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_vicmp">'<tt>vicmp</tt>' Instruction</a></li>
159 <li><a href="#i_vfcmp">'<tt>vfcmp</tt>' Instruction</a></li>
160 <li><a href="#i_phi">'<tt>phi</tt>' Instruction</a></li>
161 <li><a href="#i_select">'<tt>select</tt>' Instruction</a></li>
162 <li><a href="#i_call">'<tt>call</tt>' Instruction</a></li>
163 <li><a href="#i_va_arg">'<tt>va_arg</tt>' Instruction</a></li>
168 <li><a href="#intrinsics">Intrinsic Functions</a>
170 <li><a href="#int_varargs">Variable Argument Handling Intrinsics</a>
172 <li><a href="#int_va_start">'<tt>llvm.va_start</tt>' Intrinsic</a></li>
173 <li><a href="#int_va_end">'<tt>llvm.va_end</tt>' Intrinsic</a></li>
174 <li><a href="#int_va_copy">'<tt>llvm.va_copy</tt>' Intrinsic</a></li>
177 <li><a href="#int_gc">Accurate Garbage Collection Intrinsics</a>
179 <li><a href="#int_gcroot">'<tt>llvm.gcroot</tt>' Intrinsic</a></li>
180 <li><a href="#int_gcread">'<tt>llvm.gcread</tt>' Intrinsic</a></li>
181 <li><a href="#int_gcwrite">'<tt>llvm.gcwrite</tt>' Intrinsic</a></li>
184 <li><a href="#int_codegen">Code Generator Intrinsics</a>
186 <li><a href="#int_returnaddress">'<tt>llvm.returnaddress</tt>' Intrinsic</a></li>
187 <li><a href="#int_frameaddress">'<tt>llvm.frameaddress</tt>' Intrinsic</a></li>
188 <li><a href="#int_stacksave">'<tt>llvm.stacksave</tt>' Intrinsic</a></li>
189 <li><a href="#int_stackrestore">'<tt>llvm.stackrestore</tt>' Intrinsic</a></li>
190 <li><a href="#int_prefetch">'<tt>llvm.prefetch</tt>' Intrinsic</a></li>
191 <li><a href="#int_pcmarker">'<tt>llvm.pcmarker</tt>' Intrinsic</a></li>
192 <li><a href="#int_readcyclecounter"><tt>llvm.readcyclecounter</tt>' Intrinsic</a></li>
195 <li><a href="#int_libc">Standard C Library Intrinsics</a>
197 <li><a href="#int_memcpy">'<tt>llvm.memcpy.*</tt>' Intrinsic</a></li>
198 <li><a href="#int_memmove">'<tt>llvm.memmove.*</tt>' Intrinsic</a></li>
199 <li><a href="#int_memset">'<tt>llvm.memset.*</tt>' Intrinsic</a></li>
200 <li><a href="#int_sqrt">'<tt>llvm.sqrt.*</tt>' Intrinsic</a></li>
201 <li><a href="#int_powi">'<tt>llvm.powi.*</tt>' Intrinsic</a></li>
202 <li><a href="#int_sin">'<tt>llvm.sin.*</tt>' Intrinsic</a></li>
203 <li><a href="#int_cos">'<tt>llvm.cos.*</tt>' Intrinsic</a></li>
204 <li><a href="#int_pow">'<tt>llvm.pow.*</tt>' Intrinsic</a></li>
207 <li><a href="#int_manip">Bit Manipulation Intrinsics</a>
209 <li><a href="#int_bswap">'<tt>llvm.bswap.*</tt>' Intrinsics</a></li>
210 <li><a href="#int_ctpop">'<tt>llvm.ctpop.*</tt>' Intrinsic </a></li>
211 <li><a href="#int_ctlz">'<tt>llvm.ctlz.*</tt>' Intrinsic </a></li>
212 <li><a href="#int_cttz">'<tt>llvm.cttz.*</tt>' Intrinsic </a></li>
213 <li><a href="#int_part_select">'<tt>llvm.part.select.*</tt>' Intrinsic </a></li>
214 <li><a href="#int_part_set">'<tt>llvm.part.set.*</tt>' Intrinsic </a></li>
217 <li><a href="#int_overflow">Arithmetic with Overflow Intrinsics</a>
219 <li><a href="#int_sadd_overflow">'<tt>llvm.sadd.with.overflow.*</tt> Intrinsics</a></li>
220 <li><a href="#int_uadd_overflow">'<tt>llvm.uadd.with.overflow.*</tt> Intrinsics</a></li>
221 <li><a href="#int_ssub_overflow">'<tt>llvm.ssub.with.overflow.*</tt> Intrinsics</a></li>
222 <li><a href="#int_usub_overflow">'<tt>llvm.usub.with.overflow.*</tt> Intrinsics</a></li>
223 <li><a href="#int_smul_overflow">'<tt>llvm.smul.with.overflow.*</tt> Intrinsics</a></li>
224 <li><a href="#int_umul_overflow">'<tt>llvm.umul.with.overflow.*</tt> Intrinsics</a></li>
227 <li><a href="#int_debugger">Debugger intrinsics</a></li>
228 <li><a href="#int_eh">Exception Handling intrinsics</a></li>
229 <li><a href="#int_trampoline">Trampoline Intrinsic</a>
231 <li><a href="#int_it">'<tt>llvm.init.trampoline</tt>' Intrinsic</a></li>
234 <li><a href="#int_atomics">Atomic intrinsics</a>
236 <li><a href="#int_memory_barrier"><tt>llvm.memory_barrier</tt></a></li>
237 <li><a href="#int_atomic_cmp_swap"><tt>llvm.atomic.cmp.swap</tt></a></li>
238 <li><a href="#int_atomic_swap"><tt>llvm.atomic.swap</tt></a></li>
239 <li><a href="#int_atomic_load_add"><tt>llvm.atomic.load.add</tt></a></li>
240 <li><a href="#int_atomic_load_sub"><tt>llvm.atomic.load.sub</tt></a></li>
241 <li><a href="#int_atomic_load_and"><tt>llvm.atomic.load.and</tt></a></li>
242 <li><a href="#int_atomic_load_nand"><tt>llvm.atomic.load.nand</tt></a></li>
243 <li><a href="#int_atomic_load_or"><tt>llvm.atomic.load.or</tt></a></li>
244 <li><a href="#int_atomic_load_xor"><tt>llvm.atomic.load.xor</tt></a></li>
245 <li><a href="#int_atomic_load_max"><tt>llvm.atomic.load.max</tt></a></li>
246 <li><a href="#int_atomic_load_min"><tt>llvm.atomic.load.min</tt></a></li>
247 <li><a href="#int_atomic_load_umax"><tt>llvm.atomic.load.umax</tt></a></li>
248 <li><a href="#int_atomic_load_umin"><tt>llvm.atomic.load.umin</tt></a></li>
251 <li><a href="#int_general">General intrinsics</a>
253 <li><a href="#int_var_annotation">
254 '<tt>llvm.var.annotation</tt>' Intrinsic</a></li>
255 <li><a href="#int_annotation">
256 '<tt>llvm.annotation.*</tt>' Intrinsic</a></li>
257 <li><a href="#int_trap">
258 '<tt>llvm.trap</tt>' Intrinsic</a></li>
259 <li><a href="#int_stackprotector">
260 '<tt>llvm.stackprotector</tt>' Intrinsic</a></li>
267 <div class="doc_author">
268 <p>Written by <a href="mailto:sabre@nondot.org">Chris Lattner</a>
269 and <a href="mailto:vadve@cs.uiuc.edu">Vikram Adve</a></p>
272 <!-- *********************************************************************** -->
273 <div class="doc_section"> <a name="abstract">Abstract </a></div>
274 <!-- *********************************************************************** -->
276 <div class="doc_text">
277 <p>This document is a reference manual for the LLVM assembly language.
278 LLVM is a Static Single Assignment (SSA) based representation that provides
279 type safety, low-level operations, flexibility, and the capability of
280 representing 'all' high-level languages cleanly. It is the common code
281 representation used throughout all phases of the LLVM compilation
285 <!-- *********************************************************************** -->
286 <div class="doc_section"> <a name="introduction">Introduction</a> </div>
287 <!-- *********************************************************************** -->
289 <div class="doc_text">
291 <p>The LLVM code representation is designed to be used in three
292 different forms: as an in-memory compiler IR, as an on-disk bitcode
293 representation (suitable for fast loading by a Just-In-Time compiler),
294 and as a human readable assembly language representation. This allows
295 LLVM to provide a powerful intermediate representation for efficient
296 compiler transformations and analysis, while providing a natural means
297 to debug and visualize the transformations. The three different forms
298 of LLVM are all equivalent. This document describes the human readable
299 representation and notation.</p>
301 <p>The LLVM representation aims to be light-weight and low-level
302 while being expressive, typed, and extensible at the same time. It
303 aims to be a "universal IR" of sorts, by being at a low enough level
304 that high-level ideas may be cleanly mapped to it (similar to how
305 microprocessors are "universal IR's", allowing many source languages to
306 be mapped to them). By providing type information, LLVM can be used as
307 the target of optimizations: for example, through pointer analysis, it
308 can be proven that a C automatic variable is never accessed outside of
309 the current function... allowing it to be promoted to a simple SSA
310 value instead of a memory location.</p>
314 <!-- _______________________________________________________________________ -->
315 <div class="doc_subsubsection"> <a name="wellformed">Well-Formedness</a> </div>
317 <div class="doc_text">
319 <p>It is important to note that this document describes 'well formed'
320 LLVM assembly language. There is a difference between what the parser
321 accepts and what is considered 'well formed'. For example, the
322 following instruction is syntactically okay, but not well formed:</p>
324 <div class="doc_code">
326 %x = <a href="#i_add">add</a> i32 1, %x
330 <p>...because the definition of <tt>%x</tt> does not dominate all of
331 its uses. The LLVM infrastructure provides a verification pass that may
332 be used to verify that an LLVM module is well formed. This pass is
333 automatically run by the parser after parsing input assembly and by
334 the optimizer before it outputs bitcode. The violations pointed out
335 by the verifier pass indicate bugs in transformation passes or input to
339 <!-- Describe the typesetting conventions here. -->
341 <!-- *********************************************************************** -->
342 <div class="doc_section"> <a name="identifiers">Identifiers</a> </div>
343 <!-- *********************************************************************** -->
345 <div class="doc_text">
347 <p>LLVM identifiers come in two basic types: global and local. Global
348 identifiers (functions, global variables) begin with the @ character. Local
349 identifiers (register names, types) begin with the % character. Additionally,
350 there are three different formats for identifiers, for different purposes:</p>
353 <li>Named values are represented as a string of characters with their prefix.
354 For example, %foo, @DivisionByZero, %a.really.long.identifier. The actual
355 regular expression used is '<tt>[%@][a-zA-Z$._][a-zA-Z$._0-9]*</tt>'.
356 Identifiers which require other characters in their names can be surrounded
357 with quotes. Special characters may be escaped using "\xx" where xx is the
358 ASCII code for the character in hexadecimal. In this way, any character can
359 be used in a name value, even quotes themselves.
361 <li>Unnamed values are represented as an unsigned numeric value with their
362 prefix. For example, %12, @2, %44.</li>
364 <li>Constants, which are described in a <a href="#constants">section about
365 constants</a>, below.</li>
368 <p>LLVM requires that values start with a prefix for two reasons: Compilers
369 don't need to worry about name clashes with reserved words, and the set of
370 reserved words may be expanded in the future without penalty. Additionally,
371 unnamed identifiers allow a compiler to quickly come up with a temporary
372 variable without having to avoid symbol table conflicts.</p>
374 <p>Reserved words in LLVM are very similar to reserved words in other
375 languages. There are keywords for different opcodes
376 ('<tt><a href="#i_add">add</a></tt>',
377 '<tt><a href="#i_bitcast">bitcast</a></tt>',
378 '<tt><a href="#i_ret">ret</a></tt>', etc...), for primitive type names ('<tt><a
379 href="#t_void">void</a></tt>', '<tt><a href="#t_primitive">i32</a></tt>', etc...),
380 and others. These reserved words cannot conflict with variable names, because
381 none of them start with a prefix character ('%' or '@').</p>
383 <p>Here is an example of LLVM code to multiply the integer variable
384 '<tt>%X</tt>' by 8:</p>
388 <div class="doc_code">
390 %result = <a href="#i_mul">mul</a> i32 %X, 8
394 <p>After strength reduction:</p>
396 <div class="doc_code">
398 %result = <a href="#i_shl">shl</a> i32 %X, i8 3
402 <p>And the hard way:</p>
404 <div class="doc_code">
406 <a href="#i_add">add</a> i32 %X, %X <i>; yields {i32}:%0</i>
407 <a href="#i_add">add</a> i32 %0, %0 <i>; yields {i32}:%1</i>
408 %result = <a href="#i_add">add</a> i32 %1, %1
412 <p>This last way of multiplying <tt>%X</tt> by 8 illustrates several
413 important lexical features of LLVM:</p>
417 <li>Comments are delimited with a '<tt>;</tt>' and go until the end of
420 <li>Unnamed temporaries are created when the result of a computation is not
421 assigned to a named value.</li>
423 <li>Unnamed temporaries are numbered sequentially</li>
427 <p>...and it also shows a convention that we follow in this document. When
428 demonstrating instructions, we will follow an instruction with a comment that
429 defines the type and name of value produced. Comments are shown in italic
434 <!-- *********************************************************************** -->
435 <div class="doc_section"> <a name="highlevel">High Level Structure</a> </div>
436 <!-- *********************************************************************** -->
438 <!-- ======================================================================= -->
439 <div class="doc_subsection"> <a name="modulestructure">Module Structure</a>
442 <div class="doc_text">
444 <p>LLVM programs are composed of "Module"s, each of which is a
445 translation unit of the input programs. Each module consists of
446 functions, global variables, and symbol table entries. Modules may be
447 combined together with the LLVM linker, which merges function (and
448 global variable) definitions, resolves forward declarations, and merges
449 symbol table entries. Here is an example of the "hello world" module:</p>
451 <div class="doc_code">
452 <pre><i>; Declare the string constant as a global constant...</i>
453 <a href="#identifiers">@.LC0</a> = <a href="#linkage_internal">internal</a> <a
454 href="#globalvars">constant</a> <a href="#t_array">[13 x i8]</a> c"hello world\0A\00" <i>; [13 x i8]*</i>
456 <i>; External declaration of the puts function</i>
457 <a href="#functionstructure">declare</a> i32 @puts(i8 *) <i>; i32(i8 *)* </i>
459 <i>; Definition of main function</i>
460 define i32 @main() { <i>; i32()* </i>
461 <i>; Convert [13 x i8]* to i8 *...</i>
463 href="#i_getelementptr">getelementptr</a> [13 x i8]* @.LC0, i64 0, i64 0 <i>; i8 *</i>
465 <i>; Call puts function to write out the string to stdout...</i>
467 href="#i_call">call</a> i32 @puts(i8 * %cast210) <i>; i32</i>
469 href="#i_ret">ret</a> i32 0<br>}<br>
473 <p>This example is made up of a <a href="#globalvars">global variable</a>
474 named "<tt>.LC0</tt>", an external declaration of the "<tt>puts</tt>"
475 function, and a <a href="#functionstructure">function definition</a>
476 for "<tt>main</tt>".</p>
478 <p>In general, a module is made up of a list of global values,
479 where both functions and global variables are global values. Global values are
480 represented by a pointer to a memory location (in this case, a pointer to an
481 array of char, and a pointer to a function), and have one of the following <a
482 href="#linkage">linkage types</a>.</p>
486 <!-- ======================================================================= -->
487 <div class="doc_subsection">
488 <a name="linkage">Linkage Types</a>
491 <div class="doc_text">
494 All Global Variables and Functions have one of the following types of linkage:
499 <dt><tt><b><a name="linkage_private">private</a></b></tt>: </dt>
501 <dd>Global values with private linkage are only directly accessible by
502 objects in the current module. In particular, linking code into a module with
503 an private global value may cause the private to be renamed as necessary to
504 avoid collisions. Because the symbol is private to the module, all
505 references can be updated. This doesn't show up in any symbol table in the
509 <dt><tt><b><a name="linkage_internal">internal</a></b></tt>: </dt>
511 <dd> Similar to private, but the value shows as a local symbol (STB_LOCAL in
512 the case of ELF) in the object file. This corresponds to the notion of the
513 '<tt>static</tt>' keyword in C.
516 <dt><tt><b><a name="available_externally">available_externally</a></b></tt>:
519 <dd>Globals with "<tt>available_externally</tt>" linkage are never emitted
520 into the object file corresponding to the LLVM module. They exist to
521 allow inlining and other optimizations to take place given knowledge of the
522 definition of the global, which is known to be somewhere outside the module.
523 Globals with <tt>available_externally</tt> linkage are allowed to be discarded
524 at will, and are otherwise the same as <tt>linkonce_odr</tt>. This linkage
525 type is only allowed on definitions, not declarations.</dd>
527 <dt><tt><b><a name="linkage_linkonce">linkonce</a></b></tt>: </dt>
529 <dd>Globals with "<tt>linkonce</tt>" linkage are merged with other globals of
530 the same name when linkage occurs. This is typically used to implement
531 inline functions, templates, or other code which must be generated in each
532 translation unit that uses it. Unreferenced <tt>linkonce</tt> globals are
533 allowed to be discarded.
536 <dt><tt><b><a name="linkage_common">common</a></b></tt>: </dt>
538 <dd>"<tt>common</tt>" linkage is exactly the same as <tt>linkonce</tt>
539 linkage, except that unreferenced <tt>common</tt> globals may not be
540 discarded. This is used for globals that may be emitted in multiple
541 translation units, but that are not guaranteed to be emitted into every
542 translation unit that uses them. One example of this is tentative
543 definitions in C, such as "<tt>int X;</tt>" at global scope.
546 <dt><tt><b><a name="linkage_weak">weak</a></b></tt>: </dt>
548 <dd>"<tt>weak</tt>" linkage is the same as <tt>common</tt> linkage, except
549 that some targets may choose to emit different assembly sequences for them
550 for target-dependent reasons. This is used for globals that are declared
551 "weak" in C source code.
554 <dt><tt><b><a name="linkage_appending">appending</a></b></tt>: </dt>
556 <dd>"<tt>appending</tt>" linkage may only be applied to global variables of
557 pointer to array type. When two global variables with appending linkage are
558 linked together, the two global arrays are appended together. This is the
559 LLVM, typesafe, equivalent of having the system linker append together
560 "sections" with identical names when .o files are linked.
563 <dt><tt><b><a name="linkage_externweak">extern_weak</a></b></tt>: </dt>
565 <dd>The semantics of this linkage follow the ELF object file model: the
566 symbol is weak until linked, if not linked, the symbol becomes null instead
567 of being an undefined reference.
570 <dt><tt><b><a name="linkage_linkonce">linkonce_odr</a></b></tt>: </dt>
571 <dt><tt><b><a name="linkage_weak">weak_odr</a></b></tt>: </dt>
572 <dd>Some languages allow differing globals to be merged, such as two
573 functions with different semantics. Other languages, such as <tt>C++</tt>,
574 ensure that only equivalent globals are ever merged (the "one definition
575 rule" - "ODR"). Such languages can use the <tt>linkonce_odr</tt>
576 and <tt>weak_odr</tt> linkage types to indicate that the global will only
577 be merged with equivalent globals. These linkage types are otherwise the
578 same as their non-<tt>odr</tt> versions.
581 <dt><tt><b><a name="linkage_external">externally visible</a></b></tt>:</dt>
583 <dd>If none of the above identifiers are used, the global is externally
584 visible, meaning that it participates in linkage and can be used to resolve
585 external symbol references.
590 The next two types of linkage are targeted for Microsoft Windows platform
591 only. They are designed to support importing (exporting) symbols from (to)
592 DLLs (Dynamic Link Libraries).
596 <dt><tt><b><a name="linkage_dllimport">dllimport</a></b></tt>: </dt>
598 <dd>"<tt>dllimport</tt>" linkage causes the compiler to reference a function
599 or variable via a global pointer to a pointer that is set up by the DLL
600 exporting the symbol. On Microsoft Windows targets, the pointer name is
601 formed by combining <code>__imp_</code> and the function or variable name.
604 <dt><tt><b><a name="linkage_dllexport">dllexport</a></b></tt>: </dt>
606 <dd>"<tt>dllexport</tt>" linkage causes the compiler to provide a global
607 pointer to a pointer in a DLL, so that it can be referenced with the
608 <tt>dllimport</tt> attribute. On Microsoft Windows targets, the pointer
609 name is formed by combining <code>__imp_</code> and the function or variable
615 <p>For example, since the "<tt>.LC0</tt>"
616 variable is defined to be internal, if another module defined a "<tt>.LC0</tt>"
617 variable and was linked with this one, one of the two would be renamed,
618 preventing a collision. Since "<tt>main</tt>" and "<tt>puts</tt>" are
619 external (i.e., lacking any linkage declarations), they are accessible
620 outside of the current module.</p>
621 <p>It is illegal for a function <i>declaration</i>
622 to have any linkage type other than "externally visible", <tt>dllimport</tt>
623 or <tt>extern_weak</tt>.</p>
624 <p>Aliases can have only <tt>external</tt>, <tt>internal</tt>, <tt>weak</tt>
625 or <tt>weak_odr</tt> linkages.</p>
628 <!-- ======================================================================= -->
629 <div class="doc_subsection">
630 <a name="callingconv">Calling Conventions</a>
633 <div class="doc_text">
635 <p>LLVM <a href="#functionstructure">functions</a>, <a href="#i_call">calls</a>
636 and <a href="#i_invoke">invokes</a> can all have an optional calling convention
637 specified for the call. The calling convention of any pair of dynamic
638 caller/callee must match, or the behavior of the program is undefined. The
639 following calling conventions are supported by LLVM, and more may be added in
643 <dt><b>"<tt>ccc</tt>" - The C calling convention</b>:</dt>
645 <dd>This calling convention (the default if no other calling convention is
646 specified) matches the target C calling conventions. This calling convention
647 supports varargs function calls and tolerates some mismatch in the declared
648 prototype and implemented declaration of the function (as does normal C).
651 <dt><b>"<tt>fastcc</tt>" - The fast calling convention</b>:</dt>
653 <dd>This calling convention attempts to make calls as fast as possible
654 (e.g. by passing things in registers). This calling convention allows the
655 target to use whatever tricks it wants to produce fast code for the target,
656 without having to conform to an externally specified ABI (Application Binary
657 Interface). Implementations of this convention should allow arbitrary
658 <a href="CodeGenerator.html#tailcallopt">tail call optimization</a> to be
659 supported. This calling convention does not support varargs and requires the
660 prototype of all callees to exactly match the prototype of the function
664 <dt><b>"<tt>coldcc</tt>" - The cold calling convention</b>:</dt>
666 <dd>This calling convention attempts to make code in the caller as efficient
667 as possible under the assumption that the call is not commonly executed. As
668 such, these calls often preserve all registers so that the call does not break
669 any live ranges in the caller side. This calling convention does not support
670 varargs and requires the prototype of all callees to exactly match the
671 prototype of the function definition.
674 <dt><b>"<tt>cc <<em>n</em>></tt>" - Numbered convention</b>:</dt>
676 <dd>Any calling convention may be specified by number, allowing
677 target-specific calling conventions to be used. Target specific calling
678 conventions start at 64.
682 <p>More calling conventions can be added/defined on an as-needed basis, to
683 support pascal conventions or any other well-known target-independent
688 <!-- ======================================================================= -->
689 <div class="doc_subsection">
690 <a name="visibility">Visibility Styles</a>
693 <div class="doc_text">
696 All Global Variables and Functions have one of the following visibility styles:
700 <dt><b>"<tt>default</tt>" - Default style</b>:</dt>
702 <dd>On targets that use the ELF object file format, default visibility means
703 that the declaration is visible to other
704 modules and, in shared libraries, means that the declared entity may be
705 overridden. On Darwin, default visibility means that the declaration is
706 visible to other modules. Default visibility corresponds to "external
707 linkage" in the language.
710 <dt><b>"<tt>hidden</tt>" - Hidden style</b>:</dt>
712 <dd>Two declarations of an object with hidden visibility refer to the same
713 object if they are in the same shared object. Usually, hidden visibility
714 indicates that the symbol will not be placed into the dynamic symbol table,
715 so no other module (executable or shared library) can reference it
719 <dt><b>"<tt>protected</tt>" - Protected style</b>:</dt>
721 <dd>On ELF, protected visibility indicates that the symbol will be placed in
722 the dynamic symbol table, but that references within the defining module will
723 bind to the local symbol. That is, the symbol cannot be overridden by another
730 <!-- ======================================================================= -->
731 <div class="doc_subsection">
732 <a name="namedtypes">Named Types</a>
735 <div class="doc_text">
737 <p>LLVM IR allows you to specify name aliases for certain types. This can make
738 it easier to read the IR and make the IR more condensed (particularly when
739 recursive types are involved). An example of a name specification is:
742 <div class="doc_code">
744 %mytype = type { %mytype*, i32 }
748 <p>You may give a name to any <a href="#typesystem">type</a> except "<a
749 href="t_void">void</a>". Type name aliases may be used anywhere a type is
750 expected with the syntax "%mytype".</p>
752 <p>Note that type names are aliases for the structural type that they indicate,
753 and that you can therefore specify multiple names for the same type. This often
754 leads to confusing behavior when dumping out a .ll file. Since LLVM IR uses
755 structural typing, the name is not part of the type. When printing out LLVM IR,
756 the printer will pick <em>one name</em> to render all types of a particular
757 shape. This means that if you have code where two different source types end up
758 having the same LLVM type, that the dumper will sometimes print the "wrong" or
759 unexpected type. This is an important design point and isn't going to
764 <!-- ======================================================================= -->
765 <div class="doc_subsection">
766 <a name="globalvars">Global Variables</a>
769 <div class="doc_text">
771 <p>Global variables define regions of memory allocated at compilation time
772 instead of run-time. Global variables may optionally be initialized, may have
773 an explicit section to be placed in, and may have an optional explicit alignment
774 specified. A variable may be defined as "thread_local", which means that it
775 will not be shared by threads (each thread will have a separated copy of the
776 variable). A variable may be defined as a global "constant," which indicates
777 that the contents of the variable will <b>never</b> be modified (enabling better
778 optimization, allowing the global data to be placed in the read-only section of
779 an executable, etc). Note that variables that need runtime initialization
780 cannot be marked "constant" as there is a store to the variable.</p>
783 LLVM explicitly allows <em>declarations</em> of global variables to be marked
784 constant, even if the final definition of the global is not. This capability
785 can be used to enable slightly better optimization of the program, but requires
786 the language definition to guarantee that optimizations based on the
787 'constantness' are valid for the translation units that do not include the
791 <p>As SSA values, global variables define pointer values that are in
792 scope (i.e. they dominate) all basic blocks in the program. Global
793 variables always define a pointer to their "content" type because they
794 describe a region of memory, and all memory objects in LLVM are
795 accessed through pointers.</p>
797 <p>A global variable may be declared to reside in a target-specifc numbered
798 address space. For targets that support them, address spaces may affect how
799 optimizations are performed and/or what target instructions are used to access
800 the variable. The default address space is zero. The address space qualifier
801 must precede any other attributes.</p>
803 <p>LLVM allows an explicit section to be specified for globals. If the target
804 supports it, it will emit globals to the section specified.</p>
806 <p>An explicit alignment may be specified for a global. If not present, or if
807 the alignment is set to zero, the alignment of the global is set by the target
808 to whatever it feels convenient. If an explicit alignment is specified, the
809 global is forced to have at least that much alignment. All alignments must be
812 <p>For example, the following defines a global in a numbered address space with
813 an initializer, section, and alignment:</p>
815 <div class="doc_code">
817 @G = addrspace(5) constant float 1.0, section "foo", align 4
824 <!-- ======================================================================= -->
825 <div class="doc_subsection">
826 <a name="functionstructure">Functions</a>
829 <div class="doc_text">
831 <p>LLVM function definitions consist of the "<tt>define</tt>" keyord,
832 an optional <a href="#linkage">linkage type</a>, an optional
833 <a href="#visibility">visibility style</a>, an optional
834 <a href="#callingconv">calling convention</a>, a return type, an optional
835 <a href="#paramattrs">parameter attribute</a> for the return type, a function
836 name, a (possibly empty) argument list (each with optional
837 <a href="#paramattrs">parameter attributes</a>), optional
838 <a href="#fnattrs">function attributes</a>, an optional section,
839 an optional alignment, an optional <a href="#gc">garbage collector name</a>,
840 an opening curly brace, a list of basic blocks, and a closing curly brace.
842 LLVM function declarations consist of the "<tt>declare</tt>" keyword, an
843 optional <a href="#linkage">linkage type</a>, an optional
844 <a href="#visibility">visibility style</a>, an optional
845 <a href="#callingconv">calling convention</a>, a return type, an optional
846 <a href="#paramattrs">parameter attribute</a> for the return type, a function
847 name, a possibly empty list of arguments, an optional alignment, and an optional
848 <a href="#gc">garbage collector name</a>.</p>
850 <p>A function definition contains a list of basic blocks, forming the CFG
851 (Control Flow Graph) for
852 the function. Each basic block may optionally start with a label (giving the
853 basic block a symbol table entry), contains a list of instructions, and ends
854 with a <a href="#terminators">terminator</a> instruction (such as a branch or
855 function return).</p>
857 <p>The first basic block in a function is special in two ways: it is immediately
858 executed on entrance to the function, and it is not allowed to have predecessor
859 basic blocks (i.e. there can not be any branches to the entry block of a
860 function). Because the block can have no predecessors, it also cannot have any
861 <a href="#i_phi">PHI nodes</a>.</p>
863 <p>LLVM allows an explicit section to be specified for functions. If the target
864 supports it, it will emit functions to the section specified.</p>
866 <p>An explicit alignment may be specified for a function. If not present, or if
867 the alignment is set to zero, the alignment of the function is set by the target
868 to whatever it feels convenient. If an explicit alignment is specified, the
869 function is forced to have at least that much alignment. All alignments must be
874 <div class="doc_code">
876 define [<a href="#linkage">linkage</a>] [<a href="#visibility">visibility</a>]
877 [<a href="#callingconv">cconv</a>] [<a href="#paramattrs">ret attrs</a>]
878 <ResultType> @<FunctionName> ([argument list])
879 [<a href="#fnattrs">fn Attrs</a>] [section "name"] [align N]
880 [<a href="#gc">gc</a>] { ... }
887 <!-- ======================================================================= -->
888 <div class="doc_subsection">
889 <a name="aliasstructure">Aliases</a>
891 <div class="doc_text">
892 <p>Aliases act as "second name" for the aliasee value (which can be either
893 function, global variable, another alias or bitcast of global value). Aliases
894 may have an optional <a href="#linkage">linkage type</a>, and an
895 optional <a href="#visibility">visibility style</a>.</p>
899 <div class="doc_code">
901 @<Name> = alias [Linkage] [Visibility] <AliaseeTy> @<Aliasee>
909 <!-- ======================================================================= -->
910 <div class="doc_subsection"><a name="paramattrs">Parameter Attributes</a></div>
911 <div class="doc_text">
912 <p>The return type and each parameter of a function type may have a set of
913 <i>parameter attributes</i> associated with them. Parameter attributes are
914 used to communicate additional information about the result or parameters of
915 a function. Parameter attributes are considered to be part of the function,
916 not of the function type, so functions with different parameter attributes
917 can have the same function type.</p>
919 <p>Parameter attributes are simple keywords that follow the type specified. If
920 multiple parameter attributes are needed, they are space separated. For
923 <div class="doc_code">
925 declare i32 @printf(i8* noalias nocapture, ...)
926 declare i32 @atoi(i8 zeroext)
927 declare signext i8 @returns_signed_char()
931 <p>Note that any attributes for the function result (<tt>nounwind</tt>,
932 <tt>readonly</tt>) come immediately after the argument list.</p>
934 <p>Currently, only the following parameter attributes are defined:</p>
936 <dt><tt>zeroext</tt></dt>
937 <dd>This indicates to the code generator that the parameter or return value
938 should be zero-extended to a 32-bit value by the caller (for a parameter)
939 or the callee (for a return value).</dd>
941 <dt><tt>signext</tt></dt>
942 <dd>This indicates to the code generator that the parameter or return value
943 should be sign-extended to a 32-bit value by the caller (for a parameter)
944 or the callee (for a return value).</dd>
946 <dt><tt>inreg</tt></dt>
947 <dd>This indicates that this parameter or return value should be treated
948 in a special target-dependent fashion during while emitting code for a
949 function call or return (usually, by putting it in a register as opposed
950 to memory, though some targets use it to distinguish between two different
951 kinds of registers). Use of this attribute is target-specific.</dd>
953 <dt><tt><a name="byval">byval</a></tt></dt>
954 <dd>This indicates that the pointer parameter should really be passed by
955 value to the function. The attribute implies that a hidden copy of the
956 pointee is made between the caller and the callee, so the callee is unable
957 to modify the value in the callee. This attribute is only valid on LLVM
958 pointer arguments. It is generally used to pass structs and arrays by
959 value, but is also valid on pointers to scalars. The copy is considered to
960 belong to the caller not the callee (for example,
961 <tt><a href="#readonly">readonly</a></tt> functions should not write to
962 <tt>byval</tt> parameters). This is not a valid attribute for return
963 values. The byval attribute also supports specifying an alignment with the
964 align attribute. This has a target-specific effect on the code generator
965 that usually indicates a desired alignment for the synthesized stack
968 <dt><tt>sret</tt></dt>
969 <dd>This indicates that the pointer parameter specifies the address of a
970 structure that is the return value of the function in the source program.
971 This pointer must be guaranteed by the caller to be valid: loads and stores
972 to the structure may be assumed by the callee to not to trap. This may only
973 be applied to the first parameter. This is not a valid attribute for
976 <dt><tt>noalias</tt></dt>
977 <dd>This indicates that the pointer does not alias any global or any other
978 parameter. The caller is responsible for ensuring that this is the
979 case. On a function return value, <tt>noalias</tt> additionally indicates
980 that the pointer does not alias any other pointers visible to the
981 caller. For further details, please see the discussion of the NoAlias
983 <a href="http://llvm.org/docs/AliasAnalysis.html#MustMayNo">alias
986 <dt><tt>nocapture</tt></dt>
987 <dd>This indicates that the callee does not make any copies of the pointer
988 that outlive the callee itself. This is not a valid attribute for return
991 <dt><tt>nest</tt></dt>
992 <dd>This indicates that the pointer parameter can be excised using the
993 <a href="#int_trampoline">trampoline intrinsics</a>. This is not a valid
994 attribute for return values.</dd>
999 <!-- ======================================================================= -->
1000 <div class="doc_subsection">
1001 <a name="gc">Garbage Collector Names</a>
1004 <div class="doc_text">
1005 <p>Each function may specify a garbage collector name, which is simply a
1008 <div class="doc_code"><pre
1009 >define void @f() gc "name" { ...</pre></div>
1011 <p>The compiler declares the supported values of <i>name</i>. Specifying a
1012 collector which will cause the compiler to alter its output in order to support
1013 the named garbage collection algorithm.</p>
1016 <!-- ======================================================================= -->
1017 <div class="doc_subsection">
1018 <a name="fnattrs">Function Attributes</a>
1021 <div class="doc_text">
1023 <p>Function attributes are set to communicate additional information about
1024 a function. Function attributes are considered to be part of the function,
1025 not of the function type, so functions with different parameter attributes
1026 can have the same function type.</p>
1028 <p>Function attributes are simple keywords that follow the type specified. If
1029 multiple attributes are needed, they are space separated. For
1032 <div class="doc_code">
1034 define void @f() noinline { ... }
1035 define void @f() alwaysinline { ... }
1036 define void @f() alwaysinline optsize { ... }
1037 define void @f() optsize
1042 <dt><tt>alwaysinline</tt></dt>
1043 <dd>This attribute indicates that the inliner should attempt to inline this
1044 function into callers whenever possible, ignoring any active inlining size
1045 threshold for this caller.</dd>
1047 <dt><tt>noinline</tt></dt>
1048 <dd>This attribute indicates that the inliner should never inline this function
1049 in any situation. This attribute may not be used together with the
1050 <tt>alwaysinline</tt> attribute.</dd>
1052 <dt><tt>optsize</tt></dt>
1053 <dd>This attribute suggests that optimization passes and code generator passes
1054 make choices that keep the code size of this function low, and otherwise do
1055 optimizations specifically to reduce code size.</dd>
1057 <dt><tt>noreturn</tt></dt>
1058 <dd>This function attribute indicates that the function never returns normally.
1059 This produces undefined behavior at runtime if the function ever does
1060 dynamically return.</dd>
1062 <dt><tt>nounwind</tt></dt>
1063 <dd>This function attribute indicates that the function never returns with an
1064 unwind or exceptional control flow. If the function does unwind, its runtime
1065 behavior is undefined.</dd>
1067 <dt><tt>readnone</tt></dt>
1068 <dd>This attribute indicates that the function computes its result (or decides to
1069 unwind an exception) based strictly on its arguments, without dereferencing any
1070 pointer arguments or otherwise accessing any mutable state (e.g. memory, control
1071 registers, etc) visible to caller functions. It does not write through any
1072 pointer arguments (including <tt><a href="#byval">byval</a></tt> arguments) and
1073 never changes any state visible to callers. This means that it cannot unwind
1074 exceptions by calling the <tt>C++</tt> exception throwing methods, but could
1075 use the <tt>unwind</tt> instruction.</dd>
1077 <dt><tt><a name="readonly">readonly</a></tt></dt>
1078 <dd>This attribute indicates that the function does not write through any
1079 pointer arguments (including <tt><a href="#byval">byval</a></tt> arguments)
1080 or otherwise modify any state (e.g. memory, control registers, etc) visible to
1081 caller functions. It may dereference pointer arguments and read state that may
1082 be set in the caller. A readonly function always returns the same value (or
1083 unwinds an exception identically) when called with the same set of arguments
1084 and global state. It cannot unwind an exception by calling the <tt>C++</tt>
1085 exception throwing methods, but may use the <tt>unwind</tt> instruction.</dd>
1087 <dt><tt><a name="ssp">ssp</a></tt></dt>
1088 <dd>This attribute indicates that the function should emit a stack smashing
1089 protector. It is in the form of a "canary"—a random value placed on the
1090 stack before the local variables that's checked upon return from the function to
1091 see if it has been overwritten. A heuristic is used to determine if a function
1092 needs stack protectors or not.
1094 <p>If a function that has an <tt>ssp</tt> attribute is inlined into a function
1095 that doesn't have an <tt>ssp</tt> attribute, then the resulting function will
1096 have an <tt>ssp</tt> attribute.</p></dd>
1098 <dt><tt>sspreq</tt></dt>
1099 <dd>This attribute indicates that the function should <em>always</em> emit a
1100 stack smashing protector. This overrides the <tt><a href="#ssp">ssp</a></tt>
1103 <p>If a function that has an <tt>sspreq</tt> attribute is inlined into a
1104 function that doesn't have an <tt>sspreq</tt> attribute or which has
1105 an <tt>ssp</tt> attribute, then the resulting function will have
1106 an <tt>sspreq</tt> attribute.</p></dd>
1111 <!-- ======================================================================= -->
1112 <div class="doc_subsection">
1113 <a name="moduleasm">Module-Level Inline Assembly</a>
1116 <div class="doc_text">
1118 Modules may contain "module-level inline asm" blocks, which corresponds to the
1119 GCC "file scope inline asm" blocks. These blocks are internally concatenated by
1120 LLVM and treated as a single unit, but may be separated in the .ll file if
1121 desired. The syntax is very simple:
1124 <div class="doc_code">
1126 module asm "inline asm code goes here"
1127 module asm "more can go here"
1131 <p>The strings can contain any character by escaping non-printable characters.
1132 The escape sequence used is simply "\xx" where "xx" is the two digit hex code
1137 The inline asm code is simply printed to the machine code .s file when
1138 assembly code is generated.
1142 <!-- ======================================================================= -->
1143 <div class="doc_subsection">
1144 <a name="datalayout">Data Layout</a>
1147 <div class="doc_text">
1148 <p>A module may specify a target specific data layout string that specifies how
1149 data is to be laid out in memory. The syntax for the data layout is simply:</p>
1150 <pre> target datalayout = "<i>layout specification</i>"</pre>
1151 <p>The <i>layout specification</i> consists of a list of specifications
1152 separated by the minus sign character ('-'). Each specification starts with a
1153 letter and may include other information after the letter to define some
1154 aspect of the data layout. The specifications accepted are as follows: </p>
1157 <dd>Specifies that the target lays out data in big-endian form. That is, the
1158 bits with the most significance have the lowest address location.</dd>
1160 <dd>Specifies that the target lays out data in little-endian form. That is,
1161 the bits with the least significance have the lowest address location.</dd>
1162 <dt><tt>p:<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1163 <dd>This specifies the <i>size</i> of a pointer and its <i>abi</i> and
1164 <i>preferred</i> alignments. All sizes are in bits. Specifying the <i>pref</i>
1165 alignment is optional. If omitted, the preceding <tt>:</tt> should be omitted
1167 <dt><tt>i<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1168 <dd>This specifies the alignment for an integer type of a given bit
1169 <i>size</i>. The value of <i>size</i> must be in the range [1,2^23).</dd>
1170 <dt><tt>v<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1171 <dd>This specifies the alignment for a vector type of a given bit
1173 <dt><tt>f<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1174 <dd>This specifies the alignment for a floating point type of a given bit
1175 <i>size</i>. The value of <i>size</i> must be either 32 (float) or 64
1177 <dt><tt>a<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1178 <dd>This specifies the alignment for an aggregate type of a given bit
1181 <p>When constructing the data layout for a given target, LLVM starts with a
1182 default set of specifications which are then (possibly) overriden by the
1183 specifications in the <tt>datalayout</tt> keyword. The default specifications
1184 are given in this list:</p>
1186 <li><tt>E</tt> - big endian</li>
1187 <li><tt>p:32:64:64</tt> - 32-bit pointers with 64-bit alignment</li>
1188 <li><tt>i1:8:8</tt> - i1 is 8-bit (byte) aligned</li>
1189 <li><tt>i8:8:8</tt> - i8 is 8-bit (byte) aligned</li>
1190 <li><tt>i16:16:16</tt> - i16 is 16-bit aligned</li>
1191 <li><tt>i32:32:32</tt> - i32 is 32-bit aligned</li>
1192 <li><tt>i64:32:64</tt> - i64 has ABI alignment of 32-bits but preferred
1193 alignment of 64-bits</li>
1194 <li><tt>f32:32:32</tt> - float is 32-bit aligned</li>
1195 <li><tt>f64:64:64</tt> - double is 64-bit aligned</li>
1196 <li><tt>v64:64:64</tt> - 64-bit vector is 64-bit aligned</li>
1197 <li><tt>v128:128:128</tt> - 128-bit vector is 128-bit aligned</li>
1198 <li><tt>a0:0:1</tt> - aggregates are 8-bit aligned</li>
1200 <p>When LLVM is determining the alignment for a given type, it uses the
1201 following rules:</p>
1203 <li>If the type sought is an exact match for one of the specifications, that
1204 specification is used.</li>
1205 <li>If no match is found, and the type sought is an integer type, then the
1206 smallest integer type that is larger than the bitwidth of the sought type is
1207 used. If none of the specifications are larger than the bitwidth then the the
1208 largest integer type is used. For example, given the default specifications
1209 above, the i7 type will use the alignment of i8 (next largest) while both
1210 i65 and i256 will use the alignment of i64 (largest specified).</li>
1211 <li>If no match is found, and the type sought is a vector type, then the
1212 largest vector type that is smaller than the sought vector type will be used
1213 as a fall back. This happens because <128 x double> can be implemented
1214 in terms of 64 <2 x double>, for example.</li>
1218 <!-- *********************************************************************** -->
1219 <div class="doc_section"> <a name="typesystem">Type System</a> </div>
1220 <!-- *********************************************************************** -->
1222 <div class="doc_text">
1224 <p>The LLVM type system is one of the most important features of the
1225 intermediate representation. Being typed enables a number of
1226 optimizations to be performed on the intermediate representation directly,
1227 without having to do
1228 extra analyses on the side before the transformation. A strong type
1229 system makes it easier to read the generated code and enables novel
1230 analyses and transformations that are not feasible to perform on normal
1231 three address code representations.</p>
1235 <!-- ======================================================================= -->
1236 <div class="doc_subsection"> <a name="t_classifications">Type
1237 Classifications</a> </div>
1238 <div class="doc_text">
1239 <p>The types fall into a few useful
1240 classifications:</p>
1242 <table border="1" cellspacing="0" cellpadding="4">
1244 <tr><th>Classification</th><th>Types</th></tr>
1246 <td><a href="#t_integer">integer</a></td>
1247 <td><tt>i1, i2, i3, ... i8, ... i16, ... i32, ... i64, ... </tt></td>
1250 <td><a href="#t_floating">floating point</a></td>
1251 <td><tt>float, double, x86_fp80, fp128, ppc_fp128</tt></td>
1254 <td><a name="t_firstclass">first class</a></td>
1255 <td><a href="#t_integer">integer</a>,
1256 <a href="#t_floating">floating point</a>,
1257 <a href="#t_pointer">pointer</a>,
1258 <a href="#t_vector">vector</a>,
1259 <a href="#t_struct">structure</a>,
1260 <a href="#t_array">array</a>,
1261 <a href="#t_label">label</a>,
1262 <a href="#t_metadata">metadata</a>.
1266 <td><a href="#t_primitive">primitive</a></td>
1267 <td><a href="#t_label">label</a>,
1268 <a href="#t_void">void</a>,
1269 <a href="#t_floating">floating point</a>,
1270 <a href="#t_metadata">metadata</a>.</td>
1273 <td><a href="#t_derived">derived</a></td>
1274 <td><a href="#t_integer">integer</a>,
1275 <a href="#t_array">array</a>,
1276 <a href="#t_function">function</a>,
1277 <a href="#t_pointer">pointer</a>,
1278 <a href="#t_struct">structure</a>,
1279 <a href="#t_pstruct">packed structure</a>,
1280 <a href="#t_vector">vector</a>,
1281 <a href="#t_opaque">opaque</a>.
1287 <p>The <a href="#t_firstclass">first class</a> types are perhaps the
1288 most important. Values of these types are the only ones which can be
1289 produced by instructions, passed as arguments, or used as operands to
1293 <!-- ======================================================================= -->
1294 <div class="doc_subsection"> <a name="t_primitive">Primitive Types</a> </div>
1296 <div class="doc_text">
1297 <p>The primitive types are the fundamental building blocks of the LLVM
1302 <!-- _______________________________________________________________________ -->
1303 <div class="doc_subsubsection"> <a name="t_floating">Floating Point Types</a> </div>
1305 <div class="doc_text">
1308 <tr><th>Type</th><th>Description</th></tr>
1309 <tr><td><tt>float</tt></td><td>32-bit floating point value</td></tr>
1310 <tr><td><tt>double</tt></td><td>64-bit floating point value</td></tr>
1311 <tr><td><tt>fp128</tt></td><td>128-bit floating point value (112-bit mantissa)</td></tr>
1312 <tr><td><tt>x86_fp80</tt></td><td>80-bit floating point value (X87)</td></tr>
1313 <tr><td><tt>ppc_fp128</tt></td><td>128-bit floating point value (two 64-bits)</td></tr>
1318 <!-- _______________________________________________________________________ -->
1319 <div class="doc_subsubsection"> <a name="t_void">Void Type</a> </div>
1321 <div class="doc_text">
1323 <p>The void type does not represent any value and has no size.</p>
1332 <!-- _______________________________________________________________________ -->
1333 <div class="doc_subsubsection"> <a name="t_label">Label Type</a> </div>
1335 <div class="doc_text">
1337 <p>The label type represents code labels.</p>
1346 <!-- _______________________________________________________________________ -->
1347 <div class="doc_subsubsection"> <a name="t_metadata">Metadata Type</a> </div>
1349 <div class="doc_text">
1351 <p>The metadata type represents embedded metadata. The only derived type that
1352 may contain metadata is <tt>metadata*</tt> or a function type that returns or
1353 takes metadata typed parameters, but not pointer to metadata types.</p>
1363 <!-- ======================================================================= -->
1364 <div class="doc_subsection"> <a name="t_derived">Derived Types</a> </div>
1366 <div class="doc_text">
1368 <p>The real power in LLVM comes from the derived types in the system.
1369 This is what allows a programmer to represent arrays, functions,
1370 pointers, and other useful types. Note that these derived types may be
1371 recursive: For example, it is possible to have a two dimensional array.</p>
1375 <!-- _______________________________________________________________________ -->
1376 <div class="doc_subsubsection"> <a name="t_integer">Integer Type</a> </div>
1378 <div class="doc_text">
1381 <p>The integer type is a very simple derived type that simply specifies an
1382 arbitrary bit width for the integer type desired. Any bit width from 1 bit to
1383 2^23-1 (about 8 million) can be specified.</p>
1391 <p>The number of bits the integer will occupy is specified by the <tt>N</tt>
1395 <table class="layout">
1397 <td class="left"><tt>i1</tt></td>
1398 <td class="left">a single-bit integer.</td>
1401 <td class="left"><tt>i32</tt></td>
1402 <td class="left">a 32-bit integer.</td>
1405 <td class="left"><tt>i1942652</tt></td>
1406 <td class="left">a really big integer of over 1 million bits.</td>
1410 <p>Note that the code generator does not yet support large integer types
1411 to be used as function return types. The specific limit on how large a
1412 return type the code generator can currently handle is target-dependent;
1413 currently it's often 64 bits for 32-bit targets and 128 bits for 64-bit
1418 <!-- _______________________________________________________________________ -->
1419 <div class="doc_subsubsection"> <a name="t_array">Array Type</a> </div>
1421 <div class="doc_text">
1425 <p>The array type is a very simple derived type that arranges elements
1426 sequentially in memory. The array type requires a size (number of
1427 elements) and an underlying data type.</p>
1432 [<# elements> x <elementtype>]
1435 <p>The number of elements is a constant integer value; elementtype may
1436 be any type with a size.</p>
1439 <table class="layout">
1441 <td class="left"><tt>[40 x i32]</tt></td>
1442 <td class="left">Array of 40 32-bit integer values.</td>
1445 <td class="left"><tt>[41 x i32]</tt></td>
1446 <td class="left">Array of 41 32-bit integer values.</td>
1449 <td class="left"><tt>[4 x i8]</tt></td>
1450 <td class="left">Array of 4 8-bit integer values.</td>
1453 <p>Here are some examples of multidimensional arrays:</p>
1454 <table class="layout">
1456 <td class="left"><tt>[3 x [4 x i32]]</tt></td>
1457 <td class="left">3x4 array of 32-bit integer values.</td>
1460 <td class="left"><tt>[12 x [10 x float]]</tt></td>
1461 <td class="left">12x10 array of single precision floating point values.</td>
1464 <td class="left"><tt>[2 x [3 x [4 x i16]]]</tt></td>
1465 <td class="left">2x3x4 array of 16-bit integer values.</td>
1469 <p>Note that 'variable sized arrays' can be implemented in LLVM with a zero
1470 length array. Normally, accesses past the end of an array are undefined in
1471 LLVM (e.g. it is illegal to access the 5th element of a 3 element array).
1472 As a special case, however, zero length arrays are recognized to be variable
1473 length. This allows implementation of 'pascal style arrays' with the LLVM
1474 type "{ i32, [0 x float]}", for example.</p>
1476 <p>Note that the code generator does not yet support large aggregate types
1477 to be used as function return types. The specific limit on how large an
1478 aggregate return type the code generator can currently handle is
1479 target-dependent, and also dependent on the aggregate element types.</p>
1483 <!-- _______________________________________________________________________ -->
1484 <div class="doc_subsubsection"> <a name="t_function">Function Type</a> </div>
1485 <div class="doc_text">
1489 <p>The function type can be thought of as a function signature. It
1490 consists of a return type and a list of formal parameter types. The
1491 return type of a function type is a scalar type, a void type, or a struct type.
1492 If the return type is a struct type then all struct elements must be of first
1493 class types, and the struct must have at least one element.</p>
1498 <returntype list> (<parameter list>)
1501 <p>...where '<tt><parameter list></tt>' is a comma-separated list of type
1502 specifiers. Optionally, the parameter list may include a type <tt>...</tt>,
1503 which indicates that the function takes a variable number of arguments.
1504 Variable argument functions can access their arguments with the <a
1505 href="#int_varargs">variable argument handling intrinsic</a> functions.
1506 '<tt><returntype list></tt>' is a comma-separated list of
1507 <a href="#t_firstclass">first class</a> type specifiers.</p>
1510 <table class="layout">
1512 <td class="left"><tt>i32 (i32)</tt></td>
1513 <td class="left">function taking an <tt>i32</tt>, returning an <tt>i32</tt>
1515 </tr><tr class="layout">
1516 <td class="left"><tt>float (i16 signext, i32 *) *
1518 <td class="left"><a href="#t_pointer">Pointer</a> to a function that takes
1519 an <tt>i16</tt> that should be sign extended and a
1520 <a href="#t_pointer">pointer</a> to <tt>i32</tt>, returning
1523 </tr><tr class="layout">
1524 <td class="left"><tt>i32 (i8*, ...)</tt></td>
1525 <td class="left">A vararg function that takes at least one
1526 <a href="#t_pointer">pointer</a> to <tt>i8 </tt> (char in C),
1527 which returns an integer. This is the signature for <tt>printf</tt> in
1530 </tr><tr class="layout">
1531 <td class="left"><tt>{i32, i32} (i32)</tt></td>
1532 <td class="left">A function taking an <tt>i32</tt>, returning two
1533 <tt>i32</tt> values as an aggregate of type <tt>{ i32, i32 }</tt>
1539 <!-- _______________________________________________________________________ -->
1540 <div class="doc_subsubsection"> <a name="t_struct">Structure Type</a> </div>
1541 <div class="doc_text">
1543 <p>The structure type is used to represent a collection of data members
1544 together in memory. The packing of the field types is defined to match
1545 the ABI of the underlying processor. The elements of a structure may
1546 be any type that has a size.</p>
1547 <p>Structures are accessed using '<tt><a href="#i_load">load</a></tt>
1548 and '<tt><a href="#i_store">store</a></tt>' by getting a pointer to a
1549 field with the '<tt><a href="#i_getelementptr">getelementptr</a></tt>'
1552 <pre> { <type list> }<br></pre>
1554 <table class="layout">
1556 <td class="left"><tt>{ i32, i32, i32 }</tt></td>
1557 <td class="left">A triple of three <tt>i32</tt> values</td>
1558 </tr><tr class="layout">
1559 <td class="left"><tt>{ float, i32 (i32) * }</tt></td>
1560 <td class="left">A pair, where the first element is a <tt>float</tt> and the
1561 second element is a <a href="#t_pointer">pointer</a> to a
1562 <a href="#t_function">function</a> that takes an <tt>i32</tt>, returning
1563 an <tt>i32</tt>.</td>
1567 <p>Note that the code generator does not yet support large aggregate types
1568 to be used as function return types. The specific limit on how large an
1569 aggregate return type the code generator can currently handle is
1570 target-dependent, and also dependent on the aggregate element types.</p>
1574 <!-- _______________________________________________________________________ -->
1575 <div class="doc_subsubsection"> <a name="t_pstruct">Packed Structure Type</a>
1577 <div class="doc_text">
1579 <p>The packed structure type is used to represent a collection of data members
1580 together in memory. There is no padding between fields. Further, the alignment
1581 of a packed structure is 1 byte. The elements of a packed structure may
1582 be any type that has a size.</p>
1583 <p>Structures are accessed using '<tt><a href="#i_load">load</a></tt>
1584 and '<tt><a href="#i_store">store</a></tt>' by getting a pointer to a
1585 field with the '<tt><a href="#i_getelementptr">getelementptr</a></tt>'
1588 <pre> < { <type list> } > <br></pre>
1590 <table class="layout">
1592 <td class="left"><tt>< { i32, i32, i32 } ></tt></td>
1593 <td class="left">A triple of three <tt>i32</tt> values</td>
1594 </tr><tr class="layout">
1596 <tt>< { float, i32 (i32)* } ></tt></td>
1597 <td class="left">A pair, where the first element is a <tt>float</tt> and the
1598 second element is a <a href="#t_pointer">pointer</a> to a
1599 <a href="#t_function">function</a> that takes an <tt>i32</tt>, returning
1600 an <tt>i32</tt>.</td>
1605 <!-- _______________________________________________________________________ -->
1606 <div class="doc_subsubsection"> <a name="t_pointer">Pointer Type</a> </div>
1607 <div class="doc_text">
1609 <p>As in many languages, the pointer type represents a pointer or
1610 reference to another object, which must live in memory. Pointer types may have
1611 an optional address space attribute defining the target-specific numbered
1612 address space where the pointed-to object resides. The default address space is
1615 <p>Note that LLVM does not permit pointers to void (<tt>void*</tt>) nor does
1616 it permit pointers to labels (<tt>label*</tt>). Use <tt>i8*</tt> instead.</p>
1619 <pre> <type> *<br></pre>
1621 <table class="layout">
1623 <td class="left"><tt>[4 x i32]*</tt></td>
1624 <td class="left">A <a href="#t_pointer">pointer</a> to <a
1625 href="#t_array">array</a> of four <tt>i32</tt> values.</td>
1628 <td class="left"><tt>i32 (i32 *) *</tt></td>
1629 <td class="left"> A <a href="#t_pointer">pointer</a> to a <a
1630 href="#t_function">function</a> that takes an <tt>i32*</tt>, returning an
1634 <td class="left"><tt>i32 addrspace(5)*</tt></td>
1635 <td class="left">A <a href="#t_pointer">pointer</a> to an <tt>i32</tt> value
1636 that resides in address space #5.</td>
1641 <!-- _______________________________________________________________________ -->
1642 <div class="doc_subsubsection"> <a name="t_vector">Vector Type</a> </div>
1643 <div class="doc_text">
1647 <p>A vector type is a simple derived type that represents a vector
1648 of elements. Vector types are used when multiple primitive data
1649 are operated in parallel using a single instruction (SIMD).
1650 A vector type requires a size (number of
1651 elements) and an underlying primitive data type. Vectors must have a power
1652 of two length (1, 2, 4, 8, 16 ...). Vector types are
1653 considered <a href="#t_firstclass">first class</a>.</p>
1658 < <# elements> x <elementtype> >
1661 <p>The number of elements is a constant integer value; elementtype may
1662 be any integer or floating point type.</p>
1666 <table class="layout">
1668 <td class="left"><tt><4 x i32></tt></td>
1669 <td class="left">Vector of 4 32-bit integer values.</td>
1672 <td class="left"><tt><8 x float></tt></td>
1673 <td class="left">Vector of 8 32-bit floating-point values.</td>
1676 <td class="left"><tt><2 x i64></tt></td>
1677 <td class="left">Vector of 2 64-bit integer values.</td>
1681 <p>Note that the code generator does not yet support large vector types
1682 to be used as function return types. The specific limit on how large a
1683 vector return type codegen can currently handle is target-dependent;
1684 currently it's often a few times longer than a hardware vector register.</p>
1688 <!-- _______________________________________________________________________ -->
1689 <div class="doc_subsubsection"> <a name="t_opaque">Opaque Type</a> </div>
1690 <div class="doc_text">
1694 <p>Opaque types are used to represent unknown types in the system. This
1695 corresponds (for example) to the C notion of a forward declared structure type.
1696 In LLVM, opaque types can eventually be resolved to any type (not just a
1697 structure type).</p>
1707 <table class="layout">
1709 <td class="left"><tt>opaque</tt></td>
1710 <td class="left">An opaque type.</td>
1715 <!-- ======================================================================= -->
1716 <div class="doc_subsection">
1717 <a name="t_uprefs">Type Up-references</a>
1720 <div class="doc_text">
1723 An "up reference" allows you to refer to a lexically enclosing type without
1724 requiring it to have a name. For instance, a structure declaration may contain a
1725 pointer to any of the types it is lexically a member of. Example of up
1726 references (with their equivalent as named type declarations) include:</p>
1729 { \2 * } %x = type { %x* }
1730 { \2 }* %y = type { %y }*
1735 An up reference is needed by the asmprinter for printing out cyclic types when
1736 there is no declared name for a type in the cycle. Because the asmprinter does
1737 not want to print out an infinite type string, it needs a syntax to handle
1738 recursive types that have no names (all names are optional in llvm IR).
1747 The level is the count of the lexical type that is being referred to.
1752 <table class="layout">
1754 <td class="left"><tt>\1*</tt></td>
1755 <td class="left">Self-referential pointer.</td>
1758 <td class="left"><tt>{ { \3*, i8 }, i32 }</tt></td>
1759 <td class="left">Recursive structure where the upref refers to the out-most
1766 <!-- *********************************************************************** -->
1767 <div class="doc_section"> <a name="constants">Constants</a> </div>
1768 <!-- *********************************************************************** -->
1770 <div class="doc_text">
1772 <p>LLVM has several different basic types of constants. This section describes
1773 them all and their syntax.</p>
1777 <!-- ======================================================================= -->
1778 <div class="doc_subsection"><a name="simpleconstants">Simple Constants</a></div>
1780 <div class="doc_text">
1783 <dt><b>Boolean constants</b></dt>
1785 <dd>The two strings '<tt>true</tt>' and '<tt>false</tt>' are both valid
1786 constants of the <tt><a href="#t_primitive">i1</a></tt> type.
1789 <dt><b>Integer constants</b></dt>
1791 <dd>Standard integers (such as '4') are constants of the <a
1792 href="#t_integer">integer</a> type. Negative numbers may be used with
1796 <dt><b>Floating point constants</b></dt>
1798 <dd>Floating point constants use standard decimal notation (e.g. 123.421),
1799 exponential notation (e.g. 1.23421e+2), or a more precise hexadecimal
1800 notation (see below). The assembler requires the exact decimal value of
1801 a floating-point constant. For example, the assembler accepts 1.25 but
1802 rejects 1.3 because 1.3 is a repeating decimal in binary. Floating point
1803 constants must have a <a href="#t_floating">floating point</a> type. </dd>
1805 <dt><b>Null pointer constants</b></dt>
1807 <dd>The identifier '<tt>null</tt>' is recognized as a null pointer constant
1808 and must be of <a href="#t_pointer">pointer type</a>.</dd>
1812 <p>The one non-intuitive notation for constants is the hexadecimal form
1813 of floating point constants. For example, the form '<tt>double
1814 0x432ff973cafa8000</tt>' is equivalent to (but harder to read than) '<tt>double
1815 4.5e+15</tt>'. The only time hexadecimal floating point constants are required
1816 (and the only time that they are generated by the disassembler) is when a
1817 floating point constant must be emitted but it cannot be represented as a
1818 decimal floating point number in a reasonable number of digits. For example,
1819 NaN's, infinities, and other
1820 special values are represented in their IEEE hexadecimal format so that
1821 assembly and disassembly do not cause any bits to change in the constants.</p>
1822 <p>When using the hexadecimal form, constants of types float and double are
1823 represented using the 16-digit form shown above (which matches the IEEE754
1824 representation for double); float values must, however, be exactly representable
1825 as IEE754 single precision.
1826 Hexadecimal format is always used for long
1827 double, and there are three forms of long double. The 80-bit
1828 format used by x86 is represented as <tt>0xK</tt>
1829 followed by 20 hexadecimal digits.
1830 The 128-bit format used by PowerPC (two adjacent doubles) is represented
1831 by <tt>0xM</tt> followed by 32 hexadecimal digits. The IEEE 128-bit
1832 format is represented
1833 by <tt>0xL</tt> followed by 32 hexadecimal digits; no currently supported
1834 target uses this format. Long doubles will only work if they match
1835 the long double format on your target. All hexadecimal formats are big-endian
1836 (sign bit at the left).</p>
1839 <!-- ======================================================================= -->
1840 <div class="doc_subsection">
1841 <a name="aggregateconstants"> <!-- old anchor -->
1842 <a name="complexconstants">Complex Constants</a></a>
1845 <div class="doc_text">
1846 <p>Complex constants are a (potentially recursive) combination of simple
1847 constants and smaller complex constants.</p>
1850 <dt><b>Structure constants</b></dt>
1852 <dd>Structure constants are represented with notation similar to structure
1853 type definitions (a comma separated list of elements, surrounded by braces
1854 (<tt>{}</tt>)). For example: "<tt>{ i32 4, float 17.0, i32* @G }</tt>",
1855 where "<tt>@G</tt>" is declared as "<tt>@G = external global i32</tt>". Structure constants
1856 must have <a href="#t_struct">structure type</a>, and the number and
1857 types of elements must match those specified by the type.
1860 <dt><b>Array constants</b></dt>
1862 <dd>Array constants are represented with notation similar to array type
1863 definitions (a comma separated list of elements, surrounded by square brackets
1864 (<tt>[]</tt>)). For example: "<tt>[ i32 42, i32 11, i32 74 ]</tt>". Array
1865 constants must have <a href="#t_array">array type</a>, and the number and
1866 types of elements must match those specified by the type.
1869 <dt><b>Vector constants</b></dt>
1871 <dd>Vector constants are represented with notation similar to vector type
1872 definitions (a comma separated list of elements, surrounded by
1873 less-than/greater-than's (<tt><></tt>)). For example: "<tt>< i32 42,
1874 i32 11, i32 74, i32 100 ></tt>". Vector constants must have <a
1875 href="#t_vector">vector type</a>, and the number and types of elements must
1876 match those specified by the type.
1879 <dt><b>Zero initialization</b></dt>
1881 <dd>The string '<tt>zeroinitializer</tt>' can be used to zero initialize a
1882 value to zero of <em>any</em> type, including scalar and aggregate types.
1883 This is often used to avoid having to print large zero initializers (e.g. for
1884 large arrays) and is always exactly equivalent to using explicit zero
1888 <dt><b>Metadata node</b></dt>
1890 <dd>A metadata node is a structure-like constant with
1891 <a href="#t_metadata">metadata type</a>. For example:
1892 "<tt>metadata !{ i32 0, metadata !"test" }</tt>". Unlike other constants
1893 that are meant to be interpreted as part of the instruction stream, metadata
1894 is a place to attach additional information such as debug info.
1900 <!-- ======================================================================= -->
1901 <div class="doc_subsection">
1902 <a name="globalconstants">Global Variable and Function Addresses</a>
1905 <div class="doc_text">
1907 <p>The addresses of <a href="#globalvars">global variables</a> and <a
1908 href="#functionstructure">functions</a> are always implicitly valid (link-time)
1909 constants. These constants are explicitly referenced when the <a
1910 href="#identifiers">identifier for the global</a> is used and always have <a
1911 href="#t_pointer">pointer</a> type. For example, the following is a legal LLVM
1914 <div class="doc_code">
1918 @Z = global [2 x i32*] [ i32* @X, i32* @Y ]
1924 <!-- ======================================================================= -->
1925 <div class="doc_subsection"><a name="undefvalues">Undefined Values</a></div>
1926 <div class="doc_text">
1927 <p>The string '<tt>undef</tt>' is recognized as a type-less constant that has
1928 no specific value. Undefined values may be of any type and be used anywhere
1929 a constant is permitted.</p>
1931 <p>Undefined values indicate to the compiler that the program is well defined
1932 no matter what value is used, giving the compiler more freedom to optimize.
1936 <!-- ======================================================================= -->
1937 <div class="doc_subsection"><a name="constantexprs">Constant Expressions</a>
1940 <div class="doc_text">
1942 <p>Constant expressions are used to allow expressions involving other constants
1943 to be used as constants. Constant expressions may be of any <a
1944 href="#t_firstclass">first class</a> type and may involve any LLVM operation
1945 that does not have side effects (e.g. load and call are not supported). The
1946 following is the syntax for constant expressions:</p>
1949 <dt><b><tt>trunc ( CST to TYPE )</tt></b></dt>
1950 <dd>Truncate a constant to another type. The bit size of CST must be larger
1951 than the bit size of TYPE. Both types must be integers.</dd>
1953 <dt><b><tt>zext ( CST to TYPE )</tt></b></dt>
1954 <dd>Zero extend a constant to another type. The bit size of CST must be
1955 smaller or equal to the bit size of TYPE. Both types must be integers.</dd>
1957 <dt><b><tt>sext ( CST to TYPE )</tt></b></dt>
1958 <dd>Sign extend a constant to another type. The bit size of CST must be
1959 smaller or equal to the bit size of TYPE. Both types must be integers.</dd>
1961 <dt><b><tt>fptrunc ( CST to TYPE )</tt></b></dt>
1962 <dd>Truncate a floating point constant to another floating point type. The
1963 size of CST must be larger than the size of TYPE. Both types must be
1964 floating point.</dd>
1966 <dt><b><tt>fpext ( CST to TYPE )</tt></b></dt>
1967 <dd>Floating point extend a constant to another type. The size of CST must be
1968 smaller or equal to the size of TYPE. Both types must be floating point.</dd>
1970 <dt><b><tt>fptoui ( CST to TYPE )</tt></b></dt>
1971 <dd>Convert a floating point constant to the corresponding unsigned integer
1972 constant. TYPE must be a scalar or vector integer type. CST must be of scalar
1973 or vector floating point type. Both CST and TYPE must be scalars, or vectors
1974 of the same number of elements. If the value won't fit in the integer type,
1975 the results are undefined.</dd>
1977 <dt><b><tt>fptosi ( CST to TYPE )</tt></b></dt>
1978 <dd>Convert a floating point constant to the corresponding signed integer
1979 constant. TYPE must be a scalar or vector integer type. CST must be of scalar
1980 or vector floating point type. Both CST and TYPE must be scalars, or vectors
1981 of the same number of elements. If the value won't fit in the integer type,
1982 the results are undefined.</dd>
1984 <dt><b><tt>uitofp ( CST to TYPE )</tt></b></dt>
1985 <dd>Convert an unsigned integer constant to the corresponding floating point
1986 constant. TYPE must be a scalar or vector floating point type. CST must be of
1987 scalar or vector integer type. Both CST and TYPE must be scalars, or vectors
1988 of the same number of elements. If the value won't fit in the floating point
1989 type, the results are undefined.</dd>
1991 <dt><b><tt>sitofp ( CST to TYPE )</tt></b></dt>
1992 <dd>Convert a signed integer constant to the corresponding floating point
1993 constant. TYPE must be a scalar or vector floating point type. CST must be of
1994 scalar or vector integer type. Both CST and TYPE must be scalars, or vectors
1995 of the same number of elements. If the value won't fit in the floating point
1996 type, the results are undefined.</dd>
1998 <dt><b><tt>ptrtoint ( CST to TYPE )</tt></b></dt>
1999 <dd>Convert a pointer typed constant to the corresponding integer constant
2000 TYPE must be an integer type. CST must be of pointer type. The CST value is
2001 zero extended, truncated, or unchanged to make it fit in TYPE.</dd>
2003 <dt><b><tt>inttoptr ( CST to TYPE )</tt></b></dt>
2004 <dd>Convert a integer constant to a pointer constant. TYPE must be a
2005 pointer type. CST must be of integer type. The CST value is zero extended,
2006 truncated, or unchanged to make it fit in a pointer size. This one is
2007 <i>really</i> dangerous!</dd>
2009 <dt><b><tt>bitcast ( CST to TYPE )</tt></b></dt>
2010 <dd>Convert a constant, CST, to another TYPE. The constraints of the operands
2011 are the same as those for the <a href="#i_bitcast">bitcast
2012 instruction</a>.</dd>
2014 <dt><b><tt>getelementptr ( CSTPTR, IDX0, IDX1, ... )</tt></b></dt>
2016 <dd>Perform the <a href="#i_getelementptr">getelementptr operation</a> on
2017 constants. As with the <a href="#i_getelementptr">getelementptr</a>
2018 instruction, the index list may have zero or more indexes, which are required
2019 to make sense for the type of "CSTPTR".</dd>
2021 <dt><b><tt>select ( COND, VAL1, VAL2 )</tt></b></dt>
2023 <dd>Perform the <a href="#i_select">select operation</a> on
2026 <dt><b><tt>icmp COND ( VAL1, VAL2 )</tt></b></dt>
2027 <dd>Performs the <a href="#i_icmp">icmp operation</a> on constants.</dd>
2029 <dt><b><tt>fcmp COND ( VAL1, VAL2 )</tt></b></dt>
2030 <dd>Performs the <a href="#i_fcmp">fcmp operation</a> on constants.</dd>
2032 <dt><b><tt>vicmp COND ( VAL1, VAL2 )</tt></b></dt>
2033 <dd>Performs the <a href="#i_vicmp">vicmp operation</a> on constants.</dd>
2035 <dt><b><tt>vfcmp COND ( VAL1, VAL2 )</tt></b></dt>
2036 <dd>Performs the <a href="#i_vfcmp">vfcmp operation</a> on constants.</dd>
2038 <dt><b><tt>extractelement ( VAL, IDX )</tt></b></dt>
2040 <dd>Perform the <a href="#i_extractelement">extractelement
2041 operation</a> on constants.</dd>
2043 <dt><b><tt>insertelement ( VAL, ELT, IDX )</tt></b></dt>
2045 <dd>Perform the <a href="#i_insertelement">insertelement
2046 operation</a> on constants.</dd>
2049 <dt><b><tt>shufflevector ( VEC1, VEC2, IDXMASK )</tt></b></dt>
2051 <dd>Perform the <a href="#i_shufflevector">shufflevector
2052 operation</a> on constants.</dd>
2054 <dt><b><tt>OPCODE ( LHS, RHS )</tt></b></dt>
2056 <dd>Perform the specified operation of the LHS and RHS constants. OPCODE may
2057 be any of the <a href="#binaryops">binary</a> or <a href="#bitwiseops">bitwise
2058 binary</a> operations. The constraints on operands are the same as those for
2059 the corresponding instruction (e.g. no bitwise operations on floating point
2060 values are allowed).</dd>
2064 <!-- ======================================================================= -->
2065 <div class="doc_subsection"><a name="metadata">Embedded Metadata</a>
2068 <div class="doc_text">
2070 <p>Embedded metadata provides a way to attach arbitrary data to the
2071 instruction stream without affecting the behaviour of the program. There are
2072 two metadata primitives, strings and nodes. All metadata has the
2073 <tt>metadata</tt> type and is identified in syntax by a preceding exclamation
2074 point ('<tt>!</tt>').
2077 <p>A metadata string is a string surrounded by double quotes. It can contain
2078 any character by escaping non-printable characters with "\xx" where "xx" is
2079 the two digit hex code. For example: "<tt>!"test\00"</tt>".
2082 <p>Metadata nodes are represented with notation similar to structure constants
2083 (a comma separated list of elements, surrounded by braces and preceeded by an
2084 exclamation point). For example: "<tt>!{ metadata !"test\00", i32 10}</tt>".
2087 <p>A metadata node will attempt to track changes to the values it holds. In
2088 the event that a value is deleted, it will be replaced with a typeless
2089 "<tt>null</tt>", such as "<tt>metadata !{null, i32 10}</tt>".</p>
2091 <p>Optimizations may rely on metadata to provide additional information about
2092 the program that isn't available in the instructions, or that isn't easily
2093 computable. Similarly, the code generator may expect a certain metadata format
2094 to be used to express debugging information.</p>
2097 <!-- *********************************************************************** -->
2098 <div class="doc_section"> <a name="othervalues">Other Values</a> </div>
2099 <!-- *********************************************************************** -->
2101 <!-- ======================================================================= -->
2102 <div class="doc_subsection">
2103 <a name="inlineasm">Inline Assembler Expressions</a>
2106 <div class="doc_text">
2109 LLVM supports inline assembler expressions (as opposed to <a href="#moduleasm">
2110 Module-Level Inline Assembly</a>) through the use of a special value. This
2111 value represents the inline assembler as a string (containing the instructions
2112 to emit), a list of operand constraints (stored as a string), and a flag that
2113 indicates whether or not the inline asm expression has side effects. An example
2114 inline assembler expression is:
2117 <div class="doc_code">
2119 i32 (i32) asm "bswap $0", "=r,r"
2124 Inline assembler expressions may <b>only</b> be used as the callee operand of
2125 a <a href="#i_call"><tt>call</tt> instruction</a>. Thus, typically we have:
2128 <div class="doc_code">
2130 %X = call i32 asm "<a href="#int_bswap">bswap</a> $0", "=r,r"(i32 %Y)
2135 Inline asms with side effects not visible in the constraint list must be marked
2136 as having side effects. This is done through the use of the
2137 '<tt>sideeffect</tt>' keyword, like so:
2140 <div class="doc_code">
2142 call void asm sideeffect "eieio", ""()
2146 <p>TODO: The format of the asm and constraints string still need to be
2147 documented here. Constraints on what can be done (e.g. duplication, moving, etc
2148 need to be documented). This is probably best done by reference to another
2149 document that covers inline asm from a holistic perspective.
2154 <!-- *********************************************************************** -->
2155 <div class="doc_section"> <a name="instref">Instruction Reference</a> </div>
2156 <!-- *********************************************************************** -->
2158 <div class="doc_text">
2160 <p>The LLVM instruction set consists of several different
2161 classifications of instructions: <a href="#terminators">terminator
2162 instructions</a>, <a href="#binaryops">binary instructions</a>,
2163 <a href="#bitwiseops">bitwise binary instructions</a>, <a
2164 href="#memoryops">memory instructions</a>, and <a href="#otherops">other
2165 instructions</a>.</p>
2169 <!-- ======================================================================= -->
2170 <div class="doc_subsection"> <a name="terminators">Terminator
2171 Instructions</a> </div>
2173 <div class="doc_text">
2175 <p>As mentioned <a href="#functionstructure">previously</a>, every
2176 basic block in a program ends with a "Terminator" instruction, which
2177 indicates which block should be executed after the current block is
2178 finished. These terminator instructions typically yield a '<tt>void</tt>'
2179 value: they produce control flow, not values (the one exception being
2180 the '<a href="#i_invoke"><tt>invoke</tt></a>' instruction).</p>
2181 <p>There are six different terminator instructions: the '<a
2182 href="#i_ret"><tt>ret</tt></a>' instruction, the '<a href="#i_br"><tt>br</tt></a>'
2183 instruction, the '<a href="#i_switch"><tt>switch</tt></a>' instruction,
2184 the '<a href="#i_invoke"><tt>invoke</tt></a>' instruction, the '<a
2185 href="#i_unwind"><tt>unwind</tt></a>' instruction, and the '<a
2186 href="#i_unreachable"><tt>unreachable</tt></a>' instruction.</p>
2190 <!-- _______________________________________________________________________ -->
2191 <div class="doc_subsubsection"> <a name="i_ret">'<tt>ret</tt>'
2192 Instruction</a> </div>
2193 <div class="doc_text">
2196 ret <type> <value> <i>; Return a value from a non-void function</i>
2197 ret void <i>; Return from void function</i>
2202 <p>The '<tt>ret</tt>' instruction is used to return control flow (and
2203 optionally a value) from a function back to the caller.</p>
2204 <p>There are two forms of the '<tt>ret</tt>' instruction: one that
2205 returns a value and then causes control flow, and one that just causes
2206 control flow to occur.</p>
2210 <p>The '<tt>ret</tt>' instruction optionally accepts a single argument,
2211 the return value. The type of the return value must be a
2212 '<a href="#t_firstclass">first class</a>' type.</p>
2214 <p>A function is not <a href="#wellformed">well formed</a> if
2215 it it has a non-void return type and contains a '<tt>ret</tt>'
2216 instruction with no return value or a return value with a type that
2217 does not match its type, or if it has a void return type and contains
2218 a '<tt>ret</tt>' instruction with a return value.</p>
2222 <p>When the '<tt>ret</tt>' instruction is executed, control flow
2223 returns back to the calling function's context. If the caller is a "<a
2224 href="#i_call"><tt>call</tt></a>" instruction, execution continues at
2225 the instruction after the call. If the caller was an "<a
2226 href="#i_invoke"><tt>invoke</tt></a>" instruction, execution continues
2227 at the beginning of the "normal" destination block. If the instruction
2228 returns a value, that value shall set the call or invoke instruction's
2234 ret i32 5 <i>; Return an integer value of 5</i>
2235 ret void <i>; Return from a void function</i>
2236 ret { i32, i8 } { i32 4, i8 2 } <i>; Return a struct of values 4 and 2</i>
2239 <p>Note that the code generator does not yet fully support large
2240 return values. The specific sizes that are currently supported are
2241 dependent on the target. For integers, on 32-bit targets the limit
2242 is often 64 bits, and on 64-bit targets the limit is often 128 bits.
2243 For aggregate types, the current limits are dependent on the element
2244 types; for example targets are often limited to 2 total integer
2245 elements and 2 total floating-point elements.</p>
2248 <!-- _______________________________________________________________________ -->
2249 <div class="doc_subsubsection"> <a name="i_br">'<tt>br</tt>' Instruction</a> </div>
2250 <div class="doc_text">
2252 <pre> br i1 <cond>, label <iftrue>, label <iffalse><br> br label <dest> <i>; Unconditional branch</i>
2255 <p>The '<tt>br</tt>' instruction is used to cause control flow to
2256 transfer to a different basic block in the current function. There are
2257 two forms of this instruction, corresponding to a conditional branch
2258 and an unconditional branch.</p>
2260 <p>The conditional branch form of the '<tt>br</tt>' instruction takes a
2261 single '<tt>i1</tt>' value and two '<tt>label</tt>' values. The
2262 unconditional form of the '<tt>br</tt>' instruction takes a single
2263 '<tt>label</tt>' value as a target.</p>
2265 <p>Upon execution of a conditional '<tt>br</tt>' instruction, the '<tt>i1</tt>'
2266 argument is evaluated. If the value is <tt>true</tt>, control flows
2267 to the '<tt>iftrue</tt>' <tt>label</tt> argument. If "cond" is <tt>false</tt>,
2268 control flows to the '<tt>iffalse</tt>' <tt>label</tt> argument.</p>
2270 <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
2271 href="#i_ret">ret</a> i32 1<br>IfUnequal:<br> <a href="#i_ret">ret</a> i32 0<br></pre>
2273 <!-- _______________________________________________________________________ -->
2274 <div class="doc_subsubsection">
2275 <a name="i_switch">'<tt>switch</tt>' Instruction</a>
2278 <div class="doc_text">
2282 switch <intty> <value>, label <defaultdest> [ <intty> <val>, label <dest> ... ]
2287 <p>The '<tt>switch</tt>' instruction is used to transfer control flow to one of
2288 several different places. It is a generalization of the '<tt>br</tt>'
2289 instruction, allowing a branch to occur to one of many possible
2295 <p>The '<tt>switch</tt>' instruction uses three parameters: an integer
2296 comparison value '<tt>value</tt>', a default '<tt>label</tt>' destination, and
2297 an array of pairs of comparison value constants and '<tt>label</tt>'s. The
2298 table is not allowed to contain duplicate constant entries.</p>
2302 <p>The <tt>switch</tt> instruction specifies a table of values and
2303 destinations. When the '<tt>switch</tt>' instruction is executed, this
2304 table is searched for the given value. If the value is found, control flow is
2305 transfered to the corresponding destination; otherwise, control flow is
2306 transfered to the default destination.</p>
2308 <h5>Implementation:</h5>
2310 <p>Depending on properties of the target machine and the particular
2311 <tt>switch</tt> instruction, this instruction may be code generated in different
2312 ways. For example, it could be generated as a series of chained conditional
2313 branches or with a lookup table.</p>
2318 <i>; Emulate a conditional br instruction</i>
2319 %Val = <a href="#i_zext">zext</a> i1 %value to i32
2320 switch i32 %Val, label %truedest [ i32 0, label %falsedest ]
2322 <i>; Emulate an unconditional br instruction</i>
2323 switch i32 0, label %dest [ ]
2325 <i>; Implement a jump table:</i>
2326 switch i32 %val, label %otherwise [ i32 0, label %onzero
2328 i32 2, label %ontwo ]
2332 <!-- _______________________________________________________________________ -->
2333 <div class="doc_subsubsection">
2334 <a name="i_invoke">'<tt>invoke</tt>' Instruction</a>
2337 <div class="doc_text">
2342 <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>]
2343 to label <normal label> unwind label <exception label>
2348 <p>The '<tt>invoke</tt>' instruction causes control to transfer to a specified
2349 function, with the possibility of control flow transfer to either the
2350 '<tt>normal</tt>' label or the
2351 '<tt>exception</tt>' label. If the callee function returns with the
2352 "<tt><a href="#i_ret">ret</a></tt>" instruction, control flow will return to the
2353 "normal" label. If the callee (or any indirect callees) returns with the "<a
2354 href="#i_unwind"><tt>unwind</tt></a>" instruction, control is interrupted and
2355 continued at the dynamically nearest "exception" label.</p>
2359 <p>This instruction requires several arguments:</p>
2363 The optional "cconv" marker indicates which <a href="#callingconv">calling
2364 convention</a> the call should use. If none is specified, the call defaults
2365 to using C calling conventions.
2368 <li>The optional <a href="#paramattrs">Parameter Attributes</a> list for
2369 return values. Only '<tt>zeroext</tt>', '<tt>signext</tt>',
2370 and '<tt>inreg</tt>' attributes are valid here.</li>
2372 <li>'<tt>ptr to function ty</tt>': shall be the signature of the pointer to
2373 function value being invoked. In most cases, this is a direct function
2374 invocation, but indirect <tt>invoke</tt>s are just as possible, branching off
2375 an arbitrary pointer to function value.
2378 <li>'<tt>function ptr val</tt>': An LLVM value containing a pointer to a
2379 function to be invoked. </li>
2381 <li>'<tt>function args</tt>': argument list whose types match the function
2382 signature argument types. If the function signature indicates the function
2383 accepts a variable number of arguments, the extra arguments can be
2386 <li>'<tt>normal label</tt>': the label reached when the called function
2387 executes a '<tt><a href="#i_ret">ret</a></tt>' instruction. </li>
2389 <li>'<tt>exception label</tt>': the label reached when a callee returns with
2390 the <a href="#i_unwind"><tt>unwind</tt></a> instruction. </li>
2392 <li>The optional <a href="#fnattrs">function attributes</a> list. Only
2393 '<tt>noreturn</tt>', '<tt>nounwind</tt>', '<tt>readonly</tt>' and
2394 '<tt>readnone</tt>' attributes are valid here.</li>
2399 <p>This instruction is designed to operate as a standard '<tt><a
2400 href="#i_call">call</a></tt>' instruction in most regards. The primary
2401 difference is that it establishes an association with a label, which is used by
2402 the runtime library to unwind the stack.</p>
2404 <p>This instruction is used in languages with destructors to ensure that proper
2405 cleanup is performed in the case of either a <tt>longjmp</tt> or a thrown
2406 exception. Additionally, this is important for implementation of
2407 '<tt>catch</tt>' clauses in high-level languages that support them.</p>
2409 <p>For the purposes of the SSA form, the definition of the value
2410 returned by the '<tt>invoke</tt>' instruction is deemed to occur on
2411 the edge from the current block to the "normal" label. If the callee
2412 unwinds then no return value is available.</p>
2416 %retval = invoke i32 @Test(i32 15) to label %Continue
2417 unwind label %TestCleanup <i>; {i32}:retval set</i>
2418 %retval = invoke <a href="#callingconv">coldcc</a> i32 %Testfnptr(i32 15) to label %Continue
2419 unwind label %TestCleanup <i>; {i32}:retval set</i>
2424 <!-- _______________________________________________________________________ -->
2426 <div class="doc_subsubsection"> <a name="i_unwind">'<tt>unwind</tt>'
2427 Instruction</a> </div>
2429 <div class="doc_text">
2438 <p>The '<tt>unwind</tt>' instruction unwinds the stack, continuing control flow
2439 at the first callee in the dynamic call stack which used an <a
2440 href="#i_invoke"><tt>invoke</tt></a> instruction to perform the call. This is
2441 primarily used to implement exception handling.</p>
2445 <p>The '<tt>unwind</tt>' instruction causes execution of the current function to
2446 immediately halt. The dynamic call stack is then searched for the first <a
2447 href="#i_invoke"><tt>invoke</tt></a> instruction on the call stack. Once found,
2448 execution continues at the "exceptional" destination block specified by the
2449 <tt>invoke</tt> instruction. If there is no <tt>invoke</tt> instruction in the
2450 dynamic call chain, undefined behavior results.</p>
2453 <!-- _______________________________________________________________________ -->
2455 <div class="doc_subsubsection"> <a name="i_unreachable">'<tt>unreachable</tt>'
2456 Instruction</a> </div>
2458 <div class="doc_text">
2467 <p>The '<tt>unreachable</tt>' instruction has no defined semantics. This
2468 instruction is used to inform the optimizer that a particular portion of the
2469 code is not reachable. This can be used to indicate that the code after a
2470 no-return function cannot be reached, and other facts.</p>
2474 <p>The '<tt>unreachable</tt>' instruction has no defined semantics.</p>
2479 <!-- ======================================================================= -->
2480 <div class="doc_subsection"> <a name="binaryops">Binary Operations</a> </div>
2481 <div class="doc_text">
2482 <p>Binary operators are used to do most of the computation in a
2483 program. They require two operands of the same type, execute an operation on them, and
2484 produce a single value. The operands might represent
2485 multiple data, as is the case with the <a href="#t_vector">vector</a> data type.
2486 The result value has the same type as its operands.</p>
2487 <p>There are several different binary operators:</p>
2489 <!-- _______________________________________________________________________ -->
2490 <div class="doc_subsubsection">
2491 <a name="i_add">'<tt>add</tt>' Instruction</a>
2494 <div class="doc_text">
2499 <result> = add <ty> <op1>, <op2> <i>; yields {ty}:result</i>
2504 <p>The '<tt>add</tt>' instruction returns the sum of its two operands.</p>
2508 <p>The two arguments to the '<tt>add</tt>' instruction must be <a
2509 href="#t_integer">integer</a> or
2510 <a href="#t_vector">vector</a> of integer values. Both arguments must
2511 have identical types.</p>
2515 <p>The value produced is the integer sum of the two operands.</p>
2517 <p>If the sum has unsigned overflow, the result returned is the
2518 mathematical result modulo 2<sup>n</sup>, where n is the bit width of
2521 <p>Because LLVM integers use a two's complement representation, this
2522 instruction is appropriate for both signed and unsigned integers.</p>
2527 <result> = add i32 4, %var <i>; yields {i32}:result = 4 + %var</i>
2530 <!-- _______________________________________________________________________ -->
2531 <div class="doc_subsubsection">
2532 <a name="i_fadd">'<tt>fadd</tt>' Instruction</a>
2535 <div class="doc_text">
2540 <result> = fadd <ty> <op1>, <op2> <i>; yields {ty}:result</i>
2545 <p>The '<tt>fadd</tt>' instruction returns the sum of its two operands.</p>
2549 <p>The two arguments to the '<tt>fadd</tt>' instruction must be
2550 <a href="#t_floating">floating point</a> or <a href="#t_vector">vector</a> of
2551 floating point values. Both arguments must have identical types.</p>
2555 <p>The value produced is the floating point sum of the two operands.</p>
2560 <result> = fadd float 4.0, %var <i>; yields {float}:result = 4.0 + %var</i>
2563 <!-- _______________________________________________________________________ -->
2564 <div class="doc_subsubsection">
2565 <a name="i_sub">'<tt>sub</tt>' Instruction</a>
2568 <div class="doc_text">
2573 <result> = sub <ty> <op1>, <op2> <i>; yields {ty}:result</i>
2578 <p>The '<tt>sub</tt>' instruction returns the difference of its two
2581 <p>Note that the '<tt>sub</tt>' instruction is used to represent the
2582 '<tt>neg</tt>' instruction present in most other intermediate
2583 representations.</p>
2587 <p>The two arguments to the '<tt>sub</tt>' instruction must be <a
2588 href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of
2589 integer values. Both arguments must have identical types.</p>
2593 <p>The value produced is the integer difference of the two operands.</p>
2595 <p>If the difference has unsigned overflow, the result returned is the
2596 mathematical result modulo 2<sup>n</sup>, where n is the bit width of
2599 <p>Because LLVM integers use a two's complement representation, this
2600 instruction is appropriate for both signed and unsigned integers.</p>
2604 <result> = sub i32 4, %var <i>; yields {i32}:result = 4 - %var</i>
2605 <result> = sub i32 0, %val <i>; yields {i32}:result = -%var</i>
2609 <!-- _______________________________________________________________________ -->
2610 <div class="doc_subsubsection">
2611 <a name="i_fsub">'<tt>fsub</tt>' Instruction</a>
2614 <div class="doc_text">
2619 <result> = fsub <ty> <op1>, <op2> <i>; yields {ty}:result</i>
2624 <p>The '<tt>fsub</tt>' instruction returns the difference of its two
2627 <p>Note that the '<tt>fsub</tt>' instruction is used to represent the
2628 '<tt>fneg</tt>' instruction present in most other intermediate
2629 representations.</p>
2633 <p>The two arguments to the '<tt>fsub</tt>' instruction must be <a
2634 <a href="#t_floating">floating point</a> or <a href="#t_vector">vector</a>
2635 of floating point values. Both arguments must have identical types.</p>
2639 <p>The value produced is the floating point difference of the two operands.</p>
2643 <result> = fsub float 4.0, %var <i>; yields {float}:result = 4.0 - %var</i>
2644 <result> = fsub float -0.0, %val <i>; yields {float}:result = -%var</i>
2648 <!-- _______________________________________________________________________ -->
2649 <div class="doc_subsubsection">
2650 <a name="i_mul">'<tt>mul</tt>' Instruction</a>
2653 <div class="doc_text">
2656 <pre> <result> = mul <ty> <op1>, <op2> <i>; yields {ty}:result</i>
2659 <p>The '<tt>mul</tt>' instruction returns the product of its two
2664 <p>The two arguments to the '<tt>mul</tt>' instruction must be <a
2665 href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
2666 values. Both arguments must have identical types.</p>
2670 <p>The value produced is the integer product of the two operands.</p>
2672 <p>If the result of the multiplication has unsigned overflow,
2673 the result returned is the mathematical result modulo
2674 2<sup>n</sup>, where n is the bit width of the result.</p>
2675 <p>Because LLVM integers use a two's complement representation, and the
2676 result is the same width as the operands, this instruction returns the
2677 correct result for both signed and unsigned integers. If a full product
2678 (e.g. <tt>i32</tt>x<tt>i32</tt>-><tt>i64</tt>) is needed, the operands
2679 should be sign-extended or zero-extended as appropriate to the
2680 width of the full product.</p>
2682 <pre> <result> = mul i32 4, %var <i>; yields {i32}:result = 4 * %var</i>
2686 <!-- _______________________________________________________________________ -->
2687 <div class="doc_subsubsection">
2688 <a name="i_fmul">'<tt>fmul</tt>' Instruction</a>
2691 <div class="doc_text">
2694 <pre> <result> = fmul <ty> <op1>, <op2> <i>; yields {ty}:result</i>
2697 <p>The '<tt>fmul</tt>' instruction returns the product of its two
2702 <p>The two arguments to the '<tt>fmul</tt>' instruction must be
2703 <a href="#t_floating">floating point</a> or <a href="#t_vector">vector</a>
2704 of floating point values. Both arguments must have identical types.</p>
2708 <p>The value produced is the floating point product of the two operands.</p>
2711 <pre> <result> = fmul float 4.0, %var <i>; yields {float}:result = 4.0 * %var</i>
2715 <!-- _______________________________________________________________________ -->
2716 <div class="doc_subsubsection"> <a name="i_udiv">'<tt>udiv</tt>' Instruction
2718 <div class="doc_text">
2720 <pre> <result> = udiv <ty> <op1>, <op2> <i>; yields {ty}:result</i>
2723 <p>The '<tt>udiv</tt>' instruction returns the quotient of its two
2728 <p>The two arguments to the '<tt>udiv</tt>' instruction must be
2729 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
2730 values. Both arguments must have identical types.</p>
2734 <p>The value produced is the unsigned integer quotient of the two operands.</p>
2735 <p>Note that unsigned integer division and signed integer division are distinct
2736 operations; for signed integer division, use '<tt>sdiv</tt>'.</p>
2737 <p>Division by zero leads to undefined behavior.</p>
2739 <pre> <result> = udiv i32 4, %var <i>; yields {i32}:result = 4 / %var</i>
2742 <!-- _______________________________________________________________________ -->
2743 <div class="doc_subsubsection"> <a name="i_sdiv">'<tt>sdiv</tt>' Instruction
2745 <div class="doc_text">
2748 <result> = sdiv <ty> <op1>, <op2> <i>; yields {ty}:result</i>
2753 <p>The '<tt>sdiv</tt>' instruction returns the quotient of its two
2758 <p>The two arguments to the '<tt>sdiv</tt>' instruction must be
2759 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
2760 values. Both arguments must have identical types.</p>
2763 <p>The value produced is the signed integer quotient of the two operands rounded towards zero.</p>
2764 <p>Note that signed integer division and unsigned integer division are distinct
2765 operations; for unsigned integer division, use '<tt>udiv</tt>'.</p>
2766 <p>Division by zero leads to undefined behavior. Overflow also leads to
2767 undefined behavior; this is a rare case, but can occur, for example,
2768 by doing a 32-bit division of -2147483648 by -1.</p>
2770 <pre> <result> = sdiv i32 4, %var <i>; yields {i32}:result = 4 / %var</i>
2773 <!-- _______________________________________________________________________ -->
2774 <div class="doc_subsubsection"> <a name="i_fdiv">'<tt>fdiv</tt>'
2775 Instruction</a> </div>
2776 <div class="doc_text">
2779 <result> = fdiv <ty> <op1>, <op2> <i>; yields {ty}:result</i>
2783 <p>The '<tt>fdiv</tt>' instruction returns the quotient of its two
2788 <p>The two arguments to the '<tt>fdiv</tt>' instruction must be
2789 <a href="#t_floating">floating point</a> or <a href="#t_vector">vector</a>
2790 of floating point values. Both arguments must have identical types.</p>
2794 <p>The value produced is the floating point quotient of the two operands.</p>
2799 <result> = fdiv float 4.0, %var <i>; yields {float}:result = 4.0 / %var</i>
2803 <!-- _______________________________________________________________________ -->
2804 <div class="doc_subsubsection"> <a name="i_urem">'<tt>urem</tt>' Instruction</a>
2806 <div class="doc_text">
2808 <pre> <result> = urem <ty> <op1>, <op2> <i>; yields {ty}:result</i>
2811 <p>The '<tt>urem</tt>' instruction returns the remainder from the
2812 unsigned division of its two arguments.</p>
2814 <p>The two arguments to the '<tt>urem</tt>' instruction must be
2815 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
2816 values. Both arguments must have identical types.</p>
2818 <p>This instruction returns the unsigned integer <i>remainder</i> of a division.
2819 This instruction always performs an unsigned division to get the remainder.</p>
2820 <p>Note that unsigned integer remainder and signed integer remainder are
2821 distinct operations; for signed integer remainder, use '<tt>srem</tt>'.</p>
2822 <p>Taking the remainder of a division by zero leads to undefined behavior.</p>
2824 <pre> <result> = urem i32 4, %var <i>; yields {i32}:result = 4 % %var</i>
2828 <!-- _______________________________________________________________________ -->
2829 <div class="doc_subsubsection">
2830 <a name="i_srem">'<tt>srem</tt>' Instruction</a>
2833 <div class="doc_text">
2838 <result> = srem <ty> <op1>, <op2> <i>; yields {ty}:result</i>
2843 <p>The '<tt>srem</tt>' instruction returns the remainder from the
2844 signed division of its two operands. This instruction can also take
2845 <a href="#t_vector">vector</a> versions of the values in which case
2846 the elements must be integers.</p>
2850 <p>The two arguments to the '<tt>srem</tt>' instruction must be
2851 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
2852 values. Both arguments must have identical types.</p>
2856 <p>This instruction returns the <i>remainder</i> of a division (where the result
2857 has the same sign as the dividend, <tt>op1</tt>), not the <i>modulo</i>
2858 operator (where the result has the same sign as the divisor, <tt>op2</tt>) of
2859 a value. For more information about the difference, see <a
2860 href="http://mathforum.org/dr.math/problems/anne.4.28.99.html">The
2861 Math Forum</a>. For a table of how this is implemented in various languages,
2862 please see <a href="http://en.wikipedia.org/wiki/Modulo_operation">
2863 Wikipedia: modulo operation</a>.</p>
2864 <p>Note that signed integer remainder and unsigned integer remainder are
2865 distinct operations; for unsigned integer remainder, use '<tt>urem</tt>'.</p>
2866 <p>Taking the remainder of a division by zero leads to undefined behavior.
2867 Overflow also leads to undefined behavior; this is a rare case, but can occur,
2868 for example, by taking the remainder of a 32-bit division of -2147483648 by -1.
2869 (The remainder doesn't actually overflow, but this rule lets srem be
2870 implemented using instructions that return both the result of the division
2871 and the remainder.)</p>
2873 <pre> <result> = srem i32 4, %var <i>; yields {i32}:result = 4 % %var</i>
2877 <!-- _______________________________________________________________________ -->
2878 <div class="doc_subsubsection">
2879 <a name="i_frem">'<tt>frem</tt>' Instruction</a> </div>
2881 <div class="doc_text">
2884 <pre> <result> = frem <ty> <op1>, <op2> <i>; yields {ty}:result</i>
2887 <p>The '<tt>frem</tt>' instruction returns the remainder from the
2888 division of its two operands.</p>
2890 <p>The two arguments to the '<tt>frem</tt>' instruction must be
2891 <a href="#t_floating">floating point</a> or <a href="#t_vector">vector</a>
2892 of floating point values. Both arguments must have identical types.</p>
2896 <p>This instruction returns the <i>remainder</i> of a division.
2897 The remainder has the same sign as the dividend.</p>
2902 <result> = frem float 4.0, %var <i>; yields {float}:result = 4.0 % %var</i>
2906 <!-- ======================================================================= -->
2907 <div class="doc_subsection"> <a name="bitwiseops">Bitwise Binary
2908 Operations</a> </div>
2909 <div class="doc_text">
2910 <p>Bitwise binary operators are used to do various forms of
2911 bit-twiddling in a program. They are generally very efficient
2912 instructions and can commonly be strength reduced from other
2913 instructions. They require two operands of the same type, execute an operation on them,
2914 and produce a single value. The resulting value is the same type as its operands.</p>
2917 <!-- _______________________________________________________________________ -->
2918 <div class="doc_subsubsection"> <a name="i_shl">'<tt>shl</tt>'
2919 Instruction</a> </div>
2920 <div class="doc_text">
2922 <pre> <result> = shl <ty> <op1>, <op2> <i>; yields {ty}:result</i>
2927 <p>The '<tt>shl</tt>' instruction returns the first operand shifted to
2928 the left a specified number of bits.</p>
2932 <p>Both arguments to the '<tt>shl</tt>' instruction must be the same <a
2933 href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
2934 type. '<tt>op2</tt>' is treated as an unsigned value.</p>
2938 <p>The value produced is <tt>op1</tt> * 2<sup><tt>op2</tt></sup> mod 2<sup>n</sup>,
2939 where n is the width of the result. If <tt>op2</tt> is (statically or dynamically) negative or
2940 equal to or larger than the number of bits in <tt>op1</tt>, the result is undefined.
2941 If the arguments are vectors, each vector element of <tt>op1</tt> is shifted by the
2942 corresponding shift amount in <tt>op2</tt>.</p>
2944 <h5>Example:</h5><pre>
2945 <result> = shl i32 4, %var <i>; yields {i32}: 4 << %var</i>
2946 <result> = shl i32 4, 2 <i>; yields {i32}: 16</i>
2947 <result> = shl i32 1, 10 <i>; yields {i32}: 1024</i>
2948 <result> = shl i32 1, 32 <i>; undefined</i>
2949 <result> = shl <2 x i32> < i32 1, i32 1>, < i32 1, i32 2> <i>; yields: result=<2 x i32> < i32 2, i32 4></i>
2952 <!-- _______________________________________________________________________ -->
2953 <div class="doc_subsubsection"> <a name="i_lshr">'<tt>lshr</tt>'
2954 Instruction</a> </div>
2955 <div class="doc_text">
2957 <pre> <result> = lshr <ty> <op1>, <op2> <i>; yields {ty}:result</i>
2961 <p>The '<tt>lshr</tt>' instruction (logical shift right) returns the first
2962 operand shifted to the right a specified number of bits with zero fill.</p>
2965 <p>Both arguments to the '<tt>lshr</tt>' instruction must be the same
2966 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
2967 type. '<tt>op2</tt>' is treated as an unsigned value.</p>
2971 <p>This instruction always performs a logical shift right operation. The most
2972 significant bits of the result will be filled with zero bits after the
2973 shift. If <tt>op2</tt> is (statically or dynamically) equal to or larger than
2974 the number of bits in <tt>op1</tt>, the result is undefined. If the arguments are
2975 vectors, each vector element of <tt>op1</tt> is shifted by the corresponding shift
2976 amount in <tt>op2</tt>.</p>
2980 <result> = lshr i32 4, 1 <i>; yields {i32}:result = 2</i>
2981 <result> = lshr i32 4, 2 <i>; yields {i32}:result = 1</i>
2982 <result> = lshr i8 4, 3 <i>; yields {i8}:result = 0</i>
2983 <result> = lshr i8 -2, 1 <i>; yields {i8}:result = 0x7FFFFFFF </i>
2984 <result> = lshr i32 1, 32 <i>; undefined</i>
2985 <result> = lshr <2 x i32> < i32 -2, i32 4>, < i32 1, i32 2> <i>; yields: result=<2 x i32> < i32 0x7FFFFFFF, i32 1></i>
2989 <!-- _______________________________________________________________________ -->
2990 <div class="doc_subsubsection"> <a name="i_ashr">'<tt>ashr</tt>'
2991 Instruction</a> </div>
2992 <div class="doc_text">
2995 <pre> <result> = ashr <ty> <op1>, <op2> <i>; yields {ty}:result</i>
2999 <p>The '<tt>ashr</tt>' instruction (arithmetic shift right) returns the first
3000 operand shifted to the right a specified number of bits with sign extension.</p>
3003 <p>Both arguments to the '<tt>ashr</tt>' instruction must be the same
3004 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
3005 type. '<tt>op2</tt>' is treated as an unsigned value.</p>
3008 <p>This instruction always performs an arithmetic shift right operation,
3009 The most significant bits of the result will be filled with the sign bit
3010 of <tt>op1</tt>. If <tt>op2</tt> is (statically or dynamically) equal to or
3011 larger than the number of bits in <tt>op1</tt>, the result is undefined. If the
3012 arguments are vectors, each vector element of <tt>op1</tt> is shifted by the
3013 corresponding shift amount in <tt>op2</tt>.</p>
3017 <result> = ashr i32 4, 1 <i>; yields {i32}:result = 2</i>
3018 <result> = ashr i32 4, 2 <i>; yields {i32}:result = 1</i>
3019 <result> = ashr i8 4, 3 <i>; yields {i8}:result = 0</i>
3020 <result> = ashr i8 -2, 1 <i>; yields {i8}:result = -1</i>
3021 <result> = ashr i32 1, 32 <i>; undefined</i>
3022 <result> = ashr <2 x i32> < i32 -2, i32 4>, < i32 1, i32 3> <i>; yields: result=<2 x i32> < i32 -1, i32 0></i>
3026 <!-- _______________________________________________________________________ -->
3027 <div class="doc_subsubsection"> <a name="i_and">'<tt>and</tt>'
3028 Instruction</a> </div>
3030 <div class="doc_text">
3035 <result> = and <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3040 <p>The '<tt>and</tt>' instruction returns the bitwise logical and of
3041 its two operands.</p>
3045 <p>The two arguments to the '<tt>and</tt>' instruction must be
3046 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
3047 values. Both arguments must have identical types.</p>
3050 <p>The truth table used for the '<tt>and</tt>' instruction is:</p>
3053 <table border="1" cellspacing="0" cellpadding="4">
3085 <result> = and i32 4, %var <i>; yields {i32}:result = 4 & %var</i>
3086 <result> = and i32 15, 40 <i>; yields {i32}:result = 8</i>
3087 <result> = and i32 4, 8 <i>; yields {i32}:result = 0</i>
3090 <!-- _______________________________________________________________________ -->
3091 <div class="doc_subsubsection"> <a name="i_or">'<tt>or</tt>' Instruction</a> </div>
3092 <div class="doc_text">
3094 <pre> <result> = or <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3097 <p>The '<tt>or</tt>' instruction returns the bitwise logical inclusive
3098 or of its two operands.</p>
3101 <p>The two arguments to the '<tt>or</tt>' instruction must be
3102 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
3103 values. Both arguments must have identical types.</p>
3105 <p>The truth table used for the '<tt>or</tt>' instruction is:</p>
3108 <table border="1" cellspacing="0" cellpadding="4">
3139 <pre> <result> = or i32 4, %var <i>; yields {i32}:result = 4 | %var</i>
3140 <result> = or i32 15, 40 <i>; yields {i32}:result = 47</i>
3141 <result> = or i32 4, 8 <i>; yields {i32}:result = 12</i>
3144 <!-- _______________________________________________________________________ -->
3145 <div class="doc_subsubsection"> <a name="i_xor">'<tt>xor</tt>'
3146 Instruction</a> </div>
3147 <div class="doc_text">
3149 <pre> <result> = xor <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3152 <p>The '<tt>xor</tt>' instruction returns the bitwise logical exclusive
3153 or of its two operands. The <tt>xor</tt> is used to implement the
3154 "one's complement" operation, which is the "~" operator in C.</p>
3156 <p>The two arguments to the '<tt>xor</tt>' instruction must be
3157 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
3158 values. Both arguments must have identical types.</p>
3162 <p>The truth table used for the '<tt>xor</tt>' instruction is:</p>
3165 <table border="1" cellspacing="0" cellpadding="4">
3197 <pre> <result> = xor i32 4, %var <i>; yields {i32}:result = 4 ^ %var</i>
3198 <result> = xor i32 15, 40 <i>; yields {i32}:result = 39</i>
3199 <result> = xor i32 4, 8 <i>; yields {i32}:result = 12</i>
3200 <result> = xor i32 %V, -1 <i>; yields {i32}:result = ~%V</i>
3204 <!-- ======================================================================= -->
3205 <div class="doc_subsection">
3206 <a name="vectorops">Vector Operations</a>
3209 <div class="doc_text">
3211 <p>LLVM supports several instructions to represent vector operations in a
3212 target-independent manner. These instructions cover the element-access and
3213 vector-specific operations needed to process vectors effectively. While LLVM
3214 does directly support these vector operations, many sophisticated algorithms
3215 will want to use target-specific intrinsics to take full advantage of a specific
3220 <!-- _______________________________________________________________________ -->
3221 <div class="doc_subsubsection">
3222 <a name="i_extractelement">'<tt>extractelement</tt>' Instruction</a>
3225 <div class="doc_text">
3230 <result> = extractelement <n x <ty>> <val>, i32 <idx> <i>; yields <ty></i>
3236 The '<tt>extractelement</tt>' instruction extracts a single scalar
3237 element from a vector at a specified index.
3244 The first operand of an '<tt>extractelement</tt>' instruction is a
3245 value of <a href="#t_vector">vector</a> type. The second operand is
3246 an index indicating the position from which to extract the element.
3247 The index may be a variable.</p>
3252 The result is a scalar of the same type as the element type of
3253 <tt>val</tt>. Its value is the value at position <tt>idx</tt> of
3254 <tt>val</tt>. If <tt>idx</tt> exceeds the length of <tt>val</tt>, the
3255 results are undefined.
3261 %result = extractelement <4 x i32> %vec, i32 0 <i>; yields i32</i>
3266 <!-- _______________________________________________________________________ -->
3267 <div class="doc_subsubsection">
3268 <a name="i_insertelement">'<tt>insertelement</tt>' Instruction</a>
3271 <div class="doc_text">
3276 <result> = insertelement <n x <ty>> <val>, <ty> <elt>, i32 <idx> <i>; yields <n x <ty>></i>
3282 The '<tt>insertelement</tt>' instruction inserts a scalar
3283 element into a vector at a specified index.
3290 The first operand of an '<tt>insertelement</tt>' instruction is a
3291 value of <a href="#t_vector">vector</a> type. The second operand is a
3292 scalar value whose type must equal the element type of the first
3293 operand. The third operand is an index indicating the position at
3294 which to insert the value. The index may be a variable.</p>
3299 The result is a vector of the same type as <tt>val</tt>. Its
3300 element values are those of <tt>val</tt> except at position
3301 <tt>idx</tt>, where it gets the value <tt>elt</tt>. If <tt>idx</tt>
3302 exceeds the length of <tt>val</tt>, the results are undefined.
3308 %result = insertelement <4 x i32> %vec, i32 1, i32 0 <i>; yields <4 x i32></i>
3312 <!-- _______________________________________________________________________ -->
3313 <div class="doc_subsubsection">
3314 <a name="i_shufflevector">'<tt>shufflevector</tt>' Instruction</a>
3317 <div class="doc_text">
3322 <result> = shufflevector <n x <ty>> <v1>, <n x <ty>> <v2>, <m x i32> <mask> <i>; yields <m x <ty>></i>
3328 The '<tt>shufflevector</tt>' instruction constructs a permutation of elements
3329 from two input vectors, returning a vector with the same element type as
3330 the input and length that is the same as the shuffle mask.
3336 The first two operands of a '<tt>shufflevector</tt>' instruction are vectors
3337 with types that match each other. The third argument is a shuffle mask whose
3338 element type is always 'i32'. The result of the instruction is a vector whose
3339 length is the same as the shuffle mask and whose element type is the same as
3340 the element type of the first two operands.
3344 The shuffle mask operand is required to be a constant vector with either
3345 constant integer or undef values.
3351 The elements of the two input vectors are numbered from left to right across
3352 both of the vectors. The shuffle mask operand specifies, for each element of
3353 the result vector, which element of the two input vectors the result element
3354 gets. The element selector may be undef (meaning "don't care") and the second
3355 operand may be undef if performing a shuffle from only one vector.
3361 %result = shufflevector <4 x i32> %v1, <4 x i32> %v2,
3362 <4 x i32> <i32 0, i32 4, i32 1, i32 5> <i>; yields <4 x i32></i>
3363 %result = shufflevector <4 x i32> %v1, <4 x i32> undef,
3364 <4 x i32> <i32 0, i32 1, i32 2, i32 3> <i>; yields <4 x i32></i> - Identity shuffle.
3365 %result = shufflevector <8 x i32> %v1, <8 x i32> undef,
3366 <4 x i32> <i32 0, i32 1, i32 2, i32 3> <i>; yields <4 x i32></i>
3367 %result = shufflevector <4 x i32> %v1, <4 x i32> %v2,
3368 <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>
3373 <!-- ======================================================================= -->
3374 <div class="doc_subsection">
3375 <a name="aggregateops">Aggregate Operations</a>
3378 <div class="doc_text">
3380 <p>LLVM supports several instructions for working with aggregate values.
3385 <!-- _______________________________________________________________________ -->
3386 <div class="doc_subsubsection">
3387 <a name="i_extractvalue">'<tt>extractvalue</tt>' Instruction</a>
3390 <div class="doc_text">
3395 <result> = extractvalue <aggregate type> <val>, <idx>{, <idx>}*
3401 The '<tt>extractvalue</tt>' instruction extracts the value of a struct field
3402 or array element from an aggregate value.
3409 The first operand of an '<tt>extractvalue</tt>' instruction is a
3410 value of <a href="#t_struct">struct</a> or <a href="#t_array">array</a>
3411 type. The operands are constant indices to specify which value to extract
3412 in a similar manner as indices in a
3413 '<tt><a href="#i_getelementptr">getelementptr</a></tt>' instruction.
3419 The result is the value at the position in the aggregate specified by
3426 %result = extractvalue {i32, float} %agg, 0 <i>; yields i32</i>
3431 <!-- _______________________________________________________________________ -->
3432 <div class="doc_subsubsection">
3433 <a name="i_insertvalue">'<tt>insertvalue</tt>' Instruction</a>
3436 <div class="doc_text">
3441 <result> = insertvalue <aggregate type> <val>, <ty> <val>, <idx> <i>; yields <n x <ty>></i>
3447 The '<tt>insertvalue</tt>' instruction inserts a value
3448 into a struct field or array element in an aggregate.
3455 The first operand of an '<tt>insertvalue</tt>' instruction is a
3456 value of <a href="#t_struct">struct</a> or <a href="#t_array">array</a> type.
3457 The second operand is a first-class value to insert.
3458 The following operands are constant indices
3459 indicating the position at which to insert the value in a similar manner as
3461 '<tt><a href="#i_getelementptr">getelementptr</a></tt>' instruction.
3462 The value to insert must have the same type as the value identified
3469 The result is an aggregate of the same type as <tt>val</tt>. Its
3470 value is that of <tt>val</tt> except that the value at the position
3471 specified by the indices is that of <tt>elt</tt>.
3477 %result = insertvalue {i32, float} %agg, i32 1, 0 <i>; yields {i32, float}</i>
3482 <!-- ======================================================================= -->
3483 <div class="doc_subsection">
3484 <a name="memoryops">Memory Access and Addressing Operations</a>
3487 <div class="doc_text">
3489 <p>A key design point of an SSA-based representation is how it
3490 represents memory. In LLVM, no memory locations are in SSA form, which
3491 makes things very simple. This section describes how to read, write,
3492 allocate, and free memory in LLVM.</p>
3496 <!-- _______________________________________________________________________ -->
3497 <div class="doc_subsubsection">
3498 <a name="i_malloc">'<tt>malloc</tt>' Instruction</a>
3501 <div class="doc_text">
3506 <result> = malloc <type>[, i32 <NumElements>][, align <alignment>] <i>; yields {type*}:result</i>
3511 <p>The '<tt>malloc</tt>' instruction allocates memory from the system
3512 heap and returns a pointer to it. The object is always allocated in the generic
3513 address space (address space zero).</p>
3517 <p>The '<tt>malloc</tt>' instruction allocates
3518 <tt>sizeof(<type>)*NumElements</tt>
3519 bytes of memory from the operating system and returns a pointer of the
3520 appropriate type to the program. If "NumElements" is specified, it is the
3521 number of elements allocated, otherwise "NumElements" is defaulted to be one.
3522 If a constant alignment is specified, the value result of the allocation is guaranteed to
3523 be aligned to at least that boundary. If not specified, or if zero, the target can
3524 choose to align the allocation on any convenient boundary.</p>
3526 <p>'<tt>type</tt>' must be a sized type.</p>
3530 <p>Memory is allocated using the system "<tt>malloc</tt>" function, and
3531 a pointer is returned. The result of a zero byte allocation is undefined. The
3532 result is null if there is insufficient memory available.</p>
3537 %array = malloc [4 x i8] <i>; yields {[%4 x i8]*}:array</i>
3539 %size = <a href="#i_add">add</a> i32 2, 2 <i>; yields {i32}:size = i32 4</i>
3540 %array1 = malloc i8, i32 4 <i>; yields {i8*}:array1</i>
3541 %array2 = malloc [12 x i8], i32 %size <i>; yields {[12 x i8]*}:array2</i>
3542 %array3 = malloc i32, i32 4, align 1024 <i>; yields {i32*}:array3</i>
3543 %array4 = malloc i32, align 1024 <i>; yields {i32*}:array4</i>
3546 <p>Note that the code generator does not yet respect the
3547 alignment value.</p>
3551 <!-- _______________________________________________________________________ -->
3552 <div class="doc_subsubsection">
3553 <a name="i_free">'<tt>free</tt>' Instruction</a>
3556 <div class="doc_text">
3561 free <type> <value> <i>; yields {void}</i>
3566 <p>The '<tt>free</tt>' instruction returns memory back to the unused
3567 memory heap to be reallocated in the future.</p>
3571 <p>'<tt>value</tt>' shall be a pointer value that points to a value
3572 that was allocated with the '<tt><a href="#i_malloc">malloc</a></tt>'
3577 <p>Access to the memory pointed to by the pointer is no longer defined
3578 after this instruction executes. If the pointer is null, the operation
3584 %array = <a href="#i_malloc">malloc</a> [4 x i8] <i>; yields {[4 x i8]*}:array</i>
3585 free [4 x i8]* %array
3589 <!-- _______________________________________________________________________ -->
3590 <div class="doc_subsubsection">
3591 <a name="i_alloca">'<tt>alloca</tt>' Instruction</a>
3594 <div class="doc_text">
3599 <result> = alloca <type>[, i32 <NumElements>][, align <alignment>] <i>; yields {type*}:result</i>
3604 <p>The '<tt>alloca</tt>' instruction allocates memory on the stack frame of the
3605 currently executing function, to be automatically released when this function
3606 returns to its caller. The object is always allocated in the generic address
3607 space (address space zero).</p>
3611 <p>The '<tt>alloca</tt>' instruction allocates <tt>sizeof(<type>)*NumElements</tt>
3612 bytes of memory on the runtime stack, returning a pointer of the
3613 appropriate type to the program. If "NumElements" is specified, it is the
3614 number of elements allocated, otherwise "NumElements" is defaulted to be one.
3615 If a constant alignment is specified, the value result of the allocation is guaranteed
3616 to be aligned to at least that boundary. If not specified, or if zero, the target
3617 can choose to align the allocation on any convenient boundary.</p>
3619 <p>'<tt>type</tt>' may be any sized type.</p>
3623 <p>Memory is allocated; a pointer is returned. The operation is undefined if
3624 there is insufficient stack space for the allocation. '<tt>alloca</tt>'d
3625 memory is automatically released when the function returns. The '<tt>alloca</tt>'
3626 instruction is commonly used to represent automatic variables that must
3627 have an address available. When the function returns (either with the <tt><a
3628 href="#i_ret">ret</a></tt> or <tt><a href="#i_unwind">unwind</a></tt>
3629 instructions), the memory is reclaimed. Allocating zero bytes
3630 is legal, but the result is undefined.</p>
3635 %ptr = alloca i32 <i>; yields {i32*}:ptr</i>
3636 %ptr = alloca i32, i32 4 <i>; yields {i32*}:ptr</i>
3637 %ptr = alloca i32, i32 4, align 1024 <i>; yields {i32*}:ptr</i>
3638 %ptr = alloca i32, align 1024 <i>; yields {i32*}:ptr</i>
3642 <!-- _______________________________________________________________________ -->
3643 <div class="doc_subsubsection"> <a name="i_load">'<tt>load</tt>'
3644 Instruction</a> </div>
3645 <div class="doc_text">
3647 <pre> <result> = load <ty>* <pointer>[, align <alignment>]<br> <result> = volatile load <ty>* <pointer>[, align <alignment>]<br></pre>
3649 <p>The '<tt>load</tt>' instruction is used to read from memory.</p>
3651 <p>The argument to the '<tt>load</tt>' instruction specifies the memory
3652 address from which to load. The pointer must point to a <a
3653 href="#t_firstclass">first class</a> type. If the <tt>load</tt> is
3654 marked as <tt>volatile</tt>, then the optimizer is not allowed to modify
3655 the number or order of execution of this <tt>load</tt> with other
3656 volatile <tt>load</tt> and <tt><a href="#i_store">store</a></tt>
3659 The optional constant "align" argument specifies the alignment of the operation
3660 (that is, the alignment of the memory address). A value of 0 or an
3661 omitted "align" argument means that the operation has the preferential
3662 alignment for the target. It is the responsibility of the code emitter
3663 to ensure that the alignment information is correct. Overestimating
3664 the alignment results in an undefined behavior. Underestimating the
3665 alignment may produce less efficient code. An alignment of 1 is always
3669 <p>The location of memory pointed to is loaded. If the value being loaded
3670 is of scalar type then the number of bytes read does not exceed the minimum
3671 number of bytes needed to hold all bits of the type. For example, loading an
3672 <tt>i24</tt> reads at most three bytes. When loading a value of a type like
3673 <tt>i20</tt> with a size that is not an integral number of bytes, the result
3674 is undefined if the value was not originally written using a store of the
3677 <pre> %ptr = <a href="#i_alloca">alloca</a> i32 <i>; yields {i32*}:ptr</i>
3679 href="#i_store">store</a> i32 3, i32* %ptr <i>; yields {void}</i>
3680 %val = load i32* %ptr <i>; yields {i32}:val = i32 3</i>
3683 <!-- _______________________________________________________________________ -->
3684 <div class="doc_subsubsection"> <a name="i_store">'<tt>store</tt>'
3685 Instruction</a> </div>
3686 <div class="doc_text">
3688 <pre> store <ty> <value>, <ty>* <pointer>[, align <alignment>] <i>; yields {void}</i>
3689 volatile store <ty> <value>, <ty>* <pointer>[, align <alignment>] <i>; yields {void}</i>
3692 <p>The '<tt>store</tt>' instruction is used to write to memory.</p>
3694 <p>There are two arguments to the '<tt>store</tt>' instruction: a value
3695 to store and an address at which to store it. The type of the '<tt><pointer></tt>'
3696 operand must be a pointer to the <a href="#t_firstclass">first class</a> type
3697 of the '<tt><value></tt>'
3698 operand. If the <tt>store</tt> is marked as <tt>volatile</tt>, then the
3699 optimizer is not allowed to modify the number or order of execution of
3700 this <tt>store</tt> with other volatile <tt>load</tt> and <tt><a
3701 href="#i_store">store</a></tt> instructions.</p>
3703 The optional constant "align" argument specifies the alignment of the operation
3704 (that is, the alignment of the memory address). A value of 0 or an
3705 omitted "align" argument means that the operation has the preferential
3706 alignment for the target. It is the responsibility of the code emitter
3707 to ensure that the alignment information is correct. Overestimating
3708 the alignment results in an undefined behavior. Underestimating the
3709 alignment may produce less efficient code. An alignment of 1 is always
3713 <p>The contents of memory are updated to contain '<tt><value></tt>'
3714 at the location specified by the '<tt><pointer></tt>' operand.
3715 If '<tt><value></tt>' is of scalar type then the number of bytes
3716 written does not exceed the minimum number of bytes needed to hold all
3717 bits of the type. For example, storing an <tt>i24</tt> writes at most
3718 three bytes. When writing a value of a type like <tt>i20</tt> with a
3719 size that is not an integral number of bytes, it is unspecified what
3720 happens to the extra bits that do not belong to the type, but they will
3721 typically be overwritten.</p>
3723 <pre> %ptr = <a href="#i_alloca">alloca</a> i32 <i>; yields {i32*}:ptr</i>
3724 store i32 3, i32* %ptr <i>; yields {void}</i>
3725 %val = <a href="#i_load">load</a> i32* %ptr <i>; yields {i32}:val = i32 3</i>
3729 <!-- _______________________________________________________________________ -->
3730 <div class="doc_subsubsection">
3731 <a name="i_getelementptr">'<tt>getelementptr</tt>' Instruction</a>
3734 <div class="doc_text">
3737 <result> = getelementptr <pty>* <ptrval>{, <ty> <idx>}*
3743 The '<tt>getelementptr</tt>' instruction is used to get the address of a
3744 subelement of an aggregate data structure. It performs address calculation only
3745 and does not access memory.</p>
3749 <p>The first argument is always a pointer, and forms the basis of the
3750 calculation. The remaining arguments are indices, that indicate which of the
3751 elements of the aggregate object are indexed. The interpretation of each index
3752 is dependent on the type being indexed into. The first index always indexes the
3753 pointer value given as the first argument, the second index indexes a value of
3754 the type pointed to (not necessarily the value directly pointed to, since the
3755 first index can be non-zero), etc. The first type indexed into must be a pointer
3756 value, subsequent types can be arrays, vectors and structs. Note that subsequent
3757 types being indexed into can never be pointers, since that would require loading
3758 the pointer before continuing calculation.</p>
3760 <p>The type of each index argument depends on the type it is indexing into.
3761 When indexing into a (packed) structure, only <tt>i32</tt> integer
3762 <b>constants</b> are allowed. When indexing into an array, pointer or vector,
3763 integers of any width are allowed (also non-constants).</p>
3765 <p>For example, let's consider a C code fragment and how it gets
3766 compiled to LLVM:</p>
3768 <div class="doc_code">
3781 int *foo(struct ST *s) {
3782 return &s[1].Z.B[5][13];
3787 <p>The LLVM code generated by the GCC frontend is:</p>
3789 <div class="doc_code">
3791 %RT = <a href="#namedtypes">type</a> { i8 , [10 x [20 x i32]], i8 }
3792 %ST = <a href="#namedtypes">type</a> { i32, double, %RT }
3794 define i32* %foo(%ST* %s) {
3796 %reg = getelementptr %ST* %s, i32 1, i32 2, i32 1, i32 5, i32 13
3804 <p>In the example above, the first index is indexing into the '<tt>%ST*</tt>'
3805 type, which is a pointer, yielding a '<tt>%ST</tt>' = '<tt>{ i32, double, %RT
3806 }</tt>' type, a structure. The second index indexes into the third element of
3807 the structure, yielding a '<tt>%RT</tt>' = '<tt>{ i8 , [10 x [20 x i32]],
3808 i8 }</tt>' type, another structure. The third index indexes into the second
3809 element of the structure, yielding a '<tt>[10 x [20 x i32]]</tt>' type, an
3810 array. The two dimensions of the array are subscripted into, yielding an
3811 '<tt>i32</tt>' type. The '<tt>getelementptr</tt>' instruction returns a pointer
3812 to this element, thus computing a value of '<tt>i32*</tt>' type.</p>
3814 <p>Note that it is perfectly legal to index partially through a
3815 structure, returning a pointer to an inner element. Because of this,
3816 the LLVM code for the given testcase is equivalent to:</p>
3819 define i32* %foo(%ST* %s) {
3820 %t1 = getelementptr %ST* %s, i32 1 <i>; yields %ST*:%t1</i>
3821 %t2 = getelementptr %ST* %t1, i32 0, i32 2 <i>; yields %RT*:%t2</i>
3822 %t3 = getelementptr %RT* %t2, i32 0, i32 1 <i>; yields [10 x [20 x i32]]*:%t3</i>
3823 %t4 = getelementptr [10 x [20 x i32]]* %t3, i32 0, i32 5 <i>; yields [20 x i32]*:%t4</i>
3824 %t5 = getelementptr [20 x i32]* %t4, i32 0, i32 13 <i>; yields i32*:%t5</i>
3829 <p>Note that it is undefined to access an array out of bounds: array
3830 and pointer indexes must always be within the defined bounds of the
3831 array type when accessed with an instruction that dereferences the
3832 pointer (e.g. a load or store instruction). The one exception for
3833 this rule is zero length arrays. These arrays are defined to be
3834 accessible as variable length arrays, which requires access beyond the
3835 zero'th element.</p>
3837 <p>The getelementptr instruction is often confusing. For some more insight
3838 into how it works, see <a href="GetElementPtr.html">the getelementptr
3844 <i>; yields [12 x i8]*:aptr</i>
3845 %aptr = getelementptr {i32, [12 x i8]}* %saptr, i64 0, i32 1
3846 <i>; yields i8*:vptr</i>
3847 %vptr = getelementptr {i32, <2 x i8>}* %svptr, i64 0, i32 1, i32 1
3848 <i>; yields i8*:eptr</i>
3849 %eptr = getelementptr [12 x i8]* %aptr, i64 0, i32 1
3850 <i>; yields i32*:iptr</i>
3851 %iptr = getelementptr [10 x i32]* @arr, i16 0, i16 0
3855 <!-- ======================================================================= -->
3856 <div class="doc_subsection"> <a name="convertops">Conversion Operations</a>
3858 <div class="doc_text">
3859 <p>The instructions in this category are the conversion instructions (casting)
3860 which all take a single operand and a type. They perform various bit conversions
3864 <!-- _______________________________________________________________________ -->
3865 <div class="doc_subsubsection">
3866 <a name="i_trunc">'<tt>trunc .. to</tt>' Instruction</a>
3868 <div class="doc_text">
3872 <result> = trunc <ty> <value> to <ty2> <i>; yields ty2</i>
3877 The '<tt>trunc</tt>' instruction truncates its operand to the type <tt>ty2</tt>.
3882 The '<tt>trunc</tt>' instruction takes a <tt>value</tt> to trunc, which must
3883 be an <a href="#t_integer">integer</a> type, and a type that specifies the size
3884 and type of the result, which must be an <a href="#t_integer">integer</a>
3885 type. The bit size of <tt>value</tt> must be larger than the bit size of
3886 <tt>ty2</tt>. Equal sized types are not allowed.</p>
3890 The '<tt>trunc</tt>' instruction truncates the high order bits in <tt>value</tt>
3891 and converts the remaining bits to <tt>ty2</tt>. Since the source size must be
3892 larger than the destination size, <tt>trunc</tt> cannot be a <i>no-op cast</i>.
3893 It will always truncate bits.</p>
3897 %X = trunc i32 257 to i8 <i>; yields i8:1</i>
3898 %Y = trunc i32 123 to i1 <i>; yields i1:true</i>
3899 %Y = trunc i32 122 to i1 <i>; yields i1:false</i>
3903 <!-- _______________________________________________________________________ -->
3904 <div class="doc_subsubsection">
3905 <a name="i_zext">'<tt>zext .. to</tt>' Instruction</a>
3907 <div class="doc_text">
3911 <result> = zext <ty> <value> to <ty2> <i>; yields ty2</i>
3915 <p>The '<tt>zext</tt>' instruction zero extends its operand to type
3920 <p>The '<tt>zext</tt>' instruction takes a value to cast, which must be of
3921 <a href="#t_integer">integer</a> type, and a type to cast it to, which must
3922 also be of <a href="#t_integer">integer</a> type. The bit size of the
3923 <tt>value</tt> must be smaller than the bit size of the destination type,
3927 <p>The <tt>zext</tt> fills the high order bits of the <tt>value</tt> with zero
3928 bits until it reaches the size of the destination type, <tt>ty2</tt>.</p>
3930 <p>When zero extending from i1, the result will always be either 0 or 1.</p>
3934 %X = zext i32 257 to i64 <i>; yields i64:257</i>
3935 %Y = zext i1 true to i32 <i>; yields i32:1</i>
3939 <!-- _______________________________________________________________________ -->
3940 <div class="doc_subsubsection">
3941 <a name="i_sext">'<tt>sext .. to</tt>' Instruction</a>
3943 <div class="doc_text">
3947 <result> = sext <ty> <value> to <ty2> <i>; yields ty2</i>
3951 <p>The '<tt>sext</tt>' sign extends <tt>value</tt> to the type <tt>ty2</tt>.</p>
3955 The '<tt>sext</tt>' instruction takes a value to cast, which must be of
3956 <a href="#t_integer">integer</a> type, and a type to cast it to, which must
3957 also be of <a href="#t_integer">integer</a> type. The bit size of the
3958 <tt>value</tt> must be smaller than the bit size of the destination type,
3963 The '<tt>sext</tt>' instruction performs a sign extension by copying the sign
3964 bit (highest order bit) of the <tt>value</tt> until it reaches the bit size of
3965 the type <tt>ty2</tt>.</p>
3967 <p>When sign extending from i1, the extension always results in -1 or 0.</p>
3971 %X = sext i8 -1 to i16 <i>; yields i16 :65535</i>
3972 %Y = sext i1 true to i32 <i>; yields i32:-1</i>
3976 <!-- _______________________________________________________________________ -->
3977 <div class="doc_subsubsection">
3978 <a name="i_fptrunc">'<tt>fptrunc .. to</tt>' Instruction</a>
3981 <div class="doc_text">
3986 <result> = fptrunc <ty> <value> to <ty2> <i>; yields ty2</i>
3990 <p>The '<tt>fptrunc</tt>' instruction truncates <tt>value</tt> to type
3995 <p>The '<tt>fptrunc</tt>' instruction takes a <a href="#t_floating">floating
3996 point</a> value to cast and a <a href="#t_floating">floating point</a> type to
3997 cast it to. The size of <tt>value</tt> must be larger than the size of
3998 <tt>ty2</tt>. This implies that <tt>fptrunc</tt> cannot be used to make a
3999 <i>no-op cast</i>.</p>
4002 <p> The '<tt>fptrunc</tt>' instruction truncates a <tt>value</tt> from a larger
4003 <a href="#t_floating">floating point</a> type to a smaller
4004 <a href="#t_floating">floating point</a> type. If the value cannot fit within
4005 the destination type, <tt>ty2</tt>, then the results are undefined.</p>
4009 %X = fptrunc double 123.0 to float <i>; yields float:123.0</i>
4010 %Y = fptrunc double 1.0E+300 to float <i>; yields undefined</i>
4014 <!-- _______________________________________________________________________ -->
4015 <div class="doc_subsubsection">
4016 <a name="i_fpext">'<tt>fpext .. to</tt>' Instruction</a>
4018 <div class="doc_text">
4022 <result> = fpext <ty> <value> to <ty2> <i>; yields ty2</i>
4026 <p>The '<tt>fpext</tt>' extends a floating point <tt>value</tt> to a larger
4027 floating point value.</p>
4030 <p>The '<tt>fpext</tt>' instruction takes a
4031 <a href="#t_floating">floating point</a> <tt>value</tt> to cast,
4032 and a <a href="#t_floating">floating point</a> type to cast it to. The source
4033 type must be smaller than the destination type.</p>
4036 <p>The '<tt>fpext</tt>' instruction extends the <tt>value</tt> from a smaller
4037 <a href="#t_floating">floating point</a> type to a larger
4038 <a href="#t_floating">floating point</a> type. The <tt>fpext</tt> cannot be
4039 used to make a <i>no-op cast</i> because it always changes bits. Use
4040 <tt>bitcast</tt> to make a <i>no-op cast</i> for a floating point cast.</p>
4044 %X = fpext float 3.1415 to double <i>; yields double:3.1415</i>
4045 %Y = fpext float 1.0 to float <i>; yields float:1.0 (no-op)</i>
4049 <!-- _______________________________________________________________________ -->
4050 <div class="doc_subsubsection">
4051 <a name="i_fptoui">'<tt>fptoui .. to</tt>' Instruction</a>
4053 <div class="doc_text">
4057 <result> = fptoui <ty> <value> to <ty2> <i>; yields ty2</i>
4061 <p>The '<tt>fptoui</tt>' converts a floating point <tt>value</tt> to its
4062 unsigned integer equivalent of type <tt>ty2</tt>.
4066 <p>The '<tt>fptoui</tt>' instruction takes a value to cast, which must be a
4067 scalar or vector <a href="#t_floating">floating point</a> value, and a type
4068 to cast it to <tt>ty2</tt>, which must be an <a href="#t_integer">integer</a>
4069 type. If <tt>ty</tt> is a vector floating point type, <tt>ty2</tt> must be a
4070 vector integer type with the same number of elements as <tt>ty</tt></p>
4073 <p> The '<tt>fptoui</tt>' instruction converts its
4074 <a href="#t_floating">floating point</a> operand into the nearest (rounding
4075 towards zero) unsigned integer value. If the value cannot fit in <tt>ty2</tt>,
4076 the results are undefined.</p>
4080 %X = fptoui double 123.0 to i32 <i>; yields i32:123</i>
4081 %Y = fptoui float 1.0E+300 to i1 <i>; yields undefined:1</i>
4082 %X = fptoui float 1.04E+17 to i8 <i>; yields undefined:1</i>
4086 <!-- _______________________________________________________________________ -->
4087 <div class="doc_subsubsection">
4088 <a name="i_fptosi">'<tt>fptosi .. to</tt>' Instruction</a>
4090 <div class="doc_text">
4094 <result> = fptosi <ty> <value> to <ty2> <i>; yields ty2</i>
4098 <p>The '<tt>fptosi</tt>' instruction converts
4099 <a href="#t_floating">floating point</a> <tt>value</tt> to type <tt>ty2</tt>.
4103 <p> The '<tt>fptosi</tt>' instruction takes a value to cast, which must be a
4104 scalar or vector <a href="#t_floating">floating point</a> value, and a type
4105 to cast it to <tt>ty2</tt>, which must be an <a href="#t_integer">integer</a>
4106 type. If <tt>ty</tt> is a vector floating point type, <tt>ty2</tt> must be a
4107 vector integer type with the same number of elements as <tt>ty</tt></p>
4110 <p>The '<tt>fptosi</tt>' instruction converts its
4111 <a href="#t_floating">floating point</a> operand into the nearest (rounding
4112 towards zero) signed integer value. If the value cannot fit in <tt>ty2</tt>,
4113 the results are undefined.</p>
4117 %X = fptosi double -123.0 to i32 <i>; yields i32:-123</i>
4118 %Y = fptosi float 1.0E-247 to i1 <i>; yields undefined:1</i>
4119 %X = fptosi float 1.04E+17 to i8 <i>; yields undefined:1</i>
4123 <!-- _______________________________________________________________________ -->
4124 <div class="doc_subsubsection">
4125 <a name="i_uitofp">'<tt>uitofp .. to</tt>' Instruction</a>
4127 <div class="doc_text">
4131 <result> = uitofp <ty> <value> to <ty2> <i>; yields ty2</i>
4135 <p>The '<tt>uitofp</tt>' instruction regards <tt>value</tt> as an unsigned
4136 integer and converts that value to the <tt>ty2</tt> type.</p>
4139 <p>The '<tt>uitofp</tt>' instruction takes a value to cast, which must be a
4140 scalar or vector <a href="#t_integer">integer</a> value, and a type to cast it
4141 to <tt>ty2</tt>, which must be an <a href="#t_floating">floating point</a>
4142 type. If <tt>ty</tt> is a vector integer type, <tt>ty2</tt> must be a vector
4143 floating point type with the same number of elements as <tt>ty</tt></p>
4146 <p>The '<tt>uitofp</tt>' instruction interprets its operand as an unsigned
4147 integer quantity and converts it to the corresponding floating point value. If
4148 the value cannot fit in the floating point value, the results are undefined.</p>
4152 %X = uitofp i32 257 to float <i>; yields float:257.0</i>
4153 %Y = uitofp i8 -1 to double <i>; yields double:255.0</i>
4157 <!-- _______________________________________________________________________ -->
4158 <div class="doc_subsubsection">
4159 <a name="i_sitofp">'<tt>sitofp .. to</tt>' Instruction</a>
4161 <div class="doc_text">
4165 <result> = sitofp <ty> <value> to <ty2> <i>; yields ty2</i>
4169 <p>The '<tt>sitofp</tt>' instruction regards <tt>value</tt> as a signed
4170 integer and converts that value to the <tt>ty2</tt> type.</p>
4173 <p>The '<tt>sitofp</tt>' instruction takes a value to cast, which must be a
4174 scalar or vector <a href="#t_integer">integer</a> value, and a type to cast it
4175 to <tt>ty2</tt>, which must be an <a href="#t_floating">floating point</a>
4176 type. If <tt>ty</tt> is a vector integer type, <tt>ty2</tt> must be a vector
4177 floating point type with the same number of elements as <tt>ty</tt></p>
4180 <p>The '<tt>sitofp</tt>' instruction interprets its operand as a signed
4181 integer quantity and converts it to the corresponding floating point value. If
4182 the value cannot fit in the floating point value, the results are undefined.</p>
4186 %X = sitofp i32 257 to float <i>; yields float:257.0</i>
4187 %Y = sitofp i8 -1 to double <i>; yields double:-1.0</i>
4191 <!-- _______________________________________________________________________ -->
4192 <div class="doc_subsubsection">
4193 <a name="i_ptrtoint">'<tt>ptrtoint .. to</tt>' Instruction</a>
4195 <div class="doc_text">
4199 <result> = ptrtoint <ty> <value> to <ty2> <i>; yields ty2</i>
4203 <p>The '<tt>ptrtoint</tt>' instruction converts the pointer <tt>value</tt> to
4204 the integer type <tt>ty2</tt>.</p>
4207 <p>The '<tt>ptrtoint</tt>' instruction takes a <tt>value</tt> to cast, which
4208 must be a <a href="#t_pointer">pointer</a> value, and a type to cast it to
4209 <tt>ty2</tt>, which must be an <a href="#t_integer">integer</a> type.</p>
4212 <p>The '<tt>ptrtoint</tt>' instruction converts <tt>value</tt> to integer type
4213 <tt>ty2</tt> by interpreting the pointer value as an integer and either
4214 truncating or zero extending that value to the size of the integer type. If
4215 <tt>value</tt> is smaller than <tt>ty2</tt> then a zero extension is done. If
4216 <tt>value</tt> is larger than <tt>ty2</tt> then a truncation is done. If they
4217 are the same size, then nothing is done (<i>no-op cast</i>) other than a type
4222 %X = ptrtoint i32* %X to i8 <i>; yields truncation on 32-bit architecture</i>
4223 %Y = ptrtoint i32* %x to i64 <i>; yields zero extension on 32-bit architecture</i>
4227 <!-- _______________________________________________________________________ -->
4228 <div class="doc_subsubsection">
4229 <a name="i_inttoptr">'<tt>inttoptr .. to</tt>' Instruction</a>
4231 <div class="doc_text">
4235 <result> = inttoptr <ty> <value> to <ty2> <i>; yields ty2</i>
4239 <p>The '<tt>inttoptr</tt>' instruction converts an integer <tt>value</tt> to
4240 a pointer type, <tt>ty2</tt>.</p>
4243 <p>The '<tt>inttoptr</tt>' instruction takes an <a href="#t_integer">integer</a>
4244 value to cast, and a type to cast it to, which must be a
4245 <a href="#t_pointer">pointer</a> type.</p>
4248 <p>The '<tt>inttoptr</tt>' instruction converts <tt>value</tt> to type
4249 <tt>ty2</tt> by applying either a zero extension or a truncation depending on
4250 the size of the integer <tt>value</tt>. If <tt>value</tt> is larger than the
4251 size of a pointer then a truncation is done. If <tt>value</tt> is smaller than
4252 the size of a pointer then a zero extension is done. If they are the same size,
4253 nothing is done (<i>no-op cast</i>).</p>
4257 %X = inttoptr i32 255 to i32* <i>; yields zero extension on 64-bit architecture</i>
4258 %X = inttoptr i32 255 to i32* <i>; yields no-op on 32-bit architecture</i>
4259 %Y = inttoptr i64 0 to i32* <i>; yields truncation on 32-bit architecture</i>
4263 <!-- _______________________________________________________________________ -->
4264 <div class="doc_subsubsection">
4265 <a name="i_bitcast">'<tt>bitcast .. to</tt>' Instruction</a>
4267 <div class="doc_text">
4271 <result> = bitcast <ty> <value> to <ty2> <i>; yields ty2</i>
4276 <p>The '<tt>bitcast</tt>' instruction converts <tt>value</tt> to type
4277 <tt>ty2</tt> without changing any bits.</p>
4281 <p>The '<tt>bitcast</tt>' instruction takes a value to cast, which must be
4282 a non-aggregate first class value, and a type to cast it to, which must also be
4283 a non-aggregate <a href="#t_firstclass">first class</a> type. The bit sizes of
4285 and the destination type, <tt>ty2</tt>, must be identical. If the source
4286 type is a pointer, the destination type must also be a pointer. This
4287 instruction supports bitwise conversion of vectors to integers and to vectors
4288 of other types (as long as they have the same size).</p>
4291 <p>The '<tt>bitcast</tt>' instruction converts <tt>value</tt> to type
4292 <tt>ty2</tt>. It is always a <i>no-op cast</i> because no bits change with
4293 this conversion. The conversion is done as if the <tt>value</tt> had been
4294 stored to memory and read back as type <tt>ty2</tt>. Pointer types may only be
4295 converted to other pointer types with this instruction. To convert pointers to
4296 other types, use the <a href="#i_inttoptr">inttoptr</a> or
4297 <a href="#i_ptrtoint">ptrtoint</a> instructions first.</p>
4301 %X = bitcast i8 255 to i8 <i>; yields i8 :-1</i>
4302 %Y = bitcast i32* %x to sint* <i>; yields sint*:%x</i>
4303 %Z = bitcast <2 x int> %V to i64; <i>; yields i64: %V</i>
4307 <!-- ======================================================================= -->
4308 <div class="doc_subsection"> <a name="otherops">Other Operations</a> </div>
4309 <div class="doc_text">
4310 <p>The instructions in this category are the "miscellaneous"
4311 instructions, which defy better classification.</p>
4314 <!-- _______________________________________________________________________ -->
4315 <div class="doc_subsubsection"><a name="i_icmp">'<tt>icmp</tt>' Instruction</a>
4317 <div class="doc_text">
4319 <pre> <result> = icmp <cond> <ty> <op1>, <op2> <i>; yields {i1} or {<N x i1>}:result</i>
4322 <p>The '<tt>icmp</tt>' instruction returns a boolean value or
4323 a vector of boolean values based on comparison
4324 of its two integer, integer vector, or pointer operands.</p>
4326 <p>The '<tt>icmp</tt>' instruction takes three operands. The first operand is
4327 the condition code indicating the kind of comparison to perform. It is not
4328 a value, just a keyword. The possible condition code are:
4331 <li><tt>eq</tt>: equal</li>
4332 <li><tt>ne</tt>: not equal </li>
4333 <li><tt>ugt</tt>: unsigned greater than</li>
4334 <li><tt>uge</tt>: unsigned greater or equal</li>
4335 <li><tt>ult</tt>: unsigned less than</li>
4336 <li><tt>ule</tt>: unsigned less or equal</li>
4337 <li><tt>sgt</tt>: signed greater than</li>
4338 <li><tt>sge</tt>: signed greater or equal</li>
4339 <li><tt>slt</tt>: signed less than</li>
4340 <li><tt>sle</tt>: signed less or equal</li>
4342 <p>The remaining two arguments must be <a href="#t_integer">integer</a> or
4343 <a href="#t_pointer">pointer</a>
4344 or integer <a href="#t_vector">vector</a> typed.
4345 They must also be identical types.</p>
4347 <p>The '<tt>icmp</tt>' compares <tt>op1</tt> and <tt>op2</tt> according to
4348 the condition code given as <tt>cond</tt>. The comparison performed always
4349 yields either an <a href="#t_primitive"><tt>i1</tt></a> or vector of <tt>i1</tt> result, as follows:
4352 <li><tt>eq</tt>: yields <tt>true</tt> if the operands are equal,
4353 <tt>false</tt> otherwise. No sign interpretation is necessary or performed.
4355 <li><tt>ne</tt>: yields <tt>true</tt> if the operands are unequal,
4356 <tt>false</tt> otherwise. No sign interpretation is necessary or performed.</li>
4357 <li><tt>ugt</tt>: interprets the operands as unsigned values and yields
4358 <tt>true</tt> if <tt>op1</tt> is greater than <tt>op2</tt>.</li>
4359 <li><tt>uge</tt>: interprets the operands as unsigned values and yields
4360 <tt>true</tt> if <tt>op1</tt> is greater than or equal to <tt>op2</tt>.</li>
4361 <li><tt>ult</tt>: interprets the operands as unsigned values and yields
4362 <tt>true</tt> if <tt>op1</tt> is less than <tt>op2</tt>.</li>
4363 <li><tt>ule</tt>: interprets the operands as unsigned values and yields
4364 <tt>true</tt> if <tt>op1</tt> is less than or equal to <tt>op2</tt>.</li>
4365 <li><tt>sgt</tt>: interprets the operands as signed values and yields
4366 <tt>true</tt> if <tt>op1</tt> is greater than <tt>op2</tt>.</li>
4367 <li><tt>sge</tt>: interprets the operands as signed values and yields
4368 <tt>true</tt> if <tt>op1</tt> is greater than or equal to <tt>op2</tt>.</li>
4369 <li><tt>slt</tt>: interprets the operands as signed values and yields
4370 <tt>true</tt> if <tt>op1</tt> is less than <tt>op2</tt>.</li>
4371 <li><tt>sle</tt>: interprets the operands as signed values and yields
4372 <tt>true</tt> if <tt>op1</tt> is less than or equal to <tt>op2</tt>.</li>
4374 <p>If the operands are <a href="#t_pointer">pointer</a> typed, the pointer
4375 values are compared as if they were integers.</p>
4376 <p>If the operands are integer vectors, then they are compared
4377 element by element. The result is an <tt>i1</tt> vector with
4378 the same number of elements as the values being compared.
4379 Otherwise, the result is an <tt>i1</tt>.
4383 <pre> <result> = icmp eq i32 4, 5 <i>; yields: result=false</i>
4384 <result> = icmp ne float* %X, %X <i>; yields: result=false</i>
4385 <result> = icmp ult i16 4, 5 <i>; yields: result=true</i>
4386 <result> = icmp sgt i16 4, 5 <i>; yields: result=false</i>
4387 <result> = icmp ule i16 -4, 5 <i>; yields: result=false</i>
4388 <result> = icmp sge i16 4, 5 <i>; yields: result=false</i>
4391 <p>Note that the code generator does not yet support vector types with
4392 the <tt>icmp</tt> instruction.</p>
4396 <!-- _______________________________________________________________________ -->
4397 <div class="doc_subsubsection"><a name="i_fcmp">'<tt>fcmp</tt>' Instruction</a>
4399 <div class="doc_text">
4401 <pre> <result> = fcmp <cond> <ty> <op1>, <op2> <i>; yields {i1} or {<N x i1>}:result</i>
4404 <p>The '<tt>fcmp</tt>' instruction returns a boolean value
4405 or vector of boolean values based on comparison
4406 of its operands.</p>
4408 If the operands are floating point scalars, then the result
4409 type is a boolean (<a href="#t_primitive"><tt>i1</tt></a>).
4411 <p>If the operands are floating point vectors, then the result type
4412 is a vector of boolean with the same number of elements as the
4413 operands being compared.</p>
4415 <p>The '<tt>fcmp</tt>' instruction takes three operands. The first operand is
4416 the condition code indicating the kind of comparison to perform. It is not
4417 a value, just a keyword. The possible condition code are:</p>
4419 <li><tt>false</tt>: no comparison, always returns false</li>
4420 <li><tt>oeq</tt>: ordered and equal</li>
4421 <li><tt>ogt</tt>: ordered and greater than </li>
4422 <li><tt>oge</tt>: ordered and greater than or equal</li>
4423 <li><tt>olt</tt>: ordered and less than </li>
4424 <li><tt>ole</tt>: ordered and less than or equal</li>
4425 <li><tt>one</tt>: ordered and not equal</li>
4426 <li><tt>ord</tt>: ordered (no nans)</li>
4427 <li><tt>ueq</tt>: unordered or equal</li>
4428 <li><tt>ugt</tt>: unordered or greater than </li>
4429 <li><tt>uge</tt>: unordered or greater than or equal</li>
4430 <li><tt>ult</tt>: unordered or less than </li>
4431 <li><tt>ule</tt>: unordered or less than or equal</li>
4432 <li><tt>une</tt>: unordered or not equal</li>
4433 <li><tt>uno</tt>: unordered (either nans)</li>
4434 <li><tt>true</tt>: no comparison, always returns true</li>
4436 <p><i>Ordered</i> means that neither operand is a QNAN while
4437 <i>unordered</i> means that either operand may be a QNAN.</p>
4438 <p>Each of <tt>val1</tt> and <tt>val2</tt> arguments must be
4439 either a <a href="#t_floating">floating point</a> type
4440 or a <a href="#t_vector">vector</a> of floating point type.
4441 They must have identical types.</p>
4443 <p>The '<tt>fcmp</tt>' instruction compares <tt>op1</tt> and <tt>op2</tt>
4444 according to the condition code given as <tt>cond</tt>.
4445 If the operands are vectors, then the vectors are compared
4447 Each comparison performed
4448 always yields an <a href="#t_primitive">i1</a> result, as follows:</p>
4450 <li><tt>false</tt>: always yields <tt>false</tt>, regardless of operands.</li>
4451 <li><tt>oeq</tt>: yields <tt>true</tt> if both operands are not a QNAN and
4452 <tt>op1</tt> is equal to <tt>op2</tt>.</li>
4453 <li><tt>ogt</tt>: yields <tt>true</tt> if both operands are not a QNAN and
4454 <tt>op1</tt> is greather than <tt>op2</tt>.</li>
4455 <li><tt>oge</tt>: yields <tt>true</tt> if both operands are not a QNAN and
4456 <tt>op1</tt> is greater than or equal to <tt>op2</tt>.</li>
4457 <li><tt>olt</tt>: yields <tt>true</tt> if both operands are not a QNAN and
4458 <tt>op1</tt> is less than <tt>op2</tt>.</li>
4459 <li><tt>ole</tt>: yields <tt>true</tt> if both operands are not a QNAN and
4460 <tt>op1</tt> is less than or equal to <tt>op2</tt>.</li>
4461 <li><tt>one</tt>: yields <tt>true</tt> if both operands are not a QNAN and
4462 <tt>op1</tt> is not equal to <tt>op2</tt>.</li>
4463 <li><tt>ord</tt>: yields <tt>true</tt> if both operands are not a QNAN.</li>
4464 <li><tt>ueq</tt>: yields <tt>true</tt> if either operand is a QNAN or
4465 <tt>op1</tt> is equal to <tt>op2</tt>.</li>
4466 <li><tt>ugt</tt>: yields <tt>true</tt> if either operand is a QNAN or
4467 <tt>op1</tt> is greater than <tt>op2</tt>.</li>
4468 <li><tt>uge</tt>: yields <tt>true</tt> if either operand is a QNAN or
4469 <tt>op1</tt> is greater than or equal to <tt>op2</tt>.</li>
4470 <li><tt>ult</tt>: yields <tt>true</tt> if either operand is a QNAN or
4471 <tt>op1</tt> is less than <tt>op2</tt>.</li>
4472 <li><tt>ule</tt>: yields <tt>true</tt> if either operand is a QNAN or
4473 <tt>op1</tt> is less than or equal to <tt>op2</tt>.</li>
4474 <li><tt>une</tt>: yields <tt>true</tt> if either operand is a QNAN or
4475 <tt>op1</tt> is not equal to <tt>op2</tt>.</li>
4476 <li><tt>uno</tt>: yields <tt>true</tt> if either operand is a QNAN.</li>
4477 <li><tt>true</tt>: always yields <tt>true</tt>, regardless of operands.</li>
4481 <pre> <result> = fcmp oeq float 4.0, 5.0 <i>; yields: result=false</i>
4482 <result> = fcmp one float 4.0, 5.0 <i>; yields: result=true</i>
4483 <result> = fcmp olt float 4.0, 5.0 <i>; yields: result=true</i>
4484 <result> = fcmp ueq double 1.0, 2.0 <i>; yields: result=false</i>
4487 <p>Note that the code generator does not yet support vector types with
4488 the <tt>fcmp</tt> instruction.</p>
4492 <!-- _______________________________________________________________________ -->
4493 <div class="doc_subsubsection">
4494 <a name="i_vicmp">'<tt>vicmp</tt>' Instruction</a>
4496 <div class="doc_text">
4498 <pre> <result> = vicmp <cond> <ty> <op1>, <op2> <i>; yields {ty}:result</i>
4501 <p>The '<tt>vicmp</tt>' instruction returns an integer vector value based on
4502 element-wise comparison of its two integer vector operands.</p>
4504 <p>The '<tt>vicmp</tt>' instruction takes three operands. The first operand is
4505 the condition code indicating the kind of comparison to perform. It is not
4506 a value, just a keyword. The possible condition code are:</p>
4508 <li><tt>eq</tt>: equal</li>
4509 <li><tt>ne</tt>: not equal </li>
4510 <li><tt>ugt</tt>: unsigned greater than</li>
4511 <li><tt>uge</tt>: unsigned greater or equal</li>
4512 <li><tt>ult</tt>: unsigned less than</li>
4513 <li><tt>ule</tt>: unsigned less or equal</li>
4514 <li><tt>sgt</tt>: signed greater than</li>
4515 <li><tt>sge</tt>: signed greater or equal</li>
4516 <li><tt>slt</tt>: signed less than</li>
4517 <li><tt>sle</tt>: signed less or equal</li>
4519 <p>The remaining two arguments must be <a href="#t_vector">vector</a> or
4520 <a href="#t_integer">integer</a> typed. They must also be identical types.</p>
4522 <p>The '<tt>vicmp</tt>' instruction compares <tt>op1</tt> and <tt>op2</tt>
4523 according to the condition code given as <tt>cond</tt>. The comparison yields a
4524 <a href="#t_vector">vector</a> of <a href="#t_integer">integer</a> result, of
4525 identical type as the values being compared. The most significant bit in each
4526 element is 1 if the element-wise comparison evaluates to true, and is 0
4527 otherwise. All other bits of the result are undefined. The condition codes
4528 are evaluated identically to the <a href="#i_icmp">'<tt>icmp</tt>'
4529 instruction</a>.</p>
4533 <result> = vicmp eq <2 x i32> < i32 4, i32 0>, < i32 5, i32 0> <i>; yields: result=<2 x i32> < i32 0, i32 -1 ></i>
4534 <result> = vicmp ult <2 x i8 > < i8 1, i8 2>, < i8 2, i8 2 > <i>; yields: result=<2 x i8> < i8 -1, i8 0 ></i>
4538 <!-- _______________________________________________________________________ -->
4539 <div class="doc_subsubsection">
4540 <a name="i_vfcmp">'<tt>vfcmp</tt>' Instruction</a>
4542 <div class="doc_text">
4544 <pre> <result> = vfcmp <cond> <ty> <op1>, <op2></pre>
4546 <p>The '<tt>vfcmp</tt>' instruction returns an integer vector value based on
4547 element-wise comparison of its two floating point vector operands. The output
4548 elements have the same width as the input elements.</p>
4550 <p>The '<tt>vfcmp</tt>' instruction takes three operands. The first operand is
4551 the condition code indicating the kind of comparison to perform. It is not
4552 a value, just a keyword. The possible condition code are:</p>
4554 <li><tt>false</tt>: no comparison, always returns false</li>
4555 <li><tt>oeq</tt>: ordered and equal</li>
4556 <li><tt>ogt</tt>: ordered and greater than </li>
4557 <li><tt>oge</tt>: ordered and greater than or equal</li>
4558 <li><tt>olt</tt>: ordered and less than </li>
4559 <li><tt>ole</tt>: ordered and less than or equal</li>
4560 <li><tt>one</tt>: ordered and not equal</li>
4561 <li><tt>ord</tt>: ordered (no nans)</li>
4562 <li><tt>ueq</tt>: unordered or equal</li>
4563 <li><tt>ugt</tt>: unordered or greater than </li>
4564 <li><tt>uge</tt>: unordered or greater than or equal</li>
4565 <li><tt>ult</tt>: unordered or less than </li>
4566 <li><tt>ule</tt>: unordered or less than or equal</li>
4567 <li><tt>une</tt>: unordered or not equal</li>
4568 <li><tt>uno</tt>: unordered (either nans)</li>
4569 <li><tt>true</tt>: no comparison, always returns true</li>
4571 <p>The remaining two arguments must be <a href="#t_vector">vector</a> of
4572 <a href="#t_floating">floating point</a> typed. They must also be identical
4575 <p>The '<tt>vfcmp</tt>' instruction compares <tt>op1</tt> and <tt>op2</tt>
4576 according to the condition code given as <tt>cond</tt>. The comparison yields a
4577 <a href="#t_vector">vector</a> of <a href="#t_integer">integer</a> result, with
4578 an identical number of elements as the values being compared, and each element
4579 having identical with to the width of the floating point elements. The most
4580 significant bit in each element is 1 if the element-wise comparison evaluates to
4581 true, and is 0 otherwise. All other bits of the result are undefined. The
4582 condition codes are evaluated identically to the
4583 <a href="#i_fcmp">'<tt>fcmp</tt>' instruction</a>.</p>
4587 <i>; yields: result=<2 x i32> < i32 0, i32 -1 ></i>
4588 <result> = vfcmp oeq <2 x float> < float 4, float 0 >, < float 5, float 0 >
4590 <i>; yields: result=<2 x i64> < i64 -1, i64 0 ></i>
4591 <result> = vfcmp ult <2 x double> < double 1, double 2 >, < double 2, double 2>
4595 <!-- _______________________________________________________________________ -->
4596 <div class="doc_subsubsection">
4597 <a name="i_phi">'<tt>phi</tt>' Instruction</a>
4600 <div class="doc_text">
4604 <pre> <result> = phi <ty> [ <val0>, <label0>], ...<br></pre>
4606 <p>The '<tt>phi</tt>' instruction is used to implement the φ node in
4607 the SSA graph representing the function.</p>
4610 <p>The type of the incoming values is specified with the first type
4611 field. After this, the '<tt>phi</tt>' instruction takes a list of pairs
4612 as arguments, with one pair for each predecessor basic block of the
4613 current block. Only values of <a href="#t_firstclass">first class</a>
4614 type may be used as the value arguments to the PHI node. Only labels
4615 may be used as the label arguments.</p>
4617 <p>There must be no non-phi instructions between the start of a basic
4618 block and the PHI instructions: i.e. PHI instructions must be first in
4621 <p>For the purposes of the SSA form, the use of each incoming value is
4622 deemed to occur on the edge from the corresponding predecessor block
4623 to the current block (but after any definition of an '<tt>invoke</tt>'
4624 instruction's return value on the same edge).</p>
4628 <p>At runtime, the '<tt>phi</tt>' instruction logically takes on the value
4629 specified by the pair corresponding to the predecessor basic block that executed
4630 just prior to the current block.</p>
4634 Loop: ; Infinite loop that counts from 0 on up...
4635 %indvar = phi i32 [ 0, %LoopHeader ], [ %nextindvar, %Loop ]
4636 %nextindvar = add i32 %indvar, 1
4641 <!-- _______________________________________________________________________ -->
4642 <div class="doc_subsubsection">
4643 <a name="i_select">'<tt>select</tt>' Instruction</a>
4646 <div class="doc_text">
4651 <result> = select <i>selty</i> <cond>, <ty> <val1>, <ty> <val2> <i>; yields ty</i>
4653 <i>selty</i> is either i1 or {<N x i1>}
4659 The '<tt>select</tt>' instruction is used to choose one value based on a
4660 condition, without branching.
4667 The '<tt>select</tt>' instruction requires an 'i1' value or
4668 a vector of 'i1' values indicating the
4669 condition, and two values of the same <a href="#t_firstclass">first class</a>
4670 type. If the val1/val2 are vectors and
4671 the condition is a scalar, then entire vectors are selected, not
4672 individual elements.
4678 If the condition is an i1 and it evaluates to 1, the instruction returns the first
4679 value argument; otherwise, it returns the second value argument.
4682 If the condition is a vector of i1, then the value arguments must
4683 be vectors of the same size, and the selection is done element
4690 %X = select i1 true, i8 17, i8 42 <i>; yields i8:17</i>
4693 <p>Note that the code generator does not yet support conditions
4694 with vector type.</p>
4699 <!-- _______________________________________________________________________ -->
4700 <div class="doc_subsubsection">
4701 <a name="i_call">'<tt>call</tt>' Instruction</a>
4704 <div class="doc_text">
4708 <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>]
4713 <p>The '<tt>call</tt>' instruction represents a simple function call.</p>
4717 <p>This instruction requires several arguments:</p>
4721 <p>The optional "tail" marker indicates whether the callee function accesses
4722 any allocas or varargs in the caller. If the "tail" marker is present, the
4723 function call is eligible for tail call optimization. Note that calls may
4724 be marked "tail" even if they do not occur before a <a
4725 href="#i_ret"><tt>ret</tt></a> instruction.</p>
4728 <p>The optional "cconv" marker indicates which <a href="#callingconv">calling
4729 convention</a> the call should use. If none is specified, the call defaults
4730 to using C calling conventions.</p>
4734 <p>The optional <a href="#paramattrs">Parameter Attributes</a> list for
4735 return values. Only '<tt>zeroext</tt>', '<tt>signext</tt>',
4736 and '<tt>inreg</tt>' attributes are valid here.</p>
4740 <p>'<tt>ty</tt>': the type of the call instruction itself which is also
4741 the type of the return value. Functions that return no value are marked
4742 <tt><a href="#t_void">void</a></tt>.</p>
4745 <p>'<tt>fnty</tt>': shall be the signature of the pointer to function
4746 value being invoked. The argument types must match the types implied by
4747 this signature. This type can be omitted if the function is not varargs
4748 and if the function type does not return a pointer to a function.</p>
4751 <p>'<tt>fnptrval</tt>': An LLVM value containing a pointer to a function to
4752 be invoked. In most cases, this is a direct function invocation, but
4753 indirect <tt>call</tt>s are just as possible, calling an arbitrary pointer
4754 to function value.</p>
4757 <p>'<tt>function args</tt>': argument list whose types match the
4758 function signature argument types. All arguments must be of
4759 <a href="#t_firstclass">first class</a> type. If the function signature
4760 indicates the function accepts a variable number of arguments, the extra
4761 arguments can be specified.</p>
4764 <p>The optional <a href="#fnattrs">function attributes</a> list. Only
4765 '<tt>noreturn</tt>', '<tt>nounwind</tt>', '<tt>readonly</tt>' and
4766 '<tt>readnone</tt>' attributes are valid here.</p>
4772 <p>The '<tt>call</tt>' instruction is used to cause control flow to
4773 transfer to a specified function, with its incoming arguments bound to
4774 the specified values. Upon a '<tt><a href="#i_ret">ret</a></tt>'
4775 instruction in the called function, control flow continues with the
4776 instruction after the function call, and the return value of the
4777 function is bound to the result argument.</p>
4782 %retval = call i32 @test(i32 %argc)
4783 call i32 (i8 *, ...)* @printf(i8 * %msg, i32 12, i8 42) <i>; yields i32</i>
4784 %X = tail call i32 @foo() <i>; yields i32</i>
4785 %Y = tail call <a href="#callingconv">fastcc</a> i32 @foo() <i>; yields i32</i>
4786 call void %foo(i8 97 signext)
4788 %struct.A = type { i32, i8 }
4789 %r = call %struct.A @foo() <i>; yields { 32, i8 }</i>
4790 %gr = extractvalue %struct.A %r, 0 <i>; yields i32</i>
4791 %gr1 = extractvalue %struct.A %r, 1 <i>; yields i8</i>
4792 %Z = call void @foo() noreturn <i>; indicates that %foo never returns normally</i>
4793 %ZZ = call zeroext i32 @bar() <i>; Return value is %zero extended</i>
4798 <!-- _______________________________________________________________________ -->
4799 <div class="doc_subsubsection">
4800 <a name="i_va_arg">'<tt>va_arg</tt>' Instruction</a>
4803 <div class="doc_text">
4808 <resultval> = va_arg <va_list*> <arglist>, <argty>
4813 <p>The '<tt>va_arg</tt>' instruction is used to access arguments passed through
4814 the "variable argument" area of a function call. It is used to implement the
4815 <tt>va_arg</tt> macro in C.</p>
4819 <p>This instruction takes a <tt>va_list*</tt> value and the type of
4820 the argument. It returns a value of the specified argument type and
4821 increments the <tt>va_list</tt> to point to the next argument. The
4822 actual type of <tt>va_list</tt> is target specific.</p>
4826 <p>The '<tt>va_arg</tt>' instruction loads an argument of the specified
4827 type from the specified <tt>va_list</tt> and causes the
4828 <tt>va_list</tt> to point to the next argument. For more information,
4829 see the variable argument handling <a href="#int_varargs">Intrinsic
4832 <p>It is legal for this instruction to be called in a function which does not
4833 take a variable number of arguments, for example, the <tt>vfprintf</tt>
4836 <p><tt>va_arg</tt> is an LLVM instruction instead of an <a
4837 href="#intrinsics">intrinsic function</a> because it takes a type as an
4842 <p>See the <a href="#int_varargs">variable argument processing</a> section.</p>
4844 <p>Note that the code generator does not yet fully support va_arg
4845 on many targets. Also, it does not currently support va_arg with
4846 aggregate types on any target.</p>
4850 <!-- *********************************************************************** -->
4851 <div class="doc_section"> <a name="intrinsics">Intrinsic Functions</a> </div>
4852 <!-- *********************************************************************** -->
4854 <div class="doc_text">
4856 <p>LLVM supports the notion of an "intrinsic function". These functions have
4857 well known names and semantics and are required to follow certain restrictions.
4858 Overall, these intrinsics represent an extension mechanism for the LLVM
4859 language that does not require changing all of the transformations in LLVM when
4860 adding to the language (or the bitcode reader/writer, the parser, etc...).</p>
4862 <p>Intrinsic function names must all start with an "<tt>llvm.</tt>" prefix. This
4863 prefix is reserved in LLVM for intrinsic names; thus, function names may not
4864 begin with this prefix. Intrinsic functions must always be external functions:
4865 you cannot define the body of intrinsic functions. Intrinsic functions may
4866 only be used in call or invoke instructions: it is illegal to take the address
4867 of an intrinsic function. Additionally, because intrinsic functions are part
4868 of the LLVM language, it is required if any are added that they be documented
4871 <p>Some intrinsic functions can be overloaded, i.e., the intrinsic represents
4872 a family of functions that perform the same operation but on different data
4873 types. Because LLVM can represent over 8 million different integer types,
4874 overloading is used commonly to allow an intrinsic function to operate on any
4875 integer type. One or more of the argument types or the result type can be
4876 overloaded to accept any integer type. Argument types may also be defined as
4877 exactly matching a previous argument's type or the result type. This allows an
4878 intrinsic function which accepts multiple arguments, but needs all of them to
4879 be of the same type, to only be overloaded with respect to a single argument or
4882 <p>Overloaded intrinsics will have the names of its overloaded argument types
4883 encoded into its function name, each preceded by a period. Only those types
4884 which are overloaded result in a name suffix. Arguments whose type is matched
4885 against another type do not. For example, the <tt>llvm.ctpop</tt> function can
4886 take an integer of any width and returns an integer of exactly the same integer
4887 width. This leads to a family of functions such as
4888 <tt>i8 @llvm.ctpop.i8(i8 %val)</tt> and <tt>i29 @llvm.ctpop.i29(i29 %val)</tt>.
4889 Only one type, the return type, is overloaded, and only one type suffix is
4890 required. Because the argument's type is matched against the return type, it
4891 does not require its own name suffix.</p>
4893 <p>To learn how to add an intrinsic function, please see the
4894 <a href="ExtendingLLVM.html">Extending LLVM Guide</a>.
4899 <!-- ======================================================================= -->
4900 <div class="doc_subsection">
4901 <a name="int_varargs">Variable Argument Handling Intrinsics</a>
4904 <div class="doc_text">
4906 <p>Variable argument support is defined in LLVM with the <a
4907 href="#i_va_arg"><tt>va_arg</tt></a> instruction and these three
4908 intrinsic functions. These functions are related to the similarly
4909 named macros defined in the <tt><stdarg.h></tt> header file.</p>
4911 <p>All of these functions operate on arguments that use a
4912 target-specific value type "<tt>va_list</tt>". The LLVM assembly
4913 language reference manual does not define what this type is, so all
4914 transformations should be prepared to handle these functions regardless of
4917 <p>This example shows how the <a href="#i_va_arg"><tt>va_arg</tt></a>
4918 instruction and the variable argument handling intrinsic functions are
4921 <div class="doc_code">
4923 define i32 @test(i32 %X, ...) {
4924 ; Initialize variable argument processing
4926 %ap2 = bitcast i8** %ap to i8*
4927 call void @llvm.va_start(i8* %ap2)
4929 ; Read a single integer argument
4930 %tmp = va_arg i8** %ap, i32
4932 ; Demonstrate usage of llvm.va_copy and llvm.va_end
4934 %aq2 = bitcast i8** %aq to i8*
4935 call void @llvm.va_copy(i8* %aq2, i8* %ap2)
4936 call void @llvm.va_end(i8* %aq2)
4938 ; Stop processing of arguments.
4939 call void @llvm.va_end(i8* %ap2)
4943 declare void @llvm.va_start(i8*)
4944 declare void @llvm.va_copy(i8*, i8*)
4945 declare void @llvm.va_end(i8*)
4951 <!-- _______________________________________________________________________ -->
4952 <div class="doc_subsubsection">
4953 <a name="int_va_start">'<tt>llvm.va_start</tt>' Intrinsic</a>
4957 <div class="doc_text">
4959 <pre> declare void %llvm.va_start(i8* <arglist>)<br></pre>
4961 <p>The '<tt>llvm.va_start</tt>' intrinsic initializes
4962 <tt>*<arglist></tt> for subsequent use by <tt><a
4963 href="#i_va_arg">va_arg</a></tt>.</p>
4967 <p>The argument is a pointer to a <tt>va_list</tt> element to initialize.</p>
4971 <p>The '<tt>llvm.va_start</tt>' intrinsic works just like the <tt>va_start</tt>
4972 macro available in C. In a target-dependent way, it initializes the
4973 <tt>va_list</tt> element to which the argument points, so that the next call to
4974 <tt>va_arg</tt> will produce the first variable argument passed to the function.
4975 Unlike the C <tt>va_start</tt> macro, this intrinsic does not need to know the
4976 last argument of the function as the compiler can figure that out.</p>
4980 <!-- _______________________________________________________________________ -->
4981 <div class="doc_subsubsection">
4982 <a name="int_va_end">'<tt>llvm.va_end</tt>' Intrinsic</a>
4985 <div class="doc_text">
4987 <pre> declare void @llvm.va_end(i8* <arglist>)<br></pre>
4990 <p>The '<tt>llvm.va_end</tt>' intrinsic destroys <tt>*<arglist></tt>,
4991 which has been initialized previously with <tt><a href="#int_va_start">llvm.va_start</a></tt>
4992 or <tt><a href="#i_va_copy">llvm.va_copy</a></tt>.</p>
4996 <p>The argument is a pointer to a <tt>va_list</tt> to destroy.</p>
5000 <p>The '<tt>llvm.va_end</tt>' intrinsic works just like the <tt>va_end</tt>
5001 macro available in C. In a target-dependent way, it destroys the
5002 <tt>va_list</tt> element to which the argument points. Calls to <a
5003 href="#int_va_start"><tt>llvm.va_start</tt></a> and <a href="#int_va_copy">
5004 <tt>llvm.va_copy</tt></a> must be matched exactly with calls to
5005 <tt>llvm.va_end</tt>.</p>
5009 <!-- _______________________________________________________________________ -->
5010 <div class="doc_subsubsection">
5011 <a name="int_va_copy">'<tt>llvm.va_copy</tt>' Intrinsic</a>
5014 <div class="doc_text">
5019 declare void @llvm.va_copy(i8* <destarglist>, i8* <srcarglist>)
5024 <p>The '<tt>llvm.va_copy</tt>' intrinsic copies the current argument position
5025 from the source argument list to the destination argument list.</p>
5029 <p>The first argument is a pointer to a <tt>va_list</tt> element to initialize.
5030 The second argument is a pointer to a <tt>va_list</tt> element to copy from.</p>
5035 <p>The '<tt>llvm.va_copy</tt>' intrinsic works just like the <tt>va_copy</tt>
5036 macro available in C. In a target-dependent way, it copies the source
5037 <tt>va_list</tt> element into the destination <tt>va_list</tt> element. This
5038 intrinsic is necessary because the <tt><a href="#int_va_start">
5039 llvm.va_start</a></tt> intrinsic may be arbitrarily complex and require, for
5040 example, memory allocation.</p>
5044 <!-- ======================================================================= -->
5045 <div class="doc_subsection">
5046 <a name="int_gc">Accurate Garbage Collection Intrinsics</a>
5049 <div class="doc_text">
5052 LLVM support for <a href="GarbageCollection.html">Accurate Garbage
5053 Collection</a> (GC) requires the implementation and generation of these
5055 These intrinsics allow identification of <a href="#int_gcroot">GC roots on the
5056 stack</a>, as well as garbage collector implementations that require <a
5057 href="#int_gcread">read</a> and <a href="#int_gcwrite">write</a> barriers.
5058 Front-ends for type-safe garbage collected languages should generate these
5059 intrinsics to make use of the LLVM garbage collectors. For more details, see <a
5060 href="GarbageCollection.html">Accurate Garbage Collection with LLVM</a>.
5063 <p>The garbage collection intrinsics only operate on objects in the generic
5064 address space (address space zero).</p>
5068 <!-- _______________________________________________________________________ -->
5069 <div class="doc_subsubsection">
5070 <a name="int_gcroot">'<tt>llvm.gcroot</tt>' Intrinsic</a>
5073 <div class="doc_text">
5078 declare void @llvm.gcroot(i8** %ptrloc, i8* %metadata)
5083 <p>The '<tt>llvm.gcroot</tt>' intrinsic declares the existence of a GC root to
5084 the code generator, and allows some metadata to be associated with it.</p>
5088 <p>The first argument specifies the address of a stack object that contains the
5089 root pointer. The second pointer (which must be either a constant or a global
5090 value address) contains the meta-data to be associated with the root.</p>
5094 <p>At runtime, a call to this intrinsic stores a null pointer into the "ptrloc"
5095 location. At compile-time, the code generator generates information to allow
5096 the runtime to find the pointer at GC safe points. The '<tt>llvm.gcroot</tt>'
5097 intrinsic may only be used in a function which <a href="#gc">specifies a GC
5103 <!-- _______________________________________________________________________ -->
5104 <div class="doc_subsubsection">
5105 <a name="int_gcread">'<tt>llvm.gcread</tt>' Intrinsic</a>
5108 <div class="doc_text">
5113 declare i8* @llvm.gcread(i8* %ObjPtr, i8** %Ptr)
5118 <p>The '<tt>llvm.gcread</tt>' intrinsic identifies reads of references from heap
5119 locations, allowing garbage collector implementations that require read
5124 <p>The second argument is the address to read from, which should be an address
5125 allocated from the garbage collector. The first object is a pointer to the
5126 start of the referenced object, if needed by the language runtime (otherwise
5131 <p>The '<tt>llvm.gcread</tt>' intrinsic has the same semantics as a load
5132 instruction, but may be replaced with substantially more complex code by the
5133 garbage collector runtime, as needed. The '<tt>llvm.gcread</tt>' intrinsic
5134 may only be used in a function which <a href="#gc">specifies a GC
5140 <!-- _______________________________________________________________________ -->
5141 <div class="doc_subsubsection">
5142 <a name="int_gcwrite">'<tt>llvm.gcwrite</tt>' Intrinsic</a>
5145 <div class="doc_text">
5150 declare void @llvm.gcwrite(i8* %P1, i8* %Obj, i8** %P2)
5155 <p>The '<tt>llvm.gcwrite</tt>' intrinsic identifies writes of references to heap
5156 locations, allowing garbage collector implementations that require write
5157 barriers (such as generational or reference counting collectors).</p>
5161 <p>The first argument is the reference to store, the second is the start of the
5162 object to store it to, and the third is the address of the field of Obj to
5163 store to. If the runtime does not require a pointer to the object, Obj may be
5168 <p>The '<tt>llvm.gcwrite</tt>' intrinsic has the same semantics as a store
5169 instruction, but may be replaced with substantially more complex code by the
5170 garbage collector runtime, as needed. The '<tt>llvm.gcwrite</tt>' intrinsic
5171 may only be used in a function which <a href="#gc">specifies a GC
5178 <!-- ======================================================================= -->
5179 <div class="doc_subsection">
5180 <a name="int_codegen">Code Generator Intrinsics</a>
5183 <div class="doc_text">
5185 These intrinsics are provided by LLVM to expose special features that may only
5186 be implemented with code generator support.
5191 <!-- _______________________________________________________________________ -->
5192 <div class="doc_subsubsection">
5193 <a name="int_returnaddress">'<tt>llvm.returnaddress</tt>' Intrinsic</a>
5196 <div class="doc_text">
5200 declare i8 *@llvm.returnaddress(i32 <level>)
5206 The '<tt>llvm.returnaddress</tt>' intrinsic attempts to compute a
5207 target-specific value indicating the return address of the current function
5208 or one of its callers.
5214 The argument to this intrinsic indicates which function to return the address
5215 for. Zero indicates the calling function, one indicates its caller, etc. The
5216 argument is <b>required</b> to be a constant integer value.
5222 The '<tt>llvm.returnaddress</tt>' intrinsic either returns a pointer indicating
5223 the return address of the specified call frame, or zero if it cannot be
5224 identified. The value returned by this intrinsic is likely to be incorrect or 0
5225 for arguments other than zero, so it should only be used for debugging purposes.
5229 Note that calling this intrinsic does not prevent function inlining or other
5230 aggressive transformations, so the value returned may not be that of the obvious
5231 source-language caller.
5236 <!-- _______________________________________________________________________ -->
5237 <div class="doc_subsubsection">
5238 <a name="int_frameaddress">'<tt>llvm.frameaddress</tt>' Intrinsic</a>
5241 <div class="doc_text">
5245 declare i8 *@llvm.frameaddress(i32 <level>)
5251 The '<tt>llvm.frameaddress</tt>' intrinsic attempts to return the
5252 target-specific frame pointer value for the specified stack frame.
5258 The argument to this intrinsic indicates which function to return the frame
5259 pointer for. Zero indicates the calling function, one indicates its caller,
5260 etc. The argument is <b>required</b> to be a constant integer value.
5266 The '<tt>llvm.frameaddress</tt>' intrinsic either returns a pointer indicating
5267 the frame address of the specified call frame, or zero if it cannot be
5268 identified. The value returned by this intrinsic is likely to be incorrect or 0
5269 for arguments other than zero, so it should only be used for debugging purposes.
5273 Note that calling this intrinsic does not prevent function inlining or other
5274 aggressive transformations, so the value returned may not be that of the obvious
5275 source-language caller.
5279 <!-- _______________________________________________________________________ -->
5280 <div class="doc_subsubsection">
5281 <a name="int_stacksave">'<tt>llvm.stacksave</tt>' Intrinsic</a>
5284 <div class="doc_text">
5288 declare i8 *@llvm.stacksave()
5294 The '<tt>llvm.stacksave</tt>' intrinsic is used to remember the current state of
5295 the function stack, for use with <a href="#int_stackrestore">
5296 <tt>llvm.stackrestore</tt></a>. This is useful for implementing language
5297 features like scoped automatic variable sized arrays in C99.
5303 This intrinsic returns a opaque pointer value that can be passed to <a
5304 href="#int_stackrestore"><tt>llvm.stackrestore</tt></a>. When an
5305 <tt>llvm.stackrestore</tt> intrinsic is executed with a value saved from
5306 <tt>llvm.stacksave</tt>, it effectively restores the state of the stack to the
5307 state it was in when the <tt>llvm.stacksave</tt> intrinsic executed. In
5308 practice, this pops any <a href="#i_alloca">alloca</a> blocks from the stack
5309 that were allocated after the <tt>llvm.stacksave</tt> was executed.
5314 <!-- _______________________________________________________________________ -->
5315 <div class="doc_subsubsection">
5316 <a name="int_stackrestore">'<tt>llvm.stackrestore</tt>' Intrinsic</a>
5319 <div class="doc_text">
5323 declare void @llvm.stackrestore(i8 * %ptr)
5329 The '<tt>llvm.stackrestore</tt>' intrinsic is used to restore the state of
5330 the function stack to the state it was in when the corresponding <a
5331 href="#int_stacksave"><tt>llvm.stacksave</tt></a> intrinsic executed. This is
5332 useful for implementing language features like scoped automatic variable sized
5339 See the description for <a href="#int_stacksave"><tt>llvm.stacksave</tt></a>.
5345 <!-- _______________________________________________________________________ -->
5346 <div class="doc_subsubsection">
5347 <a name="int_prefetch">'<tt>llvm.prefetch</tt>' Intrinsic</a>
5350 <div class="doc_text">
5354 declare void @llvm.prefetch(i8* <address>, i32 <rw>, i32 <locality>)
5361 The '<tt>llvm.prefetch</tt>' intrinsic is a hint to the code generator to insert
5362 a prefetch instruction if supported; otherwise, it is a noop. Prefetches have
5364 effect on the behavior of the program but can change its performance
5371 <tt>address</tt> is the address to be prefetched, <tt>rw</tt> is the specifier
5372 determining if the fetch should be for a read (0) or write (1), and
5373 <tt>locality</tt> is a temporal locality specifier ranging from (0) - no
5374 locality, to (3) - extremely local keep in cache. The <tt>rw</tt> and
5375 <tt>locality</tt> arguments must be constant integers.
5381 This intrinsic does not modify the behavior of the program. In particular,
5382 prefetches cannot trap and do not produce a value. On targets that support this
5383 intrinsic, the prefetch can provide hints to the processor cache for better
5389 <!-- _______________________________________________________________________ -->
5390 <div class="doc_subsubsection">
5391 <a name="int_pcmarker">'<tt>llvm.pcmarker</tt>' Intrinsic</a>
5394 <div class="doc_text">
5398 declare void @llvm.pcmarker(i32 <id>)
5405 The '<tt>llvm.pcmarker</tt>' intrinsic is a method to export a Program Counter
5407 code to simulators and other tools. The method is target specific, but it is
5408 expected that the marker will use exported symbols to transmit the PC of the
5410 The marker makes no guarantees that it will remain with any specific instruction
5411 after optimizations. It is possible that the presence of a marker will inhibit
5412 optimizations. The intended use is to be inserted after optimizations to allow
5413 correlations of simulation runs.
5419 <tt>id</tt> is a numerical id identifying the marker.
5425 This intrinsic does not modify the behavior of the program. Backends that do not
5426 support this intrinisic may ignore it.
5431 <!-- _______________________________________________________________________ -->
5432 <div class="doc_subsubsection">
5433 <a name="int_readcyclecounter">'<tt>llvm.readcyclecounter</tt>' Intrinsic</a>
5436 <div class="doc_text">
5440 declare i64 @llvm.readcyclecounter( )
5447 The '<tt>llvm.readcyclecounter</tt>' intrinsic provides access to the cycle
5448 counter register (or similar low latency, high accuracy clocks) on those targets
5449 that support it. On X86, it should map to RDTSC. On Alpha, it should map to RPCC.
5450 As the backing counters overflow quickly (on the order of 9 seconds on alpha), this
5451 should only be used for small timings.
5457 When directly supported, reading the cycle counter should not modify any memory.
5458 Implementations are allowed to either return a application specific value or a
5459 system wide value. On backends without support, this is lowered to a constant 0.
5464 <!-- ======================================================================= -->
5465 <div class="doc_subsection">
5466 <a name="int_libc">Standard C Library Intrinsics</a>
5469 <div class="doc_text">
5471 LLVM provides intrinsics for a few important standard C library functions.
5472 These intrinsics allow source-language front-ends to pass information about the
5473 alignment of the pointer arguments to the code generator, providing opportunity
5474 for more efficient code generation.
5479 <!-- _______________________________________________________________________ -->
5480 <div class="doc_subsubsection">
5481 <a name="int_memcpy">'<tt>llvm.memcpy</tt>' Intrinsic</a>
5484 <div class="doc_text">
5487 <p>This is an overloaded intrinsic. You can use llvm.memcpy on any integer bit
5488 width. Not all targets support all bit widths however.</p>
5490 declare void @llvm.memcpy.i8(i8 * <dest>, i8 * <src>,
5491 i8 <len>, i32 <align>)
5492 declare void @llvm.memcpy.i16(i8 * <dest>, i8 * <src>,
5493 i16 <len>, i32 <align>)
5494 declare void @llvm.memcpy.i32(i8 * <dest>, i8 * <src>,
5495 i32 <len>, i32 <align>)
5496 declare void @llvm.memcpy.i64(i8 * <dest>, i8 * <src>,
5497 i64 <len>, i32 <align>)
5503 The '<tt>llvm.memcpy.*</tt>' intrinsics copy a block of memory from the source
5504 location to the destination location.
5508 Note that, unlike the standard libc function, the <tt>llvm.memcpy.*</tt>
5509 intrinsics do not return a value, and takes an extra alignment argument.
5515 The first argument is a pointer to the destination, the second is a pointer to
5516 the source. The third argument is an integer argument
5517 specifying the number of bytes to copy, and the fourth argument is the alignment
5518 of the source and destination locations.
5522 If the call to this intrinisic has an alignment value that is not 0 or 1, then
5523 the caller guarantees that both the source and destination pointers are aligned
5530 The '<tt>llvm.memcpy.*</tt>' intrinsics copy a block of memory from the source
5531 location to the destination location, which are not allowed to overlap. It
5532 copies "len" bytes of memory over. If the argument is known to be aligned to
5533 some boundary, this can be specified as the fourth argument, otherwise it should
5539 <!-- _______________________________________________________________________ -->
5540 <div class="doc_subsubsection">
5541 <a name="int_memmove">'<tt>llvm.memmove</tt>' Intrinsic</a>
5544 <div class="doc_text">
5547 <p>This is an overloaded intrinsic. You can use llvm.memmove on any integer bit
5548 width. Not all targets support all bit widths however.</p>
5550 declare void @llvm.memmove.i8(i8 * <dest>, i8 * <src>,
5551 i8 <len>, i32 <align>)
5552 declare void @llvm.memmove.i16(i8 * <dest>, i8 * <src>,
5553 i16 <len>, i32 <align>)
5554 declare void @llvm.memmove.i32(i8 * <dest>, i8 * <src>,
5555 i32 <len>, i32 <align>)
5556 declare void @llvm.memmove.i64(i8 * <dest>, i8 * <src>,
5557 i64 <len>, i32 <align>)
5563 The '<tt>llvm.memmove.*</tt>' intrinsics move a block of memory from the source
5564 location to the destination location. It is similar to the
5565 '<tt>llvm.memcpy</tt>' intrinsic but allows the two memory locations to overlap.
5569 Note that, unlike the standard libc function, the <tt>llvm.memmove.*</tt>
5570 intrinsics do not return a value, and takes an extra alignment argument.
5576 The first argument is a pointer to the destination, the second is a pointer to
5577 the source. The third argument is an integer argument
5578 specifying the number of bytes to copy, and the fourth argument is the alignment
5579 of the source and destination locations.
5583 If the call to this intrinisic has an alignment value that is not 0 or 1, then
5584 the caller guarantees that the source and destination pointers are aligned to
5591 The '<tt>llvm.memmove.*</tt>' intrinsics copy a block of memory from the source
5592 location to the destination location, which may overlap. It
5593 copies "len" bytes of memory over. If the argument is known to be aligned to
5594 some boundary, this can be specified as the fourth argument, otherwise it should
5600 <!-- _______________________________________________________________________ -->
5601 <div class="doc_subsubsection">
5602 <a name="int_memset">'<tt>llvm.memset.*</tt>' Intrinsics</a>
5605 <div class="doc_text">
5608 <p>This is an overloaded intrinsic. You can use llvm.memset on any integer bit
5609 width. Not all targets support all bit widths however.</p>
5611 declare void @llvm.memset.i8(i8 * <dest>, i8 <val>,
5612 i8 <len>, i32 <align>)
5613 declare void @llvm.memset.i16(i8 * <dest>, i8 <val>,
5614 i16 <len>, i32 <align>)
5615 declare void @llvm.memset.i32(i8 * <dest>, i8 <val>,
5616 i32 <len>, i32 <align>)
5617 declare void @llvm.memset.i64(i8 * <dest>, i8 <val>,
5618 i64 <len>, i32 <align>)
5624 The '<tt>llvm.memset.*</tt>' intrinsics fill a block of memory with a particular
5629 Note that, unlike the standard libc function, the <tt>llvm.memset</tt> intrinsic
5630 does not return a value, and takes an extra alignment argument.
5636 The first argument is a pointer to the destination to fill, the second is the
5637 byte value to fill it with, the third argument is an integer
5638 argument specifying the number of bytes to fill, and the fourth argument is the
5639 known alignment of destination location.
5643 If the call to this intrinisic has an alignment value that is not 0 or 1, then
5644 the caller guarantees that the destination pointer is aligned to that boundary.
5650 The '<tt>llvm.memset.*</tt>' intrinsics fill "len" bytes of memory starting at
5652 destination location. If the argument is known to be aligned to some boundary,
5653 this can be specified as the fourth argument, otherwise it should be set to 0 or
5659 <!-- _______________________________________________________________________ -->
5660 <div class="doc_subsubsection">
5661 <a name="int_sqrt">'<tt>llvm.sqrt.*</tt>' Intrinsic</a>
5664 <div class="doc_text">
5667 <p>This is an overloaded intrinsic. You can use <tt>llvm.sqrt</tt> on any
5668 floating point or vector of floating point type. Not all targets support all
5671 declare float @llvm.sqrt.f32(float %Val)
5672 declare double @llvm.sqrt.f64(double %Val)
5673 declare x86_fp80 @llvm.sqrt.f80(x86_fp80 %Val)
5674 declare fp128 @llvm.sqrt.f128(fp128 %Val)
5675 declare ppc_fp128 @llvm.sqrt.ppcf128(ppc_fp128 %Val)
5681 The '<tt>llvm.sqrt</tt>' intrinsics return the sqrt of the specified operand,
5682 returning the same value as the libm '<tt>sqrt</tt>' functions would. Unlike
5683 <tt>sqrt</tt> in libm, however, <tt>llvm.sqrt</tt> has undefined behavior for
5684 negative numbers other than -0.0 (which allows for better optimization, because
5685 there is no need to worry about errno being set). <tt>llvm.sqrt(-0.0)</tt> is
5686 defined to return -0.0 like IEEE sqrt.
5692 The argument and return value are floating point numbers of the same type.
5698 This function returns the sqrt of the specified operand if it is a nonnegative
5699 floating point number.
5703 <!-- _______________________________________________________________________ -->
5704 <div class="doc_subsubsection">
5705 <a name="int_powi">'<tt>llvm.powi.*</tt>' Intrinsic</a>
5708 <div class="doc_text">
5711 <p>This is an overloaded intrinsic. You can use <tt>llvm.powi</tt> on any
5712 floating point or vector of floating point type. Not all targets support all
5715 declare float @llvm.powi.f32(float %Val, i32 %power)
5716 declare double @llvm.powi.f64(double %Val, i32 %power)
5717 declare x86_fp80 @llvm.powi.f80(x86_fp80 %Val, i32 %power)
5718 declare fp128 @llvm.powi.f128(fp128 %Val, i32 %power)
5719 declare ppc_fp128 @llvm.powi.ppcf128(ppc_fp128 %Val, i32 %power)
5725 The '<tt>llvm.powi.*</tt>' intrinsics return the first operand raised to the
5726 specified (positive or negative) power. The order of evaluation of
5727 multiplications is not defined. When a vector of floating point type is
5728 used, the second argument remains a scalar integer value.
5734 The second argument is an integer power, and the first is a value to raise to
5741 This function returns the first value raised to the second power with an
5742 unspecified sequence of rounding operations.</p>
5745 <!-- _______________________________________________________________________ -->
5746 <div class="doc_subsubsection">
5747 <a name="int_sin">'<tt>llvm.sin.*</tt>' Intrinsic</a>
5750 <div class="doc_text">
5753 <p>This is an overloaded intrinsic. You can use <tt>llvm.sin</tt> on any
5754 floating point or vector of floating point type. Not all targets support all
5757 declare float @llvm.sin.f32(float %Val)
5758 declare double @llvm.sin.f64(double %Val)
5759 declare x86_fp80 @llvm.sin.f80(x86_fp80 %Val)
5760 declare fp128 @llvm.sin.f128(fp128 %Val)
5761 declare ppc_fp128 @llvm.sin.ppcf128(ppc_fp128 %Val)
5767 The '<tt>llvm.sin.*</tt>' intrinsics return the sine of the operand.
5773 The argument and return value are floating point numbers of the same type.
5779 This function returns the sine of the specified operand, returning the
5780 same values as the libm <tt>sin</tt> functions would, and handles error
5781 conditions in the same way.</p>
5784 <!-- _______________________________________________________________________ -->
5785 <div class="doc_subsubsection">
5786 <a name="int_cos">'<tt>llvm.cos.*</tt>' Intrinsic</a>
5789 <div class="doc_text">
5792 <p>This is an overloaded intrinsic. You can use <tt>llvm.cos</tt> on any
5793 floating point or vector of floating point type. Not all targets support all
5796 declare float @llvm.cos.f32(float %Val)
5797 declare double @llvm.cos.f64(double %Val)
5798 declare x86_fp80 @llvm.cos.f80(x86_fp80 %Val)
5799 declare fp128 @llvm.cos.f128(fp128 %Val)
5800 declare ppc_fp128 @llvm.cos.ppcf128(ppc_fp128 %Val)
5806 The '<tt>llvm.cos.*</tt>' intrinsics return the cosine of the operand.
5812 The argument and return value are floating point numbers of the same type.
5818 This function returns the cosine of the specified operand, returning the
5819 same values as the libm <tt>cos</tt> functions would, and handles error
5820 conditions in the same way.</p>
5823 <!-- _______________________________________________________________________ -->
5824 <div class="doc_subsubsection">
5825 <a name="int_pow">'<tt>llvm.pow.*</tt>' Intrinsic</a>
5828 <div class="doc_text">
5831 <p>This is an overloaded intrinsic. You can use <tt>llvm.pow</tt> on any
5832 floating point or vector of floating point type. Not all targets support all
5835 declare float @llvm.pow.f32(float %Val, float %Power)
5836 declare double @llvm.pow.f64(double %Val, double %Power)
5837 declare x86_fp80 @llvm.pow.f80(x86_fp80 %Val, x86_fp80 %Power)
5838 declare fp128 @llvm.pow.f128(fp128 %Val, fp128 %Power)
5839 declare ppc_fp128 @llvm.pow.ppcf128(ppc_fp128 %Val, ppc_fp128 Power)
5845 The '<tt>llvm.pow.*</tt>' intrinsics return the first operand raised to the
5846 specified (positive or negative) power.
5852 The second argument is a floating point power, and the first is a value to
5853 raise to that power.
5859 This function returns the first value raised to the second power,
5861 same values as the libm <tt>pow</tt> functions would, and handles error
5862 conditions in the same way.</p>
5866 <!-- ======================================================================= -->
5867 <div class="doc_subsection">
5868 <a name="int_manip">Bit Manipulation Intrinsics</a>
5871 <div class="doc_text">
5873 LLVM provides intrinsics for a few important bit manipulation operations.
5874 These allow efficient code generation for some algorithms.
5879 <!-- _______________________________________________________________________ -->
5880 <div class="doc_subsubsection">
5881 <a name="int_bswap">'<tt>llvm.bswap.*</tt>' Intrinsics</a>
5884 <div class="doc_text">
5887 <p>This is an overloaded intrinsic function. You can use bswap on any integer
5888 type that is an even number of bytes (i.e. BitWidth % 16 == 0).</p>
5890 declare i16 @llvm.bswap.i16(i16 <id>)
5891 declare i32 @llvm.bswap.i32(i32 <id>)
5892 declare i64 @llvm.bswap.i64(i64 <id>)
5898 The '<tt>llvm.bswap</tt>' family of intrinsics is used to byte swap integer
5899 values with an even number of bytes (positive multiple of 16 bits). These are
5900 useful for performing operations on data that is not in the target's native
5907 The <tt>llvm.bswap.i16</tt> intrinsic returns an i16 value that has the high
5908 and low byte of the input i16 swapped. Similarly, the <tt>llvm.bswap.i32</tt>
5909 intrinsic returns an i32 value that has the four bytes of the input i32
5910 swapped, so that if the input bytes are numbered 0, 1, 2, 3 then the returned
5911 i32 will have its bytes in 3, 2, 1, 0 order. The <tt>llvm.bswap.i48</tt>,
5912 <tt>llvm.bswap.i64</tt> and other intrinsics extend this concept to
5913 additional even-byte lengths (6 bytes, 8 bytes and more, respectively).
5918 <!-- _______________________________________________________________________ -->
5919 <div class="doc_subsubsection">
5920 <a name="int_ctpop">'<tt>llvm.ctpop.*</tt>' Intrinsic</a>
5923 <div class="doc_text">
5926 <p>This is an overloaded intrinsic. You can use llvm.ctpop on any integer bit
5927 width. Not all targets support all bit widths however.</p>
5929 declare i8 @llvm.ctpop.i8(i8 <src>)
5930 declare i16 @llvm.ctpop.i16(i16 <src>)
5931 declare i32 @llvm.ctpop.i32(i32 <src>)
5932 declare i64 @llvm.ctpop.i64(i64 <src>)
5933 declare i256 @llvm.ctpop.i256(i256 <src>)
5939 The '<tt>llvm.ctpop</tt>' family of intrinsics counts the number of bits set in a
5946 The only argument is the value to be counted. The argument may be of any
5947 integer type. The return type must match the argument type.
5953 The '<tt>llvm.ctpop</tt>' intrinsic counts the 1's in a variable.
5957 <!-- _______________________________________________________________________ -->
5958 <div class="doc_subsubsection">
5959 <a name="int_ctlz">'<tt>llvm.ctlz.*</tt>' Intrinsic</a>
5962 <div class="doc_text">
5965 <p>This is an overloaded intrinsic. You can use <tt>llvm.ctlz</tt> on any
5966 integer bit width. Not all targets support all bit widths however.</p>
5968 declare i8 @llvm.ctlz.i8 (i8 <src>)
5969 declare i16 @llvm.ctlz.i16(i16 <src>)
5970 declare i32 @llvm.ctlz.i32(i32 <src>)
5971 declare i64 @llvm.ctlz.i64(i64 <src>)
5972 declare i256 @llvm.ctlz.i256(i256 <src>)
5978 The '<tt>llvm.ctlz</tt>' family of intrinsic functions counts the number of
5979 leading zeros in a variable.
5985 The only argument is the value to be counted. The argument may be of any
5986 integer type. The return type must match the argument type.
5992 The '<tt>llvm.ctlz</tt>' intrinsic counts the leading (most significant) zeros
5993 in a variable. If the src == 0 then the result is the size in bits of the type
5994 of src. For example, <tt>llvm.ctlz(i32 2) = 30</tt>.
6000 <!-- _______________________________________________________________________ -->
6001 <div class="doc_subsubsection">
6002 <a name="int_cttz">'<tt>llvm.cttz.*</tt>' Intrinsic</a>
6005 <div class="doc_text">
6008 <p>This is an overloaded intrinsic. You can use <tt>llvm.cttz</tt> on any
6009 integer bit width. Not all targets support all bit widths however.</p>
6011 declare i8 @llvm.cttz.i8 (i8 <src>)
6012 declare i16 @llvm.cttz.i16(i16 <src>)
6013 declare i32 @llvm.cttz.i32(i32 <src>)
6014 declare i64 @llvm.cttz.i64(i64 <src>)
6015 declare i256 @llvm.cttz.i256(i256 <src>)
6021 The '<tt>llvm.cttz</tt>' family of intrinsic functions counts the number of
6028 The only argument is the value to be counted. The argument may be of any
6029 integer type. The return type must match the argument type.
6035 The '<tt>llvm.cttz</tt>' intrinsic counts the trailing (least significant) zeros
6036 in a variable. If the src == 0 then the result is the size in bits of the type
6037 of src. For example, <tt>llvm.cttz(2) = 1</tt>.
6041 <!-- _______________________________________________________________________ -->
6042 <div class="doc_subsubsection">
6043 <a name="int_part_select">'<tt>llvm.part.select.*</tt>' Intrinsic</a>
6046 <div class="doc_text">
6049 <p>This is an overloaded intrinsic. You can use <tt>llvm.part.select</tt>
6050 on any integer bit width.</p>
6052 declare i17 @llvm.part.select.i17 (i17 %val, i32 %loBit, i32 %hiBit)
6053 declare i29 @llvm.part.select.i29 (i29 %val, i32 %loBit, i32 %hiBit)
6057 <p>The '<tt>llvm.part.select</tt>' family of intrinsic functions selects a
6058 range of bits from an integer value and returns them in the same bit width as
6059 the original value.</p>
6062 <p>The first argument, <tt>%val</tt> and the result may be integer types of
6063 any bit width but they must have the same bit width. The second and third
6064 arguments must be <tt>i32</tt> type since they specify only a bit index.</p>
6067 <p>The operation of the '<tt>llvm.part.select</tt>' intrinsic has two modes
6068 of operation: forwards and reverse. If <tt>%loBit</tt> is greater than
6069 <tt>%hiBits</tt> then the intrinsic operates in reverse mode. Otherwise it
6070 operates in forward mode.</p>
6071 <p>In forward mode, this intrinsic is the equivalent of shifting <tt>%val</tt>
6072 right by <tt>%loBit</tt> bits and then ANDing it with a mask with
6073 only the <tt>%hiBit - %loBit</tt> bits set, as follows:</p>
6075 <li>The <tt>%val</tt> is shifted right (LSHR) by the number of bits specified
6076 by <tt>%loBits</tt>. This normalizes the value to the low order bits.</li>
6077 <li>The <tt>%loBits</tt> value is subtracted from the <tt>%hiBits</tt> value
6078 to determine the number of bits to retain.</li>
6079 <li>A mask of the retained bits is created by shifting a -1 value.</li>
6080 <li>The mask is ANDed with <tt>%val</tt> to produce the result.</li>
6082 <p>In reverse mode, a similar computation is made except that the bits are
6083 returned in the reverse order. So, for example, if <tt>X</tt> has the value
6084 <tt>i16 0x0ACF (101011001111)</tt> and we apply
6085 <tt>part.select(i16 X, 8, 3)</tt> to it, we get back the value
6086 <tt>i16 0x0026 (000000100110)</tt>.</p>
6089 <div class="doc_subsubsection">
6090 <a name="int_part_set">'<tt>llvm.part.set.*</tt>' Intrinsic</a>
6093 <div class="doc_text">
6096 <p>This is an overloaded intrinsic. You can use <tt>llvm.part.set</tt>
6097 on any integer bit width.</p>
6099 declare i17 @llvm.part.set.i17.i9 (i17 %val, i9 %repl, i32 %lo, i32 %hi)
6100 declare i29 @llvm.part.set.i29.i9 (i29 %val, i9 %repl, i32 %lo, i32 %hi)
6104 <p>The '<tt>llvm.part.set</tt>' family of intrinsic functions replaces a range
6105 of bits in an integer value with another integer value. It returns the integer
6106 with the replaced bits.</p>
6109 <p>The first argument, <tt>%val</tt>, and the result may be integer types of
6110 any bit width, but they must have the same bit width. <tt>%val</tt> is the value
6111 whose bits will be replaced. The second argument, <tt>%repl</tt> may be an
6112 integer of any bit width. The third and fourth arguments must be <tt>i32</tt>
6113 type since they specify only a bit index.</p>
6116 <p>The operation of the '<tt>llvm.part.set</tt>' intrinsic has two modes
6117 of operation: forwards and reverse. If <tt>%lo</tt> is greater than
6118 <tt>%hi</tt> then the intrinsic operates in reverse mode. Otherwise it
6119 operates in forward mode.</p>
6121 <p>For both modes, the <tt>%repl</tt> value is prepared for use by either
6122 truncating it down to the size of the replacement area or zero extending it
6123 up to that size.</p>
6125 <p>In forward mode, the bits between <tt>%lo</tt> and <tt>%hi</tt> (inclusive)
6126 are replaced with corresponding bits from <tt>%repl</tt>. That is the 0th bit
6127 in <tt>%repl</tt> replaces the <tt>%lo</tt>th bit in <tt>%val</tt> and etc. up
6128 to the <tt>%hi</tt>th bit.</p>
6130 <p>In reverse mode, a similar computation is made except that the bits are
6131 reversed. That is, the <tt>0</tt>th bit in <tt>%repl</tt> replaces the
6132 <tt>%hi</tt> bit in <tt>%val</tt> and etc. down to the <tt>%lo</tt>th bit.</p>
6137 llvm.part.set(0xFFFF, 0, 4, 7) -> 0xFF0F
6138 llvm.part.set(0xFFFF, 0, 7, 4) -> 0xFF0F
6139 llvm.part.set(0xFFFF, 1, 7, 4) -> 0xFF8F
6140 llvm.part.set(0xFFFF, F, 8, 3) -> 0xFFE7
6141 llvm.part.set(0xFFFF, 0, 3, 8) -> 0xFE07
6146 <!-- ======================================================================= -->
6147 <div class="doc_subsection">
6148 <a name="int_overflow">Arithmetic with Overflow Intrinsics</a>
6151 <div class="doc_text">
6153 LLVM provides intrinsics for some arithmetic with overflow operations.
6158 <!-- _______________________________________________________________________ -->
6159 <div class="doc_subsubsection">
6160 <a name="int_sadd_overflow">'<tt>llvm.sadd.with.overflow.*</tt>' Intrinsics</a>
6163 <div class="doc_text">
6167 <p>This is an overloaded intrinsic. You can use <tt>llvm.sadd.with.overflow</tt>
6168 on any integer bit width.</p>
6171 declare {i16, i1} @llvm.sadd.with.overflow.i16(i16 %a, i16 %b)
6172 declare {i32, i1} @llvm.sadd.with.overflow.i32(i32 %a, i32 %b)
6173 declare {i64, i1} @llvm.sadd.with.overflow.i64(i64 %a, i64 %b)
6178 <p>The '<tt>llvm.sadd.with.overflow</tt>' family of intrinsic functions perform
6179 a signed addition of the two arguments, and indicate whether an overflow
6180 occurred during the signed summation.</p>
6184 <p>The arguments (%a and %b) and the first element of the result structure may
6185 be of integer types of any bit width, but they must have the same bit width. The
6186 second element of the result structure must be of type <tt>i1</tt>. <tt>%a</tt>
6187 and <tt>%b</tt> are the two values that will undergo signed addition.</p>
6191 <p>The '<tt>llvm.sadd.with.overflow</tt>' family of intrinsic functions perform
6192 a signed addition of the two variables. They return a structure — the
6193 first element of which is the signed summation, and the second element of which
6194 is a bit specifying if the signed summation resulted in an overflow.</p>
6198 %res = call {i32, i1} @llvm.sadd.with.overflow.i32(i32 %a, i32 %b)
6199 %sum = extractvalue {i32, i1} %res, 0
6200 %obit = extractvalue {i32, i1} %res, 1
6201 br i1 %obit, label %overflow, label %normal
6206 <!-- _______________________________________________________________________ -->
6207 <div class="doc_subsubsection">
6208 <a name="int_uadd_overflow">'<tt>llvm.uadd.with.overflow.*</tt>' Intrinsics</a>
6211 <div class="doc_text">
6215 <p>This is an overloaded intrinsic. You can use <tt>llvm.uadd.with.overflow</tt>
6216 on any integer bit width.</p>
6219 declare {i16, i1} @llvm.uadd.with.overflow.i16(i16 %a, i16 %b)
6220 declare {i32, i1} @llvm.uadd.with.overflow.i32(i32 %a, i32 %b)
6221 declare {i64, i1} @llvm.uadd.with.overflow.i64(i64 %a, i64 %b)
6226 <p>The '<tt>llvm.uadd.with.overflow</tt>' family of intrinsic functions perform
6227 an unsigned addition of the two arguments, and indicate whether a carry occurred
6228 during the unsigned summation.</p>
6232 <p>The arguments (%a and %b) and the first element of the result structure may
6233 be of integer types of any bit width, but they must have the same bit width. The
6234 second element of the result structure must be of type <tt>i1</tt>. <tt>%a</tt>
6235 and <tt>%b</tt> are the two values that will undergo unsigned addition.</p>
6239 <p>The '<tt>llvm.uadd.with.overflow</tt>' family of intrinsic functions perform
6240 an unsigned addition of the two arguments. They return a structure — the
6241 first element of which is the sum, and the second element of which is a bit
6242 specifying if the unsigned summation resulted in a carry.</p>
6246 %res = call {i32, i1} @llvm.uadd.with.overflow.i32(i32 %a, i32 %b)
6247 %sum = extractvalue {i32, i1} %res, 0
6248 %obit = extractvalue {i32, i1} %res, 1
6249 br i1 %obit, label %carry, label %normal
6254 <!-- _______________________________________________________________________ -->
6255 <div class="doc_subsubsection">
6256 <a name="int_ssub_overflow">'<tt>llvm.ssub.with.overflow.*</tt>' Intrinsics</a>
6259 <div class="doc_text">
6263 <p>This is an overloaded intrinsic. You can use <tt>llvm.ssub.with.overflow</tt>
6264 on any integer bit width.</p>
6267 declare {i16, i1} @llvm.ssub.with.overflow.i16(i16 %a, i16 %b)
6268 declare {i32, i1} @llvm.ssub.with.overflow.i32(i32 %a, i32 %b)
6269 declare {i64, i1} @llvm.ssub.with.overflow.i64(i64 %a, i64 %b)
6274 <p>The '<tt>llvm.ssub.with.overflow</tt>' family of intrinsic functions perform
6275 a signed subtraction of the two arguments, and indicate whether an overflow
6276 occurred during the signed subtraction.</p>
6280 <p>The arguments (%a and %b) and the first element of the result structure may
6281 be of integer types of any bit width, but they must have the same bit width. The
6282 second element of the result structure must be of type <tt>i1</tt>. <tt>%a</tt>
6283 and <tt>%b</tt> are the two values that will undergo signed subtraction.</p>
6287 <p>The '<tt>llvm.ssub.with.overflow</tt>' family of intrinsic functions perform
6288 a signed subtraction of the two arguments. They return a structure — the
6289 first element of which is the subtraction, and the second element of which is a bit
6290 specifying if the signed subtraction resulted in an overflow.</p>
6294 %res = call {i32, i1} @llvm.ssub.with.overflow.i32(i32 %a, i32 %b)
6295 %sum = extractvalue {i32, i1} %res, 0
6296 %obit = extractvalue {i32, i1} %res, 1
6297 br i1 %obit, label %overflow, label %normal
6302 <!-- _______________________________________________________________________ -->
6303 <div class="doc_subsubsection">
6304 <a name="int_usub_overflow">'<tt>llvm.usub.with.overflow.*</tt>' Intrinsics</a>
6307 <div class="doc_text">
6311 <p>This is an overloaded intrinsic. You can use <tt>llvm.usub.with.overflow</tt>
6312 on any integer bit width.</p>
6315 declare {i16, i1} @llvm.usub.with.overflow.i16(i16 %a, i16 %b)
6316 declare {i32, i1} @llvm.usub.with.overflow.i32(i32 %a, i32 %b)
6317 declare {i64, i1} @llvm.usub.with.overflow.i64(i64 %a, i64 %b)
6322 <p>The '<tt>llvm.usub.with.overflow</tt>' family of intrinsic functions perform
6323 an unsigned subtraction of the two arguments, and indicate whether an overflow
6324 occurred during the unsigned subtraction.</p>
6328 <p>The arguments (%a and %b) and the first element of the result structure may
6329 be of integer types of any bit width, but they must have the same bit width. The
6330 second element of the result structure must be of type <tt>i1</tt>. <tt>%a</tt>
6331 and <tt>%b</tt> are the two values that will undergo unsigned subtraction.</p>
6335 <p>The '<tt>llvm.usub.with.overflow</tt>' family of intrinsic functions perform
6336 an unsigned subtraction of the two arguments. They return a structure — the
6337 first element of which is the subtraction, and the second element of which is a bit
6338 specifying if the unsigned subtraction resulted in an overflow.</p>
6342 %res = call {i32, i1} @llvm.usub.with.overflow.i32(i32 %a, i32 %b)
6343 %sum = extractvalue {i32, i1} %res, 0
6344 %obit = extractvalue {i32, i1} %res, 1
6345 br i1 %obit, label %overflow, label %normal
6350 <!-- _______________________________________________________________________ -->
6351 <div class="doc_subsubsection">
6352 <a name="int_smul_overflow">'<tt>llvm.smul.with.overflow.*</tt>' Intrinsics</a>
6355 <div class="doc_text">
6359 <p>This is an overloaded intrinsic. You can use <tt>llvm.smul.with.overflow</tt>
6360 on any integer bit width.</p>
6363 declare {i16, i1} @llvm.smul.with.overflow.i16(i16 %a, i16 %b)
6364 declare {i32, i1} @llvm.smul.with.overflow.i32(i32 %a, i32 %b)
6365 declare {i64, i1} @llvm.smul.with.overflow.i64(i64 %a, i64 %b)
6370 <p>The '<tt>llvm.smul.with.overflow</tt>' family of intrinsic functions perform
6371 a signed multiplication of the two arguments, and indicate whether an overflow
6372 occurred during the signed multiplication.</p>
6376 <p>The arguments (%a and %b) and the first element of the result structure may
6377 be of integer types of any bit width, but they must have the same bit width. The
6378 second element of the result structure must be of type <tt>i1</tt>. <tt>%a</tt>
6379 and <tt>%b</tt> are the two values that will undergo signed multiplication.</p>
6383 <p>The '<tt>llvm.smul.with.overflow</tt>' family of intrinsic functions perform
6384 a signed multiplication of the two arguments. They return a structure —
6385 the first element of which is the multiplication, and the second element of
6386 which is a bit specifying if the signed multiplication resulted in an
6391 %res = call {i32, i1} @llvm.smul.with.overflow.i32(i32 %a, i32 %b)
6392 %sum = extractvalue {i32, i1} %res, 0
6393 %obit = extractvalue {i32, i1} %res, 1
6394 br i1 %obit, label %overflow, label %normal
6399 <!-- _______________________________________________________________________ -->
6400 <div class="doc_subsubsection">
6401 <a name="int_umul_overflow">'<tt>llvm.umul.with.overflow.*</tt>' Intrinsics</a>
6404 <div class="doc_text">
6408 <p>This is an overloaded intrinsic. You can use <tt>llvm.umul.with.overflow</tt>
6409 on any integer bit width.</p>
6412 declare {i16, i1} @llvm.umul.with.overflow.i16(i16 %a, i16 %b)
6413 declare {i32, i1} @llvm.umul.with.overflow.i32(i32 %a, i32 %b)
6414 declare {i64, i1} @llvm.umul.with.overflow.i64(i64 %a, i64 %b)
6419 <p><i><b>Warning:</b> '<tt>llvm.umul.with.overflow</tt>' is badly broken. It is
6420 actively being fixed, but it should not currently be used!</i></p>
6422 <p>The '<tt>llvm.umul.with.overflow</tt>' family of intrinsic functions perform
6423 a unsigned multiplication of the two arguments, and indicate whether an overflow
6424 occurred during the unsigned multiplication.</p>
6428 <p>The arguments (%a and %b) and the first element of the result structure may
6429 be of integer types of any bit width, but they must have the same bit width. The
6430 second element of the result structure must be of type <tt>i1</tt>. <tt>%a</tt>
6431 and <tt>%b</tt> are the two values that will undergo unsigned
6436 <p>The '<tt>llvm.umul.with.overflow</tt>' family of intrinsic functions perform
6437 an unsigned multiplication of the two arguments. They return a structure —
6438 the first element of which is the multiplication, and the second element of
6439 which is a bit specifying if the unsigned multiplication resulted in an
6444 %res = call {i32, i1} @llvm.umul.with.overflow.i32(i32 %a, i32 %b)
6445 %sum = extractvalue {i32, i1} %res, 0
6446 %obit = extractvalue {i32, i1} %res, 1
6447 br i1 %obit, label %overflow, label %normal
6452 <!-- ======================================================================= -->
6453 <div class="doc_subsection">
6454 <a name="int_debugger">Debugger Intrinsics</a>
6457 <div class="doc_text">
6459 The LLVM debugger intrinsics (which all start with <tt>llvm.dbg.</tt> prefix),
6460 are described in the <a
6461 href="SourceLevelDebugging.html#format_common_intrinsics">LLVM Source Level
6462 Debugging</a> document.
6467 <!-- ======================================================================= -->
6468 <div class="doc_subsection">
6469 <a name="int_eh">Exception Handling Intrinsics</a>
6472 <div class="doc_text">
6473 <p> The LLVM exception handling intrinsics (which all start with
6474 <tt>llvm.eh.</tt> prefix), are described in the <a
6475 href="ExceptionHandling.html#format_common_intrinsics">LLVM Exception
6476 Handling</a> document. </p>
6479 <!-- ======================================================================= -->
6480 <div class="doc_subsection">
6481 <a name="int_trampoline">Trampoline Intrinsic</a>
6484 <div class="doc_text">
6486 This intrinsic makes it possible to excise one parameter, marked with
6487 the <tt>nest</tt> attribute, from a function. The result is a callable
6488 function pointer lacking the nest parameter - the caller does not need
6489 to provide a value for it. Instead, the value to use is stored in
6490 advance in a "trampoline", a block of memory usually allocated
6491 on the stack, which also contains code to splice the nest value into the
6492 argument list. This is used to implement the GCC nested function address
6496 For example, if the function is
6497 <tt>i32 f(i8* nest %c, i32 %x, i32 %y)</tt> then the resulting function
6498 pointer has signature <tt>i32 (i32, i32)*</tt>. It can be created as follows:</p>
6500 %tramp = alloca [10 x i8], align 4 ; size and alignment only correct for X86
6501 %tramp1 = getelementptr [10 x i8]* %tramp, i32 0, i32 0
6502 %p = call i8* @llvm.init.trampoline( i8* %tramp1, i8* bitcast (i32 (i8* nest , i32, i32)* @f to i8*), i8* %nval )
6503 %fp = bitcast i8* %p to i32 (i32, i32)*
6505 <p>The call <tt>%val = call i32 %fp( i32 %x, i32 %y )</tt> is then equivalent
6506 to <tt>%val = call i32 %f( i8* %nval, i32 %x, i32 %y )</tt>.</p>
6509 <!-- _______________________________________________________________________ -->
6510 <div class="doc_subsubsection">
6511 <a name="int_it">'<tt>llvm.init.trampoline</tt>' Intrinsic</a>
6513 <div class="doc_text">
6516 declare i8* @llvm.init.trampoline(i8* <tramp>, i8* <func>, i8* <nval>)
6520 This fills the memory pointed to by <tt>tramp</tt> with code
6521 and returns a function pointer suitable for executing it.
6525 The <tt>llvm.init.trampoline</tt> intrinsic takes three arguments, all
6526 pointers. The <tt>tramp</tt> argument must point to a sufficiently large
6527 and sufficiently aligned block of memory; this memory is written to by the
6528 intrinsic. Note that the size and the alignment are target-specific - LLVM
6529 currently provides no portable way of determining them, so a front-end that
6530 generates this intrinsic needs to have some target-specific knowledge.
6531 The <tt>func</tt> argument must hold a function bitcast to an <tt>i8*</tt>.
6535 The block of memory pointed to by <tt>tramp</tt> is filled with target
6536 dependent code, turning it into a function. A pointer to this function is
6537 returned, but needs to be bitcast to an
6538 <a href="#int_trampoline">appropriate function pointer type</a>
6539 before being called. The new function's signature is the same as that of
6540 <tt>func</tt> with any arguments marked with the <tt>nest</tt> attribute
6541 removed. At most one such <tt>nest</tt> argument is allowed, and it must be
6542 of pointer type. Calling the new function is equivalent to calling
6543 <tt>func</tt> with the same argument list, but with <tt>nval</tt> used for the
6544 missing <tt>nest</tt> argument. If, after calling
6545 <tt>llvm.init.trampoline</tt>, the memory pointed to by <tt>tramp</tt> is
6546 modified, then the effect of any later call to the returned function pointer is
6551 <!-- ======================================================================= -->
6552 <div class="doc_subsection">
6553 <a name="int_atomics">Atomic Operations and Synchronization Intrinsics</a>
6556 <div class="doc_text">
6558 These intrinsic functions expand the "universal IR" of LLVM to represent
6559 hardware constructs for atomic operations and memory synchronization. This
6560 provides an interface to the hardware, not an interface to the programmer. It
6561 is aimed at a low enough level to allow any programming models or APIs
6562 (Application Programming Interfaces) which
6563 need atomic behaviors to map cleanly onto it. It is also modeled primarily on
6564 hardware behavior. Just as hardware provides a "universal IR" for source
6565 languages, it also provides a starting point for developing a "universal"
6566 atomic operation and synchronization IR.
6569 These do <em>not</em> form an API such as high-level threading libraries,
6570 software transaction memory systems, atomic primitives, and intrinsic
6571 functions as found in BSD, GNU libc, atomic_ops, APR, and other system and
6572 application libraries. The hardware interface provided by LLVM should allow
6573 a clean implementation of all of these APIs and parallel programming models.
6574 No one model or paradigm should be selected above others unless the hardware
6575 itself ubiquitously does so.
6580 <!-- _______________________________________________________________________ -->
6581 <div class="doc_subsubsection">
6582 <a name="int_memory_barrier">'<tt>llvm.memory.barrier</tt>' Intrinsic</a>
6584 <div class="doc_text">
6587 declare void @llvm.memory.barrier( i1 <ll>, i1 <ls>, i1 <sl>, i1 <ss>,
6593 The <tt>llvm.memory.barrier</tt> intrinsic guarantees ordering between
6594 specific pairs of memory access types.
6598 The <tt>llvm.memory.barrier</tt> intrinsic requires five boolean arguments.
6599 The first four arguments enables a specific barrier as listed below. The fith
6600 argument specifies that the barrier applies to io or device or uncached memory.
6604 <li><tt>ll</tt>: load-load barrier</li>
6605 <li><tt>ls</tt>: load-store barrier</li>
6606 <li><tt>sl</tt>: store-load barrier</li>
6607 <li><tt>ss</tt>: store-store barrier</li>
6608 <li><tt>device</tt>: barrier applies to device and uncached memory also.</li>
6612 This intrinsic causes the system to enforce some ordering constraints upon
6613 the loads and stores of the program. This barrier does not indicate
6614 <em>when</em> any events will occur, it only enforces an <em>order</em> in
6615 which they occur. For any of the specified pairs of load and store operations
6616 (f.ex. load-load, or store-load), all of the first operations preceding the
6617 barrier will complete before any of the second operations succeeding the
6618 barrier begin. Specifically the semantics for each pairing is as follows:
6621 <li><tt>ll</tt>: All loads before the barrier must complete before any load
6622 after the barrier begins.</li>
6624 <li><tt>ls</tt>: All loads before the barrier must complete before any
6625 store after the barrier begins.</li>
6626 <li><tt>ss</tt>: All stores before the barrier must complete before any
6627 store after the barrier begins.</li>
6628 <li><tt>sl</tt>: All stores before the barrier must complete before any
6629 load after the barrier begins.</li>
6632 These semantics are applied with a logical "and" behavior when more than one
6633 is enabled in a single memory barrier intrinsic.
6636 Backends may implement stronger barriers than those requested when they do not
6637 support as fine grained a barrier as requested. Some architectures do not
6638 need all types of barriers and on such architectures, these become noops.
6645 %result1 = load i32* %ptr <i>; yields {i32}:result1 = 4</i>
6646 call void @llvm.memory.barrier( i1 false, i1 true, i1 false, i1 false )
6647 <i>; guarantee the above finishes</i>
6648 store i32 8, %ptr <i>; before this begins</i>
6652 <!-- _______________________________________________________________________ -->
6653 <div class="doc_subsubsection">
6654 <a name="int_atomic_cmp_swap">'<tt>llvm.atomic.cmp.swap.*</tt>' Intrinsic</a>
6656 <div class="doc_text">
6659 This is an overloaded intrinsic. You can use <tt>llvm.atomic.cmp.swap</tt> on
6660 any integer bit width and for different address spaces. Not all targets
6661 support all bit widths however.</p>
6664 declare i8 @llvm.atomic.cmp.swap.i8.p0i8( i8* <ptr>, i8 <cmp>, i8 <val> )
6665 declare i16 @llvm.atomic.cmp.swap.i16.p0i16( i16* <ptr>, i16 <cmp>, i16 <val> )
6666 declare i32 @llvm.atomic.cmp.swap.i32.p0i32( i32* <ptr>, i32 <cmp>, i32 <val> )
6667 declare i64 @llvm.atomic.cmp.swap.i64.p0i64( i64* <ptr>, i64 <cmp>, i64 <val> )
6672 This loads a value in memory and compares it to a given value. If they are
6673 equal, it stores a new value into the memory.
6677 The <tt>llvm.atomic.cmp.swap</tt> intrinsic takes three arguments. The result as
6678 well as both <tt>cmp</tt> and <tt>val</tt> must be integer values with the
6679 same bit width. The <tt>ptr</tt> argument must be a pointer to a value of
6680 this integer type. While any bit width integer may be used, targets may only
6681 lower representations they support in hardware.
6686 This entire intrinsic must be executed atomically. It first loads the value
6687 in memory pointed to by <tt>ptr</tt> and compares it with the value
6688 <tt>cmp</tt>. If they are equal, <tt>val</tt> is stored into the memory. The
6689 loaded value is yielded in all cases. This provides the equivalent of an
6690 atomic compare-and-swap operation within the SSA framework.
6698 %val1 = add i32 4, 4
6699 %result1 = call i32 @llvm.atomic.cmp.swap.i32.p0i32( i32* %ptr, i32 4, %val1 )
6700 <i>; yields {i32}:result1 = 4</i>
6701 %stored1 = icmp eq i32 %result1, 4 <i>; yields {i1}:stored1 = true</i>
6702 %memval1 = load i32* %ptr <i>; yields {i32}:memval1 = 8</i>
6704 %val2 = add i32 1, 1
6705 %result2 = call i32 @llvm.atomic.cmp.swap.i32.p0i32( i32* %ptr, i32 5, %val2 )
6706 <i>; yields {i32}:result2 = 8</i>
6707 %stored2 = icmp eq i32 %result2, 5 <i>; yields {i1}:stored2 = false</i>
6709 %memval2 = load i32* %ptr <i>; yields {i32}:memval2 = 8</i>
6713 <!-- _______________________________________________________________________ -->
6714 <div class="doc_subsubsection">
6715 <a name="int_atomic_swap">'<tt>llvm.atomic.swap.*</tt>' Intrinsic</a>
6717 <div class="doc_text">
6721 This is an overloaded intrinsic. You can use <tt>llvm.atomic.swap</tt> on any
6722 integer bit width. Not all targets support all bit widths however.</p>
6724 declare i8 @llvm.atomic.swap.i8.p0i8( i8* <ptr>, i8 <val> )
6725 declare i16 @llvm.atomic.swap.i16.p0i16( i16* <ptr>, i16 <val> )
6726 declare i32 @llvm.atomic.swap.i32.p0i32( i32* <ptr>, i32 <val> )
6727 declare i64 @llvm.atomic.swap.i64.p0i64( i64* <ptr>, i64 <val> )
6732 This intrinsic loads the value stored in memory at <tt>ptr</tt> and yields
6733 the value from memory. It then stores the value in <tt>val</tt> in the memory
6739 The <tt>llvm.atomic.swap</tt> intrinsic takes two arguments. Both the
6740 <tt>val</tt> argument and the result must be integers of the same bit width.
6741 The first argument, <tt>ptr</tt>, must be a pointer to a value of this
6742 integer type. The targets may only lower integer representations they
6747 This intrinsic loads the value pointed to by <tt>ptr</tt>, yields it, and
6748 stores <tt>val</tt> back into <tt>ptr</tt> atomically. This provides the
6749 equivalent of an atomic swap operation within the SSA framework.
6757 %val1 = add i32 4, 4
6758 %result1 = call i32 @llvm.atomic.swap.i32.p0i32( i32* %ptr, i32 %val1 )
6759 <i>; yields {i32}:result1 = 4</i>
6760 %stored1 = icmp eq i32 %result1, 4 <i>; yields {i1}:stored1 = true</i>
6761 %memval1 = load i32* %ptr <i>; yields {i32}:memval1 = 8</i>
6763 %val2 = add i32 1, 1
6764 %result2 = call i32 @llvm.atomic.swap.i32.p0i32( i32* %ptr, i32 %val2 )
6765 <i>; yields {i32}:result2 = 8</i>
6767 %stored2 = icmp eq i32 %result2, 8 <i>; yields {i1}:stored2 = true</i>
6768 %memval2 = load i32* %ptr <i>; yields {i32}:memval2 = 2</i>
6772 <!-- _______________________________________________________________________ -->
6773 <div class="doc_subsubsection">
6774 <a name="int_atomic_load_add">'<tt>llvm.atomic.load.add.*</tt>' Intrinsic</a>
6777 <div class="doc_text">
6780 This is an overloaded intrinsic. You can use <tt>llvm.atomic.load.add</tt> on any
6781 integer bit width. Not all targets support all bit widths however.</p>
6783 declare i8 @llvm.atomic.load.add.i8..p0i8( i8* <ptr>, i8 <delta> )
6784 declare i16 @llvm.atomic.load.add.i16..p0i16( i16* <ptr>, i16 <delta> )
6785 declare i32 @llvm.atomic.load.add.i32..p0i32( i32* <ptr>, i32 <delta> )
6786 declare i64 @llvm.atomic.load.add.i64..p0i64( i64* <ptr>, i64 <delta> )
6791 This intrinsic adds <tt>delta</tt> to the value stored in memory at
6792 <tt>ptr</tt>. It yields the original value at <tt>ptr</tt>.
6797 The intrinsic takes two arguments, the first a pointer to an integer value
6798 and the second an integer value. The result is also an integer value. These
6799 integer types can have any bit width, but they must all have the same bit
6800 width. The targets may only lower integer representations they support.
6804 This intrinsic does a series of operations atomically. It first loads the
6805 value stored at <tt>ptr</tt>. It then adds <tt>delta</tt>, stores the result
6806 to <tt>ptr</tt>. It yields the original value stored at <tt>ptr</tt>.
6813 %result1 = call i32 @llvm.atomic.load.add.i32.p0i32( i32* %ptr, i32 4 )
6814 <i>; yields {i32}:result1 = 4</i>
6815 %result2 = call i32 @llvm.atomic.load.add.i32.p0i32( i32* %ptr, i32 2 )
6816 <i>; yields {i32}:result2 = 8</i>
6817 %result3 = call i32 @llvm.atomic.load.add.i32.p0i32( i32* %ptr, i32 5 )
6818 <i>; yields {i32}:result3 = 10</i>
6819 %memval1 = load i32* %ptr <i>; yields {i32}:memval1 = 15</i>
6823 <!-- _______________________________________________________________________ -->
6824 <div class="doc_subsubsection">
6825 <a name="int_atomic_load_sub">'<tt>llvm.atomic.load.sub.*</tt>' Intrinsic</a>
6828 <div class="doc_text">
6831 This is an overloaded intrinsic. You can use <tt>llvm.atomic.load.sub</tt> on
6832 any integer bit width and for different address spaces. Not all targets
6833 support all bit widths however.</p>
6835 declare i8 @llvm.atomic.load.sub.i8.p0i32( i8* <ptr>, i8 <delta> )
6836 declare i16 @llvm.atomic.load.sub.i16.p0i32( i16* <ptr>, i16 <delta> )
6837 declare i32 @llvm.atomic.load.sub.i32.p0i32( i32* <ptr>, i32 <delta> )
6838 declare i64 @llvm.atomic.load.sub.i64.p0i32( i64* <ptr>, i64 <delta> )
6843 This intrinsic subtracts <tt>delta</tt> to the value stored in memory at
6844 <tt>ptr</tt>. It yields the original value at <tt>ptr</tt>.
6849 The intrinsic takes two arguments, the first a pointer to an integer value
6850 and the second an integer value. The result is also an integer value. These
6851 integer types can have any bit width, but they must all have the same bit
6852 width. The targets may only lower integer representations they support.
6856 This intrinsic does a series of operations atomically. It first loads the
6857 value stored at <tt>ptr</tt>. It then subtracts <tt>delta</tt>, stores the
6858 result to <tt>ptr</tt>. It yields the original value stored at <tt>ptr</tt>.
6865 %result1 = call i32 @llvm.atomic.load.sub.i32.p0i32( i32* %ptr, i32 4 )
6866 <i>; yields {i32}:result1 = 8</i>
6867 %result2 = call i32 @llvm.atomic.load.sub.i32.p0i32( i32* %ptr, i32 2 )
6868 <i>; yields {i32}:result2 = 4</i>
6869 %result3 = call i32 @llvm.atomic.load.sub.i32.p0i32( i32* %ptr, i32 5 )
6870 <i>; yields {i32}:result3 = 2</i>
6871 %memval1 = load i32* %ptr <i>; yields {i32}:memval1 = -3</i>
6875 <!-- _______________________________________________________________________ -->
6876 <div class="doc_subsubsection">
6877 <a name="int_atomic_load_and">'<tt>llvm.atomic.load.and.*</tt>' Intrinsic</a><br>
6878 <a name="int_atomic_load_nand">'<tt>llvm.atomic.load.nand.*</tt>' Intrinsic</a><br>
6879 <a name="int_atomic_load_or">'<tt>llvm.atomic.load.or.*</tt>' Intrinsic</a><br>
6880 <a name="int_atomic_load_xor">'<tt>llvm.atomic.load.xor.*</tt>' Intrinsic</a><br>
6883 <div class="doc_text">
6886 These are overloaded intrinsics. You can use <tt>llvm.atomic.load_and</tt>,
6887 <tt>llvm.atomic.load_nand</tt>, <tt>llvm.atomic.load_or</tt>, and
6888 <tt>llvm.atomic.load_xor</tt> on any integer bit width and for different
6889 address spaces. Not all targets support all bit widths however.</p>
6891 declare i8 @llvm.atomic.load.and.i8.p0i8( i8* <ptr>, i8 <delta> )
6892 declare i16 @llvm.atomic.load.and.i16.p0i16( i16* <ptr>, i16 <delta> )
6893 declare i32 @llvm.atomic.load.and.i32.p0i32( i32* <ptr>, i32 <delta> )
6894 declare i64 @llvm.atomic.load.and.i64.p0i64( i64* <ptr>, i64 <delta> )
6899 declare i8 @llvm.atomic.load.or.i8.p0i8( i8* <ptr>, i8 <delta> )
6900 declare i16 @llvm.atomic.load.or.i16.p0i16( i16* <ptr>, i16 <delta> )
6901 declare i32 @llvm.atomic.load.or.i32.p0i32( i32* <ptr>, i32 <delta> )
6902 declare i64 @llvm.atomic.load.or.i64.p0i64( i64* <ptr>, i64 <delta> )
6907 declare i8 @llvm.atomic.load.nand.i8.p0i32( i8* <ptr>, i8 <delta> )
6908 declare i16 @llvm.atomic.load.nand.i16.p0i32( i16* <ptr>, i16 <delta> )
6909 declare i32 @llvm.atomic.load.nand.i32.p0i32( i32* <ptr>, i32 <delta> )
6910 declare i64 @llvm.atomic.load.nand.i64.p0i32( i64* <ptr>, i64 <delta> )
6915 declare i8 @llvm.atomic.load.xor.i8.p0i32( i8* <ptr>, i8 <delta> )
6916 declare i16 @llvm.atomic.load.xor.i16.p0i32( i16* <ptr>, i16 <delta> )
6917 declare i32 @llvm.atomic.load.xor.i32.p0i32( i32* <ptr>, i32 <delta> )
6918 declare i64 @llvm.atomic.load.xor.i64.p0i32( i64* <ptr>, i64 <delta> )
6923 These intrinsics bitwise the operation (and, nand, or, xor) <tt>delta</tt> to
6924 the value stored in memory at <tt>ptr</tt>. It yields the original value
6930 These intrinsics take two arguments, the first a pointer to an integer value
6931 and the second an integer value. The result is also an integer value. These
6932 integer types can have any bit width, but they must all have the same bit
6933 width. The targets may only lower integer representations they support.
6937 These intrinsics does a series of operations atomically. They first load the
6938 value stored at <tt>ptr</tt>. They then do the bitwise operation
6939 <tt>delta</tt>, store the result to <tt>ptr</tt>. They yield the original
6940 value stored at <tt>ptr</tt>.
6946 store i32 0x0F0F, %ptr
6947 %result0 = call i32 @llvm.atomic.load.nand.i32.p0i32( i32* %ptr, i32 0xFF )
6948 <i>; yields {i32}:result0 = 0x0F0F</i>
6949 %result1 = call i32 @llvm.atomic.load.and.i32.p0i32( i32* %ptr, i32 0xFF )
6950 <i>; yields {i32}:result1 = 0xFFFFFFF0</i>
6951 %result2 = call i32 @llvm.atomic.load.or.i32.p0i32( i32* %ptr, i32 0F )
6952 <i>; yields {i32}:result2 = 0xF0</i>
6953 %result3 = call i32 @llvm.atomic.load.xor.i32.p0i32( i32* %ptr, i32 0F )
6954 <i>; yields {i32}:result3 = FF</i>
6955 %memval1 = load i32* %ptr <i>; yields {i32}:memval1 = F0</i>
6960 <!-- _______________________________________________________________________ -->
6961 <div class="doc_subsubsection">
6962 <a name="int_atomic_load_max">'<tt>llvm.atomic.load.max.*</tt>' Intrinsic</a><br>
6963 <a name="int_atomic_load_min">'<tt>llvm.atomic.load.min.*</tt>' Intrinsic</a><br>
6964 <a name="int_atomic_load_umax">'<tt>llvm.atomic.load.umax.*</tt>' Intrinsic</a><br>
6965 <a name="int_atomic_load_umin">'<tt>llvm.atomic.load.umin.*</tt>' Intrinsic</a><br>
6968 <div class="doc_text">
6971 These are overloaded intrinsics. You can use <tt>llvm.atomic.load_max</tt>,
6972 <tt>llvm.atomic.load_min</tt>, <tt>llvm.atomic.load_umax</tt>, and
6973 <tt>llvm.atomic.load_umin</tt> on any integer bit width and for different
6974 address spaces. Not all targets
6975 support all bit widths however.</p>
6977 declare i8 @llvm.atomic.load.max.i8.p0i8( i8* <ptr>, i8 <delta> )
6978 declare i16 @llvm.atomic.load.max.i16.p0i16( i16* <ptr>, i16 <delta> )
6979 declare i32 @llvm.atomic.load.max.i32.p0i32( i32* <ptr>, i32 <delta> )
6980 declare i64 @llvm.atomic.load.max.i64.p0i64( i64* <ptr>, i64 <delta> )
6985 declare i8 @llvm.atomic.load.min.i8.p0i8( i8* <ptr>, i8 <delta> )
6986 declare i16 @llvm.atomic.load.min.i16.p0i16( i16* <ptr>, i16 <delta> )
6987 declare i32 @llvm.atomic.load.min.i32..p0i32( i32* <ptr>, i32 <delta> )
6988 declare i64 @llvm.atomic.load.min.i64..p0i64( i64* <ptr>, i64 <delta> )
6993 declare i8 @llvm.atomic.load.umax.i8.p0i8( i8* <ptr>, i8 <delta> )
6994 declare i16 @llvm.atomic.load.umax.i16.p0i16( i16* <ptr>, i16 <delta> )
6995 declare i32 @llvm.atomic.load.umax.i32.p0i32( i32* <ptr>, i32 <delta> )
6996 declare i64 @llvm.atomic.load.umax.i64.p0i64( i64* <ptr>, i64 <delta> )
7001 declare i8 @llvm.atomic.load.umin.i8..p0i8( i8* <ptr>, i8 <delta> )
7002 declare i16 @llvm.atomic.load.umin.i16.p0i16( i16* <ptr>, i16 <delta> )
7003 declare i32 @llvm.atomic.load.umin.i32..p0i32( i32* <ptr>, i32 <delta> )
7004 declare i64 @llvm.atomic.load.umin.i64..p0i64( i64* <ptr>, i64 <delta> )
7009 These intrinsics takes the signed or unsigned minimum or maximum of
7010 <tt>delta</tt> and the value stored in memory at <tt>ptr</tt>. It yields the
7011 original value at <tt>ptr</tt>.
7016 These intrinsics take two arguments, the first a pointer to an integer value
7017 and the second an integer value. The result is also an integer value. These
7018 integer types can have any bit width, but they must all have the same bit
7019 width. The targets may only lower integer representations they support.
7023 These intrinsics does a series of operations atomically. They first load the
7024 value stored at <tt>ptr</tt>. They then do the signed or unsigned min or max
7025 <tt>delta</tt> and the value, store the result to <tt>ptr</tt>. They yield
7026 the original value stored at <tt>ptr</tt>.
7033 %result0 = call i32 @llvm.atomic.load.min.i32.p0i32( i32* %ptr, i32 -2 )
7034 <i>; yields {i32}:result0 = 7</i>
7035 %result1 = call i32 @llvm.atomic.load.max.i32.p0i32( i32* %ptr, i32 8 )
7036 <i>; yields {i32}:result1 = -2</i>
7037 %result2 = call i32 @llvm.atomic.load.umin.i32.p0i32( i32* %ptr, i32 10 )
7038 <i>; yields {i32}:result2 = 8</i>
7039 %result3 = call i32 @llvm.atomic.load.umax.i32.p0i32( i32* %ptr, i32 30 )
7040 <i>; yields {i32}:result3 = 8</i>
7041 %memval1 = load i32* %ptr <i>; yields {i32}:memval1 = 30</i>
7045 <!-- ======================================================================= -->
7046 <div class="doc_subsection">
7047 <a name="int_general">General Intrinsics</a>
7050 <div class="doc_text">
7051 <p> This class of intrinsics is designed to be generic and has
7052 no specific purpose. </p>
7055 <!-- _______________________________________________________________________ -->
7056 <div class="doc_subsubsection">
7057 <a name="int_var_annotation">'<tt>llvm.var.annotation</tt>' Intrinsic</a>
7060 <div class="doc_text">
7064 declare void @llvm.var.annotation(i8* <val>, i8* <str>, i8* <str>, i32 <int> )
7070 The '<tt>llvm.var.annotation</tt>' intrinsic
7076 The first argument is a pointer to a value, the second is a pointer to a
7077 global string, the third is a pointer to a global string which is the source
7078 file name, and the last argument is the line number.
7084 This intrinsic allows annotation of local variables with arbitrary strings.
7085 This can be useful for special purpose optimizations that want to look for these
7086 annotations. These have no other defined use, they are ignored by code
7087 generation and optimization.
7091 <!-- _______________________________________________________________________ -->
7092 <div class="doc_subsubsection">
7093 <a name="int_annotation">'<tt>llvm.annotation.*</tt>' Intrinsic</a>
7096 <div class="doc_text">
7099 <p>This is an overloaded intrinsic. You can use '<tt>llvm.annotation</tt>' on
7100 any integer bit width.
7103 declare i8 @llvm.annotation.i8(i8 <val>, i8* <str>, i8* <str>, i32 <int> )
7104 declare i16 @llvm.annotation.i16(i16 <val>, i8* <str>, i8* <str>, i32 <int> )
7105 declare i32 @llvm.annotation.i32(i32 <val>, i8* <str>, i8* <str>, i32 <int> )
7106 declare i64 @llvm.annotation.i64(i64 <val>, i8* <str>, i8* <str>, i32 <int> )
7107 declare i256 @llvm.annotation.i256(i256 <val>, i8* <str>, i8* <str>, i32 <int> )
7113 The '<tt>llvm.annotation</tt>' intrinsic.
7119 The first argument is an integer value (result of some expression),
7120 the second is a pointer to a global string, the third is a pointer to a global
7121 string which is the source file name, and the last argument is the line number.
7122 It returns the value of the first argument.
7128 This intrinsic allows annotations to be put on arbitrary expressions
7129 with arbitrary strings. This can be useful for special purpose optimizations
7130 that want to look for these annotations. These have no other defined use, they
7131 are ignored by code generation and optimization.
7135 <!-- _______________________________________________________________________ -->
7136 <div class="doc_subsubsection">
7137 <a name="int_trap">'<tt>llvm.trap</tt>' Intrinsic</a>
7140 <div class="doc_text">
7144 declare void @llvm.trap()
7150 The '<tt>llvm.trap</tt>' intrinsic
7162 This intrinsics is lowered to the target dependent trap instruction. If the
7163 target does not have a trap instruction, this intrinsic will be lowered to the
7164 call of the abort() function.
7168 <!-- _______________________________________________________________________ -->
7169 <div class="doc_subsubsection">
7170 <a name="int_stackprotector">'<tt>llvm.stackprotector</tt>' Intrinsic</a>
7172 <div class="doc_text">
7175 declare void @llvm.stackprotector( i8* <guard>, i8** <slot> )
7180 The <tt>llvm.stackprotector</tt> intrinsic takes the <tt>guard</tt> and stores
7181 it onto the stack at <tt>slot</tt>. The stack slot is adjusted to ensure that
7182 it is placed on the stack before local variables.
7186 The <tt>llvm.stackprotector</tt> intrinsic requires two pointer arguments. The
7187 first argument is the value loaded from the stack guard
7188 <tt>@__stack_chk_guard</tt>. The second variable is an <tt>alloca</tt> that
7189 has enough space to hold the value of the guard.
7193 This intrinsic causes the prologue/epilogue inserter to force the position of
7194 the <tt>AllocaInst</tt> stack slot to be before local variables on the
7195 stack. This is to ensure that if a local variable on the stack is overwritten,
7196 it will destroy the value of the guard. When the function exits, the guard on
7197 the stack is checked against the original guard. If they're different, then
7198 the program aborts by calling the <tt>__stack_chk_fail()</tt> function.
7202 <!-- *********************************************************************** -->
7205 <a href="http://jigsaw.w3.org/css-validator/check/referer"><img
7206 src="http://jigsaw.w3.org/css-validator/images/vcss-blue" alt="Valid CSS"></a>
7207 <a href="http://validator.w3.org/check/referer"><img
7208 src="http://www.w3.org/Icons/valid-html401-blue" alt="Valid HTML 4.01"></a>
7210 <a href="mailto:sabre@nondot.org">Chris Lattner</a><br>
7211 <a href="http://llvm.org">The LLVM Compiler Infrastructure</a><br>
7212 Last modified: $Date$