alignment of 32-bits but preferred
1149 alignment of 64-bits</li>
1150 <li><tt>f32:32:32</tt> - float is 32-bit aligned</li>
1151 <li><tt>f64:64:64</tt> - double is 64-bit aligned</li>
1152 <li><tt>v64:64:64</tt> - 64-bit vector is 64-bit aligned</li>
1153 <li><tt>v128:128:128</tt> - 128-bit vector is 128-bit aligned</li>
1154 <li><tt>a0:0:1</tt> - aggregates are 8-bit aligned</li>
1156 <p>When LLVM is determining the alignment for a given type, it uses the
1157 following rules:</p>
1159 <li>If the type sought is an exact match for one of the specifications, that
1160 specification is used.</li>
1161 <li>If no match is found, and the type sought is an integer type, then the
1162 smallest integer type that is larger than the bitwidth of the sought type is
1163 used. If none of the specifications are larger than the bitwidth then the the
1164 largest integer type is used. For example, given the default specifications
1165 above, the i7 type will use the alignment of i8 (next largest) while both
1166 i65 and i256 will use the alignment of i64 (largest specified).</li>
1167 <li>If no match is found, and the type sought is a vector type, then the
1168 largest vector type that is smaller than the sought vector type will be used
1169 as a fall back. This happens because <128 x double> can be implemented
1170 in terms of 64 <2 x double>, for example.</li>
1174 <!-- *********************************************************************** -->
1175 <div class="doc_section"> <a name="typesystem">Type System</a> </div>
1176 <!-- *********************************************************************** -->
1178 <div class="doc_text">
1180 <p>The LLVM type system is one of the most important features of the
1181 intermediate representation. Being typed enables a number of
1182 optimizations to be performed on the intermediate representation directly,
1183 without having to do
1184 extra analyses on the side before the transformation. A strong type
1185 system makes it easier to read the generated code and enables novel
1186 analyses and transformations that are not feasible to perform on normal
1187 three address code representations.</p>
1191 <!-- ======================================================================= -->
1192 <div class="doc_subsection"> <a name="t_classifications">Type
1193 Classifications</a> </div>
1194 <div class="doc_text">
1195 <p>The types fall into a few useful
1196 classifications:</p>
1198 <table border="1" cellspacing="0" cellpadding="4">
1200 <tr><th>Classification</th><th>Types</th></tr>
1202 <td><a href="#t_integer">integer</a></td>
1203 <td><tt>i1, i2, i3, ... i8, ... i16, ... i32, ... i64, ... </tt></td>
1206 <td><a href="#t_floating">floating point</a></td>
1207 <td><tt>float, double, x86_fp80, fp128, ppc_fp128</tt></td>
1210 <td><a name="t_firstclass">first class</a></td>
1211 <td><a href="#t_integer">integer</a>,
1212 <a href="#t_floating">floating point</a>,
1213 <a href="#t_pointer">pointer</a>,
1214 <a href="#t_vector">vector</a>,
1215 <a href="#t_struct">structure</a>,
1216 <a href="#t_array">array</a>,
1217 <a href="#t_label">label</a>.
1221 <td><a href="#t_primitive">primitive</a></td>
1222 <td><a href="#t_label">label</a>,
1223 <a href="#t_void">void</a>,
1224 <a href="#t_floating">floating point</a>.</td>
1227 <td><a href="#t_derived">derived</a></td>
1228 <td><a href="#t_integer">integer</a>,
1229 <a href="#t_array">array</a>,
1230 <a href="#t_function">function</a>,
1231 <a href="#t_pointer">pointer</a>,
1232 <a href="#t_struct">structure</a>,
1233 <a href="#t_pstruct">packed structure</a>,
1234 <a href="#t_vector">vector</a>,
1235 <a href="#t_opaque">opaque</a>.
1241 <p>The <a href="#t_firstclass">first class</a> types are perhaps the
1242 most important. Values of these types are the only ones which can be
1243 produced by instructions, passed as arguments, or used as operands to
1247 <!-- ======================================================================= -->
1248 <div class="doc_subsection"> <a name="t_primitive">Primitive Types</a> </div>
1250 <div class="doc_text">
1251 <p>The primitive types are the fundamental building blocks of the LLVM
1256 <!-- _______________________________________________________________________ -->
1257 <div class="doc_subsubsection"> <a name="t_floating">Floating Point Types</a> </div>
1259 <div class="doc_text">
1262 <tr><th>Type</th><th>Description</th></tr>
1263 <tr><td><tt>float</tt></td><td>32-bit floating point value</td></tr>
1264 <tr><td><tt>double</tt></td><td>64-bit floating point value</td></tr>
1265 <tr><td><tt>fp128</tt></td><td>128-bit floating point value (112-bit mantissa)</td></tr>
1266 <tr><td><tt>x86_fp80</tt></td><td>80-bit floating point value (X87)</td></tr>
1267 <tr><td><tt>ppc_fp128</tt></td><td>128-bit floating point value (two 64-bits)</td></tr>
1272 <!-- _______________________________________________________________________ -->
1273 <div class="doc_subsubsection"> <a name="t_void">Void Type</a> </div>
1275 <div class="doc_text">
1277 <p>The void type does not represent any value and has no size.</p>
1286 <!-- _______________________________________________________________________ -->
1287 <div class="doc_subsubsection"> <a name="t_label">Label Type</a> </div>
1289 <div class="doc_text">
1291 <p>The label type represents code labels.</p>
1301 <!-- ======================================================================= -->
1302 <div class="doc_subsection"> <a name="t_derived">Derived Types</a> </div>
1304 <div class="doc_text">
1306 <p>The real power in LLVM comes from the derived types in the system.
1307 This is what allows a programmer to represent arrays, functions,
1308 pointers, and other useful types. Note that these derived types may be
1309 recursive: For example, it is possible to have a two dimensional array.</p>
1313 <!-- _______________________________________________________________________ -->
1314 <div class="doc_subsubsection"> <a name="t_integer">Integer Type</a> </div>
1316 <div class="doc_text">
1319 <p>The integer type is a very simple derived type that simply specifies an
1320 arbitrary bit width for the integer type desired. Any bit width from 1 bit to
1321 2^23-1 (about 8 million) can be specified.</p>
1329 <p>The number of bits the integer will occupy is specified by the <tt>N</tt>
1333 <table class="layout">
1336 <td><tt>i1</tt></td>
1337 <td>a single-bit integer.</td>
1339 <td><tt>i32</tt></td>
1340 <td>a 32-bit integer.</td>
1342 <td><tt>i1942652</tt></td>
1343 <td>a really big integer of over 1 million bits.</td>
1348 <p>Note that the code generator does not yet support large integer types
1349 to be used as function return types. The specific limit on how large a
1350 return type the code generator can currently handle is target-dependent;
1351 currently it's often 64 bits for 32-bit targets and 128 bits for 64-bit
1356 <!-- _______________________________________________________________________ -->
1357 <div class="doc_subsubsection"> <a name="t_array">Array Type</a> </div>
1359 <div class="doc_text">
1363 <p>The array type is a very simple derived type that arranges elements
1364 sequentially in memory. The array type requires a size (number of
1365 elements) and an underlying data type.</p>
1370 [<# elements> x <elementtype>]
1373 <p>The number of elements is a constant integer value; elementtype may
1374 be any type with a size.</p>
1377 <table class="layout">
1379 <td class="left"><tt>[40 x i32]</tt></td>
1380 <td class="left">Array of 40 32-bit integer values.</td>
1383 <td class="left"><tt>[41 x i32]</tt></td>
1384 <td class="left">Array of 41 32-bit integer values.</td>
1387 <td class="left"><tt>[4 x i8]</tt></td>
1388 <td class="left">Array of 4 8-bit integer values.</td>
1391 <p>Here are some examples of multidimensional arrays:</p>
1392 <table class="layout">
1394 <td class="left"><tt>[3 x [4 x i32]]</tt></td>
1395 <td class="left">3x4 array of 32-bit integer values.</td>
1398 <td class="left"><tt>[12 x [10 x float]]</tt></td>
1399 <td class="left">12x10 array of single precision floating point values.</td>
1402 <td class="left"><tt>[2 x [3 x [4 x i16]]]</tt></td>
1403 <td class="left">2x3x4 array of 16-bit integer values.</td>
1407 <p>Note that 'variable sized arrays' can be implemented in LLVM with a zero
1408 length array. Normally, accesses past the end of an array are undefined in
1409 LLVM (e.g. it is illegal to access the 5th element of a 3 element array).
1410 As a special case, however, zero length arrays are recognized to be variable
1411 length. This allows implementation of 'pascal style arrays' with the LLVM
1412 type "{ i32, [0 x float]}", for example.</p>
1414 <p>Note that the code generator does not yet support large aggregate types
1415 to be used as function return types. The specific limit on how large an
1416 aggregate return type the code generator can currently handle is
1417 target-dependent, and also dependent on the aggregate element types.</p>
1421 <!-- _______________________________________________________________________ -->
1422 <div class="doc_subsubsection"> <a name="t_function">Function Type</a> </div>
1423 <div class="doc_text">
1427 <p>The function type can be thought of as a function signature. It
1428 consists of a return type and a list of formal parameter types. The
1429 return type of a function type is a scalar type, a void type, or a struct type.
1430 If the return type is a struct type then all struct elements must be of first
1431 class types, and the struct must have at least one element.</p>
1436 <returntype list> (<parameter list>)
1439 <p>...where '<tt><parameter list></tt>' is a comma-separated list of type
1440 specifiers. Optionally, the parameter list may include a type <tt>...</tt>,
1441 which indicates that the function takes a variable number of arguments.
1442 Variable argument functions can access their arguments with the <a
1443 href="#int_varargs">variable argument handling intrinsic</a> functions.
1444 '<tt><returntype list></tt>' is a comma-separated list of
1445 <a href="#t_firstclass">first class</a> type specifiers.</p>
1448 <table class="layout">
1450 <td class="left"><tt>i32 (i32)</tt></td>
1451 <td class="left">function taking an <tt>i32</tt>, returning an <tt>i32</tt>
1453 </tr><tr class="layout">
1454 <td class="left"><tt>float (i16 signext, i32 *) *
1456 <td class="left"><a href="#t_pointer">Pointer</a> to a function that takes
1457 an <tt>i16</tt> that should be sign extended and a
1458 <a href="#t_pointer">pointer</a> to <tt>i32</tt>, returning
1461 </tr><tr class="layout">
1462 <td class="left"><tt>i32 (i8*, ...)</tt></td>
1463 <td class="left">A vararg function that takes at least one
1464 <a href="#t_pointer">pointer</a> to <tt>i8 </tt> (char in C),
1465 which returns an integer. This is the signature for <tt>printf</tt> in
1468 </tr><tr class="layout">
1469 <td class="left"><tt>{i32, i32} (i32)</tt></td>
1470 <td class="left">A function taking an <tt>i32</tt>, returning two
1471 <tt>i32</tt> values as an aggregate of type <tt>{ i32, i32 }</tt>
1477 <!-- _______________________________________________________________________ -->
1478 <div class="doc_subsubsection"> <a name="t_struct">Structure Type</a> </div>
1479 <div class="doc_text">
1481 <p>The structure type is used to represent a collection of data members
1482 together in memory. The packing of the field types is defined to match
1483 the ABI of the underlying processor. The elements of a structure may
1484 be any type that has a size.</p>
1485 <p>Structures are accessed using '<tt><a href="#i_load">load</a></tt>
1486 and '<tt><a href="#i_store">store</a></tt>' by getting a pointer to a
1487 field with the '<tt><a href="#i_getelementptr">getelementptr</a></tt>'
1490 <pre> { <type list> }<br></pre>
1492 <table class="layout">
1494 <td class="left"><tt>{ i32, i32, i32 }</tt></td>
1495 <td class="left">A triple of three <tt>i32</tt> values</td>
1496 </tr><tr class="layout">
1497 <td class="left"><tt>{ float, i32 (i32) * }</tt></td>
1498 <td class="left">A pair, where the first element is a <tt>float</tt> and the
1499 second element is a <a href="#t_pointer">pointer</a> to a
1500 <a href="#t_function">function</a> that takes an <tt>i32</tt>, returning
1501 an <tt>i32</tt>.</td>
1505 <p>Note that the code generator does not yet support large aggregate types
1506 to be used as function return types. The specific limit on how large an
1507 aggregate return type the code generator can currently handle is
1508 target-dependent, and also dependent on the aggregate element types.</p>
1512 <!-- _______________________________________________________________________ -->
1513 <div class="doc_subsubsection"> <a name="t_pstruct">Packed Structure Type</a>
1515 <div class="doc_text">
1517 <p>The packed structure type is used to represent a collection of data members
1518 together in memory. There is no padding between fields. Further, the alignment
1519 of a packed structure is 1 byte. The elements of a packed structure may
1520 be any type that has a size.</p>
1521 <p>Structures are accessed using '<tt><a href="#i_load">load</a></tt>
1522 and '<tt><a href="#i_store">store</a></tt>' by getting a pointer to a
1523 field with the '<tt><a href="#i_getelementptr">getelementptr</a></tt>'
1526 <pre> < { <type list> } > <br></pre>
1528 <table class="layout">
1530 <td class="left"><tt>< { i32, i32, i32 } ></tt></td>
1531 <td class="left">A triple of three <tt>i32</tt> values</td>
1532 </tr><tr class="layout">
1534 <tt>< { float, i32 (i32)* } ></tt></td>
1535 <td class="left">A pair, where the first element is a <tt>float</tt> and the
1536 second element is a <a href="#t_pointer">pointer</a> to a
1537 <a href="#t_function">function</a> that takes an <tt>i32</tt>, returning
1538 an <tt>i32</tt>.</td>
1543 <!-- _______________________________________________________________________ -->
1544 <div class="doc_subsubsection"> <a name="t_pointer">Pointer Type</a> </div>
1545 <div class="doc_text">
1547 <p>As in many languages, the pointer type represents a pointer or
1548 reference to another object, which must live in memory. Pointer types may have
1549 an optional address space attribute defining the target-specific numbered
1550 address space where the pointed-to object resides. The default address space is
1553 <pre> <type> *<br></pre>
1555 <table class="layout">
1557 <td class="left"><tt>[4 x i32]*</tt></td>
1558 <td class="left">A <a href="#t_pointer">pointer</a> to <a
1559 href="#t_array">array</a> of four <tt>i32</tt> values.</td>
1562 <td class="left"><tt>i32 (i32 *) *</tt></td>
1563 <td class="left"> A <a href="#t_pointer">pointer</a> to a <a
1564 href="#t_function">function</a> that takes an <tt>i32*</tt>, returning an
1568 <td class="left"><tt>i32 addrspace(5)*</tt></td>
1569 <td class="left">A <a href="#t_pointer">pointer</a> to an <tt>i32</tt> value
1570 that resides in address space #5.</td>
1575 <!-- _______________________________________________________________________ -->
1576 <div class="doc_subsubsection"> <a name="t_vector">Vector Type</a> </div>
1577 <div class="doc_text">
1581 <p>A vector type is a simple derived type that represents a vector
1582 of elements. Vector types are used when multiple primitive data
1583 are operated in parallel using a single instruction (SIMD).
1584 A vector type requires a size (number of
1585 elements) and an underlying primitive data type. Vectors must have a power
1586 of two length (1, 2, 4, 8, 16 ...). Vector types are
1587 considered <a href="#t_firstclass">first class</a>.</p>
1592 < <# elements> x <elementtype> >
1595 <p>The number of elements is a constant integer value; elementtype may
1596 be any integer or floating point type.</p>
1600 <table class="layout">
1602 <td class="left"><tt><4 x i32></tt></td>
1603 <td class="left">Vector of 4 32-bit integer values.</td>
1606 <td class="left"><tt><8 x float></tt></td>
1607 <td class="left">Vector of 8 32-bit floating-point values.</td>
1610 <td class="left"><tt><2 x i64></tt></td>
1611 <td class="left">Vector of 2 64-bit integer values.</td>
1615 <p>Note that the code generator does not yet support large vector types
1616 to be used as function return types. The specific limit on how large a
1617 vector return type codegen can currently handle is target-dependent;
1618 currently it's often a few times longer than a hardware vector register.</p>
1622 <!-- _______________________________________________________________________ -->
1623 <div class="doc_subsubsection"> <a name="t_opaque">Opaque Type</a> </div>
1624 <div class="doc_text">
1628 <p>Opaque types are used to represent unknown types in the system. This
1629 corresponds (for example) to the C notion of a forward declared structure type.
1630 In LLVM, opaque types can eventually be resolved to any type (not just a
1631 structure type).</p>
1641 <table class="layout">
1643 <td class="left"><tt>opaque</tt></td>
1644 <td class="left">An opaque type.</td>
1649 <!-- ======================================================================= -->
1650 <div class="doc_subsection">
1651 <a name="t_uprefs">Type Up-references</a>
1654 <div class="doc_text">
1657 An "up reference" allows you to refer to a lexically enclosing type without
1658 requiring it to have a name. For instance, a structure declaration may contain a
1659 pointer to any of the types it is lexically a member of. Example of up
1660 references (with their equivalent as named type declarations) include:</p>
1663 { \2 * } %x = type { %t* }
1664 { \2 }* %y = type { %y }*
1669 An up reference is needed by the asmprinter for printing out cyclic types when
1670 there is no declared name for a type in the cycle. Because the asmprinter does
1671 not want to print out an infinite type string, it needs a syntax to handle
1672 recursive types that have no names (all names are optional in llvm IR).
1681 The level is the count of the lexical type that is being referred to.
1686 <table class="layout">
1688 <td class="left"><tt>\1*</tt></td>
1689 <td class="left">Self-referential pointer.</td>
1692 <td class="left"><tt>{ { \3*, i8 }, i32 }</tt></td>
1693 <td class="left">Recursive structure where the upref refers to the out-most
1700 <!-- *********************************************************************** -->
1701 <div class="doc_section"> <a name="constants">Constants</a> </div>
1702 <!-- *********************************************************************** -->
1704 <div class="doc_text">
1706 <p>LLVM has several different basic types of constants. This section describes
1707 them all and their syntax.</p>
1711 <!-- ======================================================================= -->
1712 <div class="doc_subsection"><a name="simpleconstants">Simple Constants</a></div>
1714 <div class="doc_text">
1717 <dt><b>Boolean constants</b></dt>
1719 <dd>The two strings '<tt>true</tt>' and '<tt>false</tt>' are both valid
1720 constants of the <tt><a href="#t_primitive">i1</a></tt> type.
1723 <dt><b>Integer constants</b></dt>
1725 <dd>Standard integers (such as '4') are constants of the <a
1726 href="#t_integer">integer</a> type. Negative numbers may be used with
1730 <dt><b>Floating point constants</b></dt>
1732 <dd>Floating point constants use standard decimal notation (e.g. 123.421),
1733 exponential notation (e.g. 1.23421e+2), or a more precise hexadecimal
1734 notation (see below). The assembler requires the exact decimal value of
1735 a floating-point constant. For example, the assembler accepts 1.25 but
1736 rejects 1.3 because 1.3 is a repeating decimal in binary. Floating point
1737 constants must have a <a href="#t_floating">floating point</a> type. </dd>
1739 <dt><b>Null pointer constants</b></dt>
1741 <dd>The identifier '<tt>null</tt>' is recognized as a null pointer constant
1742 and must be of <a href="#t_pointer">pointer type</a>.</dd>
1746 <p>The one non-intuitive notation for constants is the optional hexadecimal form
1747 of floating point constants. For example, the form '<tt>double
1748 0x432ff973cafa8000</tt>' is equivalent to (but harder to read than) '<tt>double
1749 4.5e+15</tt>'. The only time hexadecimal floating point constants are required
1750 (and the only time that they are generated by the disassembler) is when a
1751 floating point constant must be emitted but it cannot be represented as a
1752 decimal floating point number. For example, NaN's, infinities, and other
1753 special values are represented in their IEEE hexadecimal format so that
1754 assembly and disassembly do not cause any bits to change in the constants.</p>
1758 <!-- ======================================================================= -->
1759 <div class="doc_subsection"><a name="aggregateconstants">Aggregate Constants</a>
1762 <div class="doc_text">
1763 <p>Aggregate constants arise from aggregation of simple constants
1764 and smaller aggregate constants.</p>
1767 <dt><b>Structure constants</b></dt>
1769 <dd>Structure constants are represented with notation similar to structure
1770 type definitions (a comma separated list of elements, surrounded by braces
1771 (<tt>{}</tt>)). For example: "<tt>{ i32 4, float 17.0, i32* @G }</tt>",
1772 where "<tt>@G</tt>" is declared as "<tt>@G = external global i32</tt>". Structure constants
1773 must have <a href="#t_struct">structure type</a>, and the number and
1774 types of elements must match those specified by the type.
1777 <dt><b>Array constants</b></dt>
1779 <dd>Array constants are represented with notation similar to array type
1780 definitions (a comma separated list of elements, surrounded by square brackets
1781 (<tt>[]</tt>)). For example: "<tt>[ i32 42, i32 11, i32 74 ]</tt>". Array
1782 constants must have <a href="#t_array">array type</a>, and the number and
1783 types of elements must match those specified by the type.
1786 <dt><b>Vector constants</b></dt>
1788 <dd>Vector constants are represented with notation similar to vector type
1789 definitions (a comma separated list of elements, surrounded by
1790 less-than/greater-than's (<tt><></tt>)). For example: "<tt>< i32 42,
1791 i32 11, i32 74, i32 100 ></tt>". Vector constants must have <a
1792 href="#t_vector">vector type</a>, and the number and types of elements must
1793 match those specified by the type.
1796 <dt><b>Zero initialization</b></dt>
1798 <dd>The string '<tt>zeroinitializer</tt>' can be used to zero initialize a
1799 value to zero of <em>any</em> type, including scalar and aggregate types.
1800 This is often used to avoid having to print large zero initializers (e.g. for
1801 large arrays) and is always exactly equivalent to using explicit zero
1808 <!-- ======================================================================= -->
1809 <div class="doc_subsection">
1810 <a name="globalconstants">Global Variable and Function Addresses</a>
1813 <div class="doc_text">
1815 <p>The addresses of <a href="#globalvars">global variables</a> and <a
1816 href="#functionstructure">functions</a> are always implicitly valid (link-time)
1817 constants. These constants are explicitly referenced when the <a
1818 href="#identifiers">identifier for the global</a> is used and always have <a
1819 href="#t_pointer">pointer</a> type. For example, the following is a legal LLVM
1822 <div class="doc_code">
1826 @Z = global [2 x i32*] [ i32* @X, i32* @Y ]
1832 <!-- ======================================================================= -->
1833 <div class="doc_subsection"><a name="undefvalues">Undefined Values</a></div>
1834 <div class="doc_text">
1835 <p>The string '<tt>undef</tt>' is recognized as a type-less constant that has
1836 no specific value. Undefined values may be of any type and be used anywhere
1837 a constant is permitted.</p>
1839 <p>Undefined values indicate to the compiler that the program is well defined
1840 no matter what value is used, giving the compiler more freedom to optimize.
1844 <!-- ======================================================================= -->
1845 <div class="doc_subsection"><a name="constantexprs">Constant Expressions</a>
1848 <div class="doc_text">
1850 <p>Constant expressions are used to allow expressions involving other constants
1851 to be used as constants. Constant expressions may be of any <a
1852 href="#t_firstclass">first class</a> type and may involve any LLVM operation
1853 that does not have side effects (e.g. load and call are not supported). The
1854 following is the syntax for constant expressions:</p>
1857 <dt><b><tt>trunc ( CST to TYPE )</tt></b></dt>
1858 <dd>Truncate a constant to another type. The bit size of CST must be larger
1859 than the bit size of TYPE. Both types must be integers.</dd>
1861 <dt><b><tt>zext ( CST to TYPE )</tt></b></dt>
1862 <dd>Zero extend a constant to another type. The bit size of CST must be
1863 smaller or equal to the bit size of TYPE. Both types must be integers.</dd>
1865 <dt><b><tt>sext ( CST to TYPE )</tt></b></dt>
1866 <dd>Sign extend a constant to another type. The bit size of CST must be
1867 smaller or equal to the bit size of TYPE. Both types must be integers.</dd>
1869 <dt><b><tt>fptrunc ( CST to TYPE )</tt></b></dt>
1870 <dd>Truncate a floating point constant to another floating point type. The
1871 size of CST must be larger than the size of TYPE. Both types must be
1872 floating point.</dd>
1874 <dt><b><tt>fpext ( CST to TYPE )</tt></b></dt>
1875 <dd>Floating point extend a constant to another type. The size of CST must be
1876 smaller or equal to the size of TYPE. Both types must be floating point.</dd>
1878 <dt><b><tt>fptoui ( CST to TYPE )</tt></b></dt>
1879 <dd>Convert a floating point constant to the corresponding unsigned integer
1880 constant. TYPE must be a scalar or vector integer type. CST must be of scalar
1881 or vector floating point type. Both CST and TYPE must be scalars, or vectors
1882 of the same number of elements. If the value won't fit in the integer type,
1883 the results are undefined.</dd>
1885 <dt><b><tt>fptosi ( CST to TYPE )</tt></b></dt>
1886 <dd>Convert a floating point constant to the corresponding signed integer
1887 constant. TYPE must be a scalar or vector integer type. CST must be of scalar
1888 or vector floating point type. Both CST and TYPE must be scalars, or vectors
1889 of the same number of elements. If the value won't fit in the integer type,
1890 the results are undefined.</dd>
1892 <dt><b><tt>uitofp ( CST to TYPE )</tt></b></dt>
1893 <dd>Convert an unsigned integer constant to the corresponding floating point
1894 constant. TYPE must be a scalar or vector floating point type. CST must be of
1895 scalar or vector integer type. Both CST and TYPE must be scalars, or vectors
1896 of the same number of elements. If the value won't fit in the floating point
1897 type, the results are undefined.</dd>
1899 <dt><b><tt>sitofp ( CST to TYPE )</tt></b></dt>
1900 <dd>Convert a signed integer constant to the corresponding floating point
1901 constant. TYPE must be a scalar or vector floating point type. CST must be of
1902 scalar or vector integer type. Both CST and TYPE must be scalars, or vectors
1903 of the same number of elements. If the value won't fit in the floating point
1904 type, the results are undefined.</dd>
1906 <dt><b><tt>ptrtoint ( CST to TYPE )</tt></b></dt>
1907 <dd>Convert a pointer typed constant to the corresponding integer constant
1908 TYPE must be an integer type. CST must be of pointer type. The CST value is
1909 zero extended, truncated, or unchanged to make it fit in TYPE.</dd>
1911 <dt><b><tt>inttoptr ( CST to TYPE )</tt></b></dt>
1912 <dd>Convert a integer constant to a pointer constant. TYPE must be a
1913 pointer type. CST must be of integer type. The CST value is zero extended,
1914 truncated, or unchanged to make it fit in a pointer size. This one is
1915 <i>really</i> dangerous!</dd>
1917 <dt><b><tt>bitcast ( CST to TYPE )</tt></b></dt>
1918 <dd>Convert a constant, CST, to another TYPE. The size of CST and TYPE must be
1919 identical (same number of bits). The conversion is done as if the CST value
1920 was stored to memory and read back as TYPE. In other words, no bits change
1921 with this operator, just the type. This can be used for conversion of
1922 vector types to any other type, as long as they have the same bit width. For
1923 pointers it is only valid to cast to another pointer type. It is not valid
1924 to bitcast to or from an aggregate type.
1927 <dt><b><tt>getelementptr ( CSTPTR, IDX0, IDX1, ... )</tt></b></dt>
1929 <dd>Perform the <a href="#i_getelementptr">getelementptr operation</a> on
1930 constants. As with the <a href="#i_getelementptr">getelementptr</a>
1931 instruction, the index list may have zero or more indexes, which are required
1932 to make sense for the type of "CSTPTR".</dd>
1934 <dt><b><tt>select ( COND, VAL1, VAL2 )</tt></b></dt>
1936 <dd>Perform the <a href="#i_select">select operation</a> on
1939 <dt><b><tt>icmp COND ( VAL1, VAL2 )</tt></b></dt>
1940 <dd>Performs the <a href="#i_icmp">icmp operation</a> on constants.</dd>
1942 <dt><b><tt>fcmp COND ( VAL1, VAL2 )</tt></b></dt>
1943 <dd>Performs the <a href="#i_fcmp">fcmp operation</a> on constants.</dd>
1945 <dt><b><tt>vicmp COND ( VAL1, VAL2 )</tt></b></dt>
1946 <dd>Performs the <a href="#i_vicmp">vicmp operation</a> on constants.</dd>
1948 <dt><b><tt>vfcmp COND ( VAL1, VAL2 )</tt></b></dt>
1949 <dd>Performs the <a href="#i_vfcmp">vfcmp operation</a> on constants.</dd>
1951 <dt><b><tt>extractelement ( VAL, IDX )</tt></b></dt>
1953 <dd>Perform the <a href="#i_extractelement">extractelement
1954 operation</a> on constants.</dd>
1956 <dt><b><tt>insertelement ( VAL, ELT, IDX )</tt></b></dt>
1958 <dd>Perform the <a href="#i_insertelement">insertelement
1959 operation</a> on constants.</dd>
1962 <dt><b><tt>shufflevector ( VEC1, VEC2, IDXMASK )</tt></b></dt>
1964 <dd>Perform the <a href="#i_shufflevector">shufflevector
1965 operation</a> on constants.</dd>
1967 <dt><b><tt>OPCODE ( LHS, RHS )</tt></b></dt>
1969 <dd>Perform the specified operation of the LHS and RHS constants. OPCODE may
1970 be any of the <a href="#binaryops">binary</a> or <a href="#bitwiseops">bitwise
1971 binary</a> operations. The constraints on operands are the same as those for
1972 the corresponding instruction (e.g. no bitwise operations on floating point
1973 values are allowed).</dd>
1977 <!-- *********************************************************************** -->
1978 <div class="doc_section"> <a name="othervalues">Other Values</a> </div>
1979 <!-- *********************************************************************** -->
1981 <!-- ======================================================================= -->
1982 <div class="doc_subsection">
1983 <a name="inlineasm">Inline Assembler Expressions</a>
1986 <div class="doc_text">
1989 LLVM supports inline assembler expressions (as opposed to <a href="#moduleasm">
1990 Module-Level Inline Assembly</a>) through the use of a special value. This
1991 value represents the inline assembler as a string (containing the instructions
1992 to emit), a list of operand constraints (stored as a string), and a flag that
1993 indicates whether or not the inline asm expression has side effects. An example
1994 inline assembler expression is:
1997 <div class="doc_code">
1999 i32 (i32) asm "bswap $0", "=r,r"
2004 Inline assembler expressions may <b>only</b> be used as the callee operand of
2005 a <a href="#i_call"><tt>call</tt> instruction</a>. Thus, typically we have:
2008 <div class="doc_code">
2010 %X = call i32 asm "<a href="#int_bswap">bswap</a> $0", "=r,r"(i32 %Y)
2015 Inline asms with side effects not visible in the constraint list must be marked
2016 as having side effects. This is done through the use of the
2017 '<tt>sideeffect</tt>' keyword, like so:
2020 <div class="doc_code">
2022 call void asm sideeffect "eieio", ""()
2026 <p>TODO: The format of the asm and constraints string still need to be
2027 documented here. Constraints on what can be done (e.g. duplication, moving, etc
2028 need to be documented). This is probably best done by reference to another
2029 document that covers inline asm from a holistic perspective.
2034 <!-- *********************************************************************** -->
2035 <div class="doc_section"> <a name="instref">Instruction Reference</a> </div>
2036 <!-- *********************************************************************** -->
2038 <div class="doc_text">
2040 <p>The LLVM instruction set consists of several different
2041 classifications of instructions: <a href="#terminators">terminator
2042 instructions</a>, <a href="#binaryops">binary instructions</a>,
2043 <a href="#bitwiseops">bitwise binary instructions</a>, <a
2044 href="#memoryops">memory instructions</a>, and <a href="#otherops">other
2045 instructions</a>.</p>
2049 <!-- ======================================================================= -->
2050 <div class="doc_subsection"> <a name="terminators">Terminator
2051 Instructions</a> </div>
2053 <div class="doc_text">
2055 <p>As mentioned <a href="#functionstructure">previously</a>, every
2056 basic block in a program ends with a "Terminator" instruction, which
2057 indicates which block should be executed after the current block is
2058 finished. These terminator instructions typically yield a '<tt>void</tt>'
2059 value: they produce control flow, not values (the one exception being
2060 the '<a href="#i_invoke"><tt>invoke</tt></a>' instruction).</p>
2061 <p>There are six different terminator instructions: the '<a
2062 href="#i_ret"><tt>ret</tt></a>' instruction, the '<a href="#i_br"><tt>br</tt></a>'
2063 instruction, the '<a href="#i_switch"><tt>switch</tt></a>' instruction,
2064 the '<a href="#i_invoke"><tt>invoke</tt></a>' instruction, the '<a
2065 href="#i_unwind"><tt>unwind</tt></a>' instruction, and the '<a
2066 href="#i_unreachable"><tt>unreachable</tt></a>' instruction.</p>
2070 <!-- _______________________________________________________________________ -->
2071 <div class="doc_subsubsection"> <a name="i_ret">'<tt>ret</tt>'
2072 Instruction</a> </div>
2073 <div class="doc_text">
2076 ret <type> <value> <i>; Return a value from a non-void function</i>
2077 ret void <i>; Return from void function</i>
2082 <p>The '<tt>ret</tt>' instruction is used to return control flow (and
2083 optionally a value) from a function back to the caller.</p>
2084 <p>There are two forms of the '<tt>ret</tt>' instruction: one that
2085 returns a value and then causes control flow, and one that just causes
2086 control flow to occur.</p>
2090 <p>The '<tt>ret</tt>' instruction optionally accepts a single argument,
2091 the return value. The type of the return value must be a
2092 '<a href="#t_firstclass">first class</a>' type.</p>
2094 <p>A function is not <a href="#wellformed">well formed</a> if
2095 it it has a non-void return type and contains a '<tt>ret</tt>'
2096 instruction with no return value or a return value with a type that
2097 does not match its type, or if it has a void return type and contains
2098 a '<tt>ret</tt>' instruction with a return value.</p>
2102 <p>When the '<tt>ret</tt>' instruction is executed, control flow
2103 returns back to the calling function's context. If the caller is a "<a
2104 href="#i_call"><tt>call</tt></a>" instruction, execution continues at
2105 the instruction after the call. If the caller was an "<a
2106 href="#i_invoke"><tt>invoke</tt></a>" instruction, execution continues
2107 at the beginning of the "normal" destination block. If the instruction
2108 returns a value, that value shall set the call or invoke instruction's
2114 ret i32 5 <i>; Return an integer value of 5</i>
2115 ret void <i>; Return from a void function</i>
2116 ret { i32, i8 } { i32 4, i8 2 } <i>; Return an aggregate of values 4 and 2</i>
2119 <p>Note that the code generator does not yet fully support large
2120 return values. The specific sizes that are currently supported are
2121 dependent on the target. For integers, on 32-bit targets the limit
2122 is often 64 bits, and on 64-bit targets the limit is often 128 bits.
2123 For aggregate types, the current limits are dependent on the element
2124 types; for example targets are often limited to 2 total integer
2125 elements and 2 total floating-point elements.</p>
2128 <!-- _______________________________________________________________________ -->
2129 <div class="doc_subsubsection"> <a name="i_br">'<tt>br</tt>' Instruction</a> </div>
2130 <div class="doc_text">
2132 <pre> br i1 <cond>, label <iftrue>, label <iffalse><br> br label <dest> <i>; Unconditional branch</i>
2135 <p>The '<tt>br</tt>' instruction is used to cause control flow to
2136 transfer to a different basic block in the current function. There are
2137 two forms of this instruction, corresponding to a conditional branch
2138 and an unconditional branch.</p>
2140 <p>The conditional branch form of the '<tt>br</tt>' instruction takes a
2141 single '<tt>i1</tt>' value and two '<tt>label</tt>' values. The
2142 unconditional form of the '<tt>br</tt>' instruction takes a single
2143 '<tt>label</tt>' value as a target.</p>
2145 <p>Upon execution of a conditional '<tt>br</tt>' instruction, the '<tt>i1</tt>'
2146 argument is evaluated. If the value is <tt>true</tt>, control flows
2147 to the '<tt>iftrue</tt>' <tt>label</tt> argument. If "cond" is <tt>false</tt>,
2148 control flows to the '<tt>iffalse</tt>' <tt>label</tt> argument.</p>
2150 <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
2151 href="#i_ret">ret</a> i32 1<br>IfUnequal:<br> <a href="#i_ret">ret</a> i32 0<br></pre>
2153 <!-- _______________________________________________________________________ -->
2154 <div class="doc_subsubsection">
2155 <a name="i_switch">'<tt>switch</tt>' Instruction</a>
2158 <div class="doc_text">
2162 switch <intty> <value>, label <defaultdest> [ <intty> <val>, label <dest> ... ]
2167 <p>The '<tt>switch</tt>' instruction is used to transfer control flow to one of
2168 several different places. It is a generalization of the '<tt>br</tt>'
2169 instruction, allowing a branch to occur to one of many possible
2175 <p>The '<tt>switch</tt>' instruction uses three parameters: an integer
2176 comparison value '<tt>value</tt>', a default '<tt>label</tt>' destination, and
2177 an array of pairs of comparison value constants and '<tt>label</tt>'s. The
2178 table is not allowed to contain duplicate constant entries.</p>
2182 <p>The <tt>switch</tt> instruction specifies a table of values and
2183 destinations. When the '<tt>switch</tt>' instruction is executed, this
2184 table is searched for the given value. If the value is found, control flow is
2185 transfered to the corresponding destination; otherwise, control flow is
2186 transfered to the default destination.</p>
2188 <h5>Implementation:</h5>
2190 <p>Depending on properties of the target machine and the particular
2191 <tt>switch</tt> instruction, this instruction may be code generated in different
2192 ways. For example, it could be generated as a series of chained conditional
2193 branches or with a lookup table.</p>
2198 <i>; Emulate a conditional br instruction</i>
2199 %Val = <a href="#i_zext">zext</a> i1 %value to i32
2200 switch i32 %Val, label %truedest [ i32 0, label %falsedest ]
2202 <i>; Emulate an unconditional br instruction</i>
2203 switch i32 0, label %dest [ ]
2205 <i>; Implement a jump table:</i>
2206 switch i32 %val, label %otherwise [ i32 0, label %onzero
2208 i32 2, label %ontwo ]
2212 <!-- _______________________________________________________________________ -->
2213 <div class="doc_subsubsection">
2214 <a name="i_invoke">'<tt>invoke</tt>' Instruction</a>
2217 <div class="doc_text">
2222 <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>]
2223 to label <normal label> unwind label <exception label>
2228 <p>The '<tt>invoke</tt>' instruction causes control to transfer to a specified
2229 function, with the possibility of control flow transfer to either the
2230 '<tt>normal</tt>' label or the
2231 '<tt>exception</tt>' label. If the callee function returns with the
2232 "<tt><a href="#i_ret">ret</a></tt>" instruction, control flow will return to the
2233 "normal" label. If the callee (or any indirect callees) returns with the "<a
2234 href="#i_unwind"><tt>unwind</tt></a>" instruction, control is interrupted and
2235 continued at the dynamically nearest "exception" label.</p>
2239 <p>This instruction requires several arguments:</p>
2243 The optional "cconv" marker indicates which <a href="#callingconv">calling
2244 convention</a> the call should use. If none is specified, the call defaults
2245 to using C calling conventions.
2248 <li>The optional <a href="#paramattrs">Parameter Attributes</a> list for
2249 return values. Only '<tt>zeroext</tt>', '<tt>signext</tt>',
2250 and '<tt>inreg</tt>' attributes are valid here.</li>
2252 <li>'<tt>ptr to function ty</tt>': shall be the signature of the pointer to
2253 function value being invoked. In most cases, this is a direct function
2254 invocation, but indirect <tt>invoke</tt>s are just as possible, branching off
2255 an arbitrary pointer to function value.
2258 <li>'<tt>function ptr val</tt>': An LLVM value containing a pointer to a
2259 function to be invoked. </li>
2261 <li>'<tt>function args</tt>': argument list whose types match the function
2262 signature argument types. If the function signature indicates the function
2263 accepts a variable number of arguments, the extra arguments can be
2266 <li>'<tt>normal label</tt>': the label reached when the called function
2267 executes a '<tt><a href="#i_ret">ret</a></tt>' instruction. </li>
2269 <li>'<tt>exception label</tt>': the label reached when a callee returns with
2270 the <a href="#i_unwind"><tt>unwind</tt></a> instruction. </li>
2272 <li>The optional <a href="#fnattrs">function attributes</a> list. Only
2273 '<tt>noreturn</tt>', '<tt>nounwind</tt>', '<tt>readonly</tt>' and
2274 '<tt>readnone</tt>' attributes are valid here.</li>
2279 <p>This instruction is designed to operate as a standard '<tt><a
2280 href="#i_call">call</a></tt>' instruction in most regards. The primary
2281 difference is that it establishes an association with a label, which is used by
2282 the runtime library to unwind the stack.</p>
2284 <p>This instruction is used in languages with destructors to ensure that proper
2285 cleanup is performed in the case of either a <tt>longjmp</tt> or a thrown
2286 exception. Additionally, this is important for implementation of
2287 '<tt>catch</tt>' clauses in high-level languages that support them.</p>
2291 %retval = invoke i32 @Test(i32 15) to label %Continue
2292 unwind label %TestCleanup <i>; {i32}:retval set</i>
2293 %retval = invoke <a href="#callingconv">coldcc</a> i32 %Testfnptr(i32 15) to label %Continue
2294 unwind label %TestCleanup <i>; {i32}:retval set</i>
2299 <!-- _______________________________________________________________________ -->
2301 <div class="doc_subsubsection"> <a name="i_unwind">'<tt>unwind</tt>'
2302 Instruction</a> </div>
2304 <div class="doc_text">
2313 <p>The '<tt>unwind</tt>' instruction unwinds the stack, continuing control flow
2314 at the first callee in the dynamic call stack which used an <a
2315 href="#i_invoke"><tt>invoke</tt></a> instruction to perform the call. This is
2316 primarily used to implement exception handling.</p>
2320 <p>The '<tt>unwind</tt>' instruction causes execution of the current function to
2321 immediately halt. The dynamic call stack is then searched for the first <a
2322 href="#i_invoke"><tt>invoke</tt></a> instruction on the call stack. Once found,
2323 execution continues at the "exceptional" destination block specified by the
2324 <tt>invoke</tt> instruction. If there is no <tt>invoke</tt> instruction in the
2325 dynamic call chain, undefined behavior results.</p>
2328 <!-- _______________________________________________________________________ -->
2330 <div class="doc_subsubsection"> <a name="i_unreachable">'<tt>unreachable</tt>'
2331 Instruction</a> </div>
2333 <div class="doc_text">
2342 <p>The '<tt>unreachable</tt>' instruction has no defined semantics. This
2343 instruction is used to inform the optimizer that a particular portion of the
2344 code is not reachable. This can be used to indicate that the code after a
2345 no-return function cannot be reached, and other facts.</p>
2349 <p>The '<tt>unreachable</tt>' instruction has no defined semantics.</p>
2354 <!-- ======================================================================= -->
2355 <div class="doc_subsection"> <a name="binaryops">Binary Operations</a> </div>
2356 <div class="doc_text">
2357 <p>Binary operators are used to do most of the computation in a
2358 program. They require two operands of the same type, execute an operation on them, and
2359 produce a single value. The operands might represent
2360 multiple data, as is the case with the <a href="#t_vector">vector</a> data type.
2361 The result value has the same type as its operands.</p>
2362 <p>There are several different binary operators:</p>
2364 <!-- _______________________________________________________________________ -->
2365 <div class="doc_subsubsection">
2366 <a name="i_add">'<tt>add</tt>' Instruction</a>
2369 <div class="doc_text">
2374 <result> = add <ty> <op1>, <op2> <i>; yields {ty}:result</i>
2379 <p>The '<tt>add</tt>' instruction returns the sum of its two operands.</p>
2383 <p>The two arguments to the '<tt>add</tt>' instruction must be <a
2384 href="#t_integer">integer</a>, <a href="#t_floating">floating point</a>, or
2385 <a href="#t_vector">vector</a> values. Both arguments must have identical
2390 <p>The value produced is the integer or floating point sum of the two
2393 <p>If an integer sum has unsigned overflow, the result returned is the
2394 mathematical result modulo 2<sup>n</sup>, where n is the bit width of
2397 <p>Because LLVM integers use a two's complement representation, this
2398 instruction is appropriate for both signed and unsigned integers.</p>
2403 <result> = add i32 4, %var <i>; yields {i32}:result = 4 + %var</i>
2406 <!-- _______________________________________________________________________ -->
2407 <div class="doc_subsubsection">
2408 <a name="i_sub">'<tt>sub</tt>' Instruction</a>
2411 <div class="doc_text">
2416 <result> = sub <ty> <op1>, <op2> <i>; yields {ty}:result</i>
2421 <p>The '<tt>sub</tt>' instruction returns the difference of its two
2424 <p>Note that the '<tt>sub</tt>' instruction is used to represent the
2425 '<tt>neg</tt>' instruction present in most other intermediate
2426 representations.</p>
2430 <p>The two arguments to the '<tt>sub</tt>' instruction must be <a
2431 href="#t_integer">integer</a>, <a href="#t_floating">floating point</a>,
2432 or <a href="#t_vector">vector</a> values. Both arguments must have identical
2437 <p>The value produced is the integer or floating point difference of
2438 the two operands.</p>
2440 <p>If an integer difference has unsigned overflow, the result returned is the
2441 mathematical result modulo 2<sup>n</sup>, where n is the bit width of
2444 <p>Because LLVM integers use a two's complement representation, this
2445 instruction is appropriate for both signed and unsigned integers.</p>
2449 <result> = sub i32 4, %var <i>; yields {i32}:result = 4 - %var</i>
2450 <result> = sub i32 0, %val <i>; yields {i32}:result = -%var</i>
2454 <!-- _______________________________________________________________________ -->
2455 <div class="doc_subsubsection">
2456 <a name="i_mul">'<tt>mul</tt>' Instruction</a>
2459 <div class="doc_text">
2462 <pre> <result> = mul <ty> <op1>, <op2> <i>; yields {ty}:result</i>
2465 <p>The '<tt>mul</tt>' instruction returns the product of its two
2470 <p>The two arguments to the '<tt>mul</tt>' instruction must be <a
2471 href="#t_integer">integer</a>, <a href="#t_floating">floating point</a>,
2472 or <a href="#t_vector">vector</a> values. Both arguments must have identical
2477 <p>The value produced is the integer or floating point product of the
2480 <p>If the result of an integer multiplication has unsigned overflow,
2481 the result returned is the mathematical result modulo
2482 2<sup>n</sup>, where n is the bit width of the result.</p>
2483 <p>Because LLVM integers use a two's complement representation, and the
2484 result is the same width as the operands, this instruction returns the
2485 correct result for both signed and unsigned integers. If a full product
2486 (e.g. <tt>i32</tt>x<tt>i32</tt>-><tt>i64</tt>) is needed, the operands
2487 should be sign-extended or zero-extended as appropriate to the
2488 width of the full product.</p>
2490 <pre> <result> = mul i32 4, %var <i>; yields {i32}:result = 4 * %var</i>
2494 <!-- _______________________________________________________________________ -->
2495 <div class="doc_subsubsection"> <a name="i_udiv">'<tt>udiv</tt>' Instruction
2497 <div class="doc_text">
2499 <pre> <result> = udiv <ty> <op1>, <op2> <i>; yields {ty}:result</i>
2502 <p>The '<tt>udiv</tt>' instruction returns the quotient of its two
2507 <p>The two arguments to the '<tt>udiv</tt>' instruction must be
2508 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
2509 values. Both arguments must have identical types.</p>
2513 <p>The value produced is the unsigned integer quotient of the two operands.</p>
2514 <p>Note that unsigned integer division and signed integer division are distinct
2515 operations; for signed integer division, use '<tt>sdiv</tt>'.</p>
2516 <p>Division by zero leads to undefined behavior.</p>
2518 <pre> <result> = udiv i32 4, %var <i>; yields {i32}:result = 4 / %var</i>
2521 <!-- _______________________________________________________________________ -->
2522 <div class="doc_subsubsection"> <a name="i_sdiv">'<tt>sdiv</tt>' Instruction
2524 <div class="doc_text">
2527 <result> = sdiv <ty> <op1>, <op2> <i>; yields {ty}:result</i>
2532 <p>The '<tt>sdiv</tt>' instruction returns the quotient of its two
2537 <p>The two arguments to the '<tt>sdiv</tt>' instruction must be
2538 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
2539 values. Both arguments must have identical types.</p>
2542 <p>The value produced is the signed integer quotient of the two operands rounded towards zero.</p>
2543 <p>Note that signed integer division and unsigned integer division are distinct
2544 operations; for unsigned integer division, use '<tt>udiv</tt>'.</p>
2545 <p>Division by zero leads to undefined behavior. Overflow also leads to
2546 undefined behavior; this is a rare case, but can occur, for example,
2547 by doing a 32-bit division of -2147483648 by -1.</p>
2549 <pre> <result> = sdiv i32 4, %var <i>; yields {i32}:result = 4 / %var</i>
2552 <!-- _______________________________________________________________________ -->
2553 <div class="doc_subsubsection"> <a name="i_fdiv">'<tt>fdiv</tt>'
2554 Instruction</a> </div>
2555 <div class="doc_text">
2558 <result> = fdiv <ty> <op1>, <op2> <i>; yields {ty}:result</i>
2562 <p>The '<tt>fdiv</tt>' instruction returns the quotient of its two
2567 <p>The two arguments to the '<tt>fdiv</tt>' instruction must be
2568 <a href="#t_floating">floating point</a> or <a href="#t_vector">vector</a>
2569 of floating point values. Both arguments must have identical types.</p>
2573 <p>The value produced is the floating point quotient of the two operands.</p>
2578 <result> = fdiv float 4.0, %var <i>; yields {float}:result = 4.0 / %var</i>
2582 <!-- _______________________________________________________________________ -->
2583 <div class="doc_subsubsection"> <a name="i_urem">'<tt>urem</tt>' Instruction</a>
2585 <div class="doc_text">
2587 <pre> <result> = urem <ty> <op1>, <op2> <i>; yields {ty}:result</i>
2590 <p>The '<tt>urem</tt>' instruction returns the remainder from the
2591 unsigned division of its two arguments.</p>
2593 <p>The two arguments to the '<tt>urem</tt>' instruction must be
2594 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
2595 values. Both arguments must have identical types.</p>
2597 <p>This instruction returns the unsigned integer <i>remainder</i> of a division.
2598 This instruction always performs an unsigned division to get the remainder.</p>
2599 <p>Note that unsigned integer remainder and signed integer remainder are
2600 distinct operations; for signed integer remainder, use '<tt>srem</tt>'.</p>
2601 <p>Taking the remainder of a division by zero leads to undefined behavior.</p>
2603 <pre> <result> = urem i32 4, %var <i>; yields {i32}:result = 4 % %var</i>
2607 <!-- _______________________________________________________________________ -->
2608 <div class="doc_subsubsection">
2609 <a name="i_srem">'<tt>srem</tt>' Instruction</a>
2612 <div class="doc_text">
2617 <result> = srem <ty> <op1>, <op2> <i>; yields {ty}:result</i>
2622 <p>The '<tt>srem</tt>' instruction returns the remainder from the
2623 signed division of its two operands. This instruction can also take
2624 <a href="#t_vector">vector</a> versions of the values in which case
2625 the elements must be integers.</p>
2629 <p>The two arguments to the '<tt>srem</tt>' instruction must be
2630 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
2631 values. Both arguments must have identical types.</p>
2635 <p>This instruction returns the <i>remainder</i> of a division (where the result
2636 has the same sign as the dividend, <tt>op1</tt>), not the <i>modulo</i>
2637 operator (where the result has the same sign as the divisor, <tt>op2</tt>) of
2638 a value. For more information about the difference, see <a
2639 href="http://mathforum.org/dr.math/problems/anne.4.28.99.html">The
2640 Math Forum</a>. For a table of how this is implemented in various languages,
2641 please see <a href="http://en.wikipedia.org/wiki/Modulo_operation">
2642 Wikipedia: modulo operation</a>.</p>
2643 <p>Note that signed integer remainder and unsigned integer remainder are
2644 distinct operations; for unsigned integer remainder, use '<tt>urem</tt>'.</p>
2645 <p>Taking the remainder of a division by zero leads to undefined behavior.
2646 Overflow also leads to undefined behavior; this is a rare case, but can occur,
2647 for example, by taking the remainder of a 32-bit division of -2147483648 by -1.
2648 (The remainder doesn't actually overflow, but this rule lets srem be
2649 implemented using instructions that return both the result of the division
2650 and the remainder.)</p>
2652 <pre> <result> = srem i32 4, %var <i>; yields {i32}:result = 4 % %var</i>
2656 <!-- _______________________________________________________________________ -->
2657 <div class="doc_subsubsection">
2658 <a name="i_frem">'<tt>frem</tt>' Instruction</a> </div>
2660 <div class="doc_text">
2663 <pre> <result> = frem <ty> <op1>, <op2> <i>; yields {ty}:result</i>
2666 <p>The '<tt>frem</tt>' instruction returns the remainder from the
2667 division of its two operands.</p>
2669 <p>The two arguments to the '<tt>frem</tt>' instruction must be
2670 <a href="#t_floating">floating point</a> or <a href="#t_vector">vector</a>
2671 of floating point values. Both arguments must have identical types.</p>
2675 <p>This instruction returns the <i>remainder</i> of a division.
2676 The remainder has the same sign as the dividend.</p>
2681 <result> = frem float 4.0, %var <i>; yields {float}:result = 4.0 % %var</i>
2685 <!-- ======================================================================= -->
2686 <div class="doc_subsection"> <a name="bitwiseops">Bitwise Binary
2687 Operations</a> </div>
2688 <div class="doc_text">
2689 <p>Bitwise binary operators are used to do various forms of
2690 bit-twiddling in a program. They are generally very efficient
2691 instructions and can commonly be strength reduced from other
2692 instructions. They require two operands of the same type, execute an operation on them,
2693 and produce a single value. The resulting value is the same type as its operands.</p>
2696 <!-- _______________________________________________________________________ -->
2697 <div class="doc_subsubsection"> <a name="i_shl">'<tt>shl</tt>'
2698 Instruction</a> </div>
2699 <div class="doc_text">
2701 <pre> <result> = shl <ty> <op1>, <op2> <i>; yields {ty}:result</i>
2706 <p>The '<tt>shl</tt>' instruction returns the first operand shifted to
2707 the left a specified number of bits.</p>
2711 <p>Both arguments to the '<tt>shl</tt>' instruction must be the same <a
2712 href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
2713 type. '<tt>op2</tt>' is treated as an unsigned value.</p>
2717 <p>The value produced is <tt>op1</tt> * 2<sup><tt>op2</tt></sup> mod 2<sup>n</sup>,
2718 where n is the width of the result. If <tt>op2</tt> is (statically or dynamically) negative or
2719 equal to or larger than the number of bits in <tt>op1</tt>, the result is undefined.
2720 If the arguments are vectors, each vector element of <tt>op1</tt> is shifted by the
2721 corresponding shift amount in <tt>op2</tt>.</p>
2723 <h5>Example:</h5><pre>
2724 <result> = shl i32 4, %var <i>; yields {i32}: 4 << %var</i>
2725 <result> = shl i32 4, 2 <i>; yields {i32}: 16</i>
2726 <result> = shl i32 1, 10 <i>; yields {i32}: 1024</i>
2727 <result> = shl i32 1, 32 <i>; undefined</i>
2728 <result> = shl <2 x i32> < i32 1, i32 1>, < i32 1, i32 2> <i>; yields: result=<2 x i32> < i32 2, i32 4></i>
2731 <!-- _______________________________________________________________________ -->
2732 <div class="doc_subsubsection"> <a name="i_lshr">'<tt>lshr</tt>'
2733 Instruction</a> </div>
2734 <div class="doc_text">
2736 <pre> <result> = lshr <ty> <op1>, <op2> <i>; yields {ty}:result</i>
2740 <p>The '<tt>lshr</tt>' instruction (logical shift right) returns the first
2741 operand shifted to the right a specified number of bits with zero fill.</p>
2744 <p>Both arguments to the '<tt>lshr</tt>' instruction must be the same
2745 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
2746 type. '<tt>op2</tt>' is treated as an unsigned value.</p>
2750 <p>This instruction always performs a logical shift right operation. The most
2751 significant bits of the result will be filled with zero bits after the
2752 shift. If <tt>op2</tt> is (statically or dynamically) equal to or larger than
2753 the number of bits in <tt>op1</tt>, the result is undefined. If the arguments are
2754 vectors, each vector element of <tt>op1</tt> is shifted by the corresponding shift
2755 amount in <tt>op2</tt>.</p>
2759 <result> = lshr i32 4, 1 <i>; yields {i32}:result = 2</i>
2760 <result> = lshr i32 4, 2 <i>; yields {i32}:result = 1</i>
2761 <result> = lshr i8 4, 3 <i>; yields {i8}:result = 0</i>
2762 <result> = lshr i8 -2, 1 <i>; yields {i8}:result = 0x7FFFFFFF </i>
2763 <result> = lshr i32 1, 32 <i>; undefined</i>
2764 <result> = lshr <2 x i32> < i32 -2, i32 4>, < i32 1, i32 2> <i>; yields: result=<2 x i32> < i32 0x7FFFFFFF, i32 1></i>
2768 <!-- _______________________________________________________________________ -->
2769 <div class="doc_subsubsection"> <a name="i_ashr">'<tt>ashr</tt>'
2770 Instruction</a> </div>
2771 <div class="doc_text">
2774 <pre> <result> = ashr <ty> <op1>, <op2> <i>; yields {ty}:result</i>
2778 <p>The '<tt>ashr</tt>' instruction (arithmetic shift right) returns the first
2779 operand shifted to the right a specified number of bits with sign extension.</p>
2782 <p>Both arguments to the '<tt>ashr</tt>' instruction must be the same
2783 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
2784 type. '<tt>op2</tt>' is treated as an unsigned value.</p>
2787 <p>This instruction always performs an arithmetic shift right operation,
2788 The most significant bits of the result will be filled with the sign bit
2789 of <tt>op1</tt>. If <tt>op2</tt> is (statically or dynamically) equal to or
2790 larger than the number of bits in <tt>op1</tt>, the result is undefined. If the
2791 arguments are vectors, each vector element of <tt>op1</tt> is shifted by the
2792 corresponding shift amount in <tt>op2</tt>.</p>
2796 <result> = ashr i32 4, 1 <i>; yields {i32}:result = 2</i>
2797 <result> = ashr i32 4, 2 <i>; yields {i32}:result = 1</i>
2798 <result> = ashr i8 4, 3 <i>; yields {i8}:result = 0</i>
2799 <result> = ashr i8 -2, 1 <i>; yields {i8}:result = -1</i>
2800 <result> = ashr i32 1, 32 <i>; undefined</i>
2801 <result> = ashr <2 x i32> < i32 -2, i32 4>, < i32 1, i32 3> <i>; yields: result=<2 x i32> < i32 -1, i32 0></i>
2805 <!-- _______________________________________________________________________ -->
2806 <div class="doc_subsubsection"> <a name="i_and">'<tt>and</tt>'
2807 Instruction</a> </div>
2809 <div class="doc_text">
2814 <result> = and <ty> <op1>, <op2> <i>; yields {ty}:result</i>
2819 <p>The '<tt>and</tt>' instruction returns the bitwise logical and of
2820 its two operands.</p>
2824 <p>The two arguments to the '<tt>and</tt>' instruction must be
2825 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
2826 values. Both arguments must have identical types.</p>
2829 <p>The truth table used for the '<tt>and</tt>' instruction is:</p>
2832 <table border="1" cellspacing="0" cellpadding="4">
2864 <result> = and i32 4, %var <i>; yields {i32}:result = 4 & %var</i>
2865 <result> = and i32 15, 40 <i>; yields {i32}:result = 8</i>
2866 <result> = and i32 4, 8 <i>; yields {i32}:result = 0</i>
2869 <!-- _______________________________________________________________________ -->
2870 <div class="doc_subsubsection"> <a name="i_or">'<tt>or</tt>' Instruction</a> </div>
2871 <div class="doc_text">
2873 <pre> <result> = or <ty> <op1>, <op2> <i>; yields {ty}:result</i>
2876 <p>The '<tt>or</tt>' instruction returns the bitwise logical inclusive
2877 or of its two operands.</p>
2880 <p>The two arguments to the '<tt>or</tt>' instruction must be
2881 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
2882 values. Both arguments must have identical types.</p>
2884 <p>The truth table used for the '<tt>or</tt>' instruction is:</p>
2887 <table border="1" cellspacing="0" cellpadding="4">
2918 <pre> <result> = or i32 4, %var <i>; yields {i32}:result = 4 | %var</i>
2919 <result> = or i32 15, 40 <i>; yields {i32}:result = 47</i>
2920 <result> = or i32 4, 8 <i>; yields {i32}:result = 12</i>
2923 <!-- _______________________________________________________________________ -->
2924 <div class="doc_subsubsection"> <a name="i_xor">'<tt>xor</tt>'
2925 Instruction</a> </div>
2926 <div class="doc_text">
2928 <pre> <result> = xor <ty> <op1>, <op2> <i>; yields {ty}:result</i>
2931 <p>The '<tt>xor</tt>' instruction returns the bitwise logical exclusive
2932 or of its two operands. The <tt>xor</tt> is used to implement the
2933 "one's complement" operation, which is the "~" operator in C.</p>
2935 <p>The two arguments to the '<tt>xor</tt>' instruction must be
2936 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
2937 values. Both arguments must have identical types.</p>
2941 <p>The truth table used for the '<tt>xor</tt>' instruction is:</p>
2944 <table border="1" cellspacing="0" cellpadding="4">
2976 <pre> <result> = xor i32 4, %var <i>; yields {i32}:result = 4 ^ %var</i>
2977 <result> = xor i32 15, 40 <i>; yields {i32}:result = 39</i>
2978 <result> = xor i32 4, 8 <i>; yields {i32}:result = 12</i>
2979 <result> = xor i32 %V, -1 <i>; yields {i32}:result = ~%V</i>
2983 <!-- ======================================================================= -->
2984 <div class="doc_subsection">
2985 <a name="vectorops">Vector Operations</a>
2988 <div class="doc_text">
2990 <p>LLVM supports several instructions to represent vector operations in a
2991 target-independent manner. These instructions cover the element-access and
2992 vector-specific operations needed to process vectors effectively. While LLVM
2993 does directly support these vector operations, many sophisticated algorithms
2994 will want to use target-specific intrinsics to take full advantage of a specific
2999 <!-- _______________________________________________________________________ -->
3000 <div class="doc_subsubsection">
3001 <a name="i_extractelement">'<tt>extractelement</tt>' Instruction</a>
3004 <div class="doc_text">
3009 <result> = extractelement <n x <ty>> <val>, i32 <idx> <i>; yields <ty></i>
3015 The '<tt>extractelement</tt>' instruction extracts a single scalar
3016 element from a vector at a specified index.
3023 The first operand of an '<tt>extractelement</tt>' instruction is a
3024 value of <a href="#t_vector">vector</a> type. The second operand is
3025 an index indicating the position from which to extract the element.
3026 The index may be a variable.</p>
3031 The result is a scalar of the same type as the element type of
3032 <tt>val</tt>. Its value is the value at position <tt>idx</tt> of
3033 <tt>val</tt>. If <tt>idx</tt> exceeds the length of <tt>val</tt>, the
3034 results are undefined.
3040 %result = extractelement <4 x i32> %vec, i32 0 <i>; yields i32</i>
3045 <!-- _______________________________________________________________________ -->
3046 <div class="doc_subsubsection">
3047 <a name="i_insertelement">'<tt>insertelement</tt>' Instruction</a>
3050 <div class="doc_text">
3055 <result> = insertelement <n x <ty>> <val>, <ty> <elt>, i32 <idx> <i>; yields <n x <ty>></i>
3061 The '<tt>insertelement</tt>' instruction inserts a scalar
3062 element into a vector at a specified index.
3069 The first operand of an '<tt>insertelement</tt>' instruction is a
3070 value of <a href="#t_vector">vector</a> type. The second operand is a
3071 scalar value whose type must equal the element type of the first
3072 operand. The third operand is an index indicating the position at
3073 which to insert the value. The index may be a variable.</p>
3078 The result is a vector of the same type as <tt>val</tt>. Its
3079 element values are those of <tt>val</tt> except at position
3080 <tt>idx</tt>, where it gets the value <tt>elt</tt>. If <tt>idx</tt>
3081 exceeds the length of <tt>val</tt>, the results are undefined.
3087 %result = insertelement <4 x i32> %vec, i32 1, i32 0 <i>; yields <4 x i32></i>
3091 <!-- _______________________________________________________________________ -->
3092 <div class="doc_subsubsection">
3093 <a name="i_shufflevector">'<tt>shufflevector</tt>' Instruction</a>
3096 <div class="doc_text">
3101 <result> = shufflevector <n x <ty>> <v1>, <n x <ty>> <v2>, <m x i32> <mask> <i>; yields <m x <ty>></i>
3107 The '<tt>shufflevector</tt>' instruction constructs a permutation of elements
3108 from two input vectors, returning a vector with the same element type as
3109 the input and length that is the same as the shuffle mask.
3115 The first two operands of a '<tt>shufflevector</tt>' instruction are vectors
3116 with types that match each other. The third argument is a shuffle mask whose
3117 element type is always 'i32'. The result of the instruction is a vector whose
3118 length is the same as the shuffle mask and whose element type is the same as
3119 the element type of the first two operands.
3123 The shuffle mask operand is required to be a constant vector with either
3124 constant integer or undef values.
3130 The elements of the two input vectors are numbered from left to right across
3131 both of the vectors. The shuffle mask operand specifies, for each element of
3132 the result vector, which element of the two input vectors the result element
3133 gets. The element selector may be undef (meaning "don't care") and the second
3134 operand may be undef if performing a shuffle from only one vector.
3140 %result = shufflevector <4 x i32> %v1, <4 x i32> %v2,
3141 <4 x i32> <i32 0, i32 4, i32 1, i32 5> <i>; yields <4 x i32></i>
3142 %result = shufflevector <4 x i32> %v1, <4 x i32> undef,
3143 <4 x i32> <i32 0, i32 1, i32 2, i32 3> <i>; yields <4 x i32></i> - Identity shuffle.
3144 %result = shufflevector <8 x i32> %v1, <8 x i32> undef,
3145 <4 x i32> <i32 0, i32 1, i32 2, i32 3> <i>; yields <4 x i32></i>
3146 %result = shufflevector <4 x i32> %v1, <4 x i32> %v2,
3147 <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>
3152 <!-- ======================================================================= -->
3153 <div class="doc_subsection">
3154 <a name="aggregateops">Aggregate Operations</a>
3157 <div class="doc_text">
3159 <p>LLVM supports several instructions for working with aggregate values.
3164 <!-- _______________________________________________________________________ -->
3165 <div class="doc_subsubsection">
3166 <a name="i_extractvalue">'<tt>extractvalue</tt>' Instruction</a>
3169 <div class="doc_text">
3174 <result> = extractvalue <aggregate type> <val>, <idx>{, <idx>}*
3180 The '<tt>extractvalue</tt>' instruction extracts the value of a struct field
3181 or array element from an aggregate value.
3188 The first operand of an '<tt>extractvalue</tt>' instruction is a
3189 value of <a href="#t_struct">struct</a> or <a href="#t_array">array</a>
3190 type. The operands are constant indices to specify which value to extract
3191 in a similar manner as indices in a
3192 '<tt><a href="#i_getelementptr">getelementptr</a></tt>' instruction.
3198 The result is the value at the position in the aggregate specified by
3205 %result = extractvalue {i32, float} %agg, 0 <i>; yields i32</i>
3210 <!-- _______________________________________________________________________ -->
3211 <div class="doc_subsubsection">
3212 <a name="i_insertvalue">'<tt>insertvalue</tt>' Instruction</a>
3215 <div class="doc_text">
3220 <result> = insertvalue <aggregate type> <val>, <ty> <val>, <idx> <i>; yields <n x <ty>></i>
3226 The '<tt>insertvalue</tt>' instruction inserts a value
3227 into a struct field or array element in an aggregate.
3234 The first operand of an '<tt>insertvalue</tt>' instruction is a
3235 value of <a href="#t_struct">struct</a> or <a href="#t_array">array</a> type.
3236 The second operand is a first-class value to insert.
3237 The following operands are constant indices
3238 indicating the position at which to insert the value in a similar manner as
3240 '<tt><a href="#i_getelementptr">getelementptr</a></tt>' instruction.
3241 The value to insert must have the same type as the value identified
3248 The result is an aggregate of the same type as <tt>val</tt>. Its
3249 value is that of <tt>val</tt> except that the value at the position
3250 specified by the indices is that of <tt>elt</tt>.
3256 %result = insertvalue {i32, float} %agg, i32 1, 0 <i>; yields {i32, float}</i>
3261 <!-- ======================================================================= -->
3262 <div class="doc_subsection">
3263 <a name="memoryops">Memory Access and Addressing Operations</a>
3266 <div class="doc_text">
3268 <p>A key design point of an SSA-based representation is how it
3269 represents memory. In LLVM, no memory locations are in SSA form, which
3270 makes things very simple. This section describes how to read, write,
3271 allocate, and free memory in LLVM.</p>
3275 <!-- _______________________________________________________________________ -->
3276 <div class="doc_subsubsection">
3277 <a name="i_malloc">'<tt>malloc</tt>' Instruction</a>
3280 <div class="doc_text">
3285 <result> = malloc <type>[, i32 <NumElements>][, align <alignment>] <i>; yields {type*}:result</i>
3290 <p>The '<tt>malloc</tt>' instruction allocates memory from the system
3291 heap and returns a pointer to it. The object is always allocated in the generic
3292 address space (address space zero).</p>
3296 <p>The '<tt>malloc</tt>' instruction allocates
3297 <tt>sizeof(<type>)*NumElements</tt>
3298 bytes of memory from the operating system and returns a pointer of the
3299 appropriate type to the program. If "NumElements" is specified, it is the
3300 number of elements allocated, otherwise "NumElements" is defaulted to be one.
3301 If a constant alignment is specified, the value result of the allocation is guaranteed to
3302 be aligned to at least that boundary. If not specified, or if zero, the target can
3303 choose to align the allocation on any convenient boundary.</p>
3305 <p>'<tt>type</tt>' must be a sized type.</p>
3309 <p>Memory is allocated using the system "<tt>malloc</tt>" function, and
3310 a pointer is returned. The result of a zero byte allocation is undefined. The
3311 result is null if there is insufficient memory available.</p>
3316 %array = malloc [4 x i8] <i>; yields {[%4 x i8]*}:array</i>
3318 %size = <a href="#i_add">add</a> i32 2, 2 <i>; yields {i32}:size = i32 4</i>
3319 %array1 = malloc i8, i32 4 <i>; yields {i8*}:array1</i>
3320 %array2 = malloc [12 x i8], i32 %size <i>; yields {[12 x i8]*}:array2</i>
3321 %array3 = malloc i32, i32 4, align 1024 <i>; yields {i32*}:array3</i>
3322 %array4 = malloc i32, align 1024 <i>; yields {i32*}:array4</i>
3325 <p>Note that the code generator does not yet respect the
3326 alignment value.</p>
3330 <!-- _______________________________________________________________________ -->
3331 <div class="doc_subsubsection">
3332 <a name="i_free">'<tt>free</tt>' Instruction</a>
3335 <div class="doc_text">
3340 free <type> <value> <i>; yields {void}</i>
3345 <p>The '<tt>free</tt>' instruction returns memory back to the unused
3346 memory heap to be reallocated in the future.</p>
3350 <p>'<tt>value</tt>' shall be a pointer value that points to a value
3351 that was allocated with the '<tt><a href="#i_malloc">malloc</a></tt>'
3356 <p>Access to the memory pointed to by the pointer is no longer defined
3357 after this instruction executes. If the pointer is null, the operation
3363 %array = <a href="#i_malloc">malloc</a> [4 x i8] <i>; yields {[4 x i8]*}:array</i>
3364 free [4 x i8]* %array
3368 <!-- _______________________________________________________________________ -->
3369 <div class="doc_subsubsection">
3370 <a name="i_alloca">'<tt>alloca</tt>' Instruction</a>
3373 <div class="doc_text">
3378 <result> = alloca <type>[, i32 <NumElements>][, align <alignment>] <i>; yields {type*}:result</i>
3383 <p>The '<tt>alloca</tt>' instruction allocates memory on the stack frame of the
3384 currently executing function, to be automatically released when this function
3385 returns to its caller. The object is always allocated in the generic address
3386 space (address space zero).</p>
3390 <p>The '<tt>alloca</tt>' instruction allocates <tt>sizeof(<type>)*NumElements</tt>
3391 bytes of memory on the runtime stack, returning a pointer of the
3392 appropriate type to the program. If "NumElements" is specified, it is the
3393 number of elements allocated, otherwise "NumElements" is defaulted to be one.
3394 If a constant alignment is specified, the value result of the allocation is guaranteed
3395 to be aligned to at least that boundary. If not specified, or if zero, the target
3396 can choose to align the allocation on any convenient boundary.</p>
3398 <p>'<tt>type</tt>' may be any sized type.</p>
3402 <p>Memory is allocated; a pointer is returned. The operation is undefiend if
3403 there is insufficient stack space for the allocation. '<tt>alloca</tt>'d
3404 memory is automatically released when the function returns. The '<tt>alloca</tt>'
3405 instruction is commonly used to represent automatic variables that must
3406 have an address available. When the function returns (either with the <tt><a
3407 href="#i_ret">ret</a></tt> or <tt><a href="#i_unwind">unwind</a></tt>
3408 instructions), the memory is reclaimed. Allocating zero bytes
3409 is legal, but the result is undefined.</p>
3414 %ptr = alloca i32 <i>; yields {i32*}:ptr</i>
3415 %ptr = alloca i32, i32 4 <i>; yields {i32*}:ptr</i>
3416 %ptr = alloca i32, i32 4, align 1024 <i>; yields {i32*}:ptr</i>
3417 %ptr = alloca i32, align 1024 <i>; yields {i32*}:ptr</i>
3421 <!-- _______________________________________________________________________ -->
3422 <div class="doc_subsubsection"> <a name="i_load">'<tt>load</tt>'
3423 Instruction</a> </div>
3424 <div class="doc_text">
3426 <pre> <result> = load <ty>* <pointer>[, align <alignment>]<br> <result> = volatile load <ty>* <pointer>[, align <alignment>]<br></pre>
3428 <p>The '<tt>load</tt>' instruction is used to read from memory.</p>
3430 <p>The argument to the '<tt>load</tt>' instruction specifies the memory
3431 address from which to load. The pointer must point to a <a
3432 href="#t_firstclass">first class</a> type. If the <tt>load</tt> is
3433 marked as <tt>volatile</tt>, then the optimizer is not allowed to modify
3434 the number or order of execution of this <tt>load</tt> with other
3435 volatile <tt>load</tt> and <tt><a href="#i_store">store</a></tt>
3438 The optional constant "align" argument specifies the alignment of the operation
3439 (that is, the alignment of the memory address). A value of 0 or an
3440 omitted "align" argument means that the operation has the preferential
3441 alignment for the target. It is the responsibility of the code emitter
3442 to ensure that the alignment information is correct. Overestimating
3443 the alignment results in an undefined behavior. Underestimating the
3444 alignment may produce less efficient code. An alignment of 1 is always
3448 <p>The location of memory pointed to is loaded.</p>
3450 <pre> %ptr = <a href="#i_alloca">alloca</a> i32 <i>; yields {i32*}:ptr</i>
3452 href="#i_store">store</a> i32 3, i32* %ptr <i>; yields {void}</i>
3453 %val = load i32* %ptr <i>; yields {i32}:val = i32 3</i>
3456 <!-- _______________________________________________________________________ -->
3457 <div class="doc_subsubsection"> <a name="i_store">'<tt>store</tt>'
3458 Instruction</a> </div>
3459 <div class="doc_text">
3461 <pre> store <ty> <value>, <ty>* <pointer>[, align <alignment>] <i>; yields {void}</i>
3462 volatile store <ty> <value>, <ty>* <pointer>[, align <alignment>] <i>; yields {void}</i>
3465 <p>The '<tt>store</tt>' instruction is used to write to memory.</p>
3467 <p>There are two arguments to the '<tt>store</tt>' instruction: a value
3468 to store and an address at which to store it. The type of the '<tt><pointer></tt>'
3469 operand must be a pointer to the <a href="#t_firstclass">first class</a> type
3470 of the '<tt><value></tt>'
3471 operand. If the <tt>store</tt> is marked as <tt>volatile</tt>, then the
3472 optimizer is not allowed to modify the number or order of execution of
3473 this <tt>store</tt> with other volatile <tt>load</tt> and <tt><a
3474 href="#i_store">store</a></tt> instructions.</p>
3476 The optional constant "align" argument specifies the alignment of the operation
3477 (that is, the alignment of the memory address). A value of 0 or an
3478 omitted "align" argument means that the operation has the preferential
3479 alignment for the target. It is the responsibility of the code emitter
3480 to ensure that the alignment information is correct. Overestimating
3481 the alignment results in an undefined behavior. Underestimating the
3482 alignment may produce less efficient code. An alignment of 1 is always
3486 <p>The contents of memory are updated to contain '<tt><value></tt>'
3487 at the location specified by the '<tt><pointer></tt>' operand.</p>
3489 <pre> %ptr = <a href="#i_alloca">alloca</a> i32 <i>; yields {i32*}:ptr</i>
3490 store i32 3, i32* %ptr <i>; yields {void}</i>
3491 %val = <a href="#i_load">load</a> i32* %ptr <i>; yields {i32}:val = i32 3</i>
3495 <!-- _______________________________________________________________________ -->
3496 <div class="doc_subsubsection">
3497 <a name="i_getelementptr">'<tt>getelementptr</tt>' Instruction</a>
3500 <div class="doc_text">
3503 <result> = getelementptr <pty>* <ptrval>{, <ty> <idx>}*
3509 The '<tt>getelementptr</tt>' instruction is used to get the address of a
3510 subelement of an aggregate data structure. It performs address calculation only
3511 and does not access memory.</p>
3515 <p>The first argument is always a pointer, and forms the basis of the
3516 calculation. The remaining arguments are indices, that indicate which of the
3517 elements of the aggregate object are indexed. The interpretation of each index
3518 is dependent on the type being indexed into. The first index always indexes the
3519 pointer value given as the first argument, the second index indexes a value of
3520 the type pointed to (not necessarily the value directly pointed to, since the
3521 first index can be non-zero), etc. The first type indexed into must be a pointer
3522 value, subsequent types can be arrays, vectors and structs. Note that subsequent
3523 types being indexed into can never be pointers, since that would require loading
3524 the pointer before continuing calculation.</p>
3526 <p>The type of each index argument depends on the type it is indexing into.
3527 When indexing into a (packed) structure, only <tt>i32</tt> integer
3528 <b>constants</b> are allowed. When indexing into an array, pointer or vector,
3529 only integers of 32 or 64 bits are allowed (also non-constants). 32-bit values
3530 will be sign extended to 64-bits if required.</p>
3532 <p>For example, let's consider a C code fragment and how it gets
3533 compiled to LLVM:</p>
3535 <div class="doc_code">
3548 int *foo(struct ST *s) {
3549 return &s[1].Z.B[5][13];
3554 <p>The LLVM code generated by the GCC frontend is:</p>
3556 <div class="doc_code">
3558 %RT = <a href="#namedtypes">type</a> { i8 , [10 x [20 x i32]], i8 }
3559 %ST = <a href="#namedtypes">type</a> { i32, double, %RT }
3561 define i32* %foo(%ST* %s) {
3563 %reg = getelementptr %ST* %s, i32 1, i32 2, i32 1, i32 5, i32 13
3571 <p>In the example above, the first index is indexing into the '<tt>%ST*</tt>'
3572 type, which is a pointer, yielding a '<tt>%ST</tt>' = '<tt>{ i32, double, %RT
3573 }</tt>' type, a structure. The second index indexes into the third element of
3574 the structure, yielding a '<tt>%RT</tt>' = '<tt>{ i8 , [10 x [20 x i32]],
3575 i8 }</tt>' type, another structure. The third index indexes into the second
3576 element of the structure, yielding a '<tt>[10 x [20 x i32]]</tt>' type, an
3577 array. The two dimensions of the array are subscripted into, yielding an
3578 '<tt>i32</tt>' type. The '<tt>getelementptr</tt>' instruction returns a pointer
3579 to this element, thus computing a value of '<tt>i32*</tt>' type.</p>
3581 <p>Note that it is perfectly legal to index partially through a
3582 structure, returning a pointer to an inner element. Because of this,
3583 the LLVM code for the given testcase is equivalent to:</p>
3586 define i32* %foo(%ST* %s) {
3587 %t1 = getelementptr %ST* %s, i32 1 <i>; yields %ST*:%t1</i>
3588 %t2 = getelementptr %ST* %t1, i32 0, i32 2 <i>; yields %RT*:%t2</i>
3589 %t3 = getelementptr %RT* %t2, i32 0, i32 1 <i>; yields [10 x [20 x i32]]*:%t3</i>
3590 %t4 = getelementptr [10 x [20 x i32]]* %t3, i32 0, i32 5 <i>; yields [20 x i32]*:%t4</i>
3591 %t5 = getelementptr [20 x i32]* %t4, i32 0, i32 13 <i>; yields i32*:%t5</i>
3596 <p>Note that it is undefined to access an array out of bounds: array and
3597 pointer indexes must always be within the defined bounds of the array type.
3598 The one exception for this rule is zero length arrays. These arrays are
3599 defined to be accessible as variable length arrays, which requires access
3600 beyond the zero'th element.</p>
3602 <p>The getelementptr instruction is often confusing. For some more insight
3603 into how it works, see <a href="GetElementPtr.html">the getelementptr
3609 <i>; yields [12 x i8]*:aptr</i>
3610 %aptr = getelementptr {i32, [12 x i8]}* %saptr, i64 0, i32 1
3611 <i>; yields i8*:vptr</i>
3612 %vptr = getelementptr {i32, <2 x i8>}* %svptr, i64 0, i32 1, i32 1
3613 <i>; yields i8*:eptr</i>
3614 %eptr = getelementptr [12 x i8]* %aptr, i64 0, i32 1
3618 <!-- ======================================================================= -->
3619 <div class="doc_subsection"> <a name="convertops">Conversion Operations</a>
3621 <div class="doc_text">
3622 <p>The instructions in this category are the conversion instructions (casting)
3623 which all take a single operand and a type. They perform various bit conversions
3627 <!-- _______________________________________________________________________ -->
3628 <div class="doc_subsubsection">
3629 <a name="i_trunc">'<tt>trunc .. to</tt>' Instruction</a>
3631 <div class="doc_text">
3635 <result> = trunc <ty> <value> to <ty2> <i>; yields ty2</i>
3640 The '<tt>trunc</tt>' instruction truncates its operand to the type <tt>ty2</tt>.
3645 The '<tt>trunc</tt>' instruction takes a <tt>value</tt> to trunc, which must
3646 be an <a href="#t_integer">integer</a> type, and a type that specifies the size
3647 and type of the result, which must be an <a href="#t_integer">integer</a>
3648 type. The bit size of <tt>value</tt> must be larger than the bit size of
3649 <tt>ty2</tt>. Equal sized types are not allowed.</p>
3653 The '<tt>trunc</tt>' instruction truncates the high order bits in <tt>value</tt>
3654 and converts the remaining bits to <tt>ty2</tt>. Since the source size must be
3655 larger than the destination size, <tt>trunc</tt> cannot be a <i>no-op cast</i>.
3656 It will always truncate bits.</p>
3660 %X = trunc i32 257 to i8 <i>; yields i8:1</i>
3661 %Y = trunc i32 123 to i1 <i>; yields i1:true</i>
3662 %Y = trunc i32 122 to i1 <i>; yields i1:false</i>
3666 <!-- _______________________________________________________________________ -->
3667 <div class="doc_subsubsection">
3668 <a name="i_zext">'<tt>zext .. to</tt>' Instruction</a>
3670 <div class="doc_text">
3674 <result> = zext <ty> <value> to <ty2> <i>; yields ty2</i>
3678 <p>The '<tt>zext</tt>' instruction zero extends its operand to type
3683 <p>The '<tt>zext</tt>' instruction takes a value to cast, which must be of
3684 <a href="#t_integer">integer</a> type, and a type to cast it to, which must
3685 also be of <a href="#t_integer">integer</a> type. The bit size of the
3686 <tt>value</tt> must be smaller than the bit size of the destination type,
3690 <p>The <tt>zext</tt> fills the high order bits of the <tt>value</tt> with zero
3691 bits until it reaches the size of the destination type, <tt>ty2</tt>.</p>
3693 <p>When zero extending from i1, the result will always be either 0 or 1.</p>
3697 %X = zext i32 257 to i64 <i>; yields i64:257</i>
3698 %Y = zext i1 true to i32 <i>; yields i32:1</i>
3702 <!-- _______________________________________________________________________ -->
3703 <div class="doc_subsubsection">
3704 <a name="i_sext">'<tt>sext .. to</tt>' Instruction</a>
3706 <div class="doc_text">
3710 <result> = sext <ty> <value> to <ty2> <i>; yields ty2</i>
3714 <p>The '<tt>sext</tt>' sign extends <tt>value</tt> to the type <tt>ty2</tt>.</p>
3718 The '<tt>sext</tt>' instruction takes a value to cast, which must be of
3719 <a href="#t_integer">integer</a> type, and a type to cast it to, which must
3720 also be of <a href="#t_integer">integer</a> type. The bit size of the
3721 <tt>value</tt> must be smaller than the bit size of the destination type,
3726 The '<tt>sext</tt>' instruction performs a sign extension by copying the sign
3727 bit (highest order bit) of the <tt>value</tt> until it reaches the bit size of
3728 the type <tt>ty2</tt>.</p>
3730 <p>When sign extending from i1, the extension always results in -1 or 0.</p>
3734 %X = sext i8 -1 to i16 <i>; yields i16 :65535</i>
3735 %Y = sext i1 true to i32 <i>; yields i32:-1</i>
3739 <!-- _______________________________________________________________________ -->
3740 <div class="doc_subsubsection">
3741 <a name="i_fptrunc">'<tt>fptrunc .. to</tt>' Instruction</a>
3744 <div class="doc_text">
3749 <result> = fptrunc <ty> <value> to <ty2> <i>; yields ty2</i>
3753 <p>The '<tt>fptrunc</tt>' instruction truncates <tt>value</tt> to type
3758 <p>The '<tt>fptrunc</tt>' instruction takes a <a href="#t_floating">floating
3759 point</a> value to cast and a <a href="#t_floating">floating point</a> type to
3760 cast it to. The size of <tt>value</tt> must be larger than the size of
3761 <tt>ty2</tt>. This implies that <tt>fptrunc</tt> cannot be used to make a
3762 <i>no-op cast</i>.</p>
3765 <p> The '<tt>fptrunc</tt>' instruction truncates a <tt>value</tt> from a larger
3766 <a href="#t_floating">floating point</a> type to a smaller
3767 <a href="#t_floating">floating point</a> type. If the value cannot fit within
3768 the destination type, <tt>ty2</tt>, then the results are undefined.</p>
3772 %X = fptrunc double 123.0 to float <i>; yields float:123.0</i>
3773 %Y = fptrunc double 1.0E+300 to float <i>; yields undefined</i>
3777 <!-- _______________________________________________________________________ -->
3778 <div class="doc_subsubsection">
3779 <a name="i_fpext">'<tt>fpext .. to</tt>' Instruction</a>
3781 <div class="doc_text">
3785 <result> = fpext <ty> <value> to <ty2> <i>; yields ty2</i>
3789 <p>The '<tt>fpext</tt>' extends a floating point <tt>value</tt> to a larger
3790 floating point value.</p>
3793 <p>The '<tt>fpext</tt>' instruction takes a
3794 <a href="#t_floating">floating point</a> <tt>value</tt> to cast,
3795 and a <a href="#t_floating">floating point</a> type to cast it to. The source
3796 type must be smaller than the destination type.</p>
3799 <p>The '<tt>fpext</tt>' instruction extends the <tt>value</tt> from a smaller
3800 <a href="#t_floating">floating point</a> type to a larger
3801 <a href="#t_floating">floating point</a> type. The <tt>fpext</tt> cannot be
3802 used to make a <i>no-op cast</i> because it always changes bits. Use
3803 <tt>bitcast</tt> to make a <i>no-op cast</i> for a floating point cast.</p>
3807 %X = fpext float 3.1415 to double <i>; yields double:3.1415</i>
3808 %Y = fpext float 1.0 to float <i>; yields float:1.0 (no-op)</i>
3812 <!-- _______________________________________________________________________ -->
3813 <div class="doc_subsubsection">
3814 <a name="i_fptoui">'<tt>fptoui .. to</tt>' Instruction</a>
3816 <div class="doc_text">
3820 <result> = fptoui <ty> <value> to <ty2> <i>; yields ty2</i>
3824 <p>The '<tt>fptoui</tt>' converts a floating point <tt>value</tt> to its
3825 unsigned integer equivalent of type <tt>ty2</tt>.
3829 <p>The '<tt>fptoui</tt>' instruction takes a value to cast, which must be a
3830 scalar or vector <a href="#t_floating">floating point</a> value, and a type
3831 to cast it to <tt>ty2</tt>, which must be an <a href="#t_integer">integer</a>
3832 type. If <tt>ty</tt> is a vector floating point type, <tt>ty2</tt> must be a
3833 vector integer type with the same number of elements as <tt>ty</tt></p>
3836 <p> The '<tt>fptoui</tt>' instruction converts its
3837 <a href="#t_floating">floating point</a> operand into the nearest (rounding
3838 towards zero) unsigned integer value. If the value cannot fit in <tt>ty2</tt>,
3839 the results are undefined.</p>
3843 %X = fptoui double 123.0 to i32 <i>; yields i32:123</i>
3844 %Y = fptoui float 1.0E+300 to i1 <i>; yields undefined:1</i>
3845 %X = fptoui float 1.04E+17 to i8 <i>; yields undefined:1</i>
3849 <!-- _______________________________________________________________________ -->
3850 <div class="doc_subsubsection">
3851 <a name="i_fptosi">'<tt>fptosi .. to</tt>' Instruction</a>
3853 <div class="doc_text">
3857 <result> = fptosi <ty> <value> to <ty2> <i>; yields ty2</i>
3861 <p>The '<tt>fptosi</tt>' instruction converts
3862 <a href="#t_floating">floating point</a> <tt>value</tt> to type <tt>ty2</tt>.
3866 <p> The '<tt>fptosi</tt>' instruction takes a value to cast, which must be a
3867 scalar or vector <a href="#t_floating">floating point</a> value, and a type
3868 to cast it to <tt>ty2</tt>, which must be an <a href="#t_integer">integer</a>
3869 type. If <tt>ty</tt> is a vector floating point type, <tt>ty2</tt> must be a
3870 vector integer type with the same number of elements as <tt>ty</tt></p>
3873 <p>The '<tt>fptosi</tt>' instruction converts its
3874 <a href="#t_floating">floating point</a> operand into the nearest (rounding
3875 towards zero) signed integer value. If the value cannot fit in <tt>ty2</tt>,
3876 the results are undefined.</p>
3880 %X = fptosi double -123.0 to i32 <i>; yields i32:-123</i>
3881 %Y = fptosi float 1.0E-247 to i1 <i>; yields undefined:1</i>
3882 %X = fptosi float 1.04E+17 to i8 <i>; yields undefined:1</i>
3886 <!-- _______________________________________________________________________ -->
3887 <div class="doc_subsubsection">
3888 <a name="i_uitofp">'<tt>uitofp .. to</tt>' Instruction</a>
3890 <div class="doc_text">
3894 <result> = uitofp <ty> <value> to <ty2> <i>; yields ty2</i>
3898 <p>The '<tt>uitofp</tt>' instruction regards <tt>value</tt> as an unsigned
3899 integer and converts that value to the <tt>ty2</tt> type.</p>
3902 <p>The '<tt>uitofp</tt>' instruction takes a value to cast, which must be a
3903 scalar or vector <a href="#t_integer">integer</a> value, and a type to cast it
3904 to <tt>ty2</tt>, which must be an <a href="#t_floating">floating point</a>
3905 type. If <tt>ty</tt> is a vector integer type, <tt>ty2</tt> must be a vector
3906 floating point type with the same number of elements as <tt>ty</tt></p>
3909 <p>The '<tt>uitofp</tt>' instruction interprets its operand as an unsigned
3910 integer quantity and converts it to the corresponding floating point value. If
3911 the value cannot fit in the floating point value, the results are undefined.</p>
3915 %X = uitofp i32 257 to float <i>; yields float:257.0</i>
3916 %Y = uitofp i8 -1 to double <i>; yields double:255.0</i>
3920 <!-- _______________________________________________________________________ -->
3921 <div class="doc_subsubsection">
3922 <a name="i_sitofp">'<tt>sitofp .. to</tt>' Instruction</a>
3924 <div class="doc_text">
3928 <result> = sitofp <ty> <value> to <ty2> <i>; yields ty2</i>
3932 <p>The '<tt>sitofp</tt>' instruction regards <tt>value</tt> as a signed
3933 integer and converts that value to the <tt>ty2</tt> type.</p>
3936 <p>The '<tt>sitofp</tt>' instruction takes a value to cast, which must be a
3937 scalar or vector <a href="#t_integer">integer</a> value, and a type to cast it
3938 to <tt>ty2</tt>, which must be an <a href="#t_floating">floating point</a>
3939 type. If <tt>ty</tt> is a vector integer type, <tt>ty2</tt> must be a vector
3940 floating point type with the same number of elements as <tt>ty</tt></p>
3943 <p>The '<tt>sitofp</tt>' instruction interprets its operand as a signed
3944 integer quantity and converts it to the corresponding floating point value. If
3945 the value cannot fit in the floating point value, the results are undefined.</p>
3949 %X = sitofp i32 257 to float <i>; yields float:257.0</i>
3950 %Y = sitofp i8 -1 to double <i>; yields double:-1.0</i>
3954 <!-- _______________________________________________________________________ -->
3955 <div class="doc_subsubsection">
3956 <a name="i_ptrtoint">'<tt>ptrtoint .. to</tt>' Instruction</a>
3958 <div class="doc_text">
3962 <result> = ptrtoint <ty> <value> to <ty2> <i>; yields ty2</i>
3966 <p>The '<tt>ptrtoint</tt>' instruction converts the pointer <tt>value</tt> to
3967 the integer type <tt>ty2</tt>.</p>
3970 <p>The '<tt>ptrtoint</tt>' instruction takes a <tt>value</tt> to cast, which
3971 must be a <a href="#t_pointer">pointer</a> value, and a type to cast it to
3972 <tt>ty2</tt>, which must be an <a href="#t_integer">integer</a> type.</p>
3975 <p>The '<tt>ptrtoint</tt>' instruction converts <tt>value</tt> to integer type
3976 <tt>ty2</tt> by interpreting the pointer value as an integer and either
3977 truncating or zero extending that value to the size of the integer type. If
3978 <tt>value</tt> is smaller than <tt>ty2</tt> then a zero extension is done. If
3979 <tt>value</tt> is larger than <tt>ty2</tt> then a truncation is done. If they
3980 are the same size, then nothing is done (<i>no-op cast</i>) other than a type
3985 %X = ptrtoint i32* %X to i8 <i>; yields truncation on 32-bit architecture</i>
3986 %Y = ptrtoint i32* %x to i64 <i>; yields zero extension on 32-bit architecture</i>
3990 <!-- _______________________________________________________________________ -->
3991 <div class="doc_subsubsection">
3992 <a name="i_inttoptr">'<tt>inttoptr .. to</tt>' Instruction</a>
3994 <div class="doc_text">
3998 <result> = inttoptr <ty> <value> to <ty2> <i>; yields ty2</i>
4002 <p>The '<tt>inttoptr</tt>' instruction converts an integer <tt>value</tt> to
4003 a pointer type, <tt>ty2</tt>.</p>
4006 <p>The '<tt>inttoptr</tt>' instruction takes an <a href="#t_integer">integer</a>
4007 value to cast, and a type to cast it to, which must be a
4008 <a href="#t_pointer">pointer</a> type.</p>
4011 <p>The '<tt>inttoptr</tt>' instruction converts <tt>value</tt> to type
4012 <tt>ty2</tt> by applying either a zero extension or a truncation depending on
4013 the size of the integer <tt>value</tt>. If <tt>value</tt> is larger than the
4014 size of a pointer then a truncation is done. If <tt>value</tt> is smaller than
4015 the size of a pointer then a zero extension is done. If they are the same size,
4016 nothing is done (<i>no-op cast</i>).</p>
4020 %X = inttoptr i32 255 to i32* <i>; yields zero extension on 64-bit architecture</i>
4021 %X = inttoptr i32 255 to i32* <i>; yields no-op on 32-bit architecture</i>
4022 %Y = inttoptr i64 0 to i32* <i>; yields truncation on 32-bit architecture</i>
4026 <!-- _______________________________________________________________________ -->
4027 <div class="doc_subsubsection">
4028 <a name="i_bitcast">'<tt>bitcast .. to</tt>' Instruction</a>
4030 <div class="doc_text">
4034 <result> = bitcast <ty> <value> to <ty2> <i>; yields ty2</i>
4039 <p>The '<tt>bitcast</tt>' instruction converts <tt>value</tt> to type
4040 <tt>ty2</tt> without changing any bits.</p>
4044 <p>The '<tt>bitcast</tt>' instruction takes a value to cast, which must be
4045 a non-aggregate first class value, and a type to cast it to, which must also be
4046 a non-aggregate <a href="#t_firstclass">first class</a> type. The bit sizes of
4048 and the destination type, <tt>ty2</tt>, must be identical. If the source
4049 type is a pointer, the destination type must also be a pointer. This
4050 instruction supports bitwise conversion of vectors to integers and to vectors
4051 of other types (as long as they have the same size).</p>
4054 <p>The '<tt>bitcast</tt>' instruction converts <tt>value</tt> to type
4055 <tt>ty2</tt>. It is always a <i>no-op cast</i> because no bits change with
4056 this conversion. The conversion is done as if the <tt>value</tt> had been
4057 stored to memory and read back as type <tt>ty2</tt>. Pointer types may only be
4058 converted to other pointer types with this instruction. To convert pointers to
4059 other types, use the <a href="#i_inttoptr">inttoptr</a> or
4060 <a href="#i_ptrtoint">ptrtoint</a> instructions first.</p>
4064 %X = bitcast i8 255 to i8 <i>; yields i8 :-1</i>
4065 %Y = bitcast i32* %x to sint* <i>; yields sint*:%x</i>
4066 %Z = bitcast <2 x int> %V to i64; <i>; yields i64: %V</i>
4070 <!-- ======================================================================= -->
4071 <div class="doc_subsection"> <a name="otherops">Other Operations</a> </div>
4072 <div class="doc_text">
4073 <p>The instructions in this category are the "miscellaneous"
4074 instructions, which defy better classification.</p>
4077 <!-- _______________________________________________________________________ -->
4078 <div class="doc_subsubsection"><a name="i_icmp">'<tt>icmp</tt>' Instruction</a>
4080 <div class="doc_text">
4082 <pre> <result> = icmp <cond> <ty> <op1>, <op2> <i>; yields {i1} or {<N x i1>}:result</i>
4085 <p>The '<tt>icmp</tt>' instruction returns a boolean value or
4086 a vector of boolean values based on comparison
4087 of its two integer, integer vector, or pointer operands.</p>
4089 <p>The '<tt>icmp</tt>' instruction takes three operands. The first operand is
4090 the condition code indicating the kind of comparison to perform. It is not
4091 a value, just a keyword. The possible condition code are:
4094 <li><tt>eq</tt>: equal</li>
4095 <li><tt>ne</tt>: not equal </li>
4096 <li><tt>ugt</tt>: unsigned greater than</li>
4097 <li><tt>uge</tt>: unsigned greater or equal</li>
4098 <li><tt>ult</tt>: unsigned less than</li>
4099 <li><tt>ule</tt>: unsigned less or equal</li>
4100 <li><tt>sgt</tt>: signed greater than</li>
4101 <li><tt>sge</tt>: signed greater or equal</li>
4102 <li><tt>slt</tt>: signed less than</li>
4103 <li><tt>sle</tt>: signed less or equal</li>
4105 <p>The remaining two arguments must be <a href="#t_integer">integer</a> or
4106 <a href="#t_pointer">pointer</a>
4107 or integer <a href="#t_vector">vector</a> typed.
4108 They must also be identical types.</p>
4110 <p>The '<tt>icmp</tt>' compares <tt>op1</tt> and <tt>op2</tt> according to
4111 the condition code given as <tt>cond</tt>. The comparison performed always
4112 yields either an <a href="#t_primitive"><tt>i1</tt></a> or vector of <tt>i1</tt> result, as follows:
4115 <li><tt>eq</tt>: yields <tt>true</tt> if the operands are equal,
4116 <tt>false</tt> otherwise. No sign interpretation is necessary or performed.
4118 <li><tt>ne</tt>: yields <tt>true</tt> if the operands are unequal,
4119 <tt>false</tt> otherwise. No sign interpretation is necessary or performed.</li>
4120 <li><tt>ugt</tt>: interprets the operands as unsigned values and yields
4121 <tt>true</tt> if <tt>op1</tt> is greater than <tt>op2</tt>.</li>
4122 <li><tt>uge</tt>: interprets the operands as unsigned values and yields
4123 <tt>true</tt> if <tt>op1</tt> is greater than or equal to <tt>op2</tt>.</li>
4124 <li><tt>ult</tt>: interprets the operands as unsigned values and yields
4125 <tt>true</tt> if <tt>op1</tt> is less than <tt>op2</tt>.</li>
4126 <li><tt>ule</tt>: interprets the operands as unsigned values and yields
4127 <tt>true</tt> if <tt>op1</tt> is less than or equal to <tt>op2</tt>.</li>
4128 <li><tt>sgt</tt>: interprets the operands as signed values and yields
4129 <tt>true</tt> if <tt>op1</tt> is greater than <tt>op2</tt>.</li>
4130 <li><tt>sge</tt>: interprets the operands as signed values and yields
4131 <tt>true</tt> if <tt>op1</tt> is greater than or equal to <tt>op2</tt>.</li>
4132 <li><tt>slt</tt>: interprets the operands as signed values and yields
4133 <tt>true</tt> if <tt>op1</tt> is less than <tt>op2</tt>.</li>
4134 <li><tt>sle</tt>: interprets the operands as signed values and yields
4135 <tt>true</tt> if <tt>op1</tt> is less than or equal to <tt>op2</tt>.</li>
4137 <p>If the operands are <a href="#t_pointer">pointer</a> typed, the pointer
4138 values are compared as if they were integers.</p>
4139 <p>If the operands are integer vectors, then they are compared
4140 element by element. The result is an <tt>i1</tt> vector with
4141 the same number of elements as the values being compared.
4142 Otherwise, the result is an <tt>i1</tt>.
4146 <pre> <result> = icmp eq i32 4, 5 <i>; yields: result=false</i>
4147 <result> = icmp ne float* %X, %X <i>; yields: result=false</i>
4148 <result> = icmp ult i16 4, 5 <i>; yields: result=true</i>
4149 <result> = icmp sgt i16 4, 5 <i>; yields: result=false</i>
4150 <result> = icmp ule i16 -4, 5 <i>; yields: result=false</i>
4151 <result> = icmp sge i16 4, 5 <i>; yields: result=false</i>
4154 <p>Note that the code generator does not yet support vector types with
4155 the <tt>icmp</tt> instruction.</p>
4159 <!-- _______________________________________________________________________ -->
4160 <div class="doc_subsubsection"><a name="i_fcmp">'<tt>fcmp</tt>' Instruction</a>
4162 <div class="doc_text">
4164 <pre> <result> = fcmp <cond> <ty> <op1>, <op2> <i>; yields {i1} or {<N x i1>}:result</i>
4167 <p>The '<tt>fcmp</tt>' instruction returns a boolean value
4168 or vector of boolean values based on comparison
4169 of its operands.</p>
4171 If the operands are floating point scalars, then the result
4172 type is a boolean (<a href="#t_primitive"><tt>i1</tt></a>).
4174 <p>If the operands are floating point vectors, then the result type
4175 is a vector of boolean with the same number of elements as the
4176 operands being compared.</p>
4178 <p>The '<tt>fcmp</tt>' instruction takes three operands. The first operand is
4179 the condition code indicating the kind of comparison to perform. It is not
4180 a value, just a keyword. The possible condition code are:</p>
4182 <li><tt>false</tt>: no comparison, always returns false</li>
4183 <li><tt>oeq</tt>: ordered and equal</li>
4184 <li><tt>ogt</tt>: ordered and greater than </li>
4185 <li><tt>oge</tt>: ordered and greater than or equal</li>
4186 <li><tt>olt</tt>: ordered and less than </li>
4187 <li><tt>ole</tt>: ordered and less than or equal</li>
4188 <li><tt>one</tt>: ordered and not equal</li>
4189 <li><tt>ord</tt>: ordered (no nans)</li>
4190 <li><tt>ueq</tt>: unordered or equal</li>
4191 <li><tt>ugt</tt>: unordered or greater than </li>
4192 <li><tt>uge</tt>: unordered or greater than or equal</li>
4193 <li><tt>ult</tt>: unordered or less than </li>
4194 <li><tt>ule</tt>: unordered or less than or equal</li>
4195 <li><tt>une</tt>: unordered or not equal</li>
4196 <li><tt>uno</tt>: unordered (either nans)</li>
4197 <li><tt>true</tt>: no comparison, always returns true</li>
4199 <p><i>Ordered</i> means that neither operand is a QNAN while
4200 <i>unordered</i> means that either operand may be a QNAN.</p>
4201 <p>Each of <tt>val1</tt> and <tt>val2</tt> arguments must be
4202 either a <a href="#t_floating">floating point</a> type
4203 or a <a href="#t_vector">vector</a> of floating point type.
4204 They must have identical types.</p>
4206 <p>The '<tt>fcmp</tt>' instruction compares <tt>op1</tt> and <tt>op2</tt>
4207 according to the condition code given as <tt>cond</tt>.
4208 If the operands are vectors, then the vectors are compared
4210 Each comparison performed
4211 always yields an <a href="#t_primitive">i1</a> result, as follows:</p>
4213 <li><tt>false</tt>: always yields <tt>false</tt>, regardless of operands.</li>
4214 <li><tt>oeq</tt>: yields <tt>true</tt> if both operands are not a QNAN and
4215 <tt>op1</tt> is equal to <tt>op2</tt>.</li>
4216 <li><tt>ogt</tt>: yields <tt>true</tt> if both operands are not a QNAN and
4217 <tt>op1</tt> is greather than <tt>op2</tt>.</li>
4218 <li><tt>oge</tt>: yields <tt>true</tt> if both operands are not a QNAN and
4219 <tt>op1</tt> is greater than or equal to <tt>op2</tt>.</li>
4220 <li><tt>olt</tt>: yields <tt>true</tt> if both operands are not a QNAN and
4221 <tt>op1</tt> is less than <tt>op2</tt>.</li>
4222 <li><tt>ole</tt>: yields <tt>true</tt> if both operands are not a QNAN and
4223 <tt>op1</tt> is less than or equal to <tt>op2</tt>.</li>
4224 <li><tt>one</tt>: yields <tt>true</tt> if both operands are not a QNAN and
4225 <tt>op1</tt> is not equal to <tt>op2</tt>.</li>
4226 <li><tt>ord</tt>: yields <tt>true</tt> if both operands are not a QNAN.</li>
4227 <li><tt>ueq</tt>: yields <tt>true</tt> if either operand is a QNAN or
4228 <tt>op1</tt> is equal to <tt>op2</tt>.</li>
4229 <li><tt>ugt</tt>: yields <tt>true</tt> if either operand is a QNAN or
4230 <tt>op1</tt> is greater than <tt>op2</tt>.</li>
4231 <li><tt>uge</tt>: yields <tt>true</tt> if either operand is a QNAN or
4232 <tt>op1</tt> is greater than or equal to <tt>op2</tt>.</li>
4233 <li><tt>ult</tt>: yields <tt>true</tt> if either operand is a QNAN or
4234 <tt>op1</tt> is less than <tt>op2</tt>.</li>
4235 <li><tt>ule</tt>: yields <tt>true</tt> if either operand is a QNAN or
4236 <tt>op1</tt> is less than or equal to <tt>op2</tt>.</li>
4237 <li><tt>une</tt>: yields <tt>true</tt> if either operand is a QNAN or
4238 <tt>op1</tt> is not equal to <tt>op2</tt>.</li>
4239 <li><tt>uno</tt>: yields <tt>true</tt> if either operand is a QNAN.</li>
4240 <li><tt>true</tt>: always yields <tt>true</tt>, regardless of operands.</li>
4244 <pre> <result> = fcmp oeq float 4.0, 5.0 <i>; yields: result=false</i>
4245 <result> = fcmp one float 4.0, 5.0 <i>; yields: result=true</i>
4246 <result> = fcmp olt float 4.0, 5.0 <i>; yields: result=true</i>
4247 <result> = fcmp ueq double 1.0, 2.0 <i>; yields: result=false</i>
4250 <p>Note that the code generator does not yet support vector types with
4251 the <tt>fcmp</tt> instruction.</p>
4255 <!-- _______________________________________________________________________ -->
4256 <div class="doc_subsubsection">
4257 <a name="i_vicmp">'<tt>vicmp</tt>' Instruction</a>
4259 <div class="doc_text">
4261 <pre> <result> = vicmp <cond> <ty> <op1>, <op2> <i>; yields {ty}:result</i>
4264 <p>The '<tt>vicmp</tt>' instruction returns an integer vector value based on
4265 element-wise comparison of its two integer vector operands.</p>
4267 <p>The '<tt>vicmp</tt>' instruction takes three operands. The first operand is
4268 the condition code indicating the kind of comparison to perform. It is not
4269 a value, just a keyword. The possible condition code are:</p>
4271 <li><tt>eq</tt>: equal</li>
4272 <li><tt>ne</tt>: not equal </li>
4273 <li><tt>ugt</tt>: unsigned greater than</li>
4274 <li><tt>uge</tt>: unsigned greater or equal</li>
4275 <li><tt>ult</tt>: unsigned less than</li>
4276 <li><tt>ule</tt>: unsigned less or equal</li>
4277 <li><tt>sgt</tt>: signed greater than</li>
4278 <li><tt>sge</tt>: signed greater or equal</li>
4279 <li><tt>slt</tt>: signed less than</li>
4280 <li><tt>sle</tt>: signed less or equal</li>
4282 <p>The remaining two arguments must be <a href="#t_vector">vector</a> or
4283 <a href="#t_integer">integer</a> typed. They must also be identical types.</p>
4285 <p>The '<tt>vicmp</tt>' instruction compares <tt>op1</tt> and <tt>op2</tt>
4286 according to the condition code given as <tt>cond</tt>. The comparison yields a
4287 <a href="#t_vector">vector</a> of <a href="#t_integer">integer</a> result, of
4288 identical type as the values being compared. The most significant bit in each
4289 element is 1 if the element-wise comparison evaluates to true, and is 0
4290 otherwise. All other bits of the result are undefined. The condition codes
4291 are evaluated identically to the <a href="#i_icmp">'<tt>icmp</tt>'
4292 instruction</a>.</p>
4296 <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>
4297 <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>
4301 <!-- _______________________________________________________________________ -->
4302 <div class="doc_subsubsection">
4303 <a name="i_vfcmp">'<tt>vfcmp</tt>' Instruction</a>
4305 <div class="doc_text">
4307 <pre> <result> = vfcmp <cond> <ty> <op1>, <op2></pre>
4309 <p>The '<tt>vfcmp</tt>' instruction returns an integer vector value based on
4310 element-wise comparison of its two floating point vector operands. The output
4311 elements have the same width as the input elements.</p>
4313 <p>The '<tt>vfcmp</tt>' instruction takes three operands. The first operand is
4314 the condition code indicating the kind of comparison to perform. It is not
4315 a value, just a keyword. The possible condition code are:</p>
4317 <li><tt>false</tt>: no comparison, always returns false</li>
4318 <li><tt>oeq</tt>: ordered and equal</li>
4319 <li><tt>ogt</tt>: ordered and greater than </li>
4320 <li><tt>oge</tt>: ordered and greater than or equal</li>
4321 <li><tt>olt</tt>: ordered and less than </li>
4322 <li><tt>ole</tt>: ordered and less than or equal</li>
4323 <li><tt>one</tt>: ordered and not equal</li>
4324 <li><tt>ord</tt>: ordered (no nans)</li>
4325 <li><tt>ueq</tt>: unordered or equal</li>
4326 <li><tt>ugt</tt>: unordered or greater than </li>
4327 <li><tt>uge</tt>: unordered or greater than or equal</li>
4328 <li><tt>ult</tt>: unordered or less than </li>
4329 <li><tt>ule</tt>: unordered or less than or equal</li>
4330 <li><tt>une</tt>: unordered or not equal</li>
4331 <li><tt>uno</tt>: unordered (either nans)</li>
4332 <li><tt>true</tt>: no comparison, always returns true</li>
4334 <p>The remaining two arguments must be <a href="#t_vector">vector</a> of
4335 <a href="#t_floating">floating point</a> typed. They must also be identical
4338 <p>The '<tt>vfcmp</tt>' instruction compares <tt>op1</tt> and <tt>op2</tt>
4339 according to the condition code given as <tt>cond</tt>. The comparison yields a
4340 <a href="#t_vector">vector</a> of <a href="#t_integer">integer</a> result, with
4341 an identical number of elements as the values being compared, and each element
4342 having identical with to the width of the floating point elements. The most
4343 significant bit in each element is 1 if the element-wise comparison evaluates to
4344 true, and is 0 otherwise. All other bits of the result are undefined. The
4345 condition codes are evaluated identically to the
4346 <a href="#i_fcmp">'<tt>fcmp</tt>' instruction</a>.</p>
4350 <i>; yields: result=<2 x i32> < i32 0, i32 -1 ></i>
4351 <result> = vfcmp oeq <2 x float> < float 4, float 0 >, < float 5, float 0 >
4353 <i>; yields: result=<2 x i64> < i64 -1, i64 0 ></i>
4354 <result> = vfcmp ult <2 x double> < double 1, double 2 >, < double 2, double 2>
4358 <!-- _______________________________________________________________________ -->
4359 <div class="doc_subsubsection">
4360 <a name="i_phi">'<tt>phi</tt>' Instruction</a>
4363 <div class="doc_text">
4367 <pre> <result> = phi <ty> [ <val0>, <label0>], ...<br></pre>
4369 <p>The '<tt>phi</tt>' instruction is used to implement the φ node in
4370 the SSA graph representing the function.</p>
4373 <p>The type of the incoming values is specified with the first type
4374 field. After this, the '<tt>phi</tt>' instruction takes a list of pairs
4375 as arguments, with one pair for each predecessor basic block of the
4376 current block. Only values of <a href="#t_firstclass">first class</a>
4377 type may be used as the value arguments to the PHI node. Only labels
4378 may be used as the label arguments.</p>
4380 <p>There must be no non-phi instructions between the start of a basic
4381 block and the PHI instructions: i.e. PHI instructions must be first in
4386 <p>At runtime, the '<tt>phi</tt>' instruction logically takes on the value
4387 specified by the pair corresponding to the predecessor basic block that executed
4388 just prior to the current block.</p>
4392 Loop: ; Infinite loop that counts from 0 on up...
4393 %indvar = phi i32 [ 0, %LoopHeader ], [ %nextindvar, %Loop ]
4394 %nextindvar = add i32 %indvar, 1
4399 <!-- _______________________________________________________________________ -->
4400 <div class="doc_subsubsection">
4401 <a name="i_select">'<tt>select</tt>' Instruction</a>
4404 <div class="doc_text">
4409 <result> = select <i>selty</i> <cond>, <ty> <val1>, <ty> <val2> <i>; yields ty</i>
4411 <i>selty</i> is either i1 or {<N x i1>}
4417 The '<tt>select</tt>' instruction is used to choose one value based on a
4418 condition, without branching.
4425 The '<tt>select</tt>' instruction requires an 'i1' value or
4426 a vector of 'i1' values indicating the
4427 condition, and two values of the same <a href="#t_firstclass">first class</a>
4428 type. If the val1/val2 are vectors and
4429 the condition is a scalar, then entire vectors are selected, not
4430 individual elements.
4436 If the condition is an i1 and it evaluates to 1, the instruction returns the first
4437 value argument; otherwise, it returns the second value argument.
4440 If the condition is a vector of i1, then the value arguments must
4441 be vectors of the same size, and the selection is done element
4448 %X = select i1 true, i8 17, i8 42 <i>; yields i8:17</i>
4451 <p>Note that the code generator does not yet support conditions
4452 with vector type.</p>
4457 <!-- _______________________________________________________________________ -->
4458 <div class="doc_subsubsection">
4459 <a name="i_call">'<tt>call</tt>' Instruction</a>
4462 <div class="doc_text">
4466 <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>]
4471 <p>The '<tt>call</tt>' instruction represents a simple function call.</p>
4475 <p>This instruction requires several arguments:</p>
4479 <p>The optional "tail" marker indicates whether the callee function accesses
4480 any allocas or varargs in the caller. If the "tail" marker is present, the
4481 function call is eligible for tail call optimization. Note that calls may
4482 be marked "tail" even if they do not occur before a <a
4483 href="#i_ret"><tt>ret</tt></a> instruction.</p>
4486 <p>The optional "cconv" marker indicates which <a href="#callingconv">calling
4487 convention</a> the call should use. If none is specified, the call defaults
4488 to using C calling conventions.</p>
4492 <p>The optional <a href="#paramattrs">Parameter Attributes</a> list for
4493 return values. Only '<tt>zeroext</tt>', '<tt>signext</tt>',
4494 and '<tt>inreg</tt>' attributes are valid here.</p>
4498 <p>'<tt>ty</tt>': the type of the call instruction itself which is also
4499 the type of the return value. Functions that return no value are marked
4500 <tt><a href="#t_void">void</a></tt>.</p>
4503 <p>'<tt>fnty</tt>': shall be the signature of the pointer to function
4504 value being invoked. The argument types must match the types implied by
4505 this signature. This type can be omitted if the function is not varargs
4506 and if the function type does not return a pointer to a function.</p>
4509 <p>'<tt>fnptrval</tt>': An LLVM value containing a pointer to a function to
4510 be invoked. In most cases, this is a direct function invocation, but
4511 indirect <tt>call</tt>s are just as possible, calling an arbitrary pointer
4512 to function value.</p>
4515 <p>'<tt>function args</tt>': argument list whose types match the
4516 function signature argument types. All arguments must be of
4517 <a href="#t_firstclass">first class</a> type. If the function signature
4518 indicates the function accepts a variable number of arguments, the extra
4519 arguments can be specified.</p>
4522 <p>The optional <a href="#fnattrs">function attributes</a> list. Only
4523 '<tt>noreturn</tt>', '<tt>nounwind</tt>', '<tt>readonly</tt>' and
4524 '<tt>readnone</tt>' attributes are valid here.</p>
4530 <p>The '<tt>call</tt>' instruction is used to cause control flow to
4531 transfer to a specified function, with its incoming arguments bound to
4532 the specified values. Upon a '<tt><a href="#i_ret">ret</a></tt>'
4533 instruction in the called function, control flow continues with the
4534 instruction after the function call, and the return value of the
4535 function is bound to the result argument.</p>
4540 %retval = call i32 @test(i32 %argc)
4541 call i32 (i8 *, ...)* @printf(i8 * %msg, i32 12, i8 42) <i>; yields i32</i>
4542 %X = tail call i32 @foo() <i>; yields i32</i>
4543 %Y = tail call <a href="#callingconv">fastcc</a> i32 @foo() <i>; yields i32</i>
4544 call void %foo(i8 97 signext)
4546 %struct.A = type { i32, i8 }
4547 %r = call %struct.A @foo() <i>; yields { 32, i8 }</i>
4548 %gr = extractvalue %struct.A %r, 0 <i>; yields i32</i>
4549 %gr1 = extractvalue %struct.A %r, 1 <i>; yields i8</i>
4550 %Z = call void @foo() noreturn <i>; indicates that %foo never returns normally</i>
4551 %ZZ = call zeroext i32 @bar() <i>; Return value is %zero extended</i>
4556 <!-- _______________________________________________________________________ -->
4557 <div class="doc_subsubsection">
4558 <a name="i_va_arg">'<tt>va_arg</tt>' Instruction</a>
4561 <div class="doc_text">
4566 <resultval> = va_arg <va_list*> <arglist>, <argty>
4571 <p>The '<tt>va_arg</tt>' instruction is used to access arguments passed through
4572 the "variable argument" area of a function call. It is used to implement the
4573 <tt>va_arg</tt> macro in C.</p>
4577 <p>This instruction takes a <tt>va_list*</tt> value and the type of
4578 the argument. It returns a value of the specified argument type and
4579 increments the <tt>va_list</tt> to point to the next argument. The
4580 actual type of <tt>va_list</tt> is target specific.</p>
4584 <p>The '<tt>va_arg</tt>' instruction loads an argument of the specified
4585 type from the specified <tt>va_list</tt> and causes the
4586 <tt>va_list</tt> to point to the next argument. For more information,
4587 see the variable argument handling <a href="#int_varargs">Intrinsic
4590 <p>It is legal for this instruction to be called in a function which does not
4591 take a variable number of arguments, for example, the <tt>vfprintf</tt>
4594 <p><tt>va_arg</tt> is an LLVM instruction instead of an <a
4595 href="#intrinsics">intrinsic function</a> because it takes a type as an
4600 <p>See the <a href="#int_varargs">variable argument processing</a> section.</p>
4602 <p>Note that the code generator does not yet fully support va_arg
4603 on many targets. Also, it does not currently support va_arg with
4604 aggregate types on any target.</p>
4608 <!-- *********************************************************************** -->
4609 <div class="doc_section"> <a name="intrinsics">Intrinsic Functions</a> </div>
4610 <!-- *********************************************************************** -->
4612 <div class="doc_text">
4614 <p>LLVM supports the notion of an "intrinsic function". These functions have
4615 well known names and semantics and are required to follow certain restrictions.
4616 Overall, these intrinsics represent an extension mechanism for the LLVM
4617 language that does not require changing all of the transformations in LLVM when
4618 adding to the language (or the bitcode reader/writer, the parser, etc...).</p>
4620 <p>Intrinsic function names must all start with an "<tt>llvm.</tt>" prefix. This
4621 prefix is reserved in LLVM for intrinsic names; thus, function names may not
4622 begin with this prefix. Intrinsic functions must always be external functions:
4623 you cannot define the body of intrinsic functions. Intrinsic functions may
4624 only be used in call or invoke instructions: it is illegal to take the address
4625 of an intrinsic function. Additionally, because intrinsic functions are part
4626 of the LLVM language, it is required if any are added that they be documented
4629 <p>Some intrinsic functions can be overloaded, i.e., the intrinsic represents
4630 a family of functions that perform the same operation but on different data
4631 types. Because LLVM can represent over 8 million different integer types,
4632 overloading is used commonly to allow an intrinsic function to operate on any
4633 integer type. One or more of the argument types or the result type can be
4634 overloaded to accept any integer type. Argument types may also be defined as
4635 exactly matching a previous argument's type or the result type. This allows an
4636 intrinsic function which accepts multiple arguments, but needs all of them to
4637 be of the same type, to only be overloaded with respect to a single argument or
4640 <p>Overloaded intrinsics will have the names of its overloaded argument types
4641 encoded into its function name, each preceded by a period. Only those types
4642 which are overloaded result in a name suffix. Arguments whose type is matched
4643 against another type do not. For example, the <tt>llvm.ctpop</tt> function can
4644 take an integer of any width and returns an integer of exactly the same integer
4645 width. This leads to a family of functions such as
4646 <tt>i8 @llvm.ctpop.i8(i8 %val)</tt> and <tt>i29 @llvm.ctpop.i29(i29 %val)</tt>.
4647 Only one type, the return type, is overloaded, and only one type suffix is
4648 required. Because the argument's type is matched against the return type, it
4649 does not require its own name suffix.</p>
4651 <p>To learn how to add an intrinsic function, please see the
4652 <a href="ExtendingLLVM.html">Extending LLVM Guide</a>.
4657 <!-- ======================================================================= -->
4658 <div class="doc_subsection">
4659 <a name="int_varargs">Variable Argument Handling Intrinsics</a>
4662 <div class="doc_text">
4664 <p>Variable argument support is defined in LLVM with the <a
4665 href="#i_va_arg"><tt>va_arg</tt></a> instruction and these three
4666 intrinsic functions. These functions are related to the similarly
4667 named macros defined in the <tt><stdarg.h></tt> header file.</p>
4669 <p>All of these functions operate on arguments that use a
4670 target-specific value type "<tt>va_list</tt>". The LLVM assembly
4671 language reference manual does not define what this type is, so all
4672 transformations should be prepared to handle these functions regardless of
4675 <p>This example shows how the <a href="#i_va_arg"><tt>va_arg</tt></a>
4676 instruction and the variable argument handling intrinsic functions are
4679 <div class="doc_code">
4681 define i32 @test(i32 %X, ...) {
4682 ; Initialize variable argument processing
4684 %ap2 = bitcast i8** %ap to i8*
4685 call void @llvm.va_start(i8* %ap2)
4687 ; Read a single integer argument
4688 %tmp = va_arg i8** %ap, i32
4690 ; Demonstrate usage of llvm.va_copy and llvm.va_end
4692 %aq2 = bitcast i8** %aq to i8*
4693 call void @llvm.va_copy(i8* %aq2, i8* %ap2)
4694 call void @llvm.va_end(i8* %aq2)
4696 ; Stop processing of arguments.
4697 call void @llvm.va_end(i8* %ap2)
4701 declare void @llvm.va_start(i8*)
4702 declare void @llvm.va_copy(i8*, i8*)
4703 declare void @llvm.va_end(i8*)
4709 <!-- _______________________________________________________________________ -->
4710 <div class="doc_subsubsection">
4711 <a name="int_va_start">'<tt>llvm.va_start</tt>' Intrinsic</a>
4715 <div class="doc_text">
4717 <pre> declare void %llvm.va_start(i8* <arglist>)<br></pre>
4719 <p>The '<tt>llvm.va_start</tt>' intrinsic initializes
4720 <tt>*<arglist></tt> for subsequent use by <tt><a
4721 href="#i_va_arg">va_arg</a></tt>.</p>
4725 <p>The argument is a pointer to a <tt>va_list</tt> element to initialize.</p>
4729 <p>The '<tt>llvm.va_start</tt>' intrinsic works just like the <tt>va_start</tt>
4730 macro available in C. In a target-dependent way, it initializes the
4731 <tt>va_list</tt> element to which the argument points, so that the next call to
4732 <tt>va_arg</tt> will produce the first variable argument passed to the function.
4733 Unlike the C <tt>va_start</tt> macro, this intrinsic does not need to know the
4734 last argument of the function as the compiler can figure that out.</p>
4738 <!-- _______________________________________________________________________ -->
4739 <div class="doc_subsubsection">
4740 <a name="int_va_end">'<tt>llvm.va_end</tt>' Intrinsic</a>
4743 <div class="doc_text">
4745 <pre> declare void @llvm.va_end(i8* <arglist>)<br></pre>
4748 <p>The '<tt>llvm.va_end</tt>' intrinsic destroys <tt>*<arglist></tt>,
4749 which has been initialized previously with <tt><a href="#int_va_start">llvm.va_start</a></tt>
4750 or <tt><a href="#i_va_copy">llvm.va_copy</a></tt>.</p>
4754 <p>The argument is a pointer to a <tt>va_list</tt> to destroy.</p>
4758 <p>The '<tt>llvm.va_end</tt>' intrinsic works just like the <tt>va_end</tt>
4759 macro available in C. In a target-dependent way, it destroys the
4760 <tt>va_list</tt> element to which the argument points. Calls to <a
4761 href="#int_va_start"><tt>llvm.va_start</tt></a> and <a href="#int_va_copy">
4762 <tt>llvm.va_copy</tt></a> must be matched exactly with calls to
4763 <tt>llvm.va_end</tt>.</p>
4767 <!-- _______________________________________________________________________ -->
4768 <div class="doc_subsubsection">
4769 <a name="int_va_copy">'<tt>llvm.va_copy</tt>' Intrinsic</a>
4772 <div class="doc_text">
4777 declare void @llvm.va_copy(i8* <destarglist>, i8* <srcarglist>)
4782 <p>The '<tt>llvm.va_copy</tt>' intrinsic copies the current argument position
4783 from the source argument list to the destination argument list.</p>
4787 <p>The first argument is a pointer to a <tt>va_list</tt> element to initialize.
4788 The second argument is a pointer to a <tt>va_list</tt> element to copy from.</p>
4793 <p>The '<tt>llvm.va_copy</tt>' intrinsic works just like the <tt>va_copy</tt>
4794 macro available in C. In a target-dependent way, it copies the source
4795 <tt>va_list</tt> element into the destination <tt>va_list</tt> element. This
4796 intrinsic is necessary because the <tt><a href="#int_va_start">
4797 llvm.va_start</a></tt> intrinsic may be arbitrarily complex and require, for
4798 example, memory allocation.</p>
4802 <!-- ======================================================================= -->
4803 <div class="doc_subsection">
4804 <a name="int_gc">Accurate Garbage Collection Intrinsics</a>
4807 <div class="doc_text">
4810 LLVM support for <a href="GarbageCollection.html">Accurate Garbage
4811 Collection</a> (GC) requires the implementation and generation of these
4813 These intrinsics allow identification of <a href="#int_gcroot">GC roots on the
4814 stack</a>, as well as garbage collector implementations that require <a
4815 href="#int_gcread">read</a> and <a href="#int_gcwrite">write</a> barriers.
4816 Front-ends for type-safe garbage collected languages should generate these
4817 intrinsics to make use of the LLVM garbage collectors. For more details, see <a
4818 href="GarbageCollection.html">Accurate Garbage Collection with LLVM</a>.
4821 <p>The garbage collection intrinsics only operate on objects in the generic
4822 address space (address space zero).</p>
4826 <!-- _______________________________________________________________________ -->
4827 <div class="doc_subsubsection">
4828 <a name="int_gcroot">'<tt>llvm.gcroot</tt>' Intrinsic</a>
4831 <div class="doc_text">
4836 declare void @llvm.gcroot(i8** %ptrloc, i8* %metadata)
4841 <p>The '<tt>llvm.gcroot</tt>' intrinsic declares the existence of a GC root to
4842 the code generator, and allows some metadata to be associated with it.</p>
4846 <p>The first argument specifies the address of a stack object that contains the
4847 root pointer. The second pointer (which must be either a constant or a global
4848 value address) contains the meta-data to be associated with the root.</p>
4852 <p>At runtime, a call to this intrinsic stores a null pointer into the "ptrloc"
4853 location. At compile-time, the code generator generates information to allow
4854 the runtime to find the pointer at GC safe points. The '<tt>llvm.gcroot</tt>'
4855 intrinsic may only be used in a function which <a href="#gc">specifies a GC
4861 <!-- _______________________________________________________________________ -->
4862 <div class="doc_subsubsection">
4863 <a name="int_gcread">'<tt>llvm.gcread</tt>' Intrinsic</a>
4866 <div class="doc_text">
4871 declare i8* @llvm.gcread(i8* %ObjPtr, i8** %Ptr)
4876 <p>The '<tt>llvm.gcread</tt>' intrinsic identifies reads of references from heap
4877 locations, allowing garbage collector implementations that require read
4882 <p>The second argument is the address to read from, which should be an address
4883 allocated from the garbage collector. The first object is a pointer to the
4884 start of the referenced object, if needed by the language runtime (otherwise
4889 <p>The '<tt>llvm.gcread</tt>' intrinsic has the same semantics as a load
4890 instruction, but may be replaced with substantially more complex code by the
4891 garbage collector runtime, as needed. The '<tt>llvm.gcread</tt>' intrinsic
4892 may only be used in a function which <a href="#gc">specifies a GC
4898 <!-- _______________________________________________________________________ -->
4899 <div class="doc_subsubsection">
4900 <a name="int_gcwrite">'<tt>llvm.gcwrite</tt>' Intrinsic</a>
4903 <div class="doc_text">
4908 declare void @llvm.gcwrite(i8* %P1, i8* %Obj, i8** %P2)
4913 <p>The '<tt>llvm.gcwrite</tt>' intrinsic identifies writes of references to heap
4914 locations, allowing garbage collector implementations that require write
4915 barriers (such as generational or reference counting collectors).</p>
4919 <p>The first argument is the reference to store, the second is the start of the
4920 object to store it to, and the third is the address of the field of Obj to
4921 store to. If the runtime does not require a pointer to the object, Obj may be
4926 <p>The '<tt>llvm.gcwrite</tt>' intrinsic has the same semantics as a store
4927 instruction, but may be replaced with substantially more complex code by the
4928 garbage collector runtime, as needed. The '<tt>llvm.gcwrite</tt>' intrinsic
4929 may only be used in a function which <a href="#gc">specifies a GC
4936 <!-- ======================================================================= -->
4937 <div class="doc_subsection">
4938 <a name="int_codegen">Code Generator Intrinsics</a>
4941 <div class="doc_text">
4943 These intrinsics are provided by LLVM to expose special features that may only
4944 be implemented with code generator support.
4949 <!-- _______________________________________________________________________ -->
4950 <div class="doc_subsubsection">
4951 <a name="int_returnaddress">'<tt>llvm.returnaddress</tt>' Intrinsic</a>
4954 <div class="doc_text">
4958 declare i8 *@llvm.returnaddress(i32 <level>)
4964 The '<tt>llvm.returnaddress</tt>' intrinsic attempts to compute a
4965 target-specific value indicating the return address of the current function
4966 or one of its callers.
4972 The argument to this intrinsic indicates which function to return the address
4973 for. Zero indicates the calling function, one indicates its caller, etc. The
4974 argument is <b>required</b> to be a constant integer value.
4980 The '<tt>llvm.returnaddress</tt>' intrinsic either returns a pointer indicating
4981 the return address of the specified call frame, or zero if it cannot be
4982 identified. The value returned by this intrinsic is likely to be incorrect or 0
4983 for arguments other than zero, so it should only be used for debugging purposes.
4987 Note that calling this intrinsic does not prevent function inlining or other
4988 aggressive transformations, so the value returned may not be that of the obvious
4989 source-language caller.
4994 <!-- _______________________________________________________________________ -->
4995 <div class="doc_subsubsection">
4996 <a name="int_frameaddress">'<tt>llvm.frameaddress</tt>' Intrinsic</a>
4999 <div class="doc_text">
5003 declare i8 *@llvm.frameaddress(i32 <level>)
5009 The '<tt>llvm.frameaddress</tt>' intrinsic attempts to return the
5010 target-specific frame pointer value for the specified stack frame.
5016 The argument to this intrinsic indicates which function to return the frame
5017 pointer for. Zero indicates the calling function, one indicates its caller,
5018 etc. The argument is <b>required</b> to be a constant integer value.
5024 The '<tt>llvm.frameaddress</tt>' intrinsic either returns a pointer indicating
5025 the frame address of the specified call frame, or zero if it cannot be
5026 identified. The value returned by this intrinsic is likely to be incorrect or 0
5027 for arguments other than zero, so it should only be used for debugging purposes.
5031 Note that calling this intrinsic does not prevent function inlining or other
5032 aggressive transformations, so the value returned may not be that of the obvious
5033 source-language caller.
5037 <!-- _______________________________________________________________________ -->
5038 <div class="doc_subsubsection">
5039 <a name="int_stacksave">'<tt>llvm.stacksave</tt>' Intrinsic</a>
5042 <div class="doc_text">
5046 declare i8 *@llvm.stacksave()
5052 The '<tt>llvm.stacksave</tt>' intrinsic is used to remember the current state of
5053 the function stack, for use with <a href="#int_stackrestore">
5054 <tt>llvm.stackrestore</tt></a>. This is useful for implementing language
5055 features like scoped automatic variable sized arrays in C99.
5061 This intrinsic returns a opaque pointer value that can be passed to <a
5062 href="#int_stackrestore"><tt>llvm.stackrestore</tt></a>. When an
5063 <tt>llvm.stackrestore</tt> intrinsic is executed with a value saved from
5064 <tt>llvm.stacksave</tt>, it effectively restores the state of the stack to the
5065 state it was in when the <tt>llvm.stacksave</tt> intrinsic executed. In
5066 practice, this pops any <a href="#i_alloca">alloca</a> blocks from the stack
5067 that were allocated after the <tt>llvm.stacksave</tt> was executed.
5072 <!-- _______________________________________________________________________ -->
5073 <div class="doc_subsubsection">
5074 <a name="int_stackrestore">'<tt>llvm.stackrestore</tt>' Intrinsic</a>
5077 <div class="doc_text">
5081 declare void @llvm.stackrestore(i8 * %ptr)
5087 The '<tt>llvm.stackrestore</tt>' intrinsic is used to restore the state of
5088 the function stack to the state it was in when the corresponding <a
5089 href="#int_stacksave"><tt>llvm.stacksave</tt></a> intrinsic executed. This is
5090 useful for implementing language features like scoped automatic variable sized
5097 See the description for <a href="#int_stacksave"><tt>llvm.stacksave</tt></a>.
5103 <!-- _______________________________________________________________________ -->
5104 <div class="doc_subsubsection">
5105 <a name="int_prefetch">'<tt>llvm.prefetch</tt>' Intrinsic</a>
5108 <div class="doc_text">
5112 declare void @llvm.prefetch(i8* <address>, i32 <rw>, i32 <locality>)
5119 The '<tt>llvm.prefetch</tt>' intrinsic is a hint to the code generator to insert
5120 a prefetch instruction if supported; otherwise, it is a noop. Prefetches have
5122 effect on the behavior of the program but can change its performance
5129 <tt>address</tt> is the address to be prefetched, <tt>rw</tt> is the specifier
5130 determining if the fetch should be for a read (0) or write (1), and
5131 <tt>locality</tt> is a temporal locality specifier ranging from (0) - no
5132 locality, to (3) - extremely local keep in cache. The <tt>rw</tt> and
5133 <tt>locality</tt> arguments must be constant integers.
5139 This intrinsic does not modify the behavior of the program. In particular,
5140 prefetches cannot trap and do not produce a value. On targets that support this
5141 intrinsic, the prefetch can provide hints to the processor cache for better
5147 <!-- _______________________________________________________________________ -->
5148 <div class="doc_subsubsection">
5149 <a name="int_pcmarker">'<tt>llvm.pcmarker</tt>' Intrinsic</a>
5152 <div class="doc_text">
5156 declare void @llvm.pcmarker(i32 <id>)
5163 The '<tt>llvm.pcmarker</tt>' intrinsic is a method to export a Program Counter
5165 code to simulators and other tools. The method is target specific, but it is
5166 expected that the marker will use exported symbols to transmit the PC of the
5168 The marker makes no guarantees that it will remain with any specific instruction
5169 after optimizations. It is possible that the presence of a marker will inhibit
5170 optimizations. The intended use is to be inserted after optimizations to allow
5171 correlations of simulation runs.
5177 <tt>id</tt> is a numerical id identifying the marker.
5183 This intrinsic does not modify the behavior of the program. Backends that do not
5184 support this intrinisic may ignore it.
5189 <!-- _______________________________________________________________________ -->
5190 <div class="doc_subsubsection">
5191 <a name="int_readcyclecounter">'<tt>llvm.readcyclecounter</tt>' Intrinsic</a>
5194 <div class="doc_text">
5198 declare i64 @llvm.readcyclecounter( )
5205 The '<tt>llvm.readcyclecounter</tt>' intrinsic provides access to the cycle
5206 counter register (or similar low latency, high accuracy clocks) on those targets
5207 that support it. On X86, it should map to RDTSC. On Alpha, it should map to RPCC.
5208 As the backing counters overflow quickly (on the order of 9 seconds on alpha), this
5209 should only be used for small timings.
5215 When directly supported, reading the cycle counter should not modify any memory.
5216 Implementations are allowed to either return a application specific value or a
5217 system wide value. On backends without support, this is lowered to a constant 0.
5222 <!-- ======================================================================= -->
5223 <div class="doc_subsection">
5224 <a name="int_libc">Standard C Library Intrinsics</a>
5227 <div class="doc_text">
5229 LLVM provides intrinsics for a few important standard C library functions.
5230 These intrinsics allow source-language front-ends to pass information about the
5231 alignment of the pointer arguments to the code generator, providing opportunity
5232 for more efficient code generation.
5237 <!-- _______________________________________________________________________ -->
5238 <div class="doc_subsubsection">
5239 <a name="int_memcpy">'<tt>llvm.memcpy</tt>' Intrinsic</a>
5242 <div class="doc_text">
5245 <p>This is an overloaded intrinsic. You can use llvm.memcpy on any integer bit
5246 width. Not all targets support all bit widths however.</p>
5248 declare void @llvm.memcpy.i8(i8 * <dest>, i8 * <src>,
5249 i8 <len>, i32 <align>)
5250 declare void @llvm.memcpy.i16(i8 * <dest>, i8 * <src>,
5251 i16 <len>, i32 <align>)
5252 declare void @llvm.memcpy.i32(i8 * <dest>, i8 * <src>,
5253 i32 <len>, i32 <align>)
5254 declare void @llvm.memcpy.i64(i8 * <dest>, i8 * <src>,
5255 i64 <len>, i32 <align>)
5261 The '<tt>llvm.memcpy.*</tt>' intrinsics copy a block of memory from the source
5262 location to the destination location.
5266 Note that, unlike the standard libc function, the <tt>llvm.memcpy.*</tt>
5267 intrinsics do not return a value, and takes an extra alignment argument.
5273 The first argument is a pointer to the destination, the second is a pointer to
5274 the source. The third argument is an integer argument
5275 specifying the number of bytes to copy, and the fourth argument is the alignment
5276 of the source and destination locations.
5280 If the call to this intrinisic has an alignment value that is not 0 or 1, then
5281 the caller guarantees that both the source and destination pointers are aligned
5288 The '<tt>llvm.memcpy.*</tt>' intrinsics copy a block of memory from the source
5289 location to the destination location, which are not allowed to overlap. It
5290 copies "len" bytes of memory over. If the argument is known to be aligned to
5291 some boundary, this can be specified as the fourth argument, otherwise it should
5297 <!-- _______________________________________________________________________ -->
5298 <div class="doc_subsubsection">
5299 <a name="int_memmove">'<tt>llvm.memmove</tt>' Intrinsic</a>
5302 <div class="doc_text">
5305 <p>This is an overloaded intrinsic. You can use llvm.memmove on any integer bit
5306 width. Not all targets support all bit widths however.</p>
5308 declare void @llvm.memmove.i8(i8 * <dest>, i8 * <src>,
5309 i8 <len>, i32 <align>)
5310 declare void @llvm.memmove.i16(i8 * <dest>, i8 * <src>,
5311 i16 <len>, i32 <align>)
5312 declare void @llvm.memmove.i32(i8 * <dest>, i8 * <src>,
5313 i32 <len>, i32 <align>)
5314 declare void @llvm.memmove.i64(i8 * <dest>, i8 * <src>,
5315 i64 <len>, i32 <align>)
5321 The '<tt>llvm.memmove.*</tt>' intrinsics move a block of memory from the source
5322 location to the destination location. It is similar to the
5323 '<tt>llvm.memcpy</tt>' intrinsic but allows the two memory locations to overlap.
5327 Note that, unlike the standard libc function, the <tt>llvm.memmove.*</tt>
5328 intrinsics do not return a value, and takes an extra alignment argument.
5334 The first argument is a pointer to the destination, the second is a pointer to
5335 the source. The third argument is an integer argument
5336 specifying the number of bytes to copy, and the fourth argument is the alignment
5337 of the source and destination locations.
5341 If the call to this intrinisic has an alignment value that is not 0 or 1, then
5342 the caller guarantees that the source and destination pointers are aligned to
5349 The '<tt>llvm.memmove.*</tt>' intrinsics copy a block of memory from the source
5350 location to the destination location, which may overlap. It
5351 copies "len" bytes of memory over. If the argument is known to be aligned to
5352 some boundary, this can be specified as the fourth argument, otherwise it should
5358 <!-- _______________________________________________________________________ -->
5359 <div class="doc_subsubsection">
5360 <a name="int_memset">'<tt>llvm.memset.*</tt>' Intrinsics</a>
5363 <div class="doc_text">
5366 <p>This is an overloaded intrinsic. You can use llvm.memset on any integer bit
5367 width. Not all targets support all bit widths however.</p>
5369 declare void @llvm.memset.i8(i8 * <dest>, i8 <val>,
5370 i8 <len>, i32 <align>)
5371 declare void @llvm.memset.i16(i8 * <dest>, i8 <val>,
5372 i16 <len>, i32 <align>)
5373 declare void @llvm.memset.i32(i8 * <dest>, i8 <val>,
5374 i32 <len>, i32 <align>)
5375 declare void @llvm.memset.i64(i8 * <dest>, i8 <val>,
5376 i64 <len>, i32 <align>)
5382 The '<tt>llvm.memset.*</tt>' intrinsics fill a block of memory with a particular
5387 Note that, unlike the standard libc function, the <tt>llvm.memset</tt> intrinsic
5388 does not return a value, and takes an extra alignment argument.
5394 The first argument is a pointer to the destination to fill, the second is the
5395 byte value to fill it with, the third argument is an integer
5396 argument specifying the number of bytes to fill, and the fourth argument is the
5397 known alignment of destination location.
5401 If the call to this intrinisic has an alignment value that is not 0 or 1, then
5402 the caller guarantees that the destination pointer is aligned to that boundary.
5408 The '<tt>llvm.memset.*</tt>' intrinsics fill "len" bytes of memory starting at
5410 destination location. If the argument is known to be aligned to some boundary,
5411 this can be specified as the fourth argument, otherwise it should be set to 0 or
5417 <!-- _______________________________________________________________________ -->
5418 <div class="doc_subsubsection">
5419 <a name="int_sqrt">'<tt>llvm.sqrt.*</tt>' Intrinsic</a>
5422 <div class="doc_text">
5425 <p>This is an overloaded intrinsic. You can use <tt>llvm.sqrt</tt> on any
5426 floating point or vector of floating point type. Not all targets support all
5429 declare float @llvm.sqrt.f32(float %Val)
5430 declare double @llvm.sqrt.f64(double %Val)
5431 declare x86_fp80 @llvm.sqrt.f80(x86_fp80 %Val)
5432 declare fp128 @llvm.sqrt.f128(fp128 %Val)
5433 declare ppc_fp128 @llvm.sqrt.ppcf128(ppc_fp128 %Val)
5439 The '<tt>llvm.sqrt</tt>' intrinsics return the sqrt of the specified operand,
5440 returning the same value as the libm '<tt>sqrt</tt>' functions would. Unlike
5441 <tt>sqrt</tt> in libm, however, <tt>llvm.sqrt</tt> has undefined behavior for
5442 negative numbers other than -0.0 (which allows for better optimization, because
5443 there is no need to worry about errno being set). <tt>llvm.sqrt(-0.0)</tt> is
5444 defined to return -0.0 like IEEE sqrt.
5450 The argument and return value are floating point numbers of the same type.
5456 This function returns the sqrt of the specified operand if it is a nonnegative
5457 floating point number.
5461 <!-- _______________________________________________________________________ -->
5462 <div class="doc_subsubsection">
5463 <a name="int_powi">'<tt>llvm.powi.*</tt>' Intrinsic</a>
5466 <div class="doc_text">
5469 <p>This is an overloaded intrinsic. You can use <tt>llvm.powi</tt> on any
5470 floating point or vector of floating point type. Not all targets support all
5473 declare float @llvm.powi.f32(float %Val, i32 %power)
5474 declare double @llvm.powi.f64(double %Val, i32 %power)
5475 declare x86_fp80 @llvm.powi.f80(x86_fp80 %Val, i32 %power)
5476 declare fp128 @llvm.powi.f128(fp128 %Val, i32 %power)
5477 declare ppc_fp128 @llvm.powi.ppcf128(ppc_fp128 %Val, i32 %power)
5483 The '<tt>llvm.powi.*</tt>' intrinsics return the first operand raised to the
5484 specified (positive or negative) power. The order of evaluation of
5485 multiplications is not defined. When a vector of floating point type is
5486 used, the second argument remains a scalar integer value.
5492 The second argument is an integer power, and the first is a value to raise to
5499 This function returns the first value raised to the second power with an
5500 unspecified sequence of rounding operations.</p>
5503 <!-- _______________________________________________________________________ -->
5504 <div class="doc_subsubsection">
5505 <a name="int_sin">'<tt>llvm.sin.*</tt>' Intrinsic</a>
5508 <div class="doc_text">
5511 <p>This is an overloaded intrinsic. You can use <tt>llvm.sin</tt> on any
5512 floating point or vector of floating point type. Not all targets support all
5515 declare float @llvm.sin.f32(float %Val)
5516 declare double @llvm.sin.f64(double %Val)
5517 declare x86_fp80 @llvm.sin.f80(x86_fp80 %Val)
5518 declare fp128 @llvm.sin.f128(fp128 %Val)
5519 declare ppc_fp128 @llvm.sin.ppcf128(ppc_fp128 %Val)
5525 The '<tt>llvm.sin.*</tt>' intrinsics return the sine of the operand.
5531 The argument and return value are floating point numbers of the same type.
5537 This function returns the sine of the specified operand, returning the
5538 same values as the libm <tt>sin</tt> functions would, and handles error
5539 conditions in the same way.</p>
5542 <!-- _______________________________________________________________________ -->
5543 <div class="doc_subsubsection">
5544 <a name="int_cos">'<tt>llvm.cos.*</tt>' Intrinsic</a>
5547 <div class="doc_text">
5550 <p>This is an overloaded intrinsic. You can use <tt>llvm.cos</tt> on any
5551 floating point or vector of floating point type. Not all targets support all
5554 declare float @llvm.cos.f32(float %Val)
5555 declare double @llvm.cos.f64(double %Val)
5556 declare x86_fp80 @llvm.cos.f80(x86_fp80 %Val)
5557 declare fp128 @llvm.cos.f128(fp128 %Val)
5558 declare ppc_fp128 @llvm.cos.ppcf128(ppc_fp128 %Val)
5564 The '<tt>llvm.cos.*</tt>' intrinsics return the cosine of the operand.
5570 The argument and return value are floating point numbers of the same type.
5576 This function returns the cosine of the specified operand, returning the
5577 same values as the libm <tt>cos</tt> functions would, and handles error
5578 conditions in the same way.</p>
5581 <!-- _______________________________________________________________________ -->
5582 <div class="doc_subsubsection">
5583 <a name="int_pow">'<tt>llvm.pow.*</tt>' Intrinsic</a>
5586 <div class="doc_text">
5589 <p>This is an overloaded intrinsic. You can use <tt>llvm.pow</tt> on any
5590 floating point or vector of floating point type. Not all targets support all
5593 declare float @llvm.pow.f32(float %Val, float %Power)
5594 declare double @llvm.pow.f64(double %Val, double %Power)
5595 declare x86_fp80 @llvm.pow.f80(x86_fp80 %Val, x86_fp80 %Power)
5596 declare fp128 @llvm.pow.f128(fp128 %Val, fp128 %Power)
5597 declare ppc_fp128 @llvm.pow.ppcf128(ppc_fp128 %Val, ppc_fp128 Power)
5603 The '<tt>llvm.pow.*</tt>' intrinsics return the first operand raised to the
5604 specified (positive or negative) power.
5610 The second argument is a floating point power, and the first is a value to
5611 raise to that power.
5617 This function returns the first value raised to the second power,
5619 same values as the libm <tt>pow</tt> functions would, and handles error
5620 conditions in the same way.</p>
5624 <!-- ======================================================================= -->
5625 <div class="doc_subsection">
5626 <a name="int_manip">Bit Manipulation Intrinsics</a>
5629 <div class="doc_text">
5631 LLVM provides intrinsics for a few important bit manipulation operations.
5632 These allow efficient code generation for some algorithms.
5637 <!-- _______________________________________________________________________ -->
5638 <div class="doc_subsubsection">
5639 <a name="int_bswap">'<tt>llvm.bswap.*</tt>' Intrinsics</a>
5642 <div class="doc_text">
5645 <p>This is an overloaded intrinsic function. You can use bswap on any integer
5646 type that is an even number of bytes (i.e. BitWidth % 16 == 0).</p>
5648 declare i16 @llvm.bswap.i16(i16 <id>)
5649 declare i32 @llvm.bswap.i32(i32 <id>)
5650 declare i64 @llvm.bswap.i64(i64 <id>)
5656 The '<tt>llvm.bswap</tt>' family of intrinsics is used to byte swap integer
5657 values with an even number of bytes (positive multiple of 16 bits). These are
5658 useful for performing operations on data that is not in the target's native
5665 The <tt>llvm.bswap.i16</tt> intrinsic returns an i16 value that has the high
5666 and low byte of the input i16 swapped. Similarly, the <tt>llvm.bswap.i32</tt>
5667 intrinsic returns an i32 value that has the four bytes of the input i32
5668 swapped, so that if the input bytes are numbered 0, 1, 2, 3 then the returned
5669 i32 will have its bytes in 3, 2, 1, 0 order. The <tt>llvm.bswap.i48</tt>,
5670 <tt>llvm.bswap.i64</tt> and other intrinsics extend this concept to
5671 additional even-byte lengths (6 bytes, 8 bytes and more, respectively).
5676 <!-- _______________________________________________________________________ -->
5677 <div class="doc_subsubsection">
5678 <a name="int_ctpop">'<tt>llvm.ctpop.*</tt>' Intrinsic</a>
5681 <div class="doc_text">
5684 <p>This is an overloaded intrinsic. You can use llvm.ctpop on any integer bit
5685 width. Not all targets support all bit widths however.</p>
5687 declare i8 @llvm.ctpop.i8 (i8 <src>)
5688 declare i16 @llvm.ctpop.i16(i16 <src>)
5689 declare i32 @llvm.ctpop.i32(i32 <src>)
5690 declare i64 @llvm.ctpop.i64(i64 <src>)
5691 declare i256 @llvm.ctpop.i256(i256 <src>)
5697 The '<tt>llvm.ctpop</tt>' family of intrinsics counts the number of bits set in a
5704 The only argument is the value to be counted. The argument may be of any
5705 integer type. The return type must match the argument type.
5711 The '<tt>llvm.ctpop</tt>' intrinsic counts the 1's in a variable.
5715 <!-- _______________________________________________________________________ -->
5716 <div class="doc_subsubsection">
5717 <a name="int_ctlz">'<tt>llvm.ctlz.*</tt>' Intrinsic</a>
5720 <div class="doc_text">
5723 <p>This is an overloaded intrinsic. You can use <tt>llvm.ctlz</tt> on any
5724 integer bit width. Not all targets support all bit widths however.</p>
5726 declare i8 @llvm.ctlz.i8 (i8 <src>)
5727 declare i16 @llvm.ctlz.i16(i16 <src>)
5728 declare i32 @llvm.ctlz.i32(i32 <src>)
5729 declare i64 @llvm.ctlz.i64(i64 <src>)
5730 declare i256 @llvm.ctlz.i256(i256 <src>)
5736 The '<tt>llvm.ctlz</tt>' family of intrinsic functions counts the number of
5737 leading zeros in a variable.
5743 The only argument is the value to be counted. The argument may be of any
5744 integer type. The return type must match the argument type.
5750 The '<tt>llvm.ctlz</tt>' intrinsic counts the leading (most significant) zeros
5751 in a variable. If the src == 0 then the result is the size in bits of the type
5752 of src. For example, <tt>llvm.ctlz(i32 2) = 30</tt>.
5758 <!-- _______________________________________________________________________ -->
5759 <div class="doc_subsubsection">
5760 <a name="int_cttz">'<tt>llvm.cttz.*</tt>' Intrinsic</a>
5763 <div class="doc_text">
5766 <p>This is an overloaded intrinsic. You can use <tt>llvm.cttz</tt> on any
5767 integer bit width. Not all targets support all bit widths however.</p>
5769 declare i8 @llvm.cttz.i8 (i8 <src>)
5770 declare i16 @llvm.cttz.i16(i16 <src>)
5771 declare i32 @llvm.cttz.i32(i32 <src>)
5772 declare i64 @llvm.cttz.i64(i64 <src>)
5773 declare i256 @llvm.cttz.i256(i256 <src>)
5779 The '<tt>llvm.cttz</tt>' family of intrinsic functions counts the number of
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.cttz</tt>' intrinsic counts the trailing (least 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.cttz(2) = 1</tt>.
5799 <!-- _______________________________________________________________________ -->
5800 <div class="doc_subsubsection">
5801 <a name="int_part_select">'<tt>llvm.part.select.*</tt>' Intrinsic</a>
5804 <div class="doc_text">
5807 <p>This is an overloaded intrinsic. You can use <tt>llvm.part.select</tt>
5808 on any integer bit width.</p>
5810 declare i17 @llvm.part.select.i17 (i17 %val, i32 %loBit, i32 %hiBit)
5811 declare i29 @llvm.part.select.i29 (i29 %val, i32 %loBit, i32 %hiBit)
5815 <p>The '<tt>llvm.part.select</tt>' family of intrinsic functions selects a
5816 range of bits from an integer value and returns them in the same bit width as
5817 the original value.</p>
5820 <p>The first argument, <tt>%val</tt> and the result may be integer types of
5821 any bit width but they must have the same bit width. The second and third
5822 arguments must be <tt>i32</tt> type since they specify only a bit index.</p>
5825 <p>The operation of the '<tt>llvm.part.select</tt>' intrinsic has two modes
5826 of operation: forwards and reverse. If <tt>%loBit</tt> is greater than
5827 <tt>%hiBits</tt> then the intrinsic operates in reverse mode. Otherwise it
5828 operates in forward mode.</p>
5829 <p>In forward mode, this intrinsic is the equivalent of shifting <tt>%val</tt>
5830 right by <tt>%loBit</tt> bits and then ANDing it with a mask with
5831 only the <tt>%hiBit - %loBit</tt> bits set, as follows:</p>
5833 <li>The <tt>%val</tt> is shifted right (LSHR) by the number of bits specified
5834 by <tt>%loBits</tt>. This normalizes the value to the low order bits.</li>
5835 <li>The <tt>%loBits</tt> value is subtracted from the <tt>%hiBits</tt> value
5836 to determine the number of bits to retain.</li>
5837 <li>A mask of the retained bits is created by shifting a -1 value.</li>
5838 <li>The mask is ANDed with <tt>%val</tt> to produce the result.</li>
5840 <p>In reverse mode, a similar computation is made except that the bits are
5841 returned in the reverse order. So, for example, if <tt>X</tt> has the value
5842 <tt>i16 0x0ACF (101011001111)</tt> and we apply
5843 <tt>part.select(i16 X, 8, 3)</tt> to it, we get back the value
5844 <tt>i16 0x0026 (000000100110)</tt>.</p>
5847 <div class="doc_subsubsection">
5848 <a name="int_part_set">'<tt>llvm.part.set.*</tt>' Intrinsic</a>
5851 <div class="doc_text">
5854 <p>This is an overloaded intrinsic. You can use <tt>llvm.part.set</tt>
5855 on any integer bit width.</p>
5857 declare i17 @llvm.part.set.i17.i9 (i17 %val, i9 %repl, i32 %lo, i32 %hi)
5858 declare i29 @llvm.part.set.i29.i9 (i29 %val, i9 %repl, i32 %lo, i32 %hi)
5862 <p>The '<tt>llvm.part.set</tt>' family of intrinsic functions replaces a range
5863 of bits in an integer value with another integer value. It returns the integer
5864 with the replaced bits.</p>
5867 <p>The first argument, <tt>%val</tt> and the result may be integer types of
5868 any bit width but they must have the same bit width. <tt>%val</tt> is the value
5869 whose bits will be replaced. The second argument, <tt>%repl</tt> may be an
5870 integer of any bit width. The third and fourth arguments must be <tt>i32</tt>
5871 type since they specify only a bit index.</p>
5874 <p>The operation of the '<tt>llvm.part.set</tt>' intrinsic has two modes
5875 of operation: forwards and reverse. If <tt>%lo</tt> is greater than
5876 <tt>%hi</tt> then the intrinsic operates in reverse mode. Otherwise it
5877 operates in forward mode.</p>
5878 <p>For both modes, the <tt>%repl</tt> value is prepared for use by either
5879 truncating it down to the size of the replacement area or zero extending it
5880 up to that size.</p>
5881 <p>In forward mode, the bits between <tt>%lo</tt> and <tt>%hi</tt> (inclusive)
5882 are replaced with corresponding bits from <tt>%repl</tt>. That is the 0th bit
5883 in <tt>%repl</tt> replaces the <tt>%lo</tt>th bit in <tt>%val</tt> and etc. up
5884 to the <tt>%hi</tt>th bit.</p>
5885 <p>In reverse mode, a similar computation is made except that the bits are
5886 reversed. That is, the <tt>0</tt>th bit in <tt>%repl</tt> replaces the
5887 <tt>%hi</tt> bit in <tt>%val</tt> and etc. down to the <tt>%lo</tt>th bit.</p>
5890 llvm.part.set(0xFFFF, 0, 4, 7) -> 0xFF0F
5891 llvm.part.set(0xFFFF, 0, 7, 4) -> 0xFF0F
5892 llvm.part.set(0xFFFF, 1, 7, 4) -> 0xFF8F
5893 llvm.part.set(0xFFFF, F, 8, 3) -> 0xFFE7
5894 llvm.part.set(0xFFFF, 0, 3, 8) -> 0xFE07
5898 <!-- ======================================================================= -->
5899 <div class="doc_subsection">
5900 <a name="int_debugger">Debugger Intrinsics</a>
5903 <div class="doc_text">
5905 The LLVM debugger intrinsics (which all start with <tt>llvm.dbg.</tt> prefix),
5906 are described in the <a
5907 href="SourceLevelDebugging.html#format_common_intrinsics">LLVM Source Level
5908 Debugging</a> document.
5913 <!-- ======================================================================= -->
5914 <div class="doc_subsection">
5915 <a name="int_eh">Exception Handling Intrinsics</a>
5918 <div class="doc_text">
5919 <p> The LLVM exception handling intrinsics (which all start with
5920 <tt>llvm.eh.</tt> prefix), are described in the <a
5921 href="ExceptionHandling.html#format_common_intrinsics">LLVM Exception
5922 Handling</a> document. </p>
5925 <!-- ======================================================================= -->
5926 <div class="doc_subsection">
5927 <a name="int_trampoline">Trampoline Intrinsic</a>
5930 <div class="doc_text">
5932 This intrinsic makes it possible to excise one parameter, marked with
5933 the <tt>nest</tt> attribute, from a function. The result is a callable
5934 function pointer lacking the nest parameter - the caller does not need
5935 to provide a value for it. Instead, the value to use is stored in
5936 advance in a "trampoline", a block of memory usually allocated
5937 on the stack, which also contains code to splice the nest value into the
5938 argument list. This is used to implement the GCC nested function address
5942 For example, if the function is
5943 <tt>i32 f(i8* nest %c, i32 %x, i32 %y)</tt> then the resulting function
5944 pointer has signature <tt>i32 (i32, i32)*</tt>. It can be created as follows:</p>
5946 %tramp = alloca [10 x i8], align 4 ; size and alignment only correct for X86
5947 %tramp1 = getelementptr [10 x i8]* %tramp, i32 0, i32 0
5948 %p = call i8* @llvm.init.trampoline( i8* %tramp1, i8* bitcast (i32 (i8* nest , i32, i32)* @f to i8*), i8* %nval )
5949 %fp = bitcast i8* %p to i32 (i32, i32)*
5951 <p>The call <tt>%val = call i32 %fp( i32 %x, i32 %y )</tt> is then equivalent
5952 to <tt>%val = call i32 %f( i8* %nval, i32 %x, i32 %y )</tt>.</p>
5955 <!-- _______________________________________________________________________ -->
5956 <div class="doc_subsubsection">
5957 <a name="int_it">'<tt>llvm.init.trampoline</tt>' Intrinsic</a>
5959 <div class="doc_text">
5962 declare i8* @llvm.init.trampoline(i8* <tramp>, i8* <func>, i8* <nval>)
5966 This fills the memory pointed to by <tt>tramp</tt> with code
5967 and returns a function pointer suitable for executing it.
5971 The <tt>llvm.init.trampoline</tt> intrinsic takes three arguments, all
5972 pointers. The <tt>tramp</tt> argument must point to a sufficiently large
5973 and sufficiently aligned block of memory; this memory is written to by the
5974 intrinsic. Note that the size and the alignment are target-specific - LLVM
5975 currently provides no portable way of determining them, so a front-end that
5976 generates this intrinsic needs to have some target-specific knowledge.
5977 The <tt>func</tt> argument must hold a function bitcast to an <tt>i8*</tt>.
5981 The block of memory pointed to by <tt>tramp</tt> is filled with target
5982 dependent code, turning it into a function. A pointer to this function is
5983 returned, but needs to be bitcast to an
5984 <a href="#int_trampoline">appropriate function pointer type</a>
5985 before being called. The new function's signature is the same as that of
5986 <tt>func</tt> with any arguments marked with the <tt>nest</tt> attribute
5987 removed. At most one such <tt>nest</tt> argument is allowed, and it must be
5988 of pointer type. Calling the new function is equivalent to calling
5989 <tt>func</tt> with the same argument list, but with <tt>nval</tt> used for the
5990 missing <tt>nest</tt> argument. If, after calling
5991 <tt>llvm.init.trampoline</tt>, the memory pointed to by <tt>tramp</tt> is
5992 modified, then the effect of any later call to the returned function pointer is
5997 <!-- ======================================================================= -->
5998 <div class="doc_subsection">
5999 <a name="int_atomics">Atomic Operations and Synchronization Intrinsics</a>
6002 <div class="doc_text">
6004 These intrinsic functions expand the "universal IR" of LLVM to represent
6005 hardware constructs for atomic operations and memory synchronization. This
6006 provides an interface to the hardware, not an interface to the programmer. It
6007 is aimed at a low enough level to allow any programming models or APIs
6008 (Application Programming Interfaces) which
6009 need atomic behaviors to map cleanly onto it. It is also modeled primarily on
6010 hardware behavior. Just as hardware provides a "universal IR" for source
6011 languages, it also provides a starting point for developing a "universal"
6012 atomic operation and synchronization IR.
6015 These do <em>not</em> form an API such as high-level threading libraries,
6016 software transaction memory systems, atomic primitives, and intrinsic
6017 functions as found in BSD, GNU libc, atomic_ops, APR, and other system and
6018 application libraries. The hardware interface provided by LLVM should allow
6019 a clean implementation of all of these APIs and parallel programming models.
6020 No one model or paradigm should be selected above others unless the hardware
6021 itself ubiquitously does so.
6026 <!-- _______________________________________________________________________ -->
6027 <div class="doc_subsubsection">
6028 <a name="int_memory_barrier">'<tt>llvm.memory.barrier</tt>' Intrinsic</a>
6030 <div class="doc_text">
6033 declare void @llvm.memory.barrier( i1 <ll>, i1 <ls>, i1 <sl>, i1 <ss>,
6039 The <tt>llvm.memory.barrier</tt> intrinsic guarantees ordering between
6040 specific pairs of memory access types.
6044 The <tt>llvm.memory.barrier</tt> intrinsic requires five boolean arguments.
6045 The first four arguments enables a specific barrier as listed below. The fith
6046 argument specifies that the barrier applies to io or device or uncached memory.
6050 <li><tt>ll</tt>: load-load barrier</li>
6051 <li><tt>ls</tt>: load-store barrier</li>
6052 <li><tt>sl</tt>: store-load barrier</li>
6053 <li><tt>ss</tt>: store-store barrier</li>
6054 <li><tt>device</tt>: barrier applies to device and uncached memory also.</li>
6058 This intrinsic causes the system to enforce some ordering constraints upon
6059 the loads and stores of the program. This barrier does not indicate
6060 <em>when</em> any events will occur, it only enforces an <em>order</em> in
6061 which they occur. For any of the specified pairs of load and store operations
6062 (f.ex. load-load, or store-load), all of the first operations preceding the
6063 barrier will complete before any of the second operations succeeding the
6064 barrier begin. Specifically the semantics for each pairing is as follows:
6067 <li><tt>ll</tt>: All loads before the barrier must complete before any load
6068 after the barrier begins.</li>
6070 <li><tt>ls</tt>: All loads before the barrier must complete before any
6071 store after the barrier begins.</li>
6072 <li><tt>ss</tt>: All stores before the barrier must complete before any
6073 store after the barrier begins.</li>
6074 <li><tt>sl</tt>: All stores before the barrier must complete before any
6075 load after the barrier begins.</li>
6078 These semantics are applied with a logical "and" behavior when more than one
6079 is enabled in a single memory barrier intrinsic.
6082 Backends may implement stronger barriers than those requested when they do not
6083 support as fine grained a barrier as requested. Some architectures do not
6084 need all types of barriers and on such architectures, these become noops.
6091 %result1 = load i32* %ptr <i>; yields {i32}:result1 = 4</i>
6092 call void @llvm.memory.barrier( i1 false, i1 true, i1 false, i1 false )
6093 <i>; guarantee the above finishes</i>
6094 store i32 8, %ptr <i>; before this begins</i>
6098 <!-- _______________________________________________________________________ -->
6099 <div class="doc_subsubsection">
6100 <a name="int_atomic_cmp_swap">'<tt>llvm.atomic.cmp.swap.*</tt>' Intrinsic</a>
6102 <div class="doc_text">
6105 This is an overloaded intrinsic. You can use <tt>llvm.atomic.cmp.swap</tt> on
6106 any integer bit width and for different address spaces. Not all targets
6107 support all bit widths however.</p>
6110 declare i8 @llvm.atomic.cmp.swap.i8.p0i8( i8* <ptr>, i8 <cmp>, i8 <val> )
6111 declare i16 @llvm.atomic.cmp.swap.i16.p0i16( i16* <ptr>, i16 <cmp>, i16 <val> )
6112 declare i32 @llvm.atomic.cmp.swap.i32.p0i32( i32* <ptr>, i32 <cmp>, i32 <val> )
6113 declare i64 @llvm.atomic.cmp.swap.i64.p0i64( i64* <ptr>, i64 <cmp>, i64 <val> )
6118 This loads a value in memory and compares it to a given value. If they are
6119 equal, it stores a new value into the memory.
6123 The <tt>llvm.atomic.cmp.swap</tt> intrinsic takes three arguments. The result as
6124 well as both <tt>cmp</tt> and <tt>val</tt> must be integer values with the
6125 same bit width. The <tt>ptr</tt> argument must be a pointer to a value of
6126 this integer type. While any bit width integer may be used, targets may only
6127 lower representations they support in hardware.
6132 This entire intrinsic must be executed atomically. It first loads the value
6133 in memory pointed to by <tt>ptr</tt> and compares it with the value
6134 <tt>cmp</tt>. If they are equal, <tt>val</tt> is stored into the memory. The
6135 loaded value is yielded in all cases. This provides the equivalent of an
6136 atomic compare-and-swap operation within the SSA framework.
6144 %val1 = add i32 4, 4
6145 %result1 = call i32 @llvm.atomic.cmp.swap.i32.p0i32( i32* %ptr, i32 4, %val1 )
6146 <i>; yields {i32}:result1 = 4</i>
6147 %stored1 = icmp eq i32 %result1, 4 <i>; yields {i1}:stored1 = true</i>
6148 %memval1 = load i32* %ptr <i>; yields {i32}:memval1 = 8</i>
6150 %val2 = add i32 1, 1
6151 %result2 = call i32 @llvm.atomic.cmp.swap.i32.p0i32( i32* %ptr, i32 5, %val2 )
6152 <i>; yields {i32}:result2 = 8</i>
6153 %stored2 = icmp eq i32 %result2, 5 <i>; yields {i1}:stored2 = false</i>
6155 %memval2 = load i32* %ptr <i>; yields {i32}:memval2 = 8</i>
6159 <!-- _______________________________________________________________________ -->
6160 <div class="doc_subsubsection">
6161 <a name="int_atomic_swap">'<tt>llvm.atomic.swap.*</tt>' Intrinsic</a>
6163 <div class="doc_text">
6167 This is an overloaded intrinsic. You can use <tt>llvm.atomic.swap</tt> on any
6168 integer bit width. Not all targets support all bit widths however.</p>
6170 declare i8 @llvm.atomic.swap.i8.p0i8( i8* <ptr>, i8 <val> )
6171 declare i16 @llvm.atomic.swap.i16.p0i16( i16* <ptr>, i16 <val> )
6172 declare i32 @llvm.atomic.swap.i32.p0i32( i32* <ptr>, i32 <val> )
6173 declare i64 @llvm.atomic.swap.i64.p0i64( i64* <ptr>, i64 <val> )
6178 This intrinsic loads the value stored in memory at <tt>ptr</tt> and yields
6179 the value from memory. It then stores the value in <tt>val</tt> in the memory
6185 The <tt>llvm.atomic.swap</tt> intrinsic takes two arguments. Both the
6186 <tt>val</tt> argument and the result must be integers of the same bit width.
6187 The first argument, <tt>ptr</tt>, must be a pointer to a value of this
6188 integer type. The targets may only lower integer representations they
6193 This intrinsic loads the value pointed to by <tt>ptr</tt>, yields it, and
6194 stores <tt>val</tt> back into <tt>ptr</tt> atomically. This provides the
6195 equivalent of an atomic swap operation within the SSA framework.
6203 %val1 = add i32 4, 4
6204 %result1 = call i32 @llvm.atomic.swap.i32.p0i32( i32* %ptr, i32 %val1 )
6205 <i>; yields {i32}:result1 = 4</i>
6206 %stored1 = icmp eq i32 %result1, 4 <i>; yields {i1}:stored1 = true</i>
6207 %memval1 = load i32* %ptr <i>; yields {i32}:memval1 = 8</i>
6209 %val2 = add i32 1, 1
6210 %result2 = call i32 @llvm.atomic.swap.i32.p0i32( i32* %ptr, i32 %val2 )
6211 <i>; yields {i32}:result2 = 8</i>
6213 %stored2 = icmp eq i32 %result2, 8 <i>; yields {i1}:stored2 = true</i>
6214 %memval2 = load i32* %ptr <i>; yields {i32}:memval2 = 2</i>
6218 <!-- _______________________________________________________________________ -->
6219 <div class="doc_subsubsection">
6220 <a name="int_atomic_load_add">'<tt>llvm.atomic.load.add.*</tt>' Intrinsic</a>
6223 <div class="doc_text">
6226 This is an overloaded intrinsic. You can use <tt>llvm.atomic.load.add</tt> on any
6227 integer bit width. Not all targets support all bit widths however.</p>
6229 declare i8 @llvm.atomic.load.add.i8..p0i8( i8* <ptr>, i8 <delta> )
6230 declare i16 @llvm.atomic.load.add.i16..p0i16( i16* <ptr>, i16 <delta> )
6231 declare i32 @llvm.atomic.load.add.i32..p0i32( i32* <ptr>, i32 <delta> )
6232 declare i64 @llvm.atomic.load.add.i64..p0i64( i64* <ptr>, i64 <delta> )
6237 This intrinsic adds <tt>delta</tt> to the value stored in memory at
6238 <tt>ptr</tt>. It yields the original value at <tt>ptr</tt>.
6243 The intrinsic takes two arguments, the first a pointer to an integer value
6244 and the second an integer value. The result is also an integer value. These
6245 integer types can have any bit width, but they must all have the same bit
6246 width. The targets may only lower integer representations they support.
6250 This intrinsic does a series of operations atomically. It first loads the
6251 value stored at <tt>ptr</tt>. It then adds <tt>delta</tt>, stores the result
6252 to <tt>ptr</tt>. It yields the original value stored at <tt>ptr</tt>.
6259 %result1 = call i32 @llvm.atomic.load.add.i32.p0i32( i32* %ptr, i32 4 )
6260 <i>; yields {i32}:result1 = 4</i>
6261 %result2 = call i32 @llvm.atomic.load.add.i32.p0i32( i32* %ptr, i32 2 )
6262 <i>; yields {i32}:result2 = 8</i>
6263 %result3 = call i32 @llvm.atomic.load.add.i32.p0i32( i32* %ptr, i32 5 )
6264 <i>; yields {i32}:result3 = 10</i>
6265 %memval1 = load i32* %ptr <i>; yields {i32}:memval1 = 15</i>
6269 <!-- _______________________________________________________________________ -->
6270 <div class="doc_subsubsection">
6271 <a name="int_atomic_load_sub">'<tt>llvm.atomic.load.sub.*</tt>' Intrinsic</a>
6274 <div class="doc_text">
6277 This is an overloaded intrinsic. You can use <tt>llvm.atomic.load.sub</tt> on
6278 any integer bit width and for different address spaces. Not all targets
6279 support all bit widths however.</p>
6281 declare i8 @llvm.atomic.load.sub.i8.p0i32( i8* <ptr>, i8 <delta> )
6282 declare i16 @llvm.atomic.load.sub.i16.p0i32( i16* <ptr>, i16 <delta> )
6283 declare i32 @llvm.atomic.load.sub.i32.p0i32( i32* <ptr>, i32 <delta> )
6284 declare i64 @llvm.atomic.load.sub.i64.p0i32( i64* <ptr>, i64 <delta> )
6289 This intrinsic subtracts <tt>delta</tt> to the value stored in memory at
6290 <tt>ptr</tt>. It yields the original value at <tt>ptr</tt>.
6295 The intrinsic takes two arguments, the first a pointer to an integer value
6296 and the second an integer value. The result is also an integer value. These
6297 integer types can have any bit width, but they must all have the same bit
6298 width. The targets may only lower integer representations they support.
6302 This intrinsic does a series of operations atomically. It first loads the
6303 value stored at <tt>ptr</tt>. It then subtracts <tt>delta</tt>, stores the
6304 result to <tt>ptr</tt>. It yields the original value stored at <tt>ptr</tt>.
6311 %result1 = call i32 @llvm.atomic.load.sub.i32.p0i32( i32* %ptr, i32 4 )
6312 <i>; yields {i32}:result1 = 8</i>
6313 %result2 = call i32 @llvm.atomic.load.sub.i32.p0i32( i32* %ptr, i32 2 )
6314 <i>; yields {i32}:result2 = 4</i>
6315 %result3 = call i32 @llvm.atomic.load.sub.i32.p0i32( i32* %ptr, i32 5 )
6316 <i>; yields {i32}:result3 = 2</i>
6317 %memval1 = load i32* %ptr <i>; yields {i32}:memval1 = -3</i>
6321 <!-- _______________________________________________________________________ -->
6322 <div class="doc_subsubsection">
6323 <a name="int_atomic_load_and">'<tt>llvm.atomic.load.and.*</tt>' Intrinsic</a><br>
6324 <a name="int_atomic_load_nand">'<tt>llvm.atomic.load.nand.*</tt>' Intrinsic</a><br>
6325 <a name="int_atomic_load_or">'<tt>llvm.atomic.load.or.*</tt>' Intrinsic</a><br>
6326 <a name="int_atomic_load_xor">'<tt>llvm.atomic.load.xor.*</tt>' Intrinsic</a><br>
6329 <div class="doc_text">
6332 These are overloaded intrinsics. You can use <tt>llvm.atomic.load_and</tt>,
6333 <tt>llvm.atomic.load_nand</tt>, <tt>llvm.atomic.load_or</tt>, and
6334 <tt>llvm.atomic.load_xor</tt> on any integer bit width and for different
6335 address spaces. Not all targets support all bit widths however.</p>
6337 declare i8 @llvm.atomic.load.and.i8.p0i8( i8* <ptr>, i8 <delta> )
6338 declare i16 @llvm.atomic.load.and.i16.p0i16( i16* <ptr>, i16 <delta> )
6339 declare i32 @llvm.atomic.load.and.i32.p0i32( i32* <ptr>, i32 <delta> )
6340 declare i64 @llvm.atomic.load.and.i64.p0i64( i64* <ptr>, i64 <delta> )
6345 declare i8 @llvm.atomic.load.or.i8.p0i8( i8* <ptr>, i8 <delta> )
6346 declare i16 @llvm.atomic.load.or.i16.p0i16( i16* <ptr>, i16 <delta> )
6347 declare i32 @llvm.atomic.load.or.i32.p0i32( i32* <ptr>, i32 <delta> )
6348 declare i64 @llvm.atomic.load.or.i64.p0i64( i64* <ptr>, i64 <delta> )
6353 declare i8 @llvm.atomic.load.nand.i8.p0i32( i8* <ptr>, i8 <delta> )
6354 declare i16 @llvm.atomic.load.nand.i16.p0i32( i16* <ptr>, i16 <delta> )
6355 declare i32 @llvm.atomic.load.nand.i32.p0i32( i32* <ptr>, i32 <delta> )
6356 declare i64 @llvm.atomic.load.nand.i64.p0i32( i64* <ptr>, i64 <delta> )
6361 declare i8 @llvm.atomic.load.xor.i8.p0i32( i8* <ptr>, i8 <delta> )
6362 declare i16 @llvm.atomic.load.xor.i16.p0i32( i16* <ptr>, i16 <delta> )
6363 declare i32 @llvm.atomic.load.xor.i32.p0i32( i32* <ptr>, i32 <delta> )
6364 declare i64 @llvm.atomic.load.xor.i64.p0i32( i64* <ptr>, i64 <delta> )
6369 These intrinsics bitwise the operation (and, nand, or, xor) <tt>delta</tt> to
6370 the value stored in memory at <tt>ptr</tt>. It yields the original value
6376 These intrinsics take two arguments, the first a pointer to an integer value
6377 and the second an integer value. The result is also an integer value. These
6378 integer types can have any bit width, but they must all have the same bit
6379 width. The targets may only lower integer representations they support.
6383 These intrinsics does a series of operations atomically. They first load the
6384 value stored at <tt>ptr</tt>. They then do the bitwise operation
6385 <tt>delta</tt>, store the result to <tt>ptr</tt>. They yield the original
6386 value stored at <tt>ptr</tt>.
6392 store i32 0x0F0F, %ptr
6393 %result0 = call i32 @llvm.atomic.load.nand.i32.p0i32( i32* %ptr, i32 0xFF )
6394 <i>; yields {i32}:result0 = 0x0F0F</i>
6395 %result1 = call i32 @llvm.atomic.load.and.i32.p0i32( i32* %ptr, i32 0xFF )
6396 <i>; yields {i32}:result1 = 0xFFFFFFF0</i>
6397 %result2 = call i32 @llvm.atomic.load.or.i32.p0i32( i32* %ptr, i32 0F )
6398 <i>; yields {i32}:result2 = 0xF0</i>
6399 %result3 = call i32 @llvm.atomic.load.xor.i32.p0i32( i32* %ptr, i32 0F )
6400 <i>; yields {i32}:result3 = FF</i>
6401 %memval1 = load i32* %ptr <i>; yields {i32}:memval1 = F0</i>
6406 <!-- _______________________________________________________________________ -->
6407 <div class="doc_subsubsection">
6408 <a name="int_atomic_load_max">'<tt>llvm.atomic.load.max.*</tt>' Intrinsic</a><br>
6409 <a name="int_atomic_load_min">'<tt>llvm.atomic.load.min.*</tt>' Intrinsic</a><br>
6410 <a name="int_atomic_load_umax">'<tt>llvm.atomic.load.umax.*</tt>' Intrinsic</a><br>
6411 <a name="int_atomic_load_umin">'<tt>llvm.atomic.load.umin.*</tt>' Intrinsic</a><br>
6414 <div class="doc_text">
6417 These are overloaded intrinsics. You can use <tt>llvm.atomic.load_max</tt>,
6418 <tt>llvm.atomic.load_min</tt>, <tt>llvm.atomic.load_umax</tt>, and
6419 <tt>llvm.atomic.load_umin</tt> on any integer bit width and for different
6420 address spaces. Not all targets
6421 support all bit widths however.</p>
6423 declare i8 @llvm.atomic.load.max.i8.p0i8( i8* <ptr>, i8 <delta> )
6424 declare i16 @llvm.atomic.load.max.i16.p0i16( i16* <ptr>, i16 <delta> )
6425 declare i32 @llvm.atomic.load.max.i32.p0i32( i32* <ptr>, i32 <delta> )
6426 declare i64 @llvm.atomic.load.max.i64.p0i64( i64* <ptr>, i64 <delta> )
6431 declare i8 @llvm.atomic.load.min.i8.p0i8( i8* <ptr>, i8 <delta> )
6432 declare i16 @llvm.atomic.load.min.i16.p0i16( i16* <ptr>, i16 <delta> )
6433 declare i32 @llvm.atomic.load.min.i32..p0i32( i32* <ptr>, i32 <delta> )
6434 declare i64 @llvm.atomic.load.min.i64..p0i64( i64* <ptr>, i64 <delta> )
6439 declare i8 @llvm.atomic.load.umax.i8.p0i8( i8* <ptr>, i8 <delta> )
6440 declare i16 @llvm.atomic.load.umax.i16.p0i16( i16* <ptr>, i16 <delta> )
6441 declare i32 @llvm.atomic.load.umax.i32.p0i32( i32* <ptr>, i32 <delta> )
6442 declare i64 @llvm.atomic.load.umax.i64.p0i64( i64* <ptr>, i64 <delta> )
6447 declare i8 @llvm.atomic.load.umin.i8..p0i8( i8* <ptr>, i8 <delta> )
6448 declare i16 @llvm.atomic.load.umin.i16.p0i16( i16* <ptr>, i16 <delta> )
6449 declare i32 @llvm.atomic.load.umin.i32..p0i32( i32* <ptr>, i32 <delta> )
6450 declare i64 @llvm.atomic.load.umin.i64..p0i64( i64* <ptr>, i64 <delta> )
6455 These intrinsics takes the signed or unsigned minimum or maximum of
6456 <tt>delta</tt> and the value stored in memory at <tt>ptr</tt>. It yields the
6457 original value at <tt>ptr</tt>.
6462 These intrinsics take two arguments, the first a pointer to an integer value
6463 and the second an integer value. The result is also an integer value. These
6464 integer types can have any bit width, but they must all have the same bit
6465 width. The targets may only lower integer representations they support.
6469 These intrinsics does a series of operations atomically. They first load the
6470 value stored at <tt>ptr</tt>. They then do the signed or unsigned min or max
6471 <tt>delta</tt> and the value, store the result to <tt>ptr</tt>. They yield
6472 the original value stored at <tt>ptr</tt>.
6479 %result0 = call i32 @llvm.atomic.load.min.i32.p0i32( i32* %ptr, i32 -2 )
6480 <i>; yields {i32}:result0 = 7</i>
6481 %result1 = call i32 @llvm.atomic.load.max.i32.p0i32( i32* %ptr, i32 8 )
6482 <i>; yields {i32}:result1 = -2</i>
6483 %result2 = call i32 @llvm.atomic.load.umin.i32.p0i32( i32* %ptr, i32 10 )
6484 <i>; yields {i32}:result2 = 8</i>
6485 %result3 = call i32 @llvm.atomic.load.umax.i32.p0i32( i32* %ptr, i32 30 )
6486 <i>; yields {i32}:result3 = 8</i>
6487 %memval1 = load i32* %ptr <i>; yields {i32}:memval1 = 30</i>
6491 <!-- ======================================================================= -->
6492 <div class="doc_subsection">
6493 <a name="int_general">General Intrinsics</a>
6496 <div class="doc_text">
6497 <p> This class of intrinsics is designed to be generic and has
6498 no specific purpose. </p>
6501 <!-- _______________________________________________________________________ -->
6502 <div class="doc_subsubsection">
6503 <a name="int_var_annotation">'<tt>llvm.var.annotation</tt>' Intrinsic</a>
6506 <div class="doc_text">
6510 declare void @llvm.var.annotation(i8* <val>, i8* <str>, i8* <str>, i32 <int> )
6516 The '<tt>llvm.var.annotation</tt>' intrinsic
6522 The first argument is a pointer to a value, the second is a pointer to a
6523 global string, the third is a pointer to a global string which is the source
6524 file name, and the last argument is the line number.
6530 This intrinsic allows annotation of local variables with arbitrary strings.
6531 This can be useful for special purpose optimizations that want to look for these
6532 annotations. These have no other defined use, they are ignored by code
6533 generation and optimization.
6537 <!-- _______________________________________________________________________ -->
6538 <div class="doc_subsubsection">
6539 <a name="int_annotation">'<tt>llvm.annotation.*</tt>' Intrinsic</a>
6542 <div class="doc_text">
6545 <p>This is an overloaded intrinsic. You can use '<tt>llvm.annotation</tt>' on
6546 any integer bit width.
6549 declare i8 @llvm.annotation.i8(i8 <val>, i8* <str>, i8* <str>, i32 <int> )
6550 declare i16 @llvm.annotation.i16(i16 <val>, i8* <str>, i8* <str>, i32 <int> )
6551 declare i32 @llvm.annotation.i32(i32 <val>, i8* <str>, i8* <str>, i32 <int> )
6552 declare i64 @llvm.annotation.i64(i64 <val>, i8* <str>, i8* <str>, i32 <int> )
6553 declare i256 @llvm.annotation.i256(i256 <val>, i8* <str>, i8* <str>, i32 <int> )
6559 The '<tt>llvm.annotation</tt>' intrinsic.
6565 The first argument is an integer value (result of some expression),
6566 the second is a pointer to a global string, the third is a pointer to a global
6567 string which is the source file name, and the last argument is the line number.
6568 It returns the value of the first argument.
6574 This intrinsic allows annotations to be put on arbitrary expressions
6575 with arbitrary strings. This can be useful for special purpose optimizations
6576 that want to look for these annotations. These have no other defined use, they
6577 are ignored by code generation and optimization.
6581 <!-- _______________________________________________________________________ -->
6582 <div class="doc_subsubsection">
6583 <a name="int_trap">'<tt>llvm.trap</tt>' Intrinsic</a>
6586 <div class="doc_text">
6590 declare void @llvm.trap()
6596 The '<tt>llvm.trap</tt>' intrinsic
6608 This intrinsics is lowered to the target dependent trap instruction. If the
6609 target does not have a trap instruction, this intrinsic will be lowered to the
6610 call of the abort() function.
6614 <!-- _______________________________________________________________________ -->
6615 <div class="doc_subsubsection">
6616 <a name="int_stackprotector">'<tt>llvm.stackprotector</tt>' Intrinsic</a>
6618 <div class="doc_text">
6621 declare void @llvm.stackprotector( i8* <guard>, i8** <slot> )
6626 The <tt>llvm.stackprotector</tt> intrinsic takes the <tt>guard</tt> and stores
6627 it onto the stack at <tt>slot</tt>. The stack slot is adjusted to ensure that
6628 it is placed on the stack before local variables.
6632 The <tt>llvm.stackprotector</tt> intrinsic requires two pointer arguments. The
6633 first argument is the value loaded from the stack guard
6634 <tt>@__stack_chk_guard</tt>. The second variable is an <tt>alloca</tt> that
6635 has enough space to hold the value of the guard.
6639 This intrinsic causes the prologue/epilogue inserter to force the position of
6640 the <tt>AllocaInst</tt> stack slot to be before local variables on the
6641 stack. This is to ensure that if a local variable on the stack is overwritten,
6642 it will destroy the value of the guard. When the function exits, the guard on
6643 the stack is checked against the original guard. If they're different, then
6644 the program aborts by calling the <tt>__stack_chk_fail()</tt> function.
6648 <!-- *********************************************************************** -->
6651 <a href="http://jigsaw.w3.org/css-validator/check/referer"><img
6652 src="http://jigsaw.w3.org/css-validator/images/vcss-blue" alt="Valid CSS"></a>
6653 <a href="http://validator.w3.org/check/referer"><img
6654 src="http://www.w3.org/Icons/valid-html401-blue" alt="Valid HTML 4.01"></a>
6656 <a href="mailto:sabre@nondot.org">Chris Lattner</a><br>
6657 <a href="http://llvm.org">The LLVM Compiler Infrastructure</a><br>
6658 Last modified: $Date$