1 <!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01//EN"
2 "http://www.w3.org/TR/html4/strict.dtd">
5 <title>LLVM Assembly Language Reference Manual</title>
6 <meta http-equiv="Content-Type" content="text/html; charset=utf-8">
7 <meta name="author" content="Chris Lattner">
8 <meta name="description"
9 content="LLVM Assembly Language Reference Manual.">
10 <link rel="stylesheet" href="llvm.css" type="text/css">
15 <div class="doc_title"> LLVM Language Reference Manual </div>
17 <li><a href="#abstract">Abstract</a></li>
18 <li><a href="#introduction">Introduction</a></li>
19 <li><a href="#identifiers">Identifiers</a></li>
20 <li><a href="#highlevel">High Level Structure</a>
22 <li><a href="#modulestructure">Module Structure</a></li>
23 <li><a href="#linkage">Linkage Types</a></li>
24 <li><a href="#callingconv">Calling Conventions</a></li>
25 <li><a href="#namedtypes">Named Types</a></li>
26 <li><a href="#globalvars">Global Variables</a></li>
27 <li><a href="#functionstructure">Functions</a></li>
28 <li><a href="#aliasstructure">Aliases</a></li>
29 <li><a href="#paramattrs">Parameter Attributes</a></li>
30 <li><a href="#fnattrs">Function Attributes</a></li>
31 <li><a href="#gc">Garbage Collector Names</a></li>
32 <li><a href="#moduleasm">Module-Level Inline Assembly</a></li>
33 <li><a href="#datalayout">Data Layout</a></li>
36 <li><a href="#typesystem">Type System</a>
38 <li><a href="#t_classifications">Type Classifications</a></li>
39 <li><a href="#t_primitive">Primitive Types</a>
41 <li><a href="#t_floating">Floating Point Types</a></li>
42 <li><a href="#t_void">Void Type</a></li>
43 <li><a href="#t_label">Label Type</a></li>
44 <li><a href="#t_metadata">Metadata Type</a></li>
47 <li><a href="#t_derived">Derived Types</a>
49 <li><a href="#t_integer">Integer Type</a></li>
50 <li><a href="#t_array">Array Type</a></li>
51 <li><a href="#t_function">Function Type</a></li>
52 <li><a href="#t_pointer">Pointer Type</a></li>
53 <li><a href="#t_struct">Structure Type</a></li>
54 <li><a href="#t_pstruct">Packed Structure Type</a></li>
55 <li><a href="#t_vector">Vector Type</a></li>
56 <li><a href="#t_opaque">Opaque Type</a></li>
59 <li><a href="#t_uprefs">Type Up-references</a></li>
62 <li><a href="#constants">Constants</a>
64 <li><a href="#simpleconstants">Simple Constants</a></li>
65 <li><a href="#complexconstants">Complex Constants</a></li>
66 <li><a href="#globalconstants">Global Variable and Function Addresses</a></li>
67 <li><a href="#undefvalues">Undefined Values</a></li>
68 <li><a href="#constantexprs">Constant Expressions</a></li>
69 <li><a href="#metadata">Embedded Metadata</a></li>
72 <li><a href="#othervalues">Other Values</a>
74 <li><a href="#inlineasm">Inline Assembler Expressions</a></li>
77 <li><a href="#instref">Instruction Reference</a>
79 <li><a href="#terminators">Terminator Instructions</a>
81 <li><a href="#i_ret">'<tt>ret</tt>' Instruction</a></li>
82 <li><a href="#i_br">'<tt>br</tt>' Instruction</a></li>
83 <li><a href="#i_switch">'<tt>switch</tt>' Instruction</a></li>
84 <li><a href="#i_invoke">'<tt>invoke</tt>' Instruction</a></li>
85 <li><a href="#i_unwind">'<tt>unwind</tt>' Instruction</a></li>
86 <li><a href="#i_unreachable">'<tt>unreachable</tt>' Instruction</a></li>
89 <li><a href="#binaryops">Binary Operations</a>
91 <li><a href="#i_add">'<tt>add</tt>' Instruction</a></li>
92 <li><a href="#i_fadd">'<tt>fadd</tt>' Instruction</a></li>
93 <li><a href="#i_sub">'<tt>sub</tt>' Instruction</a></li>
94 <li><a href="#i_fsub">'<tt>fsub</tt>' Instruction</a></li>
95 <li><a href="#i_mul">'<tt>mul</tt>' Instruction</a></li>
96 <li><a href="#i_fmul">'<tt>fmul</tt>' Instruction</a></li>
97 <li><a href="#i_udiv">'<tt>udiv</tt>' Instruction</a></li>
98 <li><a href="#i_sdiv">'<tt>sdiv</tt>' Instruction</a></li>
99 <li><a href="#i_fdiv">'<tt>fdiv</tt>' Instruction</a></li>
100 <li><a href="#i_urem">'<tt>urem</tt>' Instruction</a></li>
101 <li><a href="#i_srem">'<tt>srem</tt>' Instruction</a></li>
102 <li><a href="#i_frem">'<tt>frem</tt>' Instruction</a></li>
105 <li><a href="#bitwiseops">Bitwise Binary Operations</a>
107 <li><a href="#i_shl">'<tt>shl</tt>' Instruction</a></li>
108 <li><a href="#i_lshr">'<tt>lshr</tt>' Instruction</a></li>
109 <li><a href="#i_ashr">'<tt>ashr</tt>' Instruction</a></li>
110 <li><a href="#i_and">'<tt>and</tt>' Instruction</a></li>
111 <li><a href="#i_or">'<tt>or</tt>' Instruction</a></li>
112 <li><a href="#i_xor">'<tt>xor</tt>' Instruction</a></li>
115 <li><a href="#vectorops">Vector Operations</a>
117 <li><a href="#i_extractelement">'<tt>extractelement</tt>' Instruction</a></li>
118 <li><a href="#i_insertelement">'<tt>insertelement</tt>' Instruction</a></li>
119 <li><a href="#i_shufflevector">'<tt>shufflevector</tt>' Instruction</a></li>
122 <li><a href="#aggregateops">Aggregate Operations</a>
124 <li><a href="#i_extractvalue">'<tt>extractvalue</tt>' Instruction</a></li>
125 <li><a href="#i_insertvalue">'<tt>insertvalue</tt>' Instruction</a></li>
128 <li><a href="#memoryops">Memory Access and Addressing Operations</a>
130 <li><a href="#i_malloc">'<tt>malloc</tt>' Instruction</a></li>
131 <li><a href="#i_free">'<tt>free</tt>' Instruction</a></li>
132 <li><a href="#i_alloca">'<tt>alloca</tt>' Instruction</a></li>
133 <li><a href="#i_load">'<tt>load</tt>' Instruction</a></li>
134 <li><a href="#i_store">'<tt>store</tt>' Instruction</a></li>
135 <li><a href="#i_getelementptr">'<tt>getelementptr</tt>' Instruction</a></li>
138 <li><a href="#convertops">Conversion Operations</a>
140 <li><a href="#i_trunc">'<tt>trunc .. to</tt>' Instruction</a></li>
141 <li><a href="#i_zext">'<tt>zext .. to</tt>' Instruction</a></li>
142 <li><a href="#i_sext">'<tt>sext .. to</tt>' Instruction</a></li>
143 <li><a href="#i_fptrunc">'<tt>fptrunc .. to</tt>' Instruction</a></li>
144 <li><a href="#i_fpext">'<tt>fpext .. to</tt>' Instruction</a></li>
145 <li><a href="#i_fptoui">'<tt>fptoui .. to</tt>' Instruction</a></li>
146 <li><a href="#i_fptosi">'<tt>fptosi .. to</tt>' Instruction</a></li>
147 <li><a href="#i_uitofp">'<tt>uitofp .. to</tt>' Instruction</a></li>
148 <li><a href="#i_sitofp">'<tt>sitofp .. to</tt>' Instruction</a></li>
149 <li><a href="#i_ptrtoint">'<tt>ptrtoint .. to</tt>' Instruction</a></li>
150 <li><a href="#i_inttoptr">'<tt>inttoptr .. to</tt>' Instruction</a></li>
151 <li><a href="#i_bitcast">'<tt>bitcast .. to</tt>' Instruction</a></li>
154 <li><a href="#otherops">Other Operations</a>
156 <li><a href="#i_icmp">'<tt>icmp</tt>' Instruction</a></li>
157 <li><a href="#i_fcmp">'<tt>fcmp</tt>' Instruction</a></li>
158 <li><a href="#i_phi">'<tt>phi</tt>' Instruction</a></li>
159 <li><a href="#i_select">'<tt>select</tt>' Instruction</a></li>
160 <li><a href="#i_call">'<tt>call</tt>' Instruction</a></li>
161 <li><a href="#i_va_arg">'<tt>va_arg</tt>' Instruction</a></li>
166 <li><a href="#intrinsics">Intrinsic Functions</a>
168 <li><a href="#int_varargs">Variable Argument Handling Intrinsics</a>
170 <li><a href="#int_va_start">'<tt>llvm.va_start</tt>' Intrinsic</a></li>
171 <li><a href="#int_va_end">'<tt>llvm.va_end</tt>' Intrinsic</a></li>
172 <li><a href="#int_va_copy">'<tt>llvm.va_copy</tt>' Intrinsic</a></li>
175 <li><a href="#int_gc">Accurate Garbage Collection Intrinsics</a>
177 <li><a href="#int_gcroot">'<tt>llvm.gcroot</tt>' Intrinsic</a></li>
178 <li><a href="#int_gcread">'<tt>llvm.gcread</tt>' Intrinsic</a></li>
179 <li><a href="#int_gcwrite">'<tt>llvm.gcwrite</tt>' Intrinsic</a></li>
182 <li><a href="#int_codegen">Code Generator Intrinsics</a>
184 <li><a href="#int_returnaddress">'<tt>llvm.returnaddress</tt>' Intrinsic</a></li>
185 <li><a href="#int_frameaddress">'<tt>llvm.frameaddress</tt>' Intrinsic</a></li>
186 <li><a href="#int_stacksave">'<tt>llvm.stacksave</tt>' Intrinsic</a></li>
187 <li><a href="#int_stackrestore">'<tt>llvm.stackrestore</tt>' Intrinsic</a></li>
188 <li><a href="#int_prefetch">'<tt>llvm.prefetch</tt>' Intrinsic</a></li>
189 <li><a href="#int_pcmarker">'<tt>llvm.pcmarker</tt>' Intrinsic</a></li>
190 <li><a href="#int_readcyclecounter"><tt>llvm.readcyclecounter</tt>' Intrinsic</a></li>
193 <li><a href="#int_libc">Standard C Library Intrinsics</a>
195 <li><a href="#int_memcpy">'<tt>llvm.memcpy.*</tt>' Intrinsic</a></li>
196 <li><a href="#int_memmove">'<tt>llvm.memmove.*</tt>' Intrinsic</a></li>
197 <li><a href="#int_memset">'<tt>llvm.memset.*</tt>' Intrinsic</a></li>
198 <li><a href="#int_sqrt">'<tt>llvm.sqrt.*</tt>' Intrinsic</a></li>
199 <li><a href="#int_powi">'<tt>llvm.powi.*</tt>' Intrinsic</a></li>
200 <li><a href="#int_sin">'<tt>llvm.sin.*</tt>' Intrinsic</a></li>
201 <li><a href="#int_cos">'<tt>llvm.cos.*</tt>' Intrinsic</a></li>
202 <li><a href="#int_pow">'<tt>llvm.pow.*</tt>' Intrinsic</a></li>
205 <li><a href="#int_manip">Bit Manipulation Intrinsics</a>
207 <li><a href="#int_bswap">'<tt>llvm.bswap.*</tt>' Intrinsics</a></li>
208 <li><a href="#int_ctpop">'<tt>llvm.ctpop.*</tt>' Intrinsic </a></li>
209 <li><a href="#int_ctlz">'<tt>llvm.ctlz.*</tt>' Intrinsic </a></li>
210 <li><a href="#int_cttz">'<tt>llvm.cttz.*</tt>' Intrinsic </a></li>
213 <li><a href="#int_overflow">Arithmetic with Overflow Intrinsics</a>
215 <li><a href="#int_sadd_overflow">'<tt>llvm.sadd.with.overflow.*</tt> Intrinsics</a></li>
216 <li><a href="#int_uadd_overflow">'<tt>llvm.uadd.with.overflow.*</tt> Intrinsics</a></li>
217 <li><a href="#int_ssub_overflow">'<tt>llvm.ssub.with.overflow.*</tt> Intrinsics</a></li>
218 <li><a href="#int_usub_overflow">'<tt>llvm.usub.with.overflow.*</tt> Intrinsics</a></li>
219 <li><a href="#int_smul_overflow">'<tt>llvm.smul.with.overflow.*</tt> Intrinsics</a></li>
220 <li><a href="#int_umul_overflow">'<tt>llvm.umul.with.overflow.*</tt> Intrinsics</a></li>
223 <li><a href="#int_debugger">Debugger intrinsics</a></li>
224 <li><a href="#int_eh">Exception Handling intrinsics</a></li>
225 <li><a href="#int_trampoline">Trampoline Intrinsic</a>
227 <li><a href="#int_it">'<tt>llvm.init.trampoline</tt>' Intrinsic</a></li>
230 <li><a href="#int_atomics">Atomic intrinsics</a>
232 <li><a href="#int_memory_barrier"><tt>llvm.memory_barrier</tt></a></li>
233 <li><a href="#int_atomic_cmp_swap"><tt>llvm.atomic.cmp.swap</tt></a></li>
234 <li><a href="#int_atomic_swap"><tt>llvm.atomic.swap</tt></a></li>
235 <li><a href="#int_atomic_load_add"><tt>llvm.atomic.load.add</tt></a></li>
236 <li><a href="#int_atomic_load_sub"><tt>llvm.atomic.load.sub</tt></a></li>
237 <li><a href="#int_atomic_load_and"><tt>llvm.atomic.load.and</tt></a></li>
238 <li><a href="#int_atomic_load_nand"><tt>llvm.atomic.load.nand</tt></a></li>
239 <li><a href="#int_atomic_load_or"><tt>llvm.atomic.load.or</tt></a></li>
240 <li><a href="#int_atomic_load_xor"><tt>llvm.atomic.load.xor</tt></a></li>
241 <li><a href="#int_atomic_load_max"><tt>llvm.atomic.load.max</tt></a></li>
242 <li><a href="#int_atomic_load_min"><tt>llvm.atomic.load.min</tt></a></li>
243 <li><a href="#int_atomic_load_umax"><tt>llvm.atomic.load.umax</tt></a></li>
244 <li><a href="#int_atomic_load_umin"><tt>llvm.atomic.load.umin</tt></a></li>
247 <li><a href="#int_general">General intrinsics</a>
249 <li><a href="#int_var_annotation">
250 '<tt>llvm.var.annotation</tt>' Intrinsic</a></li>
251 <li><a href="#int_annotation">
252 '<tt>llvm.annotation.*</tt>' Intrinsic</a></li>
253 <li><a href="#int_trap">
254 '<tt>llvm.trap</tt>' Intrinsic</a></li>
255 <li><a href="#int_stackprotector">
256 '<tt>llvm.stackprotector</tt>' Intrinsic</a></li>
263 <div class="doc_author">
264 <p>Written by <a href="mailto:sabre@nondot.org">Chris Lattner</a>
265 and <a href="mailto:vadve@cs.uiuc.edu">Vikram Adve</a></p>
268 <!-- *********************************************************************** -->
269 <div class="doc_section"> <a name="abstract">Abstract </a></div>
270 <!-- *********************************************************************** -->
272 <div class="doc_text">
273 <p>This document is a reference manual for the LLVM assembly language.
274 LLVM is a Static Single Assignment (SSA) based representation that provides
275 type safety, low-level operations, flexibility, and the capability of
276 representing 'all' high-level languages cleanly. It is the common code
277 representation used throughout all phases of the LLVM compilation
281 <!-- *********************************************************************** -->
282 <div class="doc_section"> <a name="introduction">Introduction</a> </div>
283 <!-- *********************************************************************** -->
285 <div class="doc_text">
287 <p>The LLVM code representation is designed to be used in three
288 different forms: as an in-memory compiler IR, as an on-disk bitcode
289 representation (suitable for fast loading by a Just-In-Time compiler),
290 and as a human readable assembly language representation. This allows
291 LLVM to provide a powerful intermediate representation for efficient
292 compiler transformations and analysis, while providing a natural means
293 to debug and visualize the transformations. The three different forms
294 of LLVM are all equivalent. This document describes the human readable
295 representation and notation.</p>
297 <p>The LLVM representation aims to be light-weight and low-level
298 while being expressive, typed, and extensible at the same time. It
299 aims to be a "universal IR" of sorts, by being at a low enough level
300 that high-level ideas may be cleanly mapped to it (similar to how
301 microprocessors are "universal IR's", allowing many source languages to
302 be mapped to them). By providing type information, LLVM can be used as
303 the target of optimizations: for example, through pointer analysis, it
304 can be proven that a C automatic variable is never accessed outside of
305 the current function... allowing it to be promoted to a simple SSA
306 value instead of a memory location.</p>
310 <!-- _______________________________________________________________________ -->
311 <div class="doc_subsubsection"> <a name="wellformed">Well-Formedness</a> </div>
313 <div class="doc_text">
315 <p>It is important to note that this document describes 'well formed'
316 LLVM assembly language. There is a difference between what the parser
317 accepts and what is considered 'well formed'. For example, the
318 following instruction is syntactically okay, but not well formed:</p>
320 <div class="doc_code">
322 %x = <a href="#i_add">add</a> i32 1, %x
326 <p>...because the definition of <tt>%x</tt> does not dominate all of
327 its uses. The LLVM infrastructure provides a verification pass that may
328 be used to verify that an LLVM module is well formed. This pass is
329 automatically run by the parser after parsing input assembly and by
330 the optimizer before it outputs bitcode. The violations pointed out
331 by the verifier pass indicate bugs in transformation passes or input to
335 <!-- Describe the typesetting conventions here. -->
337 <!-- *********************************************************************** -->
338 <div class="doc_section"> <a name="identifiers">Identifiers</a> </div>
339 <!-- *********************************************************************** -->
341 <div class="doc_text">
343 <p>LLVM identifiers come in two basic types: global and local. Global
344 identifiers (functions, global variables) begin with the @ character. Local
345 identifiers (register names, types) begin with the % character. Additionally,
346 there are three different formats for identifiers, for different purposes:</p>
349 <li>Named values are represented as a string of characters with their prefix.
350 For example, %foo, @DivisionByZero, %a.really.long.identifier. The actual
351 regular expression used is '<tt>[%@][a-zA-Z$._][a-zA-Z$._0-9]*</tt>'.
352 Identifiers which require other characters in their names can be surrounded
353 with quotes. Special characters may be escaped using "\xx" where xx is the
354 ASCII code for the character in hexadecimal. In this way, any character can
355 be used in a name value, even quotes themselves.
357 <li>Unnamed values are represented as an unsigned numeric value with their
358 prefix. For example, %12, @2, %44.</li>
360 <li>Constants, which are described in a <a href="#constants">section about
361 constants</a>, below.</li>
364 <p>LLVM requires that values start with a prefix for two reasons: Compilers
365 don't need to worry about name clashes with reserved words, and the set of
366 reserved words may be expanded in the future without penalty. Additionally,
367 unnamed identifiers allow a compiler to quickly come up with a temporary
368 variable without having to avoid symbol table conflicts.</p>
370 <p>Reserved words in LLVM are very similar to reserved words in other
371 languages. There are keywords for different opcodes
372 ('<tt><a href="#i_add">add</a></tt>',
373 '<tt><a href="#i_bitcast">bitcast</a></tt>',
374 '<tt><a href="#i_ret">ret</a></tt>', etc...), for primitive type names ('<tt><a
375 href="#t_void">void</a></tt>', '<tt><a href="#t_primitive">i32</a></tt>', etc...),
376 and others. These reserved words cannot conflict with variable names, because
377 none of them start with a prefix character ('%' or '@').</p>
379 <p>Here is an example of LLVM code to multiply the integer variable
380 '<tt>%X</tt>' by 8:</p>
384 <div class="doc_code">
386 %result = <a href="#i_mul">mul</a> i32 %X, 8
390 <p>After strength reduction:</p>
392 <div class="doc_code">
394 %result = <a href="#i_shl">shl</a> i32 %X, i8 3
398 <p>And the hard way:</p>
400 <div class="doc_code">
402 <a href="#i_add">add</a> i32 %X, %X <i>; yields {i32}:%0</i>
403 <a href="#i_add">add</a> i32 %0, %0 <i>; yields {i32}:%1</i>
404 %result = <a href="#i_add">add</a> i32 %1, %1
408 <p>This last way of multiplying <tt>%X</tt> by 8 illustrates several
409 important lexical features of LLVM:</p>
413 <li>Comments are delimited with a '<tt>;</tt>' and go until the end of
416 <li>Unnamed temporaries are created when the result of a computation is not
417 assigned to a named value.</li>
419 <li>Unnamed temporaries are numbered sequentially</li>
423 <p>...and it also shows a convention that we follow in this document. When
424 demonstrating instructions, we will follow an instruction with a comment that
425 defines the type and name of value produced. Comments are shown in italic
430 <!-- *********************************************************************** -->
431 <div class="doc_section"> <a name="highlevel">High Level Structure</a> </div>
432 <!-- *********************************************************************** -->
434 <!-- ======================================================================= -->
435 <div class="doc_subsection"> <a name="modulestructure">Module Structure</a>
438 <div class="doc_text">
440 <p>LLVM programs are composed of "Module"s, each of which is a
441 translation unit of the input programs. Each module consists of
442 functions, global variables, and symbol table entries. Modules may be
443 combined together with the LLVM linker, which merges function (and
444 global variable) definitions, resolves forward declarations, and merges
445 symbol table entries. Here is an example of the "hello world" module:</p>
447 <div class="doc_code">
448 <pre><i>; Declare the string constant as a global constant...</i>
449 <a href="#identifiers">@.LC0</a> = <a href="#linkage_internal">internal</a> <a
450 href="#globalvars">constant</a> <a href="#t_array">[13 x i8]</a> c"hello world\0A\00" <i>; [13 x i8]*</i>
452 <i>; External declaration of the puts function</i>
453 <a href="#functionstructure">declare</a> i32 @puts(i8 *) <i>; i32(i8 *)* </i>
455 <i>; Definition of main function</i>
456 define i32 @main() { <i>; i32()* </i>
457 <i>; Convert [13 x i8]* to i8 *...</i>
459 href="#i_getelementptr">getelementptr</a> [13 x i8]* @.LC0, i64 0, i64 0 <i>; i8 *</i>
461 <i>; Call puts function to write out the string to stdout...</i>
463 href="#i_call">call</a> i32 @puts(i8 * %cast210) <i>; i32</i>
465 href="#i_ret">ret</a> i32 0<br>}<br>
469 <p>This example is made up of a <a href="#globalvars">global variable</a>
470 named "<tt>.LC0</tt>", an external declaration of the "<tt>puts</tt>"
471 function, and a <a href="#functionstructure">function definition</a>
472 for "<tt>main</tt>".</p>
474 <p>In general, a module is made up of a list of global values,
475 where both functions and global variables are global values. Global values are
476 represented by a pointer to a memory location (in this case, a pointer to an
477 array of char, and a pointer to a function), and have one of the following <a
478 href="#linkage">linkage types</a>.</p>
482 <!-- ======================================================================= -->
483 <div class="doc_subsection">
484 <a name="linkage">Linkage Types</a>
487 <div class="doc_text">
490 All Global Variables and Functions have one of the following types of linkage:
495 <dt><tt><b><a name="linkage_private">private</a></b></tt>: </dt>
497 <dd>Global values with private linkage are only directly accessible by
498 objects in the current module. In particular, linking code into a module with
499 an private global value may cause the private to be renamed as necessary to
500 avoid collisions. Because the symbol is private to the module, all
501 references can be updated. This doesn't show up in any symbol table in the
505 <dt><tt><b><a name="linkage_internal">internal</a></b></tt>: </dt>
507 <dd> Similar to private, but the value shows as a local symbol (STB_LOCAL in
508 the case of ELF) in the object file. This corresponds to the notion of the
509 '<tt>static</tt>' keyword in C.
512 <dt><tt><b><a name="available_externally">available_externally</a></b></tt>:
515 <dd>Globals with "<tt>available_externally</tt>" linkage are never emitted
516 into the object file corresponding to the LLVM module. They exist to
517 allow inlining and other optimizations to take place given knowledge of the
518 definition of the global, which is known to be somewhere outside the module.
519 Globals with <tt>available_externally</tt> linkage are allowed to be discarded
520 at will, and are otherwise the same as <tt>linkonce_odr</tt>. This linkage
521 type is only allowed on definitions, not declarations.</dd>
523 <dt><tt><b><a name="linkage_linkonce">linkonce</a></b></tt>: </dt>
525 <dd>Globals with "<tt>linkonce</tt>" linkage are merged with other globals of
526 the same name when linkage occurs. This is typically used to implement
527 inline functions, templates, or other code which must be generated in each
528 translation unit that uses it. Unreferenced <tt>linkonce</tt> globals are
529 allowed to be discarded.
532 <dt><tt><b><a name="linkage_common">common</a></b></tt>: </dt>
534 <dd>"<tt>common</tt>" linkage is exactly the same as <tt>linkonce</tt>
535 linkage, except that unreferenced <tt>common</tt> globals may not be
536 discarded. This is used for globals that may be emitted in multiple
537 translation units, but that are not guaranteed to be emitted into every
538 translation unit that uses them. One example of this is tentative
539 definitions in C, such as "<tt>int X;</tt>" at global scope.
542 <dt><tt><b><a name="linkage_weak">weak</a></b></tt>: </dt>
544 <dd>"<tt>weak</tt>" linkage is the same as <tt>common</tt> linkage, except
545 that some targets may choose to emit different assembly sequences for them
546 for target-dependent reasons. This is used for globals that are declared
547 "weak" in C source code.
550 <dt><tt><b><a name="linkage_appending">appending</a></b></tt>: </dt>
552 <dd>"<tt>appending</tt>" linkage may only be applied to global variables of
553 pointer to array type. When two global variables with appending linkage are
554 linked together, the two global arrays are appended together. This is the
555 LLVM, typesafe, equivalent of having the system linker append together
556 "sections" with identical names when .o files are linked.
559 <dt><tt><b><a name="linkage_externweak">extern_weak</a></b></tt>: </dt>
561 <dd>The semantics of this linkage follow the ELF object file model: the
562 symbol is weak until linked, if not linked, the symbol becomes null instead
563 of being an undefined reference.
566 <dt><tt><b><a name="linkage_linkonce">linkonce_odr</a></b></tt>: </dt>
567 <dt><tt><b><a name="linkage_weak">weak_odr</a></b></tt>: </dt>
568 <dd>Some languages allow differing globals to be merged, such as two
569 functions with different semantics. Other languages, such as <tt>C++</tt>,
570 ensure that only equivalent globals are ever merged (the "one definition
571 rule" - "ODR"). Such languages can use the <tt>linkonce_odr</tt>
572 and <tt>weak_odr</tt> linkage types to indicate that the global will only
573 be merged with equivalent globals. These linkage types are otherwise the
574 same as their non-<tt>odr</tt> versions.
577 <dt><tt><b><a name="linkage_external">externally visible</a></b></tt>:</dt>
579 <dd>If none of the above identifiers are used, the global is externally
580 visible, meaning that it participates in linkage and can be used to resolve
581 external symbol references.
586 The next two types of linkage are targeted for Microsoft Windows platform
587 only. They are designed to support importing (exporting) symbols from (to)
588 DLLs (Dynamic Link Libraries).
592 <dt><tt><b><a name="linkage_dllimport">dllimport</a></b></tt>: </dt>
594 <dd>"<tt>dllimport</tt>" linkage causes the compiler to reference a function
595 or variable via a global pointer to a pointer that is set up by the DLL
596 exporting the symbol. On Microsoft Windows targets, the pointer name is
597 formed by combining <code>__imp_</code> and the function or variable name.
600 <dt><tt><b><a name="linkage_dllexport">dllexport</a></b></tt>: </dt>
602 <dd>"<tt>dllexport</tt>" linkage causes the compiler to provide a global
603 pointer to a pointer in a DLL, so that it can be referenced with the
604 <tt>dllimport</tt> attribute. On Microsoft Windows targets, the pointer
605 name is formed by combining <code>__imp_</code> and the function or variable
611 <p>For example, since the "<tt>.LC0</tt>"
612 variable is defined to be internal, if another module defined a "<tt>.LC0</tt>"
613 variable and was linked with this one, one of the two would be renamed,
614 preventing a collision. Since "<tt>main</tt>" and "<tt>puts</tt>" are
615 external (i.e., lacking any linkage declarations), they are accessible
616 outside of the current module.</p>
617 <p>It is illegal for a function <i>declaration</i>
618 to have any linkage type other than "externally visible", <tt>dllimport</tt>
619 or <tt>extern_weak</tt>.</p>
620 <p>Aliases can have only <tt>external</tt>, <tt>internal</tt>, <tt>weak</tt>
621 or <tt>weak_odr</tt> linkages.</p>
624 <!-- ======================================================================= -->
625 <div class="doc_subsection">
626 <a name="callingconv">Calling Conventions</a>
629 <div class="doc_text">
631 <p>LLVM <a href="#functionstructure">functions</a>, <a href="#i_call">calls</a>
632 and <a href="#i_invoke">invokes</a> can all have an optional calling convention
633 specified for the call. The calling convention of any pair of dynamic
634 caller/callee must match, or the behavior of the program is undefined. The
635 following calling conventions are supported by LLVM, and more may be added in
639 <dt><b>"<tt>ccc</tt>" - The C calling convention</b>:</dt>
641 <dd>This calling convention (the default if no other calling convention is
642 specified) matches the target C calling conventions. This calling convention
643 supports varargs function calls and tolerates some mismatch in the declared
644 prototype and implemented declaration of the function (as does normal C).
647 <dt><b>"<tt>fastcc</tt>" - The fast calling convention</b>:</dt>
649 <dd>This calling convention attempts to make calls as fast as possible
650 (e.g. by passing things in registers). This calling convention allows the
651 target to use whatever tricks it wants to produce fast code for the target,
652 without having to conform to an externally specified ABI (Application Binary
653 Interface). Implementations of this convention should allow arbitrary
654 <a href="CodeGenerator.html#tailcallopt">tail call optimization</a> to be
655 supported. This calling convention does not support varargs and requires the
656 prototype of all callees to exactly match the prototype of the function
660 <dt><b>"<tt>coldcc</tt>" - The cold calling convention</b>:</dt>
662 <dd>This calling convention attempts to make code in the caller as efficient
663 as possible under the assumption that the call is not commonly executed. As
664 such, these calls often preserve all registers so that the call does not break
665 any live ranges in the caller side. This calling convention does not support
666 varargs and requires the prototype of all callees to exactly match the
667 prototype of the function definition.
670 <dt><b>"<tt>cc <<em>n</em>></tt>" - Numbered convention</b>:</dt>
672 <dd>Any calling convention may be specified by number, allowing
673 target-specific calling conventions to be used. Target specific calling
674 conventions start at 64.
678 <p>More calling conventions can be added/defined on an as-needed basis, to
679 support pascal conventions or any other well-known target-independent
684 <!-- ======================================================================= -->
685 <div class="doc_subsection">
686 <a name="visibility">Visibility Styles</a>
689 <div class="doc_text">
692 All Global Variables and Functions have one of the following visibility styles:
696 <dt><b>"<tt>default</tt>" - Default style</b>:</dt>
698 <dd>On targets that use the ELF object file format, default visibility means
699 that the declaration is visible to other
700 modules and, in shared libraries, means that the declared entity may be
701 overridden. On Darwin, default visibility means that the declaration is
702 visible to other modules. Default visibility corresponds to "external
703 linkage" in the language.
706 <dt><b>"<tt>hidden</tt>" - Hidden style</b>:</dt>
708 <dd>Two declarations of an object with hidden visibility refer to the same
709 object if they are in the same shared object. Usually, hidden visibility
710 indicates that the symbol will not be placed into the dynamic symbol table,
711 so no other module (executable or shared library) can reference it
715 <dt><b>"<tt>protected</tt>" - Protected style</b>:</dt>
717 <dd>On ELF, protected visibility indicates that the symbol will be placed in
718 the dynamic symbol table, but that references within the defining module will
719 bind to the local symbol. That is, the symbol cannot be overridden by another
726 <!-- ======================================================================= -->
727 <div class="doc_subsection">
728 <a name="namedtypes">Named Types</a>
731 <div class="doc_text">
733 <p>LLVM IR allows you to specify name aliases for certain types. This can make
734 it easier to read the IR and make the IR more condensed (particularly when
735 recursive types are involved). An example of a name specification is:
738 <div class="doc_code">
740 %mytype = type { %mytype*, i32 }
744 <p>You may give a name to any <a href="#typesystem">type</a> except "<a
745 href="t_void">void</a>". Type name aliases may be used anywhere a type is
746 expected with the syntax "%mytype".</p>
748 <p>Note that type names are aliases for the structural type that they indicate,
749 and that you can therefore specify multiple names for the same type. This often
750 leads to confusing behavior when dumping out a .ll file. Since LLVM IR uses
751 structural typing, the name is not part of the type. When printing out LLVM IR,
752 the printer will pick <em>one name</em> to render all types of a particular
753 shape. This means that if you have code where two different source types end up
754 having the same LLVM type, that the dumper will sometimes print the "wrong" or
755 unexpected type. This is an important design point and isn't going to
760 <!-- ======================================================================= -->
761 <div class="doc_subsection">
762 <a name="globalvars">Global Variables</a>
765 <div class="doc_text">
767 <p>Global variables define regions of memory allocated at compilation time
768 instead of run-time. Global variables may optionally be initialized, may have
769 an explicit section to be placed in, and may have an optional explicit alignment
770 specified. A variable may be defined as "thread_local", which means that it
771 will not be shared by threads (each thread will have a separated copy of the
772 variable). A variable may be defined as a global "constant," which indicates
773 that the contents of the variable will <b>never</b> be modified (enabling better
774 optimization, allowing the global data to be placed in the read-only section of
775 an executable, etc). Note that variables that need runtime initialization
776 cannot be marked "constant" as there is a store to the variable.</p>
779 LLVM explicitly allows <em>declarations</em> of global variables to be marked
780 constant, even if the final definition of the global is not. This capability
781 can be used to enable slightly better optimization of the program, but requires
782 the language definition to guarantee that optimizations based on the
783 'constantness' are valid for the translation units that do not include the
787 <p>As SSA values, global variables define pointer values that are in
788 scope (i.e. they dominate) all basic blocks in the program. Global
789 variables always define a pointer to their "content" type because they
790 describe a region of memory, and all memory objects in LLVM are
791 accessed through pointers.</p>
793 <p>A global variable may be declared to reside in a target-specifc numbered
794 address space. For targets that support them, address spaces may affect how
795 optimizations are performed and/or what target instructions are used to access
796 the variable. The default address space is zero. The address space qualifier
797 must precede any other attributes.</p>
799 <p>LLVM allows an explicit section to be specified for globals. If the target
800 supports it, it will emit globals to the section specified.</p>
802 <p>An explicit alignment may be specified for a global. If not present, or if
803 the alignment is set to zero, the alignment of the global is set by the target
804 to whatever it feels convenient. If an explicit alignment is specified, the
805 global is forced to have at least that much alignment. All alignments must be
808 <p>For example, the following defines a global in a numbered address space with
809 an initializer, section, and alignment:</p>
811 <div class="doc_code">
813 @G = addrspace(5) constant float 1.0, section "foo", align 4
820 <!-- ======================================================================= -->
821 <div class="doc_subsection">
822 <a name="functionstructure">Functions</a>
825 <div class="doc_text">
827 <p>LLVM function definitions consist of the "<tt>define</tt>" keyord,
828 an optional <a href="#linkage">linkage type</a>, an optional
829 <a href="#visibility">visibility style</a>, an optional
830 <a href="#callingconv">calling convention</a>, a return type, an optional
831 <a href="#paramattrs">parameter attribute</a> for the return type, a function
832 name, a (possibly empty) argument list (each with optional
833 <a href="#paramattrs">parameter attributes</a>), optional
834 <a href="#fnattrs">function attributes</a>, an optional section,
835 an optional alignment, an optional <a href="#gc">garbage collector name</a>,
836 an opening curly brace, a list of basic blocks, and a closing curly brace.
838 LLVM function declarations consist of the "<tt>declare</tt>" keyword, an
839 optional <a href="#linkage">linkage type</a>, an optional
840 <a href="#visibility">visibility style</a>, an optional
841 <a href="#callingconv">calling convention</a>, a return type, an optional
842 <a href="#paramattrs">parameter attribute</a> for the return type, a function
843 name, a possibly empty list of arguments, an optional alignment, and an optional
844 <a href="#gc">garbage collector name</a>.</p>
846 <p>A function definition contains a list of basic blocks, forming the CFG
847 (Control Flow Graph) for
848 the function. Each basic block may optionally start with a label (giving the
849 basic block a symbol table entry), contains a list of instructions, and ends
850 with a <a href="#terminators">terminator</a> instruction (such as a branch or
851 function return).</p>
853 <p>The first basic block in a function is special in two ways: it is immediately
854 executed on entrance to the function, and it is not allowed to have predecessor
855 basic blocks (i.e. there can not be any branches to the entry block of a
856 function). Because the block can have no predecessors, it also cannot have any
857 <a href="#i_phi">PHI nodes</a>.</p>
859 <p>LLVM allows an explicit section to be specified for functions. If the target
860 supports it, it will emit functions to the section specified.</p>
862 <p>An explicit alignment may be specified for a function. If not present, or if
863 the alignment is set to zero, the alignment of the function is set by the target
864 to whatever it feels convenient. If an explicit alignment is specified, the
865 function is forced to have at least that much alignment. All alignments must be
870 <div class="doc_code">
872 define [<a href="#linkage">linkage</a>] [<a href="#visibility">visibility</a>]
873 [<a href="#callingconv">cconv</a>] [<a href="#paramattrs">ret attrs</a>]
874 <ResultType> @<FunctionName> ([argument list])
875 [<a href="#fnattrs">fn Attrs</a>] [section "name"] [align N]
876 [<a href="#gc">gc</a>] { ... }
883 <!-- ======================================================================= -->
884 <div class="doc_subsection">
885 <a name="aliasstructure">Aliases</a>
887 <div class="doc_text">
888 <p>Aliases act as "second name" for the aliasee value (which can be either
889 function, global variable, another alias or bitcast of global value). Aliases
890 may have an optional <a href="#linkage">linkage type</a>, and an
891 optional <a href="#visibility">visibility style</a>.</p>
895 <div class="doc_code">
897 @<Name> = alias [Linkage] [Visibility] <AliaseeTy> @<Aliasee>
905 <!-- ======================================================================= -->
906 <div class="doc_subsection"><a name="paramattrs">Parameter Attributes</a></div>
907 <div class="doc_text">
908 <p>The return type and each parameter of a function type may have a set of
909 <i>parameter attributes</i> associated with them. Parameter attributes are
910 used to communicate additional information about the result or parameters of
911 a function. Parameter attributes are considered to be part of the function,
912 not of the function type, so functions with different parameter attributes
913 can have the same function type.</p>
915 <p>Parameter attributes are simple keywords that follow the type specified. If
916 multiple parameter attributes are needed, they are space separated. For
919 <div class="doc_code">
921 declare i32 @printf(i8* noalias nocapture, ...)
922 declare i32 @atoi(i8 zeroext)
923 declare signext i8 @returns_signed_char()
927 <p>Note that any attributes for the function result (<tt>nounwind</tt>,
928 <tt>readonly</tt>) come immediately after the argument list.</p>
930 <p>Currently, only the following parameter attributes are defined:</p>
932 <dt><tt>zeroext</tt></dt>
933 <dd>This indicates to the code generator that the parameter or return value
934 should be zero-extended to a 32-bit value by the caller (for a parameter)
935 or the callee (for a return value).</dd>
937 <dt><tt>signext</tt></dt>
938 <dd>This indicates to the code generator that the parameter or return value
939 should be sign-extended to a 32-bit value by the caller (for a parameter)
940 or the callee (for a return value).</dd>
942 <dt><tt>inreg</tt></dt>
943 <dd>This indicates that this parameter or return value should be treated
944 in a special target-dependent fashion during while emitting code for a
945 function call or return (usually, by putting it in a register as opposed
946 to memory, though some targets use it to distinguish between two different
947 kinds of registers). Use of this attribute is target-specific.</dd>
949 <dt><tt><a name="byval">byval</a></tt></dt>
950 <dd>This indicates that the pointer parameter should really be passed by
951 value to the function. The attribute implies that a hidden copy of the
952 pointee is made between the caller and the callee, so the callee is unable
953 to modify the value in the callee. This attribute is only valid on LLVM
954 pointer arguments. It is generally used to pass structs and arrays by
955 value, but is also valid on pointers to scalars. The copy is considered to
956 belong to the caller not the callee (for example,
957 <tt><a href="#readonly">readonly</a></tt> functions should not write to
958 <tt>byval</tt> parameters). This is not a valid attribute for return
959 values. The byval attribute also supports specifying an alignment with the
960 align attribute. This has a target-specific effect on the code generator
961 that usually indicates a desired alignment for the synthesized stack
964 <dt><tt>sret</tt></dt>
965 <dd>This indicates that the pointer parameter specifies the address of a
966 structure that is the return value of the function in the source program.
967 This pointer must be guaranteed by the caller to be valid: loads and stores
968 to the structure may be assumed by the callee to not to trap. This may only
969 be applied to the first parameter. This is not a valid attribute for
972 <dt><tt>noalias</tt></dt>
973 <dd>This indicates that the pointer does not alias any global or any other
974 parameter. The caller is responsible for ensuring that this is the
975 case. On a function return value, <tt>noalias</tt> additionally indicates
976 that the pointer does not alias any other pointers visible to the
977 caller. For further details, please see the discussion of the NoAlias
979 <a href="http://llvm.org/docs/AliasAnalysis.html#MustMayNo">alias
982 <dt><tt>nocapture</tt></dt>
983 <dd>This indicates that the callee does not make any copies of the pointer
984 that outlive the callee itself. This is not a valid attribute for return
987 <dt><tt>nest</tt></dt>
988 <dd>This indicates that the pointer parameter can be excised using the
989 <a href="#int_trampoline">trampoline intrinsics</a>. This is not a valid
990 attribute for return values.</dd>
995 <!-- ======================================================================= -->
996 <div class="doc_subsection">
997 <a name="gc">Garbage Collector Names</a>
1000 <div class="doc_text">
1001 <p>Each function may specify a garbage collector name, which is simply a
1004 <div class="doc_code"><pre
1005 >define void @f() gc "name" { ...</pre></div>
1007 <p>The compiler declares the supported values of <i>name</i>. Specifying a
1008 collector which will cause the compiler to alter its output in order to support
1009 the named garbage collection algorithm.</p>
1012 <!-- ======================================================================= -->
1013 <div class="doc_subsection">
1014 <a name="fnattrs">Function Attributes</a>
1017 <div class="doc_text">
1019 <p>Function attributes are set to communicate additional information about
1020 a function. Function attributes are considered to be part of the function,
1021 not of the function type, so functions with different parameter attributes
1022 can have the same function type.</p>
1024 <p>Function attributes are simple keywords that follow the type specified. If
1025 multiple attributes are needed, they are space separated. For
1028 <div class="doc_code">
1030 define void @f() noinline { ... }
1031 define void @f() alwaysinline { ... }
1032 define void @f() alwaysinline optsize { ... }
1033 define void @f() optsize
1038 <dt><tt>alwaysinline</tt></dt>
1039 <dd>This attribute indicates that the inliner should attempt to inline this
1040 function into callers whenever possible, ignoring any active inlining size
1041 threshold for this caller.</dd>
1043 <dt><tt>noinline</tt></dt>
1044 <dd>This attribute indicates that the inliner should never inline this function
1045 in any situation. This attribute may not be used together with the
1046 <tt>alwaysinline</tt> attribute.</dd>
1048 <dt><tt>optsize</tt></dt>
1049 <dd>This attribute suggests that optimization passes and code generator passes
1050 make choices that keep the code size of this function low, and otherwise do
1051 optimizations specifically to reduce code size.</dd>
1053 <dt><tt>noreturn</tt></dt>
1054 <dd>This function attribute indicates that the function never returns normally.
1055 This produces undefined behavior at runtime if the function ever does
1056 dynamically return.</dd>
1058 <dt><tt>nounwind</tt></dt>
1059 <dd>This function attribute indicates that the function never returns with an
1060 unwind or exceptional control flow. If the function does unwind, its runtime
1061 behavior is undefined.</dd>
1063 <dt><tt>readnone</tt></dt>
1064 <dd>This attribute indicates that the function computes its result (or decides to
1065 unwind an exception) based strictly on its arguments, without dereferencing any
1066 pointer arguments or otherwise accessing any mutable state (e.g. memory, control
1067 registers, etc) visible to caller functions. It does not write through any
1068 pointer arguments (including <tt><a href="#byval">byval</a></tt> arguments) and
1069 never changes any state visible to callers. This means that it cannot unwind
1070 exceptions by calling the <tt>C++</tt> exception throwing methods, but could
1071 use the <tt>unwind</tt> instruction.</dd>
1073 <dt><tt><a name="readonly">readonly</a></tt></dt>
1074 <dd>This attribute indicates that the function does not write through any
1075 pointer arguments (including <tt><a href="#byval">byval</a></tt> arguments)
1076 or otherwise modify any state (e.g. memory, control registers, etc) visible to
1077 caller functions. It may dereference pointer arguments and read state that may
1078 be set in the caller. A readonly function always returns the same value (or
1079 unwinds an exception identically) when called with the same set of arguments
1080 and global state. It cannot unwind an exception by calling the <tt>C++</tt>
1081 exception throwing methods, but may use the <tt>unwind</tt> instruction.</dd>
1083 <dt><tt><a name="ssp">ssp</a></tt></dt>
1084 <dd>This attribute indicates that the function should emit a stack smashing
1085 protector. It is in the form of a "canary"—a random value placed on the
1086 stack before the local variables that's checked upon return from the function to
1087 see if it has been overwritten. A heuristic is used to determine if a function
1088 needs stack protectors or not.
1090 <br><br>If a function that has an <tt>ssp</tt> attribute is inlined into a function
1091 that doesn't have an <tt>ssp</tt> attribute, then the resulting function will
1092 have an <tt>ssp</tt> attribute.</dd>
1094 <dt><tt>sspreq</tt></dt>
1095 <dd>This attribute indicates that the function should <em>always</em> emit a
1096 stack smashing protector. This overrides the <tt><a href="#ssp">ssp</a></tt>
1099 If a function that has an <tt>sspreq</tt> attribute is inlined into a
1100 function that doesn't have an <tt>sspreq</tt> attribute or which has
1101 an <tt>ssp</tt> attribute, then the resulting function will have
1102 an <tt>sspreq</tt> attribute.</dd>
1104 <dt><tt>noredzone</tt></dt>
1105 <dd>This attribute indicates that the code generator should not use a
1106 red zone, even if the target-specific ABI normally permits it.
1109 <dt><tt>noimplicitfloat</tt></dt>
1110 <dd>This attributes disables implicit floating point instructions.</dd>
1112 <dt><tt>naked</tt></dt>
1113 <dd>This attribute disables prologue / epilogue emission for the function</dd>
1119 <!-- ======================================================================= -->
1120 <div class="doc_subsection">
1121 <a name="moduleasm">Module-Level Inline Assembly</a>
1124 <div class="doc_text">
1126 Modules may contain "module-level inline asm" blocks, which corresponds to the
1127 GCC "file scope inline asm" blocks. These blocks are internally concatenated by
1128 LLVM and treated as a single unit, but may be separated in the .ll file if
1129 desired. The syntax is very simple:
1132 <div class="doc_code">
1134 module asm "inline asm code goes here"
1135 module asm "more can go here"
1139 <p>The strings can contain any character by escaping non-printable characters.
1140 The escape sequence used is simply "\xx" where "xx" is the two digit hex code
1145 The inline asm code is simply printed to the machine code .s file when
1146 assembly code is generated.
1150 <!-- ======================================================================= -->
1151 <div class="doc_subsection">
1152 <a name="datalayout">Data Layout</a>
1155 <div class="doc_text">
1156 <p>A module may specify a target specific data layout string that specifies how
1157 data is to be laid out in memory. The syntax for the data layout is simply:</p>
1158 <pre> target datalayout = "<i>layout specification</i>"</pre>
1159 <p>The <i>layout specification</i> consists of a list of specifications
1160 separated by the minus sign character ('-'). Each specification starts with a
1161 letter and may include other information after the letter to define some
1162 aspect of the data layout. The specifications accepted are as follows: </p>
1165 <dd>Specifies that the target lays out data in big-endian form. That is, the
1166 bits with the most significance have the lowest address location.</dd>
1168 <dd>Specifies that the target lays out data in little-endian form. That is,
1169 the bits with the least significance have the lowest address location.</dd>
1170 <dt><tt>p:<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1171 <dd>This specifies the <i>size</i> of a pointer and its <i>abi</i> and
1172 <i>preferred</i> alignments. All sizes are in bits. Specifying the <i>pref</i>
1173 alignment is optional. If omitted, the preceding <tt>:</tt> should be omitted
1175 <dt><tt>i<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1176 <dd>This specifies the alignment for an integer type of a given bit
1177 <i>size</i>. The value of <i>size</i> must be in the range [1,2^23).</dd>
1178 <dt><tt>v<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1179 <dd>This specifies the alignment for a vector type of a given bit
1181 <dt><tt>f<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1182 <dd>This specifies the alignment for a floating point type of a given bit
1183 <i>size</i>. The value of <i>size</i> must be either 32 (float) or 64
1185 <dt><tt>a<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1186 <dd>This specifies the alignment for an aggregate type of a given bit
1188 <dt><tt>s<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1189 <dd>This specifies the alignment for a stack object of a given bit
1192 <p>When constructing the data layout for a given target, LLVM starts with a
1193 default set of specifications which are then (possibly) overriden by the
1194 specifications in the <tt>datalayout</tt> keyword. The default specifications
1195 are given in this list:</p>
1197 <li><tt>E</tt> - big endian</li>
1198 <li><tt>p:32:64:64</tt> - 32-bit pointers with 64-bit alignment</li>
1199 <li><tt>i1:8:8</tt> - i1 is 8-bit (byte) aligned</li>
1200 <li><tt>i8:8:8</tt> - i8 is 8-bit (byte) aligned</li>
1201 <li><tt>i16:16:16</tt> - i16 is 16-bit aligned</li>
1202 <li><tt>i32:32:32</tt> - i32 is 32-bit aligned</li>
1203 <li><tt>i64:32:64</tt> - i64 has ABI alignment of 32-bits but preferred
1204 alignment of 64-bits</li>
1205 <li><tt>f32:32:32</tt> - float is 32-bit aligned</li>
1206 <li><tt>f64:64:64</tt> - double is 64-bit aligned</li>
1207 <li><tt>v64:64:64</tt> - 64-bit vector is 64-bit aligned</li>
1208 <li><tt>v128:128:128</tt> - 128-bit vector is 128-bit aligned</li>
1209 <li><tt>a0:0:1</tt> - aggregates are 8-bit aligned</li>
1210 <li><tt>s0:64:64</tt> - stack objects are 64-bit aligned</li>
1212 <p>When LLVM is determining the alignment for a given type, it uses the
1213 following rules:</p>
1215 <li>If the type sought is an exact match for one of the specifications, that
1216 specification is used.</li>
1217 <li>If no match is found, and the type sought is an integer type, then the
1218 smallest integer type that is larger than the bitwidth of the sought type is
1219 used. If none of the specifications are larger than the bitwidth then the the
1220 largest integer type is used. For example, given the default specifications
1221 above, the i7 type will use the alignment of i8 (next largest) while both
1222 i65 and i256 will use the alignment of i64 (largest specified).</li>
1223 <li>If no match is found, and the type sought is a vector type, then the
1224 largest vector type that is smaller than the sought vector type will be used
1225 as a fall back. This happens because <128 x double> can be implemented
1226 in terms of 64 <2 x double>, for example.</li>
1230 <!-- *********************************************************************** -->
1231 <div class="doc_section"> <a name="typesystem">Type System</a> </div>
1232 <!-- *********************************************************************** -->
1234 <div class="doc_text">
1236 <p>The LLVM type system is one of the most important features of the
1237 intermediate representation. Being typed enables a number of
1238 optimizations to be performed on the intermediate representation directly,
1239 without having to do
1240 extra analyses on the side before the transformation. A strong type
1241 system makes it easier to read the generated code and enables novel
1242 analyses and transformations that are not feasible to perform on normal
1243 three address code representations.</p>
1247 <!-- ======================================================================= -->
1248 <div class="doc_subsection"> <a name="t_classifications">Type
1249 Classifications</a> </div>
1250 <div class="doc_text">
1251 <p>The types fall into a few useful
1252 classifications:</p>
1254 <table border="1" cellspacing="0" cellpadding="4">
1256 <tr><th>Classification</th><th>Types</th></tr>
1258 <td><a href="#t_integer">integer</a></td>
1259 <td><tt>i1, i2, i3, ... i8, ... i16, ... i32, ... i64, ... </tt></td>
1262 <td><a href="#t_floating">floating point</a></td>
1263 <td><tt>float, double, x86_fp80, fp128, ppc_fp128</tt></td>
1266 <td><a name="t_firstclass">first class</a></td>
1267 <td><a href="#t_integer">integer</a>,
1268 <a href="#t_floating">floating point</a>,
1269 <a href="#t_pointer">pointer</a>,
1270 <a href="#t_vector">vector</a>,
1271 <a href="#t_struct">structure</a>,
1272 <a href="#t_array">array</a>,
1273 <a href="#t_label">label</a>,
1274 <a href="#t_metadata">metadata</a>.
1278 <td><a href="#t_primitive">primitive</a></td>
1279 <td><a href="#t_label">label</a>,
1280 <a href="#t_void">void</a>,
1281 <a href="#t_floating">floating point</a>,
1282 <a href="#t_metadata">metadata</a>.</td>
1285 <td><a href="#t_derived">derived</a></td>
1286 <td><a href="#t_integer">integer</a>,
1287 <a href="#t_array">array</a>,
1288 <a href="#t_function">function</a>,
1289 <a href="#t_pointer">pointer</a>,
1290 <a href="#t_struct">structure</a>,
1291 <a href="#t_pstruct">packed structure</a>,
1292 <a href="#t_vector">vector</a>,
1293 <a href="#t_opaque">opaque</a>.
1299 <p>The <a href="#t_firstclass">first class</a> types are perhaps the
1300 most important. Values of these types are the only ones which can be
1301 produced by instructions, passed as arguments, or used as operands to
1305 <!-- ======================================================================= -->
1306 <div class="doc_subsection"> <a name="t_primitive">Primitive Types</a> </div>
1308 <div class="doc_text">
1309 <p>The primitive types are the fundamental building blocks of the LLVM
1314 <!-- _______________________________________________________________________ -->
1315 <div class="doc_subsubsection"> <a name="t_floating">Floating Point Types</a> </div>
1317 <div class="doc_text">
1320 <tr><th>Type</th><th>Description</th></tr>
1321 <tr><td><tt>float</tt></td><td>32-bit floating point value</td></tr>
1322 <tr><td><tt>double</tt></td><td>64-bit floating point value</td></tr>
1323 <tr><td><tt>fp128</tt></td><td>128-bit floating point value (112-bit mantissa)</td></tr>
1324 <tr><td><tt>x86_fp80</tt></td><td>80-bit floating point value (X87)</td></tr>
1325 <tr><td><tt>ppc_fp128</tt></td><td>128-bit floating point value (two 64-bits)</td></tr>
1330 <!-- _______________________________________________________________________ -->
1331 <div class="doc_subsubsection"> <a name="t_void">Void Type</a> </div>
1333 <div class="doc_text">
1335 <p>The void type does not represent any value and has no size.</p>
1344 <!-- _______________________________________________________________________ -->
1345 <div class="doc_subsubsection"> <a name="t_label">Label Type</a> </div>
1347 <div class="doc_text">
1349 <p>The label type represents code labels.</p>
1358 <!-- _______________________________________________________________________ -->
1359 <div class="doc_subsubsection"> <a name="t_metadata">Metadata Type</a> </div>
1361 <div class="doc_text">
1363 <p>The metadata type represents embedded metadata. The only derived type that
1364 may contain metadata is <tt>metadata*</tt> or a function type that returns or
1365 takes metadata typed parameters, but not pointer to metadata types.</p>
1375 <!-- ======================================================================= -->
1376 <div class="doc_subsection"> <a name="t_derived">Derived Types</a> </div>
1378 <div class="doc_text">
1380 <p>The real power in LLVM comes from the derived types in the system.
1381 This is what allows a programmer to represent arrays, functions,
1382 pointers, and other useful types. Note that these derived types may be
1383 recursive: For example, it is possible to have a two dimensional array.</p>
1387 <!-- _______________________________________________________________________ -->
1388 <div class="doc_subsubsection"> <a name="t_integer">Integer Type</a> </div>
1390 <div class="doc_text">
1393 <p>The integer type is a very simple derived type that simply specifies an
1394 arbitrary bit width for the integer type desired. Any bit width from 1 bit to
1395 2^23-1 (about 8 million) can be specified.</p>
1403 <p>The number of bits the integer will occupy is specified by the <tt>N</tt>
1407 <table class="layout">
1409 <td class="left"><tt>i1</tt></td>
1410 <td class="left">a single-bit integer.</td>
1413 <td class="left"><tt>i32</tt></td>
1414 <td class="left">a 32-bit integer.</td>
1417 <td class="left"><tt>i1942652</tt></td>
1418 <td class="left">a really big integer of over 1 million bits.</td>
1422 <p>Note that the code generator does not yet support large integer types
1423 to be used as function return types. The specific limit on how large a
1424 return type the code generator can currently handle is target-dependent;
1425 currently it's often 64 bits for 32-bit targets and 128 bits for 64-bit
1430 <!-- _______________________________________________________________________ -->
1431 <div class="doc_subsubsection"> <a name="t_array">Array Type</a> </div>
1433 <div class="doc_text">
1437 <p>The array type is a very simple derived type that arranges elements
1438 sequentially in memory. The array type requires a size (number of
1439 elements) and an underlying data type.</p>
1444 [<# elements> x <elementtype>]
1447 <p>The number of elements is a constant integer value; elementtype may
1448 be any type with a size.</p>
1451 <table class="layout">
1453 <td class="left"><tt>[40 x i32]</tt></td>
1454 <td class="left">Array of 40 32-bit integer values.</td>
1457 <td class="left"><tt>[41 x i32]</tt></td>
1458 <td class="left">Array of 41 32-bit integer values.</td>
1461 <td class="left"><tt>[4 x i8]</tt></td>
1462 <td class="left">Array of 4 8-bit integer values.</td>
1465 <p>Here are some examples of multidimensional arrays:</p>
1466 <table class="layout">
1468 <td class="left"><tt>[3 x [4 x i32]]</tt></td>
1469 <td class="left">3x4 array of 32-bit integer values.</td>
1472 <td class="left"><tt>[12 x [10 x float]]</tt></td>
1473 <td class="left">12x10 array of single precision floating point values.</td>
1476 <td class="left"><tt>[2 x [3 x [4 x i16]]]</tt></td>
1477 <td class="left">2x3x4 array of 16-bit integer values.</td>
1481 <p>Note that 'variable sized arrays' can be implemented in LLVM with a zero
1482 length array. Normally, accesses past the end of an array are undefined in
1483 LLVM (e.g. it is illegal to access the 5th element of a 3 element array).
1484 As a special case, however, zero length arrays are recognized to be variable
1485 length. This allows implementation of 'pascal style arrays' with the LLVM
1486 type "{ i32, [0 x float]}", for example.</p>
1488 <p>Note that the code generator does not yet support large aggregate types
1489 to be used as function return types. The specific limit on how large an
1490 aggregate return type the code generator can currently handle is
1491 target-dependent, and also dependent on the aggregate element types.</p>
1495 <!-- _______________________________________________________________________ -->
1496 <div class="doc_subsubsection"> <a name="t_function">Function Type</a> </div>
1497 <div class="doc_text">
1501 <p>The function type can be thought of as a function signature. It
1502 consists of a return type and a list of formal parameter types. The
1503 return type of a function type is a scalar type, a void type, or a struct type.
1504 If the return type is a struct type then all struct elements must be of first
1505 class types, and the struct must have at least one element.</p>
1510 <returntype list> (<parameter list>)
1513 <p>...where '<tt><parameter list></tt>' is a comma-separated list of type
1514 specifiers. Optionally, the parameter list may include a type <tt>...</tt>,
1515 which indicates that the function takes a variable number of arguments.
1516 Variable argument functions can access their arguments with the <a
1517 href="#int_varargs">variable argument handling intrinsic</a> functions.
1518 '<tt><returntype list></tt>' is a comma-separated list of
1519 <a href="#t_firstclass">first class</a> type specifiers.</p>
1522 <table class="layout">
1524 <td class="left"><tt>i32 (i32)</tt></td>
1525 <td class="left">function taking an <tt>i32</tt>, returning an <tt>i32</tt>
1527 </tr><tr class="layout">
1528 <td class="left"><tt>float (i16 signext, i32 *) *
1530 <td class="left"><a href="#t_pointer">Pointer</a> to a function that takes
1531 an <tt>i16</tt> that should be sign extended and a
1532 <a href="#t_pointer">pointer</a> to <tt>i32</tt>, returning
1535 </tr><tr class="layout">
1536 <td class="left"><tt>i32 (i8*, ...)</tt></td>
1537 <td class="left">A vararg function that takes at least one
1538 <a href="#t_pointer">pointer</a> to <tt>i8 </tt> (char in C),
1539 which returns an integer. This is the signature for <tt>printf</tt> in
1542 </tr><tr class="layout">
1543 <td class="left"><tt>{i32, i32} (i32)</tt></td>
1544 <td class="left">A function taking an <tt>i32</tt>, returning two
1545 <tt>i32</tt> values as an aggregate of type <tt>{ i32, i32 }</tt>
1551 <!-- _______________________________________________________________________ -->
1552 <div class="doc_subsubsection"> <a name="t_struct">Structure Type</a> </div>
1553 <div class="doc_text">
1555 <p>The structure type is used to represent a collection of data members
1556 together in memory. The packing of the field types is defined to match
1557 the ABI of the underlying processor. The elements of a structure may
1558 be any type that has a size.</p>
1559 <p>Structures are accessed using '<tt><a href="#i_load">load</a></tt>
1560 and '<tt><a href="#i_store">store</a></tt>' by getting a pointer to a
1561 field with the '<tt><a href="#i_getelementptr">getelementptr</a></tt>'
1564 <pre> { <type list> }<br></pre>
1566 <table class="layout">
1568 <td class="left"><tt>{ i32, i32, i32 }</tt></td>
1569 <td class="left">A triple of three <tt>i32</tt> values</td>
1570 </tr><tr class="layout">
1571 <td class="left"><tt>{ float, i32 (i32) * }</tt></td>
1572 <td class="left">A pair, where the first element is a <tt>float</tt> and the
1573 second element is a <a href="#t_pointer">pointer</a> to a
1574 <a href="#t_function">function</a> that takes an <tt>i32</tt>, returning
1575 an <tt>i32</tt>.</td>
1579 <p>Note that the code generator does not yet support large aggregate types
1580 to be used as function return types. The specific limit on how large an
1581 aggregate return type the code generator can currently handle is
1582 target-dependent, and also dependent on the aggregate element types.</p>
1586 <!-- _______________________________________________________________________ -->
1587 <div class="doc_subsubsection"> <a name="t_pstruct">Packed Structure Type</a>
1589 <div class="doc_text">
1591 <p>The packed structure type is used to represent a collection of data members
1592 together in memory. There is no padding between fields. Further, the alignment
1593 of a packed structure is 1 byte. The elements of a packed structure may
1594 be any type that has a size.</p>
1595 <p>Structures are accessed using '<tt><a href="#i_load">load</a></tt>
1596 and '<tt><a href="#i_store">store</a></tt>' by getting a pointer to a
1597 field with the '<tt><a href="#i_getelementptr">getelementptr</a></tt>'
1600 <pre> < { <type list> } > <br></pre>
1602 <table class="layout">
1604 <td class="left"><tt>< { i32, i32, i32 } ></tt></td>
1605 <td class="left">A triple of three <tt>i32</tt> values</td>
1606 </tr><tr class="layout">
1608 <tt>< { float, i32 (i32)* } ></tt></td>
1609 <td class="left">A pair, where the first element is a <tt>float</tt> and the
1610 second element is a <a href="#t_pointer">pointer</a> to a
1611 <a href="#t_function">function</a> that takes an <tt>i32</tt>, returning
1612 an <tt>i32</tt>.</td>
1617 <!-- _______________________________________________________________________ -->
1618 <div class="doc_subsubsection"> <a name="t_pointer">Pointer Type</a> </div>
1619 <div class="doc_text">
1621 <p>As in many languages, the pointer type represents a pointer or
1622 reference to another object, which must live in memory. Pointer types may have
1623 an optional address space attribute defining the target-specific numbered
1624 address space where the pointed-to object resides. The default address space is
1627 <p>Note that LLVM does not permit pointers to void (<tt>void*</tt>) nor does
1628 it permit pointers to labels (<tt>label*</tt>). Use <tt>i8*</tt> instead.</p>
1631 <pre> <type> *<br></pre>
1633 <table class="layout">
1635 <td class="left"><tt>[4 x i32]*</tt></td>
1636 <td class="left">A <a href="#t_pointer">pointer</a> to <a
1637 href="#t_array">array</a> of four <tt>i32</tt> values.</td>
1640 <td class="left"><tt>i32 (i32 *) *</tt></td>
1641 <td class="left"> A <a href="#t_pointer">pointer</a> to a <a
1642 href="#t_function">function</a> that takes an <tt>i32*</tt>, returning an
1646 <td class="left"><tt>i32 addrspace(5)*</tt></td>
1647 <td class="left">A <a href="#t_pointer">pointer</a> to an <tt>i32</tt> value
1648 that resides in address space #5.</td>
1653 <!-- _______________________________________________________________________ -->
1654 <div class="doc_subsubsection"> <a name="t_vector">Vector Type</a> </div>
1655 <div class="doc_text">
1659 <p>A vector type is a simple derived type that represents a vector
1660 of elements. Vector types are used when multiple primitive data
1661 are operated in parallel using a single instruction (SIMD).
1662 A vector type requires a size (number of
1663 elements) and an underlying primitive data type. Vectors must have a power
1664 of two length (1, 2, 4, 8, 16 ...). Vector types are
1665 considered <a href="#t_firstclass">first class</a>.</p>
1670 < <# elements> x <elementtype> >
1673 <p>The number of elements is a constant integer value; elementtype may
1674 be any integer or floating point type.</p>
1678 <table class="layout">
1680 <td class="left"><tt><4 x i32></tt></td>
1681 <td class="left">Vector of 4 32-bit integer values.</td>
1684 <td class="left"><tt><8 x float></tt></td>
1685 <td class="left">Vector of 8 32-bit floating-point values.</td>
1688 <td class="left"><tt><2 x i64></tt></td>
1689 <td class="left">Vector of 2 64-bit integer values.</td>
1693 <p>Note that the code generator does not yet support large vector types
1694 to be used as function return types. The specific limit on how large a
1695 vector return type codegen can currently handle is target-dependent;
1696 currently it's often a few times longer than a hardware vector register.</p>
1700 <!-- _______________________________________________________________________ -->
1701 <div class="doc_subsubsection"> <a name="t_opaque">Opaque Type</a> </div>
1702 <div class="doc_text">
1706 <p>Opaque types are used to represent unknown types in the system. This
1707 corresponds (for example) to the C notion of a forward declared structure type.
1708 In LLVM, opaque types can eventually be resolved to any type (not just a
1709 structure type).</p>
1719 <table class="layout">
1721 <td class="left"><tt>opaque</tt></td>
1722 <td class="left">An opaque type.</td>
1727 <!-- ======================================================================= -->
1728 <div class="doc_subsection">
1729 <a name="t_uprefs">Type Up-references</a>
1732 <div class="doc_text">
1735 An "up reference" allows you to refer to a lexically enclosing type without
1736 requiring it to have a name. For instance, a structure declaration may contain a
1737 pointer to any of the types it is lexically a member of. Example of up
1738 references (with their equivalent as named type declarations) include:</p>
1741 { \2 * } %x = type { %x* }
1742 { \2 }* %y = type { %y }*
1747 An up reference is needed by the asmprinter for printing out cyclic types when
1748 there is no declared name for a type in the cycle. Because the asmprinter does
1749 not want to print out an infinite type string, it needs a syntax to handle
1750 recursive types that have no names (all names are optional in llvm IR).
1759 The level is the count of the lexical type that is being referred to.
1764 <table class="layout">
1766 <td class="left"><tt>\1*</tt></td>
1767 <td class="left">Self-referential pointer.</td>
1770 <td class="left"><tt>{ { \3*, i8 }, i32 }</tt></td>
1771 <td class="left">Recursive structure where the upref refers to the out-most
1778 <!-- *********************************************************************** -->
1779 <div class="doc_section"> <a name="constants">Constants</a> </div>
1780 <!-- *********************************************************************** -->
1782 <div class="doc_text">
1784 <p>LLVM has several different basic types of constants. This section describes
1785 them all and their syntax.</p>
1789 <!-- ======================================================================= -->
1790 <div class="doc_subsection"><a name="simpleconstants">Simple Constants</a></div>
1792 <div class="doc_text">
1795 <dt><b>Boolean constants</b></dt>
1797 <dd>The two strings '<tt>true</tt>' and '<tt>false</tt>' are both valid
1798 constants of the <tt><a href="#t_primitive">i1</a></tt> type.
1801 <dt><b>Integer constants</b></dt>
1803 <dd>Standard integers (such as '4') are constants of the <a
1804 href="#t_integer">integer</a> type. Negative numbers may be used with
1808 <dt><b>Floating point constants</b></dt>
1810 <dd>Floating point constants use standard decimal notation (e.g. 123.421),
1811 exponential notation (e.g. 1.23421e+2), or a more precise hexadecimal
1812 notation (see below). The assembler requires the exact decimal value of
1813 a floating-point constant. For example, the assembler accepts 1.25 but
1814 rejects 1.3 because 1.3 is a repeating decimal in binary. Floating point
1815 constants must have a <a href="#t_floating">floating point</a> type. </dd>
1817 <dt><b>Null pointer constants</b></dt>
1819 <dd>The identifier '<tt>null</tt>' is recognized as a null pointer constant
1820 and must be of <a href="#t_pointer">pointer type</a>.</dd>
1824 <p>The one non-intuitive notation for constants is the hexadecimal form
1825 of floating point constants. For example, the form '<tt>double
1826 0x432ff973cafa8000</tt>' is equivalent to (but harder to read than) '<tt>double
1827 4.5e+15</tt>'. The only time hexadecimal floating point constants are required
1828 (and the only time that they are generated by the disassembler) is when a
1829 floating point constant must be emitted but it cannot be represented as a
1830 decimal floating point number in a reasonable number of digits. For example,
1831 NaN's, infinities, and other
1832 special values are represented in their IEEE hexadecimal format so that
1833 assembly and disassembly do not cause any bits to change in the constants.</p>
1834 <p>When using the hexadecimal form, constants of types float and double are
1835 represented using the 16-digit form shown above (which matches the IEEE754
1836 representation for double); float values must, however, be exactly representable
1837 as IEE754 single precision.
1838 Hexadecimal format is always used for long
1839 double, and there are three forms of long double. The 80-bit
1840 format used by x86 is represented as <tt>0xK</tt>
1841 followed by 20 hexadecimal digits.
1842 The 128-bit format used by PowerPC (two adjacent doubles) is represented
1843 by <tt>0xM</tt> followed by 32 hexadecimal digits. The IEEE 128-bit
1844 format is represented
1845 by <tt>0xL</tt> followed by 32 hexadecimal digits; no currently supported
1846 target uses this format. Long doubles will only work if they match
1847 the long double format on your target. All hexadecimal formats are big-endian
1848 (sign bit at the left).</p>
1851 <!-- ======================================================================= -->
1852 <div class="doc_subsection">
1853 <a name="aggregateconstants"> <!-- old anchor -->
1854 <a name="complexconstants">Complex Constants</a></a>
1857 <div class="doc_text">
1858 <p>Complex constants are a (potentially recursive) combination of simple
1859 constants and smaller complex constants.</p>
1862 <dt><b>Structure constants</b></dt>
1864 <dd>Structure constants are represented with notation similar to structure
1865 type definitions (a comma separated list of elements, surrounded by braces
1866 (<tt>{}</tt>)). For example: "<tt>{ i32 4, float 17.0, i32* @G }</tt>",
1867 where "<tt>@G</tt>" is declared as "<tt>@G = external global i32</tt>". Structure constants
1868 must have <a href="#t_struct">structure type</a>, and the number and
1869 types of elements must match those specified by the type.
1872 <dt><b>Array constants</b></dt>
1874 <dd>Array constants are represented with notation similar to array type
1875 definitions (a comma separated list of elements, surrounded by square brackets
1876 (<tt>[]</tt>)). For example: "<tt>[ i32 42, i32 11, i32 74 ]</tt>". Array
1877 constants must have <a href="#t_array">array type</a>, and the number and
1878 types of elements must match those specified by the type.
1881 <dt><b>Vector constants</b></dt>
1883 <dd>Vector constants are represented with notation similar to vector type
1884 definitions (a comma separated list of elements, surrounded by
1885 less-than/greater-than's (<tt><></tt>)). For example: "<tt>< i32 42,
1886 i32 11, i32 74, i32 100 ></tt>". Vector constants must have <a
1887 href="#t_vector">vector type</a>, and the number and types of elements must
1888 match those specified by the type.
1891 <dt><b>Zero initialization</b></dt>
1893 <dd>The string '<tt>zeroinitializer</tt>' can be used to zero initialize a
1894 value to zero of <em>any</em> type, including scalar and aggregate types.
1895 This is often used to avoid having to print large zero initializers (e.g. for
1896 large arrays) and is always exactly equivalent to using explicit zero
1900 <dt><b>Metadata node</b></dt>
1902 <dd>A metadata node is a structure-like constant with
1903 <a href="#t_metadata">metadata type</a>. For example:
1904 "<tt>metadata !{ i32 0, metadata !"test" }</tt>". Unlike other constants
1905 that are meant to be interpreted as part of the instruction stream, metadata
1906 is a place to attach additional information such as debug info.
1912 <!-- ======================================================================= -->
1913 <div class="doc_subsection">
1914 <a name="globalconstants">Global Variable and Function Addresses</a>
1917 <div class="doc_text">
1919 <p>The addresses of <a href="#globalvars">global variables</a> and <a
1920 href="#functionstructure">functions</a> are always implicitly valid (link-time)
1921 constants. These constants are explicitly referenced when the <a
1922 href="#identifiers">identifier for the global</a> is used and always have <a
1923 href="#t_pointer">pointer</a> type. For example, the following is a legal LLVM
1926 <div class="doc_code">
1930 @Z = global [2 x i32*] [ i32* @X, i32* @Y ]
1936 <!-- ======================================================================= -->
1937 <div class="doc_subsection"><a name="undefvalues">Undefined Values</a></div>
1938 <div class="doc_text">
1939 <p>The string '<tt>undef</tt>' is recognized as a type-less constant that has
1940 no specific value. Undefined values may be of any type and be used anywhere
1941 a constant is permitted.</p>
1943 <p>Undefined values indicate to the compiler that the program is well defined
1944 no matter what value is used, giving the compiler more freedom to optimize.
1948 <!-- ======================================================================= -->
1949 <div class="doc_subsection"><a name="constantexprs">Constant Expressions</a>
1952 <div class="doc_text">
1954 <p>Constant expressions are used to allow expressions involving other constants
1955 to be used as constants. Constant expressions may be of any <a
1956 href="#t_firstclass">first class</a> type and may involve any LLVM operation
1957 that does not have side effects (e.g. load and call are not supported). The
1958 following is the syntax for constant expressions:</p>
1961 <dt><b><tt>trunc ( CST to TYPE )</tt></b></dt>
1962 <dd>Truncate a constant to another type. The bit size of CST must be larger
1963 than the bit size of TYPE. Both types must be integers.</dd>
1965 <dt><b><tt>zext ( CST to TYPE )</tt></b></dt>
1966 <dd>Zero extend a constant to another type. The bit size of CST must be
1967 smaller or equal to the bit size of TYPE. Both types must be integers.</dd>
1969 <dt><b><tt>sext ( CST to TYPE )</tt></b></dt>
1970 <dd>Sign extend a constant to another type. The bit size of CST must be
1971 smaller or equal to the bit size of TYPE. Both types must be integers.</dd>
1973 <dt><b><tt>fptrunc ( CST to TYPE )</tt></b></dt>
1974 <dd>Truncate a floating point constant to another floating point type. The
1975 size of CST must be larger than the size of TYPE. Both types must be
1976 floating point.</dd>
1978 <dt><b><tt>fpext ( CST to TYPE )</tt></b></dt>
1979 <dd>Floating point extend a constant to another type. The size of CST must be
1980 smaller or equal to the size of TYPE. Both types must be floating point.</dd>
1982 <dt><b><tt>fptoui ( CST to TYPE )</tt></b></dt>
1983 <dd>Convert a floating point constant to the corresponding unsigned integer
1984 constant. TYPE must be a scalar or vector integer type. CST must be of scalar
1985 or vector floating point type. Both CST and TYPE must be scalars, or vectors
1986 of the same number of elements. If the value won't fit in the integer type,
1987 the results are undefined.</dd>
1989 <dt><b><tt>fptosi ( CST to TYPE )</tt></b></dt>
1990 <dd>Convert a floating point constant to the corresponding signed integer
1991 constant. TYPE must be a scalar or vector integer type. CST must be of scalar
1992 or vector floating point type. Both CST and TYPE must be scalars, or vectors
1993 of the same number of elements. If the value won't fit in the integer type,
1994 the results are undefined.</dd>
1996 <dt><b><tt>uitofp ( CST to TYPE )</tt></b></dt>
1997 <dd>Convert an unsigned integer constant to the corresponding floating point
1998 constant. TYPE must be a scalar or vector floating point type. CST must be of
1999 scalar or vector integer type. Both CST and TYPE must be scalars, or vectors
2000 of the same number of elements. If the value won't fit in the floating point
2001 type, the results are undefined.</dd>
2003 <dt><b><tt>sitofp ( CST to TYPE )</tt></b></dt>
2004 <dd>Convert a signed integer constant to the corresponding floating point
2005 constant. TYPE must be a scalar or vector floating point type. CST must be of
2006 scalar or vector integer type. Both CST and TYPE must be scalars, or vectors
2007 of the same number of elements. If the value won't fit in the floating point
2008 type, the results are undefined.</dd>
2010 <dt><b><tt>ptrtoint ( CST to TYPE )</tt></b></dt>
2011 <dd>Convert a pointer typed constant to the corresponding integer constant
2012 TYPE must be an integer type. CST must be of pointer type. The CST value is
2013 zero extended, truncated, or unchanged to make it fit in TYPE.</dd>
2015 <dt><b><tt>inttoptr ( CST to TYPE )</tt></b></dt>
2016 <dd>Convert a integer constant to a pointer constant. TYPE must be a
2017 pointer type. CST must be of integer type. The CST value is zero extended,
2018 truncated, or unchanged to make it fit in a pointer size. This one is
2019 <i>really</i> dangerous!</dd>
2021 <dt><b><tt>bitcast ( CST to TYPE )</tt></b></dt>
2022 <dd>Convert a constant, CST, to another TYPE. The constraints of the operands
2023 are the same as those for the <a href="#i_bitcast">bitcast
2024 instruction</a>.</dd>
2026 <dt><b><tt>getelementptr ( CSTPTR, IDX0, IDX1, ... )</tt></b></dt>
2028 <dd>Perform the <a href="#i_getelementptr">getelementptr operation</a> on
2029 constants. As with the <a href="#i_getelementptr">getelementptr</a>
2030 instruction, the index list may have zero or more indexes, which are required
2031 to make sense for the type of "CSTPTR".</dd>
2033 <dt><b><tt>select ( COND, VAL1, VAL2 )</tt></b></dt>
2035 <dd>Perform the <a href="#i_select">select operation</a> on
2038 <dt><b><tt>icmp COND ( VAL1, VAL2 )</tt></b></dt>
2039 <dd>Performs the <a href="#i_icmp">icmp operation</a> on constants.</dd>
2041 <dt><b><tt>fcmp COND ( VAL1, VAL2 )</tt></b></dt>
2042 <dd>Performs the <a href="#i_fcmp">fcmp operation</a> on constants.</dd>
2044 <dt><b><tt>extractelement ( VAL, IDX )</tt></b></dt>
2046 <dd>Perform the <a href="#i_extractelement">extractelement
2047 operation</a> on constants.</dd>
2049 <dt><b><tt>insertelement ( VAL, ELT, IDX )</tt></b></dt>
2051 <dd>Perform the <a href="#i_insertelement">insertelement
2052 operation</a> on constants.</dd>
2055 <dt><b><tt>shufflevector ( VEC1, VEC2, IDXMASK )</tt></b></dt>
2057 <dd>Perform the <a href="#i_shufflevector">shufflevector
2058 operation</a> on constants.</dd>
2060 <dt><b><tt>OPCODE ( LHS, RHS )</tt></b></dt>
2062 <dd>Perform the specified operation of the LHS and RHS constants. OPCODE may
2063 be any of the <a href="#binaryops">binary</a> or <a href="#bitwiseops">bitwise
2064 binary</a> operations. The constraints on operands are the same as those for
2065 the corresponding instruction (e.g. no bitwise operations on floating point
2066 values are allowed).</dd>
2070 <!-- ======================================================================= -->
2071 <div class="doc_subsection"><a name="metadata">Embedded Metadata</a>
2074 <div class="doc_text">
2076 <p>Embedded metadata provides a way to attach arbitrary data to the
2077 instruction stream without affecting the behaviour of the program. There are
2078 two metadata primitives, strings and nodes. All metadata has the
2079 <tt>metadata</tt> type and is identified in syntax by a preceding exclamation
2080 point ('<tt>!</tt>').
2083 <p>A metadata string is a string surrounded by double quotes. It can contain
2084 any character by escaping non-printable characters with "\xx" where "xx" is
2085 the two digit hex code. For example: "<tt>!"test\00"</tt>".
2088 <p>Metadata nodes are represented with notation similar to structure constants
2089 (a comma separated list of elements, surrounded by braces and preceeded by an
2090 exclamation point). For example: "<tt>!{ metadata !"test\00", i32 10}</tt>".
2093 <p>A metadata node will attempt to track changes to the values it holds. In
2094 the event that a value is deleted, it will be replaced with a typeless
2095 "<tt>null</tt>", such as "<tt>metadata !{null, i32 10}</tt>".</p>
2097 <p>Optimizations may rely on metadata to provide additional information about
2098 the program that isn't available in the instructions, or that isn't easily
2099 computable. Similarly, the code generator may expect a certain metadata format
2100 to be used to express debugging information.</p>
2103 <!-- *********************************************************************** -->
2104 <div class="doc_section"> <a name="othervalues">Other Values</a> </div>
2105 <!-- *********************************************************************** -->
2107 <!-- ======================================================================= -->
2108 <div class="doc_subsection">
2109 <a name="inlineasm">Inline Assembler Expressions</a>
2112 <div class="doc_text">
2115 LLVM supports inline assembler expressions (as opposed to <a href="#moduleasm">
2116 Module-Level Inline Assembly</a>) through the use of a special value. This
2117 value represents the inline assembler as a string (containing the instructions
2118 to emit), a list of operand constraints (stored as a string), and a flag that
2119 indicates whether or not the inline asm expression has side effects. An example
2120 inline assembler expression is:
2123 <div class="doc_code">
2125 i32 (i32) asm "bswap $0", "=r,r"
2130 Inline assembler expressions may <b>only</b> be used as the callee operand of
2131 a <a href="#i_call"><tt>call</tt> instruction</a>. Thus, typically we have:
2134 <div class="doc_code">
2136 %X = call i32 asm "<a href="#int_bswap">bswap</a> $0", "=r,r"(i32 %Y)
2141 Inline asms with side effects not visible in the constraint list must be marked
2142 as having side effects. This is done through the use of the
2143 '<tt>sideeffect</tt>' keyword, like so:
2146 <div class="doc_code">
2148 call void asm sideeffect "eieio", ""()
2152 <p>TODO: The format of the asm and constraints string still need to be
2153 documented here. Constraints on what can be done (e.g. duplication, moving, etc
2154 need to be documented). This is probably best done by reference to another
2155 document that covers inline asm from a holistic perspective.
2160 <!-- *********************************************************************** -->
2161 <div class="doc_section"> <a name="instref">Instruction Reference</a> </div>
2162 <!-- *********************************************************************** -->
2164 <div class="doc_text">
2166 <p>The LLVM instruction set consists of several different
2167 classifications of instructions: <a href="#terminators">terminator
2168 instructions</a>, <a href="#binaryops">binary instructions</a>,
2169 <a href="#bitwiseops">bitwise binary instructions</a>, <a
2170 href="#memoryops">memory instructions</a>, and <a href="#otherops">other
2171 instructions</a>.</p>
2175 <!-- ======================================================================= -->
2176 <div class="doc_subsection"> <a name="terminators">Terminator
2177 Instructions</a> </div>
2179 <div class="doc_text">
2181 <p>As mentioned <a href="#functionstructure">previously</a>, every
2182 basic block in a program ends with a "Terminator" instruction, which
2183 indicates which block should be executed after the current block is
2184 finished. These terminator instructions typically yield a '<tt>void</tt>'
2185 value: they produce control flow, not values (the one exception being
2186 the '<a href="#i_invoke"><tt>invoke</tt></a>' instruction).</p>
2187 <p>There are six different terminator instructions: the '<a
2188 href="#i_ret"><tt>ret</tt></a>' instruction, the '<a href="#i_br"><tt>br</tt></a>'
2189 instruction, the '<a href="#i_switch"><tt>switch</tt></a>' instruction,
2190 the '<a href="#i_invoke"><tt>invoke</tt></a>' instruction, the '<a
2191 href="#i_unwind"><tt>unwind</tt></a>' instruction, and the '<a
2192 href="#i_unreachable"><tt>unreachable</tt></a>' instruction.</p>
2196 <!-- _______________________________________________________________________ -->
2197 <div class="doc_subsubsection"> <a name="i_ret">'<tt>ret</tt>'
2198 Instruction</a> </div>
2199 <div class="doc_text">
2202 ret <type> <value> <i>; Return a value from a non-void function</i>
2203 ret void <i>; Return from void function</i>
2208 <p>The '<tt>ret</tt>' instruction is used to return control flow (and
2209 optionally a value) from a function back to the caller.</p>
2210 <p>There are two forms of the '<tt>ret</tt>' instruction: one that
2211 returns a value and then causes control flow, and one that just causes
2212 control flow to occur.</p>
2216 <p>The '<tt>ret</tt>' instruction optionally accepts a single argument,
2217 the return value. The type of the return value must be a
2218 '<a href="#t_firstclass">first class</a>' type.</p>
2220 <p>A function is not <a href="#wellformed">well formed</a> if
2221 it it has a non-void return type and contains a '<tt>ret</tt>'
2222 instruction with no return value or a return value with a type that
2223 does not match its type, or if it has a void return type and contains
2224 a '<tt>ret</tt>' instruction with a return value.</p>
2228 <p>When the '<tt>ret</tt>' instruction is executed, control flow
2229 returns back to the calling function's context. If the caller is a "<a
2230 href="#i_call"><tt>call</tt></a>" instruction, execution continues at
2231 the instruction after the call. If the caller was an "<a
2232 href="#i_invoke"><tt>invoke</tt></a>" instruction, execution continues
2233 at the beginning of the "normal" destination block. If the instruction
2234 returns a value, that value shall set the call or invoke instruction's
2240 ret i32 5 <i>; Return an integer value of 5</i>
2241 ret void <i>; Return from a void function</i>
2242 ret { i32, i8 } { i32 4, i8 2 } <i>; Return a struct of values 4 and 2</i>
2245 <p>Note that the code generator does not yet fully support large
2246 return values. The specific sizes that are currently supported are
2247 dependent on the target. For integers, on 32-bit targets the limit
2248 is often 64 bits, and on 64-bit targets the limit is often 128 bits.
2249 For aggregate types, the current limits are dependent on the element
2250 types; for example targets are often limited to 2 total integer
2251 elements and 2 total floating-point elements.</p>
2254 <!-- _______________________________________________________________________ -->
2255 <div class="doc_subsubsection"> <a name="i_br">'<tt>br</tt>' Instruction</a> </div>
2256 <div class="doc_text">
2258 <pre> br i1 <cond>, label <iftrue>, label <iffalse><br> br label <dest> <i>; Unconditional branch</i>
2261 <p>The '<tt>br</tt>' instruction is used to cause control flow to
2262 transfer to a different basic block in the current function. There are
2263 two forms of this instruction, corresponding to a conditional branch
2264 and an unconditional branch.</p>
2266 <p>The conditional branch form of the '<tt>br</tt>' instruction takes a
2267 single '<tt>i1</tt>' value and two '<tt>label</tt>' values. The
2268 unconditional form of the '<tt>br</tt>' instruction takes a single
2269 '<tt>label</tt>' value as a target.</p>
2271 <p>Upon execution of a conditional '<tt>br</tt>' instruction, the '<tt>i1</tt>'
2272 argument is evaluated. If the value is <tt>true</tt>, control flows
2273 to the '<tt>iftrue</tt>' <tt>label</tt> argument. If "cond" is <tt>false</tt>,
2274 control flows to the '<tt>iffalse</tt>' <tt>label</tt> argument.</p>
2276 <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
2277 href="#i_ret">ret</a> i32 1<br>IfUnequal:<br> <a href="#i_ret">ret</a> i32 0<br></pre>
2279 <!-- _______________________________________________________________________ -->
2280 <div class="doc_subsubsection">
2281 <a name="i_switch">'<tt>switch</tt>' Instruction</a>
2284 <div class="doc_text">
2288 switch <intty> <value>, label <defaultdest> [ <intty> <val>, label <dest> ... ]
2293 <p>The '<tt>switch</tt>' instruction is used to transfer control flow to one of
2294 several different places. It is a generalization of the '<tt>br</tt>'
2295 instruction, allowing a branch to occur to one of many possible
2301 <p>The '<tt>switch</tt>' instruction uses three parameters: an integer
2302 comparison value '<tt>value</tt>', a default '<tt>label</tt>' destination, and
2303 an array of pairs of comparison value constants and '<tt>label</tt>'s. The
2304 table is not allowed to contain duplicate constant entries.</p>
2308 <p>The <tt>switch</tt> instruction specifies a table of values and
2309 destinations. When the '<tt>switch</tt>' instruction is executed, this
2310 table is searched for the given value. If the value is found, control flow is
2311 transfered to the corresponding destination; otherwise, control flow is
2312 transfered to the default destination.</p>
2314 <h5>Implementation:</h5>
2316 <p>Depending on properties of the target machine and the particular
2317 <tt>switch</tt> instruction, this instruction may be code generated in different
2318 ways. For example, it could be generated as a series of chained conditional
2319 branches or with a lookup table.</p>
2324 <i>; Emulate a conditional br instruction</i>
2325 %Val = <a href="#i_zext">zext</a> i1 %value to i32
2326 switch i32 %Val, label %truedest [ i32 0, label %falsedest ]
2328 <i>; Emulate an unconditional br instruction</i>
2329 switch i32 0, label %dest [ ]
2331 <i>; Implement a jump table:</i>
2332 switch i32 %val, label %otherwise [ i32 0, label %onzero
2334 i32 2, label %ontwo ]
2338 <!-- _______________________________________________________________________ -->
2339 <div class="doc_subsubsection">
2340 <a name="i_invoke">'<tt>invoke</tt>' Instruction</a>
2343 <div class="doc_text">
2348 <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>]
2349 to label <normal label> unwind label <exception label>
2354 <p>The '<tt>invoke</tt>' instruction causes control to transfer to a specified
2355 function, with the possibility of control flow transfer to either the
2356 '<tt>normal</tt>' label or the
2357 '<tt>exception</tt>' label. If the callee function returns with the
2358 "<tt><a href="#i_ret">ret</a></tt>" instruction, control flow will return to the
2359 "normal" label. If the callee (or any indirect callees) returns with the "<a
2360 href="#i_unwind"><tt>unwind</tt></a>" instruction, control is interrupted and
2361 continued at the dynamically nearest "exception" label.</p>
2365 <p>This instruction requires several arguments:</p>
2369 The optional "cconv" marker indicates which <a href="#callingconv">calling
2370 convention</a> the call should use. If none is specified, the call defaults
2371 to using C calling conventions.
2374 <li>The optional <a href="#paramattrs">Parameter Attributes</a> list for
2375 return values. Only '<tt>zeroext</tt>', '<tt>signext</tt>',
2376 and '<tt>inreg</tt>' attributes are valid here.</li>
2378 <li>'<tt>ptr to function ty</tt>': shall be the signature of the pointer to
2379 function value being invoked. In most cases, this is a direct function
2380 invocation, but indirect <tt>invoke</tt>s are just as possible, branching off
2381 an arbitrary pointer to function value.
2384 <li>'<tt>function ptr val</tt>': An LLVM value containing a pointer to a
2385 function to be invoked. </li>
2387 <li>'<tt>function args</tt>': argument list whose types match the function
2388 signature argument types. If the function signature indicates the function
2389 accepts a variable number of arguments, the extra arguments can be
2392 <li>'<tt>normal label</tt>': the label reached when the called function
2393 executes a '<tt><a href="#i_ret">ret</a></tt>' instruction. </li>
2395 <li>'<tt>exception label</tt>': the label reached when a callee returns with
2396 the <a href="#i_unwind"><tt>unwind</tt></a> instruction. </li>
2398 <li>The optional <a href="#fnattrs">function attributes</a> list. Only
2399 '<tt>noreturn</tt>', '<tt>nounwind</tt>', '<tt>readonly</tt>' and
2400 '<tt>readnone</tt>' attributes are valid here.</li>
2405 <p>This instruction is designed to operate as a standard '<tt><a
2406 href="#i_call">call</a></tt>' instruction in most regards. The primary
2407 difference is that it establishes an association with a label, which is used by
2408 the runtime library to unwind the stack.</p>
2410 <p>This instruction is used in languages with destructors to ensure that proper
2411 cleanup is performed in the case of either a <tt>longjmp</tt> or a thrown
2412 exception. Additionally, this is important for implementation of
2413 '<tt>catch</tt>' clauses in high-level languages that support them.</p>
2415 <p>For the purposes of the SSA form, the definition of the value
2416 returned by the '<tt>invoke</tt>' instruction is deemed to occur on
2417 the edge from the current block to the "normal" label. If the callee
2418 unwinds then no return value is available.</p>
2422 %retval = invoke i32 @Test(i32 15) to label %Continue
2423 unwind label %TestCleanup <i>; {i32}:retval set</i>
2424 %retval = invoke <a href="#callingconv">coldcc</a> i32 %Testfnptr(i32 15) to label %Continue
2425 unwind label %TestCleanup <i>; {i32}:retval set</i>
2430 <!-- _______________________________________________________________________ -->
2432 <div class="doc_subsubsection"> <a name="i_unwind">'<tt>unwind</tt>'
2433 Instruction</a> </div>
2435 <div class="doc_text">
2444 <p>The '<tt>unwind</tt>' instruction unwinds the stack, continuing control flow
2445 at the first callee in the dynamic call stack which used an <a
2446 href="#i_invoke"><tt>invoke</tt></a> instruction to perform the call. This is
2447 primarily used to implement exception handling.</p>
2451 <p>The '<tt>unwind</tt>' instruction causes execution of the current function to
2452 immediately halt. The dynamic call stack is then searched for the first <a
2453 href="#i_invoke"><tt>invoke</tt></a> instruction on the call stack. Once found,
2454 execution continues at the "exceptional" destination block specified by the
2455 <tt>invoke</tt> instruction. If there is no <tt>invoke</tt> instruction in the
2456 dynamic call chain, undefined behavior results.</p>
2459 <!-- _______________________________________________________________________ -->
2461 <div class="doc_subsubsection"> <a name="i_unreachable">'<tt>unreachable</tt>'
2462 Instruction</a> </div>
2464 <div class="doc_text">
2473 <p>The '<tt>unreachable</tt>' instruction has no defined semantics. This
2474 instruction is used to inform the optimizer that a particular portion of the
2475 code is not reachable. This can be used to indicate that the code after a
2476 no-return function cannot be reached, and other facts.</p>
2480 <p>The '<tt>unreachable</tt>' instruction has no defined semantics.</p>
2485 <!-- ======================================================================= -->
2486 <div class="doc_subsection"> <a name="binaryops">Binary Operations</a> </div>
2487 <div class="doc_text">
2488 <p>Binary operators are used to do most of the computation in a
2489 program. They require two operands of the same type, execute an operation on them, and
2490 produce a single value. The operands might represent
2491 multiple data, as is the case with the <a href="#t_vector">vector</a> data type.
2492 The result value has the same type as its operands.</p>
2493 <p>There are several different binary operators:</p>
2495 <!-- _______________________________________________________________________ -->
2496 <div class="doc_subsubsection">
2497 <a name="i_add">'<tt>add</tt>' Instruction</a>
2500 <div class="doc_text">
2505 <result> = add <ty> <op1>, <op2> <i>; yields {ty}:result</i>
2510 <p>The '<tt>add</tt>' instruction returns the sum of its two operands.</p>
2514 <p>The two arguments to the '<tt>add</tt>' instruction must be <a
2515 href="#t_integer">integer</a> or
2516 <a href="#t_vector">vector</a> of integer values. Both arguments must
2517 have identical types.</p>
2521 <p>The value produced is the integer sum of the two operands.</p>
2523 <p>If the sum has unsigned overflow, the result returned is the
2524 mathematical result modulo 2<sup>n</sup>, where n is the bit width of
2527 <p>Because LLVM integers use a two's complement representation, this
2528 instruction is appropriate for both signed and unsigned integers.</p>
2533 <result> = add i32 4, %var <i>; yields {i32}:result = 4 + %var</i>
2536 <!-- _______________________________________________________________________ -->
2537 <div class="doc_subsubsection">
2538 <a name="i_fadd">'<tt>fadd</tt>' Instruction</a>
2541 <div class="doc_text">
2546 <result> = fadd <ty> <op1>, <op2> <i>; yields {ty}:result</i>
2551 <p>The '<tt>fadd</tt>' instruction returns the sum of its two operands.</p>
2555 <p>The two arguments to the '<tt>fadd</tt>' instruction must be
2556 <a href="#t_floating">floating point</a> or <a href="#t_vector">vector</a> of
2557 floating point values. Both arguments must have identical types.</p>
2561 <p>The value produced is the floating point sum of the two operands.</p>
2566 <result> = fadd float 4.0, %var <i>; yields {float}:result = 4.0 + %var</i>
2569 <!-- _______________________________________________________________________ -->
2570 <div class="doc_subsubsection">
2571 <a name="i_sub">'<tt>sub</tt>' Instruction</a>
2574 <div class="doc_text">
2579 <result> = sub <ty> <op1>, <op2> <i>; yields {ty}:result</i>
2584 <p>The '<tt>sub</tt>' instruction returns the difference of its two
2587 <p>Note that the '<tt>sub</tt>' instruction is used to represent the
2588 '<tt>neg</tt>' instruction present in most other intermediate
2589 representations.</p>
2593 <p>The two arguments to the '<tt>sub</tt>' instruction must be <a
2594 href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of
2595 integer values. Both arguments must have identical types.</p>
2599 <p>The value produced is the integer difference of the two operands.</p>
2601 <p>If the difference has unsigned overflow, the result returned is the
2602 mathematical result modulo 2<sup>n</sup>, where n is the bit width of
2605 <p>Because LLVM integers use a two's complement representation, this
2606 instruction is appropriate for both signed and unsigned integers.</p>
2610 <result> = sub i32 4, %var <i>; yields {i32}:result = 4 - %var</i>
2611 <result> = sub i32 0, %val <i>; yields {i32}:result = -%var</i>
2615 <!-- _______________________________________________________________________ -->
2616 <div class="doc_subsubsection">
2617 <a name="i_fsub">'<tt>fsub</tt>' Instruction</a>
2620 <div class="doc_text">
2625 <result> = fsub <ty> <op1>, <op2> <i>; yields {ty}:result</i>
2630 <p>The '<tt>fsub</tt>' instruction returns the difference of its two
2633 <p>Note that the '<tt>fsub</tt>' instruction is used to represent the
2634 '<tt>fneg</tt>' instruction present in most other intermediate
2635 representations.</p>
2639 <p>The two arguments to the '<tt>fsub</tt>' instruction must be <a
2640 <a href="#t_floating">floating point</a> or <a href="#t_vector">vector</a>
2641 of floating point values. Both arguments must have identical types.</p>
2645 <p>The value produced is the floating point difference of the two operands.</p>
2649 <result> = fsub float 4.0, %var <i>; yields {float}:result = 4.0 - %var</i>
2650 <result> = fsub float -0.0, %val <i>; yields {float}:result = -%var</i>
2654 <!-- _______________________________________________________________________ -->
2655 <div class="doc_subsubsection">
2656 <a name="i_mul">'<tt>mul</tt>' Instruction</a>
2659 <div class="doc_text">
2662 <pre> <result> = mul <ty> <op1>, <op2> <i>; yields {ty}:result</i>
2665 <p>The '<tt>mul</tt>' instruction returns the product of its two
2670 <p>The two arguments to the '<tt>mul</tt>' instruction must be <a
2671 href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
2672 values. Both arguments must have identical types.</p>
2676 <p>The value produced is the integer product of the two operands.</p>
2678 <p>If the result of the multiplication has unsigned overflow,
2679 the result returned is the mathematical result modulo
2680 2<sup>n</sup>, where n is the bit width of the result.</p>
2681 <p>Because LLVM integers use a two's complement representation, and the
2682 result is the same width as the operands, this instruction returns the
2683 correct result for both signed and unsigned integers. If a full product
2684 (e.g. <tt>i32</tt>x<tt>i32</tt>-><tt>i64</tt>) is needed, the operands
2685 should be sign-extended or zero-extended as appropriate to the
2686 width of the full product.</p>
2688 <pre> <result> = mul i32 4, %var <i>; yields {i32}:result = 4 * %var</i>
2692 <!-- _______________________________________________________________________ -->
2693 <div class="doc_subsubsection">
2694 <a name="i_fmul">'<tt>fmul</tt>' Instruction</a>
2697 <div class="doc_text">
2700 <pre> <result> = fmul <ty> <op1>, <op2> <i>; yields {ty}:result</i>
2703 <p>The '<tt>fmul</tt>' instruction returns the product of its two
2708 <p>The two arguments to the '<tt>fmul</tt>' instruction must be
2709 <a href="#t_floating">floating point</a> or <a href="#t_vector">vector</a>
2710 of floating point values. Both arguments must have identical types.</p>
2714 <p>The value produced is the floating point product of the two operands.</p>
2717 <pre> <result> = fmul float 4.0, %var <i>; yields {float}:result = 4.0 * %var</i>
2721 <!-- _______________________________________________________________________ -->
2722 <div class="doc_subsubsection"> <a name="i_udiv">'<tt>udiv</tt>' Instruction
2724 <div class="doc_text">
2726 <pre> <result> = udiv <ty> <op1>, <op2> <i>; yields {ty}:result</i>
2729 <p>The '<tt>udiv</tt>' instruction returns the quotient of its two
2734 <p>The two arguments to the '<tt>udiv</tt>' instruction must be
2735 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
2736 values. Both arguments must have identical types.</p>
2740 <p>The value produced is the unsigned integer quotient of the two operands.</p>
2741 <p>Note that unsigned integer division and signed integer division are distinct
2742 operations; for signed integer division, use '<tt>sdiv</tt>'.</p>
2743 <p>Division by zero leads to undefined behavior.</p>
2745 <pre> <result> = udiv i32 4, %var <i>; yields {i32}:result = 4 / %var</i>
2748 <!-- _______________________________________________________________________ -->
2749 <div class="doc_subsubsection"> <a name="i_sdiv">'<tt>sdiv</tt>' Instruction
2751 <div class="doc_text">
2754 <result> = sdiv <ty> <op1>, <op2> <i>; yields {ty}:result</i>
2759 <p>The '<tt>sdiv</tt>' instruction returns the quotient of its two
2764 <p>The two arguments to the '<tt>sdiv</tt>' instruction must be
2765 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
2766 values. Both arguments must have identical types.</p>
2769 <p>The value produced is the signed integer quotient of the two operands rounded towards zero.</p>
2770 <p>Note that signed integer division and unsigned integer division are distinct
2771 operations; for unsigned integer division, use '<tt>udiv</tt>'.</p>
2772 <p>Division by zero leads to undefined behavior. Overflow also leads to
2773 undefined behavior; this is a rare case, but can occur, for example,
2774 by doing a 32-bit division of -2147483648 by -1.</p>
2776 <pre> <result> = sdiv i32 4, %var <i>; yields {i32}:result = 4 / %var</i>
2779 <!-- _______________________________________________________________________ -->
2780 <div class="doc_subsubsection"> <a name="i_fdiv">'<tt>fdiv</tt>'
2781 Instruction</a> </div>
2782 <div class="doc_text">
2785 <result> = fdiv <ty> <op1>, <op2> <i>; yields {ty}:result</i>
2789 <p>The '<tt>fdiv</tt>' instruction returns the quotient of its two
2794 <p>The two arguments to the '<tt>fdiv</tt>' instruction must be
2795 <a href="#t_floating">floating point</a> or <a href="#t_vector">vector</a>
2796 of floating point values. Both arguments must have identical types.</p>
2800 <p>The value produced is the floating point quotient of the two operands.</p>
2805 <result> = fdiv float 4.0, %var <i>; yields {float}:result = 4.0 / %var</i>
2809 <!-- _______________________________________________________________________ -->
2810 <div class="doc_subsubsection"> <a name="i_urem">'<tt>urem</tt>' Instruction</a>
2812 <div class="doc_text">
2814 <pre> <result> = urem <ty> <op1>, <op2> <i>; yields {ty}:result</i>
2817 <p>The '<tt>urem</tt>' instruction returns the remainder from the
2818 unsigned division of its two arguments.</p>
2820 <p>The two arguments to the '<tt>urem</tt>' instruction must be
2821 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
2822 values. Both arguments must have identical types.</p>
2824 <p>This instruction returns the unsigned integer <i>remainder</i> of a division.
2825 This instruction always performs an unsigned division to get the remainder.</p>
2826 <p>Note that unsigned integer remainder and signed integer remainder are
2827 distinct operations; for signed integer remainder, use '<tt>srem</tt>'.</p>
2828 <p>Taking the remainder of a division by zero leads to undefined behavior.</p>
2830 <pre> <result> = urem i32 4, %var <i>; yields {i32}:result = 4 % %var</i>
2834 <!-- _______________________________________________________________________ -->
2835 <div class="doc_subsubsection">
2836 <a name="i_srem">'<tt>srem</tt>' Instruction</a>
2839 <div class="doc_text">
2844 <result> = srem <ty> <op1>, <op2> <i>; yields {ty}:result</i>
2849 <p>The '<tt>srem</tt>' instruction returns the remainder from the
2850 signed division of its two operands. This instruction can also take
2851 <a href="#t_vector">vector</a> versions of the values in which case
2852 the elements must be integers.</p>
2856 <p>The two arguments to the '<tt>srem</tt>' instruction must be
2857 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
2858 values. Both arguments must have identical types.</p>
2862 <p>This instruction returns the <i>remainder</i> of a division (where the result
2863 has the same sign as the dividend, <tt>op1</tt>), not the <i>modulo</i>
2864 operator (where the result has the same sign as the divisor, <tt>op2</tt>) of
2865 a value. For more information about the difference, see <a
2866 href="http://mathforum.org/dr.math/problems/anne.4.28.99.html">The
2867 Math Forum</a>. For a table of how this is implemented in various languages,
2868 please see <a href="http://en.wikipedia.org/wiki/Modulo_operation">
2869 Wikipedia: modulo operation</a>.</p>
2870 <p>Note that signed integer remainder and unsigned integer remainder are
2871 distinct operations; for unsigned integer remainder, use '<tt>urem</tt>'.</p>
2872 <p>Taking the remainder of a division by zero leads to undefined behavior.
2873 Overflow also leads to undefined behavior; this is a rare case, but can occur,
2874 for example, by taking the remainder of a 32-bit division of -2147483648 by -1.
2875 (The remainder doesn't actually overflow, but this rule lets srem be
2876 implemented using instructions that return both the result of the division
2877 and the remainder.)</p>
2879 <pre> <result> = srem i32 4, %var <i>; yields {i32}:result = 4 % %var</i>
2883 <!-- _______________________________________________________________________ -->
2884 <div class="doc_subsubsection">
2885 <a name="i_frem">'<tt>frem</tt>' Instruction</a> </div>
2887 <div class="doc_text">
2890 <pre> <result> = frem <ty> <op1>, <op2> <i>; yields {ty}:result</i>
2893 <p>The '<tt>frem</tt>' instruction returns the remainder from the
2894 division of its two operands.</p>
2896 <p>The two arguments to the '<tt>frem</tt>' instruction must be
2897 <a href="#t_floating">floating point</a> or <a href="#t_vector">vector</a>
2898 of floating point values. Both arguments must have identical types.</p>
2902 <p>This instruction returns the <i>remainder</i> of a division.
2903 The remainder has the same sign as the dividend.</p>
2908 <result> = frem float 4.0, %var <i>; yields {float}:result = 4.0 % %var</i>
2912 <!-- ======================================================================= -->
2913 <div class="doc_subsection"> <a name="bitwiseops">Bitwise Binary
2914 Operations</a> </div>
2915 <div class="doc_text">
2916 <p>Bitwise binary operators are used to do various forms of
2917 bit-twiddling in a program. They are generally very efficient
2918 instructions and can commonly be strength reduced from other
2919 instructions. They require two operands of the same type, execute an operation on them,
2920 and produce a single value. The resulting value is the same type as its operands.</p>
2923 <!-- _______________________________________________________________________ -->
2924 <div class="doc_subsubsection"> <a name="i_shl">'<tt>shl</tt>'
2925 Instruction</a> </div>
2926 <div class="doc_text">
2928 <pre> <result> = shl <ty> <op1>, <op2> <i>; yields {ty}:result</i>
2933 <p>The '<tt>shl</tt>' instruction returns the first operand shifted to
2934 the left a specified number of bits.</p>
2938 <p>Both arguments to the '<tt>shl</tt>' instruction must be the same <a
2939 href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
2940 type. '<tt>op2</tt>' is treated as an unsigned value.</p>
2944 <p>The value produced is <tt>op1</tt> * 2<sup><tt>op2</tt></sup> mod 2<sup>n</sup>,
2945 where n is the width of the result. If <tt>op2</tt> is (statically or dynamically) negative or
2946 equal to or larger than the number of bits in <tt>op1</tt>, the result is undefined.
2947 If the arguments are vectors, each vector element of <tt>op1</tt> is shifted by the
2948 corresponding shift amount in <tt>op2</tt>.</p>
2950 <h5>Example:</h5><pre>
2951 <result> = shl i32 4, %var <i>; yields {i32}: 4 << %var</i>
2952 <result> = shl i32 4, 2 <i>; yields {i32}: 16</i>
2953 <result> = shl i32 1, 10 <i>; yields {i32}: 1024</i>
2954 <result> = shl i32 1, 32 <i>; undefined</i>
2955 <result> = shl <2 x i32> < i32 1, i32 1>, < i32 1, i32 2> <i>; yields: result=<2 x i32> < i32 2, i32 4></i>
2958 <!-- _______________________________________________________________________ -->
2959 <div class="doc_subsubsection"> <a name="i_lshr">'<tt>lshr</tt>'
2960 Instruction</a> </div>
2961 <div class="doc_text">
2963 <pre> <result> = lshr <ty> <op1>, <op2> <i>; yields {ty}:result</i>
2967 <p>The '<tt>lshr</tt>' instruction (logical shift right) returns the first
2968 operand shifted to the right a specified number of bits with zero fill.</p>
2971 <p>Both arguments to the '<tt>lshr</tt>' instruction must be the same
2972 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
2973 type. '<tt>op2</tt>' is treated as an unsigned value.</p>
2977 <p>This instruction always performs a logical shift right operation. The most
2978 significant bits of the result will be filled with zero bits after the
2979 shift. If <tt>op2</tt> is (statically or dynamically) equal to or larger than
2980 the number of bits in <tt>op1</tt>, the result is undefined. If the arguments are
2981 vectors, each vector element of <tt>op1</tt> is shifted by the corresponding shift
2982 amount in <tt>op2</tt>.</p>
2986 <result> = lshr i32 4, 1 <i>; yields {i32}:result = 2</i>
2987 <result> = lshr i32 4, 2 <i>; yields {i32}:result = 1</i>
2988 <result> = lshr i8 4, 3 <i>; yields {i8}:result = 0</i>
2989 <result> = lshr i8 -2, 1 <i>; yields {i8}:result = 0x7FFFFFFF </i>
2990 <result> = lshr i32 1, 32 <i>; undefined</i>
2991 <result> = lshr <2 x i32> < i32 -2, i32 4>, < i32 1, i32 2> <i>; yields: result=<2 x i32> < i32 0x7FFFFFFF, i32 1></i>
2995 <!-- _______________________________________________________________________ -->
2996 <div class="doc_subsubsection"> <a name="i_ashr">'<tt>ashr</tt>'
2997 Instruction</a> </div>
2998 <div class="doc_text">
3001 <pre> <result> = ashr <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3005 <p>The '<tt>ashr</tt>' instruction (arithmetic shift right) returns the first
3006 operand shifted to the right a specified number of bits with sign extension.</p>
3009 <p>Both arguments to the '<tt>ashr</tt>' instruction must be the same
3010 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
3011 type. '<tt>op2</tt>' is treated as an unsigned value.</p>
3014 <p>This instruction always performs an arithmetic shift right operation,
3015 The most significant bits of the result will be filled with the sign bit
3016 of <tt>op1</tt>. If <tt>op2</tt> is (statically or dynamically) equal to or
3017 larger than the number of bits in <tt>op1</tt>, the result is undefined. If the
3018 arguments are vectors, each vector element of <tt>op1</tt> is shifted by the
3019 corresponding shift amount in <tt>op2</tt>.</p>
3023 <result> = ashr i32 4, 1 <i>; yields {i32}:result = 2</i>
3024 <result> = ashr i32 4, 2 <i>; yields {i32}:result = 1</i>
3025 <result> = ashr i8 4, 3 <i>; yields {i8}:result = 0</i>
3026 <result> = ashr i8 -2, 1 <i>; yields {i8}:result = -1</i>
3027 <result> = ashr i32 1, 32 <i>; undefined</i>
3028 <result> = ashr <2 x i32> < i32 -2, i32 4>, < i32 1, i32 3> <i>; yields: result=<2 x i32> < i32 -1, i32 0></i>
3032 <!-- _______________________________________________________________________ -->
3033 <div class="doc_subsubsection"> <a name="i_and">'<tt>and</tt>'
3034 Instruction</a> </div>
3036 <div class="doc_text">
3041 <result> = and <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3046 <p>The '<tt>and</tt>' instruction returns the bitwise logical and of
3047 its two operands.</p>
3051 <p>The two arguments to the '<tt>and</tt>' instruction must be
3052 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
3053 values. Both arguments must have identical types.</p>
3056 <p>The truth table used for the '<tt>and</tt>' instruction is:</p>
3059 <table border="1" cellspacing="0" cellpadding="4">
3091 <result> = and i32 4, %var <i>; yields {i32}:result = 4 & %var</i>
3092 <result> = and i32 15, 40 <i>; yields {i32}:result = 8</i>
3093 <result> = and i32 4, 8 <i>; yields {i32}:result = 0</i>
3096 <!-- _______________________________________________________________________ -->
3097 <div class="doc_subsubsection"> <a name="i_or">'<tt>or</tt>' Instruction</a> </div>
3098 <div class="doc_text">
3100 <pre> <result> = or <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3103 <p>The '<tt>or</tt>' instruction returns the bitwise logical inclusive
3104 or of its two operands.</p>
3107 <p>The two arguments to the '<tt>or</tt>' instruction must be
3108 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
3109 values. Both arguments must have identical types.</p>
3111 <p>The truth table used for the '<tt>or</tt>' instruction is:</p>
3114 <table border="1" cellspacing="0" cellpadding="4">
3145 <pre> <result> = or i32 4, %var <i>; yields {i32}:result = 4 | %var</i>
3146 <result> = or i32 15, 40 <i>; yields {i32}:result = 47</i>
3147 <result> = or i32 4, 8 <i>; yields {i32}:result = 12</i>
3150 <!-- _______________________________________________________________________ -->
3151 <div class="doc_subsubsection"> <a name="i_xor">'<tt>xor</tt>'
3152 Instruction</a> </div>
3153 <div class="doc_text">
3155 <pre> <result> = xor <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3158 <p>The '<tt>xor</tt>' instruction returns the bitwise logical exclusive
3159 or of its two operands. The <tt>xor</tt> is used to implement the
3160 "one's complement" operation, which is the "~" operator in C.</p>
3162 <p>The two arguments to the '<tt>xor</tt>' instruction must be
3163 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
3164 values. Both arguments must have identical types.</p>
3168 <p>The truth table used for the '<tt>xor</tt>' instruction is:</p>
3171 <table border="1" cellspacing="0" cellpadding="4">
3203 <pre> <result> = xor i32 4, %var <i>; yields {i32}:result = 4 ^ %var</i>
3204 <result> = xor i32 15, 40 <i>; yields {i32}:result = 39</i>
3205 <result> = xor i32 4, 8 <i>; yields {i32}:result = 12</i>
3206 <result> = xor i32 %V, -1 <i>; yields {i32}:result = ~%V</i>
3210 <!-- ======================================================================= -->
3211 <div class="doc_subsection">
3212 <a name="vectorops">Vector Operations</a>
3215 <div class="doc_text">
3217 <p>LLVM supports several instructions to represent vector operations in a
3218 target-independent manner. These instructions cover the element-access and
3219 vector-specific operations needed to process vectors effectively. While LLVM
3220 does directly support these vector operations, many sophisticated algorithms
3221 will want to use target-specific intrinsics to take full advantage of a specific
3226 <!-- _______________________________________________________________________ -->
3227 <div class="doc_subsubsection">
3228 <a name="i_extractelement">'<tt>extractelement</tt>' Instruction</a>
3231 <div class="doc_text">
3236 <result> = extractelement <n x <ty>> <val>, i32 <idx> <i>; yields <ty></i>
3242 The '<tt>extractelement</tt>' instruction extracts a single scalar
3243 element from a vector at a specified index.
3250 The first operand of an '<tt>extractelement</tt>' instruction is a
3251 value of <a href="#t_vector">vector</a> type. The second operand is
3252 an index indicating the position from which to extract the element.
3253 The index may be a variable.</p>
3258 The result is a scalar of the same type as the element type of
3259 <tt>val</tt>. Its value is the value at position <tt>idx</tt> of
3260 <tt>val</tt>. If <tt>idx</tt> exceeds the length of <tt>val</tt>, the
3261 results are undefined.
3267 %result = extractelement <4 x i32> %vec, i32 0 <i>; yields i32</i>
3272 <!-- _______________________________________________________________________ -->
3273 <div class="doc_subsubsection">
3274 <a name="i_insertelement">'<tt>insertelement</tt>' Instruction</a>
3277 <div class="doc_text">
3282 <result> = insertelement <n x <ty>> <val>, <ty> <elt>, i32 <idx> <i>; yields <n x <ty>></i>
3288 The '<tt>insertelement</tt>' instruction inserts a scalar
3289 element into a vector at a specified index.
3296 The first operand of an '<tt>insertelement</tt>' instruction is a
3297 value of <a href="#t_vector">vector</a> type. The second operand is a
3298 scalar value whose type must equal the element type of the first
3299 operand. The third operand is an index indicating the position at
3300 which to insert the value. The index may be a variable.</p>
3305 The result is a vector of the same type as <tt>val</tt>. Its
3306 element values are those of <tt>val</tt> except at position
3307 <tt>idx</tt>, where it gets the value <tt>elt</tt>. If <tt>idx</tt>
3308 exceeds the length of <tt>val</tt>, the results are undefined.
3314 %result = insertelement <4 x i32> %vec, i32 1, i32 0 <i>; yields <4 x i32></i>
3318 <!-- _______________________________________________________________________ -->
3319 <div class="doc_subsubsection">
3320 <a name="i_shufflevector">'<tt>shufflevector</tt>' Instruction</a>
3323 <div class="doc_text">
3328 <result> = shufflevector <n x <ty>> <v1>, <n x <ty>> <v2>, <m x i32> <mask> <i>; yields <m x <ty>></i>
3334 The '<tt>shufflevector</tt>' instruction constructs a permutation of elements
3335 from two input vectors, returning a vector with the same element type as
3336 the input and length that is the same as the shuffle mask.
3342 The first two operands of a '<tt>shufflevector</tt>' instruction are vectors
3343 with types that match each other. The third argument is a shuffle mask whose
3344 element type is always 'i32'. The result of the instruction is a vector whose
3345 length is the same as the shuffle mask and whose element type is the same as
3346 the element type of the first two operands.
3350 The shuffle mask operand is required to be a constant vector with either
3351 constant integer or undef values.
3357 The elements of the two input vectors are numbered from left to right across
3358 both of the vectors. The shuffle mask operand specifies, for each element of
3359 the result vector, which element of the two input vectors the result element
3360 gets. The element selector may be undef (meaning "don't care") and the second
3361 operand may be undef if performing a shuffle from only one vector.
3367 %result = shufflevector <4 x i32> %v1, <4 x i32> %v2,
3368 <4 x i32> <i32 0, i32 4, i32 1, i32 5> <i>; yields <4 x i32></i>
3369 %result = shufflevector <4 x i32> %v1, <4 x i32> undef,
3370 <4 x i32> <i32 0, i32 1, i32 2, i32 3> <i>; yields <4 x i32></i> - Identity shuffle.
3371 %result = shufflevector <8 x i32> %v1, <8 x i32> undef,
3372 <4 x i32> <i32 0, i32 1, i32 2, i32 3> <i>; yields <4 x i32></i>
3373 %result = shufflevector <4 x i32> %v1, <4 x i32> %v2,
3374 <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>
3379 <!-- ======================================================================= -->
3380 <div class="doc_subsection">
3381 <a name="aggregateops">Aggregate Operations</a>
3384 <div class="doc_text">
3386 <p>LLVM supports several instructions for working with aggregate values.
3391 <!-- _______________________________________________________________________ -->
3392 <div class="doc_subsubsection">
3393 <a name="i_extractvalue">'<tt>extractvalue</tt>' Instruction</a>
3396 <div class="doc_text">
3401 <result> = extractvalue <aggregate type> <val>, <idx>{, <idx>}*
3407 The '<tt>extractvalue</tt>' instruction extracts the value of a struct field
3408 or array element from an aggregate value.
3415 The first operand of an '<tt>extractvalue</tt>' instruction is a
3416 value of <a href="#t_struct">struct</a> or <a href="#t_array">array</a>
3417 type. The operands are constant indices to specify which value to extract
3418 in a similar manner as indices in a
3419 '<tt><a href="#i_getelementptr">getelementptr</a></tt>' instruction.
3425 The result is the value at the position in the aggregate specified by
3432 %result = extractvalue {i32, float} %agg, 0 <i>; yields i32</i>
3437 <!-- _______________________________________________________________________ -->
3438 <div class="doc_subsubsection">
3439 <a name="i_insertvalue">'<tt>insertvalue</tt>' Instruction</a>
3442 <div class="doc_text">
3447 <result> = insertvalue <aggregate type> <val>, <ty> <val>, <idx> <i>; yields <n x <ty>></i>
3453 The '<tt>insertvalue</tt>' instruction inserts a value
3454 into a struct field or array element in an aggregate.
3461 The first operand of an '<tt>insertvalue</tt>' instruction is a
3462 value of <a href="#t_struct">struct</a> or <a href="#t_array">array</a> type.
3463 The second operand is a first-class value to insert.
3464 The following operands are constant indices
3465 indicating the position at which to insert the value in a similar manner as
3467 '<tt><a href="#i_getelementptr">getelementptr</a></tt>' instruction.
3468 The value to insert must have the same type as the value identified
3475 The result is an aggregate of the same type as <tt>val</tt>. Its
3476 value is that of <tt>val</tt> except that the value at the position
3477 specified by the indices is that of <tt>elt</tt>.
3483 %result = insertvalue {i32, float} %agg, i32 1, 0 <i>; yields {i32, float}</i>
3488 <!-- ======================================================================= -->
3489 <div class="doc_subsection">
3490 <a name="memoryops">Memory Access and Addressing Operations</a>
3493 <div class="doc_text">
3495 <p>A key design point of an SSA-based representation is how it
3496 represents memory. In LLVM, no memory locations are in SSA form, which
3497 makes things very simple. This section describes how to read, write,
3498 allocate, and free memory in LLVM.</p>
3502 <!-- _______________________________________________________________________ -->
3503 <div class="doc_subsubsection">
3504 <a name="i_malloc">'<tt>malloc</tt>' Instruction</a>
3507 <div class="doc_text">
3512 <result> = malloc <type>[, i32 <NumElements>][, align <alignment>] <i>; yields {type*}:result</i>
3517 <p>The '<tt>malloc</tt>' instruction allocates memory from the system
3518 heap and returns a pointer to it. The object is always allocated in the generic
3519 address space (address space zero).</p>
3523 <p>The '<tt>malloc</tt>' instruction allocates
3524 <tt>sizeof(<type>)*NumElements</tt>
3525 bytes of memory from the operating system and returns a pointer of the
3526 appropriate type to the program. If "NumElements" is specified, it is the
3527 number of elements allocated, otherwise "NumElements" is defaulted to be one.
3528 If a constant alignment is specified, the value result of the allocation is
3529 guaranteed to be aligned to at least that boundary. If not specified, or if
3530 zero, the target can choose to align the allocation on any convenient boundary
3531 compatible with the type.</p>
3533 <p>'<tt>type</tt>' must be a sized type.</p>
3537 <p>Memory is allocated using the system "<tt>malloc</tt>" function, and
3538 a pointer is returned. The result of a zero byte allocation is undefined. The
3539 result is null if there is insufficient memory available.</p>
3544 %array = malloc [4 x i8] <i>; yields {[%4 x i8]*}:array</i>
3546 %size = <a href="#i_add">add</a> i32 2, 2 <i>; yields {i32}:size = i32 4</i>
3547 %array1 = malloc i8, i32 4 <i>; yields {i8*}:array1</i>
3548 %array2 = malloc [12 x i8], i32 %size <i>; yields {[12 x i8]*}:array2</i>
3549 %array3 = malloc i32, i32 4, align 1024 <i>; yields {i32*}:array3</i>
3550 %array4 = malloc i32, align 1024 <i>; yields {i32*}:array4</i>
3553 <p>Note that the code generator does not yet respect the
3554 alignment value.</p>
3558 <!-- _______________________________________________________________________ -->
3559 <div class="doc_subsubsection">
3560 <a name="i_free">'<tt>free</tt>' Instruction</a>
3563 <div class="doc_text">
3568 free <type> <value> <i>; yields {void}</i>
3573 <p>The '<tt>free</tt>' instruction returns memory back to the unused
3574 memory heap to be reallocated in the future.</p>
3578 <p>'<tt>value</tt>' shall be a pointer value that points to a value
3579 that was allocated with the '<tt><a href="#i_malloc">malloc</a></tt>'
3584 <p>Access to the memory pointed to by the pointer is no longer defined
3585 after this instruction executes. If the pointer is null, the operation
3591 %array = <a href="#i_malloc">malloc</a> [4 x i8] <i>; yields {[4 x i8]*}:array</i>
3592 free [4 x i8]* %array
3596 <!-- _______________________________________________________________________ -->
3597 <div class="doc_subsubsection">
3598 <a name="i_alloca">'<tt>alloca</tt>' Instruction</a>
3601 <div class="doc_text">
3606 <result> = alloca <type>[, i32 <NumElements>][, align <alignment>] <i>; yields {type*}:result</i>
3611 <p>The '<tt>alloca</tt>' instruction allocates memory on the stack frame of the
3612 currently executing function, to be automatically released when this function
3613 returns to its caller. The object is always allocated in the generic address
3614 space (address space zero).</p>
3618 <p>The '<tt>alloca</tt>' instruction allocates <tt>sizeof(<type>)*NumElements</tt>
3619 bytes of memory on the runtime stack, returning a pointer of the
3620 appropriate type to the program. If "NumElements" is specified, it is the
3621 number of elements allocated, otherwise "NumElements" is defaulted to be one.
3622 If a constant alignment is specified, the value result of the allocation is
3623 guaranteed to be aligned to at least that boundary. If not specified, or if
3624 zero, the target can choose to align the allocation on any convenient boundary
3625 compatible with the type.</p>
3627 <p>'<tt>type</tt>' may be any sized type.</p>
3631 <p>Memory is allocated; a pointer is returned. The operation is undefined if
3632 there is insufficient stack space for the allocation. '<tt>alloca</tt>'d
3633 memory is automatically released when the function returns. The '<tt>alloca</tt>'
3634 instruction is commonly used to represent automatic variables that must
3635 have an address available. When the function returns (either with the <tt><a
3636 href="#i_ret">ret</a></tt> or <tt><a href="#i_unwind">unwind</a></tt>
3637 instructions), the memory is reclaimed. Allocating zero bytes
3638 is legal, but the result is undefined.</p>
3643 %ptr = alloca i32 <i>; yields {i32*}:ptr</i>
3644 %ptr = alloca i32, i32 4 <i>; yields {i32*}:ptr</i>
3645 %ptr = alloca i32, i32 4, align 1024 <i>; yields {i32*}:ptr</i>
3646 %ptr = alloca i32, align 1024 <i>; yields {i32*}:ptr</i>
3650 <!-- _______________________________________________________________________ -->
3651 <div class="doc_subsubsection"> <a name="i_load">'<tt>load</tt>'
3652 Instruction</a> </div>
3653 <div class="doc_text">
3655 <pre> <result> = load <ty>* <pointer>[, align <alignment>]<br> <result> = volatile load <ty>* <pointer>[, align <alignment>]<br></pre>
3657 <p>The '<tt>load</tt>' instruction is used to read from memory.</p>
3659 <p>The argument to the '<tt>load</tt>' instruction specifies the memory
3660 address from which to load. The pointer must point to a <a
3661 href="#t_firstclass">first class</a> type. If the <tt>load</tt> is
3662 marked as <tt>volatile</tt>, then the optimizer is not allowed to modify
3663 the number or order of execution of this <tt>load</tt> with other
3664 volatile <tt>load</tt> and <tt><a href="#i_store">store</a></tt>
3667 The optional constant "align" argument specifies the alignment of the operation
3668 (that is, the alignment of the memory address). A value of 0 or an
3669 omitted "align" argument means that the operation has the preferential
3670 alignment for the target. It is the responsibility of the code emitter
3671 to ensure that the alignment information is correct. Overestimating
3672 the alignment results in an undefined behavior. Underestimating the
3673 alignment may produce less efficient code. An alignment of 1 is always
3677 <p>The location of memory pointed to is loaded. If the value being loaded
3678 is of scalar type then the number of bytes read does not exceed the minimum
3679 number of bytes needed to hold all bits of the type. For example, loading an
3680 <tt>i24</tt> reads at most three bytes. When loading a value of a type like
3681 <tt>i20</tt> with a size that is not an integral number of bytes, the result
3682 is undefined if the value was not originally written using a store of the
3685 <pre> %ptr = <a href="#i_alloca">alloca</a> i32 <i>; yields {i32*}:ptr</i>
3687 href="#i_store">store</a> i32 3, i32* %ptr <i>; yields {void}</i>
3688 %val = load i32* %ptr <i>; yields {i32}:val = i32 3</i>
3691 <!-- _______________________________________________________________________ -->
3692 <div class="doc_subsubsection"> <a name="i_store">'<tt>store</tt>'
3693 Instruction</a> </div>
3694 <div class="doc_text">
3696 <pre> store <ty> <value>, <ty>* <pointer>[, align <alignment>] <i>; yields {void}</i>
3697 volatile store <ty> <value>, <ty>* <pointer>[, align <alignment>] <i>; yields {void}</i>
3700 <p>The '<tt>store</tt>' instruction is used to write to memory.</p>
3702 <p>There are two arguments to the '<tt>store</tt>' instruction: a value
3703 to store and an address at which to store it. The type of the '<tt><pointer></tt>'
3704 operand must be a pointer to the <a href="#t_firstclass">first class</a> type
3705 of the '<tt><value></tt>'
3706 operand. If the <tt>store</tt> is marked as <tt>volatile</tt>, then the
3707 optimizer is not allowed to modify the number or order of execution of
3708 this <tt>store</tt> with other volatile <tt>load</tt> and <tt><a
3709 href="#i_store">store</a></tt> instructions.</p>
3711 The optional constant "align" argument specifies the alignment of the operation
3712 (that is, the alignment of the memory address). A value of 0 or an
3713 omitted "align" argument means that the operation has the preferential
3714 alignment for the target. It is the responsibility of the code emitter
3715 to ensure that the alignment information is correct. Overestimating
3716 the alignment results in an undefined behavior. Underestimating the
3717 alignment may produce less efficient code. An alignment of 1 is always
3721 <p>The contents of memory are updated to contain '<tt><value></tt>'
3722 at the location specified by the '<tt><pointer></tt>' operand.
3723 If '<tt><value></tt>' is of scalar type then the number of bytes
3724 written does not exceed the minimum number of bytes needed to hold all
3725 bits of the type. For example, storing an <tt>i24</tt> writes at most
3726 three bytes. When writing a value of a type like <tt>i20</tt> with a
3727 size that is not an integral number of bytes, it is unspecified what
3728 happens to the extra bits that do not belong to the type, but they will
3729 typically be overwritten.</p>
3731 <pre> %ptr = <a href="#i_alloca">alloca</a> i32 <i>; yields {i32*}:ptr</i>
3732 store i32 3, i32* %ptr <i>; yields {void}</i>
3733 %val = <a href="#i_load">load</a> i32* %ptr <i>; yields {i32}:val = i32 3</i>
3737 <!-- _______________________________________________________________________ -->
3738 <div class="doc_subsubsection">
3739 <a name="i_getelementptr">'<tt>getelementptr</tt>' Instruction</a>
3742 <div class="doc_text">
3745 <result> = getelementptr <pty>* <ptrval>{, <ty> <idx>}*
3751 The '<tt>getelementptr</tt>' instruction is used to get the address of a
3752 subelement of an aggregate data structure. It performs address calculation only
3753 and does not access memory.</p>
3757 <p>The first argument is always a pointer, and forms the basis of the
3758 calculation. The remaining arguments are indices, that indicate which of the
3759 elements of the aggregate object are indexed. The interpretation of each index
3760 is dependent on the type being indexed into. The first index always indexes the
3761 pointer value given as the first argument, the second index indexes a value of
3762 the type pointed to (not necessarily the value directly pointed to, since the
3763 first index can be non-zero), etc. The first type indexed into must be a pointer
3764 value, subsequent types can be arrays, vectors and structs. Note that subsequent
3765 types being indexed into can never be pointers, since that would require loading
3766 the pointer before continuing calculation.</p>
3768 <p>The type of each index argument depends on the type it is indexing into.
3769 When indexing into a (packed) structure, only <tt>i32</tt> integer
3770 <b>constants</b> are allowed. When indexing into an array, pointer or vector,
3771 integers of any width are allowed (also non-constants).</p>
3773 <p>For example, let's consider a C code fragment and how it gets
3774 compiled to LLVM:</p>
3776 <div class="doc_code">
3789 int *foo(struct ST *s) {
3790 return &s[1].Z.B[5][13];
3795 <p>The LLVM code generated by the GCC frontend is:</p>
3797 <div class="doc_code">
3799 %RT = <a href="#namedtypes">type</a> { i8 , [10 x [20 x i32]], i8 }
3800 %ST = <a href="#namedtypes">type</a> { i32, double, %RT }
3802 define i32* %foo(%ST* %s) {
3804 %reg = getelementptr %ST* %s, i32 1, i32 2, i32 1, i32 5, i32 13
3812 <p>In the example above, the first index is indexing into the '<tt>%ST*</tt>'
3813 type, which is a pointer, yielding a '<tt>%ST</tt>' = '<tt>{ i32, double, %RT
3814 }</tt>' type, a structure. The second index indexes into the third element of
3815 the structure, yielding a '<tt>%RT</tt>' = '<tt>{ i8 , [10 x [20 x i32]],
3816 i8 }</tt>' type, another structure. The third index indexes into the second
3817 element of the structure, yielding a '<tt>[10 x [20 x i32]]</tt>' type, an
3818 array. The two dimensions of the array are subscripted into, yielding an
3819 '<tt>i32</tt>' type. The '<tt>getelementptr</tt>' instruction returns a pointer
3820 to this element, thus computing a value of '<tt>i32*</tt>' type.</p>
3822 <p>Note that it is perfectly legal to index partially through a
3823 structure, returning a pointer to an inner element. Because of this,
3824 the LLVM code for the given testcase is equivalent to:</p>
3827 define i32* %foo(%ST* %s) {
3828 %t1 = getelementptr %ST* %s, i32 1 <i>; yields %ST*:%t1</i>
3829 %t2 = getelementptr %ST* %t1, i32 0, i32 2 <i>; yields %RT*:%t2</i>
3830 %t3 = getelementptr %RT* %t2, i32 0, i32 1 <i>; yields [10 x [20 x i32]]*:%t3</i>
3831 %t4 = getelementptr [10 x [20 x i32]]* %t3, i32 0, i32 5 <i>; yields [20 x i32]*:%t4</i>
3832 %t5 = getelementptr [20 x i32]* %t4, i32 0, i32 13 <i>; yields i32*:%t5</i>
3837 <p>Note that it is undefined to access an array out of bounds: array
3838 and pointer indexes must always be within the defined bounds of the
3839 array type when accessed with an instruction that dereferences the
3840 pointer (e.g. a load or store instruction). The one exception for
3841 this rule is zero length arrays. These arrays are defined to be
3842 accessible as variable length arrays, which requires access beyond the
3843 zero'th element.</p>
3845 <p>The getelementptr instruction is often confusing. For some more insight
3846 into how it works, see <a href="GetElementPtr.html">the getelementptr
3852 <i>; yields [12 x i8]*:aptr</i>
3853 %aptr = getelementptr {i32, [12 x i8]}* %saptr, i64 0, i32 1
3854 <i>; yields i8*:vptr</i>
3855 %vptr = getelementptr {i32, <2 x i8>}* %svptr, i64 0, i32 1, i32 1
3856 <i>; yields i8*:eptr</i>
3857 %eptr = getelementptr [12 x i8]* %aptr, i64 0, i32 1
3858 <i>; yields i32*:iptr</i>
3859 %iptr = getelementptr [10 x i32]* @arr, i16 0, i16 0
3863 <!-- ======================================================================= -->
3864 <div class="doc_subsection"> <a name="convertops">Conversion Operations</a>
3866 <div class="doc_text">
3867 <p>The instructions in this category are the conversion instructions (casting)
3868 which all take a single operand and a type. They perform various bit conversions
3872 <!-- _______________________________________________________________________ -->
3873 <div class="doc_subsubsection">
3874 <a name="i_trunc">'<tt>trunc .. to</tt>' Instruction</a>
3876 <div class="doc_text">
3880 <result> = trunc <ty> <value> to <ty2> <i>; yields ty2</i>
3885 The '<tt>trunc</tt>' instruction truncates its operand to the type <tt>ty2</tt>.
3890 The '<tt>trunc</tt>' instruction takes a <tt>value</tt> to trunc, which must
3891 be an <a href="#t_integer">integer</a> type, and a type that specifies the size
3892 and type of the result, which must be an <a href="#t_integer">integer</a>
3893 type. The bit size of <tt>value</tt> must be larger than the bit size of
3894 <tt>ty2</tt>. Equal sized types are not allowed.</p>
3898 The '<tt>trunc</tt>' instruction truncates the high order bits in <tt>value</tt>
3899 and converts the remaining bits to <tt>ty2</tt>. Since the source size must be
3900 larger than the destination size, <tt>trunc</tt> cannot be a <i>no-op cast</i>.
3901 It will always truncate bits.</p>
3905 %X = trunc i32 257 to i8 <i>; yields i8:1</i>
3906 %Y = trunc i32 123 to i1 <i>; yields i1:true</i>
3907 %Y = trunc i32 122 to i1 <i>; yields i1:false</i>
3911 <!-- _______________________________________________________________________ -->
3912 <div class="doc_subsubsection">
3913 <a name="i_zext">'<tt>zext .. to</tt>' Instruction</a>
3915 <div class="doc_text">
3919 <result> = zext <ty> <value> to <ty2> <i>; yields ty2</i>
3923 <p>The '<tt>zext</tt>' instruction zero extends its operand to type
3928 <p>The '<tt>zext</tt>' instruction takes a value to cast, which must be of
3929 <a href="#t_integer">integer</a> type, and a type to cast it to, which must
3930 also be of <a href="#t_integer">integer</a> type. The bit size of the
3931 <tt>value</tt> must be smaller than the bit size of the destination type,
3935 <p>The <tt>zext</tt> fills the high order bits of the <tt>value</tt> with zero
3936 bits until it reaches the size of the destination type, <tt>ty2</tt>.</p>
3938 <p>When zero extending from i1, the result will always be either 0 or 1.</p>
3942 %X = zext i32 257 to i64 <i>; yields i64:257</i>
3943 %Y = zext i1 true to i32 <i>; yields i32:1</i>
3947 <!-- _______________________________________________________________________ -->
3948 <div class="doc_subsubsection">
3949 <a name="i_sext">'<tt>sext .. to</tt>' Instruction</a>
3951 <div class="doc_text">
3955 <result> = sext <ty> <value> to <ty2> <i>; yields ty2</i>
3959 <p>The '<tt>sext</tt>' sign extends <tt>value</tt> to the type <tt>ty2</tt>.</p>
3963 The '<tt>sext</tt>' instruction takes a value to cast, which must be of
3964 <a href="#t_integer">integer</a> type, and a type to cast it to, which must
3965 also be of <a href="#t_integer">integer</a> type. The bit size of the
3966 <tt>value</tt> must be smaller than the bit size of the destination type,
3971 The '<tt>sext</tt>' instruction performs a sign extension by copying the sign
3972 bit (highest order bit) of the <tt>value</tt> until it reaches the bit size of
3973 the type <tt>ty2</tt>.</p>
3975 <p>When sign extending from i1, the extension always results in -1 or 0.</p>
3979 %X = sext i8 -1 to i16 <i>; yields i16 :65535</i>
3980 %Y = sext i1 true to i32 <i>; yields i32:-1</i>
3984 <!-- _______________________________________________________________________ -->
3985 <div class="doc_subsubsection">
3986 <a name="i_fptrunc">'<tt>fptrunc .. to</tt>' Instruction</a>
3989 <div class="doc_text">
3994 <result> = fptrunc <ty> <value> to <ty2> <i>; yields ty2</i>
3998 <p>The '<tt>fptrunc</tt>' instruction truncates <tt>value</tt> to type
4003 <p>The '<tt>fptrunc</tt>' instruction takes a <a href="#t_floating">floating
4004 point</a> value to cast and a <a href="#t_floating">floating point</a> type to
4005 cast it to. The size of <tt>value</tt> must be larger than the size of
4006 <tt>ty2</tt>. This implies that <tt>fptrunc</tt> cannot be used to make a
4007 <i>no-op cast</i>.</p>
4010 <p> The '<tt>fptrunc</tt>' instruction truncates a <tt>value</tt> from a larger
4011 <a href="#t_floating">floating point</a> type to a smaller
4012 <a href="#t_floating">floating point</a> type. If the value cannot fit within
4013 the destination type, <tt>ty2</tt>, then the results are undefined.</p>
4017 %X = fptrunc double 123.0 to float <i>; yields float:123.0</i>
4018 %Y = fptrunc double 1.0E+300 to float <i>; yields undefined</i>
4022 <!-- _______________________________________________________________________ -->
4023 <div class="doc_subsubsection">
4024 <a name="i_fpext">'<tt>fpext .. to</tt>' Instruction</a>
4026 <div class="doc_text">
4030 <result> = fpext <ty> <value> to <ty2> <i>; yields ty2</i>
4034 <p>The '<tt>fpext</tt>' extends a floating point <tt>value</tt> to a larger
4035 floating point value.</p>
4038 <p>The '<tt>fpext</tt>' instruction takes a
4039 <a href="#t_floating">floating point</a> <tt>value</tt> to cast,
4040 and a <a href="#t_floating">floating point</a> type to cast it to. The source
4041 type must be smaller than the destination type.</p>
4044 <p>The '<tt>fpext</tt>' instruction extends the <tt>value</tt> from a smaller
4045 <a href="#t_floating">floating point</a> type to a larger
4046 <a href="#t_floating">floating point</a> type. The <tt>fpext</tt> cannot be
4047 used to make a <i>no-op cast</i> because it always changes bits. Use
4048 <tt>bitcast</tt> to make a <i>no-op cast</i> for a floating point cast.</p>
4052 %X = fpext float 3.1415 to double <i>; yields double:3.1415</i>
4053 %Y = fpext float 1.0 to float <i>; yields float:1.0 (no-op)</i>
4057 <!-- _______________________________________________________________________ -->
4058 <div class="doc_subsubsection">
4059 <a name="i_fptoui">'<tt>fptoui .. to</tt>' Instruction</a>
4061 <div class="doc_text">
4065 <result> = fptoui <ty> <value> to <ty2> <i>; yields ty2</i>
4069 <p>The '<tt>fptoui</tt>' converts a floating point <tt>value</tt> to its
4070 unsigned integer equivalent of type <tt>ty2</tt>.
4074 <p>The '<tt>fptoui</tt>' instruction takes a value to cast, which must be a
4075 scalar or vector <a href="#t_floating">floating point</a> value, and a type
4076 to cast it to <tt>ty2</tt>, which must be an <a href="#t_integer">integer</a>
4077 type. If <tt>ty</tt> is a vector floating point type, <tt>ty2</tt> must be a
4078 vector integer type with the same number of elements as <tt>ty</tt></p>
4081 <p> The '<tt>fptoui</tt>' instruction converts its
4082 <a href="#t_floating">floating point</a> operand into the nearest (rounding
4083 towards zero) unsigned integer value. If the value cannot fit in <tt>ty2</tt>,
4084 the results are undefined.</p>
4088 %X = fptoui double 123.0 to i32 <i>; yields i32:123</i>
4089 %Y = fptoui float 1.0E+300 to i1 <i>; yields undefined:1</i>
4090 %X = fptoui float 1.04E+17 to i8 <i>; yields undefined:1</i>
4094 <!-- _______________________________________________________________________ -->
4095 <div class="doc_subsubsection">
4096 <a name="i_fptosi">'<tt>fptosi .. to</tt>' Instruction</a>
4098 <div class="doc_text">
4102 <result> = fptosi <ty> <value> to <ty2> <i>; yields ty2</i>
4106 <p>The '<tt>fptosi</tt>' instruction converts
4107 <a href="#t_floating">floating point</a> <tt>value</tt> to type <tt>ty2</tt>.
4111 <p> The '<tt>fptosi</tt>' instruction takes a value to cast, which must be a
4112 scalar or vector <a href="#t_floating">floating point</a> value, and a type
4113 to cast it to <tt>ty2</tt>, which must be an <a href="#t_integer">integer</a>
4114 type. If <tt>ty</tt> is a vector floating point type, <tt>ty2</tt> must be a
4115 vector integer type with the same number of elements as <tt>ty</tt></p>
4118 <p>The '<tt>fptosi</tt>' instruction converts its
4119 <a href="#t_floating">floating point</a> operand into the nearest (rounding
4120 towards zero) signed integer value. If the value cannot fit in <tt>ty2</tt>,
4121 the results are undefined.</p>
4125 %X = fptosi double -123.0 to i32 <i>; yields i32:-123</i>
4126 %Y = fptosi float 1.0E-247 to i1 <i>; yields undefined:1</i>
4127 %X = fptosi float 1.04E+17 to i8 <i>; yields undefined:1</i>
4131 <!-- _______________________________________________________________________ -->
4132 <div class="doc_subsubsection">
4133 <a name="i_uitofp">'<tt>uitofp .. to</tt>' Instruction</a>
4135 <div class="doc_text">
4139 <result> = uitofp <ty> <value> to <ty2> <i>; yields ty2</i>
4143 <p>The '<tt>uitofp</tt>' instruction regards <tt>value</tt> as an unsigned
4144 integer and converts that value to the <tt>ty2</tt> type.</p>
4147 <p>The '<tt>uitofp</tt>' instruction takes a value to cast, which must be a
4148 scalar or vector <a href="#t_integer">integer</a> value, and a type to cast it
4149 to <tt>ty2</tt>, which must be an <a href="#t_floating">floating point</a>
4150 type. If <tt>ty</tt> is a vector integer type, <tt>ty2</tt> must be a vector
4151 floating point type with the same number of elements as <tt>ty</tt></p>
4154 <p>The '<tt>uitofp</tt>' instruction interprets its operand as an unsigned
4155 integer quantity and converts it to the corresponding floating point value. If
4156 the value cannot fit in the floating point value, the results are undefined.</p>
4160 %X = uitofp i32 257 to float <i>; yields float:257.0</i>
4161 %Y = uitofp i8 -1 to double <i>; yields double:255.0</i>
4165 <!-- _______________________________________________________________________ -->
4166 <div class="doc_subsubsection">
4167 <a name="i_sitofp">'<tt>sitofp .. to</tt>' Instruction</a>
4169 <div class="doc_text">
4173 <result> = sitofp <ty> <value> to <ty2> <i>; yields ty2</i>
4177 <p>The '<tt>sitofp</tt>' instruction regards <tt>value</tt> as a signed
4178 integer and converts that value to the <tt>ty2</tt> type.</p>
4181 <p>The '<tt>sitofp</tt>' instruction takes a value to cast, which must be a
4182 scalar or vector <a href="#t_integer">integer</a> value, and a type to cast it
4183 to <tt>ty2</tt>, which must be an <a href="#t_floating">floating point</a>
4184 type. If <tt>ty</tt> is a vector integer type, <tt>ty2</tt> must be a vector
4185 floating point type with the same number of elements as <tt>ty</tt></p>
4188 <p>The '<tt>sitofp</tt>' instruction interprets its operand as a signed
4189 integer quantity and converts it to the corresponding floating point value. If
4190 the value cannot fit in the floating point value, the results are undefined.</p>
4194 %X = sitofp i32 257 to float <i>; yields float:257.0</i>
4195 %Y = sitofp i8 -1 to double <i>; yields double:-1.0</i>
4199 <!-- _______________________________________________________________________ -->
4200 <div class="doc_subsubsection">
4201 <a name="i_ptrtoint">'<tt>ptrtoint .. to</tt>' Instruction</a>
4203 <div class="doc_text">
4207 <result> = ptrtoint <ty> <value> to <ty2> <i>; yields ty2</i>
4211 <p>The '<tt>ptrtoint</tt>' instruction converts the pointer <tt>value</tt> to
4212 the integer type <tt>ty2</tt>.</p>
4215 <p>The '<tt>ptrtoint</tt>' instruction takes a <tt>value</tt> to cast, which
4216 must be a <a href="#t_pointer">pointer</a> value, and a type to cast it to
4217 <tt>ty2</tt>, which must be an <a href="#t_integer">integer</a> type.</p>
4220 <p>The '<tt>ptrtoint</tt>' instruction converts <tt>value</tt> to integer type
4221 <tt>ty2</tt> by interpreting the pointer value as an integer and either
4222 truncating or zero extending that value to the size of the integer type. If
4223 <tt>value</tt> is smaller than <tt>ty2</tt> then a zero extension is done. If
4224 <tt>value</tt> is larger than <tt>ty2</tt> then a truncation is done. If they
4225 are the same size, then nothing is done (<i>no-op cast</i>) other than a type
4230 %X = ptrtoint i32* %X to i8 <i>; yields truncation on 32-bit architecture</i>
4231 %Y = ptrtoint i32* %x to i64 <i>; yields zero extension on 32-bit architecture</i>
4235 <!-- _______________________________________________________________________ -->
4236 <div class="doc_subsubsection">
4237 <a name="i_inttoptr">'<tt>inttoptr .. to</tt>' Instruction</a>
4239 <div class="doc_text">
4243 <result> = inttoptr <ty> <value> to <ty2> <i>; yields ty2</i>
4247 <p>The '<tt>inttoptr</tt>' instruction converts an integer <tt>value</tt> to
4248 a pointer type, <tt>ty2</tt>.</p>
4251 <p>The '<tt>inttoptr</tt>' instruction takes an <a href="#t_integer">integer</a>
4252 value to cast, and a type to cast it to, which must be a
4253 <a href="#t_pointer">pointer</a> type.</p>
4256 <p>The '<tt>inttoptr</tt>' instruction converts <tt>value</tt> to type
4257 <tt>ty2</tt> by applying either a zero extension or a truncation depending on
4258 the size of the integer <tt>value</tt>. If <tt>value</tt> is larger than the
4259 size of a pointer then a truncation is done. If <tt>value</tt> is smaller than
4260 the size of a pointer then a zero extension is done. If they are the same size,
4261 nothing is done (<i>no-op cast</i>).</p>
4265 %X = inttoptr i32 255 to i32* <i>; yields zero extension on 64-bit architecture</i>
4266 %X = inttoptr i32 255 to i32* <i>; yields no-op on 32-bit architecture</i>
4267 %Y = inttoptr i64 0 to i32* <i>; yields truncation on 32-bit architecture</i>
4271 <!-- _______________________________________________________________________ -->
4272 <div class="doc_subsubsection">
4273 <a name="i_bitcast">'<tt>bitcast .. to</tt>' Instruction</a>
4275 <div class="doc_text">
4279 <result> = bitcast <ty> <value> to <ty2> <i>; yields ty2</i>
4284 <p>The '<tt>bitcast</tt>' instruction converts <tt>value</tt> to type
4285 <tt>ty2</tt> without changing any bits.</p>
4289 <p>The '<tt>bitcast</tt>' instruction takes a value to cast, which must be
4290 a non-aggregate first class value, and a type to cast it to, which must also be
4291 a non-aggregate <a href="#t_firstclass">first class</a> type. The bit sizes of
4293 and the destination type, <tt>ty2</tt>, must be identical. If the source
4294 type is a pointer, the destination type must also be a pointer. This
4295 instruction supports bitwise conversion of vectors to integers and to vectors
4296 of other types (as long as they have the same size).</p>
4299 <p>The '<tt>bitcast</tt>' instruction converts <tt>value</tt> to type
4300 <tt>ty2</tt>. It is always a <i>no-op cast</i> because no bits change with
4301 this conversion. The conversion is done as if the <tt>value</tt> had been
4302 stored to memory and read back as type <tt>ty2</tt>. Pointer types may only be
4303 converted to other pointer types with this instruction. To convert pointers to
4304 other types, use the <a href="#i_inttoptr">inttoptr</a> or
4305 <a href="#i_ptrtoint">ptrtoint</a> instructions first.</p>
4309 %X = bitcast i8 255 to i8 <i>; yields i8 :-1</i>
4310 %Y = bitcast i32* %x to sint* <i>; yields sint*:%x</i>
4311 %Z = bitcast <2 x int> %V to i64; <i>; yields i64: %V</i>
4315 <!-- ======================================================================= -->
4316 <div class="doc_subsection"> <a name="otherops">Other Operations</a> </div>
4317 <div class="doc_text">
4318 <p>The instructions in this category are the "miscellaneous"
4319 instructions, which defy better classification.</p>
4322 <!-- _______________________________________________________________________ -->
4323 <div class="doc_subsubsection"><a name="i_icmp">'<tt>icmp</tt>' Instruction</a>
4325 <div class="doc_text">
4327 <pre> <result> = icmp <cond> <ty> <op1>, <op2> <i>; yields {i1} or {<N x i1>}:result</i>
4330 <p>The '<tt>icmp</tt>' instruction returns a boolean value or
4331 a vector of boolean values based on comparison
4332 of its two integer, integer vector, or pointer operands.</p>
4334 <p>The '<tt>icmp</tt>' instruction takes three operands. The first operand is
4335 the condition code indicating the kind of comparison to perform. It is not
4336 a value, just a keyword. The possible condition code are:
4339 <li><tt>eq</tt>: equal</li>
4340 <li><tt>ne</tt>: not equal </li>
4341 <li><tt>ugt</tt>: unsigned greater than</li>
4342 <li><tt>uge</tt>: unsigned greater or equal</li>
4343 <li><tt>ult</tt>: unsigned less than</li>
4344 <li><tt>ule</tt>: unsigned less or equal</li>
4345 <li><tt>sgt</tt>: signed greater than</li>
4346 <li><tt>sge</tt>: signed greater or equal</li>
4347 <li><tt>slt</tt>: signed less than</li>
4348 <li><tt>sle</tt>: signed less or equal</li>
4350 <p>The remaining two arguments must be <a href="#t_integer">integer</a> or
4351 <a href="#t_pointer">pointer</a>
4352 or integer <a href="#t_vector">vector</a> typed.
4353 They must also be identical types.</p>
4355 <p>The '<tt>icmp</tt>' compares <tt>op1</tt> and <tt>op2</tt> according to
4356 the condition code given as <tt>cond</tt>. The comparison performed always
4357 yields either an <a href="#t_primitive"><tt>i1</tt></a> or vector of <tt>i1</tt> result, as follows:
4360 <li><tt>eq</tt>: yields <tt>true</tt> if the operands are equal,
4361 <tt>false</tt> otherwise. No sign interpretation is necessary or performed.
4363 <li><tt>ne</tt>: yields <tt>true</tt> if the operands are unequal,
4364 <tt>false</tt> otherwise. No sign interpretation is necessary or performed.</li>
4365 <li><tt>ugt</tt>: interprets the operands as unsigned values and yields
4366 <tt>true</tt> if <tt>op1</tt> is greater than <tt>op2</tt>.</li>
4367 <li><tt>uge</tt>: interprets the operands as unsigned values and yields
4368 <tt>true</tt> if <tt>op1</tt> is greater than or equal to <tt>op2</tt>.</li>
4369 <li><tt>ult</tt>: interprets the operands as unsigned values and yields
4370 <tt>true</tt> if <tt>op1</tt> is less than <tt>op2</tt>.</li>
4371 <li><tt>ule</tt>: interprets the operands as unsigned values and yields
4372 <tt>true</tt> if <tt>op1</tt> is less than or equal to <tt>op2</tt>.</li>
4373 <li><tt>sgt</tt>: interprets the operands as signed values and yields
4374 <tt>true</tt> if <tt>op1</tt> is greater than <tt>op2</tt>.</li>
4375 <li><tt>sge</tt>: interprets the operands as signed values and yields
4376 <tt>true</tt> if <tt>op1</tt> is greater than or equal to <tt>op2</tt>.</li>
4377 <li><tt>slt</tt>: interprets the operands as signed values and yields
4378 <tt>true</tt> if <tt>op1</tt> is less than <tt>op2</tt>.</li>
4379 <li><tt>sle</tt>: interprets the operands as signed values and yields
4380 <tt>true</tt> if <tt>op1</tt> is less than or equal to <tt>op2</tt>.</li>
4382 <p>If the operands are <a href="#t_pointer">pointer</a> typed, the pointer
4383 values are compared as if they were integers.</p>
4384 <p>If the operands are integer vectors, then they are compared
4385 element by element. The result is an <tt>i1</tt> vector with
4386 the same number of elements as the values being compared.
4387 Otherwise, the result is an <tt>i1</tt>.
4391 <pre> <result> = icmp eq i32 4, 5 <i>; yields: result=false</i>
4392 <result> = icmp ne float* %X, %X <i>; yields: result=false</i>
4393 <result> = icmp ult i16 4, 5 <i>; yields: result=true</i>
4394 <result> = icmp sgt i16 4, 5 <i>; yields: result=false</i>
4395 <result> = icmp ule i16 -4, 5 <i>; yields: result=false</i>
4396 <result> = icmp sge i16 4, 5 <i>; yields: result=false</i>
4399 <p>Note that the code generator does not yet support vector types with
4400 the <tt>icmp</tt> instruction.</p>
4404 <!-- _______________________________________________________________________ -->
4405 <div class="doc_subsubsection"><a name="i_fcmp">'<tt>fcmp</tt>' Instruction</a>
4407 <div class="doc_text">
4409 <pre> <result> = fcmp <cond> <ty> <op1>, <op2> <i>; yields {i1} or {<N x i1>}:result</i>
4412 <p>The '<tt>fcmp</tt>' instruction returns a boolean value
4413 or vector of boolean values based on comparison
4414 of its operands.</p>
4416 If the operands are floating point scalars, then the result
4417 type is a boolean (<a href="#t_primitive"><tt>i1</tt></a>).
4419 <p>If the operands are floating point vectors, then the result type
4420 is a vector of boolean with the same number of elements as the
4421 operands being compared.</p>
4423 <p>The '<tt>fcmp</tt>' instruction takes three operands. The first operand is
4424 the condition code indicating the kind of comparison to perform. It is not
4425 a value, just a keyword. The possible condition code are:</p>
4427 <li><tt>false</tt>: no comparison, always returns false</li>
4428 <li><tt>oeq</tt>: ordered and equal</li>
4429 <li><tt>ogt</tt>: ordered and greater than </li>
4430 <li><tt>oge</tt>: ordered and greater than or equal</li>
4431 <li><tt>olt</tt>: ordered and less than </li>
4432 <li><tt>ole</tt>: ordered and less than or equal</li>
4433 <li><tt>one</tt>: ordered and not equal</li>
4434 <li><tt>ord</tt>: ordered (no nans)</li>
4435 <li><tt>ueq</tt>: unordered or equal</li>
4436 <li><tt>ugt</tt>: unordered or greater than </li>
4437 <li><tt>uge</tt>: unordered or greater than or equal</li>
4438 <li><tt>ult</tt>: unordered or less than </li>
4439 <li><tt>ule</tt>: unordered or less than or equal</li>
4440 <li><tt>une</tt>: unordered or not equal</li>
4441 <li><tt>uno</tt>: unordered (either nans)</li>
4442 <li><tt>true</tt>: no comparison, always returns true</li>
4444 <p><i>Ordered</i> means that neither operand is a QNAN while
4445 <i>unordered</i> means that either operand may be a QNAN.</p>
4446 <p>Each of <tt>val1</tt> and <tt>val2</tt> arguments must be
4447 either a <a href="#t_floating">floating point</a> type
4448 or a <a href="#t_vector">vector</a> of floating point type.
4449 They must have identical types.</p>
4451 <p>The '<tt>fcmp</tt>' instruction compares <tt>op1</tt> and <tt>op2</tt>
4452 according to the condition code given as <tt>cond</tt>.
4453 If the operands are vectors, then the vectors are compared
4455 Each comparison performed
4456 always yields an <a href="#t_primitive">i1</a> result, as follows:</p>
4458 <li><tt>false</tt>: always yields <tt>false</tt>, regardless of operands.</li>
4459 <li><tt>oeq</tt>: yields <tt>true</tt> if both operands are not a QNAN and
4460 <tt>op1</tt> is equal to <tt>op2</tt>.</li>
4461 <li><tt>ogt</tt>: yields <tt>true</tt> if both operands are not a QNAN and
4462 <tt>op1</tt> is greather than <tt>op2</tt>.</li>
4463 <li><tt>oge</tt>: yields <tt>true</tt> if both operands are not a QNAN and
4464 <tt>op1</tt> is greater than or equal to <tt>op2</tt>.</li>
4465 <li><tt>olt</tt>: yields <tt>true</tt> if both operands are not a QNAN and
4466 <tt>op1</tt> is less than <tt>op2</tt>.</li>
4467 <li><tt>ole</tt>: yields <tt>true</tt> if both operands are not a QNAN and
4468 <tt>op1</tt> is less than or equal to <tt>op2</tt>.</li>
4469 <li><tt>one</tt>: yields <tt>true</tt> if both operands are not a QNAN and
4470 <tt>op1</tt> is not equal to <tt>op2</tt>.</li>
4471 <li><tt>ord</tt>: yields <tt>true</tt> if both operands are not a QNAN.</li>
4472 <li><tt>ueq</tt>: yields <tt>true</tt> if either operand is a QNAN or
4473 <tt>op1</tt> is equal to <tt>op2</tt>.</li>
4474 <li><tt>ugt</tt>: yields <tt>true</tt> if either operand is a QNAN or
4475 <tt>op1</tt> is greater than <tt>op2</tt>.</li>
4476 <li><tt>uge</tt>: yields <tt>true</tt> if either operand is a QNAN or
4477 <tt>op1</tt> is greater than or equal to <tt>op2</tt>.</li>
4478 <li><tt>ult</tt>: yields <tt>true</tt> if either operand is a QNAN or
4479 <tt>op1</tt> is less than <tt>op2</tt>.</li>
4480 <li><tt>ule</tt>: yields <tt>true</tt> if either operand is a QNAN or
4481 <tt>op1</tt> is less than or equal to <tt>op2</tt>.</li>
4482 <li><tt>une</tt>: yields <tt>true</tt> if either operand is a QNAN or
4483 <tt>op1</tt> is not equal to <tt>op2</tt>.</li>
4484 <li><tt>uno</tt>: yields <tt>true</tt> if either operand is a QNAN.</li>
4485 <li><tt>true</tt>: always yields <tt>true</tt>, regardless of operands.</li>
4489 <pre> <result> = fcmp oeq float 4.0, 5.0 <i>; yields: result=false</i>
4490 <result> = fcmp one float 4.0, 5.0 <i>; yields: result=true</i>
4491 <result> = fcmp olt float 4.0, 5.0 <i>; yields: result=true</i>
4492 <result> = fcmp ueq double 1.0, 2.0 <i>; yields: result=false</i>
4495 <p>Note that the code generator does not yet support vector types with
4496 the <tt>fcmp</tt> instruction.</p>
4500 <!-- _______________________________________________________________________ -->
4501 <div class="doc_subsubsection">
4502 <a name="i_phi">'<tt>phi</tt>' Instruction</a>
4505 <div class="doc_text">
4509 <pre> <result> = phi <ty> [ <val0>, <label0>], ...<br></pre>
4511 <p>The '<tt>phi</tt>' instruction is used to implement the φ node in
4512 the SSA graph representing the function.</p>
4515 <p>The type of the incoming values is specified with the first type
4516 field. After this, the '<tt>phi</tt>' instruction takes a list of pairs
4517 as arguments, with one pair for each predecessor basic block of the
4518 current block. Only values of <a href="#t_firstclass">first class</a>
4519 type may be used as the value arguments to the PHI node. Only labels
4520 may be used as the label arguments.</p>
4522 <p>There must be no non-phi instructions between the start of a basic
4523 block and the PHI instructions: i.e. PHI instructions must be first in
4526 <p>For the purposes of the SSA form, the use of each incoming value is
4527 deemed to occur on the edge from the corresponding predecessor block
4528 to the current block (but after any definition of an '<tt>invoke</tt>'
4529 instruction's return value on the same edge).</p>
4533 <p>At runtime, the '<tt>phi</tt>' instruction logically takes on the value
4534 specified by the pair corresponding to the predecessor basic block that executed
4535 just prior to the current block.</p>
4539 Loop: ; Infinite loop that counts from 0 on up...
4540 %indvar = phi i32 [ 0, %LoopHeader ], [ %nextindvar, %Loop ]
4541 %nextindvar = add i32 %indvar, 1
4546 <!-- _______________________________________________________________________ -->
4547 <div class="doc_subsubsection">
4548 <a name="i_select">'<tt>select</tt>' Instruction</a>
4551 <div class="doc_text">
4556 <result> = select <i>selty</i> <cond>, <ty> <val1>, <ty> <val2> <i>; yields ty</i>
4558 <i>selty</i> is either i1 or {<N x i1>}
4564 The '<tt>select</tt>' instruction is used to choose one value based on a
4565 condition, without branching.
4572 The '<tt>select</tt>' instruction requires an 'i1' value or
4573 a vector of 'i1' values indicating the
4574 condition, and two values of the same <a href="#t_firstclass">first class</a>
4575 type. If the val1/val2 are vectors and
4576 the condition is a scalar, then entire vectors are selected, not
4577 individual elements.
4583 If the condition is an i1 and it evaluates to 1, the instruction returns the first
4584 value argument; otherwise, it returns the second value argument.
4587 If the condition is a vector of i1, then the value arguments must
4588 be vectors of the same size, and the selection is done element
4595 %X = select i1 true, i8 17, i8 42 <i>; yields i8:17</i>
4598 <p>Note that the code generator does not yet support conditions
4599 with vector type.</p>
4604 <!-- _______________________________________________________________________ -->
4605 <div class="doc_subsubsection">
4606 <a name="i_call">'<tt>call</tt>' Instruction</a>
4609 <div class="doc_text">
4613 <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>]
4618 <p>The '<tt>call</tt>' instruction represents a simple function call.</p>
4622 <p>This instruction requires several arguments:</p>
4626 <p>The optional "tail" marker indicates whether the callee function accesses
4627 any allocas or varargs in the caller. If the "tail" marker is present, the
4628 function call is eligible for tail call optimization. Note that calls may
4629 be marked "tail" even if they do not occur before a <a
4630 href="#i_ret"><tt>ret</tt></a> instruction.</p>
4633 <p>The optional "cconv" marker indicates which <a href="#callingconv">calling
4634 convention</a> the call should use. If none is specified, the call defaults
4635 to using C calling conventions.</p>
4639 <p>The optional <a href="#paramattrs">Parameter Attributes</a> list for
4640 return values. Only '<tt>zeroext</tt>', '<tt>signext</tt>',
4641 and '<tt>inreg</tt>' attributes are valid here.</p>
4645 <p>'<tt>ty</tt>': the type of the call instruction itself which is also
4646 the type of the return value. Functions that return no value are marked
4647 <tt><a href="#t_void">void</a></tt>.</p>
4650 <p>'<tt>fnty</tt>': shall be the signature of the pointer to function
4651 value being invoked. The argument types must match the types implied by
4652 this signature. This type can be omitted if the function is not varargs
4653 and if the function type does not return a pointer to a function.</p>
4656 <p>'<tt>fnptrval</tt>': An LLVM value containing a pointer to a function to
4657 be invoked. In most cases, this is a direct function invocation, but
4658 indirect <tt>call</tt>s are just as possible, calling an arbitrary pointer
4659 to function value.</p>
4662 <p>'<tt>function args</tt>': argument list whose types match the
4663 function signature argument types. All arguments must be of
4664 <a href="#t_firstclass">first class</a> type. If the function signature
4665 indicates the function accepts a variable number of arguments, the extra
4666 arguments can be specified.</p>
4669 <p>The optional <a href="#fnattrs">function attributes</a> list. Only
4670 '<tt>noreturn</tt>', '<tt>nounwind</tt>', '<tt>readonly</tt>' and
4671 '<tt>readnone</tt>' attributes are valid here.</p>
4677 <p>The '<tt>call</tt>' instruction is used to cause control flow to
4678 transfer to a specified function, with its incoming arguments bound to
4679 the specified values. Upon a '<tt><a href="#i_ret">ret</a></tt>'
4680 instruction in the called function, control flow continues with the
4681 instruction after the function call, and the return value of the
4682 function is bound to the result argument.</p>
4687 %retval = call i32 @test(i32 %argc)
4688 call i32 (i8 *, ...)* @printf(i8 * %msg, i32 12, i8 42) <i>; yields i32</i>
4689 %X = tail call i32 @foo() <i>; yields i32</i>
4690 %Y = tail call <a href="#callingconv">fastcc</a> i32 @foo() <i>; yields i32</i>
4691 call void %foo(i8 97 signext)
4693 %struct.A = type { i32, i8 }
4694 %r = call %struct.A @foo() <i>; yields { 32, i8 }</i>
4695 %gr = extractvalue %struct.A %r, 0 <i>; yields i32</i>
4696 %gr1 = extractvalue %struct.A %r, 1 <i>; yields i8</i>
4697 %Z = call void @foo() noreturn <i>; indicates that %foo never returns normally</i>
4698 %ZZ = call zeroext i32 @bar() <i>; Return value is %zero extended</i>
4703 <!-- _______________________________________________________________________ -->
4704 <div class="doc_subsubsection">
4705 <a name="i_va_arg">'<tt>va_arg</tt>' Instruction</a>
4708 <div class="doc_text">
4713 <resultval> = va_arg <va_list*> <arglist>, <argty>
4718 <p>The '<tt>va_arg</tt>' instruction is used to access arguments passed through
4719 the "variable argument" area of a function call. It is used to implement the
4720 <tt>va_arg</tt> macro in C.</p>
4724 <p>This instruction takes a <tt>va_list*</tt> value and the type of
4725 the argument. It returns a value of the specified argument type and
4726 increments the <tt>va_list</tt> to point to the next argument. The
4727 actual type of <tt>va_list</tt> is target specific.</p>
4731 <p>The '<tt>va_arg</tt>' instruction loads an argument of the specified
4732 type from the specified <tt>va_list</tt> and causes the
4733 <tt>va_list</tt> to point to the next argument. For more information,
4734 see the variable argument handling <a href="#int_varargs">Intrinsic
4737 <p>It is legal for this instruction to be called in a function which does not
4738 take a variable number of arguments, for example, the <tt>vfprintf</tt>
4741 <p><tt>va_arg</tt> is an LLVM instruction instead of an <a
4742 href="#intrinsics">intrinsic function</a> because it takes a type as an
4747 <p>See the <a href="#int_varargs">variable argument processing</a> section.</p>
4749 <p>Note that the code generator does not yet fully support va_arg
4750 on many targets. Also, it does not currently support va_arg with
4751 aggregate types on any target.</p>
4755 <!-- *********************************************************************** -->
4756 <div class="doc_section"> <a name="intrinsics">Intrinsic Functions</a> </div>
4757 <!-- *********************************************************************** -->
4759 <div class="doc_text">
4761 <p>LLVM supports the notion of an "intrinsic function". These functions have
4762 well known names and semantics and are required to follow certain restrictions.
4763 Overall, these intrinsics represent an extension mechanism for the LLVM
4764 language that does not require changing all of the transformations in LLVM when
4765 adding to the language (or the bitcode reader/writer, the parser, etc...).</p>
4767 <p>Intrinsic function names must all start with an "<tt>llvm.</tt>" prefix. This
4768 prefix is reserved in LLVM for intrinsic names; thus, function names may not
4769 begin with this prefix. Intrinsic functions must always be external functions:
4770 you cannot define the body of intrinsic functions. Intrinsic functions may
4771 only be used in call or invoke instructions: it is illegal to take the address
4772 of an intrinsic function. Additionally, because intrinsic functions are part
4773 of the LLVM language, it is required if any are added that they be documented
4776 <p>Some intrinsic functions can be overloaded, i.e., the intrinsic represents
4777 a family of functions that perform the same operation but on different data
4778 types. Because LLVM can represent over 8 million different integer types,
4779 overloading is used commonly to allow an intrinsic function to operate on any
4780 integer type. One or more of the argument types or the result type can be
4781 overloaded to accept any integer type. Argument types may also be defined as
4782 exactly matching a previous argument's type or the result type. This allows an
4783 intrinsic function which accepts multiple arguments, but needs all of them to
4784 be of the same type, to only be overloaded with respect to a single argument or
4787 <p>Overloaded intrinsics will have the names of its overloaded argument types
4788 encoded into its function name, each preceded by a period. Only those types
4789 which are overloaded result in a name suffix. Arguments whose type is matched
4790 against another type do not. For example, the <tt>llvm.ctpop</tt> function can
4791 take an integer of any width and returns an integer of exactly the same integer
4792 width. This leads to a family of functions such as
4793 <tt>i8 @llvm.ctpop.i8(i8 %val)</tt> and <tt>i29 @llvm.ctpop.i29(i29 %val)</tt>.
4794 Only one type, the return type, is overloaded, and only one type suffix is
4795 required. Because the argument's type is matched against the return type, it
4796 does not require its own name suffix.</p>
4798 <p>To learn how to add an intrinsic function, please see the
4799 <a href="ExtendingLLVM.html">Extending LLVM Guide</a>.
4804 <!-- ======================================================================= -->
4805 <div class="doc_subsection">
4806 <a name="int_varargs">Variable Argument Handling Intrinsics</a>
4809 <div class="doc_text">
4811 <p>Variable argument support is defined in LLVM with the <a
4812 href="#i_va_arg"><tt>va_arg</tt></a> instruction and these three
4813 intrinsic functions. These functions are related to the similarly
4814 named macros defined in the <tt><stdarg.h></tt> header file.</p>
4816 <p>All of these functions operate on arguments that use a
4817 target-specific value type "<tt>va_list</tt>". The LLVM assembly
4818 language reference manual does not define what this type is, so all
4819 transformations should be prepared to handle these functions regardless of
4822 <p>This example shows how the <a href="#i_va_arg"><tt>va_arg</tt></a>
4823 instruction and the variable argument handling intrinsic functions are
4826 <div class="doc_code">
4828 define i32 @test(i32 %X, ...) {
4829 ; Initialize variable argument processing
4831 %ap2 = bitcast i8** %ap to i8*
4832 call void @llvm.va_start(i8* %ap2)
4834 ; Read a single integer argument
4835 %tmp = va_arg i8** %ap, i32
4837 ; Demonstrate usage of llvm.va_copy and llvm.va_end
4839 %aq2 = bitcast i8** %aq to i8*
4840 call void @llvm.va_copy(i8* %aq2, i8* %ap2)
4841 call void @llvm.va_end(i8* %aq2)
4843 ; Stop processing of arguments.
4844 call void @llvm.va_end(i8* %ap2)
4848 declare void @llvm.va_start(i8*)
4849 declare void @llvm.va_copy(i8*, i8*)
4850 declare void @llvm.va_end(i8*)
4856 <!-- _______________________________________________________________________ -->
4857 <div class="doc_subsubsection">
4858 <a name="int_va_start">'<tt>llvm.va_start</tt>' Intrinsic</a>
4862 <div class="doc_text">
4864 <pre> declare void %llvm.va_start(i8* <arglist>)<br></pre>
4866 <p>The '<tt>llvm.va_start</tt>' intrinsic initializes
4867 <tt>*<arglist></tt> for subsequent use by <tt><a
4868 href="#i_va_arg">va_arg</a></tt>.</p>
4872 <p>The argument is a pointer to a <tt>va_list</tt> element to initialize.</p>
4876 <p>The '<tt>llvm.va_start</tt>' intrinsic works just like the <tt>va_start</tt>
4877 macro available in C. In a target-dependent way, it initializes the
4878 <tt>va_list</tt> element to which the argument points, so that the next call to
4879 <tt>va_arg</tt> will produce the first variable argument passed to the function.
4880 Unlike the C <tt>va_start</tt> macro, this intrinsic does not need to know the
4881 last argument of the function as the compiler can figure that out.</p>
4885 <!-- _______________________________________________________________________ -->
4886 <div class="doc_subsubsection">
4887 <a name="int_va_end">'<tt>llvm.va_end</tt>' Intrinsic</a>
4890 <div class="doc_text">
4892 <pre> declare void @llvm.va_end(i8* <arglist>)<br></pre>
4895 <p>The '<tt>llvm.va_end</tt>' intrinsic destroys <tt>*<arglist></tt>,
4896 which has been initialized previously with <tt><a href="#int_va_start">llvm.va_start</a></tt>
4897 or <tt><a href="#i_va_copy">llvm.va_copy</a></tt>.</p>
4901 <p>The argument is a pointer to a <tt>va_list</tt> to destroy.</p>
4905 <p>The '<tt>llvm.va_end</tt>' intrinsic works just like the <tt>va_end</tt>
4906 macro available in C. In a target-dependent way, it destroys the
4907 <tt>va_list</tt> element to which the argument points. Calls to <a
4908 href="#int_va_start"><tt>llvm.va_start</tt></a> and <a href="#int_va_copy">
4909 <tt>llvm.va_copy</tt></a> must be matched exactly with calls to
4910 <tt>llvm.va_end</tt>.</p>
4914 <!-- _______________________________________________________________________ -->
4915 <div class="doc_subsubsection">
4916 <a name="int_va_copy">'<tt>llvm.va_copy</tt>' Intrinsic</a>
4919 <div class="doc_text">
4924 declare void @llvm.va_copy(i8* <destarglist>, i8* <srcarglist>)
4929 <p>The '<tt>llvm.va_copy</tt>' intrinsic copies the current argument position
4930 from the source argument list to the destination argument list.</p>
4934 <p>The first argument is a pointer to a <tt>va_list</tt> element to initialize.
4935 The second argument is a pointer to a <tt>va_list</tt> element to copy from.</p>
4940 <p>The '<tt>llvm.va_copy</tt>' intrinsic works just like the <tt>va_copy</tt>
4941 macro available in C. In a target-dependent way, it copies the source
4942 <tt>va_list</tt> element into the destination <tt>va_list</tt> element. This
4943 intrinsic is necessary because the <tt><a href="#int_va_start">
4944 llvm.va_start</a></tt> intrinsic may be arbitrarily complex and require, for
4945 example, memory allocation.</p>
4949 <!-- ======================================================================= -->
4950 <div class="doc_subsection">
4951 <a name="int_gc">Accurate Garbage Collection Intrinsics</a>
4954 <div class="doc_text">
4957 LLVM support for <a href="GarbageCollection.html">Accurate Garbage
4958 Collection</a> (GC) requires the implementation and generation of these
4960 These intrinsics allow identification of <a href="#int_gcroot">GC roots on the
4961 stack</a>, as well as garbage collector implementations that require <a
4962 href="#int_gcread">read</a> and <a href="#int_gcwrite">write</a> barriers.
4963 Front-ends for type-safe garbage collected languages should generate these
4964 intrinsics to make use of the LLVM garbage collectors. For more details, see <a
4965 href="GarbageCollection.html">Accurate Garbage Collection with LLVM</a>.
4968 <p>The garbage collection intrinsics only operate on objects in the generic
4969 address space (address space zero).</p>
4973 <!-- _______________________________________________________________________ -->
4974 <div class="doc_subsubsection">
4975 <a name="int_gcroot">'<tt>llvm.gcroot</tt>' Intrinsic</a>
4978 <div class="doc_text">
4983 declare void @llvm.gcroot(i8** %ptrloc, i8* %metadata)
4988 <p>The '<tt>llvm.gcroot</tt>' intrinsic declares the existence of a GC root to
4989 the code generator, and allows some metadata to be associated with it.</p>
4993 <p>The first argument specifies the address of a stack object that contains the
4994 root pointer. The second pointer (which must be either a constant or a global
4995 value address) contains the meta-data to be associated with the root.</p>
4999 <p>At runtime, a call to this intrinsic stores a null pointer into the "ptrloc"
5000 location. At compile-time, the code generator generates information to allow
5001 the runtime to find the pointer at GC safe points. The '<tt>llvm.gcroot</tt>'
5002 intrinsic may only be used in a function which <a href="#gc">specifies a GC
5008 <!-- _______________________________________________________________________ -->
5009 <div class="doc_subsubsection">
5010 <a name="int_gcread">'<tt>llvm.gcread</tt>' Intrinsic</a>
5013 <div class="doc_text">
5018 declare i8* @llvm.gcread(i8* %ObjPtr, i8** %Ptr)
5023 <p>The '<tt>llvm.gcread</tt>' intrinsic identifies reads of references from heap
5024 locations, allowing garbage collector implementations that require read
5029 <p>The second argument is the address to read from, which should be an address
5030 allocated from the garbage collector. The first object is a pointer to the
5031 start of the referenced object, if needed by the language runtime (otherwise
5036 <p>The '<tt>llvm.gcread</tt>' intrinsic has the same semantics as a load
5037 instruction, but may be replaced with substantially more complex code by the
5038 garbage collector runtime, as needed. The '<tt>llvm.gcread</tt>' intrinsic
5039 may only be used in a function which <a href="#gc">specifies a GC
5045 <!-- _______________________________________________________________________ -->
5046 <div class="doc_subsubsection">
5047 <a name="int_gcwrite">'<tt>llvm.gcwrite</tt>' Intrinsic</a>
5050 <div class="doc_text">
5055 declare void @llvm.gcwrite(i8* %P1, i8* %Obj, i8** %P2)
5060 <p>The '<tt>llvm.gcwrite</tt>' intrinsic identifies writes of references to heap
5061 locations, allowing garbage collector implementations that require write
5062 barriers (such as generational or reference counting collectors).</p>
5066 <p>The first argument is the reference to store, the second is the start of the
5067 object to store it to, and the third is the address of the field of Obj to
5068 store to. If the runtime does not require a pointer to the object, Obj may be
5073 <p>The '<tt>llvm.gcwrite</tt>' intrinsic has the same semantics as a store
5074 instruction, but may be replaced with substantially more complex code by the
5075 garbage collector runtime, as needed. The '<tt>llvm.gcwrite</tt>' intrinsic
5076 may only be used in a function which <a href="#gc">specifies a GC
5083 <!-- ======================================================================= -->
5084 <div class="doc_subsection">
5085 <a name="int_codegen">Code Generator Intrinsics</a>
5088 <div class="doc_text">
5090 These intrinsics are provided by LLVM to expose special features that may only
5091 be implemented with code generator support.
5096 <!-- _______________________________________________________________________ -->
5097 <div class="doc_subsubsection">
5098 <a name="int_returnaddress">'<tt>llvm.returnaddress</tt>' Intrinsic</a>
5101 <div class="doc_text">
5105 declare i8 *@llvm.returnaddress(i32 <level>)
5111 The '<tt>llvm.returnaddress</tt>' intrinsic attempts to compute a
5112 target-specific value indicating the return address of the current function
5113 or one of its callers.
5119 The argument to this intrinsic indicates which function to return the address
5120 for. Zero indicates the calling function, one indicates its caller, etc. The
5121 argument is <b>required</b> to be a constant integer value.
5127 The '<tt>llvm.returnaddress</tt>' intrinsic either returns a pointer indicating
5128 the return address of the specified call frame, or zero if it cannot be
5129 identified. The value returned by this intrinsic is likely to be incorrect or 0
5130 for arguments other than zero, so it should only be used for debugging purposes.
5134 Note that calling this intrinsic does not prevent function inlining or other
5135 aggressive transformations, so the value returned may not be that of the obvious
5136 source-language caller.
5141 <!-- _______________________________________________________________________ -->
5142 <div class="doc_subsubsection">
5143 <a name="int_frameaddress">'<tt>llvm.frameaddress</tt>' Intrinsic</a>
5146 <div class="doc_text">
5150 declare i8 *@llvm.frameaddress(i32 <level>)
5156 The '<tt>llvm.frameaddress</tt>' intrinsic attempts to return the
5157 target-specific frame pointer value for the specified stack frame.
5163 The argument to this intrinsic indicates which function to return the frame
5164 pointer for. Zero indicates the calling function, one indicates its caller,
5165 etc. The argument is <b>required</b> to be a constant integer value.
5171 The '<tt>llvm.frameaddress</tt>' intrinsic either returns a pointer indicating
5172 the frame address of the specified call frame, or zero if it cannot be
5173 identified. The value returned by this intrinsic is likely to be incorrect or 0
5174 for arguments other than zero, so it should only be used for debugging purposes.
5178 Note that calling this intrinsic does not prevent function inlining or other
5179 aggressive transformations, so the value returned may not be that of the obvious
5180 source-language caller.
5184 <!-- _______________________________________________________________________ -->
5185 <div class="doc_subsubsection">
5186 <a name="int_stacksave">'<tt>llvm.stacksave</tt>' Intrinsic</a>
5189 <div class="doc_text">
5193 declare i8 *@llvm.stacksave()
5199 The '<tt>llvm.stacksave</tt>' intrinsic is used to remember the current state of
5200 the function stack, for use with <a href="#int_stackrestore">
5201 <tt>llvm.stackrestore</tt></a>. This is useful for implementing language
5202 features like scoped automatic variable sized arrays in C99.
5208 This intrinsic returns a opaque pointer value that can be passed to <a
5209 href="#int_stackrestore"><tt>llvm.stackrestore</tt></a>. When an
5210 <tt>llvm.stackrestore</tt> intrinsic is executed with a value saved from
5211 <tt>llvm.stacksave</tt>, it effectively restores the state of the stack to the
5212 state it was in when the <tt>llvm.stacksave</tt> intrinsic executed. In
5213 practice, this pops any <a href="#i_alloca">alloca</a> blocks from the stack
5214 that were allocated after the <tt>llvm.stacksave</tt> was executed.
5219 <!-- _______________________________________________________________________ -->
5220 <div class="doc_subsubsection">
5221 <a name="int_stackrestore">'<tt>llvm.stackrestore</tt>' Intrinsic</a>
5224 <div class="doc_text">
5228 declare void @llvm.stackrestore(i8 * %ptr)
5234 The '<tt>llvm.stackrestore</tt>' intrinsic is used to restore the state of
5235 the function stack to the state it was in when the corresponding <a
5236 href="#int_stacksave"><tt>llvm.stacksave</tt></a> intrinsic executed. This is
5237 useful for implementing language features like scoped automatic variable sized
5244 See the description for <a href="#int_stacksave"><tt>llvm.stacksave</tt></a>.
5250 <!-- _______________________________________________________________________ -->
5251 <div class="doc_subsubsection">
5252 <a name="int_prefetch">'<tt>llvm.prefetch</tt>' Intrinsic</a>
5255 <div class="doc_text">
5259 declare void @llvm.prefetch(i8* <address>, i32 <rw>, i32 <locality>)
5266 The '<tt>llvm.prefetch</tt>' intrinsic is a hint to the code generator to insert
5267 a prefetch instruction if supported; otherwise, it is a noop. Prefetches have
5269 effect on the behavior of the program but can change its performance
5276 <tt>address</tt> is the address to be prefetched, <tt>rw</tt> is the specifier
5277 determining if the fetch should be for a read (0) or write (1), and
5278 <tt>locality</tt> is a temporal locality specifier ranging from (0) - no
5279 locality, to (3) - extremely local keep in cache. The <tt>rw</tt> and
5280 <tt>locality</tt> arguments must be constant integers.
5286 This intrinsic does not modify the behavior of the program. In particular,
5287 prefetches cannot trap and do not produce a value. On targets that support this
5288 intrinsic, the prefetch can provide hints to the processor cache for better
5294 <!-- _______________________________________________________________________ -->
5295 <div class="doc_subsubsection">
5296 <a name="int_pcmarker">'<tt>llvm.pcmarker</tt>' Intrinsic</a>
5299 <div class="doc_text">
5303 declare void @llvm.pcmarker(i32 <id>)
5310 The '<tt>llvm.pcmarker</tt>' intrinsic is a method to export a Program Counter
5312 code to simulators and other tools. The method is target specific, but it is
5313 expected that the marker will use exported symbols to transmit the PC of the
5315 The marker makes no guarantees that it will remain with any specific instruction
5316 after optimizations. It is possible that the presence of a marker will inhibit
5317 optimizations. The intended use is to be inserted after optimizations to allow
5318 correlations of simulation runs.
5324 <tt>id</tt> is a numerical id identifying the marker.
5330 This intrinsic does not modify the behavior of the program. Backends that do not
5331 support this intrinisic may ignore it.
5336 <!-- _______________________________________________________________________ -->
5337 <div class="doc_subsubsection">
5338 <a name="int_readcyclecounter">'<tt>llvm.readcyclecounter</tt>' Intrinsic</a>
5341 <div class="doc_text">
5345 declare i64 @llvm.readcyclecounter( )
5352 The '<tt>llvm.readcyclecounter</tt>' intrinsic provides access to the cycle
5353 counter register (or similar low latency, high accuracy clocks) on those targets
5354 that support it. On X86, it should map to RDTSC. On Alpha, it should map to RPCC.
5355 As the backing counters overflow quickly (on the order of 9 seconds on alpha), this
5356 should only be used for small timings.
5362 When directly supported, reading the cycle counter should not modify any memory.
5363 Implementations are allowed to either return a application specific value or a
5364 system wide value. On backends without support, this is lowered to a constant 0.
5369 <!-- ======================================================================= -->
5370 <div class="doc_subsection">
5371 <a name="int_libc">Standard C Library Intrinsics</a>
5374 <div class="doc_text">
5376 LLVM provides intrinsics for a few important standard C library functions.
5377 These intrinsics allow source-language front-ends to pass information about the
5378 alignment of the pointer arguments to the code generator, providing opportunity
5379 for more efficient code generation.
5384 <!-- _______________________________________________________________________ -->
5385 <div class="doc_subsubsection">
5386 <a name="int_memcpy">'<tt>llvm.memcpy</tt>' Intrinsic</a>
5389 <div class="doc_text">
5392 <p>This is an overloaded intrinsic. You can use llvm.memcpy on any integer bit
5393 width. Not all targets support all bit widths however.</p>
5395 declare void @llvm.memcpy.i8(i8 * <dest>, i8 * <src>,
5396 i8 <len>, i32 <align>)
5397 declare void @llvm.memcpy.i16(i8 * <dest>, i8 * <src>,
5398 i16 <len>, i32 <align>)
5399 declare void @llvm.memcpy.i32(i8 * <dest>, i8 * <src>,
5400 i32 <len>, i32 <align>)
5401 declare void @llvm.memcpy.i64(i8 * <dest>, i8 * <src>,
5402 i64 <len>, i32 <align>)
5408 The '<tt>llvm.memcpy.*</tt>' intrinsics copy a block of memory from the source
5409 location to the destination location.
5413 Note that, unlike the standard libc function, the <tt>llvm.memcpy.*</tt>
5414 intrinsics do not return a value, and takes an extra alignment argument.
5420 The first argument is a pointer to the destination, the second is a pointer to
5421 the source. The third argument is an integer argument
5422 specifying the number of bytes to copy, and the fourth argument is the alignment
5423 of the source and destination locations.
5427 If the call to this intrinisic has an alignment value that is not 0 or 1, then
5428 the caller guarantees that both the source and destination pointers are aligned
5435 The '<tt>llvm.memcpy.*</tt>' intrinsics copy a block of memory from the source
5436 location to the destination location, which are not allowed to overlap. It
5437 copies "len" bytes of memory over. If the argument is known to be aligned to
5438 some boundary, this can be specified as the fourth argument, otherwise it should
5444 <!-- _______________________________________________________________________ -->
5445 <div class="doc_subsubsection">
5446 <a name="int_memmove">'<tt>llvm.memmove</tt>' Intrinsic</a>
5449 <div class="doc_text">
5452 <p>This is an overloaded intrinsic. You can use llvm.memmove on any integer bit
5453 width. Not all targets support all bit widths however.</p>
5455 declare void @llvm.memmove.i8(i8 * <dest>, i8 * <src>,
5456 i8 <len>, i32 <align>)
5457 declare void @llvm.memmove.i16(i8 * <dest>, i8 * <src>,
5458 i16 <len>, i32 <align>)
5459 declare void @llvm.memmove.i32(i8 * <dest>, i8 * <src>,
5460 i32 <len>, i32 <align>)
5461 declare void @llvm.memmove.i64(i8 * <dest>, i8 * <src>,
5462 i64 <len>, i32 <align>)
5468 The '<tt>llvm.memmove.*</tt>' intrinsics move a block of memory from the source
5469 location to the destination location. It is similar to the
5470 '<tt>llvm.memcpy</tt>' intrinsic but allows the two memory locations to overlap.
5474 Note that, unlike the standard libc function, the <tt>llvm.memmove.*</tt>
5475 intrinsics do not return a value, and takes an extra alignment argument.
5481 The first argument is a pointer to the destination, the second is a pointer to
5482 the source. The third argument is an integer argument
5483 specifying the number of bytes to copy, and the fourth argument is the alignment
5484 of the source and destination locations.
5488 If the call to this intrinisic has an alignment value that is not 0 or 1, then
5489 the caller guarantees that the source and destination pointers are aligned to
5496 The '<tt>llvm.memmove.*</tt>' intrinsics copy a block of memory from the source
5497 location to the destination location, which may overlap. It
5498 copies "len" bytes of memory over. If the argument is known to be aligned to
5499 some boundary, this can be specified as the fourth argument, otherwise it should
5505 <!-- _______________________________________________________________________ -->
5506 <div class="doc_subsubsection">
5507 <a name="int_memset">'<tt>llvm.memset.*</tt>' Intrinsics</a>
5510 <div class="doc_text">
5513 <p>This is an overloaded intrinsic. You can use llvm.memset on any integer bit
5514 width. Not all targets support all bit widths however.</p>
5516 declare void @llvm.memset.i8(i8 * <dest>, i8 <val>,
5517 i8 <len>, i32 <align>)
5518 declare void @llvm.memset.i16(i8 * <dest>, i8 <val>,
5519 i16 <len>, i32 <align>)
5520 declare void @llvm.memset.i32(i8 * <dest>, i8 <val>,
5521 i32 <len>, i32 <align>)
5522 declare void @llvm.memset.i64(i8 * <dest>, i8 <val>,
5523 i64 <len>, i32 <align>)
5529 The '<tt>llvm.memset.*</tt>' intrinsics fill a block of memory with a particular
5534 Note that, unlike the standard libc function, the <tt>llvm.memset</tt> intrinsic
5535 does not return a value, and takes an extra alignment argument.
5541 The first argument is a pointer to the destination to fill, the second is the
5542 byte value to fill it with, the third argument is an integer
5543 argument specifying the number of bytes to fill, and the fourth argument is the
5544 known alignment of destination location.
5548 If the call to this intrinisic has an alignment value that is not 0 or 1, then
5549 the caller guarantees that the destination pointer is aligned to that boundary.
5555 The '<tt>llvm.memset.*</tt>' intrinsics fill "len" bytes of memory starting at
5557 destination location. If the argument is known to be aligned to some boundary,
5558 this can be specified as the fourth argument, otherwise it should be set to 0 or
5564 <!-- _______________________________________________________________________ -->
5565 <div class="doc_subsubsection">
5566 <a name="int_sqrt">'<tt>llvm.sqrt.*</tt>' Intrinsic</a>
5569 <div class="doc_text">
5572 <p>This is an overloaded intrinsic. You can use <tt>llvm.sqrt</tt> on any
5573 floating point or vector of floating point type. Not all targets support all
5576 declare float @llvm.sqrt.f32(float %Val)
5577 declare double @llvm.sqrt.f64(double %Val)
5578 declare x86_fp80 @llvm.sqrt.f80(x86_fp80 %Val)
5579 declare fp128 @llvm.sqrt.f128(fp128 %Val)
5580 declare ppc_fp128 @llvm.sqrt.ppcf128(ppc_fp128 %Val)
5586 The '<tt>llvm.sqrt</tt>' intrinsics return the sqrt of the specified operand,
5587 returning the same value as the libm '<tt>sqrt</tt>' functions would. Unlike
5588 <tt>sqrt</tt> in libm, however, <tt>llvm.sqrt</tt> has undefined behavior for
5589 negative numbers other than -0.0 (which allows for better optimization, because
5590 there is no need to worry about errno being set). <tt>llvm.sqrt(-0.0)</tt> is
5591 defined to return -0.0 like IEEE sqrt.
5597 The argument and return value are floating point numbers of the same type.
5603 This function returns the sqrt of the specified operand if it is a nonnegative
5604 floating point number.
5608 <!-- _______________________________________________________________________ -->
5609 <div class="doc_subsubsection">
5610 <a name="int_powi">'<tt>llvm.powi.*</tt>' Intrinsic</a>
5613 <div class="doc_text">
5616 <p>This is an overloaded intrinsic. You can use <tt>llvm.powi</tt> on any
5617 floating point or vector of floating point type. Not all targets support all
5620 declare float @llvm.powi.f32(float %Val, i32 %power)
5621 declare double @llvm.powi.f64(double %Val, i32 %power)
5622 declare x86_fp80 @llvm.powi.f80(x86_fp80 %Val, i32 %power)
5623 declare fp128 @llvm.powi.f128(fp128 %Val, i32 %power)
5624 declare ppc_fp128 @llvm.powi.ppcf128(ppc_fp128 %Val, i32 %power)
5630 The '<tt>llvm.powi.*</tt>' intrinsics return the first operand raised to the
5631 specified (positive or negative) power. The order of evaluation of
5632 multiplications is not defined. When a vector of floating point type is
5633 used, the second argument remains a scalar integer value.
5639 The second argument is an integer power, and the first is a value to raise to
5646 This function returns the first value raised to the second power with an
5647 unspecified sequence of rounding operations.</p>
5650 <!-- _______________________________________________________________________ -->
5651 <div class="doc_subsubsection">
5652 <a name="int_sin">'<tt>llvm.sin.*</tt>' Intrinsic</a>
5655 <div class="doc_text">
5658 <p>This is an overloaded intrinsic. You can use <tt>llvm.sin</tt> on any
5659 floating point or vector of floating point type. Not all targets support all
5662 declare float @llvm.sin.f32(float %Val)
5663 declare double @llvm.sin.f64(double %Val)
5664 declare x86_fp80 @llvm.sin.f80(x86_fp80 %Val)
5665 declare fp128 @llvm.sin.f128(fp128 %Val)
5666 declare ppc_fp128 @llvm.sin.ppcf128(ppc_fp128 %Val)
5672 The '<tt>llvm.sin.*</tt>' intrinsics return the sine of the operand.
5678 The argument and return value are floating point numbers of the same type.
5684 This function returns the sine of the specified operand, returning the
5685 same values as the libm <tt>sin</tt> functions would, and handles error
5686 conditions in the same way.</p>
5689 <!-- _______________________________________________________________________ -->
5690 <div class="doc_subsubsection">
5691 <a name="int_cos">'<tt>llvm.cos.*</tt>' Intrinsic</a>
5694 <div class="doc_text">
5697 <p>This is an overloaded intrinsic. You can use <tt>llvm.cos</tt> on any
5698 floating point or vector of floating point type. Not all targets support all
5701 declare float @llvm.cos.f32(float %Val)
5702 declare double @llvm.cos.f64(double %Val)
5703 declare x86_fp80 @llvm.cos.f80(x86_fp80 %Val)
5704 declare fp128 @llvm.cos.f128(fp128 %Val)
5705 declare ppc_fp128 @llvm.cos.ppcf128(ppc_fp128 %Val)
5711 The '<tt>llvm.cos.*</tt>' intrinsics return the cosine of the operand.
5717 The argument and return value are floating point numbers of the same type.
5723 This function returns the cosine of the specified operand, returning the
5724 same values as the libm <tt>cos</tt> functions would, and handles error
5725 conditions in the same way.</p>
5728 <!-- _______________________________________________________________________ -->
5729 <div class="doc_subsubsection">
5730 <a name="int_pow">'<tt>llvm.pow.*</tt>' Intrinsic</a>
5733 <div class="doc_text">
5736 <p>This is an overloaded intrinsic. You can use <tt>llvm.pow</tt> on any
5737 floating point or vector of floating point type. Not all targets support all
5740 declare float @llvm.pow.f32(float %Val, float %Power)
5741 declare double @llvm.pow.f64(double %Val, double %Power)
5742 declare x86_fp80 @llvm.pow.f80(x86_fp80 %Val, x86_fp80 %Power)
5743 declare fp128 @llvm.pow.f128(fp128 %Val, fp128 %Power)
5744 declare ppc_fp128 @llvm.pow.ppcf128(ppc_fp128 %Val, ppc_fp128 Power)
5750 The '<tt>llvm.pow.*</tt>' intrinsics return the first operand raised to the
5751 specified (positive or negative) power.
5757 The second argument is a floating point power, and the first is a value to
5758 raise to that power.
5764 This function returns the first value raised to the second power,
5766 same values as the libm <tt>pow</tt> functions would, and handles error
5767 conditions in the same way.</p>
5771 <!-- ======================================================================= -->
5772 <div class="doc_subsection">
5773 <a name="int_manip">Bit Manipulation Intrinsics</a>
5776 <div class="doc_text">
5778 LLVM provides intrinsics for a few important bit manipulation operations.
5779 These allow efficient code generation for some algorithms.
5784 <!-- _______________________________________________________________________ -->
5785 <div class="doc_subsubsection">
5786 <a name="int_bswap">'<tt>llvm.bswap.*</tt>' Intrinsics</a>
5789 <div class="doc_text">
5792 <p>This is an overloaded intrinsic function. You can use bswap on any integer
5793 type that is an even number of bytes (i.e. BitWidth % 16 == 0).</p>
5795 declare i16 @llvm.bswap.i16(i16 <id>)
5796 declare i32 @llvm.bswap.i32(i32 <id>)
5797 declare i64 @llvm.bswap.i64(i64 <id>)
5803 The '<tt>llvm.bswap</tt>' family of intrinsics is used to byte swap integer
5804 values with an even number of bytes (positive multiple of 16 bits). These are
5805 useful for performing operations on data that is not in the target's native
5812 The <tt>llvm.bswap.i16</tt> intrinsic returns an i16 value that has the high
5813 and low byte of the input i16 swapped. Similarly, the <tt>llvm.bswap.i32</tt>
5814 intrinsic returns an i32 value that has the four bytes of the input i32
5815 swapped, so that if the input bytes are numbered 0, 1, 2, 3 then the returned
5816 i32 will have its bytes in 3, 2, 1, 0 order. The <tt>llvm.bswap.i48</tt>,
5817 <tt>llvm.bswap.i64</tt> and other intrinsics extend this concept to
5818 additional even-byte lengths (6 bytes, 8 bytes and more, respectively).
5823 <!-- _______________________________________________________________________ -->
5824 <div class="doc_subsubsection">
5825 <a name="int_ctpop">'<tt>llvm.ctpop.*</tt>' Intrinsic</a>
5828 <div class="doc_text">
5831 <p>This is an overloaded intrinsic. You can use llvm.ctpop on any integer bit
5832 width. Not all targets support all bit widths however.</p>
5834 declare i8 @llvm.ctpop.i8(i8 <src>)
5835 declare i16 @llvm.ctpop.i16(i16 <src>)
5836 declare i32 @llvm.ctpop.i32(i32 <src>)
5837 declare i64 @llvm.ctpop.i64(i64 <src>)
5838 declare i256 @llvm.ctpop.i256(i256 <src>)
5844 The '<tt>llvm.ctpop</tt>' family of intrinsics counts the number of bits set in a
5851 The only argument is the value to be counted. The argument may be of any
5852 integer type. The return type must match the argument type.
5858 The '<tt>llvm.ctpop</tt>' intrinsic counts the 1's in a variable.
5862 <!-- _______________________________________________________________________ -->
5863 <div class="doc_subsubsection">
5864 <a name="int_ctlz">'<tt>llvm.ctlz.*</tt>' Intrinsic</a>
5867 <div class="doc_text">
5870 <p>This is an overloaded intrinsic. You can use <tt>llvm.ctlz</tt> on any
5871 integer bit width. Not all targets support all bit widths however.</p>
5873 declare i8 @llvm.ctlz.i8 (i8 <src>)
5874 declare i16 @llvm.ctlz.i16(i16 <src>)
5875 declare i32 @llvm.ctlz.i32(i32 <src>)
5876 declare i64 @llvm.ctlz.i64(i64 <src>)
5877 declare i256 @llvm.ctlz.i256(i256 <src>)
5883 The '<tt>llvm.ctlz</tt>' family of intrinsic functions counts the number of
5884 leading zeros in a variable.
5890 The only argument is the value to be counted. The argument may be of any
5891 integer type. The return type must match the argument type.
5897 The '<tt>llvm.ctlz</tt>' intrinsic counts the leading (most significant) zeros
5898 in a variable. If the src == 0 then the result is the size in bits of the type
5899 of src. For example, <tt>llvm.ctlz(i32 2) = 30</tt>.
5905 <!-- _______________________________________________________________________ -->
5906 <div class="doc_subsubsection">
5907 <a name="int_cttz">'<tt>llvm.cttz.*</tt>' Intrinsic</a>
5910 <div class="doc_text">
5913 <p>This is an overloaded intrinsic. You can use <tt>llvm.cttz</tt> on any
5914 integer bit width. Not all targets support all bit widths however.</p>
5916 declare i8 @llvm.cttz.i8 (i8 <src>)
5917 declare i16 @llvm.cttz.i16(i16 <src>)
5918 declare i32 @llvm.cttz.i32(i32 <src>)
5919 declare i64 @llvm.cttz.i64(i64 <src>)
5920 declare i256 @llvm.cttz.i256(i256 <src>)
5926 The '<tt>llvm.cttz</tt>' family of intrinsic functions counts the number of
5933 The only argument is the value to be counted. The argument may be of any
5934 integer type. The return type must match the argument type.
5940 The '<tt>llvm.cttz</tt>' intrinsic counts the trailing (least significant) zeros
5941 in a variable. If the src == 0 then the result is the size in bits of the type
5942 of src. For example, <tt>llvm.cttz(2) = 1</tt>.
5947 <!-- ======================================================================= -->
5948 <div class="doc_subsection">
5949 <a name="int_overflow">Arithmetic with Overflow Intrinsics</a>
5952 <div class="doc_text">
5954 LLVM provides intrinsics for some arithmetic with overflow operations.
5959 <!-- _______________________________________________________________________ -->
5960 <div class="doc_subsubsection">
5961 <a name="int_sadd_overflow">'<tt>llvm.sadd.with.overflow.*</tt>' Intrinsics</a>
5964 <div class="doc_text">
5968 <p>This is an overloaded intrinsic. You can use <tt>llvm.sadd.with.overflow</tt>
5969 on any integer bit width.</p>
5972 declare {i16, i1} @llvm.sadd.with.overflow.i16(i16 %a, i16 %b)
5973 declare {i32, i1} @llvm.sadd.with.overflow.i32(i32 %a, i32 %b)
5974 declare {i64, i1} @llvm.sadd.with.overflow.i64(i64 %a, i64 %b)
5979 <p>The '<tt>llvm.sadd.with.overflow</tt>' family of intrinsic functions perform
5980 a signed addition of the two arguments, and indicate whether an overflow
5981 occurred during the signed summation.</p>
5985 <p>The arguments (%a and %b) and the first element of the result structure may
5986 be of integer types of any bit width, but they must have the same bit width. The
5987 second element of the result structure must be of type <tt>i1</tt>. <tt>%a</tt>
5988 and <tt>%b</tt> are the two values that will undergo signed addition.</p>
5992 <p>The '<tt>llvm.sadd.with.overflow</tt>' family of intrinsic functions perform
5993 a signed addition of the two variables. They return a structure — the
5994 first element of which is the signed summation, and the second element of which
5995 is a bit specifying if the signed summation resulted in an overflow.</p>
5999 %res = call {i32, i1} @llvm.sadd.with.overflow.i32(i32 %a, i32 %b)
6000 %sum = extractvalue {i32, i1} %res, 0
6001 %obit = extractvalue {i32, i1} %res, 1
6002 br i1 %obit, label %overflow, label %normal
6007 <!-- _______________________________________________________________________ -->
6008 <div class="doc_subsubsection">
6009 <a name="int_uadd_overflow">'<tt>llvm.uadd.with.overflow.*</tt>' Intrinsics</a>
6012 <div class="doc_text">
6016 <p>This is an overloaded intrinsic. You can use <tt>llvm.uadd.with.overflow</tt>
6017 on any integer bit width.</p>
6020 declare {i16, i1} @llvm.uadd.with.overflow.i16(i16 %a, i16 %b)
6021 declare {i32, i1} @llvm.uadd.with.overflow.i32(i32 %a, i32 %b)
6022 declare {i64, i1} @llvm.uadd.with.overflow.i64(i64 %a, i64 %b)
6027 <p>The '<tt>llvm.uadd.with.overflow</tt>' family of intrinsic functions perform
6028 an unsigned addition of the two arguments, and indicate whether a carry occurred
6029 during the unsigned summation.</p>
6033 <p>The arguments (%a and %b) and the first element of the result structure may
6034 be of integer types of any bit width, but they must have the same bit width. The
6035 second element of the result structure must be of type <tt>i1</tt>. <tt>%a</tt>
6036 and <tt>%b</tt> are the two values that will undergo unsigned addition.</p>
6040 <p>The '<tt>llvm.uadd.with.overflow</tt>' family of intrinsic functions perform
6041 an unsigned addition of the two arguments. They return a structure — the
6042 first element of which is the sum, and the second element of which is a bit
6043 specifying if the unsigned summation resulted in a carry.</p>
6047 %res = call {i32, i1} @llvm.uadd.with.overflow.i32(i32 %a, i32 %b)
6048 %sum = extractvalue {i32, i1} %res, 0
6049 %obit = extractvalue {i32, i1} %res, 1
6050 br i1 %obit, label %carry, label %normal
6055 <!-- _______________________________________________________________________ -->
6056 <div class="doc_subsubsection">
6057 <a name="int_ssub_overflow">'<tt>llvm.ssub.with.overflow.*</tt>' Intrinsics</a>
6060 <div class="doc_text">
6064 <p>This is an overloaded intrinsic. You can use <tt>llvm.ssub.with.overflow</tt>
6065 on any integer bit width.</p>
6068 declare {i16, i1} @llvm.ssub.with.overflow.i16(i16 %a, i16 %b)
6069 declare {i32, i1} @llvm.ssub.with.overflow.i32(i32 %a, i32 %b)
6070 declare {i64, i1} @llvm.ssub.with.overflow.i64(i64 %a, i64 %b)
6075 <p>The '<tt>llvm.ssub.with.overflow</tt>' family of intrinsic functions perform
6076 a signed subtraction of the two arguments, and indicate whether an overflow
6077 occurred during the signed subtraction.</p>
6081 <p>The arguments (%a and %b) and the first element of the result structure may
6082 be of integer types of any bit width, but they must have the same bit width. The
6083 second element of the result structure must be of type <tt>i1</tt>. <tt>%a</tt>
6084 and <tt>%b</tt> are the two values that will undergo signed subtraction.</p>
6088 <p>The '<tt>llvm.ssub.with.overflow</tt>' family of intrinsic functions perform
6089 a signed subtraction of the two arguments. They return a structure — the
6090 first element of which is the subtraction, and the second element of which is a bit
6091 specifying if the signed subtraction resulted in an overflow.</p>
6095 %res = call {i32, i1} @llvm.ssub.with.overflow.i32(i32 %a, i32 %b)
6096 %sum = extractvalue {i32, i1} %res, 0
6097 %obit = extractvalue {i32, i1} %res, 1
6098 br i1 %obit, label %overflow, label %normal
6103 <!-- _______________________________________________________________________ -->
6104 <div class="doc_subsubsection">
6105 <a name="int_usub_overflow">'<tt>llvm.usub.with.overflow.*</tt>' Intrinsics</a>
6108 <div class="doc_text">
6112 <p>This is an overloaded intrinsic. You can use <tt>llvm.usub.with.overflow</tt>
6113 on any integer bit width.</p>
6116 declare {i16, i1} @llvm.usub.with.overflow.i16(i16 %a, i16 %b)
6117 declare {i32, i1} @llvm.usub.with.overflow.i32(i32 %a, i32 %b)
6118 declare {i64, i1} @llvm.usub.with.overflow.i64(i64 %a, i64 %b)
6123 <p>The '<tt>llvm.usub.with.overflow</tt>' family of intrinsic functions perform
6124 an unsigned subtraction of the two arguments, and indicate whether an overflow
6125 occurred during the unsigned subtraction.</p>
6129 <p>The arguments (%a and %b) and the first element of the result structure may
6130 be of integer types of any bit width, but they must have the same bit width. The
6131 second element of the result structure must be of type <tt>i1</tt>. <tt>%a</tt>
6132 and <tt>%b</tt> are the two values that will undergo unsigned subtraction.</p>
6136 <p>The '<tt>llvm.usub.with.overflow</tt>' family of intrinsic functions perform
6137 an unsigned subtraction of the two arguments. They return a structure — the
6138 first element of which is the subtraction, and the second element of which is a bit
6139 specifying if the unsigned subtraction resulted in an overflow.</p>
6143 %res = call {i32, i1} @llvm.usub.with.overflow.i32(i32 %a, i32 %b)
6144 %sum = extractvalue {i32, i1} %res, 0
6145 %obit = extractvalue {i32, i1} %res, 1
6146 br i1 %obit, label %overflow, label %normal
6151 <!-- _______________________________________________________________________ -->
6152 <div class="doc_subsubsection">
6153 <a name="int_smul_overflow">'<tt>llvm.smul.with.overflow.*</tt>' Intrinsics</a>
6156 <div class="doc_text">
6160 <p>This is an overloaded intrinsic. You can use <tt>llvm.smul.with.overflow</tt>
6161 on any integer bit width.</p>
6164 declare {i16, i1} @llvm.smul.with.overflow.i16(i16 %a, i16 %b)
6165 declare {i32, i1} @llvm.smul.with.overflow.i32(i32 %a, i32 %b)
6166 declare {i64, i1} @llvm.smul.with.overflow.i64(i64 %a, i64 %b)
6171 <p>The '<tt>llvm.smul.with.overflow</tt>' family of intrinsic functions perform
6172 a signed multiplication of the two arguments, and indicate whether an overflow
6173 occurred during the signed multiplication.</p>
6177 <p>The arguments (%a and %b) and the first element of the result structure may
6178 be of integer types of any bit width, but they must have the same bit width. The
6179 second element of the result structure must be of type <tt>i1</tt>. <tt>%a</tt>
6180 and <tt>%b</tt> are the two values that will undergo signed multiplication.</p>
6184 <p>The '<tt>llvm.smul.with.overflow</tt>' family of intrinsic functions perform
6185 a signed multiplication of the two arguments. They return a structure —
6186 the first element of which is the multiplication, and the second element of
6187 which is a bit specifying if the signed multiplication resulted in an
6192 %res = call {i32, i1} @llvm.smul.with.overflow.i32(i32 %a, i32 %b)
6193 %sum = extractvalue {i32, i1} %res, 0
6194 %obit = extractvalue {i32, i1} %res, 1
6195 br i1 %obit, label %overflow, label %normal
6200 <!-- _______________________________________________________________________ -->
6201 <div class="doc_subsubsection">
6202 <a name="int_umul_overflow">'<tt>llvm.umul.with.overflow.*</tt>' Intrinsics</a>
6205 <div class="doc_text">
6209 <p>This is an overloaded intrinsic. You can use <tt>llvm.umul.with.overflow</tt>
6210 on any integer bit width.</p>
6213 declare {i16, i1} @llvm.umul.with.overflow.i16(i16 %a, i16 %b)
6214 declare {i32, i1} @llvm.umul.with.overflow.i32(i32 %a, i32 %b)
6215 declare {i64, i1} @llvm.umul.with.overflow.i64(i64 %a, i64 %b)
6220 <p>The '<tt>llvm.umul.with.overflow</tt>' family of intrinsic functions perform
6221 a unsigned multiplication of the two arguments, and indicate whether an overflow
6222 occurred during the unsigned multiplication.</p>
6226 <p>The arguments (%a and %b) and the first element of the result structure may
6227 be of integer types of any bit width, but they must have the same bit width. The
6228 second element of the result structure must be of type <tt>i1</tt>. <tt>%a</tt>
6229 and <tt>%b</tt> are the two values that will undergo unsigned
6234 <p>The '<tt>llvm.umul.with.overflow</tt>' family of intrinsic functions perform
6235 an unsigned multiplication of the two arguments. They return a structure —
6236 the first element of which is the multiplication, and the second element of
6237 which is a bit specifying if the unsigned multiplication resulted in an
6242 %res = call {i32, i1} @llvm.umul.with.overflow.i32(i32 %a, i32 %b)
6243 %sum = extractvalue {i32, i1} %res, 0
6244 %obit = extractvalue {i32, i1} %res, 1
6245 br i1 %obit, label %overflow, label %normal
6250 <!-- ======================================================================= -->
6251 <div class="doc_subsection">
6252 <a name="int_debugger">Debugger Intrinsics</a>
6255 <div class="doc_text">
6257 The LLVM debugger intrinsics (which all start with <tt>llvm.dbg.</tt> prefix),
6258 are described in the <a
6259 href="SourceLevelDebugging.html#format_common_intrinsics">LLVM Source Level
6260 Debugging</a> document.
6265 <!-- ======================================================================= -->
6266 <div class="doc_subsection">
6267 <a name="int_eh">Exception Handling Intrinsics</a>
6270 <div class="doc_text">
6271 <p> The LLVM exception handling intrinsics (which all start with
6272 <tt>llvm.eh.</tt> prefix), are described in the <a
6273 href="ExceptionHandling.html#format_common_intrinsics">LLVM Exception
6274 Handling</a> document. </p>
6277 <!-- ======================================================================= -->
6278 <div class="doc_subsection">
6279 <a name="int_trampoline">Trampoline Intrinsic</a>
6282 <div class="doc_text">
6284 This intrinsic makes it possible to excise one parameter, marked with
6285 the <tt>nest</tt> attribute, from a function. The result is a callable
6286 function pointer lacking the nest parameter - the caller does not need
6287 to provide a value for it. Instead, the value to use is stored in
6288 advance in a "trampoline", a block of memory usually allocated
6289 on the stack, which also contains code to splice the nest value into the
6290 argument list. This is used to implement the GCC nested function address
6294 For example, if the function is
6295 <tt>i32 f(i8* nest %c, i32 %x, i32 %y)</tt> then the resulting function
6296 pointer has signature <tt>i32 (i32, i32)*</tt>. It can be created as follows:</p>
6298 %tramp = alloca [10 x i8], align 4 ; size and alignment only correct for X86
6299 %tramp1 = getelementptr [10 x i8]* %tramp, i32 0, i32 0
6300 %p = call i8* @llvm.init.trampoline( i8* %tramp1, i8* bitcast (i32 (i8* nest , i32, i32)* @f to i8*), i8* %nval )
6301 %fp = bitcast i8* %p to i32 (i32, i32)*
6303 <p>The call <tt>%val = call i32 %fp( i32 %x, i32 %y )</tt> is then equivalent
6304 to <tt>%val = call i32 %f( i8* %nval, i32 %x, i32 %y )</tt>.</p>
6307 <!-- _______________________________________________________________________ -->
6308 <div class="doc_subsubsection">
6309 <a name="int_it">'<tt>llvm.init.trampoline</tt>' Intrinsic</a>
6311 <div class="doc_text">
6314 declare i8* @llvm.init.trampoline(i8* <tramp>, i8* <func>, i8* <nval>)
6318 This fills the memory pointed to by <tt>tramp</tt> with code
6319 and returns a function pointer suitable for executing it.
6323 The <tt>llvm.init.trampoline</tt> intrinsic takes three arguments, all
6324 pointers. The <tt>tramp</tt> argument must point to a sufficiently large
6325 and sufficiently aligned block of memory; this memory is written to by the
6326 intrinsic. Note that the size and the alignment are target-specific - LLVM
6327 currently provides no portable way of determining them, so a front-end that
6328 generates this intrinsic needs to have some target-specific knowledge.
6329 The <tt>func</tt> argument must hold a function bitcast to an <tt>i8*</tt>.
6333 The block of memory pointed to by <tt>tramp</tt> is filled with target
6334 dependent code, turning it into a function. A pointer to this function is
6335 returned, but needs to be bitcast to an
6336 <a href="#int_trampoline">appropriate function pointer type</a>
6337 before being called. The new function's signature is the same as that of
6338 <tt>func</tt> with any arguments marked with the <tt>nest</tt> attribute
6339 removed. At most one such <tt>nest</tt> argument is allowed, and it must be
6340 of pointer type. Calling the new function is equivalent to calling
6341 <tt>func</tt> with the same argument list, but with <tt>nval</tt> used for the
6342 missing <tt>nest</tt> argument. If, after calling
6343 <tt>llvm.init.trampoline</tt>, the memory pointed to by <tt>tramp</tt> is
6344 modified, then the effect of any later call to the returned function pointer is
6349 <!-- ======================================================================= -->
6350 <div class="doc_subsection">
6351 <a name="int_atomics">Atomic Operations and Synchronization Intrinsics</a>
6354 <div class="doc_text">
6356 These intrinsic functions expand the "universal IR" of LLVM to represent
6357 hardware constructs for atomic operations and memory synchronization. This
6358 provides an interface to the hardware, not an interface to the programmer. It
6359 is aimed at a low enough level to allow any programming models or APIs
6360 (Application Programming Interfaces) which
6361 need atomic behaviors to map cleanly onto it. It is also modeled primarily on
6362 hardware behavior. Just as hardware provides a "universal IR" for source
6363 languages, it also provides a starting point for developing a "universal"
6364 atomic operation and synchronization IR.
6367 These do <em>not</em> form an API such as high-level threading libraries,
6368 software transaction memory systems, atomic primitives, and intrinsic
6369 functions as found in BSD, GNU libc, atomic_ops, APR, and other system and
6370 application libraries. The hardware interface provided by LLVM should allow
6371 a clean implementation of all of these APIs and parallel programming models.
6372 No one model or paradigm should be selected above others unless the hardware
6373 itself ubiquitously does so.
6378 <!-- _______________________________________________________________________ -->
6379 <div class="doc_subsubsection">
6380 <a name="int_memory_barrier">'<tt>llvm.memory.barrier</tt>' Intrinsic</a>
6382 <div class="doc_text">
6385 declare void @llvm.memory.barrier( i1 <ll>, i1 <ls>, i1 <sl>, i1 <ss>,
6391 The <tt>llvm.memory.barrier</tt> intrinsic guarantees ordering between
6392 specific pairs of memory access types.
6396 The <tt>llvm.memory.barrier</tt> intrinsic requires five boolean arguments.
6397 The first four arguments enables a specific barrier as listed below. The fith
6398 argument specifies that the barrier applies to io or device or uncached memory.
6402 <li><tt>ll</tt>: load-load barrier</li>
6403 <li><tt>ls</tt>: load-store barrier</li>
6404 <li><tt>sl</tt>: store-load barrier</li>
6405 <li><tt>ss</tt>: store-store barrier</li>
6406 <li><tt>device</tt>: barrier applies to device and uncached memory also.</li>
6410 This intrinsic causes the system to enforce some ordering constraints upon
6411 the loads and stores of the program. This barrier does not indicate
6412 <em>when</em> any events will occur, it only enforces an <em>order</em> in
6413 which they occur. For any of the specified pairs of load and store operations
6414 (f.ex. load-load, or store-load), all of the first operations preceding the
6415 barrier will complete before any of the second operations succeeding the
6416 barrier begin. Specifically the semantics for each pairing is as follows:
6419 <li><tt>ll</tt>: All loads before the barrier must complete before any load
6420 after the barrier begins.</li>
6422 <li><tt>ls</tt>: All loads before the barrier must complete before any
6423 store after the barrier begins.</li>
6424 <li><tt>ss</tt>: All stores before the barrier must complete before any
6425 store after the barrier begins.</li>
6426 <li><tt>sl</tt>: All stores before the barrier must complete before any
6427 load after the barrier begins.</li>
6430 These semantics are applied with a logical "and" behavior when more than one
6431 is enabled in a single memory barrier intrinsic.
6434 Backends may implement stronger barriers than those requested when they do not
6435 support as fine grained a barrier as requested. Some architectures do not
6436 need all types of barriers and on such architectures, these become noops.
6443 %result1 = load i32* %ptr <i>; yields {i32}:result1 = 4</i>
6444 call void @llvm.memory.barrier( i1 false, i1 true, i1 false, i1 false )
6445 <i>; guarantee the above finishes</i>
6446 store i32 8, %ptr <i>; before this begins</i>
6450 <!-- _______________________________________________________________________ -->
6451 <div class="doc_subsubsection">
6452 <a name="int_atomic_cmp_swap">'<tt>llvm.atomic.cmp.swap.*</tt>' Intrinsic</a>
6454 <div class="doc_text">
6457 This is an overloaded intrinsic. You can use <tt>llvm.atomic.cmp.swap</tt> on
6458 any integer bit width and for different address spaces. Not all targets
6459 support all bit widths however.</p>
6462 declare i8 @llvm.atomic.cmp.swap.i8.p0i8( i8* <ptr>, i8 <cmp>, i8 <val> )
6463 declare i16 @llvm.atomic.cmp.swap.i16.p0i16( i16* <ptr>, i16 <cmp>, i16 <val> )
6464 declare i32 @llvm.atomic.cmp.swap.i32.p0i32( i32* <ptr>, i32 <cmp>, i32 <val> )
6465 declare i64 @llvm.atomic.cmp.swap.i64.p0i64( i64* <ptr>, i64 <cmp>, i64 <val> )
6470 This loads a value in memory and compares it to a given value. If they are
6471 equal, it stores a new value into the memory.
6475 The <tt>llvm.atomic.cmp.swap</tt> intrinsic takes three arguments. The result as
6476 well as both <tt>cmp</tt> and <tt>val</tt> must be integer values with the
6477 same bit width. The <tt>ptr</tt> argument must be a pointer to a value of
6478 this integer type. While any bit width integer may be used, targets may only
6479 lower representations they support in hardware.
6484 This entire intrinsic must be executed atomically. It first loads the value
6485 in memory pointed to by <tt>ptr</tt> and compares it with the value
6486 <tt>cmp</tt>. If they are equal, <tt>val</tt> is stored into the memory. The
6487 loaded value is yielded in all cases. This provides the equivalent of an
6488 atomic compare-and-swap operation within the SSA framework.
6496 %val1 = add i32 4, 4
6497 %result1 = call i32 @llvm.atomic.cmp.swap.i32.p0i32( i32* %ptr, i32 4, %val1 )
6498 <i>; yields {i32}:result1 = 4</i>
6499 %stored1 = icmp eq i32 %result1, 4 <i>; yields {i1}:stored1 = true</i>
6500 %memval1 = load i32* %ptr <i>; yields {i32}:memval1 = 8</i>
6502 %val2 = add i32 1, 1
6503 %result2 = call i32 @llvm.atomic.cmp.swap.i32.p0i32( i32* %ptr, i32 5, %val2 )
6504 <i>; yields {i32}:result2 = 8</i>
6505 %stored2 = icmp eq i32 %result2, 5 <i>; yields {i1}:stored2 = false</i>
6507 %memval2 = load i32* %ptr <i>; yields {i32}:memval2 = 8</i>
6511 <!-- _______________________________________________________________________ -->
6512 <div class="doc_subsubsection">
6513 <a name="int_atomic_swap">'<tt>llvm.atomic.swap.*</tt>' Intrinsic</a>
6515 <div class="doc_text">
6519 This is an overloaded intrinsic. You can use <tt>llvm.atomic.swap</tt> on any
6520 integer bit width. Not all targets support all bit widths however.</p>
6522 declare i8 @llvm.atomic.swap.i8.p0i8( i8* <ptr>, i8 <val> )
6523 declare i16 @llvm.atomic.swap.i16.p0i16( i16* <ptr>, i16 <val> )
6524 declare i32 @llvm.atomic.swap.i32.p0i32( i32* <ptr>, i32 <val> )
6525 declare i64 @llvm.atomic.swap.i64.p0i64( i64* <ptr>, i64 <val> )
6530 This intrinsic loads the value stored in memory at <tt>ptr</tt> and yields
6531 the value from memory. It then stores the value in <tt>val</tt> in the memory
6537 The <tt>llvm.atomic.swap</tt> intrinsic takes two arguments. Both the
6538 <tt>val</tt> argument and the result must be integers of the same bit width.
6539 The first argument, <tt>ptr</tt>, must be a pointer to a value of this
6540 integer type. The targets may only lower integer representations they
6545 This intrinsic loads the value pointed to by <tt>ptr</tt>, yields it, and
6546 stores <tt>val</tt> back into <tt>ptr</tt> atomically. This provides the
6547 equivalent of an atomic swap operation within the SSA framework.
6555 %val1 = add i32 4, 4
6556 %result1 = call i32 @llvm.atomic.swap.i32.p0i32( i32* %ptr, i32 %val1 )
6557 <i>; yields {i32}:result1 = 4</i>
6558 %stored1 = icmp eq i32 %result1, 4 <i>; yields {i1}:stored1 = true</i>
6559 %memval1 = load i32* %ptr <i>; yields {i32}:memval1 = 8</i>
6561 %val2 = add i32 1, 1
6562 %result2 = call i32 @llvm.atomic.swap.i32.p0i32( i32* %ptr, i32 %val2 )
6563 <i>; yields {i32}:result2 = 8</i>
6565 %stored2 = icmp eq i32 %result2, 8 <i>; yields {i1}:stored2 = true</i>
6566 %memval2 = load i32* %ptr <i>; yields {i32}:memval2 = 2</i>
6570 <!-- _______________________________________________________________________ -->
6571 <div class="doc_subsubsection">
6572 <a name="int_atomic_load_add">'<tt>llvm.atomic.load.add.*</tt>' Intrinsic</a>
6575 <div class="doc_text">
6578 This is an overloaded intrinsic. You can use <tt>llvm.atomic.load.add</tt> on any
6579 integer bit width. Not all targets support all bit widths however.</p>
6581 declare i8 @llvm.atomic.load.add.i8..p0i8( i8* <ptr>, i8 <delta> )
6582 declare i16 @llvm.atomic.load.add.i16..p0i16( i16* <ptr>, i16 <delta> )
6583 declare i32 @llvm.atomic.load.add.i32..p0i32( i32* <ptr>, i32 <delta> )
6584 declare i64 @llvm.atomic.load.add.i64..p0i64( i64* <ptr>, i64 <delta> )
6589 This intrinsic adds <tt>delta</tt> to the value stored in memory at
6590 <tt>ptr</tt>. It yields the original value at <tt>ptr</tt>.
6595 The intrinsic takes two arguments, the first a pointer to an integer value
6596 and the second an integer value. The result is also an integer value. These
6597 integer types can have any bit width, but they must all have the same bit
6598 width. The targets may only lower integer representations they support.
6602 This intrinsic does a series of operations atomically. It first loads the
6603 value stored at <tt>ptr</tt>. It then adds <tt>delta</tt>, stores the result
6604 to <tt>ptr</tt>. It yields the original value stored at <tt>ptr</tt>.
6611 %result1 = call i32 @llvm.atomic.load.add.i32.p0i32( i32* %ptr, i32 4 )
6612 <i>; yields {i32}:result1 = 4</i>
6613 %result2 = call i32 @llvm.atomic.load.add.i32.p0i32( i32* %ptr, i32 2 )
6614 <i>; yields {i32}:result2 = 8</i>
6615 %result3 = call i32 @llvm.atomic.load.add.i32.p0i32( i32* %ptr, i32 5 )
6616 <i>; yields {i32}:result3 = 10</i>
6617 %memval1 = load i32* %ptr <i>; yields {i32}:memval1 = 15</i>
6621 <!-- _______________________________________________________________________ -->
6622 <div class="doc_subsubsection">
6623 <a name="int_atomic_load_sub">'<tt>llvm.atomic.load.sub.*</tt>' Intrinsic</a>
6626 <div class="doc_text">
6629 This is an overloaded intrinsic. You can use <tt>llvm.atomic.load.sub</tt> on
6630 any integer bit width and for different address spaces. Not all targets
6631 support all bit widths however.</p>
6633 declare i8 @llvm.atomic.load.sub.i8.p0i32( i8* <ptr>, i8 <delta> )
6634 declare i16 @llvm.atomic.load.sub.i16.p0i32( i16* <ptr>, i16 <delta> )
6635 declare i32 @llvm.atomic.load.sub.i32.p0i32( i32* <ptr>, i32 <delta> )
6636 declare i64 @llvm.atomic.load.sub.i64.p0i32( i64* <ptr>, i64 <delta> )
6641 This intrinsic subtracts <tt>delta</tt> to the value stored in memory at
6642 <tt>ptr</tt>. It yields the original value at <tt>ptr</tt>.
6647 The intrinsic takes two arguments, the first a pointer to an integer value
6648 and the second an integer value. The result is also an integer value. These
6649 integer types can have any bit width, but they must all have the same bit
6650 width. The targets may only lower integer representations they support.
6654 This intrinsic does a series of operations atomically. It first loads the
6655 value stored at <tt>ptr</tt>. It then subtracts <tt>delta</tt>, stores the
6656 result to <tt>ptr</tt>. It yields the original value stored at <tt>ptr</tt>.
6663 %result1 = call i32 @llvm.atomic.load.sub.i32.p0i32( i32* %ptr, i32 4 )
6664 <i>; yields {i32}:result1 = 8</i>
6665 %result2 = call i32 @llvm.atomic.load.sub.i32.p0i32( i32* %ptr, i32 2 )
6666 <i>; yields {i32}:result2 = 4</i>
6667 %result3 = call i32 @llvm.atomic.load.sub.i32.p0i32( i32* %ptr, i32 5 )
6668 <i>; yields {i32}:result3 = 2</i>
6669 %memval1 = load i32* %ptr <i>; yields {i32}:memval1 = -3</i>
6673 <!-- _______________________________________________________________________ -->
6674 <div class="doc_subsubsection">
6675 <a name="int_atomic_load_and">'<tt>llvm.atomic.load.and.*</tt>' Intrinsic</a><br>
6676 <a name="int_atomic_load_nand">'<tt>llvm.atomic.load.nand.*</tt>' Intrinsic</a><br>
6677 <a name="int_atomic_load_or">'<tt>llvm.atomic.load.or.*</tt>' Intrinsic</a><br>
6678 <a name="int_atomic_load_xor">'<tt>llvm.atomic.load.xor.*</tt>' Intrinsic</a><br>
6681 <div class="doc_text">
6684 These are overloaded intrinsics. You can use <tt>llvm.atomic.load_and</tt>,
6685 <tt>llvm.atomic.load_nand</tt>, <tt>llvm.atomic.load_or</tt>, and
6686 <tt>llvm.atomic.load_xor</tt> on any integer bit width and for different
6687 address spaces. Not all targets support all bit widths however.</p>
6689 declare i8 @llvm.atomic.load.and.i8.p0i8( i8* <ptr>, i8 <delta> )
6690 declare i16 @llvm.atomic.load.and.i16.p0i16( i16* <ptr>, i16 <delta> )
6691 declare i32 @llvm.atomic.load.and.i32.p0i32( i32* <ptr>, i32 <delta> )
6692 declare i64 @llvm.atomic.load.and.i64.p0i64( i64* <ptr>, i64 <delta> )
6697 declare i8 @llvm.atomic.load.or.i8.p0i8( i8* <ptr>, i8 <delta> )
6698 declare i16 @llvm.atomic.load.or.i16.p0i16( i16* <ptr>, i16 <delta> )
6699 declare i32 @llvm.atomic.load.or.i32.p0i32( i32* <ptr>, i32 <delta> )
6700 declare i64 @llvm.atomic.load.or.i64.p0i64( i64* <ptr>, i64 <delta> )
6705 declare i8 @llvm.atomic.load.nand.i8.p0i32( i8* <ptr>, i8 <delta> )
6706 declare i16 @llvm.atomic.load.nand.i16.p0i32( i16* <ptr>, i16 <delta> )
6707 declare i32 @llvm.atomic.load.nand.i32.p0i32( i32* <ptr>, i32 <delta> )
6708 declare i64 @llvm.atomic.load.nand.i64.p0i32( i64* <ptr>, i64 <delta> )
6713 declare i8 @llvm.atomic.load.xor.i8.p0i32( i8* <ptr>, i8 <delta> )
6714 declare i16 @llvm.atomic.load.xor.i16.p0i32( i16* <ptr>, i16 <delta> )
6715 declare i32 @llvm.atomic.load.xor.i32.p0i32( i32* <ptr>, i32 <delta> )
6716 declare i64 @llvm.atomic.load.xor.i64.p0i32( i64* <ptr>, i64 <delta> )
6721 These intrinsics bitwise the operation (and, nand, or, xor) <tt>delta</tt> to
6722 the value stored in memory at <tt>ptr</tt>. It yields the original value
6728 These intrinsics take two arguments, the first a pointer to an integer value
6729 and the second an integer value. The result is also an integer value. These
6730 integer types can have any bit width, but they must all have the same bit
6731 width. The targets may only lower integer representations they support.
6735 These intrinsics does a series of operations atomically. They first load the
6736 value stored at <tt>ptr</tt>. They then do the bitwise operation
6737 <tt>delta</tt>, store the result to <tt>ptr</tt>. They yield the original
6738 value stored at <tt>ptr</tt>.
6744 store i32 0x0F0F, %ptr
6745 %result0 = call i32 @llvm.atomic.load.nand.i32.p0i32( i32* %ptr, i32 0xFF )
6746 <i>; yields {i32}:result0 = 0x0F0F</i>
6747 %result1 = call i32 @llvm.atomic.load.and.i32.p0i32( i32* %ptr, i32 0xFF )
6748 <i>; yields {i32}:result1 = 0xFFFFFFF0</i>
6749 %result2 = call i32 @llvm.atomic.load.or.i32.p0i32( i32* %ptr, i32 0F )
6750 <i>; yields {i32}:result2 = 0xF0</i>
6751 %result3 = call i32 @llvm.atomic.load.xor.i32.p0i32( i32* %ptr, i32 0F )
6752 <i>; yields {i32}:result3 = FF</i>
6753 %memval1 = load i32* %ptr <i>; yields {i32}:memval1 = F0</i>
6758 <!-- _______________________________________________________________________ -->
6759 <div class="doc_subsubsection">
6760 <a name="int_atomic_load_max">'<tt>llvm.atomic.load.max.*</tt>' Intrinsic</a><br>
6761 <a name="int_atomic_load_min">'<tt>llvm.atomic.load.min.*</tt>' Intrinsic</a><br>
6762 <a name="int_atomic_load_umax">'<tt>llvm.atomic.load.umax.*</tt>' Intrinsic</a><br>
6763 <a name="int_atomic_load_umin">'<tt>llvm.atomic.load.umin.*</tt>' Intrinsic</a><br>
6766 <div class="doc_text">
6769 These are overloaded intrinsics. You can use <tt>llvm.atomic.load_max</tt>,
6770 <tt>llvm.atomic.load_min</tt>, <tt>llvm.atomic.load_umax</tt>, and
6771 <tt>llvm.atomic.load_umin</tt> on any integer bit width and for different
6772 address spaces. Not all targets
6773 support all bit widths however.</p>
6775 declare i8 @llvm.atomic.load.max.i8.p0i8( i8* <ptr>, i8 <delta> )
6776 declare i16 @llvm.atomic.load.max.i16.p0i16( i16* <ptr>, i16 <delta> )
6777 declare i32 @llvm.atomic.load.max.i32.p0i32( i32* <ptr>, i32 <delta> )
6778 declare i64 @llvm.atomic.load.max.i64.p0i64( i64* <ptr>, i64 <delta> )
6783 declare i8 @llvm.atomic.load.min.i8.p0i8( i8* <ptr>, i8 <delta> )
6784 declare i16 @llvm.atomic.load.min.i16.p0i16( i16* <ptr>, i16 <delta> )
6785 declare i32 @llvm.atomic.load.min.i32..p0i32( i32* <ptr>, i32 <delta> )
6786 declare i64 @llvm.atomic.load.min.i64..p0i64( i64* <ptr>, i64 <delta> )
6791 declare i8 @llvm.atomic.load.umax.i8.p0i8( i8* <ptr>, i8 <delta> )
6792 declare i16 @llvm.atomic.load.umax.i16.p0i16( i16* <ptr>, i16 <delta> )
6793 declare i32 @llvm.atomic.load.umax.i32.p0i32( i32* <ptr>, i32 <delta> )
6794 declare i64 @llvm.atomic.load.umax.i64.p0i64( i64* <ptr>, i64 <delta> )
6799 declare i8 @llvm.atomic.load.umin.i8..p0i8( i8* <ptr>, i8 <delta> )
6800 declare i16 @llvm.atomic.load.umin.i16.p0i16( i16* <ptr>, i16 <delta> )
6801 declare i32 @llvm.atomic.load.umin.i32..p0i32( i32* <ptr>, i32 <delta> )
6802 declare i64 @llvm.atomic.load.umin.i64..p0i64( i64* <ptr>, i64 <delta> )
6807 These intrinsics takes the signed or unsigned minimum or maximum of
6808 <tt>delta</tt> and the value stored in memory at <tt>ptr</tt>. It yields the
6809 original value at <tt>ptr</tt>.
6814 These intrinsics take two arguments, the first a pointer to an integer value
6815 and the second an integer value. The result is also an integer value. These
6816 integer types can have any bit width, but they must all have the same bit
6817 width. The targets may only lower integer representations they support.
6821 These intrinsics does a series of operations atomically. They first load the
6822 value stored at <tt>ptr</tt>. They then do the signed or unsigned min or max
6823 <tt>delta</tt> and the value, store the result to <tt>ptr</tt>. They yield
6824 the original value stored at <tt>ptr</tt>.
6831 %result0 = call i32 @llvm.atomic.load.min.i32.p0i32( i32* %ptr, i32 -2 )
6832 <i>; yields {i32}:result0 = 7</i>
6833 %result1 = call i32 @llvm.atomic.load.max.i32.p0i32( i32* %ptr, i32 8 )
6834 <i>; yields {i32}:result1 = -2</i>
6835 %result2 = call i32 @llvm.atomic.load.umin.i32.p0i32( i32* %ptr, i32 10 )
6836 <i>; yields {i32}:result2 = 8</i>
6837 %result3 = call i32 @llvm.atomic.load.umax.i32.p0i32( i32* %ptr, i32 30 )
6838 <i>; yields {i32}:result3 = 8</i>
6839 %memval1 = load i32* %ptr <i>; yields {i32}:memval1 = 30</i>
6843 <!-- ======================================================================= -->
6844 <div class="doc_subsection">
6845 <a name="int_general">General Intrinsics</a>
6848 <div class="doc_text">
6849 <p> This class of intrinsics is designed to be generic and has
6850 no specific purpose. </p>
6853 <!-- _______________________________________________________________________ -->
6854 <div class="doc_subsubsection">
6855 <a name="int_var_annotation">'<tt>llvm.var.annotation</tt>' Intrinsic</a>
6858 <div class="doc_text">
6862 declare void @llvm.var.annotation(i8* <val>, i8* <str>, i8* <str>, i32 <int> )
6868 The '<tt>llvm.var.annotation</tt>' intrinsic
6874 The first argument is a pointer to a value, the second is a pointer to a
6875 global string, the third is a pointer to a global string which is the source
6876 file name, and the last argument is the line number.
6882 This intrinsic allows annotation of local variables with arbitrary strings.
6883 This can be useful for special purpose optimizations that want to look for these
6884 annotations. These have no other defined use, they are ignored by code
6885 generation and optimization.
6889 <!-- _______________________________________________________________________ -->
6890 <div class="doc_subsubsection">
6891 <a name="int_annotation">'<tt>llvm.annotation.*</tt>' Intrinsic</a>
6894 <div class="doc_text">
6897 <p>This is an overloaded intrinsic. You can use '<tt>llvm.annotation</tt>' on
6898 any integer bit width.
6901 declare i8 @llvm.annotation.i8(i8 <val>, i8* <str>, i8* <str>, i32 <int> )
6902 declare i16 @llvm.annotation.i16(i16 <val>, i8* <str>, i8* <str>, i32 <int> )
6903 declare i32 @llvm.annotation.i32(i32 <val>, i8* <str>, i8* <str>, i32 <int> )
6904 declare i64 @llvm.annotation.i64(i64 <val>, i8* <str>, i8* <str>, i32 <int> )
6905 declare i256 @llvm.annotation.i256(i256 <val>, i8* <str>, i8* <str>, i32 <int> )
6911 The '<tt>llvm.annotation</tt>' intrinsic.
6917 The first argument is an integer value (result of some expression),
6918 the second is a pointer to a global string, the third is a pointer to a global
6919 string which is the source file name, and the last argument is the line number.
6920 It returns the value of the first argument.
6926 This intrinsic allows annotations to be put on arbitrary expressions
6927 with arbitrary strings. This can be useful for special purpose optimizations
6928 that want to look for these annotations. These have no other defined use, they
6929 are ignored by code generation and optimization.
6933 <!-- _______________________________________________________________________ -->
6934 <div class="doc_subsubsection">
6935 <a name="int_trap">'<tt>llvm.trap</tt>' Intrinsic</a>
6938 <div class="doc_text">
6942 declare void @llvm.trap()
6948 The '<tt>llvm.trap</tt>' intrinsic
6960 This intrinsics is lowered to the target dependent trap instruction. If the
6961 target does not have a trap instruction, this intrinsic will be lowered to the
6962 call of the abort() function.
6966 <!-- _______________________________________________________________________ -->
6967 <div class="doc_subsubsection">
6968 <a name="int_stackprotector">'<tt>llvm.stackprotector</tt>' Intrinsic</a>
6970 <div class="doc_text">
6973 declare void @llvm.stackprotector( i8* <guard>, i8** <slot> )
6978 The <tt>llvm.stackprotector</tt> intrinsic takes the <tt>guard</tt> and stores
6979 it onto the stack at <tt>slot</tt>. The stack slot is adjusted to ensure that
6980 it is placed on the stack before local variables.
6984 The <tt>llvm.stackprotector</tt> intrinsic requires two pointer arguments. The
6985 first argument is the value loaded from the stack guard
6986 <tt>@__stack_chk_guard</tt>. The second variable is an <tt>alloca</tt> that
6987 has enough space to hold the value of the guard.
6991 This intrinsic causes the prologue/epilogue inserter to force the position of
6992 the <tt>AllocaInst</tt> stack slot to be before local variables on the
6993 stack. This is to ensure that if a local variable on the stack is overwritten,
6994 it will destroy the value of the guard. When the function exits, the guard on
6995 the stack is checked against the original guard. If they're different, then
6996 the program aborts by calling the <tt>__stack_chk_fail()</tt> function.
7000 <!-- *********************************************************************** -->
7003 <a href="http://jigsaw.w3.org/css-validator/check/referer"><img
7004 src="http://jigsaw.w3.org/css-validator/images/vcss-blue" alt="Valid CSS"></a>
7005 <a href="http://validator.w3.org/check/referer"><img
7006 src="http://www.w3.org/Icons/valid-html401-blue" alt="Valid HTML 4.01"></a>
7008 <a href="mailto:sabre@nondot.org">Chris Lattner</a><br>
7009 <a href="http://llvm.org">The LLVM Compiler Infrastructure</a><br>
7010 Last modified: $Date$