1 <!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01//EN"
2 "http://www.w3.org/TR/html4/strict.dtd">
5 <title>LLVM Assembly Language Reference Manual</title>
6 <meta http-equiv="Content-Type" content="text/html; charset=utf-8">
7 <meta name="author" content="Chris Lattner">
8 <meta name="description"
9 content="LLVM Assembly Language Reference Manual.">
10 <link rel="stylesheet" href="llvm.css" type="text/css">
15 <div class="doc_title"> LLVM Language Reference Manual </div>
17 <li><a href="#abstract">Abstract</a></li>
18 <li><a href="#introduction">Introduction</a></li>
19 <li><a href="#identifiers">Identifiers</a></li>
20 <li><a href="#highlevel">High Level Structure</a>
22 <li><a href="#modulestructure">Module Structure</a></li>
23 <li><a href="#linkage">Linkage Types</a></li>
24 <li><a href="#callingconv">Calling Conventions</a></li>
25 <li><a href="#globalvars">Global Variables</a></li>
26 <li><a href="#functionstructure">Functions</a></li>
27 <li><a href="#aliasstructure">Aliases</a></li>
28 <li><a href="#paramattrs">Parameter Attributes</a></li>
29 <li><a href="#fnattrs">Function Attributes</a></li>
30 <li><a href="#gc">Garbage Collector Names</a></li>
31 <li><a href="#moduleasm">Module-Level Inline Assembly</a></li>
32 <li><a href="#datalayout">Data Layout</a></li>
35 <li><a href="#typesystem">Type System</a>
37 <li><a href="#t_classifications">Type Classifications</a></li>
38 <li><a href="#t_primitive">Primitive Types</a>
40 <li><a href="#t_floating">Floating Point Types</a></li>
41 <li><a href="#t_void">Void Type</a></li>
42 <li><a href="#t_label">Label Type</a></li>
45 <li><a href="#t_derived">Derived Types</a>
47 <li><a href="#t_integer">Integer Type</a></li>
48 <li><a href="#t_array">Array Type</a></li>
49 <li><a href="#t_function">Function Type</a></li>
50 <li><a href="#t_pointer">Pointer Type</a></li>
51 <li><a href="#t_struct">Structure Type</a></li>
52 <li><a href="#t_pstruct">Packed Structure Type</a></li>
53 <li><a href="#t_vector">Vector Type</a></li>
54 <li><a href="#t_opaque">Opaque Type</a></li>
59 <li><a href="#constants">Constants</a>
61 <li><a href="#simpleconstants">Simple Constants</a></li>
62 <li><a href="#aggregateconstants">Aggregate Constants</a></li>
63 <li><a href="#globalconstants">Global Variable and Function Addresses</a></li>
64 <li><a href="#undefvalues">Undefined Values</a></li>
65 <li><a href="#constantexprs">Constant Expressions</a></li>
68 <li><a href="#othervalues">Other Values</a>
70 <li><a href="#inlineasm">Inline Assembler Expressions</a></li>
73 <li><a href="#instref">Instruction Reference</a>
75 <li><a href="#terminators">Terminator Instructions</a>
77 <li><a href="#i_ret">'<tt>ret</tt>' Instruction</a></li>
78 <li><a href="#i_br">'<tt>br</tt>' Instruction</a></li>
79 <li><a href="#i_switch">'<tt>switch</tt>' Instruction</a></li>
80 <li><a href="#i_invoke">'<tt>invoke</tt>' Instruction</a></li>
81 <li><a href="#i_unwind">'<tt>unwind</tt>' Instruction</a></li>
82 <li><a href="#i_unreachable">'<tt>unreachable</tt>' Instruction</a></li>
85 <li><a href="#binaryops">Binary Operations</a>
87 <li><a href="#i_add">'<tt>add</tt>' Instruction</a></li>
88 <li><a href="#i_sub">'<tt>sub</tt>' Instruction</a></li>
89 <li><a href="#i_mul">'<tt>mul</tt>' Instruction</a></li>
90 <li><a href="#i_udiv">'<tt>udiv</tt>' Instruction</a></li>
91 <li><a href="#i_sdiv">'<tt>sdiv</tt>' Instruction</a></li>
92 <li><a href="#i_fdiv">'<tt>fdiv</tt>' Instruction</a></li>
93 <li><a href="#i_urem">'<tt>urem</tt>' Instruction</a></li>
94 <li><a href="#i_srem">'<tt>srem</tt>' Instruction</a></li>
95 <li><a href="#i_frem">'<tt>frem</tt>' Instruction</a></li>
98 <li><a href="#bitwiseops">Bitwise Binary Operations</a>
100 <li><a href="#i_shl">'<tt>shl</tt>' Instruction</a></li>
101 <li><a href="#i_lshr">'<tt>lshr</tt>' Instruction</a></li>
102 <li><a href="#i_ashr">'<tt>ashr</tt>' Instruction</a></li>
103 <li><a href="#i_and">'<tt>and</tt>' Instruction</a></li>
104 <li><a href="#i_or">'<tt>or</tt>' Instruction</a></li>
105 <li><a href="#i_xor">'<tt>xor</tt>' Instruction</a></li>
108 <li><a href="#vectorops">Vector Operations</a>
110 <li><a href="#i_extractelement">'<tt>extractelement</tt>' Instruction</a></li>
111 <li><a href="#i_insertelement">'<tt>insertelement</tt>' Instruction</a></li>
112 <li><a href="#i_shufflevector">'<tt>shufflevector</tt>' Instruction</a></li>
115 <li><a href="#aggregateops">Aggregate Operations</a>
117 <li><a href="#i_extractvalue">'<tt>extractvalue</tt>' Instruction</a></li>
118 <li><a href="#i_insertvalue">'<tt>insertvalue</tt>' Instruction</a></li>
121 <li><a href="#memoryops">Memory Access and Addressing Operations</a>
123 <li><a href="#i_malloc">'<tt>malloc</tt>' Instruction</a></li>
124 <li><a href="#i_free">'<tt>free</tt>' Instruction</a></li>
125 <li><a href="#i_alloca">'<tt>alloca</tt>' Instruction</a></li>
126 <li><a href="#i_load">'<tt>load</tt>' Instruction</a></li>
127 <li><a href="#i_store">'<tt>store</tt>' Instruction</a></li>
128 <li><a href="#i_getelementptr">'<tt>getelementptr</tt>' Instruction</a></li>
131 <li><a href="#convertops">Conversion Operations</a>
133 <li><a href="#i_trunc">'<tt>trunc .. to</tt>' Instruction</a></li>
134 <li><a href="#i_zext">'<tt>zext .. to</tt>' Instruction</a></li>
135 <li><a href="#i_sext">'<tt>sext .. to</tt>' Instruction</a></li>
136 <li><a href="#i_fptrunc">'<tt>fptrunc .. to</tt>' Instruction</a></li>
137 <li><a href="#i_fpext">'<tt>fpext .. to</tt>' Instruction</a></li>
138 <li><a href="#i_fptoui">'<tt>fptoui .. to</tt>' Instruction</a></li>
139 <li><a href="#i_fptosi">'<tt>fptosi .. to</tt>' Instruction</a></li>
140 <li><a href="#i_uitofp">'<tt>uitofp .. to</tt>' Instruction</a></li>
141 <li><a href="#i_sitofp">'<tt>sitofp .. to</tt>' Instruction</a></li>
142 <li><a href="#i_ptrtoint">'<tt>ptrtoint .. to</tt>' Instruction</a></li>
143 <li><a href="#i_inttoptr">'<tt>inttoptr .. to</tt>' Instruction</a></li>
144 <li><a href="#i_bitcast">'<tt>bitcast .. to</tt>' Instruction</a></li>
147 <li><a href="#otherops">Other Operations</a>
149 <li><a href="#i_icmp">'<tt>icmp</tt>' Instruction</a></li>
150 <li><a href="#i_fcmp">'<tt>fcmp</tt>' Instruction</a></li>
151 <li><a href="#i_vicmp">'<tt>vicmp</tt>' Instruction</a></li>
152 <li><a href="#i_vfcmp">'<tt>vfcmp</tt>' Instruction</a></li>
153 <li><a href="#i_phi">'<tt>phi</tt>' Instruction</a></li>
154 <li><a href="#i_select">'<tt>select</tt>' Instruction</a></li>
155 <li><a href="#i_call">'<tt>call</tt>' Instruction</a></li>
156 <li><a href="#i_va_arg">'<tt>va_arg</tt>' Instruction</a></li>
161 <li><a href="#intrinsics">Intrinsic Functions</a>
163 <li><a href="#int_varargs">Variable Argument Handling Intrinsics</a>
165 <li><a href="#int_va_start">'<tt>llvm.va_start</tt>' Intrinsic</a></li>
166 <li><a href="#int_va_end">'<tt>llvm.va_end</tt>' Intrinsic</a></li>
167 <li><a href="#int_va_copy">'<tt>llvm.va_copy</tt>' Intrinsic</a></li>
170 <li><a href="#int_gc">Accurate Garbage Collection Intrinsics</a>
172 <li><a href="#int_gcroot">'<tt>llvm.gcroot</tt>' Intrinsic</a></li>
173 <li><a href="#int_gcread">'<tt>llvm.gcread</tt>' Intrinsic</a></li>
174 <li><a href="#int_gcwrite">'<tt>llvm.gcwrite</tt>' Intrinsic</a></li>
177 <li><a href="#int_codegen">Code Generator Intrinsics</a>
179 <li><a href="#int_returnaddress">'<tt>llvm.returnaddress</tt>' Intrinsic</a></li>
180 <li><a href="#int_frameaddress">'<tt>llvm.frameaddress</tt>' Intrinsic</a></li>
181 <li><a href="#int_stacksave">'<tt>llvm.stacksave</tt>' Intrinsic</a></li>
182 <li><a href="#int_stackrestore">'<tt>llvm.stackrestore</tt>' Intrinsic</a></li>
183 <li><a href="#int_prefetch">'<tt>llvm.prefetch</tt>' Intrinsic</a></li>
184 <li><a href="#int_pcmarker">'<tt>llvm.pcmarker</tt>' Intrinsic</a></li>
185 <li><a href="#int_readcyclecounter"><tt>llvm.readcyclecounter</tt>' Intrinsic</a></li>
188 <li><a href="#int_libc">Standard C Library Intrinsics</a>
190 <li><a href="#int_memcpy">'<tt>llvm.memcpy.*</tt>' Intrinsic</a></li>
191 <li><a href="#int_memmove">'<tt>llvm.memmove.*</tt>' Intrinsic</a></li>
192 <li><a href="#int_memset">'<tt>llvm.memset.*</tt>' Intrinsic</a></li>
193 <li><a href="#int_sqrt">'<tt>llvm.sqrt.*</tt>' Intrinsic</a></li>
194 <li><a href="#int_powi">'<tt>llvm.powi.*</tt>' Intrinsic</a></li>
195 <li><a href="#int_sin">'<tt>llvm.sin.*</tt>' Intrinsic</a></li>
196 <li><a href="#int_cos">'<tt>llvm.cos.*</tt>' Intrinsic</a></li>
197 <li><a href="#int_pow">'<tt>llvm.pow.*</tt>' Intrinsic</a></li>
200 <li><a href="#int_manip">Bit Manipulation Intrinsics</a>
202 <li><a href="#int_bswap">'<tt>llvm.bswap.*</tt>' Intrinsics</a></li>
203 <li><a href="#int_ctpop">'<tt>llvm.ctpop.*</tt>' Intrinsic </a></li>
204 <li><a href="#int_ctlz">'<tt>llvm.ctlz.*</tt>' Intrinsic </a></li>
205 <li><a href="#int_cttz">'<tt>llvm.cttz.*</tt>' Intrinsic </a></li>
206 <li><a href="#int_part_select">'<tt>llvm.part.select.*</tt>' Intrinsic </a></li>
207 <li><a href="#int_part_set">'<tt>llvm.part.set.*</tt>' Intrinsic </a></li>
210 <li><a href="#int_debugger">Debugger intrinsics</a></li>
211 <li><a href="#int_eh">Exception Handling intrinsics</a></li>
212 <li><a href="#int_trampoline">Trampoline Intrinsic</a>
214 <li><a href="#int_it">'<tt>llvm.init.trampoline</tt>' Intrinsic</a></li>
217 <li><a href="#int_atomics">Atomic intrinsics</a>
219 <li><a href="#int_memory_barrier"><tt>llvm.memory_barrier</tt></a></li>
220 <li><a href="#int_atomic_cmp_swap"><tt>llvm.atomic.cmp.swap</tt></a></li>
221 <li><a href="#int_atomic_swap"><tt>llvm.atomic.swap</tt></a></li>
222 <li><a href="#int_atomic_load_add"><tt>llvm.atomic.load.add</tt></a></li>
223 <li><a href="#int_atomic_load_sub"><tt>llvm.atomic.load.sub</tt></a></li>
224 <li><a href="#int_atomic_load_and"><tt>llvm.atomic.load.and</tt></a></li>
225 <li><a href="#int_atomic_load_nand"><tt>llvm.atomic.load.nand</tt></a></li>
226 <li><a href="#int_atomic_load_or"><tt>llvm.atomic.load.or</tt></a></li>
227 <li><a href="#int_atomic_load_xor"><tt>llvm.atomic.load.xor</tt></a></li>
228 <li><a href="#int_atomic_load_max"><tt>llvm.atomic.load.max</tt></a></li>
229 <li><a href="#int_atomic_load_min"><tt>llvm.atomic.load.min</tt></a></li>
230 <li><a href="#int_atomic_load_umax"><tt>llvm.atomic.load.umax</tt></a></li>
231 <li><a href="#int_atomic_load_umin"><tt>llvm.atomic.load.umin</tt></a></li>
234 <li><a href="#int_general">General intrinsics</a>
236 <li><a href="#int_var_annotation">
237 '<tt>llvm.var.annotation</tt>' Intrinsic</a></li>
238 <li><a href="#int_annotation">
239 '<tt>llvm.annotation.*</tt>' Intrinsic</a></li>
240 <li><a href="#int_trap">
241 '<tt>llvm.trap</tt>' Intrinsic</a></li>
242 <li><a href="#int_stackprotector">
243 '<tt>llvm.stackprotector</tt>' Intrinsic</a></li>
250 <div class="doc_author">
251 <p>Written by <a href="mailto:sabre@nondot.org">Chris Lattner</a>
252 and <a href="mailto:vadve@cs.uiuc.edu">Vikram Adve</a></p>
255 <!-- *********************************************************************** -->
256 <div class="doc_section"> <a name="abstract">Abstract </a></div>
257 <!-- *********************************************************************** -->
259 <div class="doc_text">
260 <p>This document is a reference manual for the LLVM assembly language.
261 LLVM is a Static Single Assignment (SSA) based representation that provides
262 type safety, low-level operations, flexibility, and the capability of
263 representing 'all' high-level languages cleanly. It is the common code
264 representation used throughout all phases of the LLVM compilation
268 <!-- *********************************************************************** -->
269 <div class="doc_section"> <a name="introduction">Introduction</a> </div>
270 <!-- *********************************************************************** -->
272 <div class="doc_text">
274 <p>The LLVM code representation is designed to be used in three
275 different forms: as an in-memory compiler IR, as an on-disk bitcode
276 representation (suitable for fast loading by a Just-In-Time compiler),
277 and as a human readable assembly language representation. This allows
278 LLVM to provide a powerful intermediate representation for efficient
279 compiler transformations and analysis, while providing a natural means
280 to debug and visualize the transformations. The three different forms
281 of LLVM are all equivalent. This document describes the human readable
282 representation and notation.</p>
284 <p>The LLVM representation aims to be light-weight and low-level
285 while being expressive, typed, and extensible at the same time. It
286 aims to be a "universal IR" of sorts, by being at a low enough level
287 that high-level ideas may be cleanly mapped to it (similar to how
288 microprocessors are "universal IR's", allowing many source languages to
289 be mapped to them). By providing type information, LLVM can be used as
290 the target of optimizations: for example, through pointer analysis, it
291 can be proven that a C automatic variable is never accessed outside of
292 the current function... allowing it to be promoted to a simple SSA
293 value instead of a memory location.</p>
297 <!-- _______________________________________________________________________ -->
298 <div class="doc_subsubsection"> <a name="wellformed">Well-Formedness</a> </div>
300 <div class="doc_text">
302 <p>It is important to note that this document describes 'well formed'
303 LLVM assembly language. There is a difference between what the parser
304 accepts and what is considered 'well formed'. For example, the
305 following instruction is syntactically okay, but not well formed:</p>
307 <div class="doc_code">
309 %x = <a href="#i_add">add</a> i32 1, %x
313 <p>...because the definition of <tt>%x</tt> does not dominate all of
314 its uses. The LLVM infrastructure provides a verification pass that may
315 be used to verify that an LLVM module is well formed. This pass is
316 automatically run by the parser after parsing input assembly and by
317 the optimizer before it outputs bitcode. The violations pointed out
318 by the verifier pass indicate bugs in transformation passes or input to
322 <!-- Describe the typesetting conventions here. -->
324 <!-- *********************************************************************** -->
325 <div class="doc_section"> <a name="identifiers">Identifiers</a> </div>
326 <!-- *********************************************************************** -->
328 <div class="doc_text">
330 <p>LLVM identifiers come in two basic types: global and local. Global
331 identifiers (functions, global variables) begin with the @ character. Local
332 identifiers (register names, types) begin with the % character. Additionally,
333 there are three different formats for identifiers, for different purposes:</p>
336 <li>Named values are represented as a string of characters with their prefix.
337 For example, %foo, @DivisionByZero, %a.really.long.identifier. The actual
338 regular expression used is '<tt>[%@][a-zA-Z$._][a-zA-Z$._0-9]*</tt>'.
339 Identifiers which require other characters in their names can be surrounded
340 with quotes. Special characters may be escaped using "\xx" where xx is the
341 ASCII code for the character in hexadecimal. In this way, any character can
342 be used in a name value, even quotes themselves.
344 <li>Unnamed values are represented as an unsigned numeric value with their
345 prefix. For example, %12, @2, %44.</li>
347 <li>Constants, which are described in a <a href="#constants">section about
348 constants</a>, below.</li>
351 <p>LLVM requires that values start with a prefix for two reasons: Compilers
352 don't need to worry about name clashes with reserved words, and the set of
353 reserved words may be expanded in the future without penalty. Additionally,
354 unnamed identifiers allow a compiler to quickly come up with a temporary
355 variable without having to avoid symbol table conflicts.</p>
357 <p>Reserved words in LLVM are very similar to reserved words in other
358 languages. There are keywords for different opcodes
359 ('<tt><a href="#i_add">add</a></tt>',
360 '<tt><a href="#i_bitcast">bitcast</a></tt>',
361 '<tt><a href="#i_ret">ret</a></tt>', etc...), for primitive type names ('<tt><a
362 href="#t_void">void</a></tt>', '<tt><a href="#t_primitive">i32</a></tt>', etc...),
363 and others. These reserved words cannot conflict with variable names, because
364 none of them start with a prefix character ('%' or '@').</p>
366 <p>Here is an example of LLVM code to multiply the integer variable
367 '<tt>%X</tt>' by 8:</p>
371 <div class="doc_code">
373 %result = <a href="#i_mul">mul</a> i32 %X, 8
377 <p>After strength reduction:</p>
379 <div class="doc_code">
381 %result = <a href="#i_shl">shl</a> i32 %X, i8 3
385 <p>And the hard way:</p>
387 <div class="doc_code">
389 <a href="#i_add">add</a> i32 %X, %X <i>; yields {i32}:%0</i>
390 <a href="#i_add">add</a> i32 %0, %0 <i>; yields {i32}:%1</i>
391 %result = <a href="#i_add">add</a> i32 %1, %1
395 <p>This last way of multiplying <tt>%X</tt> by 8 illustrates several
396 important lexical features of LLVM:</p>
400 <li>Comments are delimited with a '<tt>;</tt>' and go until the end of
403 <li>Unnamed temporaries are created when the result of a computation is not
404 assigned to a named value.</li>
406 <li>Unnamed temporaries are numbered sequentially</li>
410 <p>...and it also shows a convention that we follow in this document. When
411 demonstrating instructions, we will follow an instruction with a comment that
412 defines the type and name of value produced. Comments are shown in italic
417 <!-- *********************************************************************** -->
418 <div class="doc_section"> <a name="highlevel">High Level Structure</a> </div>
419 <!-- *********************************************************************** -->
421 <!-- ======================================================================= -->
422 <div class="doc_subsection"> <a name="modulestructure">Module Structure</a>
425 <div class="doc_text">
427 <p>LLVM programs are composed of "Module"s, each of which is a
428 translation unit of the input programs. Each module consists of
429 functions, global variables, and symbol table entries. Modules may be
430 combined together with the LLVM linker, which merges function (and
431 global variable) definitions, resolves forward declarations, and merges
432 symbol table entries. Here is an example of the "hello world" module:</p>
434 <div class="doc_code">
435 <pre><i>; Declare the string constant as a global constant...</i>
436 <a href="#identifiers">@.LC0</a> = <a href="#linkage_internal">internal</a> <a
437 href="#globalvars">constant</a> <a href="#t_array">[13 x i8]</a> c"hello world\0A\00" <i>; [13 x i8]*</i>
439 <i>; External declaration of the puts function</i>
440 <a href="#functionstructure">declare</a> i32 @puts(i8 *) <i>; i32(i8 *)* </i>
442 <i>; Definition of main function</i>
443 define i32 @main() { <i>; i32()* </i>
444 <i>; Convert [13x i8 ]* to i8 *...</i>
446 href="#i_getelementptr">getelementptr</a> [13 x i8 ]* @.LC0, i64 0, i64 0 <i>; i8 *</i>
448 <i>; Call puts function to write out the string to stdout...</i>
450 href="#i_call">call</a> i32 @puts(i8 * %cast210) <i>; i32</i>
452 href="#i_ret">ret</a> i32 0<br>}<br>
456 <p>This example is made up of a <a href="#globalvars">global variable</a>
457 named "<tt>.LC0</tt>", an external declaration of the "<tt>puts</tt>"
458 function, and a <a href="#functionstructure">function definition</a>
459 for "<tt>main</tt>".</p>
461 <p>In general, a module is made up of a list of global values,
462 where both functions and global variables are global values. Global values are
463 represented by a pointer to a memory location (in this case, a pointer to an
464 array of char, and a pointer to a function), and have one of the following <a
465 href="#linkage">linkage types</a>.</p>
469 <!-- ======================================================================= -->
470 <div class="doc_subsection">
471 <a name="linkage">Linkage Types</a>
474 <div class="doc_text">
477 All Global Variables and Functions have one of the following types of linkage:
482 <dt><tt><b><a name="linkage_internal">internal</a></b></tt>: </dt>
484 <dd>Global values with internal linkage are only directly accessible by
485 objects in the current module. In particular, linking code into a module with
486 an internal global value may cause the internal to be renamed as necessary to
487 avoid collisions. Because the symbol is internal to the module, all
488 references can be updated. This corresponds to the notion of the
489 '<tt>static</tt>' keyword in C.
492 <dt><tt><b><a name="linkage_linkonce">linkonce</a></b></tt>: </dt>
494 <dd>Globals with "<tt>linkonce</tt>" linkage are merged with other globals of
495 the same name when linkage occurs. This is typically used to implement
496 inline functions, templates, or other code which must be generated in each
497 translation unit that uses it. Unreferenced <tt>linkonce</tt> globals are
498 allowed to be discarded.
501 <dt><tt><b><a name="linkage_common">common</a></b></tt>: </dt>
503 <dd>"<tt>common</tt>" linkage is exactly the same as <tt>linkonce</tt>
504 linkage, except that unreferenced <tt>common</tt> globals may not be
505 discarded. This is used for globals that may be emitted in multiple
506 translation units, but that are not guaranteed to be emitted into every
507 translation unit that uses them. One example of this is tentative
508 definitions in C, such as "<tt>int X;</tt>" at global scope.
511 <dt><tt><b><a name="linkage_weak">weak</a></b></tt>: </dt>
513 <dd>"<tt>weak</tt>" linkage is the same as <tt>common</tt> linkage, except
514 that some targets may choose to emit different assembly sequences for them
515 for target-dependent reasons. This is used for globals that are declared
516 "weak" in C source code.
519 <dt><tt><b><a name="linkage_appending">appending</a></b></tt>: </dt>
521 <dd>"<tt>appending</tt>" linkage may only be applied to global variables of
522 pointer to array type. When two global variables with appending linkage are
523 linked together, the two global arrays are appended together. This is the
524 LLVM, typesafe, equivalent of having the system linker append together
525 "sections" with identical names when .o files are linked.
528 <dt><tt><b><a name="linkage_externweak">extern_weak</a></b></tt>: </dt>
529 <dd>The semantics of this linkage follow the ELF object file model: the
530 symbol is weak until linked, if not linked, the symbol becomes null instead
531 of being an undefined reference.
534 <dt><tt><b><a name="linkage_external">externally visible</a></b></tt>:</dt>
536 <dd>If none of the above identifiers are used, the global is externally
537 visible, meaning that it participates in linkage and can be used to resolve
538 external symbol references.
543 The next two types of linkage are targeted for Microsoft Windows platform
544 only. They are designed to support importing (exporting) symbols from (to)
545 DLLs (Dynamic Link Libraries).
549 <dt><tt><b><a name="linkage_dllimport">dllimport</a></b></tt>: </dt>
551 <dd>"<tt>dllimport</tt>" linkage causes the compiler to reference a function
552 or variable via a global pointer to a pointer that is set up by the DLL
553 exporting the symbol. On Microsoft Windows targets, the pointer name is
554 formed by combining <code>_imp__</code> and the function or variable name.
557 <dt><tt><b><a name="linkage_dllexport">dllexport</a></b></tt>: </dt>
559 <dd>"<tt>dllexport</tt>" linkage causes the compiler to provide a global
560 pointer to a pointer in a DLL, so that it can be referenced with the
561 <tt>dllimport</tt> attribute. On Microsoft Windows targets, the pointer
562 name is formed by combining <code>_imp__</code> and the function or variable
568 <p><a name="linkage_external"></a>For example, since the "<tt>.LC0</tt>"
569 variable is defined to be internal, if another module defined a "<tt>.LC0</tt>"
570 variable and was linked with this one, one of the two would be renamed,
571 preventing a collision. Since "<tt>main</tt>" and "<tt>puts</tt>" are
572 external (i.e., lacking any linkage declarations), they are accessible
573 outside of the current module.</p>
574 <p>It is illegal for a function <i>declaration</i>
575 to have any linkage type other than "externally visible", <tt>dllimport</tt>,
576 or <tt>extern_weak</tt>.</p>
577 <p>Aliases can have only <tt>external</tt>, <tt>internal</tt> and <tt>weak</tt>
581 <!-- ======================================================================= -->
582 <div class="doc_subsection">
583 <a name="callingconv">Calling Conventions</a>
586 <div class="doc_text">
588 <p>LLVM <a href="#functionstructure">functions</a>, <a href="#i_call">calls</a>
589 and <a href="#i_invoke">invokes</a> can all have an optional calling convention
590 specified for the call. The calling convention of any pair of dynamic
591 caller/callee must match, or the behavior of the program is undefined. The
592 following calling conventions are supported by LLVM, and more may be added in
596 <dt><b>"<tt>ccc</tt>" - The C calling convention</b>:</dt>
598 <dd>This calling convention (the default if no other calling convention is
599 specified) matches the target C calling conventions. This calling convention
600 supports varargs function calls and tolerates some mismatch in the declared
601 prototype and implemented declaration of the function (as does normal C).
604 <dt><b>"<tt>fastcc</tt>" - The fast calling convention</b>:</dt>
606 <dd>This calling convention attempts to make calls as fast as possible
607 (e.g. by passing things in registers). This calling convention allows the
608 target to use whatever tricks it wants to produce fast code for the target,
609 without having to conform to an externally specified ABI (Application Binary
610 Interface). Implementations of this convention should allow arbitrary
611 <a href="CodeGenerator.html#tailcallopt">tail call optimization</a> to be
612 supported. This calling convention does not support varargs and requires the
613 prototype of all callees to exactly match the prototype of the function
617 <dt><b>"<tt>coldcc</tt>" - The cold calling convention</b>:</dt>
619 <dd>This calling convention attempts to make code in the caller as efficient
620 as possible under the assumption that the call is not commonly executed. As
621 such, these calls often preserve all registers so that the call does not break
622 any live ranges in the caller side. This calling convention does not support
623 varargs and requires the prototype of all callees to exactly match the
624 prototype of the function definition.
627 <dt><b>"<tt>cc <<em>n</em>></tt>" - Numbered convention</b>:</dt>
629 <dd>Any calling convention may be specified by number, allowing
630 target-specific calling conventions to be used. Target specific calling
631 conventions start at 64.
635 <p>More calling conventions can be added/defined on an as-needed basis, to
636 support pascal conventions or any other well-known target-independent
641 <!-- ======================================================================= -->
642 <div class="doc_subsection">
643 <a name="visibility">Visibility Styles</a>
646 <div class="doc_text">
649 All Global Variables and Functions have one of the following visibility styles:
653 <dt><b>"<tt>default</tt>" - Default style</b>:</dt>
655 <dd>On targets that use the ELF object file format, default visibility means
656 that the declaration is visible to other
657 modules and, in shared libraries, means that the declared entity may be
658 overridden. On Darwin, default visibility means that the declaration is
659 visible to other modules. Default visibility corresponds to "external
660 linkage" in the language.
663 <dt><b>"<tt>hidden</tt>" - Hidden style</b>:</dt>
665 <dd>Two declarations of an object with hidden visibility refer to the same
666 object if they are in the same shared object. Usually, hidden visibility
667 indicates that the symbol will not be placed into the dynamic symbol table,
668 so no other module (executable or shared library) can reference it
672 <dt><b>"<tt>protected</tt>" - Protected style</b>:</dt>
674 <dd>On ELF, protected visibility indicates that the symbol will be placed in
675 the dynamic symbol table, but that references within the defining module will
676 bind to the local symbol. That is, the symbol cannot be overridden by another
683 <!-- ======================================================================= -->
684 <div class="doc_subsection">
685 <a name="globalvars">Global Variables</a>
688 <div class="doc_text">
690 <p>Global variables define regions of memory allocated at compilation time
691 instead of run-time. Global variables may optionally be initialized, may have
692 an explicit section to be placed in, and may have an optional explicit alignment
693 specified. A variable may be defined as "thread_local", which means that it
694 will not be shared by threads (each thread will have a separated copy of the
695 variable). A variable may be defined as a global "constant," which indicates
696 that the contents of the variable will <b>never</b> be modified (enabling better
697 optimization, allowing the global data to be placed in the read-only section of
698 an executable, etc). Note that variables that need runtime initialization
699 cannot be marked "constant" as there is a store to the variable.</p>
702 LLVM explicitly allows <em>declarations</em> of global variables to be marked
703 constant, even if the final definition of the global is not. This capability
704 can be used to enable slightly better optimization of the program, but requires
705 the language definition to guarantee that optimizations based on the
706 'constantness' are valid for the translation units that do not include the
710 <p>As SSA values, global variables define pointer values that are in
711 scope (i.e. they dominate) all basic blocks in the program. Global
712 variables always define a pointer to their "content" type because they
713 describe a region of memory, and all memory objects in LLVM are
714 accessed through pointers.</p>
716 <p>A global variable may be declared to reside in a target-specifc numbered
717 address space. For targets that support them, address spaces may affect how
718 optimizations are performed and/or what target instructions are used to access
719 the variable. The default address space is zero. The address space qualifier
720 must precede any other attributes.</p>
722 <p>LLVM allows an explicit section to be specified for globals. If the target
723 supports it, it will emit globals to the section specified.</p>
725 <p>An explicit alignment may be specified for a global. If not present, or if
726 the alignment is set to zero, the alignment of the global is set by the target
727 to whatever it feels convenient. If an explicit alignment is specified, the
728 global is forced to have at least that much alignment. All alignments must be
731 <p>For example, the following defines a global in a numbered address space with
732 an initializer, section, and alignment:</p>
734 <div class="doc_code">
736 @G = constant float 1.0 addrspace(5), section "foo", align 4
743 <!-- ======================================================================= -->
744 <div class="doc_subsection">
745 <a name="functionstructure">Functions</a>
748 <div class="doc_text">
750 <p>LLVM function definitions consist of the "<tt>define</tt>" keyord,
751 an optional <a href="#linkage">linkage type</a>, an optional
752 <a href="#visibility">visibility style</a>, an optional
753 <a href="#callingconv">calling convention</a>, a return type, an optional
754 <a href="#paramattrs">parameter attribute</a> for the return type, a function
755 name, a (possibly empty) argument list (each with optional
756 <a href="#paramattrs">parameter attributes</a>), optional
757 <a href="#fnattrs">function attributes</a>, an optional section,
758 an optional alignment, an optional <a href="#gc">garbage collector name</a>,
759 an opening curly brace, a list of basic blocks, and a closing curly brace.
761 LLVM function declarations consist of the "<tt>declare</tt>" keyword, an
762 optional <a href="#linkage">linkage type</a>, an optional
763 <a href="#visibility">visibility style</a>, an optional
764 <a href="#callingconv">calling convention</a>, a return type, an optional
765 <a href="#paramattrs">parameter attribute</a> for the return type, a function
766 name, a possibly empty list of arguments, an optional alignment, and an optional
767 <a href="#gc">garbage collector name</a>.</p>
769 <p>A function definition contains a list of basic blocks, forming the CFG
770 (Control Flow Graph) for
771 the function. Each basic block may optionally start with a label (giving the
772 basic block a symbol table entry), contains a list of instructions, and ends
773 with a <a href="#terminators">terminator</a> instruction (such as a branch or
774 function return).</p>
776 <p>The first basic block in a function is special in two ways: it is immediately
777 executed on entrance to the function, and it is not allowed to have predecessor
778 basic blocks (i.e. there can not be any branches to the entry block of a
779 function). Because the block can have no predecessors, it also cannot have any
780 <a href="#i_phi">PHI nodes</a>.</p>
782 <p>LLVM allows an explicit section to be specified for functions. If the target
783 supports it, it will emit functions to the section specified.</p>
785 <p>An explicit alignment may be specified for a function. If not present, or if
786 the alignment is set to zero, the alignment of the function is set by the target
787 to whatever it feels convenient. If an explicit alignment is specified, the
788 function is forced to have at least that much alignment. All alignments must be
793 <div class="doc_code">
795 define [<a href="#linkage">linkage</a>] [<a href="#visibility">visibility</a>]
796 [<a href="#callingconv">cconv</a>] [<a href="#paramattrs">ret attrs</a>]
797 <ResultType> @<FunctionName> ([argument list])
798 [<a href="#fnattrs">fn Attrs</a>] [section "name"] [align N]
799 [<a href="#gc">gc</a>] { ... }
806 <!-- ======================================================================= -->
807 <div class="doc_subsection">
808 <a name="aliasstructure">Aliases</a>
810 <div class="doc_text">
811 <p>Aliases act as "second name" for the aliasee value (which can be either
812 function, global variable, another alias or bitcast of global value). Aliases
813 may have an optional <a href="#linkage">linkage type</a>, and an
814 optional <a href="#visibility">visibility style</a>.</p>
818 <div class="doc_code">
820 @<Name> = alias [Linkage] [Visibility] <AliaseeTy> @<Aliasee>
828 <!-- ======================================================================= -->
829 <div class="doc_subsection"><a name="paramattrs">Parameter Attributes</a></div>
830 <div class="doc_text">
831 <p>The return type and each parameter of a function type may have a set of
832 <i>parameter attributes</i> associated with them. Parameter attributes are
833 used to communicate additional information about the result or parameters of
834 a function. Parameter attributes are considered to be part of the function,
835 not of the function type, so functions with different parameter attributes
836 can have the same function type.</p>
838 <p>Parameter attributes are simple keywords that follow the type specified. If
839 multiple parameter attributes are needed, they are space separated. For
842 <div class="doc_code">
844 declare i32 @printf(i8* noalias , ...)
845 declare i32 @atoi(i8 zeroext)
846 declare signext i8 @returns_signed_char()
850 <p>Note that any attributes for the function result (<tt>nounwind</tt>,
851 <tt>readonly</tt>) come immediately after the argument list.</p>
853 <p>Currently, only the following parameter attributes are defined:</p>
855 <dt><tt>zeroext</tt></dt>
856 <dd>This indicates to the code generator that the parameter or return value
857 should be zero-extended to a 32-bit value by the caller (for a parameter)
858 or the callee (for a return value).</dd>
860 <dt><tt>signext</tt></dt>
861 <dd>This indicates to the code generator that the parameter or return value
862 should be sign-extended to a 32-bit value by the caller (for a parameter)
863 or the callee (for a return value).</dd>
865 <dt><tt>inreg</tt></dt>
866 <dd>This indicates that this parameter or return value should be treated
867 in a special target-dependent fashion during while emitting code for a
868 function call or return (usually, by putting it in a register as opposed
869 to memory, though some targets use it to distinguish between two different
870 kinds of registers). Use of this attribute is target-specific.</dd>
872 <dt><tt><a name="byval">byval</a></tt></dt>
873 <dd>This indicates that the pointer parameter should really be passed by
874 value to the function. The attribute implies that a hidden copy of the
875 pointee is made between the caller and the callee, so the callee is unable
876 to modify the value in the callee. This attribute is only valid on LLVM
877 pointer arguments. It is generally used to pass structs and arrays by
878 value, but is also valid on pointers to scalars. The copy is considered to
879 belong to the caller not the callee (for example,
880 <tt><a href="#readonly">readonly</a></tt> functions should not write to
881 <tt>byval</tt> parameters). This is not a valid attribute for return
884 <dt><tt>sret</tt></dt>
885 <dd>This indicates that the pointer parameter specifies the address of a
886 structure that is the return value of the function in the source program.
887 This pointer must be guaranteed by the caller to be valid: loads and stores
888 to the structure may be assumed by the callee to not to trap. This may only
889 be applied to the first parameter. This is not a valid attribute for
892 <dt><tt>noalias</tt></dt>
893 <dd>This indicates that the pointer does not alias any global or any other
894 parameter. The caller is responsible for ensuring that this is the
895 case. On a function return value, <tt>noalias</tt> additionally indicates
896 that the pointer does not alias any other pointers visible to the
899 <dt><tt>nest</tt></dt>
900 <dd>This indicates that the pointer parameter can be excised using the
901 <a href="#int_trampoline">trampoline intrinsics</a>. This is not a valid
902 attribute for return values.</dd>
907 <!-- ======================================================================= -->
908 <div class="doc_subsection">
909 <a name="gc">Garbage Collector Names</a>
912 <div class="doc_text">
913 <p>Each function may specify a garbage collector name, which is simply a
916 <div class="doc_code"><pre
917 >define void @f() gc "name" { ...</pre></div>
919 <p>The compiler declares the supported values of <i>name</i>. Specifying a
920 collector which will cause the compiler to alter its output in order to support
921 the named garbage collection algorithm.</p>
924 <!-- ======================================================================= -->
925 <div class="doc_subsection">
926 <a name="fnattrs">Function Attributes</a>
929 <div class="doc_text">
931 <p>Function attributes are set to communicate additional information about
932 a function. Function attributes are considered to be part of the function,
933 not of the function type, so functions with different parameter attributes
934 can have the same function type.</p>
936 <p>Function attributes are simple keywords that follow the type specified. If
937 multiple attributes are needed, they are space separated. For
940 <div class="doc_code">
942 define void @f() noinline { ... }
943 define void @f() alwaysinline { ... }
944 define void @f() alwaysinline optsize { ... }
945 define void @f() optsize
950 <dt><tt>alwaysinline</tt></dt>
951 <dd>This attribute indicates that the inliner should attempt to inline this
952 function into callers whenever possible, ignoring any active inlining size
953 threshold for this caller.</dd>
955 <dt><tt>noinline</tt></dt>
956 <dd>This attribute indicates that the inliner should never inline this function
957 in any situation. This attribute may not be used together with the
958 <tt>alwaysinline</tt> attribute.</dd>
960 <dt><tt>optsize</tt></dt>
961 <dd>This attribute suggests that optimization passes and code generator passes
962 make choices that keep the code size of this function low, and otherwise do
963 optimizations specifically to reduce code size.</dd>
965 <dt><tt>noreturn</tt></dt>
966 <dd>This function attribute indicates that the function never returns normally.
967 This produces undefined behavior at runtime if the function ever does
968 dynamically return.</dd>
970 <dt><tt>nounwind</tt></dt>
971 <dd>This function attribute indicates that the function never returns with an
972 unwind or exceptional control flow. If the function does unwind, its runtime
973 behavior is undefined.</dd>
975 <dt><tt>readnone</tt></dt>
976 <dd>This attribute indicates that the function computes its result (or the
977 exception it throws) based strictly on its arguments, without dereferencing any
978 pointer arguments or otherwise accessing any mutable state (e.g. memory, control
979 registers, etc) visible to caller functions. It does not write through any
980 pointer arguments (including <tt><a href="#byval">byval</a></tt> arguments) and
981 never changes any state visible to callers.</dd>
983 <dt><tt><a name="readonly">readonly</a></tt></dt>
984 <dd>This attribute indicates that the function does not write through any
985 pointer arguments (including <tt><a href="#byval">byval</a></tt> arguments)
986 or otherwise modify any state (e.g. memory, control registers, etc) visible to
987 caller functions. It may dereference pointer arguments and read state that may
988 be set in the caller. A readonly function always returns the same value (or
989 throws the same exception) when called with the same set of arguments and global
992 <dt><tt><a name="ssp">ssp</a></tt></dt>
993 <dd>This attribute indicates that the function should emit a stack smashing
994 protector. It is in the form of a "canary"—a random value placed on the
995 stack before the local variables that's checked upon return from the function to
996 see if it has been overwritten. A heuristic is used to determine if a function
997 needs stack protectors or not.</dd>
999 <dt><tt>ssp-req</tt></dt>
1000 <dd>This attribute indicates that the function should <em>always</em> emit a
1001 stack smashing protector. This overrides the <tt><a href="#ssp">ssp</a></tt>
1002 function attribute.</dd>
1007 <!-- ======================================================================= -->
1008 <div class="doc_subsection">
1009 <a name="moduleasm">Module-Level Inline Assembly</a>
1012 <div class="doc_text">
1014 Modules may contain "module-level inline asm" blocks, which corresponds to the
1015 GCC "file scope inline asm" blocks. These blocks are internally concatenated by
1016 LLVM and treated as a single unit, but may be separated in the .ll file if
1017 desired. The syntax is very simple:
1020 <div class="doc_code">
1022 module asm "inline asm code goes here"
1023 module asm "more can go here"
1027 <p>The strings can contain any character by escaping non-printable characters.
1028 The escape sequence used is simply "\xx" where "xx" is the two digit hex code
1033 The inline asm code is simply printed to the machine code .s file when
1034 assembly code is generated.
1038 <!-- ======================================================================= -->
1039 <div class="doc_subsection">
1040 <a name="datalayout">Data Layout</a>
1043 <div class="doc_text">
1044 <p>A module may specify a target specific data layout string that specifies how
1045 data is to be laid out in memory. The syntax for the data layout is simply:</p>
1046 <pre> target datalayout = "<i>layout specification</i>"</pre>
1047 <p>The <i>layout specification</i> consists of a list of specifications
1048 separated by the minus sign character ('-'). Each specification starts with a
1049 letter and may include other information after the letter to define some
1050 aspect of the data layout. The specifications accepted are as follows: </p>
1053 <dd>Specifies that the target lays out data in big-endian form. That is, the
1054 bits with the most significance have the lowest address location.</dd>
1056 <dd>Specifies that the target lays out data in little-endian form. That is,
1057 the bits with the least significance have the lowest address location.</dd>
1058 <dt><tt>p:<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1059 <dd>This specifies the <i>size</i> of a pointer and its <i>abi</i> and
1060 <i>preferred</i> alignments. All sizes are in bits. Specifying the <i>pref</i>
1061 alignment is optional. If omitted, the preceding <tt>:</tt> should be omitted
1063 <dt><tt>i<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1064 <dd>This specifies the alignment for an integer type of a given bit
1065 <i>size</i>. The value of <i>size</i> must be in the range [1,2^23).</dd>
1066 <dt><tt>v<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1067 <dd>This specifies the alignment for a vector type of a given bit
1069 <dt><tt>f<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1070 <dd>This specifies the alignment for a floating point type of a given bit
1071 <i>size</i>. The value of <i>size</i> must be either 32 (float) or 64
1073 <dt><tt>a<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1074 <dd>This specifies the alignment for an aggregate type of a given bit
1077 <p>When constructing the data layout for a given target, LLVM starts with a
1078 default set of specifications which are then (possibly) overriden by the
1079 specifications in the <tt>datalayout</tt> keyword. The default specifications
1080 are given in this list:</p>
1082 <li><tt>E</tt> - big endian</li>
1083 <li><tt>p:32:64:64</tt> - 32-bit pointers with 64-bit alignment</li>
1084 <li><tt>i1:8:8</tt> - i1 is 8-bit (byte) aligned</li>
1085 <li><tt>i8:8:8</tt> - i8 is 8-bit (byte) aligned</li>
1086 <li><tt>i16:16:16</tt> - i16 is 16-bit aligned</li>
1087 <li><tt>i32:32:32</tt> - i32 is 32-bit aligned</li>
1088 <li><tt>i64:32:64</tt> - i64 has ABI alignment of 32-bits but preferred
1089 alignment of 64-bits</li>
1090 <li><tt>f32:32:32</tt> - float is 32-bit aligned</li>
1091 <li><tt>f64:64:64</tt> - double is 64-bit aligned</li>
1092 <li><tt>v64:64:64</tt> - 64-bit vector is 64-bit aligned</li>
1093 <li><tt>v128:128:128</tt> - 128-bit vector is 128-bit aligned</li>
1094 <li><tt>a0:0:1</tt> - aggregates are 8-bit aligned</li>
1096 <p>When LLVM is determining the alignment for a given type, it uses the
1097 following rules:</p>
1099 <li>If the type sought is an exact match for one of the specifications, that
1100 specification is used.</li>
1101 <li>If no match is found, and the type sought is an integer type, then the
1102 smallest integer type that is larger than the bitwidth of the sought type is
1103 used. If none of the specifications are larger than the bitwidth then the the
1104 largest integer type is used. For example, given the default specifications
1105 above, the i7 type will use the alignment of i8 (next largest) while both
1106 i65 and i256 will use the alignment of i64 (largest specified).</li>
1107 <li>If no match is found, and the type sought is a vector type, then the
1108 largest vector type that is smaller than the sought vector type will be used
1109 as a fall back. This happens because <128 x double> can be implemented
1110 in terms of 64 <2 x double>, for example.</li>
1114 <!-- *********************************************************************** -->
1115 <div class="doc_section"> <a name="typesystem">Type System</a> </div>
1116 <!-- *********************************************************************** -->
1118 <div class="doc_text">
1120 <p>The LLVM type system is one of the most important features of the
1121 intermediate representation. Being typed enables a number of
1122 optimizations to be performed on the intermediate representation directly,
1123 without having to do
1124 extra analyses on the side before the transformation. A strong type
1125 system makes it easier to read the generated code and enables novel
1126 analyses and transformations that are not feasible to perform on normal
1127 three address code representations.</p>
1131 <!-- ======================================================================= -->
1132 <div class="doc_subsection"> <a name="t_classifications">Type
1133 Classifications</a> </div>
1134 <div class="doc_text">
1135 <p>The types fall into a few useful
1136 classifications:</p>
1138 <table border="1" cellspacing="0" cellpadding="4">
1140 <tr><th>Classification</th><th>Types</th></tr>
1142 <td><a href="#t_integer">integer</a></td>
1143 <td><tt>i1, i2, i3, ... i8, ... i16, ... i32, ... i64, ... </tt></td>
1146 <td><a href="#t_floating">floating point</a></td>
1147 <td><tt>float, double, x86_fp80, fp128, ppc_fp128</tt></td>
1150 <td><a name="t_firstclass">first class</a></td>
1151 <td><a href="#t_integer">integer</a>,
1152 <a href="#t_floating">floating point</a>,
1153 <a href="#t_pointer">pointer</a>,
1154 <a href="#t_vector">vector</a>,
1155 <a href="#t_struct">structure</a>,
1156 <a href="#t_array">array</a>,
1157 <a href="#t_label">label</a>.
1161 <td><a href="#t_primitive">primitive</a></td>
1162 <td><a href="#t_label">label</a>,
1163 <a href="#t_void">void</a>,
1164 <a href="#t_floating">floating point</a>.</td>
1167 <td><a href="#t_derived">derived</a></td>
1168 <td><a href="#t_integer">integer</a>,
1169 <a href="#t_array">array</a>,
1170 <a href="#t_function">function</a>,
1171 <a href="#t_pointer">pointer</a>,
1172 <a href="#t_struct">structure</a>,
1173 <a href="#t_pstruct">packed structure</a>,
1174 <a href="#t_vector">vector</a>,
1175 <a href="#t_opaque">opaque</a>.
1181 <p>The <a href="#t_firstclass">first class</a> types are perhaps the
1182 most important. Values of these types are the only ones which can be
1183 produced by instructions, passed as arguments, or used as operands to
1187 <!-- ======================================================================= -->
1188 <div class="doc_subsection"> <a name="t_primitive">Primitive Types</a> </div>
1190 <div class="doc_text">
1191 <p>The primitive types are the fundamental building blocks of the LLVM
1196 <!-- _______________________________________________________________________ -->
1197 <div class="doc_subsubsection"> <a name="t_floating">Floating Point Types</a> </div>
1199 <div class="doc_text">
1202 <tr><th>Type</th><th>Description</th></tr>
1203 <tr><td><tt>float</tt></td><td>32-bit floating point value</td></tr>
1204 <tr><td><tt>double</tt></td><td>64-bit floating point value</td></tr>
1205 <tr><td><tt>fp128</tt></td><td>128-bit floating point value (112-bit mantissa)</td></tr>
1206 <tr><td><tt>x86_fp80</tt></td><td>80-bit floating point value (X87)</td></tr>
1207 <tr><td><tt>ppc_fp128</tt></td><td>128-bit floating point value (two 64-bits)</td></tr>
1212 <!-- _______________________________________________________________________ -->
1213 <div class="doc_subsubsection"> <a name="t_void">Void Type</a> </div>
1215 <div class="doc_text">
1217 <p>The void type does not represent any value and has no size.</p>
1226 <!-- _______________________________________________________________________ -->
1227 <div class="doc_subsubsection"> <a name="t_label">Label Type</a> </div>
1229 <div class="doc_text">
1231 <p>The label type represents code labels.</p>
1241 <!-- ======================================================================= -->
1242 <div class="doc_subsection"> <a name="t_derived">Derived Types</a> </div>
1244 <div class="doc_text">
1246 <p>The real power in LLVM comes from the derived types in the system.
1247 This is what allows a programmer to represent arrays, functions,
1248 pointers, and other useful types. Note that these derived types may be
1249 recursive: For example, it is possible to have a two dimensional array.</p>
1253 <!-- _______________________________________________________________________ -->
1254 <div class="doc_subsubsection"> <a name="t_integer">Integer Type</a> </div>
1256 <div class="doc_text">
1259 <p>The integer type is a very simple derived type that simply specifies an
1260 arbitrary bit width for the integer type desired. Any bit width from 1 bit to
1261 2^23-1 (about 8 million) can be specified.</p>
1269 <p>The number of bits the integer will occupy is specified by the <tt>N</tt>
1273 <table class="layout">
1276 <td><tt>i1</tt></td>
1277 <td>a single-bit integer.</td>
1279 <td><tt>i32</tt></td>
1280 <td>a 32-bit integer.</td>
1282 <td><tt>i1942652</tt></td>
1283 <td>a really big integer of over 1 million bits.</td>
1289 <!-- _______________________________________________________________________ -->
1290 <div class="doc_subsubsection"> <a name="t_array">Array Type</a> </div>
1292 <div class="doc_text">
1296 <p>The array type is a very simple derived type that arranges elements
1297 sequentially in memory. The array type requires a size (number of
1298 elements) and an underlying data type.</p>
1303 [<# elements> x <elementtype>]
1306 <p>The number of elements is a constant integer value; elementtype may
1307 be any type with a size.</p>
1310 <table class="layout">
1312 <td class="left"><tt>[40 x i32]</tt></td>
1313 <td class="left">Array of 40 32-bit integer values.</td>
1316 <td class="left"><tt>[41 x i32]</tt></td>
1317 <td class="left">Array of 41 32-bit integer values.</td>
1320 <td class="left"><tt>[4 x i8]</tt></td>
1321 <td class="left">Array of 4 8-bit integer values.</td>
1324 <p>Here are some examples of multidimensional arrays:</p>
1325 <table class="layout">
1327 <td class="left"><tt>[3 x [4 x i32]]</tt></td>
1328 <td class="left">3x4 array of 32-bit integer values.</td>
1331 <td class="left"><tt>[12 x [10 x float]]</tt></td>
1332 <td class="left">12x10 array of single precision floating point values.</td>
1335 <td class="left"><tt>[2 x [3 x [4 x i16]]]</tt></td>
1336 <td class="left">2x3x4 array of 16-bit integer values.</td>
1340 <p>Note that 'variable sized arrays' can be implemented in LLVM with a zero
1341 length array. Normally, accesses past the end of an array are undefined in
1342 LLVM (e.g. it is illegal to access the 5th element of a 3 element array).
1343 As a special case, however, zero length arrays are recognized to be variable
1344 length. This allows implementation of 'pascal style arrays' with the LLVM
1345 type "{ i32, [0 x float]}", for example.</p>
1349 <!-- _______________________________________________________________________ -->
1350 <div class="doc_subsubsection"> <a name="t_function">Function Type</a> </div>
1351 <div class="doc_text">
1355 <p>The function type can be thought of as a function signature. It
1356 consists of a return type and a list of formal parameter types. The
1357 return type of a function type is a scalar type, a void type, or a struct type.
1358 If the return type is a struct type then all struct elements must be of first
1359 class types, and the struct must have at least one element.</p>
1364 <returntype list> (<parameter list>)
1367 <p>...where '<tt><parameter list></tt>' is a comma-separated list of type
1368 specifiers. Optionally, the parameter list may include a type <tt>...</tt>,
1369 which indicates that the function takes a variable number of arguments.
1370 Variable argument functions can access their arguments with the <a
1371 href="#int_varargs">variable argument handling intrinsic</a> functions.
1372 '<tt><returntype list></tt>' is a comma-separated list of
1373 <a href="#t_firstclass">first class</a> type specifiers.</p>
1376 <table class="layout">
1378 <td class="left"><tt>i32 (i32)</tt></td>
1379 <td class="left">function taking an <tt>i32</tt>, returning an <tt>i32</tt>
1381 </tr><tr class="layout">
1382 <td class="left"><tt>float (i16 signext, i32 *) *
1384 <td class="left"><a href="#t_pointer">Pointer</a> to a function that takes
1385 an <tt>i16</tt> that should be sign extended and a
1386 <a href="#t_pointer">pointer</a> to <tt>i32</tt>, returning
1389 </tr><tr class="layout">
1390 <td class="left"><tt>i32 (i8*, ...)</tt></td>
1391 <td class="left">A vararg function that takes at least one
1392 <a href="#t_pointer">pointer</a> to <tt>i8 </tt> (char in C),
1393 which returns an integer. This is the signature for <tt>printf</tt> in
1396 </tr><tr class="layout">
1397 <td class="left"><tt>{i32, i32} (i32)</tt></td>
1398 <td class="left">A function taking an <tt>i32></tt>, returning two
1399 <tt> i32 </tt> values as an aggregate of type <tt>{ i32, i32 }</tt>
1405 <!-- _______________________________________________________________________ -->
1406 <div class="doc_subsubsection"> <a name="t_struct">Structure Type</a> </div>
1407 <div class="doc_text">
1409 <p>The structure type is used to represent a collection of data members
1410 together in memory. The packing of the field types is defined to match
1411 the ABI of the underlying processor. The elements of a structure may
1412 be any type that has a size.</p>
1413 <p>Structures are accessed using '<tt><a href="#i_load">load</a></tt>
1414 and '<tt><a href="#i_store">store</a></tt>' by getting a pointer to a
1415 field with the '<tt><a href="#i_getelementptr">getelementptr</a></tt>'
1418 <pre> { <type list> }<br></pre>
1420 <table class="layout">
1422 <td class="left"><tt>{ i32, i32, i32 }</tt></td>
1423 <td class="left">A triple of three <tt>i32</tt> values</td>
1424 </tr><tr class="layout">
1425 <td class="left"><tt>{ float, i32 (i32) * }</tt></td>
1426 <td class="left">A pair, where the first element is a <tt>float</tt> and the
1427 second element is a <a href="#t_pointer">pointer</a> to a
1428 <a href="#t_function">function</a> that takes an <tt>i32</tt>, returning
1429 an <tt>i32</tt>.</td>
1434 <!-- _______________________________________________________________________ -->
1435 <div class="doc_subsubsection"> <a name="t_pstruct">Packed Structure Type</a>
1437 <div class="doc_text">
1439 <p>The packed structure type is used to represent a collection of data members
1440 together in memory. There is no padding between fields. Further, the alignment
1441 of a packed structure is 1 byte. The elements of a packed structure may
1442 be any type that has a size.</p>
1443 <p>Structures are accessed using '<tt><a href="#i_load">load</a></tt>
1444 and '<tt><a href="#i_store">store</a></tt>' by getting a pointer to a
1445 field with the '<tt><a href="#i_getelementptr">getelementptr</a></tt>'
1448 <pre> < { <type list> } > <br></pre>
1450 <table class="layout">
1452 <td class="left"><tt>< { i32, i32, i32 } ></tt></td>
1453 <td class="left">A triple of three <tt>i32</tt> values</td>
1454 </tr><tr class="layout">
1456 <tt>< { float, i32 (i32)* } ></tt></td>
1457 <td class="left">A pair, where the first element is a <tt>float</tt> and the
1458 second element is a <a href="#t_pointer">pointer</a> to a
1459 <a href="#t_function">function</a> that takes an <tt>i32</tt>, returning
1460 an <tt>i32</tt>.</td>
1465 <!-- _______________________________________________________________________ -->
1466 <div class="doc_subsubsection"> <a name="t_pointer">Pointer Type</a> </div>
1467 <div class="doc_text">
1469 <p>As in many languages, the pointer type represents a pointer or
1470 reference to another object, which must live in memory. Pointer types may have
1471 an optional address space attribute defining the target-specific numbered
1472 address space where the pointed-to object resides. The default address space is
1475 <pre> <type> *<br></pre>
1477 <table class="layout">
1479 <td class="left"><tt>[4x i32]*</tt></td>
1480 <td class="left">A <a href="#t_pointer">pointer</a> to <a
1481 href="#t_array">array</a> of four <tt>i32</tt> values.</td>
1484 <td class="left"><tt>i32 (i32 *) *</tt></td>
1485 <td class="left"> A <a href="#t_pointer">pointer</a> to a <a
1486 href="#t_function">function</a> that takes an <tt>i32*</tt>, returning an
1490 <td class="left"><tt>i32 addrspace(5)*</tt></td>
1491 <td class="left">A <a href="#t_pointer">pointer</a> to an <tt>i32</tt> value
1492 that resides in address space #5.</td>
1497 <!-- _______________________________________________________________________ -->
1498 <div class="doc_subsubsection"> <a name="t_vector">Vector Type</a> </div>
1499 <div class="doc_text">
1503 <p>A vector type is a simple derived type that represents a vector
1504 of elements. Vector types are used when multiple primitive data
1505 are operated in parallel using a single instruction (SIMD).
1506 A vector type requires a size (number of
1507 elements) and an underlying primitive data type. Vectors must have a power
1508 of two length (1, 2, 4, 8, 16 ...). Vector types are
1509 considered <a href="#t_firstclass">first class</a>.</p>
1514 < <# elements> x <elementtype> >
1517 <p>The number of elements is a constant integer value; elementtype may
1518 be any integer or floating point type.</p>
1522 <table class="layout">
1524 <td class="left"><tt><4 x i32></tt></td>
1525 <td class="left">Vector of 4 32-bit integer values.</td>
1528 <td class="left"><tt><8 x float></tt></td>
1529 <td class="left">Vector of 8 32-bit floating-point values.</td>
1532 <td class="left"><tt><2 x i64></tt></td>
1533 <td class="left">Vector of 2 64-bit integer values.</td>
1538 <!-- _______________________________________________________________________ -->
1539 <div class="doc_subsubsection"> <a name="t_opaque">Opaque Type</a> </div>
1540 <div class="doc_text">
1544 <p>Opaque types are used to represent unknown types in the system. This
1545 corresponds (for example) to the C notion of a forward declared structure type.
1546 In LLVM, opaque types can eventually be resolved to any type (not just a
1547 structure type).</p>
1557 <table class="layout">
1559 <td class="left"><tt>opaque</tt></td>
1560 <td class="left">An opaque type.</td>
1566 <!-- *********************************************************************** -->
1567 <div class="doc_section"> <a name="constants">Constants</a> </div>
1568 <!-- *********************************************************************** -->
1570 <div class="doc_text">
1572 <p>LLVM has several different basic types of constants. This section describes
1573 them all and their syntax.</p>
1577 <!-- ======================================================================= -->
1578 <div class="doc_subsection"><a name="simpleconstants">Simple Constants</a></div>
1580 <div class="doc_text">
1583 <dt><b>Boolean constants</b></dt>
1585 <dd>The two strings '<tt>true</tt>' and '<tt>false</tt>' are both valid
1586 constants of the <tt><a href="#t_primitive">i1</a></tt> type.
1589 <dt><b>Integer constants</b></dt>
1591 <dd>Standard integers (such as '4') are constants of the <a
1592 href="#t_integer">integer</a> type. Negative numbers may be used with
1596 <dt><b>Floating point constants</b></dt>
1598 <dd>Floating point constants use standard decimal notation (e.g. 123.421),
1599 exponential notation (e.g. 1.23421e+2), or a more precise hexadecimal
1600 notation (see below). The assembler requires the exact decimal value of
1601 a floating-point constant. For example, the assembler accepts 1.25 but
1602 rejects 1.3 because 1.3 is a repeating decimal in binary. Floating point
1603 constants must have a <a href="#t_floating">floating point</a> type. </dd>
1605 <dt><b>Null pointer constants</b></dt>
1607 <dd>The identifier '<tt>null</tt>' is recognized as a null pointer constant
1608 and must be of <a href="#t_pointer">pointer type</a>.</dd>
1612 <p>The one non-intuitive notation for constants is the optional hexadecimal form
1613 of floating point constants. For example, the form '<tt>double
1614 0x432ff973cafa8000</tt>' is equivalent to (but harder to read than) '<tt>double
1615 4.5e+15</tt>'. The only time hexadecimal floating point constants are required
1616 (and the only time that they are generated by the disassembler) is when a
1617 floating point constant must be emitted but it cannot be represented as a
1618 decimal floating point number. For example, NaN's, infinities, and other
1619 special values are represented in their IEEE hexadecimal format so that
1620 assembly and disassembly do not cause any bits to change in the constants.</p>
1624 <!-- ======================================================================= -->
1625 <div class="doc_subsection"><a name="aggregateconstants">Aggregate Constants</a>
1628 <div class="doc_text">
1629 <p>Aggregate constants arise from aggregation of simple constants
1630 and smaller aggregate constants.</p>
1633 <dt><b>Structure constants</b></dt>
1635 <dd>Structure constants are represented with notation similar to structure
1636 type definitions (a comma separated list of elements, surrounded by braces
1637 (<tt>{}</tt>)). For example: "<tt>{ i32 4, float 17.0, i32* @G }</tt>",
1638 where "<tt>@G</tt>" is declared as "<tt>@G = external global i32</tt>". Structure constants
1639 must have <a href="#t_struct">structure type</a>, and the number and
1640 types of elements must match those specified by the type.
1643 <dt><b>Array constants</b></dt>
1645 <dd>Array constants are represented with notation similar to array type
1646 definitions (a comma separated list of elements, surrounded by square brackets
1647 (<tt>[]</tt>)). For example: "<tt>[ i32 42, i32 11, i32 74 ]</tt>". Array
1648 constants must have <a href="#t_array">array type</a>, and the number and
1649 types of elements must match those specified by the type.
1652 <dt><b>Vector constants</b></dt>
1654 <dd>Vector constants are represented with notation similar to vector type
1655 definitions (a comma separated list of elements, surrounded by
1656 less-than/greater-than's (<tt><></tt>)). For example: "<tt>< i32 42,
1657 i32 11, i32 74, i32 100 ></tt>". Vector constants must have <a
1658 href="#t_vector">vector type</a>, and the number and types of elements must
1659 match those specified by the type.
1662 <dt><b>Zero initialization</b></dt>
1664 <dd>The string '<tt>zeroinitializer</tt>' can be used to zero initialize a
1665 value to zero of <em>any</em> type, including scalar and aggregate types.
1666 This is often used to avoid having to print large zero initializers (e.g. for
1667 large arrays) and is always exactly equivalent to using explicit zero
1674 <!-- ======================================================================= -->
1675 <div class="doc_subsection">
1676 <a name="globalconstants">Global Variable and Function Addresses</a>
1679 <div class="doc_text">
1681 <p>The addresses of <a href="#globalvars">global variables</a> and <a
1682 href="#functionstructure">functions</a> are always implicitly valid (link-time)
1683 constants. These constants are explicitly referenced when the <a
1684 href="#identifiers">identifier for the global</a> is used and always have <a
1685 href="#t_pointer">pointer</a> type. For example, the following is a legal LLVM
1688 <div class="doc_code">
1692 @Z = global [2 x i32*] [ i32* @X, i32* @Y ]
1698 <!-- ======================================================================= -->
1699 <div class="doc_subsection"><a name="undefvalues">Undefined Values</a></div>
1700 <div class="doc_text">
1701 <p>The string '<tt>undef</tt>' is recognized as a type-less constant that has
1702 no specific value. Undefined values may be of any type and be used anywhere
1703 a constant is permitted.</p>
1705 <p>Undefined values indicate to the compiler that the program is well defined
1706 no matter what value is used, giving the compiler more freedom to optimize.
1710 <!-- ======================================================================= -->
1711 <div class="doc_subsection"><a name="constantexprs">Constant Expressions</a>
1714 <div class="doc_text">
1716 <p>Constant expressions are used to allow expressions involving other constants
1717 to be used as constants. Constant expressions may be of any <a
1718 href="#t_firstclass">first class</a> type and may involve any LLVM operation
1719 that does not have side effects (e.g. load and call are not supported). The
1720 following is the syntax for constant expressions:</p>
1723 <dt><b><tt>trunc ( CST to TYPE )</tt></b></dt>
1724 <dd>Truncate a constant to another type. The bit size of CST must be larger
1725 than the bit size of TYPE. Both types must be integers.</dd>
1727 <dt><b><tt>zext ( CST to TYPE )</tt></b></dt>
1728 <dd>Zero extend a constant to another type. The bit size of CST must be
1729 smaller or equal to the bit size of TYPE. Both types must be integers.</dd>
1731 <dt><b><tt>sext ( CST to TYPE )</tt></b></dt>
1732 <dd>Sign extend a constant to another type. The bit size of CST must be
1733 smaller or equal to the bit size of TYPE. Both types must be integers.</dd>
1735 <dt><b><tt>fptrunc ( CST to TYPE )</tt></b></dt>
1736 <dd>Truncate a floating point constant to another floating point type. The
1737 size of CST must be larger than the size of TYPE. Both types must be
1738 floating point.</dd>
1740 <dt><b><tt>fpext ( CST to TYPE )</tt></b></dt>
1741 <dd>Floating point extend a constant to another type. The size of CST must be
1742 smaller or equal to the size of TYPE. Both types must be floating point.</dd>
1744 <dt><b><tt>fptoui ( CST to TYPE )</tt></b></dt>
1745 <dd>Convert a floating point constant to the corresponding unsigned integer
1746 constant. TYPE must be a scalar or vector integer type. CST must be of scalar
1747 or vector floating point type. Both CST and TYPE must be scalars, or vectors
1748 of the same number of elements. If the value won't fit in the integer type,
1749 the results are undefined.</dd>
1751 <dt><b><tt>fptosi ( CST to TYPE )</tt></b></dt>
1752 <dd>Convert a floating point constant to the corresponding signed integer
1753 constant. TYPE must be a scalar or vector integer type. CST must be of scalar
1754 or vector floating point type. Both CST and TYPE must be scalars, or vectors
1755 of the same number of elements. If the value won't fit in the integer type,
1756 the results are undefined.</dd>
1758 <dt><b><tt>uitofp ( CST to TYPE )</tt></b></dt>
1759 <dd>Convert an unsigned integer constant to the corresponding floating point
1760 constant. TYPE must be a scalar or vector floating point type. CST must be of
1761 scalar or vector integer type. Both CST and TYPE must be scalars, or vectors
1762 of the same number of elements. If the value won't fit in the floating point
1763 type, the results are undefined.</dd>
1765 <dt><b><tt>sitofp ( CST to TYPE )</tt></b></dt>
1766 <dd>Convert a signed integer constant to the corresponding floating point
1767 constant. TYPE must be a scalar or vector floating point type. CST must be of
1768 scalar or vector integer type. Both CST and TYPE must be scalars, or vectors
1769 of the same number of elements. If the value won't fit in the floating point
1770 type, the results are undefined.</dd>
1772 <dt><b><tt>ptrtoint ( CST to TYPE )</tt></b></dt>
1773 <dd>Convert a pointer typed constant to the corresponding integer constant
1774 TYPE must be an integer type. CST must be of pointer type. The CST value is
1775 zero extended, truncated, or unchanged to make it fit in TYPE.</dd>
1777 <dt><b><tt>inttoptr ( CST to TYPE )</tt></b></dt>
1778 <dd>Convert a integer constant to a pointer constant. TYPE must be a
1779 pointer type. CST must be of integer type. The CST value is zero extended,
1780 truncated, or unchanged to make it fit in a pointer size. This one is
1781 <i>really</i> dangerous!</dd>
1783 <dt><b><tt>bitcast ( CST to TYPE )</tt></b></dt>
1784 <dd>Convert a constant, CST, to another TYPE. The size of CST and TYPE must be
1785 identical (same number of bits). The conversion is done as if the CST value
1786 was stored to memory and read back as TYPE. In other words, no bits change
1787 with this operator, just the type. This can be used for conversion of
1788 vector types to any other type, as long as they have the same bit width. For
1789 pointers it is only valid to cast to another pointer type. It is not valid
1790 to bitcast to or from an aggregate type.
1793 <dt><b><tt>getelementptr ( CSTPTR, IDX0, IDX1, ... )</tt></b></dt>
1795 <dd>Perform the <a href="#i_getelementptr">getelementptr operation</a> on
1796 constants. As with the <a href="#i_getelementptr">getelementptr</a>
1797 instruction, the index list may have zero or more indexes, which are required
1798 to make sense for the type of "CSTPTR".</dd>
1800 <dt><b><tt>select ( COND, VAL1, VAL2 )</tt></b></dt>
1802 <dd>Perform the <a href="#i_select">select operation</a> on
1805 <dt><b><tt>icmp COND ( VAL1, VAL2 )</tt></b></dt>
1806 <dd>Performs the <a href="#i_icmp">icmp operation</a> on constants.</dd>
1808 <dt><b><tt>fcmp COND ( VAL1, VAL2 )</tt></b></dt>
1809 <dd>Performs the <a href="#i_fcmp">fcmp operation</a> on constants.</dd>
1811 <dt><b><tt>vicmp COND ( VAL1, VAL2 )</tt></b></dt>
1812 <dd>Performs the <a href="#i_vicmp">vicmp operation</a> on constants.</dd>
1814 <dt><b><tt>vfcmp COND ( VAL1, VAL2 )</tt></b></dt>
1815 <dd>Performs the <a href="#i_vfcmp">vfcmp operation</a> on constants.</dd>
1817 <dt><b><tt>extractelement ( VAL, IDX )</tt></b></dt>
1819 <dd>Perform the <a href="#i_extractelement">extractelement
1820 operation</a> on constants.</dd>
1822 <dt><b><tt>insertelement ( VAL, ELT, IDX )</tt></b></dt>
1824 <dd>Perform the <a href="#i_insertelement">insertelement
1825 operation</a> on constants.</dd>
1828 <dt><b><tt>shufflevector ( VEC1, VEC2, IDXMASK )</tt></b></dt>
1830 <dd>Perform the <a href="#i_shufflevector">shufflevector
1831 operation</a> on constants.</dd>
1833 <dt><b><tt>OPCODE ( LHS, RHS )</tt></b></dt>
1835 <dd>Perform the specified operation of the LHS and RHS constants. OPCODE may
1836 be any of the <a href="#binaryops">binary</a> or <a href="#bitwiseops">bitwise
1837 binary</a> operations. The constraints on operands are the same as those for
1838 the corresponding instruction (e.g. no bitwise operations on floating point
1839 values are allowed).</dd>
1843 <!-- *********************************************************************** -->
1844 <div class="doc_section"> <a name="othervalues">Other Values</a> </div>
1845 <!-- *********************************************************************** -->
1847 <!-- ======================================================================= -->
1848 <div class="doc_subsection">
1849 <a name="inlineasm">Inline Assembler Expressions</a>
1852 <div class="doc_text">
1855 LLVM supports inline assembler expressions (as opposed to <a href="#moduleasm">
1856 Module-Level Inline Assembly</a>) through the use of a special value. This
1857 value represents the inline assembler as a string (containing the instructions
1858 to emit), a list of operand constraints (stored as a string), and a flag that
1859 indicates whether or not the inline asm expression has side effects. An example
1860 inline assembler expression is:
1863 <div class="doc_code">
1865 i32 (i32) asm "bswap $0", "=r,r"
1870 Inline assembler expressions may <b>only</b> be used as the callee operand of
1871 a <a href="#i_call"><tt>call</tt> instruction</a>. Thus, typically we have:
1874 <div class="doc_code">
1876 %X = call i32 asm "<a href="#int_bswap">bswap</a> $0", "=r,r"(i32 %Y)
1881 Inline asms with side effects not visible in the constraint list must be marked
1882 as having side effects. This is done through the use of the
1883 '<tt>sideeffect</tt>' keyword, like so:
1886 <div class="doc_code">
1888 call void asm sideeffect "eieio", ""()
1892 <p>TODO: The format of the asm and constraints string still need to be
1893 documented here. Constraints on what can be done (e.g. duplication, moving, etc
1894 need to be documented). This is probably best done by reference to another
1895 document that covers inline asm from a holistic perspective.
1900 <!-- *********************************************************************** -->
1901 <div class="doc_section"> <a name="instref">Instruction Reference</a> </div>
1902 <!-- *********************************************************************** -->
1904 <div class="doc_text">
1906 <p>The LLVM instruction set consists of several different
1907 classifications of instructions: <a href="#terminators">terminator
1908 instructions</a>, <a href="#binaryops">binary instructions</a>,
1909 <a href="#bitwiseops">bitwise binary instructions</a>, <a
1910 href="#memoryops">memory instructions</a>, and <a href="#otherops">other
1911 instructions</a>.</p>
1915 <!-- ======================================================================= -->
1916 <div class="doc_subsection"> <a name="terminators">Terminator
1917 Instructions</a> </div>
1919 <div class="doc_text">
1921 <p>As mentioned <a href="#functionstructure">previously</a>, every
1922 basic block in a program ends with a "Terminator" instruction, which
1923 indicates which block should be executed after the current block is
1924 finished. These terminator instructions typically yield a '<tt>void</tt>'
1925 value: they produce control flow, not values (the one exception being
1926 the '<a href="#i_invoke"><tt>invoke</tt></a>' instruction).</p>
1927 <p>There are six different terminator instructions: the '<a
1928 href="#i_ret"><tt>ret</tt></a>' instruction, the '<a href="#i_br"><tt>br</tt></a>'
1929 instruction, the '<a href="#i_switch"><tt>switch</tt></a>' instruction,
1930 the '<a href="#i_invoke"><tt>invoke</tt></a>' instruction, the '<a
1931 href="#i_unwind"><tt>unwind</tt></a>' instruction, and the '<a
1932 href="#i_unreachable"><tt>unreachable</tt></a>' instruction.</p>
1936 <!-- _______________________________________________________________________ -->
1937 <div class="doc_subsubsection"> <a name="i_ret">'<tt>ret</tt>'
1938 Instruction</a> </div>
1939 <div class="doc_text">
1942 ret <type> <value> <i>; Return a value from a non-void function</i>
1943 ret void <i>; Return from void function</i>
1948 <p>The '<tt>ret</tt>' instruction is used to return control flow (and
1949 optionally a value) from a function back to the caller.</p>
1950 <p>There are two forms of the '<tt>ret</tt>' instruction: one that
1951 returns a value and then causes control flow, and one that just causes
1952 control flow to occur.</p>
1956 <p>The '<tt>ret</tt>' instruction optionally accepts a single argument,
1957 the return value. The type of the return value must be a
1958 '<a href="#t_firstclass">first class</a>' type.</p>
1960 <p>A function is not <a href="#wellformed">well formed</a> if
1961 it it has a non-void return type and contains a '<tt>ret</tt>'
1962 instruction with no return value or a return value with a type that
1963 does not match its type, or if it has a void return type and contains
1964 a '<tt>ret</tt>' instruction with a return value.</p>
1968 <p>When the '<tt>ret</tt>' instruction is executed, control flow
1969 returns back to the calling function's context. If the caller is a "<a
1970 href="#i_call"><tt>call</tt></a>" instruction, execution continues at
1971 the instruction after the call. If the caller was an "<a
1972 href="#i_invoke"><tt>invoke</tt></a>" instruction, execution continues
1973 at the beginning of the "normal" destination block. If the instruction
1974 returns a value, that value shall set the call or invoke instruction's
1980 ret i32 5 <i>; Return an integer value of 5</i>
1981 ret void <i>; Return from a void function</i>
1982 ret { i32, i8 } { i32 4, i8 2 } <i>; Return an aggregate of values 4 and 2</i>
1985 <!-- _______________________________________________________________________ -->
1986 <div class="doc_subsubsection"> <a name="i_br">'<tt>br</tt>' Instruction</a> </div>
1987 <div class="doc_text">
1989 <pre> br i1 <cond>, label <iftrue>, label <iffalse><br> br label <dest> <i>; Unconditional branch</i>
1992 <p>The '<tt>br</tt>' instruction is used to cause control flow to
1993 transfer to a different basic block in the current function. There are
1994 two forms of this instruction, corresponding to a conditional branch
1995 and an unconditional branch.</p>
1997 <p>The conditional branch form of the '<tt>br</tt>' instruction takes a
1998 single '<tt>i1</tt>' value and two '<tt>label</tt>' values. The
1999 unconditional form of the '<tt>br</tt>' instruction takes a single
2000 '<tt>label</tt>' value as a target.</p>
2002 <p>Upon execution of a conditional '<tt>br</tt>' instruction, the '<tt>i1</tt>'
2003 argument is evaluated. If the value is <tt>true</tt>, control flows
2004 to the '<tt>iftrue</tt>' <tt>label</tt> argument. If "cond" is <tt>false</tt>,
2005 control flows to the '<tt>iffalse</tt>' <tt>label</tt> argument.</p>
2007 <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
2008 href="#i_ret">ret</a> i32 1<br>IfUnequal:<br> <a href="#i_ret">ret</a> i32 0<br></pre>
2010 <!-- _______________________________________________________________________ -->
2011 <div class="doc_subsubsection">
2012 <a name="i_switch">'<tt>switch</tt>' Instruction</a>
2015 <div class="doc_text">
2019 switch <intty> <value>, label <defaultdest> [ <intty> <val>, label <dest> ... ]
2024 <p>The '<tt>switch</tt>' instruction is used to transfer control flow to one of
2025 several different places. It is a generalization of the '<tt>br</tt>'
2026 instruction, allowing a branch to occur to one of many possible
2032 <p>The '<tt>switch</tt>' instruction uses three parameters: an integer
2033 comparison value '<tt>value</tt>', a default '<tt>label</tt>' destination, and
2034 an array of pairs of comparison value constants and '<tt>label</tt>'s. The
2035 table is not allowed to contain duplicate constant entries.</p>
2039 <p>The <tt>switch</tt> instruction specifies a table of values and
2040 destinations. When the '<tt>switch</tt>' instruction is executed, this
2041 table is searched for the given value. If the value is found, control flow is
2042 transfered to the corresponding destination; otherwise, control flow is
2043 transfered to the default destination.</p>
2045 <h5>Implementation:</h5>
2047 <p>Depending on properties of the target machine and the particular
2048 <tt>switch</tt> instruction, this instruction may be code generated in different
2049 ways. For example, it could be generated as a series of chained conditional
2050 branches or with a lookup table.</p>
2055 <i>; Emulate a conditional br instruction</i>
2056 %Val = <a href="#i_zext">zext</a> i1 %value to i32
2057 switch i32 %Val, label %truedest [i32 0, label %falsedest ]
2059 <i>; Emulate an unconditional br instruction</i>
2060 switch i32 0, label %dest [ ]
2062 <i>; Implement a jump table:</i>
2063 switch i32 %val, label %otherwise [ i32 0, label %onzero
2065 i32 2, label %ontwo ]
2069 <!-- _______________________________________________________________________ -->
2070 <div class="doc_subsubsection">
2071 <a name="i_invoke">'<tt>invoke</tt>' Instruction</a>
2074 <div class="doc_text">
2079 <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>]
2080 to label <normal label> unwind label <exception label>
2085 <p>The '<tt>invoke</tt>' instruction causes control to transfer to a specified
2086 function, with the possibility of control flow transfer to either the
2087 '<tt>normal</tt>' label or the
2088 '<tt>exception</tt>' label. If the callee function returns with the
2089 "<tt><a href="#i_ret">ret</a></tt>" instruction, control flow will return to the
2090 "normal" label. If the callee (or any indirect callees) returns with the "<a
2091 href="#i_unwind"><tt>unwind</tt></a>" instruction, control is interrupted and
2092 continued at the dynamically nearest "exception" label.</p>
2096 <p>This instruction requires several arguments:</p>
2100 The optional "cconv" marker indicates which <a href="#callingconv">calling
2101 convention</a> the call should use. If none is specified, the call defaults
2102 to using C calling conventions.
2105 <li>The optional <a href="#paramattrs">Parameter Attributes</a> list for
2106 return values. Only '<tt>zeroext</tt>', '<tt>signext</tt>',
2107 and '<tt>inreg</tt>' attributes are valid here.</li>
2109 <li>'<tt>ptr to function ty</tt>': shall be the signature of the pointer to
2110 function value being invoked. In most cases, this is a direct function
2111 invocation, but indirect <tt>invoke</tt>s are just as possible, branching off
2112 an arbitrary pointer to function value.
2115 <li>'<tt>function ptr val</tt>': An LLVM value containing a pointer to a
2116 function to be invoked. </li>
2118 <li>'<tt>function args</tt>': argument list whose types match the function
2119 signature argument types. If the function signature indicates the function
2120 accepts a variable number of arguments, the extra arguments can be
2123 <li>'<tt>normal label</tt>': the label reached when the called function
2124 executes a '<tt><a href="#i_ret">ret</a></tt>' instruction. </li>
2126 <li>'<tt>exception label</tt>': the label reached when a callee returns with
2127 the <a href="#i_unwind"><tt>unwind</tt></a> instruction. </li>
2129 <li>The optional <a href="#fnattrs">function attributes</a> list. Only
2130 '<tt>noreturn</tt>', '<tt>nounwind</tt>', '<tt>readonly</tt>' and
2131 '<tt>readnone</tt>' attributes are valid here.</li>
2136 <p>This instruction is designed to operate as a standard '<tt><a
2137 href="#i_call">call</a></tt>' instruction in most regards. The primary
2138 difference is that it establishes an association with a label, which is used by
2139 the runtime library to unwind the stack.</p>
2141 <p>This instruction is used in languages with destructors to ensure that proper
2142 cleanup is performed in the case of either a <tt>longjmp</tt> or a thrown
2143 exception. Additionally, this is important for implementation of
2144 '<tt>catch</tt>' clauses in high-level languages that support them.</p>
2148 %retval = invoke i32 @Test(i32 15) to label %Continue
2149 unwind label %TestCleanup <i>; {i32}:retval set</i>
2150 %retval = invoke <a href="#callingconv">coldcc</a> i32 %Testfnptr(i32 15) to label %Continue
2151 unwind label %TestCleanup <i>; {i32}:retval set</i>
2156 <!-- _______________________________________________________________________ -->
2158 <div class="doc_subsubsection"> <a name="i_unwind">'<tt>unwind</tt>'
2159 Instruction</a> </div>
2161 <div class="doc_text">
2170 <p>The '<tt>unwind</tt>' instruction unwinds the stack, continuing control flow
2171 at the first callee in the dynamic call stack which used an <a
2172 href="#i_invoke"><tt>invoke</tt></a> instruction to perform the call. This is
2173 primarily used to implement exception handling.</p>
2177 <p>The '<tt>unwind</tt>' instruction causes execution of the current function to
2178 immediately halt. The dynamic call stack is then searched for the first <a
2179 href="#i_invoke"><tt>invoke</tt></a> instruction on the call stack. Once found,
2180 execution continues at the "exceptional" destination block specified by the
2181 <tt>invoke</tt> instruction. If there is no <tt>invoke</tt> instruction in the
2182 dynamic call chain, undefined behavior results.</p>
2185 <!-- _______________________________________________________________________ -->
2187 <div class="doc_subsubsection"> <a name="i_unreachable">'<tt>unreachable</tt>'
2188 Instruction</a> </div>
2190 <div class="doc_text">
2199 <p>The '<tt>unreachable</tt>' instruction has no defined semantics. This
2200 instruction is used to inform the optimizer that a particular portion of the
2201 code is not reachable. This can be used to indicate that the code after a
2202 no-return function cannot be reached, and other facts.</p>
2206 <p>The '<tt>unreachable</tt>' instruction has no defined semantics.</p>
2211 <!-- ======================================================================= -->
2212 <div class="doc_subsection"> <a name="binaryops">Binary Operations</a> </div>
2213 <div class="doc_text">
2214 <p>Binary operators are used to do most of the computation in a
2215 program. They require two operands of the same type, execute an operation on them, and
2216 produce a single value. The operands might represent
2217 multiple data, as is the case with the <a href="#t_vector">vector</a> data type.
2218 The result value has the same type as its operands.</p>
2219 <p>There are several different binary operators:</p>
2221 <!-- _______________________________________________________________________ -->
2222 <div class="doc_subsubsection">
2223 <a name="i_add">'<tt>add</tt>' Instruction</a>
2226 <div class="doc_text">
2231 <result> = add <ty> <op1>, <op2> <i>; yields {ty}:result</i>
2236 <p>The '<tt>add</tt>' instruction returns the sum of its two operands.</p>
2240 <p>The two arguments to the '<tt>add</tt>' instruction must be <a
2241 href="#t_integer">integer</a>, <a href="#t_floating">floating point</a>, or
2242 <a href="#t_vector">vector</a> values. Both arguments must have identical
2247 <p>The value produced is the integer or floating point sum of the two
2250 <p>If an integer sum has unsigned overflow, the result returned is the
2251 mathematical result modulo 2<sup>n</sup>, where n is the bit width of
2254 <p>Because LLVM integers use a two's complement representation, this
2255 instruction is appropriate for both signed and unsigned integers.</p>
2260 <result> = add i32 4, %var <i>; yields {i32}:result = 4 + %var</i>
2263 <!-- _______________________________________________________________________ -->
2264 <div class="doc_subsubsection">
2265 <a name="i_sub">'<tt>sub</tt>' Instruction</a>
2268 <div class="doc_text">
2273 <result> = sub <ty> <op1>, <op2> <i>; yields {ty}:result</i>
2278 <p>The '<tt>sub</tt>' instruction returns the difference of its two
2281 <p>Note that the '<tt>sub</tt>' instruction is used to represent the
2282 '<tt>neg</tt>' instruction present in most other intermediate
2283 representations.</p>
2287 <p>The two arguments to the '<tt>sub</tt>' instruction must be <a
2288 href="#t_integer">integer</a>, <a href="#t_floating">floating point</a>,
2289 or <a href="#t_vector">vector</a> values. Both arguments must have identical
2294 <p>The value produced is the integer or floating point difference of
2295 the two operands.</p>
2297 <p>If an integer difference has unsigned overflow, the result returned is the
2298 mathematical result modulo 2<sup>n</sup>, where n is the bit width of
2301 <p>Because LLVM integers use a two's complement representation, this
2302 instruction is appropriate for both signed and unsigned integers.</p>
2306 <result> = sub i32 4, %var <i>; yields {i32}:result = 4 - %var</i>
2307 <result> = sub i32 0, %val <i>; yields {i32}:result = -%var</i>
2311 <!-- _______________________________________________________________________ -->
2312 <div class="doc_subsubsection">
2313 <a name="i_mul">'<tt>mul</tt>' Instruction</a>
2316 <div class="doc_text">
2319 <pre> <result> = mul <ty> <op1>, <op2> <i>; yields {ty}:result</i>
2322 <p>The '<tt>mul</tt>' instruction returns the product of its two
2327 <p>The two arguments to the '<tt>mul</tt>' instruction must be <a
2328 href="#t_integer">integer</a>, <a href="#t_floating">floating point</a>,
2329 or <a href="#t_vector">vector</a> values. Both arguments must have identical
2334 <p>The value produced is the integer or floating point product of the
2337 <p>If the result of an integer multiplication has unsigned overflow,
2338 the result returned is the mathematical result modulo
2339 2<sup>n</sup>, where n is the bit width of the result.</p>
2340 <p>Because LLVM integers use a two's complement representation, and the
2341 result is the same width as the operands, this instruction returns the
2342 correct result for both signed and unsigned integers. If a full product
2343 (e.g. <tt>i32</tt>x<tt>i32</tt>-><tt>i64</tt>) is needed, the operands
2344 should be sign-extended or zero-extended as appropriate to the
2345 width of the full product.</p>
2347 <pre> <result> = mul i32 4, %var <i>; yields {i32}:result = 4 * %var</i>
2351 <!-- _______________________________________________________________________ -->
2352 <div class="doc_subsubsection"> <a name="i_udiv">'<tt>udiv</tt>' Instruction
2354 <div class="doc_text">
2356 <pre> <result> = udiv <ty> <op1>, <op2> <i>; yields {ty}:result</i>
2359 <p>The '<tt>udiv</tt>' instruction returns the quotient of its two
2364 <p>The two arguments to the '<tt>udiv</tt>' instruction must be
2365 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
2366 values. Both arguments must have identical types.</p>
2370 <p>The value produced is the unsigned integer quotient of the two operands.</p>
2371 <p>Note that unsigned integer division and signed integer division are distinct
2372 operations; for signed integer division, use '<tt>sdiv</tt>'.</p>
2373 <p>Division by zero leads to undefined behavior.</p>
2375 <pre> <result> = udiv i32 4, %var <i>; yields {i32}:result = 4 / %var</i>
2378 <!-- _______________________________________________________________________ -->
2379 <div class="doc_subsubsection"> <a name="i_sdiv">'<tt>sdiv</tt>' Instruction
2381 <div class="doc_text">
2384 <result> = sdiv <ty> <op1>, <op2> <i>; yields {ty}:result</i>
2389 <p>The '<tt>sdiv</tt>' instruction returns the quotient of its two
2394 <p>The two arguments to the '<tt>sdiv</tt>' instruction must be
2395 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
2396 values. Both arguments must have identical types.</p>
2399 <p>The value produced is the signed integer quotient of the two operands rounded towards zero.</p>
2400 <p>Note that signed integer division and unsigned integer division are distinct
2401 operations; for unsigned integer division, use '<tt>udiv</tt>'.</p>
2402 <p>Division by zero leads to undefined behavior. Overflow also leads to
2403 undefined behavior; this is a rare case, but can occur, for example,
2404 by doing a 32-bit division of -2147483648 by -1.</p>
2406 <pre> <result> = sdiv i32 4, %var <i>; yields {i32}:result = 4 / %var</i>
2409 <!-- _______________________________________________________________________ -->
2410 <div class="doc_subsubsection"> <a name="i_fdiv">'<tt>fdiv</tt>'
2411 Instruction</a> </div>
2412 <div class="doc_text">
2415 <result> = fdiv <ty> <op1>, <op2> <i>; yields {ty}:result</i>
2419 <p>The '<tt>fdiv</tt>' instruction returns the quotient of its two
2424 <p>The two arguments to the '<tt>fdiv</tt>' instruction must be
2425 <a href="#t_floating">floating point</a> or <a href="#t_vector">vector</a>
2426 of floating point values. Both arguments must have identical types.</p>
2430 <p>The value produced is the floating point quotient of the two operands.</p>
2435 <result> = fdiv float 4.0, %var <i>; yields {float}:result = 4.0 / %var</i>
2439 <!-- _______________________________________________________________________ -->
2440 <div class="doc_subsubsection"> <a name="i_urem">'<tt>urem</tt>' Instruction</a>
2442 <div class="doc_text">
2444 <pre> <result> = urem <ty> <op1>, <op2> <i>; yields {ty}:result</i>
2447 <p>The '<tt>urem</tt>' instruction returns the remainder from the
2448 unsigned division of its two arguments.</p>
2450 <p>The two arguments to the '<tt>urem</tt>' instruction must be
2451 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
2452 values. Both arguments must have identical types.</p>
2454 <p>This instruction returns the unsigned integer <i>remainder</i> of a division.
2455 This instruction always performs an unsigned division to get the remainder.</p>
2456 <p>Note that unsigned integer remainder and signed integer remainder are
2457 distinct operations; for signed integer remainder, use '<tt>srem</tt>'.</p>
2458 <p>Taking the remainder of a division by zero leads to undefined behavior.</p>
2460 <pre> <result> = urem i32 4, %var <i>; yields {i32}:result = 4 % %var</i>
2464 <!-- _______________________________________________________________________ -->
2465 <div class="doc_subsubsection">
2466 <a name="i_srem">'<tt>srem</tt>' Instruction</a>
2469 <div class="doc_text">
2474 <result> = srem <ty> <op1>, <op2> <i>; yields {ty}:result</i>
2479 <p>The '<tt>srem</tt>' instruction returns the remainder from the
2480 signed division of its two operands. This instruction can also take
2481 <a href="#t_vector">vector</a> versions of the values in which case
2482 the elements must be integers.</p>
2486 <p>The two arguments to the '<tt>srem</tt>' instruction must be
2487 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
2488 values. Both arguments must have identical types.</p>
2492 <p>This instruction returns the <i>remainder</i> of a division (where the result
2493 has the same sign as the dividend, <tt>op1</tt>), not the <i>modulo</i>
2494 operator (where the result has the same sign as the divisor, <tt>op2</tt>) of
2495 a value. For more information about the difference, see <a
2496 href="http://mathforum.org/dr.math/problems/anne.4.28.99.html">The
2497 Math Forum</a>. For a table of how this is implemented in various languages,
2498 please see <a href="http://en.wikipedia.org/wiki/Modulo_operation">
2499 Wikipedia: modulo operation</a>.</p>
2500 <p>Note that signed integer remainder and unsigned integer remainder are
2501 distinct operations; for unsigned integer remainder, use '<tt>urem</tt>'.</p>
2502 <p>Taking the remainder of a division by zero leads to undefined behavior.
2503 Overflow also leads to undefined behavior; this is a rare case, but can occur,
2504 for example, by taking the remainder of a 32-bit division of -2147483648 by -1.
2505 (The remainder doesn't actually overflow, but this rule lets srem be
2506 implemented using instructions that return both the result of the division
2507 and the remainder.)</p>
2509 <pre> <result> = srem i32 4, %var <i>; yields {i32}:result = 4 % %var</i>
2513 <!-- _______________________________________________________________________ -->
2514 <div class="doc_subsubsection">
2515 <a name="i_frem">'<tt>frem</tt>' Instruction</a> </div>
2517 <div class="doc_text">
2520 <pre> <result> = frem <ty> <op1>, <op2> <i>; yields {ty}:result</i>
2523 <p>The '<tt>frem</tt>' instruction returns the remainder from the
2524 division of its two operands.</p>
2526 <p>The two arguments to the '<tt>frem</tt>' instruction must be
2527 <a href="#t_floating">floating point</a> or <a href="#t_vector">vector</a>
2528 of floating point values. Both arguments must have identical types.</p>
2532 <p>This instruction returns the <i>remainder</i> of a division.
2533 The remainder has the same sign as the dividend.</p>
2538 <result> = frem float 4.0, %var <i>; yields {float}:result = 4.0 % %var</i>
2542 <!-- ======================================================================= -->
2543 <div class="doc_subsection"> <a name="bitwiseops">Bitwise Binary
2544 Operations</a> </div>
2545 <div class="doc_text">
2546 <p>Bitwise binary operators are used to do various forms of
2547 bit-twiddling in a program. They are generally very efficient
2548 instructions and can commonly be strength reduced from other
2549 instructions. They require two operands of the same type, execute an operation on them,
2550 and produce a single value. The resulting value is the same type as its operands.</p>
2553 <!-- _______________________________________________________________________ -->
2554 <div class="doc_subsubsection"> <a name="i_shl">'<tt>shl</tt>'
2555 Instruction</a> </div>
2556 <div class="doc_text">
2558 <pre> <result> = shl <ty> <op1>, <op2> <i>; yields {ty}:result</i>
2563 <p>The '<tt>shl</tt>' instruction returns the first operand shifted to
2564 the left a specified number of bits.</p>
2568 <p>Both arguments to the '<tt>shl</tt>' instruction must be the same <a
2569 href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
2570 type. '<tt>op2</tt>' is treated as an unsigned value.</p>
2574 <p>The value produced is <tt>op1</tt> * 2<sup><tt>op2</tt></sup> mod 2<sup>n</sup>,
2575 where n is the width of the result. If <tt>op2</tt> is (statically or dynamically) negative or
2576 equal to or larger than the number of bits in <tt>op1</tt>, the result is undefined.</p>
2578 <h5>Example:</h5><pre>
2579 <result> = shl i32 4, %var <i>; yields {i32}: 4 << %var</i>
2580 <result> = shl i32 4, 2 <i>; yields {i32}: 16</i>
2581 <result> = shl i32 1, 10 <i>; yields {i32}: 1024</i>
2582 <result> = shl i32 1, 32 <i>; undefined</i>
2585 <!-- _______________________________________________________________________ -->
2586 <div class="doc_subsubsection"> <a name="i_lshr">'<tt>lshr</tt>'
2587 Instruction</a> </div>
2588 <div class="doc_text">
2590 <pre> <result> = lshr <ty> <op1>, <op2> <i>; yields {ty}:result</i>
2594 <p>The '<tt>lshr</tt>' instruction (logical shift right) returns the first
2595 operand shifted to the right a specified number of bits with zero fill.</p>
2598 <p>Both arguments to the '<tt>lshr</tt>' instruction must be the same
2599 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
2600 type. '<tt>op2</tt>' is treated as an unsigned value.</p>
2604 <p>This instruction always performs a logical shift right operation. The most
2605 significant bits of the result will be filled with zero bits after the
2606 shift. If <tt>op2</tt> is (statically or dynamically) equal to or larger than
2607 the number of bits in <tt>op1</tt>, the result is undefined.</p>
2611 <result> = lshr i32 4, 1 <i>; yields {i32}:result = 2</i>
2612 <result> = lshr i32 4, 2 <i>; yields {i32}:result = 1</i>
2613 <result> = lshr i8 4, 3 <i>; yields {i8}:result = 0</i>
2614 <result> = lshr i8 -2, 1 <i>; yields {i8}:result = 0x7FFFFFFF </i>
2615 <result> = lshr i32 1, 32 <i>; undefined</i>
2619 <!-- _______________________________________________________________________ -->
2620 <div class="doc_subsubsection"> <a name="i_ashr">'<tt>ashr</tt>'
2621 Instruction</a> </div>
2622 <div class="doc_text">
2625 <pre> <result> = ashr <ty> <op1>, <op2> <i>; yields {ty}:result</i>
2629 <p>The '<tt>ashr</tt>' instruction (arithmetic shift right) returns the first
2630 operand shifted to the right a specified number of bits with sign extension.</p>
2633 <p>Both arguments to the '<tt>ashr</tt>' instruction must be the same
2634 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
2635 type. '<tt>op2</tt>' is treated as an unsigned value.</p>
2638 <p>This instruction always performs an arithmetic shift right operation,
2639 The most significant bits of the result will be filled with the sign bit
2640 of <tt>op1</tt>. If <tt>op2</tt> is (statically or dynamically) equal to or
2641 larger than the number of bits in <tt>op1</tt>, the result is undefined.
2646 <result> = ashr i32 4, 1 <i>; yields {i32}:result = 2</i>
2647 <result> = ashr i32 4, 2 <i>; yields {i32}:result = 1</i>
2648 <result> = ashr i8 4, 3 <i>; yields {i8}:result = 0</i>
2649 <result> = ashr i8 -2, 1 <i>; yields {i8}:result = -1</i>
2650 <result> = ashr i32 1, 32 <i>; undefined</i>
2654 <!-- _______________________________________________________________________ -->
2655 <div class="doc_subsubsection"> <a name="i_and">'<tt>and</tt>'
2656 Instruction</a> </div>
2658 <div class="doc_text">
2663 <result> = and <ty> <op1>, <op2> <i>; yields {ty}:result</i>
2668 <p>The '<tt>and</tt>' instruction returns the bitwise logical and of
2669 its two operands.</p>
2673 <p>The two arguments to the '<tt>and</tt>' instruction must be
2674 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
2675 values. Both arguments must have identical types.</p>
2678 <p>The truth table used for the '<tt>and</tt>' instruction is:</p>
2681 <table border="1" cellspacing="0" cellpadding="4">
2713 <result> = and i32 4, %var <i>; yields {i32}:result = 4 & %var</i>
2714 <result> = and i32 15, 40 <i>; yields {i32}:result = 8</i>
2715 <result> = and i32 4, 8 <i>; yields {i32}:result = 0</i>
2718 <!-- _______________________________________________________________________ -->
2719 <div class="doc_subsubsection"> <a name="i_or">'<tt>or</tt>' Instruction</a> </div>
2720 <div class="doc_text">
2722 <pre> <result> = or <ty> <op1>, <op2> <i>; yields {ty}:result</i>
2725 <p>The '<tt>or</tt>' instruction returns the bitwise logical inclusive
2726 or of its two operands.</p>
2729 <p>The two arguments to the '<tt>or</tt>' instruction must be
2730 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
2731 values. Both arguments must have identical types.</p>
2733 <p>The truth table used for the '<tt>or</tt>' instruction is:</p>
2736 <table border="1" cellspacing="0" cellpadding="4">
2767 <pre> <result> = or i32 4, %var <i>; yields {i32}:result = 4 | %var</i>
2768 <result> = or i32 15, 40 <i>; yields {i32}:result = 47</i>
2769 <result> = or i32 4, 8 <i>; yields {i32}:result = 12</i>
2772 <!-- _______________________________________________________________________ -->
2773 <div class="doc_subsubsection"> <a name="i_xor">'<tt>xor</tt>'
2774 Instruction</a> </div>
2775 <div class="doc_text">
2777 <pre> <result> = xor <ty> <op1>, <op2> <i>; yields {ty}:result</i>
2780 <p>The '<tt>xor</tt>' instruction returns the bitwise logical exclusive
2781 or of its two operands. The <tt>xor</tt> is used to implement the
2782 "one's complement" operation, which is the "~" operator in C.</p>
2784 <p>The two arguments to the '<tt>xor</tt>' instruction must be
2785 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
2786 values. Both arguments must have identical types.</p>
2790 <p>The truth table used for the '<tt>xor</tt>' instruction is:</p>
2793 <table border="1" cellspacing="0" cellpadding="4">
2825 <pre> <result> = xor i32 4, %var <i>; yields {i32}:result = 4 ^ %var</i>
2826 <result> = xor i32 15, 40 <i>; yields {i32}:result = 39</i>
2827 <result> = xor i32 4, 8 <i>; yields {i32}:result = 12</i>
2828 <result> = xor i32 %V, -1 <i>; yields {i32}:result = ~%V</i>
2832 <!-- ======================================================================= -->
2833 <div class="doc_subsection">
2834 <a name="vectorops">Vector Operations</a>
2837 <div class="doc_text">
2839 <p>LLVM supports several instructions to represent vector operations in a
2840 target-independent manner. These instructions cover the element-access and
2841 vector-specific operations needed to process vectors effectively. While LLVM
2842 does directly support these vector operations, many sophisticated algorithms
2843 will want to use target-specific intrinsics to take full advantage of a specific
2848 <!-- _______________________________________________________________________ -->
2849 <div class="doc_subsubsection">
2850 <a name="i_extractelement">'<tt>extractelement</tt>' Instruction</a>
2853 <div class="doc_text">
2858 <result> = extractelement <n x <ty>> <val>, i32 <idx> <i>; yields <ty></i>
2864 The '<tt>extractelement</tt>' instruction extracts a single scalar
2865 element from a vector at a specified index.
2872 The first operand of an '<tt>extractelement</tt>' instruction is a
2873 value of <a href="#t_vector">vector</a> type. The second operand is
2874 an index indicating the position from which to extract the element.
2875 The index may be a variable.</p>
2880 The result is a scalar of the same type as the element type of
2881 <tt>val</tt>. Its value is the value at position <tt>idx</tt> of
2882 <tt>val</tt>. If <tt>idx</tt> exceeds the length of <tt>val</tt>, the
2883 results are undefined.
2889 %result = extractelement <4 x i32> %vec, i32 0 <i>; yields i32</i>
2894 <!-- _______________________________________________________________________ -->
2895 <div class="doc_subsubsection">
2896 <a name="i_insertelement">'<tt>insertelement</tt>' Instruction</a>
2899 <div class="doc_text">
2904 <result> = insertelement <n x <ty>> <val>, <ty> <elt>, i32 <idx> <i>; yields <n x <ty>></i>
2910 The '<tt>insertelement</tt>' instruction inserts a scalar
2911 element into a vector at a specified index.
2918 The first operand of an '<tt>insertelement</tt>' instruction is a
2919 value of <a href="#t_vector">vector</a> type. The second operand is a
2920 scalar value whose type must equal the element type of the first
2921 operand. The third operand is an index indicating the position at
2922 which to insert the value. The index may be a variable.</p>
2927 The result is a vector of the same type as <tt>val</tt>. Its
2928 element values are those of <tt>val</tt> except at position
2929 <tt>idx</tt>, where it gets the value <tt>elt</tt>. If <tt>idx</tt>
2930 exceeds the length of <tt>val</tt>, the results are undefined.
2936 %result = insertelement <4 x i32> %vec, i32 1, i32 0 <i>; yields <4 x i32></i>
2940 <!-- _______________________________________________________________________ -->
2941 <div class="doc_subsubsection">
2942 <a name="i_shufflevector">'<tt>shufflevector</tt>' Instruction</a>
2945 <div class="doc_text">
2950 <result> = shufflevector <n x <ty>> <v1>, <n x <ty>> <v2>, <m x i32> <mask> <i>; yields <m x <ty>></i>
2956 The '<tt>shufflevector</tt>' instruction constructs a permutation of elements
2957 from two input vectors, returning a vector with the same element type as
2958 the input and length that is the same as the shuffle mask.
2964 The first two operands of a '<tt>shufflevector</tt>' instruction are vectors
2965 with types that match each other. The third argument is a shuffle mask whose
2966 element type is always 'i32'. The result of the instruction is a vector whose
2967 length is the same as the shuffle mask and whose element type is the same as
2968 the element type of the first two operands.
2972 The shuffle mask operand is required to be a constant vector with either
2973 constant integer or undef values.
2979 The elements of the two input vectors are numbered from left to right across
2980 both of the vectors. The shuffle mask operand specifies, for each element of
2981 the result vector, which element of the two input vectors the result element
2982 gets. The element selector may be undef (meaning "don't care") and the second
2983 operand may be undef if performing a shuffle from only one vector.
2989 %result = shufflevector <4 x i32> %v1, <4 x i32> %v2,
2990 <4 x i32> <i32 0, i32 4, i32 1, i32 5> <i>; yields <4 x i32></i>
2991 %result = shufflevector <4 x i32> %v1, <4 x i32> undef,
2992 <4 x i32> <i32 0, i32 1, i32 2, i32 3> <i>; yields <4 x i32></i> - Identity shuffle.
2993 %result = shufflevector <8 x i32> %v1, <8 x i32> undef,
2994 <4 x i32> <i32 0, i32 1, i32 2, i32 3> <i>; yields <4 x i32></i>
2995 %result = shufflevector <4 x i32> %v1, <4 x i32> %v2,
2996 <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>
3001 <!-- ======================================================================= -->
3002 <div class="doc_subsection">
3003 <a name="aggregateops">Aggregate Operations</a>
3006 <div class="doc_text">
3008 <p>LLVM supports several instructions for working with aggregate values.
3013 <!-- _______________________________________________________________________ -->
3014 <div class="doc_subsubsection">
3015 <a name="i_extractvalue">'<tt>extractvalue</tt>' Instruction</a>
3018 <div class="doc_text">
3023 <result> = extractvalue <aggregate type> <val>, <idx>{, <idx>}*
3029 The '<tt>extractvalue</tt>' instruction extracts the value of a struct field
3030 or array element from an aggregate value.
3037 The first operand of an '<tt>extractvalue</tt>' instruction is a
3038 value of <a href="#t_struct">struct</a> or <a href="#t_array">array</a>
3039 type. The operands are constant indices to specify which value to extract
3040 in a similar manner as indices in a
3041 '<tt><a href="#i_getelementptr">getelementptr</a></tt>' instruction.
3047 The result is the value at the position in the aggregate specified by
3054 %result = extractvalue {i32, float} %agg, 0 <i>; yields i32</i>
3059 <!-- _______________________________________________________________________ -->
3060 <div class="doc_subsubsection">
3061 <a name="i_insertvalue">'<tt>insertvalue</tt>' Instruction</a>
3064 <div class="doc_text">
3069 <result> = insertvalue <aggregate type> <val>, <ty> <val>, <idx> <i>; yields <n x <ty>></i>
3075 The '<tt>insertvalue</tt>' instruction inserts a value
3076 into a struct field or array element in an aggregate.
3083 The first operand of an '<tt>insertvalue</tt>' instruction is a
3084 value of <a href="#t_struct">struct</a> or <a href="#t_array">array</a> type.
3085 The second operand is a first-class value to insert.
3086 The following operands are constant indices
3087 indicating the position at which to insert the value in a similar manner as
3089 '<tt><a href="#i_getelementptr">getelementptr</a></tt>' instruction.
3090 The value to insert must have the same type as the value identified
3097 The result is an aggregate of the same type as <tt>val</tt>. Its
3098 value is that of <tt>val</tt> except that the value at the position
3099 specified by the indices is that of <tt>elt</tt>.
3105 %result = insertvalue {i32, float} %agg, i32 1, 0 <i>; yields {i32, float}</i>
3110 <!-- ======================================================================= -->
3111 <div class="doc_subsection">
3112 <a name="memoryops">Memory Access and Addressing Operations</a>
3115 <div class="doc_text">
3117 <p>A key design point of an SSA-based representation is how it
3118 represents memory. In LLVM, no memory locations are in SSA form, which
3119 makes things very simple. This section describes how to read, write,
3120 allocate, and free memory in LLVM.</p>
3124 <!-- _______________________________________________________________________ -->
3125 <div class="doc_subsubsection">
3126 <a name="i_malloc">'<tt>malloc</tt>' Instruction</a>
3129 <div class="doc_text">
3134 <result> = malloc <type>[, i32 <NumElements>][, align <alignment>] <i>; yields {type*}:result</i>
3139 <p>The '<tt>malloc</tt>' instruction allocates memory from the system
3140 heap and returns a pointer to it. The object is always allocated in the generic
3141 address space (address space zero).</p>
3145 <p>The '<tt>malloc</tt>' instruction allocates
3146 <tt>sizeof(<type>)*NumElements</tt>
3147 bytes of memory from the operating system and returns a pointer of the
3148 appropriate type to the program. If "NumElements" is specified, it is the
3149 number of elements allocated, otherwise "NumElements" is defaulted to be one.
3150 If a constant alignment is specified, the value result of the allocation is guaranteed to
3151 be aligned to at least that boundary. If not specified, or if zero, the target can
3152 choose to align the allocation on any convenient boundary.</p>
3154 <p>'<tt>type</tt>' must be a sized type.</p>
3158 <p>Memory is allocated using the system "<tt>malloc</tt>" function, and
3159 a pointer is returned. The result of a zero byte allocation is undefined. The
3160 result is null if there is insufficient memory available.</p>
3165 %array = malloc [4 x i8 ] <i>; yields {[%4 x i8]*}:array</i>
3167 %size = <a href="#i_add">add</a> i32 2, 2 <i>; yields {i32}:size = i32 4</i>
3168 %array1 = malloc i8, i32 4 <i>; yields {i8*}:array1</i>
3169 %array2 = malloc [12 x i8], i32 %size <i>; yields {[12 x i8]*}:array2</i>
3170 %array3 = malloc i32, i32 4, align 1024 <i>; yields {i32*}:array3</i>
3171 %array4 = malloc i32, align 1024 <i>; yields {i32*}:array4</i>
3175 <!-- _______________________________________________________________________ -->
3176 <div class="doc_subsubsection">
3177 <a name="i_free">'<tt>free</tt>' Instruction</a>
3180 <div class="doc_text">
3185 free <type> <value> <i>; yields {void}</i>
3190 <p>The '<tt>free</tt>' instruction returns memory back to the unused
3191 memory heap to be reallocated in the future.</p>
3195 <p>'<tt>value</tt>' shall be a pointer value that points to a value
3196 that was allocated with the '<tt><a href="#i_malloc">malloc</a></tt>'
3201 <p>Access to the memory pointed to by the pointer is no longer defined
3202 after this instruction executes. If the pointer is null, the operation
3208 %array = <a href="#i_malloc">malloc</a> [4 x i8] <i>; yields {[4 x i8]*}:array</i>
3209 free [4 x i8]* %array
3213 <!-- _______________________________________________________________________ -->
3214 <div class="doc_subsubsection">
3215 <a name="i_alloca">'<tt>alloca</tt>' Instruction</a>
3218 <div class="doc_text">
3223 <result> = alloca <type>[, i32 <NumElements>][, align <alignment>] <i>; yields {type*}:result</i>
3228 <p>The '<tt>alloca</tt>' instruction allocates memory on the stack frame of the
3229 currently executing function, to be automatically released when this function
3230 returns to its caller. The object is always allocated in the generic address
3231 space (address space zero).</p>
3235 <p>The '<tt>alloca</tt>' instruction allocates <tt>sizeof(<type>)*NumElements</tt>
3236 bytes of memory on the runtime stack, returning a pointer of the
3237 appropriate type to the program. If "NumElements" is specified, it is the
3238 number of elements allocated, otherwise "NumElements" is defaulted to be one.
3239 If a constant alignment is specified, the value result of the allocation is guaranteed
3240 to be aligned to at least that boundary. If not specified, or if zero, the target
3241 can choose to align the allocation on any convenient boundary.</p>
3243 <p>'<tt>type</tt>' may be any sized type.</p>
3247 <p>Memory is allocated; a pointer is returned. The operation is undefiend if
3248 there is insufficient stack space for the allocation. '<tt>alloca</tt>'d
3249 memory is automatically released when the function returns. The '<tt>alloca</tt>'
3250 instruction is commonly used to represent automatic variables that must
3251 have an address available. When the function returns (either with the <tt><a
3252 href="#i_ret">ret</a></tt> or <tt><a href="#i_unwind">unwind</a></tt>
3253 instructions), the memory is reclaimed. Allocating zero bytes
3254 is legal, but the result is undefined.</p>
3259 %ptr = alloca i32 <i>; yields {i32*}:ptr</i>
3260 %ptr = alloca i32, i32 4 <i>; yields {i32*}:ptr</i>
3261 %ptr = alloca i32, i32 4, align 1024 <i>; yields {i32*}:ptr</i>
3262 %ptr = alloca i32, align 1024 <i>; yields {i32*}:ptr</i>
3266 <!-- _______________________________________________________________________ -->
3267 <div class="doc_subsubsection"> <a name="i_load">'<tt>load</tt>'
3268 Instruction</a> </div>
3269 <div class="doc_text">
3271 <pre> <result> = load <ty>* <pointer>[, align <alignment>]<br> <result> = volatile load <ty>* <pointer>[, align <alignment>]<br></pre>
3273 <p>The '<tt>load</tt>' instruction is used to read from memory.</p>
3275 <p>The argument to the '<tt>load</tt>' instruction specifies the memory
3276 address from which to load. The pointer must point to a <a
3277 href="#t_firstclass">first class</a> type. If the <tt>load</tt> is
3278 marked as <tt>volatile</tt>, then the optimizer is not allowed to modify
3279 the number or order of execution of this <tt>load</tt> with other
3280 volatile <tt>load</tt> and <tt><a href="#i_store">store</a></tt>
3283 The optional constant "align" argument specifies the alignment of the operation
3284 (that is, the alignment of the memory address). A value of 0 or an
3285 omitted "align" argument means that the operation has the preferential
3286 alignment for the target. It is the responsibility of the code emitter
3287 to ensure that the alignment information is correct. Overestimating
3288 the alignment results in an undefined behavior. Underestimating the
3289 alignment may produce less efficient code. An alignment of 1 is always
3293 <p>The location of memory pointed to is loaded.</p>
3295 <pre> %ptr = <a href="#i_alloca">alloca</a> i32 <i>; yields {i32*}:ptr</i>
3297 href="#i_store">store</a> i32 3, i32* %ptr <i>; yields {void}</i>
3298 %val = load i32* %ptr <i>; yields {i32}:val = i32 3</i>
3301 <!-- _______________________________________________________________________ -->
3302 <div class="doc_subsubsection"> <a name="i_store">'<tt>store</tt>'
3303 Instruction</a> </div>
3304 <div class="doc_text">
3306 <pre> store <ty> <value>, <ty>* <pointer>[, align <alignment>] <i>; yields {void}</i>
3307 volatile store <ty> <value>, <ty>* <pointer>[, align <alignment>] <i>; yields {void}</i>
3310 <p>The '<tt>store</tt>' instruction is used to write to memory.</p>
3312 <p>There are two arguments to the '<tt>store</tt>' instruction: a value
3313 to store and an address at which to store it. The type of the '<tt><pointer></tt>'
3314 operand must be a pointer to the <a href="#t_firstclass">first class</a> type
3315 of the '<tt><value></tt>'
3316 operand. If the <tt>store</tt> is marked as <tt>volatile</tt>, then the
3317 optimizer is not allowed to modify the number or order of execution of
3318 this <tt>store</tt> with other volatile <tt>load</tt> and <tt><a
3319 href="#i_store">store</a></tt> instructions.</p>
3321 The optional constant "align" argument specifies the alignment of the operation
3322 (that is, the alignment of the memory address). A value of 0 or an
3323 omitted "align" argument means that the operation has the preferential
3324 alignment for the target. It is the responsibility of the code emitter
3325 to ensure that the alignment information is correct. Overestimating
3326 the alignment results in an undefined behavior. Underestimating the
3327 alignment may produce less efficient code. An alignment of 1 is always
3331 <p>The contents of memory are updated to contain '<tt><value></tt>'
3332 at the location specified by the '<tt><pointer></tt>' operand.</p>
3334 <pre> %ptr = <a href="#i_alloca">alloca</a> i32 <i>; yields {i32*}:ptr</i>
3335 store i32 3, i32* %ptr <i>; yields {void}</i>
3336 %val = <a href="#i_load">load</a> i32* %ptr <i>; yields {i32}:val = i32 3</i>
3340 <!-- _______________________________________________________________________ -->
3341 <div class="doc_subsubsection">
3342 <a name="i_getelementptr">'<tt>getelementptr</tt>' Instruction</a>
3345 <div class="doc_text">
3348 <result> = getelementptr <pty>* <ptrval>{, <ty> <idx>}*
3354 The '<tt>getelementptr</tt>' instruction is used to get the address of a
3355 subelement of an aggregate data structure. It performs address calculation only
3356 and does not access memory.</p>
3360 <p>The first argument is always a pointer, and forms the basis of the
3361 calculation. The remaining arguments are indices, that indicate which of the
3362 elements of the aggregate object are indexed. The interpretation of each index
3363 is dependent on the type being indexed into. The first index always indexes the
3364 pointer value given as the first argument, the second index indexes a value of
3365 the type pointed to (not necessarily the value directly pointed to, since the
3366 first index can be non-zero), etc. The first type indexed into must be a pointer
3367 value, subsequent types can be arrays, vectors and structs. Note that subsequent
3368 types being indexed into can never be pointers, since that would require loading
3369 the pointer before continuing calculation.</p>
3371 <p>The type of each index argument depends on the type it is indexing into.
3372 When indexing into a (packed) structure, only <tt>i32</tt> integer
3373 <b>constants</b> are allowed. When indexing into an array, pointer or vector,
3374 only integers of 32 or 64 bits are allowed (also non-constants). 32-bit values
3375 will be sign extended to 64-bits if required.</p>
3377 <p>For example, let's consider a C code fragment and how it gets
3378 compiled to LLVM:</p>
3380 <div class="doc_code">
3393 int *foo(struct ST *s) {
3394 return &s[1].Z.B[5][13];
3399 <p>The LLVM code generated by the GCC frontend is:</p>
3401 <div class="doc_code">
3403 %RT = type { i8 , [10 x [20 x i32]], i8 }
3404 %ST = type { i32, double, %RT }
3406 define i32* %foo(%ST* %s) {
3408 %reg = getelementptr %ST* %s, i32 1, i32 2, i32 1, i32 5, i32 13
3416 <p>In the example above, the first index is indexing into the '<tt>%ST*</tt>'
3417 type, which is a pointer, yielding a '<tt>%ST</tt>' = '<tt>{ i32, double, %RT
3418 }</tt>' type, a structure. The second index indexes into the third element of
3419 the structure, yielding a '<tt>%RT</tt>' = '<tt>{ i8 , [10 x [20 x i32]],
3420 i8 }</tt>' type, another structure. The third index indexes into the second
3421 element of the structure, yielding a '<tt>[10 x [20 x i32]]</tt>' type, an
3422 array. The two dimensions of the array are subscripted into, yielding an
3423 '<tt>i32</tt>' type. The '<tt>getelementptr</tt>' instruction returns a pointer
3424 to this element, thus computing a value of '<tt>i32*</tt>' type.</p>
3426 <p>Note that it is perfectly legal to index partially through a
3427 structure, returning a pointer to an inner element. Because of this,
3428 the LLVM code for the given testcase is equivalent to:</p>
3431 define i32* %foo(%ST* %s) {
3432 %t1 = getelementptr %ST* %s, i32 1 <i>; yields %ST*:%t1</i>
3433 %t2 = getelementptr %ST* %t1, i32 0, i32 2 <i>; yields %RT*:%t2</i>
3434 %t3 = getelementptr %RT* %t2, i32 0, i32 1 <i>; yields [10 x [20 x i32]]*:%t3</i>
3435 %t4 = getelementptr [10 x [20 x i32]]* %t3, i32 0, i32 5 <i>; yields [20 x i32]*:%t4</i>
3436 %t5 = getelementptr [20 x i32]* %t4, i32 0, i32 13 <i>; yields i32*:%t5</i>
3441 <p>Note that it is undefined to access an array out of bounds: array and
3442 pointer indexes must always be within the defined bounds of the array type.
3443 The one exception for this rule is zero length arrays. These arrays are
3444 defined to be accessible as variable length arrays, which requires access
3445 beyond the zero'th element.</p>
3447 <p>The getelementptr instruction is often confusing. For some more insight
3448 into how it works, see <a href="GetElementPtr.html">the getelementptr
3454 <i>; yields [12 x i8]*:aptr</i>
3455 %aptr = getelementptr {i32, [12 x i8]}* %saptr, i64 0, i32 1
3456 <i>; yields i8*:vptr</i>
3457 %vptr = getelementptr {i32, <2 x i8>}* %svptr, i64 0, i32 1, i32 1
3458 <i>; yields i8*:eptr</i>
3459 %eptr = getelementptr [12 x i8]* %aptr, i64 0, i32 1
3463 <!-- ======================================================================= -->
3464 <div class="doc_subsection"> <a name="convertops">Conversion Operations</a>
3466 <div class="doc_text">
3467 <p>The instructions in this category are the conversion instructions (casting)
3468 which all take a single operand and a type. They perform various bit conversions
3472 <!-- _______________________________________________________________________ -->
3473 <div class="doc_subsubsection">
3474 <a name="i_trunc">'<tt>trunc .. to</tt>' Instruction</a>
3476 <div class="doc_text">
3480 <result> = trunc <ty> <value> to <ty2> <i>; yields ty2</i>
3485 The '<tt>trunc</tt>' instruction truncates its operand to the type <tt>ty2</tt>.
3490 The '<tt>trunc</tt>' instruction takes a <tt>value</tt> to trunc, which must
3491 be an <a href="#t_integer">integer</a> type, and a type that specifies the size
3492 and type of the result, which must be an <a href="#t_integer">integer</a>
3493 type. The bit size of <tt>value</tt> must be larger than the bit size of
3494 <tt>ty2</tt>. Equal sized types are not allowed.</p>
3498 The '<tt>trunc</tt>' instruction truncates the high order bits in <tt>value</tt>
3499 and converts the remaining bits to <tt>ty2</tt>. Since the source size must be
3500 larger than the destination size, <tt>trunc</tt> cannot be a <i>no-op cast</i>.
3501 It will always truncate bits.</p>
3505 %X = trunc i32 257 to i8 <i>; yields i8:1</i>
3506 %Y = trunc i32 123 to i1 <i>; yields i1:true</i>
3507 %Y = trunc i32 122 to i1 <i>; yields i1:false</i>
3511 <!-- _______________________________________________________________________ -->
3512 <div class="doc_subsubsection">
3513 <a name="i_zext">'<tt>zext .. to</tt>' Instruction</a>
3515 <div class="doc_text">
3519 <result> = zext <ty> <value> to <ty2> <i>; yields ty2</i>
3523 <p>The '<tt>zext</tt>' instruction zero extends its operand to type
3528 <p>The '<tt>zext</tt>' instruction takes a value to cast, which must be of
3529 <a href="#t_integer">integer</a> type, and a type to cast it to, which must
3530 also be of <a href="#t_integer">integer</a> type. The bit size of the
3531 <tt>value</tt> must be smaller than the bit size of the destination type,
3535 <p>The <tt>zext</tt> fills the high order bits of the <tt>value</tt> with zero
3536 bits until it reaches the size of the destination type, <tt>ty2</tt>.</p>
3538 <p>When zero extending from i1, the result will always be either 0 or 1.</p>
3542 %X = zext i32 257 to i64 <i>; yields i64:257</i>
3543 %Y = zext i1 true to i32 <i>; yields i32:1</i>
3547 <!-- _______________________________________________________________________ -->
3548 <div class="doc_subsubsection">
3549 <a name="i_sext">'<tt>sext .. to</tt>' Instruction</a>
3551 <div class="doc_text">
3555 <result> = sext <ty> <value> to <ty2> <i>; yields ty2</i>
3559 <p>The '<tt>sext</tt>' sign extends <tt>value</tt> to the type <tt>ty2</tt>.</p>
3563 The '<tt>sext</tt>' instruction takes a value to cast, which must be of
3564 <a href="#t_integer">integer</a> type, and a type to cast it to, which must
3565 also be of <a href="#t_integer">integer</a> type. The bit size of the
3566 <tt>value</tt> must be smaller than the bit size of the destination type,
3571 The '<tt>sext</tt>' instruction performs a sign extension by copying the sign
3572 bit (highest order bit) of the <tt>value</tt> until it reaches the bit size of
3573 the type <tt>ty2</tt>.</p>
3575 <p>When sign extending from i1, the extension always results in -1 or 0.</p>
3579 %X = sext i8 -1 to i16 <i>; yields i16 :65535</i>
3580 %Y = sext i1 true to i32 <i>; yields i32:-1</i>
3584 <!-- _______________________________________________________________________ -->
3585 <div class="doc_subsubsection">
3586 <a name="i_fptrunc">'<tt>fptrunc .. to</tt>' Instruction</a>
3589 <div class="doc_text">
3594 <result> = fptrunc <ty> <value> to <ty2> <i>; yields ty2</i>
3598 <p>The '<tt>fptrunc</tt>' instruction truncates <tt>value</tt> to type
3603 <p>The '<tt>fptrunc</tt>' instruction takes a <a href="#t_floating">floating
3604 point</a> value to cast and a <a href="#t_floating">floating point</a> type to
3605 cast it to. The size of <tt>value</tt> must be larger than the size of
3606 <tt>ty2</tt>. This implies that <tt>fptrunc</tt> cannot be used to make a
3607 <i>no-op cast</i>.</p>
3610 <p> The '<tt>fptrunc</tt>' instruction truncates a <tt>value</tt> from a larger
3611 <a href="#t_floating">floating point</a> type to a smaller
3612 <a href="#t_floating">floating point</a> type. If the value cannot fit within
3613 the destination type, <tt>ty2</tt>, then the results are undefined.</p>
3617 %X = fptrunc double 123.0 to float <i>; yields float:123.0</i>
3618 %Y = fptrunc double 1.0E+300 to float <i>; yields undefined</i>
3622 <!-- _______________________________________________________________________ -->
3623 <div class="doc_subsubsection">
3624 <a name="i_fpext">'<tt>fpext .. to</tt>' Instruction</a>
3626 <div class="doc_text">
3630 <result> = fpext <ty> <value> to <ty2> <i>; yields ty2</i>
3634 <p>The '<tt>fpext</tt>' extends a floating point <tt>value</tt> to a larger
3635 floating point value.</p>
3638 <p>The '<tt>fpext</tt>' instruction takes a
3639 <a href="#t_floating">floating point</a> <tt>value</tt> to cast,
3640 and a <a href="#t_floating">floating point</a> type to cast it to. The source
3641 type must be smaller than the destination type.</p>
3644 <p>The '<tt>fpext</tt>' instruction extends the <tt>value</tt> from a smaller
3645 <a href="#t_floating">floating point</a> type to a larger
3646 <a href="#t_floating">floating point</a> type. The <tt>fpext</tt> cannot be
3647 used to make a <i>no-op cast</i> because it always changes bits. Use
3648 <tt>bitcast</tt> to make a <i>no-op cast</i> for a floating point cast.</p>
3652 %X = fpext float 3.1415 to double <i>; yields double:3.1415</i>
3653 %Y = fpext float 1.0 to float <i>; yields float:1.0 (no-op)</i>
3657 <!-- _______________________________________________________________________ -->
3658 <div class="doc_subsubsection">
3659 <a name="i_fptoui">'<tt>fptoui .. to</tt>' Instruction</a>
3661 <div class="doc_text">
3665 <result> = fptoui <ty> <value> to <ty2> <i>; yields ty2</i>
3669 <p>The '<tt>fptoui</tt>' converts a floating point <tt>value</tt> to its
3670 unsigned integer equivalent of type <tt>ty2</tt>.
3674 <p>The '<tt>fptoui</tt>' instruction takes a value to cast, which must be a
3675 scalar or vector <a href="#t_floating">floating point</a> value, and a type
3676 to cast it to <tt>ty2</tt>, which must be an <a href="#t_integer">integer</a>
3677 type. If <tt>ty</tt> is a vector floating point type, <tt>ty2</tt> must be a
3678 vector integer type with the same number of elements as <tt>ty</tt></p>
3681 <p> The '<tt>fptoui</tt>' instruction converts its
3682 <a href="#t_floating">floating point</a> operand into the nearest (rounding
3683 towards zero) unsigned integer value. If the value cannot fit in <tt>ty2</tt>,
3684 the results are undefined.</p>
3688 %X = fptoui double 123.0 to i32 <i>; yields i32:123</i>
3689 %Y = fptoui float 1.0E+300 to i1 <i>; yields undefined:1</i>
3690 %X = fptoui float 1.04E+17 to i8 <i>; yields undefined:1</i>
3694 <!-- _______________________________________________________________________ -->
3695 <div class="doc_subsubsection">
3696 <a name="i_fptosi">'<tt>fptosi .. to</tt>' Instruction</a>
3698 <div class="doc_text">
3702 <result> = fptosi <ty> <value> to <ty2> <i>; yields ty2</i>
3706 <p>The '<tt>fptosi</tt>' instruction converts
3707 <a href="#t_floating">floating point</a> <tt>value</tt> to type <tt>ty2</tt>.
3711 <p> The '<tt>fptosi</tt>' instruction takes a value to cast, which must be a
3712 scalar or vector <a href="#t_floating">floating point</a> value, and a type
3713 to cast it to <tt>ty2</tt>, which must be an <a href="#t_integer">integer</a>
3714 type. If <tt>ty</tt> is a vector floating point type, <tt>ty2</tt> must be a
3715 vector integer type with the same number of elements as <tt>ty</tt></p>
3718 <p>The '<tt>fptosi</tt>' instruction converts its
3719 <a href="#t_floating">floating point</a> operand into the nearest (rounding
3720 towards zero) signed integer value. If the value cannot fit in <tt>ty2</tt>,
3721 the results are undefined.</p>
3725 %X = fptosi double -123.0 to i32 <i>; yields i32:-123</i>
3726 %Y = fptosi float 1.0E-247 to i1 <i>; yields undefined:1</i>
3727 %X = fptosi float 1.04E+17 to i8 <i>; yields undefined:1</i>
3731 <!-- _______________________________________________________________________ -->
3732 <div class="doc_subsubsection">
3733 <a name="i_uitofp">'<tt>uitofp .. to</tt>' Instruction</a>
3735 <div class="doc_text">
3739 <result> = uitofp <ty> <value> to <ty2> <i>; yields ty2</i>
3743 <p>The '<tt>uitofp</tt>' instruction regards <tt>value</tt> as an unsigned
3744 integer and converts that value to the <tt>ty2</tt> type.</p>
3747 <p>The '<tt>uitofp</tt>' instruction takes a value to cast, which must be a
3748 scalar or vector <a href="#t_integer">integer</a> value, and a type to cast it
3749 to <tt>ty2</tt>, which must be an <a href="#t_floating">floating point</a>
3750 type. If <tt>ty</tt> is a vector integer type, <tt>ty2</tt> must be a vector
3751 floating point type with the same number of elements as <tt>ty</tt></p>
3754 <p>The '<tt>uitofp</tt>' instruction interprets its operand as an unsigned
3755 integer quantity and converts it to the corresponding floating point value. If
3756 the value cannot fit in the floating point value, the results are undefined.</p>
3760 %X = uitofp i32 257 to float <i>; yields float:257.0</i>
3761 %Y = uitofp i8 -1 to double <i>; yields double:255.0</i>
3765 <!-- _______________________________________________________________________ -->
3766 <div class="doc_subsubsection">
3767 <a name="i_sitofp">'<tt>sitofp .. to</tt>' Instruction</a>
3769 <div class="doc_text">
3773 <result> = sitofp <ty> <value> to <ty2> <i>; yields ty2</i>
3777 <p>The '<tt>sitofp</tt>' instruction regards <tt>value</tt> as a signed
3778 integer and converts that value to the <tt>ty2</tt> type.</p>
3781 <p>The '<tt>sitofp</tt>' instruction takes a value to cast, which must be a
3782 scalar or vector <a href="#t_integer">integer</a> value, and a type to cast it
3783 to <tt>ty2</tt>, which must be an <a href="#t_floating">floating point</a>
3784 type. If <tt>ty</tt> is a vector integer type, <tt>ty2</tt> must be a vector
3785 floating point type with the same number of elements as <tt>ty</tt></p>
3788 <p>The '<tt>sitofp</tt>' instruction interprets its operand as a signed
3789 integer quantity and converts it to the corresponding floating point value. If
3790 the value cannot fit in the floating point value, the results are undefined.</p>
3794 %X = sitofp i32 257 to float <i>; yields float:257.0</i>
3795 %Y = sitofp i8 -1 to double <i>; yields double:-1.0</i>
3799 <!-- _______________________________________________________________________ -->
3800 <div class="doc_subsubsection">
3801 <a name="i_ptrtoint">'<tt>ptrtoint .. to</tt>' Instruction</a>
3803 <div class="doc_text">
3807 <result> = ptrtoint <ty> <value> to <ty2> <i>; yields ty2</i>
3811 <p>The '<tt>ptrtoint</tt>' instruction converts the pointer <tt>value</tt> to
3812 the integer type <tt>ty2</tt>.</p>
3815 <p>The '<tt>ptrtoint</tt>' instruction takes a <tt>value</tt> to cast, which
3816 must be a <a href="#t_pointer">pointer</a> value, and a type to cast it to
3817 <tt>ty2</tt>, which must be an <a href="#t_integer">integer</a> type.</p>
3820 <p>The '<tt>ptrtoint</tt>' instruction converts <tt>value</tt> to integer type
3821 <tt>ty2</tt> by interpreting the pointer value as an integer and either
3822 truncating or zero extending that value to the size of the integer type. If
3823 <tt>value</tt> is smaller than <tt>ty2</tt> then a zero extension is done. If
3824 <tt>value</tt> is larger than <tt>ty2</tt> then a truncation is done. If they
3825 are the same size, then nothing is done (<i>no-op cast</i>) other than a type
3830 %X = ptrtoint i32* %X to i8 <i>; yields truncation on 32-bit architecture</i>
3831 %Y = ptrtoint i32* %x to i64 <i>; yields zero extension on 32-bit architecture</i>
3835 <!-- _______________________________________________________________________ -->
3836 <div class="doc_subsubsection">
3837 <a name="i_inttoptr">'<tt>inttoptr .. to</tt>' Instruction</a>
3839 <div class="doc_text">
3843 <result> = inttoptr <ty> <value> to <ty2> <i>; yields ty2</i>
3847 <p>The '<tt>inttoptr</tt>' instruction converts an integer <tt>value</tt> to
3848 a pointer type, <tt>ty2</tt>.</p>
3851 <p>The '<tt>inttoptr</tt>' instruction takes an <a href="#t_integer">integer</a>
3852 value to cast, and a type to cast it to, which must be a
3853 <a href="#t_pointer">pointer</a> type.</p>
3856 <p>The '<tt>inttoptr</tt>' instruction converts <tt>value</tt> to type
3857 <tt>ty2</tt> by applying either a zero extension or a truncation depending on
3858 the size of the integer <tt>value</tt>. If <tt>value</tt> is larger than the
3859 size of a pointer then a truncation is done. If <tt>value</tt> is smaller than
3860 the size of a pointer then a zero extension is done. If they are the same size,
3861 nothing is done (<i>no-op cast</i>).</p>
3865 %X = inttoptr i32 255 to i32* <i>; yields zero extension on 64-bit architecture</i>
3866 %X = inttoptr i32 255 to i32* <i>; yields no-op on 32-bit architecture</i>
3867 %Y = inttoptr i64 0 to i32* <i>; yields truncation on 32-bit architecture</i>
3871 <!-- _______________________________________________________________________ -->
3872 <div class="doc_subsubsection">
3873 <a name="i_bitcast">'<tt>bitcast .. to</tt>' Instruction</a>
3875 <div class="doc_text">
3879 <result> = bitcast <ty> <value> to <ty2> <i>; yields ty2</i>
3884 <p>The '<tt>bitcast</tt>' instruction converts <tt>value</tt> to type
3885 <tt>ty2</tt> without changing any bits.</p>
3889 <p>The '<tt>bitcast</tt>' instruction takes a value to cast, which must be
3890 a non-aggregate first class value, and a type to cast it to, which must also be
3891 a non-aggregate <a href="#t_firstclass">first class</a> type. The bit sizes of
3893 and the destination type, <tt>ty2</tt>, must be identical. If the source
3894 type is a pointer, the destination type must also be a pointer. This
3895 instruction supports bitwise conversion of vectors to integers and to vectors
3896 of other types (as long as they have the same size).</p>
3899 <p>The '<tt>bitcast</tt>' instruction converts <tt>value</tt> to type
3900 <tt>ty2</tt>. It is always a <i>no-op cast</i> because no bits change with
3901 this conversion. The conversion is done as if the <tt>value</tt> had been
3902 stored to memory and read back as type <tt>ty2</tt>. Pointer types may only be
3903 converted to other pointer types with this instruction. To convert pointers to
3904 other types, use the <a href="#i_inttoptr">inttoptr</a> or
3905 <a href="#i_ptrtoint">ptrtoint</a> instructions first.</p>
3909 %X = bitcast i8 255 to i8 <i>; yields i8 :-1</i>
3910 %Y = bitcast i32* %x to sint* <i>; yields sint*:%x</i>
3911 %Z = bitcast <2 x int> %V to i64; <i>; yields i64: %V</i>
3915 <!-- ======================================================================= -->
3916 <div class="doc_subsection"> <a name="otherops">Other Operations</a> </div>
3917 <div class="doc_text">
3918 <p>The instructions in this category are the "miscellaneous"
3919 instructions, which defy better classification.</p>
3922 <!-- _______________________________________________________________________ -->
3923 <div class="doc_subsubsection"><a name="i_icmp">'<tt>icmp</tt>' Instruction</a>
3925 <div class="doc_text">
3927 <pre> <result> = icmp <cond> <ty> <op1>, <op2> <i>; yields {i1} or {<N x i1>}:result</i>
3930 <p>The '<tt>icmp</tt>' instruction returns a boolean value or
3931 a vector of boolean values based on comparison
3932 of its two integer, integer vector, or pointer operands.</p>
3934 <p>The '<tt>icmp</tt>' instruction takes three operands. The first operand is
3935 the condition code indicating the kind of comparison to perform. It is not
3936 a value, just a keyword. The possible condition code are:
3939 <li><tt>eq</tt>: equal</li>
3940 <li><tt>ne</tt>: not equal </li>
3941 <li><tt>ugt</tt>: unsigned greater than</li>
3942 <li><tt>uge</tt>: unsigned greater or equal</li>
3943 <li><tt>ult</tt>: unsigned less than</li>
3944 <li><tt>ule</tt>: unsigned less or equal</li>
3945 <li><tt>sgt</tt>: signed greater than</li>
3946 <li><tt>sge</tt>: signed greater or equal</li>
3947 <li><tt>slt</tt>: signed less than</li>
3948 <li><tt>sle</tt>: signed less or equal</li>
3950 <p>The remaining two arguments must be <a href="#t_integer">integer</a> or
3951 <a href="#t_pointer">pointer</a>
3952 or integer <a href="#t_vector">vector</a> typed.
3953 They must also be identical types.</p>
3955 <p>The '<tt>icmp</tt>' compares <tt>op1</tt> and <tt>op2</tt> according to
3956 the condition code given as <tt>cond</tt>. The comparison performed always
3957 yields either an <a href="#t_primitive"><tt>i1</tt></a> or vector of <tt>i1</tt> result, as follows:
3960 <li><tt>eq</tt>: yields <tt>true</tt> if the operands are equal,
3961 <tt>false</tt> otherwise. No sign interpretation is necessary or performed.
3963 <li><tt>ne</tt>: yields <tt>true</tt> if the operands are unequal,
3964 <tt>false</tt> otherwise. No sign interpretation is necessary or performed.</li>
3965 <li><tt>ugt</tt>: interprets the operands as unsigned values and yields
3966 <tt>true</tt> if <tt>op1</tt> is greater than <tt>op2</tt>.</li>
3967 <li><tt>uge</tt>: interprets the operands as unsigned values and yields
3968 <tt>true</tt> if <tt>op1</tt> is greater than or equal to <tt>op2</tt>.</li>
3969 <li><tt>ult</tt>: interprets the operands as unsigned values and yields
3970 <tt>true</tt> if <tt>op1</tt> is less than <tt>op2</tt>.</li>
3971 <li><tt>ule</tt>: interprets the operands as unsigned values and yields
3972 <tt>true</tt> if <tt>op1</tt> is less than or equal to <tt>op2</tt>.</li>
3973 <li><tt>sgt</tt>: interprets the operands as signed values and yields
3974 <tt>true</tt> if <tt>op1</tt> is greater than <tt>op2</tt>.</li>
3975 <li><tt>sge</tt>: interprets the operands as signed values and yields
3976 <tt>true</tt> if <tt>op1</tt> is greater than or equal to <tt>op2</tt>.</li>
3977 <li><tt>slt</tt>: interprets the operands as signed values and yields
3978 <tt>true</tt> if <tt>op1</tt> is less than <tt>op2</tt>.</li>
3979 <li><tt>sle</tt>: interprets the operands as signed values and yields
3980 <tt>true</tt> if <tt>op1</tt> is less than or equal to <tt>op2</tt>.</li>
3982 <p>If the operands are <a href="#t_pointer">pointer</a> typed, the pointer
3983 values are compared as if they were integers.</p>
3984 <p>If the operands are integer vectors, then they are compared
3985 element by element. The result is an <tt>i1</tt> vector with
3986 the same number of elements as the values being compared.
3987 Otherwise, the result is an <tt>i1</tt>.
3991 <pre> <result> = icmp eq i32 4, 5 <i>; yields: result=false</i>
3992 <result> = icmp ne float* %X, %X <i>; yields: result=false</i>
3993 <result> = icmp ult i16 4, 5 <i>; yields: result=true</i>
3994 <result> = icmp sgt i16 4, 5 <i>; yields: result=false</i>
3995 <result> = icmp ule i16 -4, 5 <i>; yields: result=false</i>
3996 <result> = icmp sge i16 4, 5 <i>; yields: result=false</i>
4000 <!-- _______________________________________________________________________ -->
4001 <div class="doc_subsubsection"><a name="i_fcmp">'<tt>fcmp</tt>' Instruction</a>
4003 <div class="doc_text">
4005 <pre> <result> = fcmp <cond> <ty> <op1>, <op2> <i>; yields {i1} or {<N x i1>}:result</i>
4008 <p>The '<tt>fcmp</tt>' instruction returns a boolean value
4009 or vector of boolean values based on comparison
4010 of its operands.</p>
4012 If the operands are floating point scalars, then the result
4013 type is a boolean (<a href="#t_primitive"><tt>i1</tt></a>).
4015 <p>If the operands are floating point vectors, then the result type
4016 is a vector of boolean with the same number of elements as the
4017 operands being compared.</p>
4019 <p>The '<tt>fcmp</tt>' instruction takes three operands. The first operand is
4020 the condition code indicating the kind of comparison to perform. It is not
4021 a value, just a keyword. The possible condition code are:</p>
4023 <li><tt>false</tt>: no comparison, always returns false</li>
4024 <li><tt>oeq</tt>: ordered and equal</li>
4025 <li><tt>ogt</tt>: ordered and greater than </li>
4026 <li><tt>oge</tt>: ordered and greater than or equal</li>
4027 <li><tt>olt</tt>: ordered and less than </li>
4028 <li><tt>ole</tt>: ordered and less than or equal</li>
4029 <li><tt>one</tt>: ordered and not equal</li>
4030 <li><tt>ord</tt>: ordered (no nans)</li>
4031 <li><tt>ueq</tt>: unordered or equal</li>
4032 <li><tt>ugt</tt>: unordered or greater than </li>
4033 <li><tt>uge</tt>: unordered or greater than or equal</li>
4034 <li><tt>ult</tt>: unordered or less than </li>
4035 <li><tt>ule</tt>: unordered or less than or equal</li>
4036 <li><tt>une</tt>: unordered or not equal</li>
4037 <li><tt>uno</tt>: unordered (either nans)</li>
4038 <li><tt>true</tt>: no comparison, always returns true</li>
4040 <p><i>Ordered</i> means that neither operand is a QNAN while
4041 <i>unordered</i> means that either operand may be a QNAN.</p>
4042 <p>Each of <tt>val1</tt> and <tt>val2</tt> arguments must be
4043 either a <a href="#t_floating">floating point</a> type
4044 or a <a href="#t_vector">vector</a> of floating point type.
4045 They must have identical types.</p>
4047 <p>The '<tt>fcmp</tt>' instruction compares <tt>op1</tt> and <tt>op2</tt>
4048 according to the condition code given as <tt>cond</tt>.
4049 If the operands are vectors, then the vectors are compared
4051 Each comparison performed
4052 always yields an <a href="#t_primitive">i1</a> result, as follows:</p>
4054 <li><tt>false</tt>: always yields <tt>false</tt>, regardless of operands.</li>
4055 <li><tt>oeq</tt>: yields <tt>true</tt> if both operands are not a QNAN and
4056 <tt>op1</tt> is equal to <tt>op2</tt>.</li>
4057 <li><tt>ogt</tt>: yields <tt>true</tt> if both operands are not a QNAN and
4058 <tt>op1</tt> is greather than <tt>op2</tt>.</li>
4059 <li><tt>oge</tt>: yields <tt>true</tt> if both operands are not a QNAN and
4060 <tt>op1</tt> is greater than or equal to <tt>op2</tt>.</li>
4061 <li><tt>olt</tt>: yields <tt>true</tt> if both operands are not a QNAN and
4062 <tt>op1</tt> is less than <tt>op2</tt>.</li>
4063 <li><tt>ole</tt>: yields <tt>true</tt> if both operands are not a QNAN and
4064 <tt>op1</tt> is less than or equal to <tt>op2</tt>.</li>
4065 <li><tt>one</tt>: yields <tt>true</tt> if both operands are not a QNAN and
4066 <tt>op1</tt> is not equal to <tt>op2</tt>.</li>
4067 <li><tt>ord</tt>: yields <tt>true</tt> if both operands are not a QNAN.</li>
4068 <li><tt>ueq</tt>: yields <tt>true</tt> if either operand is a QNAN or
4069 <tt>op1</tt> is equal to <tt>op2</tt>.</li>
4070 <li><tt>ugt</tt>: yields <tt>true</tt> if either operand is a QNAN or
4071 <tt>op1</tt> is greater than <tt>op2</tt>.</li>
4072 <li><tt>uge</tt>: yields <tt>true</tt> if either operand is a QNAN or
4073 <tt>op1</tt> is greater than or equal to <tt>op2</tt>.</li>
4074 <li><tt>ult</tt>: yields <tt>true</tt> if either operand is a QNAN or
4075 <tt>op1</tt> is less than <tt>op2</tt>.</li>
4076 <li><tt>ule</tt>: yields <tt>true</tt> if either operand is a QNAN or
4077 <tt>op1</tt> is less than or equal to <tt>op2</tt>.</li>
4078 <li><tt>une</tt>: yields <tt>true</tt> if either operand is a QNAN or
4079 <tt>op1</tt> is not equal to <tt>op2</tt>.</li>
4080 <li><tt>uno</tt>: yields <tt>true</tt> if either operand is a QNAN.</li>
4081 <li><tt>true</tt>: always yields <tt>true</tt>, regardless of operands.</li>
4085 <pre> <result> = fcmp oeq float 4.0, 5.0 <i>; yields: result=false</i>
4086 <result> = fcmp one float 4.0, 5.0 <i>; yields: result=true</i>
4087 <result> = fcmp olt float 4.0, 5.0 <i>; yields: result=true</i>
4088 <result> = fcmp ueq double 1.0, 2.0 <i>; yields: result=false</i>
4092 <!-- _______________________________________________________________________ -->
4093 <div class="doc_subsubsection">
4094 <a name="i_vicmp">'<tt>vicmp</tt>' Instruction</a>
4096 <div class="doc_text">
4098 <pre> <result> = vicmp <cond> <ty> <op1>, <op2> <i>; yields {ty}:result</i>
4101 <p>The '<tt>vicmp</tt>' instruction returns an integer vector value based on
4102 element-wise comparison of its two integer vector operands.</p>
4104 <p>The '<tt>vicmp</tt>' instruction takes three operands. The first operand is
4105 the condition code indicating the kind of comparison to perform. It is not
4106 a value, just a keyword. The possible condition code are:</p>
4108 <li><tt>eq</tt>: equal</li>
4109 <li><tt>ne</tt>: not equal </li>
4110 <li><tt>ugt</tt>: unsigned greater than</li>
4111 <li><tt>uge</tt>: unsigned greater or equal</li>
4112 <li><tt>ult</tt>: unsigned less than</li>
4113 <li><tt>ule</tt>: unsigned less or equal</li>
4114 <li><tt>sgt</tt>: signed greater than</li>
4115 <li><tt>sge</tt>: signed greater or equal</li>
4116 <li><tt>slt</tt>: signed less than</li>
4117 <li><tt>sle</tt>: signed less or equal</li>
4119 <p>The remaining two arguments must be <a href="#t_vector">vector</a> or
4120 <a href="#t_integer">integer</a> typed. They must also be identical types.</p>
4122 <p>The '<tt>vicmp</tt>' instruction compares <tt>op1</tt> and <tt>op2</tt>
4123 according to the condition code given as <tt>cond</tt>. The comparison yields a
4124 <a href="#t_vector">vector</a> of <a href="#t_integer">integer</a> result, of
4125 identical type as the values being compared. The most significant bit in each
4126 element is 1 if the element-wise comparison evaluates to true, and is 0
4127 otherwise. All other bits of the result are undefined. The condition codes
4128 are evaluated identically to the <a href="#i_icmp">'<tt>icmp</tt>'
4129 instruction</a>.</p>
4133 <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>
4134 <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>
4138 <!-- _______________________________________________________________________ -->
4139 <div class="doc_subsubsection">
4140 <a name="i_vfcmp">'<tt>vfcmp</tt>' Instruction</a>
4142 <div class="doc_text">
4144 <pre> <result> = vfcmp <cond> <ty> <op1>, <op2></pre>
4146 <p>The '<tt>vfcmp</tt>' instruction returns an integer vector value based on
4147 element-wise comparison of its two floating point vector operands. The output
4148 elements have the same width as the input elements.</p>
4150 <p>The '<tt>vfcmp</tt>' instruction takes three operands. The first operand is
4151 the condition code indicating the kind of comparison to perform. It is not
4152 a value, just a keyword. The possible condition code are:</p>
4154 <li><tt>false</tt>: no comparison, always returns false</li>
4155 <li><tt>oeq</tt>: ordered and equal</li>
4156 <li><tt>ogt</tt>: ordered and greater than </li>
4157 <li><tt>oge</tt>: ordered and greater than or equal</li>
4158 <li><tt>olt</tt>: ordered and less than </li>
4159 <li><tt>ole</tt>: ordered and less than or equal</li>
4160 <li><tt>one</tt>: ordered and not equal</li>
4161 <li><tt>ord</tt>: ordered (no nans)</li>
4162 <li><tt>ueq</tt>: unordered or equal</li>
4163 <li><tt>ugt</tt>: unordered or greater than </li>
4164 <li><tt>uge</tt>: unordered or greater than or equal</li>
4165 <li><tt>ult</tt>: unordered or less than </li>
4166 <li><tt>ule</tt>: unordered or less than or equal</li>
4167 <li><tt>une</tt>: unordered or not equal</li>
4168 <li><tt>uno</tt>: unordered (either nans)</li>
4169 <li><tt>true</tt>: no comparison, always returns true</li>
4171 <p>The remaining two arguments must be <a href="#t_vector">vector</a> of
4172 <a href="#t_floating">floating point</a> typed. They must also be identical
4175 <p>The '<tt>vfcmp</tt>' instruction compares <tt>op1</tt> and <tt>op2</tt>
4176 according to the condition code given as <tt>cond</tt>. The comparison yields a
4177 <a href="#t_vector">vector</a> of <a href="#t_integer">integer</a> result, with
4178 an identical number of elements as the values being compared, and each element
4179 having identical with to the width of the floating point elements. The most
4180 significant bit in each element is 1 if the element-wise comparison evaluates to
4181 true, and is 0 otherwise. All other bits of the result are undefined. The
4182 condition codes are evaluated identically to the
4183 <a href="#i_fcmp">'<tt>fcmp</tt>' instruction</a>.</p>
4187 <i>; yields: result=<2 x i32> < i32 0, i32 -1 ></i>
4188 <result> = vfcmp oeq <2 x float> < float 4, float 0 >, < float 5, float 0 >
4190 <i>; yields: result=<2 x i64> < i64 -1, i64 0 ></i>
4191 <result> = vfcmp ult <2 x double> < double 1, double 2 >, < double 2, double 2>
4195 <!-- _______________________________________________________________________ -->
4196 <div class="doc_subsubsection">
4197 <a name="i_phi">'<tt>phi</tt>' Instruction</a>
4200 <div class="doc_text">
4204 <pre> <result> = phi <ty> [ <val0>, <label0>], ...<br></pre>
4206 <p>The '<tt>phi</tt>' instruction is used to implement the φ node in
4207 the SSA graph representing the function.</p>
4210 <p>The type of the incoming values is specified with the first type
4211 field. After this, the '<tt>phi</tt>' instruction takes a list of pairs
4212 as arguments, with one pair for each predecessor basic block of the
4213 current block. Only values of <a href="#t_firstclass">first class</a>
4214 type may be used as the value arguments to the PHI node. Only labels
4215 may be used as the label arguments.</p>
4217 <p>There must be no non-phi instructions between the start of a basic
4218 block and the PHI instructions: i.e. PHI instructions must be first in
4223 <p>At runtime, the '<tt>phi</tt>' instruction logically takes on the value
4224 specified by the pair corresponding to the predecessor basic block that executed
4225 just prior to the current block.</p>
4229 Loop: ; Infinite loop that counts from 0 on up...
4230 %indvar = phi i32 [ 0, %LoopHeader ], [ %nextindvar, %Loop ]
4231 %nextindvar = add i32 %indvar, 1
4236 <!-- _______________________________________________________________________ -->
4237 <div class="doc_subsubsection">
4238 <a name="i_select">'<tt>select</tt>' Instruction</a>
4241 <div class="doc_text">
4246 <result> = select <i>selty</i> <cond>, <ty> <val1>, <ty> <val2> <i>; yields ty</i>
4248 <i>selty</i> is either i1 or {<N x i1>}
4254 The '<tt>select</tt>' instruction is used to choose one value based on a
4255 condition, without branching.
4262 The '<tt>select</tt>' instruction requires an 'i1' value or
4263 a vector of 'i1' values indicating the
4264 condition, and two values of the same <a href="#t_firstclass">first class</a>
4265 type. If the val1/val2 are vectors and
4266 the condition is a scalar, then entire vectors are selected, not
4267 individual elements.
4273 If the condition is an i1 and it evaluates to 1, the instruction returns the first
4274 value argument; otherwise, it returns the second value argument.
4277 If the condition is a vector of i1, then the value arguments must
4278 be vectors of the same size, and the selection is done element
4285 %X = select i1 true, i8 17, i8 42 <i>; yields i8:17</i>
4290 <!-- _______________________________________________________________________ -->
4291 <div class="doc_subsubsection">
4292 <a name="i_call">'<tt>call</tt>' Instruction</a>
4295 <div class="doc_text">
4299 <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>]
4304 <p>The '<tt>call</tt>' instruction represents a simple function call.</p>
4308 <p>This instruction requires several arguments:</p>
4312 <p>The optional "tail" marker indicates whether the callee function accesses
4313 any allocas or varargs in the caller. If the "tail" marker is present, the
4314 function call is eligible for tail call optimization. Note that calls may
4315 be marked "tail" even if they do not occur before a <a
4316 href="#i_ret"><tt>ret</tt></a> instruction.</p>
4319 <p>The optional "cconv" marker indicates which <a href="#callingconv">calling
4320 convention</a> the call should use. If none is specified, the call defaults
4321 to using C calling conventions.</p>
4325 <p>The optional <a href="#paramattrs">Parameter Attributes</a> list for
4326 return values. Only '<tt>zeroext</tt>', '<tt>signext</tt>',
4327 and '<tt>inreg</tt>' attributes are valid here.</p>
4331 <p>'<tt>ty</tt>': the type of the call instruction itself which is also
4332 the type of the return value. Functions that return no value are marked
4333 <tt><a href="#t_void">void</a></tt>.</p>
4336 <p>'<tt>fnty</tt>': shall be the signature of the pointer to function
4337 value being invoked. The argument types must match the types implied by
4338 this signature. This type can be omitted if the function is not varargs
4339 and if the function type does not return a pointer to a function.</p>
4342 <p>'<tt>fnptrval</tt>': An LLVM value containing a pointer to a function to
4343 be invoked. In most cases, this is a direct function invocation, but
4344 indirect <tt>call</tt>s are just as possible, calling an arbitrary pointer
4345 to function value.</p>
4348 <p>'<tt>function args</tt>': argument list whose types match the
4349 function signature argument types. All arguments must be of
4350 <a href="#t_firstclass">first class</a> type. If the function signature
4351 indicates the function accepts a variable number of arguments, the extra
4352 arguments can be specified.</p>
4355 <p>The optional <a href="#fnattrs">function attributes</a> list. Only
4356 '<tt>noreturn</tt>', '<tt>nounwind</tt>', '<tt>readonly</tt>' and
4357 '<tt>readnone</tt>' attributes are valid here.</p>
4363 <p>The '<tt>call</tt>' instruction is used to cause control flow to
4364 transfer to a specified function, with its incoming arguments bound to
4365 the specified values. Upon a '<tt><a href="#i_ret">ret</a></tt>'
4366 instruction in the called function, control flow continues with the
4367 instruction after the function call, and the return value of the
4368 function is bound to the result argument.</p>
4373 %retval = call i32 @test(i32 %argc)
4374 call i32 (i8 *, ...)* @printf(i8 * %msg, i32 12, i8 42) <i>; yields i32</i>
4375 %X = tail call i32 @foo() <i>; yields i32</i>
4376 %Y = tail call <a href="#callingconv">fastcc</a> i32 @foo() <i>; yields i32</i>
4377 call void %foo(i8 97 signext)
4379 %struct.A = type { i32, i8 }
4380 %r = call %struct.A @foo() <i>; yields { 32, i8 }</i>
4381 %gr = extractvalue %struct.A %r, 0 <i>; yields i32</i>
4382 %gr1 = extractvalue %struct.A %r, 1 <i>; yields i8</i>
4383 %Z = call void @foo() noreturn <i>; indicates that %foo never returns normally</i>
4384 %ZZ = call zeroext i32 @bar() <i>; Return value is %zero extended</i>
4389 <!-- _______________________________________________________________________ -->
4390 <div class="doc_subsubsection">
4391 <a name="i_va_arg">'<tt>va_arg</tt>' Instruction</a>
4394 <div class="doc_text">
4399 <resultval> = va_arg <va_list*> <arglist>, <argty>
4404 <p>The '<tt>va_arg</tt>' instruction is used to access arguments passed through
4405 the "variable argument" area of a function call. It is used to implement the
4406 <tt>va_arg</tt> macro in C.</p>
4410 <p>This instruction takes a <tt>va_list*</tt> value and the type of
4411 the argument. It returns a value of the specified argument type and
4412 increments the <tt>va_list</tt> to point to the next argument. The
4413 actual type of <tt>va_list</tt> is target specific.</p>
4417 <p>The '<tt>va_arg</tt>' instruction loads an argument of the specified
4418 type from the specified <tt>va_list</tt> and causes the
4419 <tt>va_list</tt> to point to the next argument. For more information,
4420 see the variable argument handling <a href="#int_varargs">Intrinsic
4423 <p>It is legal for this instruction to be called in a function which does not
4424 take a variable number of arguments, for example, the <tt>vfprintf</tt>
4427 <p><tt>va_arg</tt> is an LLVM instruction instead of an <a
4428 href="#intrinsics">intrinsic function</a> because it takes a type as an
4433 <p>See the <a href="#int_varargs">variable argument processing</a> section.</p>
4437 <!-- *********************************************************************** -->
4438 <div class="doc_section"> <a name="intrinsics">Intrinsic Functions</a> </div>
4439 <!-- *********************************************************************** -->
4441 <div class="doc_text">
4443 <p>LLVM supports the notion of an "intrinsic function". These functions have
4444 well known names and semantics and are required to follow certain restrictions.
4445 Overall, these intrinsics represent an extension mechanism for the LLVM
4446 language that does not require changing all of the transformations in LLVM when
4447 adding to the language (or the bitcode reader/writer, the parser, etc...).</p>
4449 <p>Intrinsic function names must all start with an "<tt>llvm.</tt>" prefix. This
4450 prefix is reserved in LLVM for intrinsic names; thus, function names may not
4451 begin with this prefix. Intrinsic functions must always be external functions:
4452 you cannot define the body of intrinsic functions. Intrinsic functions may
4453 only be used in call or invoke instructions: it is illegal to take the address
4454 of an intrinsic function. Additionally, because intrinsic functions are part
4455 of the LLVM language, it is required if any are added that they be documented
4458 <p>Some intrinsic functions can be overloaded, i.e., the intrinsic represents
4459 a family of functions that perform the same operation but on different data
4460 types. Because LLVM can represent over 8 million different integer types,
4461 overloading is used commonly to allow an intrinsic function to operate on any
4462 integer type. One or more of the argument types or the result type can be
4463 overloaded to accept any integer type. Argument types may also be defined as
4464 exactly matching a previous argument's type or the result type. This allows an
4465 intrinsic function which accepts multiple arguments, but needs all of them to
4466 be of the same type, to only be overloaded with respect to a single argument or
4469 <p>Overloaded intrinsics will have the names of its overloaded argument types
4470 encoded into its function name, each preceded by a period. Only those types
4471 which are overloaded result in a name suffix. Arguments whose type is matched
4472 against another type do not. For example, the <tt>llvm.ctpop</tt> function can
4473 take an integer of any width and returns an integer of exactly the same integer
4474 width. This leads to a family of functions such as
4475 <tt>i8 @llvm.ctpop.i8(i8 %val)</tt> and <tt>i29 @llvm.ctpop.i29(i29 %val)</tt>.
4476 Only one type, the return type, is overloaded, and only one type suffix is
4477 required. Because the argument's type is matched against the return type, it
4478 does not require its own name suffix.</p>
4480 <p>To learn how to add an intrinsic function, please see the
4481 <a href="ExtendingLLVM.html">Extending LLVM Guide</a>.
4486 <!-- ======================================================================= -->
4487 <div class="doc_subsection">
4488 <a name="int_varargs">Variable Argument Handling Intrinsics</a>
4491 <div class="doc_text">
4493 <p>Variable argument support is defined in LLVM with the <a
4494 href="#i_va_arg"><tt>va_arg</tt></a> instruction and these three
4495 intrinsic functions. These functions are related to the similarly
4496 named macros defined in the <tt><stdarg.h></tt> header file.</p>
4498 <p>All of these functions operate on arguments that use a
4499 target-specific value type "<tt>va_list</tt>". The LLVM assembly
4500 language reference manual does not define what this type is, so all
4501 transformations should be prepared to handle these functions regardless of
4504 <p>This example shows how the <a href="#i_va_arg"><tt>va_arg</tt></a>
4505 instruction and the variable argument handling intrinsic functions are
4508 <div class="doc_code">
4510 define i32 @test(i32 %X, ...) {
4511 ; Initialize variable argument processing
4513 %ap2 = bitcast i8** %ap to i8*
4514 call void @llvm.va_start(i8* %ap2)
4516 ; Read a single integer argument
4517 %tmp = va_arg i8** %ap, i32
4519 ; Demonstrate usage of llvm.va_copy and llvm.va_end
4521 %aq2 = bitcast i8** %aq to i8*
4522 call void @llvm.va_copy(i8* %aq2, i8* %ap2)
4523 call void @llvm.va_end(i8* %aq2)
4525 ; Stop processing of arguments.
4526 call void @llvm.va_end(i8* %ap2)
4530 declare void @llvm.va_start(i8*)
4531 declare void @llvm.va_copy(i8*, i8*)
4532 declare void @llvm.va_end(i8*)
4538 <!-- _______________________________________________________________________ -->
4539 <div class="doc_subsubsection">
4540 <a name="int_va_start">'<tt>llvm.va_start</tt>' Intrinsic</a>
4544 <div class="doc_text">
4546 <pre> declare void %llvm.va_start(i8* <arglist>)<br></pre>
4548 <p>The '<tt>llvm.va_start</tt>' intrinsic initializes
4549 <tt>*<arglist></tt> for subsequent use by <tt><a
4550 href="#i_va_arg">va_arg</a></tt>.</p>
4554 <p>The argument is a pointer to a <tt>va_list</tt> element to initialize.</p>
4558 <p>The '<tt>llvm.va_start</tt>' intrinsic works just like the <tt>va_start</tt>
4559 macro available in C. In a target-dependent way, it initializes the
4560 <tt>va_list</tt> element to which the argument points, so that the next call to
4561 <tt>va_arg</tt> will produce the first variable argument passed to the function.
4562 Unlike the C <tt>va_start</tt> macro, this intrinsic does not need to know the
4563 last argument of the function as the compiler can figure that out.</p>
4567 <!-- _______________________________________________________________________ -->
4568 <div class="doc_subsubsection">
4569 <a name="int_va_end">'<tt>llvm.va_end</tt>' Intrinsic</a>
4572 <div class="doc_text">
4574 <pre> declare void @llvm.va_end(i8* <arglist>)<br></pre>
4577 <p>The '<tt>llvm.va_end</tt>' intrinsic destroys <tt>*<arglist></tt>,
4578 which has been initialized previously with <tt><a href="#int_va_start">llvm.va_start</a></tt>
4579 or <tt><a href="#i_va_copy">llvm.va_copy</a></tt>.</p>
4583 <p>The argument is a pointer to a <tt>va_list</tt> to destroy.</p>
4587 <p>The '<tt>llvm.va_end</tt>' intrinsic works just like the <tt>va_end</tt>
4588 macro available in C. In a target-dependent way, it destroys the
4589 <tt>va_list</tt> element to which the argument points. Calls to <a
4590 href="#int_va_start"><tt>llvm.va_start</tt></a> and <a href="#int_va_copy">
4591 <tt>llvm.va_copy</tt></a> must be matched exactly with calls to
4592 <tt>llvm.va_end</tt>.</p>
4596 <!-- _______________________________________________________________________ -->
4597 <div class="doc_subsubsection">
4598 <a name="int_va_copy">'<tt>llvm.va_copy</tt>' Intrinsic</a>
4601 <div class="doc_text">
4606 declare void @llvm.va_copy(i8* <destarglist>, i8* <srcarglist>)
4611 <p>The '<tt>llvm.va_copy</tt>' intrinsic copies the current argument position
4612 from the source argument list to the destination argument list.</p>
4616 <p>The first argument is a pointer to a <tt>va_list</tt> element to initialize.
4617 The second argument is a pointer to a <tt>va_list</tt> element to copy from.</p>
4622 <p>The '<tt>llvm.va_copy</tt>' intrinsic works just like the <tt>va_copy</tt>
4623 macro available in C. In a target-dependent way, it copies the source
4624 <tt>va_list</tt> element into the destination <tt>va_list</tt> element. This
4625 intrinsic is necessary because the <tt><a href="#int_va_start">
4626 llvm.va_start</a></tt> intrinsic may be arbitrarily complex and require, for
4627 example, memory allocation.</p>
4631 <!-- ======================================================================= -->
4632 <div class="doc_subsection">
4633 <a name="int_gc">Accurate Garbage Collection Intrinsics</a>
4636 <div class="doc_text">
4639 LLVM support for <a href="GarbageCollection.html">Accurate Garbage
4640 Collection</a> (GC) requires the implementation and generation of these
4642 These intrinsics allow identification of <a href="#int_gcroot">GC roots on the
4643 stack</a>, as well as garbage collector implementations that require <a
4644 href="#int_gcread">read</a> and <a href="#int_gcwrite">write</a> barriers.
4645 Front-ends for type-safe garbage collected languages should generate these
4646 intrinsics to make use of the LLVM garbage collectors. For more details, see <a
4647 href="GarbageCollection.html">Accurate Garbage Collection with LLVM</a>.
4650 <p>The garbage collection intrinsics only operate on objects in the generic
4651 address space (address space zero).</p>
4655 <!-- _______________________________________________________________________ -->
4656 <div class="doc_subsubsection">
4657 <a name="int_gcroot">'<tt>llvm.gcroot</tt>' Intrinsic</a>
4660 <div class="doc_text">
4665 declare void @llvm.gcroot(i8** %ptrloc, i8* %metadata)
4670 <p>The '<tt>llvm.gcroot</tt>' intrinsic declares the existence of a GC root to
4671 the code generator, and allows some metadata to be associated with it.</p>
4675 <p>The first argument specifies the address of a stack object that contains the
4676 root pointer. The second pointer (which must be either a constant or a global
4677 value address) contains the meta-data to be associated with the root.</p>
4681 <p>At runtime, a call to this intrinsic stores a null pointer into the "ptrloc"
4682 location. At compile-time, the code generator generates information to allow
4683 the runtime to find the pointer at GC safe points. The '<tt>llvm.gcroot</tt>'
4684 intrinsic may only be used in a function which <a href="#gc">specifies a GC
4690 <!-- _______________________________________________________________________ -->
4691 <div class="doc_subsubsection">
4692 <a name="int_gcread">'<tt>llvm.gcread</tt>' Intrinsic</a>
4695 <div class="doc_text">
4700 declare i8* @llvm.gcread(i8* %ObjPtr, i8** %Ptr)
4705 <p>The '<tt>llvm.gcread</tt>' intrinsic identifies reads of references from heap
4706 locations, allowing garbage collector implementations that require read
4711 <p>The second argument is the address to read from, which should be an address
4712 allocated from the garbage collector. The first object is a pointer to the
4713 start of the referenced object, if needed by the language runtime (otherwise
4718 <p>The '<tt>llvm.gcread</tt>' intrinsic has the same semantics as a load
4719 instruction, but may be replaced with substantially more complex code by the
4720 garbage collector runtime, as needed. The '<tt>llvm.gcread</tt>' intrinsic
4721 may only be used in a function which <a href="#gc">specifies a GC
4727 <!-- _______________________________________________________________________ -->
4728 <div class="doc_subsubsection">
4729 <a name="int_gcwrite">'<tt>llvm.gcwrite</tt>' Intrinsic</a>
4732 <div class="doc_text">
4737 declare void @llvm.gcwrite(i8* %P1, i8* %Obj, i8** %P2)
4742 <p>The '<tt>llvm.gcwrite</tt>' intrinsic identifies writes of references to heap
4743 locations, allowing garbage collector implementations that require write
4744 barriers (such as generational or reference counting collectors).</p>
4748 <p>The first argument is the reference to store, the second is the start of the
4749 object to store it to, and the third is the address of the field of Obj to
4750 store to. If the runtime does not require a pointer to the object, Obj may be
4755 <p>The '<tt>llvm.gcwrite</tt>' intrinsic has the same semantics as a store
4756 instruction, but may be replaced with substantially more complex code by the
4757 garbage collector runtime, as needed. The '<tt>llvm.gcwrite</tt>' intrinsic
4758 may only be used in a function which <a href="#gc">specifies a GC
4765 <!-- ======================================================================= -->
4766 <div class="doc_subsection">
4767 <a name="int_codegen">Code Generator Intrinsics</a>
4770 <div class="doc_text">
4772 These intrinsics are provided by LLVM to expose special features that may only
4773 be implemented with code generator support.
4778 <!-- _______________________________________________________________________ -->
4779 <div class="doc_subsubsection">
4780 <a name="int_returnaddress">'<tt>llvm.returnaddress</tt>' Intrinsic</a>
4783 <div class="doc_text">
4787 declare i8 *@llvm.returnaddress(i32 <level>)
4793 The '<tt>llvm.returnaddress</tt>' intrinsic attempts to compute a
4794 target-specific value indicating the return address of the current function
4795 or one of its callers.
4801 The argument to this intrinsic indicates which function to return the address
4802 for. Zero indicates the calling function, one indicates its caller, etc. The
4803 argument is <b>required</b> to be a constant integer value.
4809 The '<tt>llvm.returnaddress</tt>' intrinsic either returns a pointer indicating
4810 the return address of the specified call frame, or zero if it cannot be
4811 identified. The value returned by this intrinsic is likely to be incorrect or 0
4812 for arguments other than zero, so it should only be used for debugging purposes.
4816 Note that calling this intrinsic does not prevent function inlining or other
4817 aggressive transformations, so the value returned may not be that of the obvious
4818 source-language caller.
4823 <!-- _______________________________________________________________________ -->
4824 <div class="doc_subsubsection">
4825 <a name="int_frameaddress">'<tt>llvm.frameaddress</tt>' Intrinsic</a>
4828 <div class="doc_text">
4832 declare i8 *@llvm.frameaddress(i32 <level>)
4838 The '<tt>llvm.frameaddress</tt>' intrinsic attempts to return the
4839 target-specific frame pointer value for the specified stack frame.
4845 The argument to this intrinsic indicates which function to return the frame
4846 pointer for. Zero indicates the calling function, one indicates its caller,
4847 etc. The argument is <b>required</b> to be a constant integer value.
4853 The '<tt>llvm.frameaddress</tt>' intrinsic either returns a pointer indicating
4854 the frame address of the specified call frame, or zero if it cannot be
4855 identified. The value returned by this intrinsic is likely to be incorrect or 0
4856 for arguments other than zero, so it should only be used for debugging purposes.
4860 Note that calling this intrinsic does not prevent function inlining or other
4861 aggressive transformations, so the value returned may not be that of the obvious
4862 source-language caller.
4866 <!-- _______________________________________________________________________ -->
4867 <div class="doc_subsubsection">
4868 <a name="int_stacksave">'<tt>llvm.stacksave</tt>' Intrinsic</a>
4871 <div class="doc_text">
4875 declare i8 *@llvm.stacksave()
4881 The '<tt>llvm.stacksave</tt>' intrinsic is used to remember the current state of
4882 the function stack, for use with <a href="#int_stackrestore">
4883 <tt>llvm.stackrestore</tt></a>. This is useful for implementing language
4884 features like scoped automatic variable sized arrays in C99.
4890 This intrinsic returns a opaque pointer value that can be passed to <a
4891 href="#int_stackrestore"><tt>llvm.stackrestore</tt></a>. When an
4892 <tt>llvm.stackrestore</tt> intrinsic is executed with a value saved from
4893 <tt>llvm.stacksave</tt>, it effectively restores the state of the stack to the
4894 state it was in when the <tt>llvm.stacksave</tt> intrinsic executed. In
4895 practice, this pops any <a href="#i_alloca">alloca</a> blocks from the stack
4896 that were allocated after the <tt>llvm.stacksave</tt> was executed.
4901 <!-- _______________________________________________________________________ -->
4902 <div class="doc_subsubsection">
4903 <a name="int_stackrestore">'<tt>llvm.stackrestore</tt>' Intrinsic</a>
4906 <div class="doc_text">
4910 declare void @llvm.stackrestore(i8 * %ptr)
4916 The '<tt>llvm.stackrestore</tt>' intrinsic is used to restore the state of
4917 the function stack to the state it was in when the corresponding <a
4918 href="#int_stacksave"><tt>llvm.stacksave</tt></a> intrinsic executed. This is
4919 useful for implementing language features like scoped automatic variable sized
4926 See the description for <a href="#int_stacksave"><tt>llvm.stacksave</tt></a>.
4932 <!-- _______________________________________________________________________ -->
4933 <div class="doc_subsubsection">
4934 <a name="int_prefetch">'<tt>llvm.prefetch</tt>' Intrinsic</a>
4937 <div class="doc_text">
4941 declare void @llvm.prefetch(i8* <address>, i32 <rw>, i32 <locality>)
4948 The '<tt>llvm.prefetch</tt>' intrinsic is a hint to the code generator to insert
4949 a prefetch instruction if supported; otherwise, it is a noop. Prefetches have
4951 effect on the behavior of the program but can change its performance
4958 <tt>address</tt> is the address to be prefetched, <tt>rw</tt> is the specifier
4959 determining if the fetch should be for a read (0) or write (1), and
4960 <tt>locality</tt> is a temporal locality specifier ranging from (0) - no
4961 locality, to (3) - extremely local keep in cache. The <tt>rw</tt> and
4962 <tt>locality</tt> arguments must be constant integers.
4968 This intrinsic does not modify the behavior of the program. In particular,
4969 prefetches cannot trap and do not produce a value. On targets that support this
4970 intrinsic, the prefetch can provide hints to the processor cache for better
4976 <!-- _______________________________________________________________________ -->
4977 <div class="doc_subsubsection">
4978 <a name="int_pcmarker">'<tt>llvm.pcmarker</tt>' Intrinsic</a>
4981 <div class="doc_text">
4985 declare void @llvm.pcmarker(i32 <id>)
4992 The '<tt>llvm.pcmarker</tt>' intrinsic is a method to export a Program Counter
4994 code to simulators and other tools. The method is target specific, but it is
4995 expected that the marker will use exported symbols to transmit the PC of the
4997 The marker makes no guarantees that it will remain with any specific instruction
4998 after optimizations. It is possible that the presence of a marker will inhibit
4999 optimizations. The intended use is to be inserted after optimizations to allow
5000 correlations of simulation runs.
5006 <tt>id</tt> is a numerical id identifying the marker.
5012 This intrinsic does not modify the behavior of the program. Backends that do not
5013 support this intrinisic may ignore it.
5018 <!-- _______________________________________________________________________ -->
5019 <div class="doc_subsubsection">
5020 <a name="int_readcyclecounter">'<tt>llvm.readcyclecounter</tt>' Intrinsic</a>
5023 <div class="doc_text">
5027 declare i64 @llvm.readcyclecounter( )
5034 The '<tt>llvm.readcyclecounter</tt>' intrinsic provides access to the cycle
5035 counter register (or similar low latency, high accuracy clocks) on those targets
5036 that support it. On X86, it should map to RDTSC. On Alpha, it should map to RPCC.
5037 As the backing counters overflow quickly (on the order of 9 seconds on alpha), this
5038 should only be used for small timings.
5044 When directly supported, reading the cycle counter should not modify any memory.
5045 Implementations are allowed to either return a application specific value or a
5046 system wide value. On backends without support, this is lowered to a constant 0.
5051 <!-- ======================================================================= -->
5052 <div class="doc_subsection">
5053 <a name="int_libc">Standard C Library Intrinsics</a>
5056 <div class="doc_text">
5058 LLVM provides intrinsics for a few important standard C library functions.
5059 These intrinsics allow source-language front-ends to pass information about the
5060 alignment of the pointer arguments to the code generator, providing opportunity
5061 for more efficient code generation.
5066 <!-- _______________________________________________________________________ -->
5067 <div class="doc_subsubsection">
5068 <a name="int_memcpy">'<tt>llvm.memcpy</tt>' Intrinsic</a>
5071 <div class="doc_text">
5074 <p>This is an overloaded intrinsic. You can use llvm.memcpy on any integer bit
5075 width. Not all targets support all bit widths however.</p>
5077 declare void @llvm.memcpy.i8(i8 * <dest>, i8 * <src>,
5078 i8 <len>, i32 <align>)
5079 declare void @llvm.memcpy.i16(i8 * <dest>, i8 * <src>,
5080 i16 <len>, i32 <align>)
5081 declare void @llvm.memcpy.i32(i8 * <dest>, i8 * <src>,
5082 i32 <len>, i32 <align>)
5083 declare void @llvm.memcpy.i64(i8 * <dest>, i8 * <src>,
5084 i64 <len>, i32 <align>)
5090 The '<tt>llvm.memcpy.*</tt>' intrinsics copy a block of memory from the source
5091 location to the destination location.
5095 Note that, unlike the standard libc function, the <tt>llvm.memcpy.*</tt>
5096 intrinsics do not return a value, and takes an extra alignment argument.
5102 The first argument is a pointer to the destination, the second is a pointer to
5103 the source. The third argument is an integer argument
5104 specifying the number of bytes to copy, and the fourth argument is the alignment
5105 of the source and destination locations.
5109 If the call to this intrinisic has an alignment value that is not 0 or 1, then
5110 the caller guarantees that both the source and destination pointers are aligned
5117 The '<tt>llvm.memcpy.*</tt>' intrinsics copy a block of memory from the source
5118 location to the destination location, which are not allowed to overlap. It
5119 copies "len" bytes of memory over. If the argument is known to be aligned to
5120 some boundary, this can be specified as the fourth argument, otherwise it should
5126 <!-- _______________________________________________________________________ -->
5127 <div class="doc_subsubsection">
5128 <a name="int_memmove">'<tt>llvm.memmove</tt>' Intrinsic</a>
5131 <div class="doc_text">
5134 <p>This is an overloaded intrinsic. You can use llvm.memmove on any integer bit
5135 width. Not all targets support all bit widths however.</p>
5137 declare void @llvm.memmove.i8(i8 * <dest>, i8 * <src>,
5138 i8 <len>, i32 <align>)
5139 declare void @llvm.memmove.i16(i8 * <dest>, i8 * <src>,
5140 i16 <len>, i32 <align>)
5141 declare void @llvm.memmove.i32(i8 * <dest>, i8 * <src>,
5142 i32 <len>, i32 <align>)
5143 declare void @llvm.memmove.i64(i8 * <dest>, i8 * <src>,
5144 i64 <len>, i32 <align>)
5150 The '<tt>llvm.memmove.*</tt>' intrinsics move a block of memory from the source
5151 location to the destination location. It is similar to the
5152 '<tt>llvm.memcpy</tt>' intrinsic but allows the two memory locations to overlap.
5156 Note that, unlike the standard libc function, the <tt>llvm.memmove.*</tt>
5157 intrinsics do not return a value, and takes an extra alignment argument.
5163 The first argument is a pointer to the destination, the second is a pointer to
5164 the source. The third argument is an integer argument
5165 specifying the number of bytes to copy, and the fourth argument is the alignment
5166 of the source and destination locations.
5170 If the call to this intrinisic has an alignment value that is not 0 or 1, then
5171 the caller guarantees that the source and destination pointers are aligned to
5178 The '<tt>llvm.memmove.*</tt>' intrinsics copy a block of memory from the source
5179 location to the destination location, which may overlap. It
5180 copies "len" bytes of memory over. If the argument is known to be aligned to
5181 some boundary, this can be specified as the fourth argument, otherwise it should
5187 <!-- _______________________________________________________________________ -->
5188 <div class="doc_subsubsection">
5189 <a name="int_memset">'<tt>llvm.memset.*</tt>' Intrinsics</a>
5192 <div class="doc_text">
5195 <p>This is an overloaded intrinsic. You can use llvm.memset on any integer bit
5196 width. Not all targets support all bit widths however.</p>
5198 declare void @llvm.memset.i8(i8 * <dest>, i8 <val>,
5199 i8 <len>, i32 <align>)
5200 declare void @llvm.memset.i16(i8 * <dest>, i8 <val>,
5201 i16 <len>, i32 <align>)
5202 declare void @llvm.memset.i32(i8 * <dest>, i8 <val>,
5203 i32 <len>, i32 <align>)
5204 declare void @llvm.memset.i64(i8 * <dest>, i8 <val>,
5205 i64 <len>, i32 <align>)
5211 The '<tt>llvm.memset.*</tt>' intrinsics fill a block of memory with a particular
5216 Note that, unlike the standard libc function, the <tt>llvm.memset</tt> intrinsic
5217 does not return a value, and takes an extra alignment argument.
5223 The first argument is a pointer to the destination to fill, the second is the
5224 byte value to fill it with, the third argument is an integer
5225 argument specifying the number of bytes to fill, and the fourth argument is the
5226 known alignment of destination location.
5230 If the call to this intrinisic has an alignment value that is not 0 or 1, then
5231 the caller guarantees that the destination pointer is aligned to that boundary.
5237 The '<tt>llvm.memset.*</tt>' intrinsics fill "len" bytes of memory starting at
5239 destination location. If the argument is known to be aligned to some boundary,
5240 this can be specified as the fourth argument, otherwise it should be set to 0 or
5246 <!-- _______________________________________________________________________ -->
5247 <div class="doc_subsubsection">
5248 <a name="int_sqrt">'<tt>llvm.sqrt.*</tt>' Intrinsic</a>
5251 <div class="doc_text">
5254 <p>This is an overloaded intrinsic. You can use <tt>llvm.sqrt</tt> on any
5255 floating point or vector of floating point type. Not all targets support all
5258 declare float @llvm.sqrt.f32(float %Val)
5259 declare double @llvm.sqrt.f64(double %Val)
5260 declare x86_fp80 @llvm.sqrt.f80(x86_fp80 %Val)
5261 declare fp128 @llvm.sqrt.f128(fp128 %Val)
5262 declare ppc_fp128 @llvm.sqrt.ppcf128(ppc_fp128 %Val)
5268 The '<tt>llvm.sqrt</tt>' intrinsics return the sqrt of the specified operand,
5269 returning the same value as the libm '<tt>sqrt</tt>' functions would. Unlike
5270 <tt>sqrt</tt> in libm, however, <tt>llvm.sqrt</tt> has undefined behavior for
5271 negative numbers other than -0.0 (which allows for better optimization, because
5272 there is no need to worry about errno being set). <tt>llvm.sqrt(-0.0)</tt> is
5273 defined to return -0.0 like IEEE sqrt.
5279 The argument and return value are floating point numbers of the same type.
5285 This function returns the sqrt of the specified operand if it is a nonnegative
5286 floating point number.
5290 <!-- _______________________________________________________________________ -->
5291 <div class="doc_subsubsection">
5292 <a name="int_powi">'<tt>llvm.powi.*</tt>' Intrinsic</a>
5295 <div class="doc_text">
5298 <p>This is an overloaded intrinsic. You can use <tt>llvm.powi</tt> on any
5299 floating point or vector of floating point type. Not all targets support all
5302 declare float @llvm.powi.f32(float %Val, i32 %power)
5303 declare double @llvm.powi.f64(double %Val, i32 %power)
5304 declare x86_fp80 @llvm.powi.f80(x86_fp80 %Val, i32 %power)
5305 declare fp128 @llvm.powi.f128(fp128 %Val, i32 %power)
5306 declare ppc_fp128 @llvm.powi.ppcf128(ppc_fp128 %Val, i32 %power)
5312 The '<tt>llvm.powi.*</tt>' intrinsics return the first operand raised to the
5313 specified (positive or negative) power. The order of evaluation of
5314 multiplications is not defined. When a vector of floating point type is
5315 used, the second argument remains a scalar integer value.
5321 The second argument is an integer power, and the first is a value to raise to
5328 This function returns the first value raised to the second power with an
5329 unspecified sequence of rounding operations.</p>
5332 <!-- _______________________________________________________________________ -->
5333 <div class="doc_subsubsection">
5334 <a name="int_sin">'<tt>llvm.sin.*</tt>' Intrinsic</a>
5337 <div class="doc_text">
5340 <p>This is an overloaded intrinsic. You can use <tt>llvm.sin</tt> on any
5341 floating point or vector of floating point type. Not all targets support all
5344 declare float @llvm.sin.f32(float %Val)
5345 declare double @llvm.sin.f64(double %Val)
5346 declare x86_fp80 @llvm.sin.f80(x86_fp80 %Val)
5347 declare fp128 @llvm.sin.f128(fp128 %Val)
5348 declare ppc_fp128 @llvm.sin.ppcf128(ppc_fp128 %Val)
5354 The '<tt>llvm.sin.*</tt>' intrinsics return the sine of the operand.
5360 The argument and return value are floating point numbers of the same type.
5366 This function returns the sine of the specified operand, returning the
5367 same values as the libm <tt>sin</tt> functions would, and handles error
5368 conditions in the same way.</p>
5371 <!-- _______________________________________________________________________ -->
5372 <div class="doc_subsubsection">
5373 <a name="int_cos">'<tt>llvm.cos.*</tt>' Intrinsic</a>
5376 <div class="doc_text">
5379 <p>This is an overloaded intrinsic. You can use <tt>llvm.cos</tt> on any
5380 floating point or vector of floating point type. Not all targets support all
5383 declare float @llvm.cos.f32(float %Val)
5384 declare double @llvm.cos.f64(double %Val)
5385 declare x86_fp80 @llvm.cos.f80(x86_fp80 %Val)
5386 declare fp128 @llvm.cos.f128(fp128 %Val)
5387 declare ppc_fp128 @llvm.cos.ppcf128(ppc_fp128 %Val)
5393 The '<tt>llvm.cos.*</tt>' intrinsics return the cosine of the operand.
5399 The argument and return value are floating point numbers of the same type.
5405 This function returns the cosine of the specified operand, returning the
5406 same values as the libm <tt>cos</tt> functions would, and handles error
5407 conditions in the same way.</p>
5410 <!-- _______________________________________________________________________ -->
5411 <div class="doc_subsubsection">
5412 <a name="int_pow">'<tt>llvm.pow.*</tt>' Intrinsic</a>
5415 <div class="doc_text">
5418 <p>This is an overloaded intrinsic. You can use <tt>llvm.pow</tt> on any
5419 floating point or vector of floating point type. Not all targets support all
5422 declare float @llvm.pow.f32(float %Val, float %Power)
5423 declare double @llvm.pow.f64(double %Val, double %Power)
5424 declare x86_fp80 @llvm.pow.f80(x86_fp80 %Val, x86_fp80 %Power)
5425 declare fp128 @llvm.pow.f128(fp128 %Val, fp128 %Power)
5426 declare ppc_fp128 @llvm.pow.ppcf128(ppc_fp128 %Val, ppc_fp128 Power)
5432 The '<tt>llvm.pow.*</tt>' intrinsics return the first operand raised to the
5433 specified (positive or negative) power.
5439 The second argument is a floating point power, and the first is a value to
5440 raise to that power.
5446 This function returns the first value raised to the second power,
5448 same values as the libm <tt>pow</tt> functions would, and handles error
5449 conditions in the same way.</p>
5453 <!-- ======================================================================= -->
5454 <div class="doc_subsection">
5455 <a name="int_manip">Bit Manipulation Intrinsics</a>
5458 <div class="doc_text">
5460 LLVM provides intrinsics for a few important bit manipulation operations.
5461 These allow efficient code generation for some algorithms.
5466 <!-- _______________________________________________________________________ -->
5467 <div class="doc_subsubsection">
5468 <a name="int_bswap">'<tt>llvm.bswap.*</tt>' Intrinsics</a>
5471 <div class="doc_text">
5474 <p>This is an overloaded intrinsic function. You can use bswap on any integer
5475 type that is an even number of bytes (i.e. BitWidth % 16 == 0).</p>
5477 declare i16 @llvm.bswap.i16(i16 <id>)
5478 declare i32 @llvm.bswap.i32(i32 <id>)
5479 declare i64 @llvm.bswap.i64(i64 <id>)
5485 The '<tt>llvm.bswap</tt>' family of intrinsics is used to byte swap integer
5486 values with an even number of bytes (positive multiple of 16 bits). These are
5487 useful for performing operations on data that is not in the target's native
5494 The <tt>llvm.bswap.i16</tt> intrinsic returns an i16 value that has the high
5495 and low byte of the input i16 swapped. Similarly, the <tt>llvm.bswap.i32</tt>
5496 intrinsic returns an i32 value that has the four bytes of the input i32
5497 swapped, so that if the input bytes are numbered 0, 1, 2, 3 then the returned
5498 i32 will have its bytes in 3, 2, 1, 0 order. The <tt>llvm.bswap.i48</tt>,
5499 <tt>llvm.bswap.i64</tt> and other intrinsics extend this concept to
5500 additional even-byte lengths (6 bytes, 8 bytes and more, respectively).
5505 <!-- _______________________________________________________________________ -->
5506 <div class="doc_subsubsection">
5507 <a name="int_ctpop">'<tt>llvm.ctpop.*</tt>' Intrinsic</a>
5510 <div class="doc_text">
5513 <p>This is an overloaded intrinsic. You can use llvm.ctpop on any integer bit
5514 width. Not all targets support all bit widths however.</p>
5516 declare i8 @llvm.ctpop.i8 (i8 <src>)
5517 declare i16 @llvm.ctpop.i16(i16 <src>)
5518 declare i32 @llvm.ctpop.i32(i32 <src>)
5519 declare i64 @llvm.ctpop.i64(i64 <src>)
5520 declare i256 @llvm.ctpop.i256(i256 <src>)
5526 The '<tt>llvm.ctpop</tt>' family of intrinsics counts the number of bits set in a
5533 The only argument is the value to be counted. The argument may be of any
5534 integer type. The return type must match the argument type.
5540 The '<tt>llvm.ctpop</tt>' intrinsic counts the 1's in a variable.
5544 <!-- _______________________________________________________________________ -->
5545 <div class="doc_subsubsection">
5546 <a name="int_ctlz">'<tt>llvm.ctlz.*</tt>' Intrinsic</a>
5549 <div class="doc_text">
5552 <p>This is an overloaded intrinsic. You can use <tt>llvm.ctlz</tt> on any
5553 integer bit width. Not all targets support all bit widths however.</p>
5555 declare i8 @llvm.ctlz.i8 (i8 <src>)
5556 declare i16 @llvm.ctlz.i16(i16 <src>)
5557 declare i32 @llvm.ctlz.i32(i32 <src>)
5558 declare i64 @llvm.ctlz.i64(i64 <src>)
5559 declare i256 @llvm.ctlz.i256(i256 <src>)
5565 The '<tt>llvm.ctlz</tt>' family of intrinsic functions counts the number of
5566 leading zeros in a variable.
5572 The only argument is the value to be counted. The argument may be of any
5573 integer type. The return type must match the argument type.
5579 The '<tt>llvm.ctlz</tt>' intrinsic counts the leading (most significant) zeros
5580 in a variable. If the src == 0 then the result is the size in bits of the type
5581 of src. For example, <tt>llvm.ctlz(i32 2) = 30</tt>.
5587 <!-- _______________________________________________________________________ -->
5588 <div class="doc_subsubsection">
5589 <a name="int_cttz">'<tt>llvm.cttz.*</tt>' Intrinsic</a>
5592 <div class="doc_text">
5595 <p>This is an overloaded intrinsic. You can use <tt>llvm.cttz</tt> on any
5596 integer bit width. Not all targets support all bit widths however.</p>
5598 declare i8 @llvm.cttz.i8 (i8 <src>)
5599 declare i16 @llvm.cttz.i16(i16 <src>)
5600 declare i32 @llvm.cttz.i32(i32 <src>)
5601 declare i64 @llvm.cttz.i64(i64 <src>)
5602 declare i256 @llvm.cttz.i256(i256 <src>)
5608 The '<tt>llvm.cttz</tt>' family of intrinsic functions counts the number of
5615 The only argument is the value to be counted. The argument may be of any
5616 integer type. The return type must match the argument type.
5622 The '<tt>llvm.cttz</tt>' intrinsic counts the trailing (least significant) zeros
5623 in a variable. If the src == 0 then the result is the size in bits of the type
5624 of src. For example, <tt>llvm.cttz(2) = 1</tt>.
5628 <!-- _______________________________________________________________________ -->
5629 <div class="doc_subsubsection">
5630 <a name="int_part_select">'<tt>llvm.part.select.*</tt>' Intrinsic</a>
5633 <div class="doc_text">
5636 <p>This is an overloaded intrinsic. You can use <tt>llvm.part.select</tt>
5637 on any integer bit width.</p>
5639 declare i17 @llvm.part.select.i17 (i17 %val, i32 %loBit, i32 %hiBit)
5640 declare i29 @llvm.part.select.i29 (i29 %val, i32 %loBit, i32 %hiBit)
5644 <p>The '<tt>llvm.part.select</tt>' family of intrinsic functions selects a
5645 range of bits from an integer value and returns them in the same bit width as
5646 the original value.</p>
5649 <p>The first argument, <tt>%val</tt> and the result may be integer types of
5650 any bit width but they must have the same bit width. The second and third
5651 arguments must be <tt>i32</tt> type since they specify only a bit index.</p>
5654 <p>The operation of the '<tt>llvm.part.select</tt>' intrinsic has two modes
5655 of operation: forwards and reverse. If <tt>%loBit</tt> is greater than
5656 <tt>%hiBits</tt> then the intrinsic operates in reverse mode. Otherwise it
5657 operates in forward mode.</p>
5658 <p>In forward mode, this intrinsic is the equivalent of shifting <tt>%val</tt>
5659 right by <tt>%loBit</tt> bits and then ANDing it with a mask with
5660 only the <tt>%hiBit - %loBit</tt> bits set, as follows:</p>
5662 <li>The <tt>%val</tt> is shifted right (LSHR) by the number of bits specified
5663 by <tt>%loBits</tt>. This normalizes the value to the low order bits.</li>
5664 <li>The <tt>%loBits</tt> value is subtracted from the <tt>%hiBits</tt> value
5665 to determine the number of bits to retain.</li>
5666 <li>A mask of the retained bits is created by shifting a -1 value.</li>
5667 <li>The mask is ANDed with <tt>%val</tt> to produce the result.</li>
5669 <p>In reverse mode, a similar computation is made except that the bits are
5670 returned in the reverse order. So, for example, if <tt>X</tt> has the value
5671 <tt>i16 0x0ACF (101011001111)</tt> and we apply
5672 <tt>part.select(i16 X, 8, 3)</tt> to it, we get back the value
5673 <tt>i16 0x0026 (000000100110)</tt>.</p>
5676 <div class="doc_subsubsection">
5677 <a name="int_part_set">'<tt>llvm.part.set.*</tt>' Intrinsic</a>
5680 <div class="doc_text">
5683 <p>This is an overloaded intrinsic. You can use <tt>llvm.part.set</tt>
5684 on any integer bit width.</p>
5686 declare i17 @llvm.part.set.i17.i9 (i17 %val, i9 %repl, i32 %lo, i32 %hi)
5687 declare i29 @llvm.part.set.i29.i9 (i29 %val, i9 %repl, i32 %lo, i32 %hi)
5691 <p>The '<tt>llvm.part.set</tt>' family of intrinsic functions replaces a range
5692 of bits in an integer value with another integer value. It returns the integer
5693 with the replaced bits.</p>
5696 <p>The first argument, <tt>%val</tt> and the result may be integer types of
5697 any bit width but they must have the same bit width. <tt>%val</tt> is the value
5698 whose bits will be replaced. The second argument, <tt>%repl</tt> may be an
5699 integer of any bit width. The third and fourth arguments must be <tt>i32</tt>
5700 type since they specify only a bit index.</p>
5703 <p>The operation of the '<tt>llvm.part.set</tt>' intrinsic has two modes
5704 of operation: forwards and reverse. If <tt>%lo</tt> is greater than
5705 <tt>%hi</tt> then the intrinsic operates in reverse mode. Otherwise it
5706 operates in forward mode.</p>
5707 <p>For both modes, the <tt>%repl</tt> value is prepared for use by either
5708 truncating it down to the size of the replacement area or zero extending it
5709 up to that size.</p>
5710 <p>In forward mode, the bits between <tt>%lo</tt> and <tt>%hi</tt> (inclusive)
5711 are replaced with corresponding bits from <tt>%repl</tt>. That is the 0th bit
5712 in <tt>%repl</tt> replaces the <tt>%lo</tt>th bit in <tt>%val</tt> and etc. up
5713 to the <tt>%hi</tt>th bit.</p>
5714 <p>In reverse mode, a similar computation is made except that the bits are
5715 reversed. That is, the <tt>0</tt>th bit in <tt>%repl</tt> replaces the
5716 <tt>%hi</tt> bit in <tt>%val</tt> and etc. down to the <tt>%lo</tt>th bit.</p>
5719 llvm.part.set(0xFFFF, 0, 4, 7) -> 0xFF0F
5720 llvm.part.set(0xFFFF, 0, 7, 4) -> 0xFF0F
5721 llvm.part.set(0xFFFF, 1, 7, 4) -> 0xFF8F
5722 llvm.part.set(0xFFFF, F, 8, 3) -> 0xFFE7
5723 llvm.part.set(0xFFFF, 0, 3, 8) -> 0xFE07
5727 <!-- ======================================================================= -->
5728 <div class="doc_subsection">
5729 <a name="int_debugger">Debugger Intrinsics</a>
5732 <div class="doc_text">
5734 The LLVM debugger intrinsics (which all start with <tt>llvm.dbg.</tt> prefix),
5735 are described in the <a
5736 href="SourceLevelDebugging.html#format_common_intrinsics">LLVM Source Level
5737 Debugging</a> document.
5742 <!-- ======================================================================= -->
5743 <div class="doc_subsection">
5744 <a name="int_eh">Exception Handling Intrinsics</a>
5747 <div class="doc_text">
5748 <p> The LLVM exception handling intrinsics (which all start with
5749 <tt>llvm.eh.</tt> prefix), are described in the <a
5750 href="ExceptionHandling.html#format_common_intrinsics">LLVM Exception
5751 Handling</a> document. </p>
5754 <!-- ======================================================================= -->
5755 <div class="doc_subsection">
5756 <a name="int_trampoline">Trampoline Intrinsic</a>
5759 <div class="doc_text">
5761 This intrinsic makes it possible to excise one parameter, marked with
5762 the <tt>nest</tt> attribute, from a function. The result is a callable
5763 function pointer lacking the nest parameter - the caller does not need
5764 to provide a value for it. Instead, the value to use is stored in
5765 advance in a "trampoline", a block of memory usually allocated
5766 on the stack, which also contains code to splice the nest value into the
5767 argument list. This is used to implement the GCC nested function address
5771 For example, if the function is
5772 <tt>i32 f(i8* nest %c, i32 %x, i32 %y)</tt> then the resulting function
5773 pointer has signature <tt>i32 (i32, i32)*</tt>. It can be created as follows:</p>
5775 %tramp = alloca [10 x i8], align 4 ; size and alignment only correct for X86
5776 %tramp1 = getelementptr [10 x i8]* %tramp, i32 0, i32 0
5777 %p = call i8* @llvm.init.trampoline( i8* %tramp1, i8* bitcast (i32 (i8* nest , i32, i32)* @f to i8*), i8* %nval )
5778 %fp = bitcast i8* %p to i32 (i32, i32)*
5780 <p>The call <tt>%val = call i32 %fp( i32 %x, i32 %y )</tt> is then equivalent
5781 to <tt>%val = call i32 %f( i8* %nval, i32 %x, i32 %y )</tt>.</p>
5784 <!-- _______________________________________________________________________ -->
5785 <div class="doc_subsubsection">
5786 <a name="int_it">'<tt>llvm.init.trampoline</tt>' Intrinsic</a>
5788 <div class="doc_text">
5791 declare i8* @llvm.init.trampoline(i8* <tramp>, i8* <func>, i8* <nval>)
5795 This fills the memory pointed to by <tt>tramp</tt> with code
5796 and returns a function pointer suitable for executing it.
5800 The <tt>llvm.init.trampoline</tt> intrinsic takes three arguments, all
5801 pointers. The <tt>tramp</tt> argument must point to a sufficiently large
5802 and sufficiently aligned block of memory; this memory is written to by the
5803 intrinsic. Note that the size and the alignment are target-specific - LLVM
5804 currently provides no portable way of determining them, so a front-end that
5805 generates this intrinsic needs to have some target-specific knowledge.
5806 The <tt>func</tt> argument must hold a function bitcast to an <tt>i8*</tt>.
5810 The block of memory pointed to by <tt>tramp</tt> is filled with target
5811 dependent code, turning it into a function. A pointer to this function is
5812 returned, but needs to be bitcast to an
5813 <a href="#int_trampoline">appropriate function pointer type</a>
5814 before being called. The new function's signature is the same as that of
5815 <tt>func</tt> with any arguments marked with the <tt>nest</tt> attribute
5816 removed. At most one such <tt>nest</tt> argument is allowed, and it must be
5817 of pointer type. Calling the new function is equivalent to calling
5818 <tt>func</tt> with the same argument list, but with <tt>nval</tt> used for the
5819 missing <tt>nest</tt> argument. If, after calling
5820 <tt>llvm.init.trampoline</tt>, the memory pointed to by <tt>tramp</tt> is
5821 modified, then the effect of any later call to the returned function pointer is
5826 <!-- ======================================================================= -->
5827 <div class="doc_subsection">
5828 <a name="int_atomics">Atomic Operations and Synchronization Intrinsics</a>
5831 <div class="doc_text">
5833 These intrinsic functions expand the "universal IR" of LLVM to represent
5834 hardware constructs for atomic operations and memory synchronization. This
5835 provides an interface to the hardware, not an interface to the programmer. It
5836 is aimed at a low enough level to allow any programming models or APIs
5837 (Application Programming Interfaces) which
5838 need atomic behaviors to map cleanly onto it. It is also modeled primarily on
5839 hardware behavior. Just as hardware provides a "universal IR" for source
5840 languages, it also provides a starting point for developing a "universal"
5841 atomic operation and synchronization IR.
5844 These do <em>not</em> form an API such as high-level threading libraries,
5845 software transaction memory systems, atomic primitives, and intrinsic
5846 functions as found in BSD, GNU libc, atomic_ops, APR, and other system and
5847 application libraries. The hardware interface provided by LLVM should allow
5848 a clean implementation of all of these APIs and parallel programming models.
5849 No one model or paradigm should be selected above others unless the hardware
5850 itself ubiquitously does so.
5855 <!-- _______________________________________________________________________ -->
5856 <div class="doc_subsubsection">
5857 <a name="int_memory_barrier">'<tt>llvm.memory.barrier</tt>' Intrinsic</a>
5859 <div class="doc_text">
5862 declare void @llvm.memory.barrier( i1 <ll>, i1 <ls>, i1 <sl>, i1 <ss>,
5868 The <tt>llvm.memory.barrier</tt> intrinsic guarantees ordering between
5869 specific pairs of memory access types.
5873 The <tt>llvm.memory.barrier</tt> intrinsic requires five boolean arguments.
5874 The first four arguments enables a specific barrier as listed below. The fith
5875 argument specifies that the barrier applies to io or device or uncached memory.
5879 <li><tt>ll</tt>: load-load barrier</li>
5880 <li><tt>ls</tt>: load-store barrier</li>
5881 <li><tt>sl</tt>: store-load barrier</li>
5882 <li><tt>ss</tt>: store-store barrier</li>
5883 <li><tt>device</tt>: barrier applies to device and uncached memory also.</li>
5887 This intrinsic causes the system to enforce some ordering constraints upon
5888 the loads and stores of the program. This barrier does not indicate
5889 <em>when</em> any events will occur, it only enforces an <em>order</em> in
5890 which they occur. For any of the specified pairs of load and store operations
5891 (f.ex. load-load, or store-load), all of the first operations preceding the
5892 barrier will complete before any of the second operations succeeding the
5893 barrier begin. Specifically the semantics for each pairing is as follows:
5896 <li><tt>ll</tt>: All loads before the barrier must complete before any load
5897 after the barrier begins.</li>
5899 <li><tt>ls</tt>: All loads before the barrier must complete before any
5900 store after the barrier begins.</li>
5901 <li><tt>ss</tt>: All stores before the barrier must complete before any
5902 store after the barrier begins.</li>
5903 <li><tt>sl</tt>: All stores before the barrier must complete before any
5904 load after the barrier begins.</li>
5907 These semantics are applied with a logical "and" behavior when more than one
5908 is enabled in a single memory barrier intrinsic.
5911 Backends may implement stronger barriers than those requested when they do not
5912 support as fine grained a barrier as requested. Some architectures do not
5913 need all types of barriers and on such architectures, these become noops.
5920 %result1 = load i32* %ptr <i>; yields {i32}:result1 = 4</i>
5921 call void @llvm.memory.barrier( i1 false, i1 true, i1 false, i1 false )
5922 <i>; guarantee the above finishes</i>
5923 store i32 8, %ptr <i>; before this begins</i>
5927 <!-- _______________________________________________________________________ -->
5928 <div class="doc_subsubsection">
5929 <a name="int_atomic_cmp_swap">'<tt>llvm.atomic.cmp.swap.*</tt>' Intrinsic</a>
5931 <div class="doc_text">
5934 This is an overloaded intrinsic. You can use <tt>llvm.atomic.cmp.swap</tt> on
5935 any integer bit width and for different address spaces. Not all targets
5936 support all bit widths however.</p>
5939 declare i8 @llvm.atomic.cmp.swap.i8.p0i8( i8* <ptr>, i8 <cmp>, i8 <val> )
5940 declare i16 @llvm.atomic.cmp.swap.i16.p0i16( i16* <ptr>, i16 <cmp>, i16 <val> )
5941 declare i32 @llvm.atomic.cmp.swap.i32.p0i32( i32* <ptr>, i32 <cmp>, i32 <val> )
5942 declare i64 @llvm.atomic.cmp.swap.i64.p0i64( i64* <ptr>, i64 <cmp>, i64 <val> )
5947 This loads a value in memory and compares it to a given value. If they are
5948 equal, it stores a new value into the memory.
5952 The <tt>llvm.atomic.cmp.swap</tt> intrinsic takes three arguments. The result as
5953 well as both <tt>cmp</tt> and <tt>val</tt> must be integer values with the
5954 same bit width. The <tt>ptr</tt> argument must be a pointer to a value of
5955 this integer type. While any bit width integer may be used, targets may only
5956 lower representations they support in hardware.
5961 This entire intrinsic must be executed atomically. It first loads the value
5962 in memory pointed to by <tt>ptr</tt> and compares it with the value
5963 <tt>cmp</tt>. If they are equal, <tt>val</tt> is stored into the memory. The
5964 loaded value is yielded in all cases. This provides the equivalent of an
5965 atomic compare-and-swap operation within the SSA framework.
5973 %val1 = add i32 4, 4
5974 %result1 = call i32 @llvm.atomic.cmp.swap.i32.p0i32( i32* %ptr, i32 4, %val1 )
5975 <i>; yields {i32}:result1 = 4</i>
5976 %stored1 = icmp eq i32 %result1, 4 <i>; yields {i1}:stored1 = true</i>
5977 %memval1 = load i32* %ptr <i>; yields {i32}:memval1 = 8</i>
5979 %val2 = add i32 1, 1
5980 %result2 = call i32 @llvm.atomic.cmp.swap.i32.p0i32( i32* %ptr, i32 5, %val2 )
5981 <i>; yields {i32}:result2 = 8</i>
5982 %stored2 = icmp eq i32 %result2, 5 <i>; yields {i1}:stored2 = false</i>
5984 %memval2 = load i32* %ptr <i>; yields {i32}:memval2 = 8</i>
5988 <!-- _______________________________________________________________________ -->
5989 <div class="doc_subsubsection">
5990 <a name="int_atomic_swap">'<tt>llvm.atomic.swap.*</tt>' Intrinsic</a>
5992 <div class="doc_text">
5996 This is an overloaded intrinsic. You can use <tt>llvm.atomic.swap</tt> on any
5997 integer bit width. Not all targets support all bit widths however.</p>
5999 declare i8 @llvm.atomic.swap.i8.p0i8( i8* <ptr>, i8 <val> )
6000 declare i16 @llvm.atomic.swap.i16.p0i16( i16* <ptr>, i16 <val> )
6001 declare i32 @llvm.atomic.swap.i32.p0i32( i32* <ptr>, i32 <val> )
6002 declare i64 @llvm.atomic.swap.i64.p0i64( i64* <ptr>, i64 <val> )
6007 This intrinsic loads the value stored in memory at <tt>ptr</tt> and yields
6008 the value from memory. It then stores the value in <tt>val</tt> in the memory
6014 The <tt>llvm.atomic.swap</tt> intrinsic takes two arguments. Both the
6015 <tt>val</tt> argument and the result must be integers of the same bit width.
6016 The first argument, <tt>ptr</tt>, must be a pointer to a value of this
6017 integer type. The targets may only lower integer representations they
6022 This intrinsic loads the value pointed to by <tt>ptr</tt>, yields it, and
6023 stores <tt>val</tt> back into <tt>ptr</tt> atomically. This provides the
6024 equivalent of an atomic swap operation within the SSA framework.
6032 %val1 = add i32 4, 4
6033 %result1 = call i32 @llvm.atomic.swap.i32.p0i32( i32* %ptr, i32 %val1 )
6034 <i>; yields {i32}:result1 = 4</i>
6035 %stored1 = icmp eq i32 %result1, 4 <i>; yields {i1}:stored1 = true</i>
6036 %memval1 = load i32* %ptr <i>; yields {i32}:memval1 = 8</i>
6038 %val2 = add i32 1, 1
6039 %result2 = call i32 @llvm.atomic.swap.i32.p0i32( i32* %ptr, i32 %val2 )
6040 <i>; yields {i32}:result2 = 8</i>
6042 %stored2 = icmp eq i32 %result2, 8 <i>; yields {i1}:stored2 = true</i>
6043 %memval2 = load i32* %ptr <i>; yields {i32}:memval2 = 2</i>
6047 <!-- _______________________________________________________________________ -->
6048 <div class="doc_subsubsection">
6049 <a name="int_atomic_load_add">'<tt>llvm.atomic.load.add.*</tt>' Intrinsic</a>
6052 <div class="doc_text">
6055 This is an overloaded intrinsic. You can use <tt>llvm.atomic.load.add</tt> on any
6056 integer bit width. Not all targets support all bit widths however.</p>
6058 declare i8 @llvm.atomic.load.add.i8..p0i8( i8* <ptr>, i8 <delta> )
6059 declare i16 @llvm.atomic.load.add.i16..p0i16( i16* <ptr>, i16 <delta> )
6060 declare i32 @llvm.atomic.load.add.i32..p0i32( i32* <ptr>, i32 <delta> )
6061 declare i64 @llvm.atomic.load.add.i64..p0i64( i64* <ptr>, i64 <delta> )
6066 This intrinsic adds <tt>delta</tt> to the value stored in memory at
6067 <tt>ptr</tt>. It yields the original value at <tt>ptr</tt>.
6072 The intrinsic takes two arguments, the first a pointer to an integer value
6073 and the second an integer value. The result is also an integer value. These
6074 integer types can have any bit width, but they must all have the same bit
6075 width. The targets may only lower integer representations they support.
6079 This intrinsic does a series of operations atomically. It first loads the
6080 value stored at <tt>ptr</tt>. It then adds <tt>delta</tt>, stores the result
6081 to <tt>ptr</tt>. It yields the original value stored at <tt>ptr</tt>.
6088 %result1 = call i32 @llvm.atomic.load.add.i32.p0i32( i32* %ptr, i32 4 )
6089 <i>; yields {i32}:result1 = 4</i>
6090 %result2 = call i32 @llvm.atomic.load.add.i32.p0i32( i32* %ptr, i32 2 )
6091 <i>; yields {i32}:result2 = 8</i>
6092 %result3 = call i32 @llvm.atomic.load.add.i32.p0i32( i32* %ptr, i32 5 )
6093 <i>; yields {i32}:result3 = 10</i>
6094 %memval1 = load i32* %ptr <i>; yields {i32}:memval1 = 15</i>
6098 <!-- _______________________________________________________________________ -->
6099 <div class="doc_subsubsection">
6100 <a name="int_atomic_load_sub">'<tt>llvm.atomic.load.sub.*</tt>' Intrinsic</a>
6103 <div class="doc_text">
6106 This is an overloaded intrinsic. You can use <tt>llvm.atomic.load.sub</tt> on
6107 any integer bit width and for different address spaces. Not all targets
6108 support all bit widths however.</p>
6110 declare i8 @llvm.atomic.load.sub.i8.p0i32( i8* <ptr>, i8 <delta> )
6111 declare i16 @llvm.atomic.load.sub.i16.p0i32( i16* <ptr>, i16 <delta> )
6112 declare i32 @llvm.atomic.load.sub.i32.p0i32( i32* <ptr>, i32 <delta> )
6113 declare i64 @llvm.atomic.load.sub.i64.p0i32( i64* <ptr>, i64 <delta> )
6118 This intrinsic subtracts <tt>delta</tt> to the value stored in memory at
6119 <tt>ptr</tt>. It yields the original value at <tt>ptr</tt>.
6124 The intrinsic takes two arguments, the first a pointer to an integer value
6125 and the second an integer value. The result is also an integer value. These
6126 integer types can have any bit width, but they must all have the same bit
6127 width. The targets may only lower integer representations they support.
6131 This intrinsic does a series of operations atomically. It first loads the
6132 value stored at <tt>ptr</tt>. It then subtracts <tt>delta</tt>, stores the
6133 result to <tt>ptr</tt>. It yields the original value stored at <tt>ptr</tt>.
6140 %result1 = call i32 @llvm.atomic.load.sub.i32.p0i32( i32* %ptr, i32 4 )
6141 <i>; yields {i32}:result1 = 8</i>
6142 %result2 = call i32 @llvm.atomic.load.sub.i32.p0i32( i32* %ptr, i32 2 )
6143 <i>; yields {i32}:result2 = 4</i>
6144 %result3 = call i32 @llvm.atomic.load.sub.i32.p0i32( i32* %ptr, i32 5 )
6145 <i>; yields {i32}:result3 = 2</i>
6146 %memval1 = load i32* %ptr <i>; yields {i32}:memval1 = -3</i>
6150 <!-- _______________________________________________________________________ -->
6151 <div class="doc_subsubsection">
6152 <a name="int_atomic_load_and">'<tt>llvm.atomic.load.and.*</tt>' Intrinsic</a><br>
6153 <a name="int_atomic_load_nand">'<tt>llvm.atomic.load.nand.*</tt>' Intrinsic</a><br>
6154 <a name="int_atomic_load_or">'<tt>llvm.atomic.load.or.*</tt>' Intrinsic</a><br>
6155 <a name="int_atomic_load_xor">'<tt>llvm.atomic.load.xor.*</tt>' Intrinsic</a><br>
6158 <div class="doc_text">
6161 These are overloaded intrinsics. You can use <tt>llvm.atomic.load_and</tt>,
6162 <tt>llvm.atomic.load_nand</tt>, <tt>llvm.atomic.load_or</tt>, and
6163 <tt>llvm.atomic.load_xor</tt> on any integer bit width and for different
6164 address spaces. Not all targets support all bit widths however.</p>
6166 declare i8 @llvm.atomic.load.and.i8.p0i8( i8* <ptr>, i8 <delta> )
6167 declare i16 @llvm.atomic.load.and.i16.p0i16( i16* <ptr>, i16 <delta> )
6168 declare i32 @llvm.atomic.load.and.i32.p0i32( i32* <ptr>, i32 <delta> )
6169 declare i64 @llvm.atomic.load.and.i64.p0i64( i64* <ptr>, i64 <delta> )
6174 declare i8 @llvm.atomic.load.or.i8.p0i8( i8* <ptr>, i8 <delta> )
6175 declare i16 @llvm.atomic.load.or.i16.p0i16( i16* <ptr>, i16 <delta> )
6176 declare i32 @llvm.atomic.load.or.i32.p0i32( i32* <ptr>, i32 <delta> )
6177 declare i64 @llvm.atomic.load.or.i64.p0i64( i64* <ptr>, i64 <delta> )
6182 declare i8 @llvm.atomic.load.nand.i8.p0i32( i8* <ptr>, i8 <delta> )
6183 declare i16 @llvm.atomic.load.nand.i16.p0i32( i16* <ptr>, i16 <delta> )
6184 declare i32 @llvm.atomic.load.nand.i32.p0i32( i32* <ptr>, i32 <delta> )
6185 declare i64 @llvm.atomic.load.nand.i64.p0i32( i64* <ptr>, i64 <delta> )
6190 declare i8 @llvm.atomic.load.xor.i8.p0i32( i8* <ptr>, i8 <delta> )
6191 declare i16 @llvm.atomic.load.xor.i16.p0i32( i16* <ptr>, i16 <delta> )
6192 declare i32 @llvm.atomic.load.xor.i32.p0i32( i32* <ptr>, i32 <delta> )
6193 declare i64 @llvm.atomic.load.xor.i64.p0i32( i64* <ptr>, i64 <delta> )
6198 These intrinsics bitwise the operation (and, nand, or, xor) <tt>delta</tt> to
6199 the value stored in memory at <tt>ptr</tt>. It yields the original value
6205 These intrinsics take two arguments, the first a pointer to an integer value
6206 and the second an integer value. The result is also an integer value. These
6207 integer types can have any bit width, but they must all have the same bit
6208 width. The targets may only lower integer representations they support.
6212 These intrinsics does a series of operations atomically. They first load the
6213 value stored at <tt>ptr</tt>. They then do the bitwise operation
6214 <tt>delta</tt>, store the result to <tt>ptr</tt>. They yield the original
6215 value stored at <tt>ptr</tt>.
6221 store i32 0x0F0F, %ptr
6222 %result0 = call i32 @llvm.atomic.load.nand.i32.p0i32( i32* %ptr, i32 0xFF )
6223 <i>; yields {i32}:result0 = 0x0F0F</i>
6224 %result1 = call i32 @llvm.atomic.load.and.i32.p0i32( i32* %ptr, i32 0xFF )
6225 <i>; yields {i32}:result1 = 0xFFFFFFF0</i>
6226 %result2 = call i32 @llvm.atomic.load.or.i32.p0i32( i32* %ptr, i32 0F )
6227 <i>; yields {i32}:result2 = 0xF0</i>
6228 %result3 = call i32 @llvm.atomic.load.xor.i32.p0i32( i32* %ptr, i32 0F )
6229 <i>; yields {i32}:result3 = FF</i>
6230 %memval1 = load i32* %ptr <i>; yields {i32}:memval1 = F0</i>
6235 <!-- _______________________________________________________________________ -->
6236 <div class="doc_subsubsection">
6237 <a name="int_atomic_load_max">'<tt>llvm.atomic.load.max.*</tt>' Intrinsic</a><br>
6238 <a name="int_atomic_load_min">'<tt>llvm.atomic.load.min.*</tt>' Intrinsic</a><br>
6239 <a name="int_atomic_load_umax">'<tt>llvm.atomic.load.umax.*</tt>' Intrinsic</a><br>
6240 <a name="int_atomic_load_umin">'<tt>llvm.atomic.load.umin.*</tt>' Intrinsic</a><br>
6243 <div class="doc_text">
6246 These are overloaded intrinsics. You can use <tt>llvm.atomic.load_max</tt>,
6247 <tt>llvm.atomic.load_min</tt>, <tt>llvm.atomic.load_umax</tt>, and
6248 <tt>llvm.atomic.load_umin</tt> on any integer bit width and for different
6249 address spaces. Not all targets
6250 support all bit widths however.</p>
6252 declare i8 @llvm.atomic.load.max.i8.p0i8( i8* <ptr>, i8 <delta> )
6253 declare i16 @llvm.atomic.load.max.i16.p0i16( i16* <ptr>, i16 <delta> )
6254 declare i32 @llvm.atomic.load.max.i32.p0i32( i32* <ptr>, i32 <delta> )
6255 declare i64 @llvm.atomic.load.max.i64.p0i64( i64* <ptr>, i64 <delta> )
6260 declare i8 @llvm.atomic.load.min.i8.p0i8( i8* <ptr>, i8 <delta> )
6261 declare i16 @llvm.atomic.load.min.i16.p0i16( i16* <ptr>, i16 <delta> )
6262 declare i32 @llvm.atomic.load.min.i32..p0i32( i32* <ptr>, i32 <delta> )
6263 declare i64 @llvm.atomic.load.min.i64..p0i64( i64* <ptr>, i64 <delta> )
6268 declare i8 @llvm.atomic.load.umax.i8.p0i8( i8* <ptr>, i8 <delta> )
6269 declare i16 @llvm.atomic.load.umax.i16.p0i16( i16* <ptr>, i16 <delta> )
6270 declare i32 @llvm.atomic.load.umax.i32.p0i32( i32* <ptr>, i32 <delta> )
6271 declare i64 @llvm.atomic.load.umax.i64.p0i64( i64* <ptr>, i64 <delta> )
6276 declare i8 @llvm.atomic.load.umin.i8..p0i8( i8* <ptr>, i8 <delta> )
6277 declare i16 @llvm.atomic.load.umin.i16.p0i16( i16* <ptr>, i16 <delta> )
6278 declare i32 @llvm.atomic.load.umin.i32..p0i32( i32* <ptr>, i32 <delta> )
6279 declare i64 @llvm.atomic.load.umin.i64..p0i64( i64* <ptr>, i64 <delta> )
6284 These intrinsics takes the signed or unsigned minimum or maximum of
6285 <tt>delta</tt> and the value stored in memory at <tt>ptr</tt>. It yields the
6286 original value at <tt>ptr</tt>.
6291 These intrinsics take two arguments, the first a pointer to an integer value
6292 and the second an integer value. The result is also an integer value. These
6293 integer types can have any bit width, but they must all have the same bit
6294 width. The targets may only lower integer representations they support.
6298 These intrinsics does a series of operations atomically. They first load the
6299 value stored at <tt>ptr</tt>. They then do the signed or unsigned min or max
6300 <tt>delta</tt> and the value, store the result to <tt>ptr</tt>. They yield
6301 the original value stored at <tt>ptr</tt>.
6308 %result0 = call i32 @llvm.atomic.load.min.i32.p0i32( i32* %ptr, i32 -2 )
6309 <i>; yields {i32}:result0 = 7</i>
6310 %result1 = call i32 @llvm.atomic.load.max.i32.p0i32( i32* %ptr, i32 8 )
6311 <i>; yields {i32}:result1 = -2</i>
6312 %result2 = call i32 @llvm.atomic.load.umin.i32.p0i32( i32* %ptr, i32 10 )
6313 <i>; yields {i32}:result2 = 8</i>
6314 %result3 = call i32 @llvm.atomic.load.umax.i32.p0i32( i32* %ptr, i32 30 )
6315 <i>; yields {i32}:result3 = 8</i>
6316 %memval1 = load i32* %ptr <i>; yields {i32}:memval1 = 30</i>
6320 <!-- ======================================================================= -->
6321 <div class="doc_subsection">
6322 <a name="int_general">General Intrinsics</a>
6325 <div class="doc_text">
6326 <p> This class of intrinsics is designed to be generic and has
6327 no specific purpose. </p>
6330 <!-- _______________________________________________________________________ -->
6331 <div class="doc_subsubsection">
6332 <a name="int_var_annotation">'<tt>llvm.var.annotation</tt>' Intrinsic</a>
6335 <div class="doc_text">
6339 declare void @llvm.var.annotation(i8* <val>, i8* <str>, i8* <str>, i32 <int> )
6345 The '<tt>llvm.var.annotation</tt>' intrinsic
6351 The first argument is a pointer to a value, the second is a pointer to a
6352 global string, the third is a pointer to a global string which is the source
6353 file name, and the last argument is the line number.
6359 This intrinsic allows annotation of local variables with arbitrary strings.
6360 This can be useful for special purpose optimizations that want to look for these
6361 annotations. These have no other defined use, they are ignored by code
6362 generation and optimization.
6366 <!-- _______________________________________________________________________ -->
6367 <div class="doc_subsubsection">
6368 <a name="int_annotation">'<tt>llvm.annotation.*</tt>' Intrinsic</a>
6371 <div class="doc_text">
6374 <p>This is an overloaded intrinsic. You can use '<tt>llvm.annotation</tt>' on
6375 any integer bit width.
6378 declare i8 @llvm.annotation.i8(i8 <val>, i8* <str>, i8* <str>, i32 <int> )
6379 declare i16 @llvm.annotation.i16(i16 <val>, i8* <str>, i8* <str>, i32 <int> )
6380 declare i32 @llvm.annotation.i32(i32 <val>, i8* <str>, i8* <str>, i32 <int> )
6381 declare i64 @llvm.annotation.i64(i64 <val>, i8* <str>, i8* <str>, i32 <int> )
6382 declare i256 @llvm.annotation.i256(i256 <val>, i8* <str>, i8* <str>, i32 <int> )
6388 The '<tt>llvm.annotation</tt>' intrinsic.
6394 The first argument is an integer value (result of some expression),
6395 the second is a pointer to a global string, the third is a pointer to a global
6396 string which is the source file name, and the last argument is the line number.
6397 It returns the value of the first argument.
6403 This intrinsic allows annotations to be put on arbitrary expressions
6404 with arbitrary strings. This can be useful for special purpose optimizations
6405 that want to look for these annotations. These have no other defined use, they
6406 are ignored by code generation and optimization.
6410 <!-- _______________________________________________________________________ -->
6411 <div class="doc_subsubsection">
6412 <a name="int_trap">'<tt>llvm.trap</tt>' Intrinsic</a>
6415 <div class="doc_text">
6419 declare void @llvm.trap()
6425 The '<tt>llvm.trap</tt>' intrinsic
6437 This intrinsics is lowered to the target dependent trap instruction. If the
6438 target does not have a trap instruction, this intrinsic will be lowered to the
6439 call of the abort() function.
6443 <!-- _______________________________________________________________________ -->
6444 <div class="doc_subsubsection">
6445 <a name="int_stackprotector">'<tt>llvm.stackprotector</tt>' Intrinsic</a>
6447 <div class="doc_text">
6450 declare void @llvm.stackprotector( i8* <guard>, i8** <slot> )
6455 The <tt>llvm.stackprotector</tt> intrinsic takes the <tt>guard</tt> and stores
6456 it onto the stack at <tt>slot</tt>. The stack slot is adjusted to ensure that
6457 it is placed on the stack before local variables.
6461 The <tt>llvm.stackprotector</tt> intrinsic requires two pointer arguments. The
6462 first argument is the value loaded from the stack guard
6463 <tt>@__stack_chk_guard</tt>. The second variable is an <tt>alloca</tt> that
6464 has enough space to hold the value of the guard.
6468 This intrinsic causes the prologue/epilogue inserter to force the position of
6469 the <tt>AllocaInst</tt> stack slot to be before local variables on the
6470 stack. This is to ensure that if a local variable on the stack is overwritten,
6471 it will destroy the value of the guard. When the function exits, the guard on
6472 the stack is checked against the original guard. If they're different, then
6473 the program aborts by calling the <tt>__stack_chk_fail()</tt> function.
6477 <!-- *********************************************************************** -->
6480 <a href="http://jigsaw.w3.org/css-validator/check/referer"><img
6481 src="http://jigsaw.w3.org/css-validator/images/vcss" alt="Valid CSS!"></a>
6482 <a href="http://validator.w3.org/check/referer"><img
6483 src="http://www.w3.org/Icons/valid-html401" alt="Valid HTML 4.01!"></a>
6485 <a href="mailto:sabre@nondot.org">Chris Lattner</a><br>
6486 <a href="http://llvm.org">The LLVM Compiler Infrastructure</a><br>
6487 Last modified: $Date$