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>
25 <li><a href="#linkage_private">'<tt>private</tt>' Linkage</a></li>
26 <li><a href="#linkage_linker_private">'<tt>linker_private</tt>' Linkage</a></li>
27 <li><a href="#linkage_internal">'<tt>internal</tt>' Linkage</a></li>
28 <li><a href="#linkage_available_externally">'<tt>available_externally</tt>' Linkage</a></li>
29 <li><a href="#linkage_linkonce">'<tt>linkonce</tt>' Linkage</a></li>
30 <li><a href="#linkage_common">'<tt>common</tt>' Linkage</a></li>
31 <li><a href="#linkage_weak">'<tt>weak</tt>' Linkage</a></li>
32 <li><a href="#linkage_appending">'<tt>appending</tt>' Linkage</a></li>
33 <li><a href="#linkage_externweak">'<tt>extern_weak</tt>' Linkage</a></li>
34 <li><a href="#linkage_linkonce">'<tt>linkonce_odr</tt>' Linkage</a></li>
35 <li><a href="#linkage_weak">'<tt>weak_odr</tt>' Linkage</a></li>
36 <li><a href="#linkage_external">'<tt>externally visible</tt>' Linkage</a></li>
37 <li><a href="#linkage_dllimport">'<tt>dllimport</tt>' Linkage</a></li>
38 <li><a href="#linkage_dllexport">'<tt>dllexport</tt>' Linkage</a></li>
41 <li><a href="#callingconv">Calling Conventions</a></li>
42 <li><a href="#namedtypes">Named Types</a></li>
43 <li><a href="#globalvars">Global Variables</a></li>
44 <li><a href="#functionstructure">Functions</a></li>
45 <li><a href="#aliasstructure">Aliases</a></li>
46 <li><a href="#paramattrs">Parameter Attributes</a></li>
47 <li><a href="#fnattrs">Function Attributes</a></li>
48 <li><a href="#gc">Garbage Collector Names</a></li>
49 <li><a href="#moduleasm">Module-Level Inline Assembly</a></li>
50 <li><a href="#datalayout">Data Layout</a></li>
51 <li><a href="#pointeraliasing">Pointer Aliasing Rules</a></li>
54 <li><a href="#typesystem">Type System</a>
56 <li><a href="#t_classifications">Type Classifications</a></li>
57 <li><a href="#t_primitive">Primitive Types</a>
59 <li><a href="#t_floating">Floating Point Types</a></li>
60 <li><a href="#t_void">Void Type</a></li>
61 <li><a href="#t_label">Label Type</a></li>
62 <li><a href="#t_metadata">Metadata Type</a></li>
65 <li><a href="#t_derived">Derived Types</a>
67 <li><a href="#t_integer">Integer Type</a></li>
68 <li><a href="#t_array">Array Type</a></li>
69 <li><a href="#t_function">Function Type</a></li>
70 <li><a href="#t_pointer">Pointer Type</a></li>
71 <li><a href="#t_struct">Structure Type</a></li>
72 <li><a href="#t_pstruct">Packed Structure Type</a></li>
73 <li><a href="#t_vector">Vector Type</a></li>
74 <li><a href="#t_opaque">Opaque Type</a></li>
77 <li><a href="#t_uprefs">Type Up-references</a></li>
80 <li><a href="#constants">Constants</a>
82 <li><a href="#simpleconstants">Simple Constants</a></li>
83 <li><a href="#complexconstants">Complex Constants</a></li>
84 <li><a href="#globalconstants">Global Variable and Function Addresses</a></li>
85 <li><a href="#undefvalues">Undefined Values</a></li>
86 <li><a href="#constantexprs">Constant Expressions</a></li>
87 <li><a href="#metadata">Embedded Metadata</a></li>
90 <li><a href="#othervalues">Other Values</a>
92 <li><a href="#inlineasm">Inline Assembler Expressions</a></li>
95 <li><a href="#intrinsic_globals">Intrinsic Global Variables</a>
97 <li><a href="#intg_used">The '<tt>llvm.used</tt>' Global Variable</a></li>
98 <li><a href="#intg_compiler_used">The '<tt>llvm.compiler.used</tt>'
99 Global Variable</a></li>
100 <li><a href="#intg_global_ctors">The '<tt>llvm.global_ctors</tt>'
101 Global Variable</a></li>
102 <li><a href="#intg_global_dtors">The '<tt>llvm.global_dtors</tt>'
103 Global Variable</a></li>
106 <li><a href="#instref">Instruction Reference</a>
108 <li><a href="#terminators">Terminator Instructions</a>
110 <li><a href="#i_ret">'<tt>ret</tt>' Instruction</a></li>
111 <li><a href="#i_br">'<tt>br</tt>' Instruction</a></li>
112 <li><a href="#i_switch">'<tt>switch</tt>' Instruction</a></li>
113 <li><a href="#i_invoke">'<tt>invoke</tt>' Instruction</a></li>
114 <li><a href="#i_unwind">'<tt>unwind</tt>' Instruction</a></li>
115 <li><a href="#i_unreachable">'<tt>unreachable</tt>' Instruction</a></li>
118 <li><a href="#binaryops">Binary Operations</a>
120 <li><a href="#i_add">'<tt>add</tt>' Instruction</a></li>
121 <li><a href="#i_fadd">'<tt>fadd</tt>' Instruction</a></li>
122 <li><a href="#i_sub">'<tt>sub</tt>' Instruction</a></li>
123 <li><a href="#i_fsub">'<tt>fsub</tt>' Instruction</a></li>
124 <li><a href="#i_mul">'<tt>mul</tt>' Instruction</a></li>
125 <li><a href="#i_fmul">'<tt>fmul</tt>' Instruction</a></li>
126 <li><a href="#i_udiv">'<tt>udiv</tt>' Instruction</a></li>
127 <li><a href="#i_sdiv">'<tt>sdiv</tt>' Instruction</a></li>
128 <li><a href="#i_fdiv">'<tt>fdiv</tt>' Instruction</a></li>
129 <li><a href="#i_urem">'<tt>urem</tt>' Instruction</a></li>
130 <li><a href="#i_srem">'<tt>srem</tt>' Instruction</a></li>
131 <li><a href="#i_frem">'<tt>frem</tt>' Instruction</a></li>
134 <li><a href="#bitwiseops">Bitwise Binary Operations</a>
136 <li><a href="#i_shl">'<tt>shl</tt>' Instruction</a></li>
137 <li><a href="#i_lshr">'<tt>lshr</tt>' Instruction</a></li>
138 <li><a href="#i_ashr">'<tt>ashr</tt>' Instruction</a></li>
139 <li><a href="#i_and">'<tt>and</tt>' Instruction</a></li>
140 <li><a href="#i_or">'<tt>or</tt>' Instruction</a></li>
141 <li><a href="#i_xor">'<tt>xor</tt>' Instruction</a></li>
144 <li><a href="#vectorops">Vector Operations</a>
146 <li><a href="#i_extractelement">'<tt>extractelement</tt>' Instruction</a></li>
147 <li><a href="#i_insertelement">'<tt>insertelement</tt>' Instruction</a></li>
148 <li><a href="#i_shufflevector">'<tt>shufflevector</tt>' Instruction</a></li>
151 <li><a href="#aggregateops">Aggregate Operations</a>
153 <li><a href="#i_extractvalue">'<tt>extractvalue</tt>' Instruction</a></li>
154 <li><a href="#i_insertvalue">'<tt>insertvalue</tt>' Instruction</a></li>
157 <li><a href="#memoryops">Memory Access and Addressing Operations</a>
159 <li><a href="#i_malloc">'<tt>malloc</tt>' Instruction</a></li>
160 <li><a href="#i_free">'<tt>free</tt>' Instruction</a></li>
161 <li><a href="#i_alloca">'<tt>alloca</tt>' Instruction</a></li>
162 <li><a href="#i_load">'<tt>load</tt>' Instruction</a></li>
163 <li><a href="#i_store">'<tt>store</tt>' Instruction</a></li>
164 <li><a href="#i_getelementptr">'<tt>getelementptr</tt>' Instruction</a></li>
167 <li><a href="#convertops">Conversion Operations</a>
169 <li><a href="#i_trunc">'<tt>trunc .. to</tt>' Instruction</a></li>
170 <li><a href="#i_zext">'<tt>zext .. to</tt>' Instruction</a></li>
171 <li><a href="#i_sext">'<tt>sext .. to</tt>' Instruction</a></li>
172 <li><a href="#i_fptrunc">'<tt>fptrunc .. to</tt>' Instruction</a></li>
173 <li><a href="#i_fpext">'<tt>fpext .. to</tt>' Instruction</a></li>
174 <li><a href="#i_fptoui">'<tt>fptoui .. to</tt>' Instruction</a></li>
175 <li><a href="#i_fptosi">'<tt>fptosi .. to</tt>' Instruction</a></li>
176 <li><a href="#i_uitofp">'<tt>uitofp .. to</tt>' Instruction</a></li>
177 <li><a href="#i_sitofp">'<tt>sitofp .. to</tt>' Instruction</a></li>
178 <li><a href="#i_ptrtoint">'<tt>ptrtoint .. to</tt>' Instruction</a></li>
179 <li><a href="#i_inttoptr">'<tt>inttoptr .. to</tt>' Instruction</a></li>
180 <li><a href="#i_bitcast">'<tt>bitcast .. to</tt>' Instruction</a></li>
183 <li><a href="#otherops">Other Operations</a>
185 <li><a href="#i_icmp">'<tt>icmp</tt>' Instruction</a></li>
186 <li><a href="#i_fcmp">'<tt>fcmp</tt>' Instruction</a></li>
187 <li><a href="#i_phi">'<tt>phi</tt>' Instruction</a></li>
188 <li><a href="#i_select">'<tt>select</tt>' Instruction</a></li>
189 <li><a href="#i_call">'<tt>call</tt>' Instruction</a></li>
190 <li><a href="#i_va_arg">'<tt>va_arg</tt>' Instruction</a></li>
195 <li><a href="#intrinsics">Intrinsic Functions</a>
197 <li><a href="#int_varargs">Variable Argument Handling Intrinsics</a>
199 <li><a href="#int_va_start">'<tt>llvm.va_start</tt>' Intrinsic</a></li>
200 <li><a href="#int_va_end">'<tt>llvm.va_end</tt>' Intrinsic</a></li>
201 <li><a href="#int_va_copy">'<tt>llvm.va_copy</tt>' Intrinsic</a></li>
204 <li><a href="#int_gc">Accurate Garbage Collection Intrinsics</a>
206 <li><a href="#int_gcroot">'<tt>llvm.gcroot</tt>' Intrinsic</a></li>
207 <li><a href="#int_gcread">'<tt>llvm.gcread</tt>' Intrinsic</a></li>
208 <li><a href="#int_gcwrite">'<tt>llvm.gcwrite</tt>' Intrinsic</a></li>
211 <li><a href="#int_codegen">Code Generator Intrinsics</a>
213 <li><a href="#int_returnaddress">'<tt>llvm.returnaddress</tt>' Intrinsic</a></li>
214 <li><a href="#int_frameaddress">'<tt>llvm.frameaddress</tt>' Intrinsic</a></li>
215 <li><a href="#int_stacksave">'<tt>llvm.stacksave</tt>' Intrinsic</a></li>
216 <li><a href="#int_stackrestore">'<tt>llvm.stackrestore</tt>' Intrinsic</a></li>
217 <li><a href="#int_prefetch">'<tt>llvm.prefetch</tt>' Intrinsic</a></li>
218 <li><a href="#int_pcmarker">'<tt>llvm.pcmarker</tt>' Intrinsic</a></li>
219 <li><a href="#int_readcyclecounter"><tt>llvm.readcyclecounter</tt>' Intrinsic</a></li>
222 <li><a href="#int_libc">Standard C Library Intrinsics</a>
224 <li><a href="#int_memcpy">'<tt>llvm.memcpy.*</tt>' Intrinsic</a></li>
225 <li><a href="#int_memmove">'<tt>llvm.memmove.*</tt>' Intrinsic</a></li>
226 <li><a href="#int_memset">'<tt>llvm.memset.*</tt>' Intrinsic</a></li>
227 <li><a href="#int_sqrt">'<tt>llvm.sqrt.*</tt>' Intrinsic</a></li>
228 <li><a href="#int_powi">'<tt>llvm.powi.*</tt>' Intrinsic</a></li>
229 <li><a href="#int_sin">'<tt>llvm.sin.*</tt>' Intrinsic</a></li>
230 <li><a href="#int_cos">'<tt>llvm.cos.*</tt>' Intrinsic</a></li>
231 <li><a href="#int_pow">'<tt>llvm.pow.*</tt>' Intrinsic</a></li>
234 <li><a href="#int_manip">Bit Manipulation Intrinsics</a>
236 <li><a href="#int_bswap">'<tt>llvm.bswap.*</tt>' Intrinsics</a></li>
237 <li><a href="#int_ctpop">'<tt>llvm.ctpop.*</tt>' Intrinsic </a></li>
238 <li><a href="#int_ctlz">'<tt>llvm.ctlz.*</tt>' Intrinsic </a></li>
239 <li><a href="#int_cttz">'<tt>llvm.cttz.*</tt>' Intrinsic </a></li>
242 <li><a href="#int_overflow">Arithmetic with Overflow Intrinsics</a>
244 <li><a href="#int_sadd_overflow">'<tt>llvm.sadd.with.overflow.*</tt> Intrinsics</a></li>
245 <li><a href="#int_uadd_overflow">'<tt>llvm.uadd.with.overflow.*</tt> Intrinsics</a></li>
246 <li><a href="#int_ssub_overflow">'<tt>llvm.ssub.with.overflow.*</tt> Intrinsics</a></li>
247 <li><a href="#int_usub_overflow">'<tt>llvm.usub.with.overflow.*</tt> Intrinsics</a></li>
248 <li><a href="#int_smul_overflow">'<tt>llvm.smul.with.overflow.*</tt> Intrinsics</a></li>
249 <li><a href="#int_umul_overflow">'<tt>llvm.umul.with.overflow.*</tt> Intrinsics</a></li>
252 <li><a href="#int_debugger">Debugger intrinsics</a></li>
253 <li><a href="#int_eh">Exception Handling intrinsics</a></li>
254 <li><a href="#int_trampoline">Trampoline Intrinsic</a>
256 <li><a href="#int_it">'<tt>llvm.init.trampoline</tt>' Intrinsic</a></li>
259 <li><a href="#int_atomics">Atomic intrinsics</a>
261 <li><a href="#int_memory_barrier"><tt>llvm.memory_barrier</tt></a></li>
262 <li><a href="#int_atomic_cmp_swap"><tt>llvm.atomic.cmp.swap</tt></a></li>
263 <li><a href="#int_atomic_swap"><tt>llvm.atomic.swap</tt></a></li>
264 <li><a href="#int_atomic_load_add"><tt>llvm.atomic.load.add</tt></a></li>
265 <li><a href="#int_atomic_load_sub"><tt>llvm.atomic.load.sub</tt></a></li>
266 <li><a href="#int_atomic_load_and"><tt>llvm.atomic.load.and</tt></a></li>
267 <li><a href="#int_atomic_load_nand"><tt>llvm.atomic.load.nand</tt></a></li>
268 <li><a href="#int_atomic_load_or"><tt>llvm.atomic.load.or</tt></a></li>
269 <li><a href="#int_atomic_load_xor"><tt>llvm.atomic.load.xor</tt></a></li>
270 <li><a href="#int_atomic_load_max"><tt>llvm.atomic.load.max</tt></a></li>
271 <li><a href="#int_atomic_load_min"><tt>llvm.atomic.load.min</tt></a></li>
272 <li><a href="#int_atomic_load_umax"><tt>llvm.atomic.load.umax</tt></a></li>
273 <li><a href="#int_atomic_load_umin"><tt>llvm.atomic.load.umin</tt></a></li>
276 <li><a href="#int_general">General intrinsics</a>
278 <li><a href="#int_var_annotation">
279 '<tt>llvm.var.annotation</tt>' Intrinsic</a></li>
280 <li><a href="#int_annotation">
281 '<tt>llvm.annotation.*</tt>' Intrinsic</a></li>
282 <li><a href="#int_trap">
283 '<tt>llvm.trap</tt>' Intrinsic</a></li>
284 <li><a href="#int_stackprotector">
285 '<tt>llvm.stackprotector</tt>' Intrinsic</a></li>
292 <div class="doc_author">
293 <p>Written by <a href="mailto:sabre@nondot.org">Chris Lattner</a>
294 and <a href="mailto:vadve@cs.uiuc.edu">Vikram Adve</a></p>
297 <!-- *********************************************************************** -->
298 <div class="doc_section"> <a name="abstract">Abstract </a></div>
299 <!-- *********************************************************************** -->
301 <div class="doc_text">
303 <p>This document is a reference manual for the LLVM assembly language. LLVM is
304 a Static Single Assignment (SSA) based representation that provides type
305 safety, low-level operations, flexibility, and the capability of representing
306 'all' high-level languages cleanly. It is the common code representation
307 used throughout all phases of the LLVM compilation strategy.</p>
311 <!-- *********************************************************************** -->
312 <div class="doc_section"> <a name="introduction">Introduction</a> </div>
313 <!-- *********************************************************************** -->
315 <div class="doc_text">
317 <p>The LLVM code representation is designed to be used in three different forms:
318 as an in-memory compiler IR, as an on-disk bitcode representation (suitable
319 for fast loading by a Just-In-Time compiler), and as a human readable
320 assembly language representation. This allows LLVM to provide a powerful
321 intermediate representation for efficient compiler transformations and
322 analysis, while providing a natural means to debug and visualize the
323 transformations. The three different forms of LLVM are all equivalent. This
324 document describes the human readable representation and notation.</p>
326 <p>The LLVM representation aims to be light-weight and low-level while being
327 expressive, typed, and extensible at the same time. It aims to be a
328 "universal IR" of sorts, by being at a low enough level that high-level ideas
329 may be cleanly mapped to it (similar to how microprocessors are "universal
330 IR's", allowing many source languages to be mapped to them). By providing
331 type information, LLVM can be used as the target of optimizations: for
332 example, through pointer analysis, it can be proven that a C automatic
333 variable is never accessed outside of the current function... allowing it to
334 be promoted to a simple SSA value instead of a memory location.</p>
338 <!-- _______________________________________________________________________ -->
339 <div class="doc_subsubsection"> <a name="wellformed">Well-Formedness</a> </div>
341 <div class="doc_text">
343 <p>It is important to note that this document describes 'well formed' LLVM
344 assembly language. There is a difference between what the parser accepts and
345 what is considered 'well formed'. For example, the following instruction is
346 syntactically okay, but not well formed:</p>
348 <div class="doc_code">
350 %x = <a href="#i_add">add</a> i32 1, %x
354 <p>...because the definition of <tt>%x</tt> does not dominate all of its
355 uses. The LLVM infrastructure provides a verification pass that may be used
356 to verify that an LLVM module is well formed. This pass is automatically run
357 by the parser after parsing input assembly and by the optimizer before it
358 outputs bitcode. The violations pointed out by the verifier pass indicate
359 bugs in transformation passes or input to the parser.</p>
363 <!-- Describe the typesetting conventions here. -->
365 <!-- *********************************************************************** -->
366 <div class="doc_section"> <a name="identifiers">Identifiers</a> </div>
367 <!-- *********************************************************************** -->
369 <div class="doc_text">
371 <p>LLVM identifiers come in two basic types: global and local. Global
372 identifiers (functions, global variables) begin with the <tt>'@'</tt>
373 character. Local identifiers (register names, types) begin with
374 the <tt>'%'</tt> character. Additionally, there are three different formats
375 for identifiers, for different purposes:</p>
378 <li>Named values are represented as a string of characters with their prefix.
379 For example, <tt>%foo</tt>, <tt>@DivisionByZero</tt>,
380 <tt>%a.really.long.identifier</tt>. The actual regular expression used is
381 '<tt>[%@][a-zA-Z$._][a-zA-Z$._0-9]*</tt>'. Identifiers which require
382 other characters in their names can be surrounded with quotes. Special
383 characters may be escaped using <tt>"\xx"</tt> where <tt>xx</tt> is the
384 ASCII code for the character in hexadecimal. In this way, any character
385 can be used in a name value, even quotes themselves.</li>
387 <li>Unnamed values are represented as an unsigned numeric value with their
388 prefix. For example, <tt>%12</tt>, <tt>@2</tt>, <tt>%44</tt>.</li>
390 <li>Constants, which are described in a <a href="#constants">section about
391 constants</a>, below.</li>
394 <p>LLVM requires that values start with a prefix for two reasons: Compilers
395 don't need to worry about name clashes with reserved words, and the set of
396 reserved words may be expanded in the future without penalty. Additionally,
397 unnamed identifiers allow a compiler to quickly come up with a temporary
398 variable without having to avoid symbol table conflicts.</p>
400 <p>Reserved words in LLVM are very similar to reserved words in other
401 languages. There are keywords for different opcodes
402 ('<tt><a href="#i_add">add</a></tt>',
403 '<tt><a href="#i_bitcast">bitcast</a></tt>',
404 '<tt><a href="#i_ret">ret</a></tt>', etc...), for primitive type names
405 ('<tt><a href="#t_void">void</a></tt>',
406 '<tt><a href="#t_primitive">i32</a></tt>', etc...), and others. These
407 reserved words cannot conflict with variable names, because none of them
408 start with a prefix character (<tt>'%'</tt> or <tt>'@'</tt>).</p>
410 <p>Here is an example of LLVM code to multiply the integer variable
411 '<tt>%X</tt>' by 8:</p>
415 <div class="doc_code">
417 %result = <a href="#i_mul">mul</a> i32 %X, 8
421 <p>After strength reduction:</p>
423 <div class="doc_code">
425 %result = <a href="#i_shl">shl</a> i32 %X, i8 3
429 <p>And the hard way:</p>
431 <div class="doc_code">
433 <a href="#i_add">add</a> i32 %X, %X <i>; yields {i32}:%0</i>
434 <a href="#i_add">add</a> i32 %0, %0 <i>; yields {i32}:%1</i>
435 %result = <a href="#i_add">add</a> i32 %1, %1
439 <p>This last way of multiplying <tt>%X</tt> by 8 illustrates several important
440 lexical features of LLVM:</p>
443 <li>Comments are delimited with a '<tt>;</tt>' and go until the end of
446 <li>Unnamed temporaries are created when the result of a computation is not
447 assigned to a named value.</li>
449 <li>Unnamed temporaries are numbered sequentially</li>
452 <p>...and it also shows a convention that we follow in this document. When
453 demonstrating instructions, we will follow an instruction with a comment that
454 defines the type and name of value produced. Comments are shown in italic
459 <!-- *********************************************************************** -->
460 <div class="doc_section"> <a name="highlevel">High Level Structure</a> </div>
461 <!-- *********************************************************************** -->
463 <!-- ======================================================================= -->
464 <div class="doc_subsection"> <a name="modulestructure">Module Structure</a>
467 <div class="doc_text">
469 <p>LLVM programs are composed of "Module"s, each of which is a translation unit
470 of the input programs. Each module consists of functions, global variables,
471 and symbol table entries. Modules may be combined together with the LLVM
472 linker, which merges function (and global variable) definitions, resolves
473 forward declarations, and merges symbol table entries. Here is an example of
474 the "hello world" module:</p>
476 <div class="doc_code">
477 <pre><i>; Declare the string constant as a global constant...</i>
478 <a href="#identifiers">@.LC0</a> = <a href="#linkage_internal">internal</a> <a
479 href="#globalvars">constant</a> <a href="#t_array">[13 x i8]</a> c"hello world\0A\00" <i>; [13 x i8]*</i>
481 <i>; External declaration of the puts function</i>
482 <a href="#functionstructure">declare</a> i32 @puts(i8 *) <i>; i32(i8 *)* </i>
484 <i>; Definition of main function</i>
485 define i32 @main() { <i>; i32()* </i>
486 <i>; Convert [13 x i8]* to i8 *...</i>
488 href="#i_getelementptr">getelementptr</a> [13 x i8]* @.LC0, i64 0, i64 0 <i>; i8 *</i>
490 <i>; Call puts function to write out the string to stdout...</i>
492 href="#i_call">call</a> i32 @puts(i8 * %cast210) <i>; i32</i>
494 href="#i_ret">ret</a> i32 0<br>}<br>
498 <p>This example is made up of a <a href="#globalvars">global variable</a> named
499 "<tt>.LC0</tt>", an external declaration of the "<tt>puts</tt>" function, and
500 a <a href="#functionstructure">function definition</a> for
503 <p>In general, a module is made up of a list of global values, where both
504 functions and global variables are global values. Global values are
505 represented by a pointer to a memory location (in this case, a pointer to an
506 array of char, and a pointer to a function), and have one of the
507 following <a href="#linkage">linkage types</a>.</p>
511 <!-- ======================================================================= -->
512 <div class="doc_subsection">
513 <a name="linkage">Linkage Types</a>
516 <div class="doc_text">
518 <p>All Global Variables and Functions have one of the following types of
522 <dt><tt><b><a name="linkage_private">private</a></b></tt>: </dt>
523 <dd>Global values with private linkage are only directly accessible by objects
524 in the current module. In particular, linking code into a module with an
525 private global value may cause the private to be renamed as necessary to
526 avoid collisions. Because the symbol is private to the module, all
527 references can be updated. This doesn't show up in any symbol table in the
530 <dt><tt><b><a name="linkage_linker_private">linker_private</a></b></tt>: </dt>
531 <dd>Similar to private, but the symbol is passed through the assembler and
532 removed by the linker after evaluation.</dd>
534 <dt><tt><b><a name="linkage_internal">internal</a></b></tt>: </dt>
535 <dd>Similar to private, but the value shows as a local symbol
536 (<tt>STB_LOCAL</tt> in the case of ELF) in the object file. This
537 corresponds to the notion of the '<tt>static</tt>' keyword in C.</dd>
539 <dt><tt><b><a name="linkage_available_externally">available_externally</a></b></tt>: </dt>
540 <dd>Globals with "<tt>available_externally</tt>" linkage are never emitted
541 into the object file corresponding to the LLVM module. They exist to
542 allow inlining and other optimizations to take place given knowledge of
543 the definition of the global, which is known to be somewhere outside the
544 module. Globals with <tt>available_externally</tt> linkage are allowed to
545 be discarded at will, and are otherwise the same as <tt>linkonce_odr</tt>.
546 This linkage type is only allowed on definitions, not declarations.</dd>
548 <dt><tt><b><a name="linkage_linkonce">linkonce</a></b></tt>: </dt>
549 <dd>Globals with "<tt>linkonce</tt>" linkage are merged with other globals of
550 the same name when linkage occurs. This is typically used to implement
551 inline functions, templates, or other code which must be generated in each
552 translation unit that uses it. Unreferenced <tt>linkonce</tt> globals are
553 allowed to be discarded.</dd>
555 <dt><tt><b><a name="linkage_weak">weak</a></b></tt>: </dt>
556 <dd>"<tt>weak</tt>" linkage has the same merging semantics as
557 <tt>linkonce</tt> linkage, except that unreferenced globals with
558 <tt>weak</tt> linkage may not be discarded. This is used for globals that
559 are declared "weak" in C source code.</dd>
561 <dt><tt><b><a name="linkage_common">common</a></b></tt>: </dt>
562 <dd>"<tt>common</tt>" linkage is most similar to "<tt>weak</tt>" linkage, but
563 they are used for tentative definitions in C, such as "<tt>int X;</tt>" at
565 Symbols with "<tt>common</tt>" linkage are merged in the same way as
566 <tt>weak symbols</tt>, and they may not be deleted if unreferenced.
567 <tt>common</tt> symbols may not have an explicit section,
568 must have a zero initializer, and may not be marked '<a
569 href="#globalvars"><tt>constant</tt></a>'. Functions and aliases may not
570 have common linkage.</dd>
573 <dt><tt><b><a name="linkage_appending">appending</a></b></tt>: </dt>
574 <dd>"<tt>appending</tt>" linkage may only be applied to global variables of
575 pointer to array type. When two global variables with appending linkage
576 are linked together, the two global arrays are appended together. This is
577 the LLVM, typesafe, equivalent of having the system linker append together
578 "sections" with identical names when .o files are linked.</dd>
580 <dt><tt><b><a name="linkage_externweak">extern_weak</a></b></tt>: </dt>
581 <dd>The semantics of this linkage follow the ELF object file model: the symbol
582 is weak until linked, if not linked, the symbol becomes null instead of
583 being an undefined reference.</dd>
585 <dt><tt><b><a name="linkage_linkonce">linkonce_odr</a></b></tt>: </dt>
586 <dt><tt><b><a name="linkage_weak">weak_odr</a></b></tt>: </dt>
587 <dd>Some languages allow differing globals to be merged, such as two functions
588 with different semantics. Other languages, such as <tt>C++</tt>, ensure
589 that only equivalent globals are ever merged (the "one definition rule" -
590 "ODR"). Such languages can use the <tt>linkonce_odr</tt>
591 and <tt>weak_odr</tt> linkage types to indicate that the global will only
592 be merged with equivalent globals. These linkage types are otherwise the
593 same as their non-<tt>odr</tt> versions.</dd>
595 <dt><tt><b><a name="linkage_external">externally visible</a></b></tt>:</dt>
596 <dd>If none of the above identifiers are used, the global is externally
597 visible, meaning that it participates in linkage and can be used to
598 resolve external symbol references.</dd>
601 <p>The next two types of linkage are targeted for Microsoft Windows platform
602 only. They are designed to support importing (exporting) symbols from (to)
603 DLLs (Dynamic Link Libraries).</p>
606 <dt><tt><b><a name="linkage_dllimport">dllimport</a></b></tt>: </dt>
607 <dd>"<tt>dllimport</tt>" linkage causes the compiler to reference a function
608 or variable via a global pointer to a pointer that is set up by the DLL
609 exporting the symbol. On Microsoft Windows targets, the pointer name is
610 formed by combining <code>__imp_</code> and the function or variable
613 <dt><tt><b><a name="linkage_dllexport">dllexport</a></b></tt>: </dt>
614 <dd>"<tt>dllexport</tt>" linkage causes the compiler to provide a global
615 pointer to a pointer in a DLL, so that it can be referenced with the
616 <tt>dllimport</tt> attribute. On Microsoft Windows targets, the pointer
617 name is formed by combining <code>__imp_</code> and the function or
621 <p>For example, since the "<tt>.LC0</tt>" variable is defined to be internal, if
622 another module defined a "<tt>.LC0</tt>" variable and was linked with this
623 one, one of the two would be renamed, preventing a collision. Since
624 "<tt>main</tt>" and "<tt>puts</tt>" are external (i.e., lacking any linkage
625 declarations), they are accessible outside of the current module.</p>
627 <p>It is illegal for a function <i>declaration</i> to have any linkage type
628 other than "externally visible", <tt>dllimport</tt>
629 or <tt>extern_weak</tt>.</p>
631 <p>Aliases can have only <tt>external</tt>, <tt>internal</tt>, <tt>weak</tt>
632 or <tt>weak_odr</tt> linkages.</p>
636 <!-- ======================================================================= -->
637 <div class="doc_subsection">
638 <a name="callingconv">Calling Conventions</a>
641 <div class="doc_text">
643 <p>LLVM <a href="#functionstructure">functions</a>, <a href="#i_call">calls</a>
644 and <a href="#i_invoke">invokes</a> can all have an optional calling
645 convention specified for the call. The calling convention of any pair of
646 dynamic caller/callee must match, or the behavior of the program is
647 undefined. The following calling conventions are supported by LLVM, and more
648 may be added in the future:</p>
651 <dt><b>"<tt>ccc</tt>" - The C calling convention</b>:</dt>
652 <dd>This calling convention (the default if no other calling convention is
653 specified) matches the target C calling conventions. This calling
654 convention supports varargs function calls and tolerates some mismatch in
655 the declared prototype and implemented declaration of the function (as
658 <dt><b>"<tt>fastcc</tt>" - The fast calling convention</b>:</dt>
659 <dd>This calling convention attempts to make calls as fast as possible
660 (e.g. by passing things in registers). This calling convention allows the
661 target to use whatever tricks it wants to produce fast code for the
662 target, without having to conform to an externally specified ABI
663 (Application Binary Interface). Implementations of this convention should
664 allow arbitrary <a href="CodeGenerator.html#tailcallopt">tail call
665 optimization</a> to be supported. This calling convention does not
666 support varargs and requires the prototype of all callees to exactly match
667 the prototype of the function definition.</dd>
669 <dt><b>"<tt>coldcc</tt>" - The cold calling convention</b>:</dt>
670 <dd>This calling convention attempts to make code in the caller as efficient
671 as possible under the assumption that the call is not commonly executed.
672 As such, these calls often preserve all registers so that the call does
673 not break any live ranges in the caller side. This calling convention
674 does not support varargs and requires the prototype of all callees to
675 exactly match the prototype of the function definition.</dd>
677 <dt><b>"<tt>cc <<em>n</em>></tt>" - Numbered convention</b>:</dt>
678 <dd>Any calling convention may be specified by number, allowing
679 target-specific calling conventions to be used. Target specific calling
680 conventions start at 64.</dd>
683 <p>More calling conventions can be added/defined on an as-needed basis, to
684 support Pascal conventions or any other well-known target-independent
689 <!-- ======================================================================= -->
690 <div class="doc_subsection">
691 <a name="visibility">Visibility Styles</a>
694 <div class="doc_text">
696 <p>All Global Variables and Functions have one of the following visibility
700 <dt><b>"<tt>default</tt>" - Default style</b>:</dt>
701 <dd>On targets that use the ELF object file format, default visibility means
702 that the declaration is visible to other modules and, in shared libraries,
703 means that the declared entity may be overridden. On Darwin, default
704 visibility means that the declaration is visible to other modules. Default
705 visibility corresponds to "external linkage" in the language.</dd>
707 <dt><b>"<tt>hidden</tt>" - Hidden style</b>:</dt>
708 <dd>Two declarations of an object with hidden visibility refer to the same
709 object if they are in the same shared object. Usually, hidden visibility
710 indicates that the symbol will not be placed into the dynamic symbol
711 table, so no other module (executable or shared library) can reference it
714 <dt><b>"<tt>protected</tt>" - Protected style</b>:</dt>
715 <dd>On ELF, protected visibility indicates that the symbol will be placed in
716 the dynamic symbol table, but that references within the defining module
717 will bind to the local symbol. That is, the symbol cannot be overridden by
723 <!-- ======================================================================= -->
724 <div class="doc_subsection">
725 <a name="namedtypes">Named Types</a>
728 <div class="doc_text">
730 <p>LLVM IR allows you to specify name aliases for certain types. This can make
731 it easier to read the IR and make the IR more condensed (particularly when
732 recursive types are involved). An example of a name specification is:</p>
734 <div class="doc_code">
736 %mytype = type { %mytype*, i32 }
740 <p>You may give a name to any <a href="#typesystem">type</a> except
741 "<a href="t_void">void</a>". Type name aliases may be used anywhere a type
742 is expected with the syntax "%mytype".</p>
744 <p>Note that type names are aliases for the structural type that they indicate,
745 and that you can therefore specify multiple names for the same type. This
746 often leads to confusing behavior when dumping out a .ll file. Since LLVM IR
747 uses structural typing, the name is not part of the type. When printing out
748 LLVM IR, the printer will pick <em>one name</em> to render all types of a
749 particular shape. This means that if you have code where two different
750 source types end up having the same LLVM type, that the dumper will sometimes
751 print the "wrong" or unexpected type. This is an important design point and
752 isn't going to change.</p>
756 <!-- ======================================================================= -->
757 <div class="doc_subsection">
758 <a name="globalvars">Global Variables</a>
761 <div class="doc_text">
763 <p>Global variables define regions of memory allocated at compilation time
764 instead of run-time. Global variables may optionally be initialized, may
765 have an explicit section to be placed in, and may have an optional explicit
766 alignment specified. A variable may be defined as "thread_local", which
767 means that it will not be shared by threads (each thread will have a
768 separated copy of the variable). A variable may be defined as a global
769 "constant," which indicates that the contents of the variable
770 will <b>never</b> be modified (enabling better optimization, allowing the
771 global data to be placed in the read-only section of an executable, etc).
772 Note that variables that need runtime initialization cannot be marked
773 "constant" as there is a store to the variable.</p>
775 <p>LLVM explicitly allows <em>declarations</em> of global variables to be marked
776 constant, even if the final definition of the global is not. This capability
777 can be used to enable slightly better optimization of the program, but
778 requires the language definition to guarantee that optimizations based on the
779 'constantness' are valid for the translation units that do not include the
782 <p>As SSA values, global variables define pointer values that are in scope
783 (i.e. they dominate) all basic blocks in the program. Global variables
784 always define a pointer to their "content" type because they describe a
785 region of memory, and all memory objects in LLVM are accessed through
788 <p>A global variable may be declared to reside in a target-specific numbered
789 address space. For targets that support them, address spaces may affect how
790 optimizations are performed and/or what target instructions are used to
791 access the variable. The default address space is zero. The address space
792 qualifier must precede any other attributes.</p>
794 <p>LLVM allows an explicit section to be specified for globals. If the target
795 supports it, it will emit globals to the section specified.</p>
797 <p>An explicit alignment may be specified for a global. If not present, or if
798 the alignment is set to zero, the alignment of the global is set by the
799 target to whatever it feels convenient. If an explicit alignment is
800 specified, the global is forced to have at least that much alignment. All
801 alignments must be a power of 2.</p>
803 <p>For example, the following defines a global in a numbered address space with
804 an initializer, section, and alignment:</p>
806 <div class="doc_code">
808 @G = addrspace(5) constant float 1.0, section "foo", align 4
815 <!-- ======================================================================= -->
816 <div class="doc_subsection">
817 <a name="functionstructure">Functions</a>
820 <div class="doc_text">
822 <p>LLVM function definitions consist of the "<tt>define</tt>" keyord, an
823 optional <a href="#linkage">linkage type</a>, an optional
824 <a href="#visibility">visibility style</a>, an optional
825 <a href="#callingconv">calling convention</a>, a return type, an optional
826 <a href="#paramattrs">parameter attribute</a> for the return type, a function
827 name, a (possibly empty) argument list (each with optional
828 <a href="#paramattrs">parameter attributes</a>), optional
829 <a href="#fnattrs">function attributes</a>, an optional section, an optional
830 alignment, an optional <a href="#gc">garbage collector name</a>, an opening
831 curly brace, a list of basic blocks, and a closing curly brace.</p>
833 <p>LLVM function declarations consist of the "<tt>declare</tt>" keyword, an
834 optional <a href="#linkage">linkage type</a>, an optional
835 <a href="#visibility">visibility style</a>, an optional
836 <a href="#callingconv">calling convention</a>, a return type, an optional
837 <a href="#paramattrs">parameter attribute</a> for the return type, a function
838 name, a possibly empty list of arguments, an optional alignment, and an
839 optional <a href="#gc">garbage collector name</a>.</p>
841 <p>A function definition contains a list of basic blocks, forming the CFG
842 (Control Flow Graph) for the function. Each basic block may optionally start
843 with a label (giving the basic block a symbol table entry), contains a list
844 of instructions, and ends with a <a href="#terminators">terminator</a>
845 instruction (such as a branch or function return).</p>
847 <p>The first basic block in a function is special in two ways: it is immediately
848 executed on entrance to the function, and it is not allowed to have
849 predecessor basic blocks (i.e. there can not be any branches to the entry
850 block of a function). Because the block can have no predecessors, it also
851 cannot have any <a href="#i_phi">PHI nodes</a>.</p>
853 <p>LLVM allows an explicit section to be specified for functions. If the target
854 supports it, it will emit functions to the section specified.</p>
856 <p>An explicit alignment may be specified for a function. If not present, or if
857 the alignment is set to zero, the alignment of the function is set by the
858 target to whatever it feels convenient. If an explicit alignment is
859 specified, the function is forced to have at least that much alignment. All
860 alignments must be a power of 2.</p>
863 <div class="doc_code">
865 define [<a href="#linkage">linkage</a>] [<a href="#visibility">visibility</a>]
866 [<a href="#callingconv">cconv</a>] [<a href="#paramattrs">ret attrs</a>]
867 <ResultType> @<FunctionName> ([argument list])
868 [<a href="#fnattrs">fn Attrs</a>] [section "name"] [align N]
869 [<a href="#gc">gc</a>] { ... }
875 <!-- ======================================================================= -->
876 <div class="doc_subsection">
877 <a name="aliasstructure">Aliases</a>
880 <div class="doc_text">
882 <p>Aliases act as "second name" for the aliasee value (which can be either
883 function, global variable, another alias or bitcast of global value). Aliases
884 may have an optional <a href="#linkage">linkage type</a>, and an
885 optional <a href="#visibility">visibility style</a>.</p>
888 <div class="doc_code">
890 @<Name> = alias [Linkage] [Visibility] <AliaseeTy> @<Aliasee>
896 <!-- ======================================================================= -->
897 <div class="doc_subsection"><a name="paramattrs">Parameter Attributes</a></div>
899 <div class="doc_text">
901 <p>The return type and each parameter of a function type may have a set of
902 <i>parameter attributes</i> associated with them. Parameter attributes are
903 used to communicate additional information about the result or parameters of
904 a function. Parameter attributes are considered to be part of the function,
905 not of the function type, so functions with different parameter attributes
906 can have the same function type.</p>
908 <p>Parameter attributes are simple keywords that follow the type specified. If
909 multiple parameter attributes are needed, they are space separated. For
912 <div class="doc_code">
914 declare i32 @printf(i8* noalias nocapture, ...)
915 declare i32 @atoi(i8 zeroext)
916 declare signext i8 @returns_signed_char()
920 <p>Note that any attributes for the function result (<tt>nounwind</tt>,
921 <tt>readonly</tt>) come immediately after the argument list.</p>
923 <p>Currently, only the following parameter attributes are defined:</p>
926 <dt><tt>zeroext</tt></dt>
927 <dd>This indicates to the code generator that the parameter or return value
928 should be zero-extended to a 32-bit value by the caller (for a parameter)
929 or the callee (for a return value).</dd>
931 <dt><tt>signext</tt></dt>
932 <dd>This indicates to the code generator that the parameter or return value
933 should be sign-extended to a 32-bit value by the caller (for a parameter)
934 or the callee (for a return value).</dd>
936 <dt><tt>inreg</tt></dt>
937 <dd>This indicates that this parameter or return value should be treated in a
938 special target-dependent fashion during while emitting code for a function
939 call or return (usually, by putting it in a register as opposed to memory,
940 though some targets use it to distinguish between two different kinds of
941 registers). Use of this attribute is target-specific.</dd>
943 <dt><tt><a name="byval">byval</a></tt></dt>
944 <dd>This indicates that the pointer parameter should really be passed by value
945 to the function. The attribute implies that a hidden copy of the pointee
946 is made between the caller and the callee, so the callee is unable to
947 modify the value in the callee. This attribute is only valid on LLVM
948 pointer arguments. It is generally used to pass structs and arrays by
949 value, but is also valid on pointers to scalars. The copy is considered
950 to belong to the caller not the callee (for example,
951 <tt><a href="#readonly">readonly</a></tt> functions should not write to
952 <tt>byval</tt> parameters). This is not a valid attribute for return
953 values. The byval attribute also supports specifying an alignment with
954 the align attribute. This has a target-specific effect on the code
955 generator that usually indicates a desired alignment for the synthesized
958 <dt><tt>sret</tt></dt>
959 <dd>This indicates that the pointer parameter specifies the address of a
960 structure that is the return value of the function in the source program.
961 This pointer must be guaranteed by the caller to be valid: loads and
962 stores to the structure may be assumed by the callee to not to trap. This
963 may only be applied to the first parameter. This is not a valid attribute
964 for return values. </dd>
966 <dt><tt>noalias</tt></dt>
967 <dd>This indicates that the pointer does not alias any global or any other
968 parameter. The caller is responsible for ensuring that this is the
969 case. On a function return value, <tt>noalias</tt> additionally indicates
970 that the pointer does not alias any other pointers visible to the
971 caller. For further details, please see the discussion of the NoAlias
973 <a href="http://llvm.org/docs/AliasAnalysis.html#MustMayNo">alias
976 <dt><tt>nocapture</tt></dt>
977 <dd>This indicates that the callee does not make any copies of the pointer
978 that outlive the callee itself. This is not a valid attribute for return
981 <dt><tt>nest</tt></dt>
982 <dd>This indicates that the pointer parameter can be excised using the
983 <a href="#int_trampoline">trampoline intrinsics</a>. This is not a valid
984 attribute for return values.</dd>
989 <!-- ======================================================================= -->
990 <div class="doc_subsection">
991 <a name="gc">Garbage Collector Names</a>
994 <div class="doc_text">
996 <p>Each function may specify a garbage collector name, which is simply a
999 <div class="doc_code">
1001 define void @f() gc "name" { ...
1005 <p>The compiler declares the supported values of <i>name</i>. Specifying a
1006 collector which will cause the compiler to alter its output in order to
1007 support the named garbage collection algorithm.</p>
1011 <!-- ======================================================================= -->
1012 <div class="doc_subsection">
1013 <a name="fnattrs">Function Attributes</a>
1016 <div class="doc_text">
1018 <p>Function attributes are set to communicate additional information about a
1019 function. Function attributes are considered to be part of the function, not
1020 of the function type, so functions with different parameter attributes can
1021 have the same function type.</p>
1023 <p>Function attributes are simple keywords that follow the type specified. If
1024 multiple attributes are needed, they are space separated. For example:</p>
1026 <div class="doc_code">
1028 define void @f() noinline { ... }
1029 define void @f() alwaysinline { ... }
1030 define void @f() alwaysinline optsize { ... }
1031 define void @f() optsize
1036 <dt><tt>alwaysinline</tt></dt>
1037 <dd>This attribute indicates that the inliner should attempt to inline this
1038 function into callers whenever possible, ignoring any active inlining size
1039 threshold for this caller.</dd>
1041 <dt><tt>noinline</tt></dt>
1042 <dd>This attribute indicates that the inliner should never inline this
1043 function in any situation. This attribute may not be used together with
1044 the <tt>alwaysinline</tt> attribute.</dd>
1046 <dt><tt>optsize</tt></dt>
1047 <dd>This attribute suggests that optimization passes and code generator passes
1048 make choices that keep the code size of this function low, and otherwise
1049 do optimizations specifically to reduce code size.</dd>
1051 <dt><tt>noreturn</tt></dt>
1052 <dd>This function attribute indicates that the function never returns
1053 normally. This produces undefined behavior at runtime if the function
1054 ever does dynamically return.</dd>
1056 <dt><tt>nounwind</tt></dt>
1057 <dd>This function attribute indicates that the function never returns with an
1058 unwind or exceptional control flow. If the function does unwind, its
1059 runtime behavior is undefined.</dd>
1061 <dt><tt>readnone</tt></dt>
1062 <dd>This attribute indicates that the function computes its result (or decides
1063 to unwind an exception) based strictly on its arguments, without
1064 dereferencing any pointer arguments or otherwise accessing any mutable
1065 state (e.g. memory, control registers, etc) visible to caller functions.
1066 It does not write through any pointer arguments
1067 (including <tt><a href="#byval">byval</a></tt> arguments) and never
1068 changes any state visible to callers. This means that it cannot unwind
1069 exceptions by calling the <tt>C++</tt> exception throwing methods, but
1070 could use the <tt>unwind</tt> instruction.</dd>
1072 <dt><tt><a name="readonly">readonly</a></tt></dt>
1073 <dd>This attribute indicates that the function does not write through any
1074 pointer arguments (including <tt><a href="#byval">byval</a></tt>
1075 arguments) or otherwise modify any state (e.g. memory, control registers,
1076 etc) visible to caller functions. It may dereference pointer arguments
1077 and read state that may be set in the caller. A readonly function always
1078 returns the same value (or unwinds an exception identically) when called
1079 with the same set of arguments and global state. It cannot unwind an
1080 exception by calling the <tt>C++</tt> exception throwing methods, but may
1081 use the <tt>unwind</tt> instruction.</dd>
1083 <dt><tt><a name="ssp">ssp</a></tt></dt>
1084 <dd>This attribute indicates that the function should emit a stack smashing
1085 protector. It is in the form of a "canary"—a random value placed on
1086 the stack before the local variables that's checked upon return from the
1087 function to see if it has been overwritten. A heuristic is used to
1088 determine if a function needs stack protectors or not.<br>
1090 If a function that has an <tt>ssp</tt> attribute is inlined into a
1091 function that doesn't have an <tt>ssp</tt> attribute, then the resulting
1092 function will have an <tt>ssp</tt> attribute.</dd>
1094 <dt><tt>sspreq</tt></dt>
1095 <dd>This attribute indicates that the function should <em>always</em> emit a
1096 stack smashing protector. This overrides
1097 the <tt><a href="#ssp">ssp</a></tt> function attribute.<br>
1099 If a function that has an <tt>sspreq</tt> attribute is inlined into a
1100 function that doesn't have an <tt>sspreq</tt> attribute or which has
1101 an <tt>ssp</tt> attribute, then the resulting function will have
1102 an <tt>sspreq</tt> attribute.</dd>
1104 <dt><tt>noredzone</tt></dt>
1105 <dd>This attribute indicates that the code generator should not use a red
1106 zone, even if the target-specific ABI normally permits it.</dd>
1108 <dt><tt>noimplicitfloat</tt></dt>
1109 <dd>This attributes disables implicit floating point instructions.</dd>
1111 <dt><tt>naked</tt></dt>
1112 <dd>This attribute disables prologue / epilogue emission for the function.
1113 This can have very system-specific consequences.</dd>
1118 <!-- ======================================================================= -->
1119 <div class="doc_subsection">
1120 <a name="moduleasm">Module-Level Inline Assembly</a>
1123 <div class="doc_text">
1125 <p>Modules may contain "module-level inline asm" blocks, which corresponds to
1126 the GCC "file scope inline asm" blocks. These blocks are internally
1127 concatenated by LLVM and treated as a single unit, but may be separated in
1128 the <tt>.ll</tt> file if desired. The syntax is very simple:</p>
1130 <div class="doc_code">
1132 module asm "inline asm code goes here"
1133 module asm "more can go here"
1137 <p>The strings can contain any character by escaping non-printable characters.
1138 The escape sequence used is simply "\xx" where "xx" is the two digit hex code
1141 <p>The inline asm code is simply printed to the machine code .s file when
1142 assembly code is generated.</p>
1146 <!-- ======================================================================= -->
1147 <div class="doc_subsection">
1148 <a name="datalayout">Data Layout</a>
1151 <div class="doc_text">
1153 <p>A module may specify a target specific data layout string that specifies how
1154 data is to be laid out in memory. The syntax for the data layout is
1157 <div class="doc_code">
1159 target datalayout = "<i>layout specification</i>"
1163 <p>The <i>layout specification</i> consists of a list of specifications
1164 separated by the minus sign character ('-'). Each specification starts with
1165 a letter and may include other information after the letter to define some
1166 aspect of the data layout. The specifications accepted are as follows:</p>
1170 <dd>Specifies that the target lays out data in big-endian form. That is, the
1171 bits with the most significance have the lowest address location.</dd>
1174 <dd>Specifies that the target lays out data in little-endian form. That is,
1175 the bits with the least significance have the lowest address
1178 <dt><tt>p:<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1179 <dd>This specifies the <i>size</i> of a pointer and its <i>abi</i> and
1180 <i>preferred</i> alignments. All sizes are in bits. Specifying
1181 the <i>pref</i> alignment is optional. If omitted, the
1182 preceding <tt>:</tt> should be omitted too.</dd>
1184 <dt><tt>i<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1185 <dd>This specifies the alignment for an integer type of a given bit
1186 <i>size</i>. The value of <i>size</i> must be in the range [1,2^23).</dd>
1188 <dt><tt>v<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1189 <dd>This specifies the alignment for a vector type of a given bit
1192 <dt><tt>f<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1193 <dd>This specifies the alignment for a floating point type of a given bit
1194 <i>size</i>. The value of <i>size</i> must be either 32 (float) or 64
1197 <dt><tt>a<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1198 <dd>This specifies the alignment for an aggregate type of a given bit
1201 <dt><tt>s<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1202 <dd>This specifies the alignment for a stack object of a given bit
1206 <p>When constructing the data layout for a given target, LLVM starts with a
1207 default set of specifications which are then (possibly) overriden by the
1208 specifications in the <tt>datalayout</tt> keyword. The default specifications
1209 are given in this list:</p>
1212 <li><tt>E</tt> - big endian</li>
1213 <li><tt>p:32:64:64</tt> - 32-bit pointers with 64-bit alignment</li>
1214 <li><tt>i1:8:8</tt> - i1 is 8-bit (byte) aligned</li>
1215 <li><tt>i8:8:8</tt> - i8 is 8-bit (byte) aligned</li>
1216 <li><tt>i16:16:16</tt> - i16 is 16-bit aligned</li>
1217 <li><tt>i32:32:32</tt> - i32 is 32-bit aligned</li>
1218 <li><tt>i64:32:64</tt> - i64 has ABI alignment of 32-bits but preferred
1219 alignment of 64-bits</li>
1220 <li><tt>f32:32:32</tt> - float is 32-bit aligned</li>
1221 <li><tt>f64:64:64</tt> - double is 64-bit aligned</li>
1222 <li><tt>v64:64:64</tt> - 64-bit vector is 64-bit aligned</li>
1223 <li><tt>v128:128:128</tt> - 128-bit vector is 128-bit aligned</li>
1224 <li><tt>a0:0:1</tt> - aggregates are 8-bit aligned</li>
1225 <li><tt>s0:64:64</tt> - stack objects are 64-bit aligned</li>
1228 <p>When LLVM is determining the alignment for a given type, it uses the
1229 following rules:</p>
1232 <li>If the type sought is an exact match for one of the specifications, that
1233 specification is used.</li>
1235 <li>If no match is found, and the type sought is an integer type, then the
1236 smallest integer type that is larger than the bitwidth of the sought type
1237 is used. If none of the specifications are larger than the bitwidth then
1238 the the largest integer type is used. For example, given the default
1239 specifications above, the i7 type will use the alignment of i8 (next
1240 largest) while both i65 and i256 will use the alignment of i64 (largest
1243 <li>If no match is found, and the type sought is a vector type, then the
1244 largest vector type that is smaller than the sought vector type will be
1245 used as a fall back. This happens because <128 x double> can be
1246 implemented in terms of 64 <2 x double>, for example.</li>
1251 <!-- ======================================================================= -->
1252 <div class="doc_subsection">
1253 <a name="pointeraliasing">Pointer Aliasing Rules</a>
1256 <div class="doc_text">
1258 <p>Any memory access must be done through a pointer value associated
1259 with an address range of the memory access, otherwise the behavior
1260 is undefined. Pointer values are associated with address ranges
1261 according to the following rules:</p>
1264 <li>A pointer value formed from a
1265 <tt><a href="#i_getelementptr">getelementptr</a></tt> instruction
1266 is associated with the addresses associated with the first operand
1267 of the <tt>getelementptr</tt>.</li>
1268 <li>An address of a global variable is associated with the address
1269 range of the variable's storage.</li>
1270 <li>The result value of an allocation instruction is associated with
1271 the address range of the allocated storage.</li>
1272 <li>A null pointer in the default address-space is associated with
1274 <li>A pointer value formed by an
1275 <tt><a href="#i_inttoptr">inttoptr</a></tt> is associated with all
1276 address ranges of all pointer values that contribute (directly or
1277 indirectly) to the computation of the pointer's value.</li>
1278 <li>The result value of a
1279 <tt><a href="#i_bitcast">bitcast</a></tt> is associated with all
1280 addresses associated with the operand of the <tt>bitcast</tt>.</li>
1281 <li>An integer constant other than zero or a pointer value returned
1282 from a function not defined within LLVM may be associated with address
1283 ranges allocated through mechanisms other than those provided by
1284 LLVM. Such ranges shall not overlap with any ranges of addresses
1285 allocated by mechanisms provided by LLVM.</li>
1288 <p>LLVM IR does not associate types with memory. The result type of a
1289 <tt><a href="#i_load">load</a></tt> merely indicates the size and
1290 alignment of the memory from which to load, as well as the
1291 interpretation of the value. The first operand of a
1292 <tt><a href="#i_store">store</a></tt> similarly only indicates the size
1293 and alignment of the store.</p>
1295 <p>Consequently, type-based alias analysis, aka TBAA, aka
1296 <tt>-fstrict-aliasing</tt>, is not applicable to general unadorned
1297 LLVM IR. <a href="#metadata">Metadata</a> may be used to encode
1298 additional information which specialized optimization passes may use
1299 to implement type-based alias analysis.</p>
1303 <!-- *********************************************************************** -->
1304 <div class="doc_section"> <a name="typesystem">Type System</a> </div>
1305 <!-- *********************************************************************** -->
1307 <div class="doc_text">
1309 <p>The LLVM type system is one of the most important features of the
1310 intermediate representation. Being typed enables a number of optimizations
1311 to be performed on the intermediate representation directly, without having
1312 to do extra analyses on the side before the transformation. A strong type
1313 system makes it easier to read the generated code and enables novel analyses
1314 and transformations that are not feasible to perform on normal three address
1315 code representations.</p>
1319 <!-- ======================================================================= -->
1320 <div class="doc_subsection"> <a name="t_classifications">Type
1321 Classifications</a> </div>
1323 <div class="doc_text">
1325 <p>The types fall into a few useful classifications:</p>
1327 <table border="1" cellspacing="0" cellpadding="4">
1329 <tr><th>Classification</th><th>Types</th></tr>
1331 <td><a href="#t_integer">integer</a></td>
1332 <td><tt>i1, i2, i3, ... i8, ... i16, ... i32, ... i64, ... </tt></td>
1335 <td><a href="#t_floating">floating point</a></td>
1336 <td><tt>float, double, x86_fp80, fp128, ppc_fp128</tt></td>
1339 <td><a name="t_firstclass">first class</a></td>
1340 <td><a href="#t_integer">integer</a>,
1341 <a href="#t_floating">floating point</a>,
1342 <a href="#t_pointer">pointer</a>,
1343 <a href="#t_vector">vector</a>,
1344 <a href="#t_struct">structure</a>,
1345 <a href="#t_array">array</a>,
1346 <a href="#t_label">label</a>,
1347 <a href="#t_metadata">metadata</a>.
1351 <td><a href="#t_primitive">primitive</a></td>
1352 <td><a href="#t_label">label</a>,
1353 <a href="#t_void">void</a>,
1354 <a href="#t_floating">floating point</a>,
1355 <a href="#t_metadata">metadata</a>.</td>
1358 <td><a href="#t_derived">derived</a></td>
1359 <td><a href="#t_integer">integer</a>,
1360 <a href="#t_array">array</a>,
1361 <a href="#t_function">function</a>,
1362 <a href="#t_pointer">pointer</a>,
1363 <a href="#t_struct">structure</a>,
1364 <a href="#t_pstruct">packed structure</a>,
1365 <a href="#t_vector">vector</a>,
1366 <a href="#t_opaque">opaque</a>.
1372 <p>The <a href="#t_firstclass">first class</a> types are perhaps the most
1373 important. Values of these types are the only ones which can be produced by
1374 instructions, passed as arguments, or used as operands to instructions.</p>
1378 <!-- ======================================================================= -->
1379 <div class="doc_subsection"> <a name="t_primitive">Primitive Types</a> </div>
1381 <div class="doc_text">
1383 <p>The primitive types are the fundamental building blocks of the LLVM
1388 <!-- _______________________________________________________________________ -->
1389 <div class="doc_subsubsection"> <a name="t_floating">Floating Point Types</a> </div>
1391 <div class="doc_text">
1395 <tr><th>Type</th><th>Description</th></tr>
1396 <tr><td><tt>float</tt></td><td>32-bit floating point value</td></tr>
1397 <tr><td><tt>double</tt></td><td>64-bit floating point value</td></tr>
1398 <tr><td><tt>fp128</tt></td><td>128-bit floating point value (112-bit mantissa)</td></tr>
1399 <tr><td><tt>x86_fp80</tt></td><td>80-bit floating point value (X87)</td></tr>
1400 <tr><td><tt>ppc_fp128</tt></td><td>128-bit floating point value (two 64-bits)</td></tr>
1406 <!-- _______________________________________________________________________ -->
1407 <div class="doc_subsubsection"> <a name="t_void">Void Type</a> </div>
1409 <div class="doc_text">
1412 <p>The void type does not represent any value and has no size.</p>
1421 <!-- _______________________________________________________________________ -->
1422 <div class="doc_subsubsection"> <a name="t_label">Label Type</a> </div>
1424 <div class="doc_text">
1427 <p>The label type represents code labels.</p>
1436 <!-- _______________________________________________________________________ -->
1437 <div class="doc_subsubsection"> <a name="t_metadata">Metadata Type</a> </div>
1439 <div class="doc_text">
1442 <p>The metadata type represents embedded metadata. The only derived type that
1443 may contain metadata is <tt>metadata*</tt> or a function type that returns or
1444 takes metadata typed parameters, but not pointer to metadata types.</p>
1454 <!-- ======================================================================= -->
1455 <div class="doc_subsection"> <a name="t_derived">Derived Types</a> </div>
1457 <div class="doc_text">
1459 <p>The real power in LLVM comes from the derived types in the system. This is
1460 what allows a programmer to represent arrays, functions, pointers, and other
1461 useful types. Note that these derived types may be recursive: For example,
1462 it is possible to have a two dimensional array.</p>
1466 <!-- _______________________________________________________________________ -->
1467 <div class="doc_subsubsection"> <a name="t_integer">Integer Type</a> </div>
1469 <div class="doc_text">
1472 <p>The integer type is a very simple derived type that simply specifies an
1473 arbitrary bit width for the integer type desired. Any bit width from 1 bit to
1474 2^23-1 (about 8 million) can be specified.</p>
1481 <p>The number of bits the integer will occupy is specified by the <tt>N</tt>
1485 <table class="layout">
1487 <td class="left"><tt>i1</tt></td>
1488 <td class="left">a single-bit integer.</td>
1491 <td class="left"><tt>i32</tt></td>
1492 <td class="left">a 32-bit integer.</td>
1495 <td class="left"><tt>i1942652</tt></td>
1496 <td class="left">a really big integer of over 1 million bits.</td>
1500 <p>Note that the code generator does not yet support large integer types to be
1501 used as function return types. The specific limit on how large a return type
1502 the code generator can currently handle is target-dependent; currently it's
1503 often 64 bits for 32-bit targets and 128 bits for 64-bit targets.</p>
1507 <!-- _______________________________________________________________________ -->
1508 <div class="doc_subsubsection"> <a name="t_array">Array Type</a> </div>
1510 <div class="doc_text">
1513 <p>The array type is a very simple derived type that arranges elements
1514 sequentially in memory. The array type requires a size (number of elements)
1515 and an underlying data type.</p>
1519 [<# elements> x <elementtype>]
1522 <p>The number of elements is a constant integer value; <tt>elementtype</tt> may
1523 be any type with a size.</p>
1526 <table class="layout">
1528 <td class="left"><tt>[40 x i32]</tt></td>
1529 <td class="left">Array of 40 32-bit integer values.</td>
1532 <td class="left"><tt>[41 x i32]</tt></td>
1533 <td class="left">Array of 41 32-bit integer values.</td>
1536 <td class="left"><tt>[4 x i8]</tt></td>
1537 <td class="left">Array of 4 8-bit integer values.</td>
1540 <p>Here are some examples of multidimensional arrays:</p>
1541 <table class="layout">
1543 <td class="left"><tt>[3 x [4 x i32]]</tt></td>
1544 <td class="left">3x4 array of 32-bit integer values.</td>
1547 <td class="left"><tt>[12 x [10 x float]]</tt></td>
1548 <td class="left">12x10 array of single precision floating point values.</td>
1551 <td class="left"><tt>[2 x [3 x [4 x i16]]]</tt></td>
1552 <td class="left">2x3x4 array of 16-bit integer values.</td>
1556 <p>Note that 'variable sized arrays' can be implemented in LLVM with a zero
1557 length array. Normally, accesses past the end of an array are undefined in
1558 LLVM (e.g. it is illegal to access the 5th element of a 3 element array). As
1559 a special case, however, zero length arrays are recognized to be variable
1560 length. This allows implementation of 'pascal style arrays' with the LLVM
1561 type "<tt>{ i32, [0 x float]}</tt>", for example.</p>
1563 <p>Note that the code generator does not yet support large aggregate types to be
1564 used as function return types. The specific limit on how large an aggregate
1565 return type the code generator can currently handle is target-dependent, and
1566 also dependent on the aggregate element types.</p>
1570 <!-- _______________________________________________________________________ -->
1571 <div class="doc_subsubsection"> <a name="t_function">Function Type</a> </div>
1573 <div class="doc_text">
1576 <p>The function type can be thought of as a function signature. It consists of
1577 a return type and a list of formal parameter types. The return type of a
1578 function type is a scalar type, a void type, or a struct type. If the return
1579 type is a struct type then all struct elements must be of first class types,
1580 and the struct must have at least one element.</p>
1584 <returntype list> (<parameter list>)
1587 <p>...where '<tt><parameter list></tt>' is a comma-separated list of type
1588 specifiers. Optionally, the parameter list may include a type <tt>...</tt>,
1589 which indicates that the function takes a variable number of arguments.
1590 Variable argument functions can access their arguments with
1591 the <a href="#int_varargs">variable argument handling intrinsic</a>
1592 functions. '<tt><returntype list></tt>' is a comma-separated list of
1593 <a href="#t_firstclass">first class</a> type specifiers.</p>
1596 <table class="layout">
1598 <td class="left"><tt>i32 (i32)</tt></td>
1599 <td class="left">function taking an <tt>i32</tt>, returning an <tt>i32</tt>
1601 </tr><tr class="layout">
1602 <td class="left"><tt>float (i16 signext, i32 *) *
1604 <td class="left"><a href="#t_pointer">Pointer</a> to a function that takes
1605 an <tt>i16</tt> that should be sign extended and a
1606 <a href="#t_pointer">pointer</a> to <tt>i32</tt>, returning
1609 </tr><tr class="layout">
1610 <td class="left"><tt>i32 (i8*, ...)</tt></td>
1611 <td class="left">A vararg function that takes at least one
1612 <a href="#t_pointer">pointer</a> to <tt>i8 </tt> (char in C),
1613 which returns an integer. This is the signature for <tt>printf</tt> in
1616 </tr><tr class="layout">
1617 <td class="left"><tt>{i32, i32} (i32)</tt></td>
1618 <td class="left">A function taking an <tt>i32</tt>, returning two
1619 <tt>i32</tt> values as an aggregate of type <tt>{ i32, i32 }</tt>
1626 <!-- _______________________________________________________________________ -->
1627 <div class="doc_subsubsection"> <a name="t_struct">Structure Type</a> </div>
1629 <div class="doc_text">
1632 <p>The structure type is used to represent a collection of data members together
1633 in memory. The packing of the field types is defined to match the ABI of the
1634 underlying processor. The elements of a structure may be any type that has a
1637 <p>Structures are accessed using '<tt><a href="#i_load">load</a></tt> and
1638 '<tt><a href="#i_store">store</a></tt>' by getting a pointer to a field with
1639 the '<tt><a href="#i_getelementptr">getelementptr</a></tt>' instruction.</p>
1643 { <type list> }
1647 <table class="layout">
1649 <td class="left"><tt>{ i32, i32, i32 }</tt></td>
1650 <td class="left">A triple of three <tt>i32</tt> values</td>
1651 </tr><tr class="layout">
1652 <td class="left"><tt>{ float, i32 (i32) * }</tt></td>
1653 <td class="left">A pair, where the first element is a <tt>float</tt> and the
1654 second element is a <a href="#t_pointer">pointer</a> to a
1655 <a href="#t_function">function</a> that takes an <tt>i32</tt>, returning
1656 an <tt>i32</tt>.</td>
1660 <p>Note that the code generator does not yet support large aggregate types to be
1661 used as function return types. The specific limit on how large an aggregate
1662 return type the code generator can currently handle is target-dependent, and
1663 also dependent on the aggregate element types.</p>
1667 <!-- _______________________________________________________________________ -->
1668 <div class="doc_subsubsection"> <a name="t_pstruct">Packed Structure Type</a>
1671 <div class="doc_text">
1674 <p>The packed structure type is used to represent a collection of data members
1675 together in memory. There is no padding between fields. Further, the
1676 alignment of a packed structure is 1 byte. The elements of a packed
1677 structure may be any type that has a size.</p>
1679 <p>Structures are accessed using '<tt><a href="#i_load">load</a></tt> and
1680 '<tt><a href="#i_store">store</a></tt>' by getting a pointer to a field with
1681 the '<tt><a href="#i_getelementptr">getelementptr</a></tt>' instruction.</p>
1685 < { <type list> } >
1689 <table class="layout">
1691 <td class="left"><tt>< { i32, i32, i32 } ></tt></td>
1692 <td class="left">A triple of three <tt>i32</tt> values</td>
1693 </tr><tr class="layout">
1695 <tt>< { float, i32 (i32)* } ></tt></td>
1696 <td class="left">A pair, where the first element is a <tt>float</tt> and the
1697 second element is a <a href="#t_pointer">pointer</a> to a
1698 <a href="#t_function">function</a> that takes an <tt>i32</tt>, returning
1699 an <tt>i32</tt>.</td>
1705 <!-- _______________________________________________________________________ -->
1706 <div class="doc_subsubsection"> <a name="t_pointer">Pointer Type</a> </div>
1708 <div class="doc_text">
1711 <p>As in many languages, the pointer type represents a pointer or reference to
1712 another object, which must live in memory. Pointer types may have an optional
1713 address space attribute defining the target-specific numbered address space
1714 where the pointed-to object resides. The default address space is zero.</p>
1716 <p>Note that LLVM does not permit pointers to void (<tt>void*</tt>) nor does it
1717 permit pointers to labels (<tt>label*</tt>). Use <tt>i8*</tt> instead.</p>
1725 <table class="layout">
1727 <td class="left"><tt>[4 x i32]*</tt></td>
1728 <td class="left">A <a href="#t_pointer">pointer</a> to <a
1729 href="#t_array">array</a> of four <tt>i32</tt> values.</td>
1732 <td class="left"><tt>i32 (i32 *) *</tt></td>
1733 <td class="left"> A <a href="#t_pointer">pointer</a> to a <a
1734 href="#t_function">function</a> that takes an <tt>i32*</tt>, returning an
1738 <td class="left"><tt>i32 addrspace(5)*</tt></td>
1739 <td class="left">A <a href="#t_pointer">pointer</a> to an <tt>i32</tt> value
1740 that resides in address space #5.</td>
1746 <!-- _______________________________________________________________________ -->
1747 <div class="doc_subsubsection"> <a name="t_vector">Vector Type</a> </div>
1749 <div class="doc_text">
1752 <p>A vector type is a simple derived type that represents a vector of elements.
1753 Vector types are used when multiple primitive data are operated in parallel
1754 using a single instruction (SIMD). A vector type requires a size (number of
1755 elements) and an underlying primitive data type. Vectors must have a power
1756 of two length (1, 2, 4, 8, 16 ...). Vector types are considered
1757 <a href="#t_firstclass">first class</a>.</p>
1761 < <# elements> x <elementtype> >
1764 <p>The number of elements is a constant integer value; elementtype may be any
1765 integer or floating point type.</p>
1768 <table class="layout">
1770 <td class="left"><tt><4 x i32></tt></td>
1771 <td class="left">Vector of 4 32-bit integer values.</td>
1774 <td class="left"><tt><8 x float></tt></td>
1775 <td class="left">Vector of 8 32-bit floating-point values.</td>
1778 <td class="left"><tt><2 x i64></tt></td>
1779 <td class="left">Vector of 2 64-bit integer values.</td>
1783 <p>Note that the code generator does not yet support large vector types to be
1784 used as function return types. The specific limit on how large a vector
1785 return type codegen can currently handle is target-dependent; currently it's
1786 often a few times longer than a hardware vector register.</p>
1790 <!-- _______________________________________________________________________ -->
1791 <div class="doc_subsubsection"> <a name="t_opaque">Opaque Type</a> </div>
1792 <div class="doc_text">
1795 <p>Opaque types are used to represent unknown types in the system. This
1796 corresponds (for example) to the C notion of a forward declared structure
1797 type. In LLVM, opaque types can eventually be resolved to any type (not just
1798 a structure type).</p>
1806 <table class="layout">
1808 <td class="left"><tt>opaque</tt></td>
1809 <td class="left">An opaque type.</td>
1815 <!-- ======================================================================= -->
1816 <div class="doc_subsection">
1817 <a name="t_uprefs">Type Up-references</a>
1820 <div class="doc_text">
1823 <p>An "up reference" allows you to refer to a lexically enclosing type without
1824 requiring it to have a name. For instance, a structure declaration may
1825 contain a pointer to any of the types it is lexically a member of. Example
1826 of up references (with their equivalent as named type declarations)
1830 { \2 * } %x = type { %x* }
1831 { \2 }* %y = type { %y }*
1835 <p>An up reference is needed by the asmprinter for printing out cyclic types
1836 when there is no declared name for a type in the cycle. Because the
1837 asmprinter does not want to print out an infinite type string, it needs a
1838 syntax to handle recursive types that have no names (all names are optional
1846 <p>The level is the count of the lexical type that is being referred to.</p>
1849 <table class="layout">
1851 <td class="left"><tt>\1*</tt></td>
1852 <td class="left">Self-referential pointer.</td>
1855 <td class="left"><tt>{ { \3*, i8 }, i32 }</tt></td>
1856 <td class="left">Recursive structure where the upref refers to the out-most
1863 <!-- *********************************************************************** -->
1864 <div class="doc_section"> <a name="constants">Constants</a> </div>
1865 <!-- *********************************************************************** -->
1867 <div class="doc_text">
1869 <p>LLVM has several different basic types of constants. This section describes
1870 them all and their syntax.</p>
1874 <!-- ======================================================================= -->
1875 <div class="doc_subsection"><a name="simpleconstants">Simple Constants</a></div>
1877 <div class="doc_text">
1880 <dt><b>Boolean constants</b></dt>
1881 <dd>The two strings '<tt>true</tt>' and '<tt>false</tt>' are both valid
1882 constants of the <tt><a href="#t_primitive">i1</a></tt> type.</dd>
1884 <dt><b>Integer constants</b></dt>
1885 <dd>Standard integers (such as '4') are constants of
1886 the <a href="#t_integer">integer</a> type. Negative numbers may be used
1887 with integer types.</dd>
1889 <dt><b>Floating point constants</b></dt>
1890 <dd>Floating point constants use standard decimal notation (e.g. 123.421),
1891 exponential notation (e.g. 1.23421e+2), or a more precise hexadecimal
1892 notation (see below). The assembler requires the exact decimal value of a
1893 floating-point constant. For example, the assembler accepts 1.25 but
1894 rejects 1.3 because 1.3 is a repeating decimal in binary. Floating point
1895 constants must have a <a href="#t_floating">floating point</a> type. </dd>
1897 <dt><b>Null pointer constants</b></dt>
1898 <dd>The identifier '<tt>null</tt>' is recognized as a null pointer constant
1899 and must be of <a href="#t_pointer">pointer type</a>.</dd>
1902 <p>The one non-intuitive notation for constants is the hexadecimal form of
1903 floating point constants. For example, the form '<tt>double
1904 0x432ff973cafa8000</tt>' is equivalent to (but harder to read than)
1905 '<tt>double 4.5e+15</tt>'. The only time hexadecimal floating point
1906 constants are required (and the only time that they are generated by the
1907 disassembler) is when a floating point constant must be emitted but it cannot
1908 be represented as a decimal floating point number in a reasonable number of
1909 digits. For example, NaN's, infinities, and other special values are
1910 represented in their IEEE hexadecimal format so that assembly and disassembly
1911 do not cause any bits to change in the constants.</p>
1913 <p>When using the hexadecimal form, constants of types float and double are
1914 represented using the 16-digit form shown above (which matches the IEEE754
1915 representation for double); float values must, however, be exactly
1916 representable as IEE754 single precision. Hexadecimal format is always used
1917 for long double, and there are three forms of long double. The 80-bit format
1918 used by x86 is represented as <tt>0xK</tt> followed by 20 hexadecimal digits.
1919 The 128-bit format used by PowerPC (two adjacent doubles) is represented
1920 by <tt>0xM</tt> followed by 32 hexadecimal digits. The IEEE 128-bit format
1921 is represented by <tt>0xL</tt> followed by 32 hexadecimal digits; no
1922 currently supported target uses this format. Long doubles will only work if
1923 they match the long double format on your target. All hexadecimal formats
1924 are big-endian (sign bit at the left).</p>
1928 <!-- ======================================================================= -->
1929 <div class="doc_subsection">
1930 <a name="aggregateconstants"></a> <!-- old anchor -->
1931 <a name="complexconstants">Complex Constants</a>
1934 <div class="doc_text">
1936 <p>Complex constants are a (potentially recursive) combination of simple
1937 constants and smaller complex constants.</p>
1940 <dt><b>Structure constants</b></dt>
1941 <dd>Structure constants are represented with notation similar to structure
1942 type definitions (a comma separated list of elements, surrounded by braces
1943 (<tt>{}</tt>)). For example: "<tt>{ i32 4, float 17.0, i32* @G }</tt>",
1944 where "<tt>@G</tt>" is declared as "<tt>@G = external global i32</tt>".
1945 Structure constants must have <a href="#t_struct">structure type</a>, and
1946 the number and types of elements must match those specified by the
1949 <dt><b>Array constants</b></dt>
1950 <dd>Array constants are represented with notation similar to array type
1951 definitions (a comma separated list of elements, surrounded by square
1952 brackets (<tt>[]</tt>)). For example: "<tt>[ i32 42, i32 11, i32 74
1953 ]</tt>". Array constants must have <a href="#t_array">array type</a>, and
1954 the number and types of elements must match those specified by the
1957 <dt><b>Vector constants</b></dt>
1958 <dd>Vector constants are represented with notation similar to vector type
1959 definitions (a comma separated list of elements, surrounded by
1960 less-than/greater-than's (<tt><></tt>)). For example: "<tt>< i32
1961 42, i32 11, i32 74, i32 100 ></tt>". Vector constants must
1962 have <a href="#t_vector">vector type</a>, and the number and types of
1963 elements must match those specified by the type.</dd>
1965 <dt><b>Zero initialization</b></dt>
1966 <dd>The string '<tt>zeroinitializer</tt>' can be used to zero initialize a
1967 value to zero of <em>any</em> type, including scalar and aggregate types.
1968 This is often used to avoid having to print large zero initializers
1969 (e.g. for large arrays) and is always exactly equivalent to using explicit
1970 zero initializers.</dd>
1972 <dt><b>Metadata node</b></dt>
1973 <dd>A metadata node is a structure-like constant with
1974 <a href="#t_metadata">metadata type</a>. For example: "<tt>metadata !{
1975 i32 0, metadata !"test" }</tt>". Unlike other constants that are meant to
1976 be interpreted as part of the instruction stream, metadata is a place to
1977 attach additional information such as debug info.</dd>
1982 <!-- ======================================================================= -->
1983 <div class="doc_subsection">
1984 <a name="globalconstants">Global Variable and Function Addresses</a>
1987 <div class="doc_text">
1989 <p>The addresses of <a href="#globalvars">global variables</a>
1990 and <a href="#functionstructure">functions</a> are always implicitly valid
1991 (link-time) constants. These constants are explicitly referenced when
1992 the <a href="#identifiers">identifier for the global</a> is used and always
1993 have <a href="#t_pointer">pointer</a> type. For example, the following is a
1994 legal LLVM file:</p>
1996 <div class="doc_code">
2000 @Z = global [2 x i32*] [ i32* @X, i32* @Y ]
2006 <!-- ======================================================================= -->
2007 <div class="doc_subsection"><a name="undefvalues">Undefined Values</a></div>
2008 <div class="doc_text">
2010 <p>The string '<tt>undef</tt>' is recognized as a type-less constant that has no
2011 specific value. Undefined values may be of any type and be used anywhere a
2012 constant is permitted.</p>
2014 <p>Undefined values indicate to the compiler that the program is well defined no
2015 matter what value is used, giving the compiler more freedom to optimize.</p>
2019 <!-- ======================================================================= -->
2020 <div class="doc_subsection"><a name="constantexprs">Constant Expressions</a>
2023 <div class="doc_text">
2025 <p>Constant expressions are used to allow expressions involving other constants
2026 to be used as constants. Constant expressions may be of
2027 any <a href="#t_firstclass">first class</a> type and may involve any LLVM
2028 operation that does not have side effects (e.g. load and call are not
2029 supported). The following is the syntax for constant expressions:</p>
2032 <dt><b><tt>trunc ( CST to TYPE )</tt></b></dt>
2033 <dd>Truncate a constant to another type. The bit size of CST must be larger
2034 than the bit size of TYPE. Both types must be integers.</dd>
2036 <dt><b><tt>zext ( CST to TYPE )</tt></b></dt>
2037 <dd>Zero extend a constant to another type. The bit size of CST must be
2038 smaller or equal to the bit size of TYPE. Both types must be
2041 <dt><b><tt>sext ( CST to TYPE )</tt></b></dt>
2042 <dd>Sign extend a constant to another type. The bit size of CST must be
2043 smaller or equal to the bit size of TYPE. Both types must be
2046 <dt><b><tt>fptrunc ( CST to TYPE )</tt></b></dt>
2047 <dd>Truncate a floating point constant to another floating point type. The
2048 size of CST must be larger than the size of TYPE. Both types must be
2049 floating point.</dd>
2051 <dt><b><tt>fpext ( CST to TYPE )</tt></b></dt>
2052 <dd>Floating point extend a constant to another type. The size of CST must be
2053 smaller or equal to the size of TYPE. Both types must be floating
2056 <dt><b><tt>fptoui ( CST to TYPE )</tt></b></dt>
2057 <dd>Convert a floating point constant to the corresponding unsigned integer
2058 constant. TYPE must be a scalar or vector integer type. CST must be of
2059 scalar or vector floating point type. Both CST and TYPE must be scalars,
2060 or vectors of the same number of elements. If the value won't fit in the
2061 integer type, the results are undefined.</dd>
2063 <dt><b><tt>fptosi ( CST to TYPE )</tt></b></dt>
2064 <dd>Convert a floating point constant to the corresponding signed integer
2065 constant. TYPE must be a scalar or vector integer type. CST must be of
2066 scalar or vector floating point type. Both CST and TYPE must be scalars,
2067 or vectors of the same number of elements. If the value won't fit in the
2068 integer type, the results are undefined.</dd>
2070 <dt><b><tt>uitofp ( CST to TYPE )</tt></b></dt>
2071 <dd>Convert an unsigned integer constant to the corresponding floating point
2072 constant. TYPE must be a scalar or vector floating point type. CST must be
2073 of scalar or vector integer type. Both CST and TYPE must be scalars, or
2074 vectors of the same number of elements. If the value won't fit in the
2075 floating point type, the results are undefined.</dd>
2077 <dt><b><tt>sitofp ( CST to TYPE )</tt></b></dt>
2078 <dd>Convert a signed integer constant to the corresponding floating point
2079 constant. TYPE must be a scalar or vector floating point type. CST must be
2080 of scalar or vector integer type. Both CST and TYPE must be scalars, or
2081 vectors of the same number of elements. If the value won't fit in the
2082 floating point type, the results are undefined.</dd>
2084 <dt><b><tt>ptrtoint ( CST to TYPE )</tt></b></dt>
2085 <dd>Convert a pointer typed constant to the corresponding integer constant
2086 <tt>TYPE</tt> must be an integer type. <tt>CST</tt> must be of pointer
2087 type. The <tt>CST</tt> value is zero extended, truncated, or unchanged to
2088 make it fit in <tt>TYPE</tt>.</dd>
2090 <dt><b><tt>inttoptr ( CST to TYPE )</tt></b></dt>
2091 <dd>Convert a integer constant to a pointer constant. TYPE must be a pointer
2092 type. CST must be of integer type. The CST value is zero extended,
2093 truncated, or unchanged to make it fit in a pointer size. This one is
2094 <i>really</i> dangerous!</dd>
2096 <dt><b><tt>bitcast ( CST to TYPE )</tt></b></dt>
2097 <dd>Convert a constant, CST, to another TYPE. The constraints of the operands
2098 are the same as those for the <a href="#i_bitcast">bitcast
2099 instruction</a>.</dd>
2101 <dt><b><tt>getelementptr ( CSTPTR, IDX0, IDX1, ... )</tt></b></dt>
2102 <dt><b><tt>getelementptr inbounds ( CSTPTR, IDX0, IDX1, ... )</tt></b></dt>
2103 <dd>Perform the <a href="#i_getelementptr">getelementptr operation</a> on
2104 constants. As with the <a href="#i_getelementptr">getelementptr</a>
2105 instruction, the index list may have zero or more indexes, which are
2106 required to make sense for the type of "CSTPTR".</dd>
2108 <dt><b><tt>select ( COND, VAL1, VAL2 )</tt></b></dt>
2109 <dd>Perform the <a href="#i_select">select operation</a> on constants.</dd>
2111 <dt><b><tt>icmp COND ( VAL1, VAL2 )</tt></b></dt>
2112 <dd>Performs the <a href="#i_icmp">icmp operation</a> on constants.</dd>
2114 <dt><b><tt>fcmp COND ( VAL1, VAL2 )</tt></b></dt>
2115 <dd>Performs the <a href="#i_fcmp">fcmp operation</a> on constants.</dd>
2117 <dt><b><tt>extractelement ( VAL, IDX )</tt></b></dt>
2118 <dd>Perform the <a href="#i_extractelement">extractelement operation</a> on
2121 <dt><b><tt>insertelement ( VAL, ELT, IDX )</tt></b></dt>
2122 <dd>Perform the <a href="#i_insertelement">insertelement operation</a> on
2125 <dt><b><tt>shufflevector ( VEC1, VEC2, IDXMASK )</tt></b></dt>
2126 <dd>Perform the <a href="#i_shufflevector">shufflevector operation</a> on
2129 <dt><b><tt>OPCODE ( LHS, RHS )</tt></b></dt>
2130 <dd>Perform the specified operation of the LHS and RHS constants. OPCODE may
2131 be any of the <a href="#binaryops">binary</a>
2132 or <a href="#bitwiseops">bitwise binary</a> operations. The constraints
2133 on operands are the same as those for the corresponding instruction
2134 (e.g. no bitwise operations on floating point values are allowed).</dd>
2139 <!-- ======================================================================= -->
2140 <div class="doc_subsection"><a name="metadata">Embedded Metadata</a>
2143 <div class="doc_text">
2145 <p>Embedded metadata provides a way to attach arbitrary data to the instruction
2146 stream without affecting the behaviour of the program. There are two
2147 metadata primitives, strings and nodes. All metadata has the
2148 <tt>metadata</tt> type and is identified in syntax by a preceding exclamation
2149 point ('<tt>!</tt>').</p>
2151 <p>A metadata string is a string surrounded by double quotes. It can contain
2152 any character by escaping non-printable characters with "\xx" where "xx" is
2153 the two digit hex code. For example: "<tt>!"test\00"</tt>".</p>
2155 <p>Metadata nodes are represented with notation similar to structure constants
2156 (a comma separated list of elements, surrounded by braces and preceeded by an
2157 exclamation point). For example: "<tt>!{ metadata !"test\00", i32
2160 <p>A metadata node will attempt to track changes to the values it holds. In the
2161 event that a value is deleted, it will be replaced with a typeless
2162 "<tt>null</tt>", such as "<tt>metadata !{null, i32 10}</tt>".</p>
2164 <p>Optimizations may rely on metadata to provide additional information about
2165 the program that isn't available in the instructions, or that isn't easily
2166 computable. Similarly, the code generator may expect a certain metadata
2167 format to be used to express debugging information.</p>
2171 <!-- *********************************************************************** -->
2172 <div class="doc_section"> <a name="othervalues">Other Values</a> </div>
2173 <!-- *********************************************************************** -->
2175 <!-- ======================================================================= -->
2176 <div class="doc_subsection">
2177 <a name="inlineasm">Inline Assembler Expressions</a>
2180 <div class="doc_text">
2182 <p>LLVM supports inline assembler expressions (as opposed
2183 to <a href="#moduleasm"> Module-Level Inline Assembly</a>) through the use of
2184 a special value. This value represents the inline assembler as a string
2185 (containing the instructions to emit), a list of operand constraints (stored
2186 as a string), and a flag that indicates whether or not the inline asm
2187 expression has side effects. An example inline assembler expression is:</p>
2189 <div class="doc_code">
2191 i32 (i32) asm "bswap $0", "=r,r"
2195 <p>Inline assembler expressions may <b>only</b> be used as the callee operand of
2196 a <a href="#i_call"><tt>call</tt> instruction</a>. Thus, typically we
2199 <div class="doc_code">
2201 %X = call i32 asm "<a href="#int_bswap">bswap</a> $0", "=r,r"(i32 %Y)
2205 <p>Inline asms with side effects not visible in the constraint list must be
2206 marked as having side effects. This is done through the use of the
2207 '<tt>sideeffect</tt>' keyword, like so:</p>
2209 <div class="doc_code">
2211 call void asm sideeffect "eieio", ""()
2215 <p>TODO: The format of the asm and constraints string still need to be
2216 documented here. Constraints on what can be done (e.g. duplication, moving,
2217 etc need to be documented). This is probably best done by reference to
2218 another document that covers inline asm from a holistic perspective.</p>
2223 <!-- *********************************************************************** -->
2224 <div class="doc_section">
2225 <a name="intrinsic_globals">Intrinsic Global Variables</a>
2227 <!-- *********************************************************************** -->
2229 <p>LLVM has a number of "magic" global variables that contain data that affect
2230 code generation or other IR semantics. These are documented here. All globals
2231 of this sort should have a section specified as "<tt>llvm.metadata</tt>". This
2232 section and all globals that start with "<tt>llvm.</tt>" are reserved for use
2235 <!-- ======================================================================= -->
2236 <div class="doc_subsection">
2237 <a name="intg_used">The '<tt>llvm.used</tt>' Global Variable</a>
2240 <div class="doc_text">
2242 <p>The <tt>@llvm.used</tt> global is an array with i8* element type which has <a
2243 href="#linkage_appending">appending linkage</a>. This array contains a list of
2244 pointers to global variables and functions which may optionally have a pointer
2245 cast formed of bitcast or getelementptr. For example, a legal use of it is:</p>
2251 @llvm.used = appending global [2 x i8*] [
2253 i8* bitcast (i32* @Y to i8*)
2254 ], section "llvm.metadata"
2257 <p>If a global variable appears in the <tt>@llvm.used</tt> list, then the
2258 compiler, assembler, and linker are required to treat the symbol as if there is
2259 a reference to the global that it cannot see. For example, if a variable has
2260 internal linkage and no references other than that from the <tt>@llvm.used</tt>
2261 list, it cannot be deleted. This is commonly used to represent references from
2262 inline asms and other things the compiler cannot "see", and corresponds to
2263 "attribute((used))" in GNU C.</p>
2265 <p>On some targets, the code generator must emit a directive to the assembler or
2266 object file to prevent the assembler and linker from molesting the symbol.</p>
2270 <!-- ======================================================================= -->
2271 <div class="doc_subsection">
2272 <a name="intg_compiler_used">The '<tt>llvm.compiler.used</tt>' Global Variable</a>
2275 <div class="doc_text">
2277 <p>The <tt>@llvm.compiler.used</tt> directive is the same as the
2278 <tt>@llvm.used</tt> directive, except that it only prevents the compiler from
2279 touching the symbol. On targets that support it, this allows an intelligent
2280 linker to optimize references to the symbol without being impeded as it would be
2281 by <tt>@llvm.used</tt>.</p>
2283 <p>This is a rare construct that should only be used in rare circumstances, and
2284 should not be exposed to source languages.</p>
2288 <!-- ======================================================================= -->
2289 <div class="doc_subsection">
2290 <a name="intg_global_ctors">The '<tt>llvm.global_ctors</tt>' Global Variable</a>
2293 <div class="doc_text">
2295 <p>TODO: Describe this.</p>
2299 <!-- ======================================================================= -->
2300 <div class="doc_subsection">
2301 <a name="intg_global_dtors">The '<tt>llvm.global_dtors</tt>' Global Variable</a>
2304 <div class="doc_text">
2306 <p>TODO: Describe this.</p>
2311 <!-- *********************************************************************** -->
2312 <div class="doc_section"> <a name="instref">Instruction Reference</a> </div>
2313 <!-- *********************************************************************** -->
2315 <div class="doc_text">
2317 <p>The LLVM instruction set consists of several different classifications of
2318 instructions: <a href="#terminators">terminator
2319 instructions</a>, <a href="#binaryops">binary instructions</a>,
2320 <a href="#bitwiseops">bitwise binary instructions</a>,
2321 <a href="#memoryops">memory instructions</a>, and
2322 <a href="#otherops">other instructions</a>.</p>
2326 <!-- ======================================================================= -->
2327 <div class="doc_subsection"> <a name="terminators">Terminator
2328 Instructions</a> </div>
2330 <div class="doc_text">
2332 <p>As mentioned <a href="#functionstructure">previously</a>, every basic block
2333 in a program ends with a "Terminator" instruction, which indicates which
2334 block should be executed after the current block is finished. These
2335 terminator instructions typically yield a '<tt>void</tt>' value: they produce
2336 control flow, not values (the one exception being the
2337 '<a href="#i_invoke"><tt>invoke</tt></a>' instruction).</p>
2339 <p>There are six different terminator instructions: the
2340 '<a href="#i_ret"><tt>ret</tt></a>' instruction, the
2341 '<a href="#i_br"><tt>br</tt></a>' instruction, the
2342 '<a href="#i_switch"><tt>switch</tt></a>' instruction, the
2343 '<a href="#i_invoke"><tt>invoke</tt></a>' instruction, the
2344 '<a href="#i_unwind"><tt>unwind</tt></a>' instruction, and the
2345 '<a href="#i_unreachable"><tt>unreachable</tt></a>' instruction.</p>
2349 <!-- _______________________________________________________________________ -->
2350 <div class="doc_subsubsection"> <a name="i_ret">'<tt>ret</tt>'
2351 Instruction</a> </div>
2353 <div class="doc_text">
2357 ret <type> <value> <i>; Return a value from a non-void function</i>
2358 ret void <i>; Return from void function</i>
2362 <p>The '<tt>ret</tt>' instruction is used to return control flow (and optionally
2363 a value) from a function back to the caller.</p>
2365 <p>There are two forms of the '<tt>ret</tt>' instruction: one that returns a
2366 value and then causes control flow, and one that just causes control flow to
2370 <p>The '<tt>ret</tt>' instruction optionally accepts a single argument, the
2371 return value. The type of the return value must be a
2372 '<a href="#t_firstclass">first class</a>' type.</p>
2374 <p>A function is not <a href="#wellformed">well formed</a> if it it has a
2375 non-void return type and contains a '<tt>ret</tt>' instruction with no return
2376 value or a return value with a type that does not match its type, or if it
2377 has a void return type and contains a '<tt>ret</tt>' instruction with a
2381 <p>When the '<tt>ret</tt>' instruction is executed, control flow returns back to
2382 the calling function's context. If the caller is a
2383 "<a href="#i_call"><tt>call</tt></a>" instruction, execution continues at the
2384 instruction after the call. If the caller was an
2385 "<a href="#i_invoke"><tt>invoke</tt></a>" instruction, execution continues at
2386 the beginning of the "normal" destination block. If the instruction returns
2387 a value, that value shall set the call or invoke instruction's return
2392 ret i32 5 <i>; Return an integer value of 5</i>
2393 ret void <i>; Return from a void function</i>
2394 ret { i32, i8 } { i32 4, i8 2 } <i>; Return a struct of values 4 and 2</i>
2397 <p>Note that the code generator does not yet fully support large
2398 return values. The specific sizes that are currently supported are
2399 dependent on the target. For integers, on 32-bit targets the limit
2400 is often 64 bits, and on 64-bit targets the limit is often 128 bits.
2401 For aggregate types, the current limits are dependent on the element
2402 types; for example targets are often limited to 2 total integer
2403 elements and 2 total floating-point elements.</p>
2406 <!-- _______________________________________________________________________ -->
2407 <div class="doc_subsubsection"> <a name="i_br">'<tt>br</tt>' Instruction</a> </div>
2409 <div class="doc_text">
2413 br i1 <cond>, label <iftrue>, label <iffalse><br> br label <dest> <i>; Unconditional branch</i>
2417 <p>The '<tt>br</tt>' instruction is used to cause control flow to transfer to a
2418 different basic block in the current function. There are two forms of this
2419 instruction, corresponding to a conditional branch and an unconditional
2423 <p>The conditional branch form of the '<tt>br</tt>' instruction takes a single
2424 '<tt>i1</tt>' value and two '<tt>label</tt>' values. The unconditional form
2425 of the '<tt>br</tt>' instruction takes a single '<tt>label</tt>' value as a
2429 <p>Upon execution of a conditional '<tt>br</tt>' instruction, the '<tt>i1</tt>'
2430 argument is evaluated. If the value is <tt>true</tt>, control flows to the
2431 '<tt>iftrue</tt>' <tt>label</tt> argument. If "cond" is <tt>false</tt>,
2432 control flows to the '<tt>iffalse</tt>' <tt>label</tt> argument.</p>
2437 %cond = <a href="#i_icmp">icmp</a> eq i32 %a, %b
2438 br i1 %cond, label %IfEqual, label %IfUnequal
2440 <a href="#i_ret">ret</a> i32 1
2442 <a href="#i_ret">ret</a> i32 0
2447 <!-- _______________________________________________________________________ -->
2448 <div class="doc_subsubsection">
2449 <a name="i_switch">'<tt>switch</tt>' Instruction</a>
2452 <div class="doc_text">
2456 switch <intty> <value>, label <defaultdest> [ <intty> <val>, label <dest> ... ]
2460 <p>The '<tt>switch</tt>' instruction is used to transfer control flow to one of
2461 several different places. It is a generalization of the '<tt>br</tt>'
2462 instruction, allowing a branch to occur to one of many possible
2466 <p>The '<tt>switch</tt>' instruction uses three parameters: an integer
2467 comparison value '<tt>value</tt>', a default '<tt>label</tt>' destination,
2468 and an array of pairs of comparison value constants and '<tt>label</tt>'s.
2469 The table is not allowed to contain duplicate constant entries.</p>
2472 <p>The <tt>switch</tt> instruction specifies a table of values and
2473 destinations. When the '<tt>switch</tt>' instruction is executed, this table
2474 is searched for the given value. If the value is found, control flow is
2475 transfered to the corresponding destination; otherwise, control flow is
2476 transfered to the default destination.</p>
2478 <h5>Implementation:</h5>
2479 <p>Depending on properties of the target machine and the particular
2480 <tt>switch</tt> instruction, this instruction may be code generated in
2481 different ways. For example, it could be generated as a series of chained
2482 conditional branches or with a lookup table.</p>
2486 <i>; Emulate a conditional br instruction</i>
2487 %Val = <a href="#i_zext">zext</a> i1 %value to i32
2488 switch i32 %Val, label %truedest [ i32 0, label %falsedest ]
2490 <i>; Emulate an unconditional br instruction</i>
2491 switch i32 0, label %dest [ ]
2493 <i>; Implement a jump table:</i>
2494 switch i32 %val, label %otherwise [ i32 0, label %onzero
2496 i32 2, label %ontwo ]
2501 <!-- _______________________________________________________________________ -->
2502 <div class="doc_subsubsection">
2503 <a name="i_invoke">'<tt>invoke</tt>' Instruction</a>
2506 <div class="doc_text">
2510 <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>]
2511 to label <normal label> unwind label <exception label>
2515 <p>The '<tt>invoke</tt>' instruction causes control to transfer to a specified
2516 function, with the possibility of control flow transfer to either the
2517 '<tt>normal</tt>' label or the '<tt>exception</tt>' label. If the callee
2518 function returns with the "<tt><a href="#i_ret">ret</a></tt>" instruction,
2519 control flow will return to the "normal" label. If the callee (or any
2520 indirect callees) returns with the "<a href="#i_unwind"><tt>unwind</tt></a>"
2521 instruction, control is interrupted and continued at the dynamically nearest
2522 "exception" label.</p>
2525 <p>This instruction requires several arguments:</p>
2528 <li>The optional "cconv" marker indicates which <a href="#callingconv">calling
2529 convention</a> the call should use. If none is specified, the call
2530 defaults to using C calling conventions.</li>
2532 <li>The optional <a href="#paramattrs">Parameter Attributes</a> list for
2533 return values. Only '<tt>zeroext</tt>', '<tt>signext</tt>', and
2534 '<tt>inreg</tt>' attributes are valid here.</li>
2536 <li>'<tt>ptr to function ty</tt>': shall be the signature of the pointer to
2537 function value being invoked. In most cases, this is a direct function
2538 invocation, but indirect <tt>invoke</tt>s are just as possible, branching
2539 off an arbitrary pointer to function value.</li>
2541 <li>'<tt>function ptr val</tt>': An LLVM value containing a pointer to a
2542 function to be invoked. </li>
2544 <li>'<tt>function args</tt>': argument list whose types match the function
2545 signature argument types. If the function signature indicates the
2546 function accepts a variable number of arguments, the extra arguments can
2549 <li>'<tt>normal label</tt>': the label reached when the called function
2550 executes a '<tt><a href="#i_ret">ret</a></tt>' instruction. </li>
2552 <li>'<tt>exception label</tt>': the label reached when a callee returns with
2553 the <a href="#i_unwind"><tt>unwind</tt></a> instruction. </li>
2555 <li>The optional <a href="#fnattrs">function attributes</a> list. Only
2556 '<tt>noreturn</tt>', '<tt>nounwind</tt>', '<tt>readonly</tt>' and
2557 '<tt>readnone</tt>' attributes are valid here.</li>
2561 <p>This instruction is designed to operate as a standard
2562 '<tt><a href="#i_call">call</a></tt>' instruction in most regards. The
2563 primary difference is that it establishes an association with a label, which
2564 is used by the runtime library to unwind the stack.</p>
2566 <p>This instruction is used in languages with destructors to ensure that proper
2567 cleanup is performed in the case of either a <tt>longjmp</tt> or a thrown
2568 exception. Additionally, this is important for implementation of
2569 '<tt>catch</tt>' clauses in high-level languages that support them.</p>
2571 <p>For the purposes of the SSA form, the definition of the value returned by the
2572 '<tt>invoke</tt>' instruction is deemed to occur on the edge from the current
2573 block to the "normal" label. If the callee unwinds then no return value is
2578 %retval = invoke i32 @Test(i32 15) to label %Continue
2579 unwind label %TestCleanup <i>; {i32}:retval set</i>
2580 %retval = invoke <a href="#callingconv">coldcc</a> i32 %Testfnptr(i32 15) to label %Continue
2581 unwind label %TestCleanup <i>; {i32}:retval set</i>
2586 <!-- _______________________________________________________________________ -->
2588 <div class="doc_subsubsection"> <a name="i_unwind">'<tt>unwind</tt>'
2589 Instruction</a> </div>
2591 <div class="doc_text">
2599 <p>The '<tt>unwind</tt>' instruction unwinds the stack, continuing control flow
2600 at the first callee in the dynamic call stack which used
2601 an <a href="#i_invoke"><tt>invoke</tt></a> instruction to perform the call.
2602 This is primarily used to implement exception handling.</p>
2605 <p>The '<tt>unwind</tt>' instruction causes execution of the current function to
2606 immediately halt. The dynamic call stack is then searched for the
2607 first <a href="#i_invoke"><tt>invoke</tt></a> instruction on the call stack.
2608 Once found, execution continues at the "exceptional" destination block
2609 specified by the <tt>invoke</tt> instruction. If there is no <tt>invoke</tt>
2610 instruction in the dynamic call chain, undefined behavior results.</p>
2614 <!-- _______________________________________________________________________ -->
2616 <div class="doc_subsubsection"> <a name="i_unreachable">'<tt>unreachable</tt>'
2617 Instruction</a> </div>
2619 <div class="doc_text">
2627 <p>The '<tt>unreachable</tt>' instruction has no defined semantics. This
2628 instruction is used to inform the optimizer that a particular portion of the
2629 code is not reachable. This can be used to indicate that the code after a
2630 no-return function cannot be reached, and other facts.</p>
2633 <p>The '<tt>unreachable</tt>' instruction has no defined semantics.</p>
2637 <!-- ======================================================================= -->
2638 <div class="doc_subsection"> <a name="binaryops">Binary Operations</a> </div>
2640 <div class="doc_text">
2642 <p>Binary operators are used to do most of the computation in a program. They
2643 require two operands of the same type, execute an operation on them, and
2644 produce a single value. The operands might represent multiple data, as is
2645 the case with the <a href="#t_vector">vector</a> data type. The result value
2646 has the same type as its operands.</p>
2648 <p>There are several different binary operators:</p>
2652 <!-- _______________________________________________________________________ -->
2653 <div class="doc_subsubsection">
2654 <a name="i_add">'<tt>add</tt>' Instruction</a>
2657 <div class="doc_text">
2661 <result> = add <ty> <op1>, <op2> <i>; yields {ty}:result</i>
2662 <result> = nuw add <ty> <op1>, <op2> <i>; yields {ty}:result</i>
2663 <result> = nsw add <ty> <op1>, <op2> <i>; yields {ty}:result</i>
2664 <result> = nuw nsw add <ty> <op1>, <op2> <i>; yields {ty}:result</i>
2668 <p>The '<tt>add</tt>' instruction returns the sum of its two operands.</p>
2671 <p>The two arguments to the '<tt>add</tt>' instruction must
2672 be <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of
2673 integer values. Both arguments must have identical types.</p>
2676 <p>The value produced is the integer sum of the two operands.</p>
2678 <p>If the sum has unsigned overflow, the result returned is the mathematical
2679 result modulo 2<sup>n</sup>, where n is the bit width of the result.</p>
2681 <p>Because LLVM integers use a two's complement representation, this instruction
2682 is appropriate for both signed and unsigned integers.</p>
2684 <p><tt>nuw</tt> and <tt>nsw</tt> stand for "No Unsigned Wrap"
2685 and "No Signed Wrap", respectively. If the <tt>nuw</tt> and/or
2686 <tt>nsw</tt> keywords are present, the result value of the <tt>add</tt>
2687 is undefined if unsigned and/or signed overflow, respectively, occurs.</p>
2691 <result> = add i32 4, %var <i>; yields {i32}:result = 4 + %var</i>
2696 <!-- _______________________________________________________________________ -->
2697 <div class="doc_subsubsection">
2698 <a name="i_fadd">'<tt>fadd</tt>' Instruction</a>
2701 <div class="doc_text">
2705 <result> = fadd <ty> <op1>, <op2> <i>; yields {ty}:result</i>
2709 <p>The '<tt>fadd</tt>' instruction returns the sum of its two operands.</p>
2712 <p>The two arguments to the '<tt>fadd</tt>' instruction must be
2713 <a href="#t_floating">floating point</a> or <a href="#t_vector">vector</a> of
2714 floating point values. Both arguments must have identical types.</p>
2717 <p>The value produced is the floating point sum of the two operands.</p>
2721 <result> = fadd float 4.0, %var <i>; yields {float}:result = 4.0 + %var</i>
2726 <!-- _______________________________________________________________________ -->
2727 <div class="doc_subsubsection">
2728 <a name="i_sub">'<tt>sub</tt>' Instruction</a>
2731 <div class="doc_text">
2735 <result> = sub <ty> <op1>, <op2> <i>; yields {ty}:result</i>
2736 <result> = nuw sub <ty> <op1>, <op2> <i>; yields {ty}:result</i>
2737 <result> = nsw sub <ty> <op1>, <op2> <i>; yields {ty}:result</i>
2738 <result> = nuw nsw sub <ty> <op1>, <op2> <i>; yields {ty}:result</i>
2742 <p>The '<tt>sub</tt>' instruction returns the difference of its two
2745 <p>Note that the '<tt>sub</tt>' instruction is used to represent the
2746 '<tt>neg</tt>' instruction present in most other intermediate
2747 representations.</p>
2750 <p>The two arguments to the '<tt>sub</tt>' instruction must
2751 be <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of
2752 integer values. Both arguments must have identical types.</p>
2755 <p>The value produced is the integer difference of the two operands.</p>
2757 <p>If the difference has unsigned overflow, the result returned is the
2758 mathematical result modulo 2<sup>n</sup>, where n is the bit width of the
2761 <p>Because LLVM integers use a two's complement representation, this instruction
2762 is appropriate for both signed and unsigned integers.</p>
2764 <p><tt>nuw</tt> and <tt>nsw</tt> stand for "No Unsigned Wrap"
2765 and "No Signed Wrap", respectively. If the <tt>nuw</tt> and/or
2766 <tt>nsw</tt> keywords are present, the result value of the <tt>sub</tt>
2767 is undefined if unsigned and/or signed overflow, respectively, occurs.</p>
2771 <result> = sub i32 4, %var <i>; yields {i32}:result = 4 - %var</i>
2772 <result> = sub i32 0, %val <i>; yields {i32}:result = -%var</i>
2777 <!-- _______________________________________________________________________ -->
2778 <div class="doc_subsubsection">
2779 <a name="i_fsub">'<tt>fsub</tt>' Instruction</a>
2782 <div class="doc_text">
2786 <result> = fsub <ty> <op1>, <op2> <i>; yields {ty}:result</i>
2790 <p>The '<tt>fsub</tt>' instruction returns the difference of its two
2793 <p>Note that the '<tt>fsub</tt>' instruction is used to represent the
2794 '<tt>fneg</tt>' instruction present in most other intermediate
2795 representations.</p>
2798 <p>The two arguments to the '<tt>fsub</tt>' instruction must be
2799 <a href="#t_floating">floating point</a> or <a href="#t_vector">vector</a> of
2800 floating point values. Both arguments must have identical types.</p>
2803 <p>The value produced is the floating point difference of the two operands.</p>
2807 <result> = fsub float 4.0, %var <i>; yields {float}:result = 4.0 - %var</i>
2808 <result> = fsub float -0.0, %val <i>; yields {float}:result = -%var</i>
2813 <!-- _______________________________________________________________________ -->
2814 <div class="doc_subsubsection">
2815 <a name="i_mul">'<tt>mul</tt>' Instruction</a>
2818 <div class="doc_text">
2822 <result> = mul <ty> <op1>, <op2> <i>; yields {ty}:result</i>
2823 <result> = nuw mul <ty> <op1>, <op2> <i>; yields {ty}:result</i>
2824 <result> = nsw mul <ty> <op1>, <op2> <i>; yields {ty}:result</i>
2825 <result> = nuw nsw mul <ty> <op1>, <op2> <i>; yields {ty}:result</i>
2829 <p>The '<tt>mul</tt>' instruction returns the product of its two operands.</p>
2832 <p>The two arguments to the '<tt>mul</tt>' instruction must
2833 be <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of
2834 integer values. Both arguments must have identical types.</p>
2837 <p>The value produced is the integer product of the two operands.</p>
2839 <p>If the result of the multiplication has unsigned overflow, the result
2840 returned is the mathematical result modulo 2<sup>n</sup>, where n is the bit
2841 width of the result.</p>
2843 <p>Because LLVM integers use a two's complement representation, and the result
2844 is the same width as the operands, this instruction returns the correct
2845 result for both signed and unsigned integers. If a full product
2846 (e.g. <tt>i32</tt>x<tt>i32</tt>-><tt>i64</tt>) is needed, the operands should
2847 be sign-extended or zero-extended as appropriate to the width of the full
2850 <p><tt>nuw</tt> and <tt>nsw</tt> stand for "No Unsigned Wrap"
2851 and "No Signed Wrap", respectively. If the <tt>nuw</tt> and/or
2852 <tt>nsw</tt> keywords are present, the result value of the <tt>mul</tt>
2853 is undefined if unsigned and/or signed overflow, respectively, occurs.</p>
2857 <result> = mul i32 4, %var <i>; yields {i32}:result = 4 * %var</i>
2862 <!-- _______________________________________________________________________ -->
2863 <div class="doc_subsubsection">
2864 <a name="i_fmul">'<tt>fmul</tt>' Instruction</a>
2867 <div class="doc_text">
2871 <result> = fmul <ty> <op1>, <op2> <i>; yields {ty}:result</i>
2875 <p>The '<tt>fmul</tt>' instruction returns the product of its two operands.</p>
2878 <p>The two arguments to the '<tt>fmul</tt>' instruction must be
2879 <a href="#t_floating">floating point</a> or <a href="#t_vector">vector</a> of
2880 floating point values. Both arguments must have identical types.</p>
2883 <p>The value produced is the floating point product of the two operands.</p>
2887 <result> = fmul float 4.0, %var <i>; yields {float}:result = 4.0 * %var</i>
2892 <!-- _______________________________________________________________________ -->
2893 <div class="doc_subsubsection"> <a name="i_udiv">'<tt>udiv</tt>' Instruction
2896 <div class="doc_text">
2900 <result> = udiv <ty> <op1>, <op2> <i>; yields {ty}:result</i>
2904 <p>The '<tt>udiv</tt>' instruction returns the quotient of its two operands.</p>
2907 <p>The two arguments to the '<tt>udiv</tt>' instruction must be
2908 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
2909 values. Both arguments must have identical types.</p>
2912 <p>The value produced is the unsigned integer quotient of the two operands.</p>
2914 <p>Note that unsigned integer division and signed integer division are distinct
2915 operations; for signed integer division, use '<tt>sdiv</tt>'.</p>
2917 <p>Division by zero leads to undefined behavior.</p>
2921 <result> = udiv i32 4, %var <i>; yields {i32}:result = 4 / %var</i>
2926 <!-- _______________________________________________________________________ -->
2927 <div class="doc_subsubsection"> <a name="i_sdiv">'<tt>sdiv</tt>' Instruction
2930 <div class="doc_text">
2934 <result> = sdiv <ty> <op1>, <op2> <i>; yields {ty}:result</i>
2935 <result> = exact sdiv <ty> <op1>, <op2> <i>; yields {ty}:result</i>
2939 <p>The '<tt>sdiv</tt>' instruction returns the quotient of its two operands.</p>
2942 <p>The two arguments to the '<tt>sdiv</tt>' instruction must be
2943 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
2944 values. Both arguments must have identical types.</p>
2947 <p>The value produced is the signed integer quotient of the two operands rounded
2950 <p>Note that signed integer division and unsigned integer division are distinct
2951 operations; for unsigned integer division, use '<tt>udiv</tt>'.</p>
2953 <p>Division by zero leads to undefined behavior. Overflow also leads to
2954 undefined behavior; this is a rare case, but can occur, for example, by doing
2955 a 32-bit division of -2147483648 by -1.</p>
2957 <p>If the <tt>exact</tt> keyword is present, the result value of the
2958 <tt>sdiv</tt> is undefined if the result would be rounded or if overflow
2963 <result> = sdiv i32 4, %var <i>; yields {i32}:result = 4 / %var</i>
2968 <!-- _______________________________________________________________________ -->
2969 <div class="doc_subsubsection"> <a name="i_fdiv">'<tt>fdiv</tt>'
2970 Instruction</a> </div>
2972 <div class="doc_text">
2976 <result> = fdiv <ty> <op1>, <op2> <i>; yields {ty}:result</i>
2980 <p>The '<tt>fdiv</tt>' instruction returns the quotient of its two operands.</p>
2983 <p>The two arguments to the '<tt>fdiv</tt>' instruction must be
2984 <a href="#t_floating">floating point</a> or <a href="#t_vector">vector</a> of
2985 floating point values. Both arguments must have identical types.</p>
2988 <p>The value produced is the floating point quotient of the two operands.</p>
2992 <result> = fdiv float 4.0, %var <i>; yields {float}:result = 4.0 / %var</i>
2997 <!-- _______________________________________________________________________ -->
2998 <div class="doc_subsubsection"> <a name="i_urem">'<tt>urem</tt>' Instruction</a>
3001 <div class="doc_text">
3005 <result> = urem <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3009 <p>The '<tt>urem</tt>' instruction returns the remainder from the unsigned
3010 division of its two arguments.</p>
3013 <p>The two arguments to the '<tt>urem</tt>' instruction must be
3014 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
3015 values. Both arguments must have identical types.</p>
3018 <p>This instruction returns the unsigned integer <i>remainder</i> of a division.
3019 This instruction always performs an unsigned division to get the
3022 <p>Note that unsigned integer remainder and signed integer remainder are
3023 distinct operations; for signed integer remainder, use '<tt>srem</tt>'.</p>
3025 <p>Taking the remainder of a division by zero leads to undefined behavior.</p>
3029 <result> = urem i32 4, %var <i>; yields {i32}:result = 4 % %var</i>
3034 <!-- _______________________________________________________________________ -->
3035 <div class="doc_subsubsection">
3036 <a name="i_srem">'<tt>srem</tt>' Instruction</a>
3039 <div class="doc_text">
3043 <result> = srem <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3047 <p>The '<tt>srem</tt>' instruction returns the remainder from the signed
3048 division of its two operands. This instruction can also take
3049 <a href="#t_vector">vector</a> versions of the values in which case the
3050 elements must be integers.</p>
3053 <p>The two arguments to the '<tt>srem</tt>' instruction must be
3054 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
3055 values. Both arguments must have identical types.</p>
3058 <p>This instruction returns the <i>remainder</i> of a division (where the result
3059 has the same sign as the dividend, <tt>op1</tt>), not the <i>modulo</i>
3060 operator (where the result has the same sign as the divisor, <tt>op2</tt>) of
3061 a value. For more information about the difference,
3062 see <a href="http://mathforum.org/dr.math/problems/anne.4.28.99.html">The
3063 Math Forum</a>. For a table of how this is implemented in various languages,
3064 please see <a href="http://en.wikipedia.org/wiki/Modulo_operation">
3065 Wikipedia: modulo operation</a>.</p>
3067 <p>Note that signed integer remainder and unsigned integer remainder are
3068 distinct operations; for unsigned integer remainder, use '<tt>urem</tt>'.</p>
3070 <p>Taking the remainder of a division by zero leads to undefined behavior.
3071 Overflow also leads to undefined behavior; this is a rare case, but can
3072 occur, for example, by taking the remainder of a 32-bit division of
3073 -2147483648 by -1. (The remainder doesn't actually overflow, but this rule
3074 lets srem be implemented using instructions that return both the result of
3075 the division and the remainder.)</p>
3079 <result> = srem i32 4, %var <i>; yields {i32}:result = 4 % %var</i>
3084 <!-- _______________________________________________________________________ -->
3085 <div class="doc_subsubsection">
3086 <a name="i_frem">'<tt>frem</tt>' Instruction</a> </div>
3088 <div class="doc_text">
3092 <result> = frem <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3096 <p>The '<tt>frem</tt>' instruction returns the remainder from the division of
3097 its two operands.</p>
3100 <p>The two arguments to the '<tt>frem</tt>' instruction must be
3101 <a href="#t_floating">floating point</a> or <a href="#t_vector">vector</a> of
3102 floating point values. Both arguments must have identical types.</p>
3105 <p>This instruction returns the <i>remainder</i> of a division. The remainder
3106 has the same sign as the dividend.</p>
3110 <result> = frem float 4.0, %var <i>; yields {float}:result = 4.0 % %var</i>
3115 <!-- ======================================================================= -->
3116 <div class="doc_subsection"> <a name="bitwiseops">Bitwise Binary
3117 Operations</a> </div>
3119 <div class="doc_text">
3121 <p>Bitwise binary operators are used to do various forms of bit-twiddling in a
3122 program. They are generally very efficient instructions and can commonly be
3123 strength reduced from other instructions. They require two operands of the
3124 same type, execute an operation on them, and produce a single value. The
3125 resulting value is the same type as its operands.</p>
3129 <!-- _______________________________________________________________________ -->
3130 <div class="doc_subsubsection"> <a name="i_shl">'<tt>shl</tt>'
3131 Instruction</a> </div>
3133 <div class="doc_text">
3137 <result> = shl <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3141 <p>The '<tt>shl</tt>' instruction returns the first operand shifted to the left
3142 a specified number of bits.</p>
3145 <p>Both arguments to the '<tt>shl</tt>' instruction must be the
3146 same <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of
3147 integer type. '<tt>op2</tt>' is treated as an unsigned value.</p>
3150 <p>The value produced is <tt>op1</tt> * 2<sup><tt>op2</tt></sup> mod
3151 2<sup>n</sup>, where <tt>n</tt> is the width of the result. If <tt>op2</tt>
3152 is (statically or dynamically) negative or equal to or larger than the number
3153 of bits in <tt>op1</tt>, the result is undefined. If the arguments are
3154 vectors, each vector element of <tt>op1</tt> is shifted by the corresponding
3155 shift amount in <tt>op2</tt>.</p>
3159 <result> = shl i32 4, %var <i>; yields {i32}: 4 << %var</i>
3160 <result> = shl i32 4, 2 <i>; yields {i32}: 16</i>
3161 <result> = shl i32 1, 10 <i>; yields {i32}: 1024</i>
3162 <result> = shl i32 1, 32 <i>; undefined</i>
3163 <result> = shl <2 x i32> < i32 1, i32 1>, < i32 1, i32 2> <i>; yields: result=<2 x i32> < i32 2, i32 4></i>
3168 <!-- _______________________________________________________________________ -->
3169 <div class="doc_subsubsection"> <a name="i_lshr">'<tt>lshr</tt>'
3170 Instruction</a> </div>
3172 <div class="doc_text">
3176 <result> = lshr <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3180 <p>The '<tt>lshr</tt>' instruction (logical shift right) returns the first
3181 operand shifted to the right a specified number of bits with zero fill.</p>
3184 <p>Both arguments to the '<tt>lshr</tt>' instruction must be the same
3185 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
3186 type. '<tt>op2</tt>' is treated as an unsigned value.</p>
3189 <p>This instruction always performs a logical shift right operation. The most
3190 significant bits of the result will be filled with zero bits after the shift.
3191 If <tt>op2</tt> is (statically or dynamically) equal to or larger than the
3192 number of bits in <tt>op1</tt>, the result is undefined. If the arguments are
3193 vectors, each vector element of <tt>op1</tt> is shifted by the corresponding
3194 shift amount in <tt>op2</tt>.</p>
3198 <result> = lshr i32 4, 1 <i>; yields {i32}:result = 2</i>
3199 <result> = lshr i32 4, 2 <i>; yields {i32}:result = 1</i>
3200 <result> = lshr i8 4, 3 <i>; yields {i8}:result = 0</i>
3201 <result> = lshr i8 -2, 1 <i>; yields {i8}:result = 0x7FFFFFFF </i>
3202 <result> = lshr i32 1, 32 <i>; undefined</i>
3203 <result> = lshr <2 x i32> < i32 -2, i32 4>, < i32 1, i32 2> <i>; yields: result=<2 x i32> < i32 0x7FFFFFFF, i32 1></i>
3208 <!-- _______________________________________________________________________ -->
3209 <div class="doc_subsubsection"> <a name="i_ashr">'<tt>ashr</tt>'
3210 Instruction</a> </div>
3211 <div class="doc_text">
3215 <result> = ashr <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3219 <p>The '<tt>ashr</tt>' instruction (arithmetic shift right) returns the first
3220 operand shifted to the right a specified number of bits with sign
3224 <p>Both arguments to the '<tt>ashr</tt>' instruction must be the same
3225 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
3226 type. '<tt>op2</tt>' is treated as an unsigned value.</p>
3229 <p>This instruction always performs an arithmetic shift right operation, The
3230 most significant bits of the result will be filled with the sign bit
3231 of <tt>op1</tt>. If <tt>op2</tt> is (statically or dynamically) equal to or
3232 larger than the number of bits in <tt>op1</tt>, the result is undefined. If
3233 the arguments are vectors, each vector element of <tt>op1</tt> is shifted by
3234 the corresponding shift amount in <tt>op2</tt>.</p>
3238 <result> = ashr i32 4, 1 <i>; yields {i32}:result = 2</i>
3239 <result> = ashr i32 4, 2 <i>; yields {i32}:result = 1</i>
3240 <result> = ashr i8 4, 3 <i>; yields {i8}:result = 0</i>
3241 <result> = ashr i8 -2, 1 <i>; yields {i8}:result = -1</i>
3242 <result> = ashr i32 1, 32 <i>; undefined</i>
3243 <result> = ashr <2 x i32> < i32 -2, i32 4>, < i32 1, i32 3> <i>; yields: result=<2 x i32> < i32 -1, i32 0></i>
3248 <!-- _______________________________________________________________________ -->
3249 <div class="doc_subsubsection"> <a name="i_and">'<tt>and</tt>'
3250 Instruction</a> </div>
3252 <div class="doc_text">
3256 <result> = and <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3260 <p>The '<tt>and</tt>' instruction returns the bitwise logical and of its two
3264 <p>The two arguments to the '<tt>and</tt>' instruction must be
3265 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
3266 values. Both arguments must have identical types.</p>
3269 <p>The truth table used for the '<tt>and</tt>' instruction is:</p>
3271 <table border="1" cellspacing="0" cellpadding="4">
3303 <result> = and i32 4, %var <i>; yields {i32}:result = 4 & %var</i>
3304 <result> = and i32 15, 40 <i>; yields {i32}:result = 8</i>
3305 <result> = and i32 4, 8 <i>; yields {i32}:result = 0</i>
3308 <!-- _______________________________________________________________________ -->
3309 <div class="doc_subsubsection"> <a name="i_or">'<tt>or</tt>' Instruction</a> </div>
3311 <div class="doc_text">
3315 <result> = or <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3319 <p>The '<tt>or</tt>' instruction returns the bitwise logical inclusive or of its
3323 <p>The two arguments to the '<tt>or</tt>' instruction must be
3324 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
3325 values. Both arguments must have identical types.</p>
3328 <p>The truth table used for the '<tt>or</tt>' instruction is:</p>
3330 <table border="1" cellspacing="0" cellpadding="4">
3362 <result> = or i32 4, %var <i>; yields {i32}:result = 4 | %var</i>
3363 <result> = or i32 15, 40 <i>; yields {i32}:result = 47</i>
3364 <result> = or i32 4, 8 <i>; yields {i32}:result = 12</i>
3369 <!-- _______________________________________________________________________ -->
3370 <div class="doc_subsubsection"> <a name="i_xor">'<tt>xor</tt>'
3371 Instruction</a> </div>
3373 <div class="doc_text">
3377 <result> = xor <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3381 <p>The '<tt>xor</tt>' instruction returns the bitwise logical exclusive or of
3382 its two operands. The <tt>xor</tt> is used to implement the "one's
3383 complement" operation, which is the "~" operator in C.</p>
3386 <p>The two arguments to the '<tt>xor</tt>' instruction must be
3387 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
3388 values. Both arguments must have identical types.</p>
3391 <p>The truth table used for the '<tt>xor</tt>' instruction is:</p>
3393 <table border="1" cellspacing="0" cellpadding="4">
3425 <result> = xor i32 4, %var <i>; yields {i32}:result = 4 ^ %var</i>
3426 <result> = xor i32 15, 40 <i>; yields {i32}:result = 39</i>
3427 <result> = xor i32 4, 8 <i>; yields {i32}:result = 12</i>
3428 <result> = xor i32 %V, -1 <i>; yields {i32}:result = ~%V</i>
3433 <!-- ======================================================================= -->
3434 <div class="doc_subsection">
3435 <a name="vectorops">Vector Operations</a>
3438 <div class="doc_text">
3440 <p>LLVM supports several instructions to represent vector operations in a
3441 target-independent manner. These instructions cover the element-access and
3442 vector-specific operations needed to process vectors effectively. While LLVM
3443 does directly support these vector operations, many sophisticated algorithms
3444 will want to use target-specific intrinsics to take full advantage of a
3445 specific target.</p>
3449 <!-- _______________________________________________________________________ -->
3450 <div class="doc_subsubsection">
3451 <a name="i_extractelement">'<tt>extractelement</tt>' Instruction</a>
3454 <div class="doc_text">
3458 <result> = extractelement <n x <ty>> <val>, i32 <idx> <i>; yields <ty></i>
3462 <p>The '<tt>extractelement</tt>' instruction extracts a single scalar element
3463 from a vector at a specified index.</p>
3467 <p>The first operand of an '<tt>extractelement</tt>' instruction is a value
3468 of <a href="#t_vector">vector</a> type. The second operand is an index
3469 indicating the position from which to extract the element. The index may be
3473 <p>The result is a scalar of the same type as the element type of
3474 <tt>val</tt>. Its value is the value at position <tt>idx</tt> of
3475 <tt>val</tt>. If <tt>idx</tt> exceeds the length of <tt>val</tt>, the
3476 results are undefined.</p>
3480 %result = extractelement <4 x i32> %vec, i32 0 <i>; yields i32</i>
3485 <!-- _______________________________________________________________________ -->
3486 <div class="doc_subsubsection">
3487 <a name="i_insertelement">'<tt>insertelement</tt>' Instruction</a>
3490 <div class="doc_text">
3494 <result> = insertelement <n x <ty>> <val>, <ty> <elt>, i32 <idx> <i>; yields <n x <ty>></i>
3498 <p>The '<tt>insertelement</tt>' instruction inserts a scalar element into a
3499 vector at a specified index.</p>
3502 <p>The first operand of an '<tt>insertelement</tt>' instruction is a value
3503 of <a href="#t_vector">vector</a> type. The second operand is a scalar value
3504 whose type must equal the element type of the first operand. The third
3505 operand is an index indicating the position at which to insert the value.
3506 The index may be a variable.</p>
3509 <p>The result is a vector of the same type as <tt>val</tt>. Its element values
3510 are those of <tt>val</tt> except at position <tt>idx</tt>, where it gets the
3511 value <tt>elt</tt>. If <tt>idx</tt> exceeds the length of <tt>val</tt>, the
3512 results are undefined.</p>
3516 %result = insertelement <4 x i32> %vec, i32 1, i32 0 <i>; yields <4 x i32></i>
3521 <!-- _______________________________________________________________________ -->
3522 <div class="doc_subsubsection">
3523 <a name="i_shufflevector">'<tt>shufflevector</tt>' Instruction</a>
3526 <div class="doc_text">
3530 <result> = shufflevector <n x <ty>> <v1>, <n x <ty>> <v2>, <m x i32> <mask> <i>; yields <m x <ty>></i>
3534 <p>The '<tt>shufflevector</tt>' instruction constructs a permutation of elements
3535 from two input vectors, returning a vector with the same element type as the
3536 input and length that is the same as the shuffle mask.</p>
3539 <p>The first two operands of a '<tt>shufflevector</tt>' instruction are vectors
3540 with types that match each other. The third argument is a shuffle mask whose
3541 element type is always 'i32'. The result of the instruction is a vector
3542 whose length is the same as the shuffle mask and whose element type is the
3543 same as the element type of the first two operands.</p>
3545 <p>The shuffle mask operand is required to be a constant vector with either
3546 constant integer or undef values.</p>
3549 <p>The elements of the two input vectors are numbered from left to right across
3550 both of the vectors. The shuffle mask operand specifies, for each element of
3551 the result vector, which element of the two input vectors the result element
3552 gets. The element selector may be undef (meaning "don't care") and the
3553 second operand may be undef if performing a shuffle from only one vector.</p>
3557 %result = shufflevector <4 x i32> %v1, <4 x i32> %v2,
3558 <4 x i32> <i32 0, i32 4, i32 1, i32 5> <i>; yields <4 x i32></i>
3559 %result = shufflevector <4 x i32> %v1, <4 x i32> undef,
3560 <4 x i32> <i32 0, i32 1, i32 2, i32 3> <i>; yields <4 x i32></i> - Identity shuffle.
3561 %result = shufflevector <8 x i32> %v1, <8 x i32> undef,
3562 <4 x i32> <i32 0, i32 1, i32 2, i32 3> <i>; yields <4 x i32></i>
3563 %result = shufflevector <4 x i32> %v1, <4 x i32> %v2,
3564 <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>
3569 <!-- ======================================================================= -->
3570 <div class="doc_subsection">
3571 <a name="aggregateops">Aggregate Operations</a>
3574 <div class="doc_text">
3576 <p>LLVM supports several instructions for working with aggregate values.</p>
3580 <!-- _______________________________________________________________________ -->
3581 <div class="doc_subsubsection">
3582 <a name="i_extractvalue">'<tt>extractvalue</tt>' Instruction</a>
3585 <div class="doc_text">
3589 <result> = extractvalue <aggregate type> <val>, <idx>{, <idx>}*
3593 <p>The '<tt>extractvalue</tt>' instruction extracts the value of a struct field
3594 or array element from an aggregate value.</p>
3597 <p>The first operand of an '<tt>extractvalue</tt>' instruction is a value
3598 of <a href="#t_struct">struct</a> or <a href="#t_array">array</a> type. The
3599 operands are constant indices to specify which value to extract in a similar
3600 manner as indices in a
3601 '<tt><a href="#i_getelementptr">getelementptr</a></tt>' instruction.</p>
3604 <p>The result is the value at the position in the aggregate specified by the
3609 %result = extractvalue {i32, float} %agg, 0 <i>; yields i32</i>
3614 <!-- _______________________________________________________________________ -->
3615 <div class="doc_subsubsection">
3616 <a name="i_insertvalue">'<tt>insertvalue</tt>' Instruction</a>
3619 <div class="doc_text">
3623 <result> = insertvalue <aggregate type> <val>, <ty> <val>, <idx> <i>; yields <n x <ty>></i>
3627 <p>The '<tt>insertvalue</tt>' instruction inserts a value into a struct field or
3628 array element in an aggregate.</p>
3632 <p>The first operand of an '<tt>insertvalue</tt>' instruction is a value
3633 of <a href="#t_struct">struct</a> or <a href="#t_array">array</a> type. The
3634 second operand is a first-class value to insert. The following operands are
3635 constant indices indicating the position at which to insert the value in a
3636 similar manner as indices in a
3637 '<tt><a href="#i_getelementptr">getelementptr</a></tt>' instruction. The
3638 value to insert must have the same type as the value identified by the
3642 <p>The result is an aggregate of the same type as <tt>val</tt>. Its value is
3643 that of <tt>val</tt> except that the value at the position specified by the
3644 indices is that of <tt>elt</tt>.</p>
3648 %result = insertvalue {i32, float} %agg, i32 1, 0 <i>; yields {i32, float}</i>
3654 <!-- ======================================================================= -->
3655 <div class="doc_subsection">
3656 <a name="memoryops">Memory Access and Addressing Operations</a>
3659 <div class="doc_text">
3661 <p>A key design point of an SSA-based representation is how it represents
3662 memory. In LLVM, no memory locations are in SSA form, which makes things
3663 very simple. This section describes how to read, write, allocate, and free
3668 <!-- _______________________________________________________________________ -->
3669 <div class="doc_subsubsection">
3670 <a name="i_malloc">'<tt>malloc</tt>' Instruction</a>
3673 <div class="doc_text">
3677 <result> = malloc <type>[, i32 <NumElements>][, align <alignment>] <i>; yields {type*}:result</i>
3681 <p>The '<tt>malloc</tt>' instruction allocates memory from the system heap and
3682 returns a pointer to it. The object is always allocated in the generic
3683 address space (address space zero).</p>
3686 <p>The '<tt>malloc</tt>' instruction allocates
3687 <tt>sizeof(<type>)*NumElements</tt> bytes of memory from the operating
3688 system and returns a pointer of the appropriate type to the program. If
3689 "NumElements" is specified, it is the number of elements allocated, otherwise
3690 "NumElements" is defaulted to be one. If a constant alignment is specified,
3691 the value result of the allocation is guaranteed to be aligned to at least
3692 that boundary. If not specified, or if zero, the target can choose to align
3693 the allocation on any convenient boundary compatible with the type.</p>
3695 <p>'<tt>type</tt>' must be a sized type.</p>
3698 <p>Memory is allocated using the system "<tt>malloc</tt>" function, and a
3699 pointer is returned. The result of a zero byte allocation is undefined. The
3700 result is null if there is insufficient memory available.</p>
3704 %array = malloc [4 x i8] <i>; yields {[%4 x i8]*}:array</i>
3706 %size = <a href="#i_add">add</a> i32 2, 2 <i>; yields {i32}:size = i32 4</i>
3707 %array1 = malloc i8, i32 4 <i>; yields {i8*}:array1</i>
3708 %array2 = malloc [12 x i8], i32 %size <i>; yields {[12 x i8]*}:array2</i>
3709 %array3 = malloc i32, i32 4, align 1024 <i>; yields {i32*}:array3</i>
3710 %array4 = malloc i32, align 1024 <i>; yields {i32*}:array4</i>
3713 <p>Note that the code generator does not yet respect the alignment value.</p>
3717 <!-- _______________________________________________________________________ -->
3718 <div class="doc_subsubsection">
3719 <a name="i_free">'<tt>free</tt>' Instruction</a>
3722 <div class="doc_text">
3726 free <type> <value> <i>; yields {void}</i>
3730 <p>The '<tt>free</tt>' instruction returns memory back to the unused memory heap
3731 to be reallocated in the future.</p>
3734 <p>'<tt>value</tt>' shall be a pointer value that points to a value that was
3735 allocated with the '<tt><a href="#i_malloc">malloc</a></tt>' instruction.</p>
3738 <p>Access to the memory pointed to by the pointer is no longer defined after
3739 this instruction executes. If the pointer is null, the operation is a
3744 %array = <a href="#i_malloc">malloc</a> [4 x i8] <i>; yields {[4 x i8]*}:array</i>
3745 free [4 x i8]* %array
3750 <!-- _______________________________________________________________________ -->
3751 <div class="doc_subsubsection">
3752 <a name="i_alloca">'<tt>alloca</tt>' Instruction</a>
3755 <div class="doc_text">
3759 <result> = alloca <type>[, i32 <NumElements>][, align <alignment>] <i>; yields {type*}:result</i>
3763 <p>The '<tt>alloca</tt>' instruction allocates memory on the stack frame of the
3764 currently executing function, to be automatically released when this function
3765 returns to its caller. The object is always allocated in the generic address
3766 space (address space zero).</p>
3769 <p>The '<tt>alloca</tt>' instruction
3770 allocates <tt>sizeof(<type>)*NumElements</tt> bytes of memory on the
3771 runtime stack, returning a pointer of the appropriate type to the program.
3772 If "NumElements" is specified, it is the number of elements allocated,
3773 otherwise "NumElements" is defaulted to be one. If a constant alignment is
3774 specified, the value result of the allocation is guaranteed to be aligned to
3775 at least that boundary. If not specified, or if zero, the target can choose
3776 to align the allocation on any convenient boundary compatible with the
3779 <p>'<tt>type</tt>' may be any sized type.</p>
3782 <p>Memory is allocated; a pointer is returned. The operation is undefined if
3783 there is insufficient stack space for the allocation. '<tt>alloca</tt>'d
3784 memory is automatically released when the function returns. The
3785 '<tt>alloca</tt>' instruction is commonly used to represent automatic
3786 variables that must have an address available. When the function returns
3787 (either with the <tt><a href="#i_ret">ret</a></tt>
3788 or <tt><a href="#i_unwind">unwind</a></tt> instructions), the memory is
3789 reclaimed. Allocating zero bytes is legal, but the result is undefined.</p>
3793 %ptr = alloca i32 <i>; yields {i32*}:ptr</i>
3794 %ptr = alloca i32, i32 4 <i>; yields {i32*}:ptr</i>
3795 %ptr = alloca i32, i32 4, align 1024 <i>; yields {i32*}:ptr</i>
3796 %ptr = alloca i32, align 1024 <i>; yields {i32*}:ptr</i>
3801 <!-- _______________________________________________________________________ -->
3802 <div class="doc_subsubsection"> <a name="i_load">'<tt>load</tt>'
3803 Instruction</a> </div>
3805 <div class="doc_text">
3809 <result> = load <ty>* <pointer>[, align <alignment>]
3810 <result> = volatile load <ty>* <pointer>[, align <alignment>]
3814 <p>The '<tt>load</tt>' instruction is used to read from memory.</p>
3817 <p>The argument to the '<tt>load</tt>' instruction specifies the memory address
3818 from which to load. The pointer must point to
3819 a <a href="#t_firstclass">first class</a> type. If the <tt>load</tt> is
3820 marked as <tt>volatile</tt>, then the optimizer is not allowed to modify the
3821 number or order of execution of this <tt>load</tt> with other
3822 volatile <tt>load</tt> and <tt><a href="#i_store">store</a></tt>
3825 <p>The optional constant "align" argument specifies the alignment of the
3826 operation (that is, the alignment of the memory address). A value of 0 or an
3827 omitted "align" argument means that the operation has the preferential
3828 alignment for the target. It is the responsibility of the code emitter to
3829 ensure that the alignment information is correct. Overestimating the
3830 alignment results in an undefined behavior. Underestimating the alignment may
3831 produce less efficient code. An alignment of 1 is always safe.</p>
3834 <p>The location of memory pointed to is loaded. If the value being loaded is of
3835 scalar type then the number of bytes read does not exceed the minimum number
3836 of bytes needed to hold all bits of the type. For example, loading an
3837 <tt>i24</tt> reads at most three bytes. When loading a value of a type like
3838 <tt>i20</tt> with a size that is not an integral number of bytes, the result
3839 is undefined if the value was not originally written using a store of the
3844 %ptr = <a href="#i_alloca">alloca</a> i32 <i>; yields {i32*}:ptr</i>
3845 <a href="#i_store">store</a> i32 3, i32* %ptr <i>; yields {void}</i>
3846 %val = load i32* %ptr <i>; yields {i32}:val = i32 3</i>
3851 <!-- _______________________________________________________________________ -->
3852 <div class="doc_subsubsection"> <a name="i_store">'<tt>store</tt>'
3853 Instruction</a> </div>
3855 <div class="doc_text">
3859 store <ty> <value>, <ty>* <pointer>[, align <alignment>] <i>; yields {void}</i>
3860 volatile store <ty> <value>, <ty>* <pointer>[, align <alignment>] <i>; yields {void}</i>
3864 <p>The '<tt>store</tt>' instruction is used to write to memory.</p>
3867 <p>There are two arguments to the '<tt>store</tt>' instruction: a value to store
3868 and an address at which to store it. The type of the
3869 '<tt><pointer></tt>' operand must be a pointer to
3870 the <a href="#t_firstclass">first class</a> type of the
3871 '<tt><value></tt>' operand. If the <tt>store</tt> is marked
3872 as <tt>volatile</tt>, then the optimizer is not allowed to modify the number
3873 or order of execution of this <tt>store</tt> with other
3874 volatile <tt>load</tt> and <tt><a href="#i_store">store</a></tt>
3877 <p>The optional constant "align" argument specifies the alignment of the
3878 operation (that is, the alignment of the memory address). A value of 0 or an
3879 omitted "align" argument means that the operation has the preferential
3880 alignment for the target. It is the responsibility of the code emitter to
3881 ensure that the alignment information is correct. Overestimating the
3882 alignment results in an undefined behavior. Underestimating the alignment may
3883 produce less efficient code. An alignment of 1 is always safe.</p>
3886 <p>The contents of memory are updated to contain '<tt><value></tt>' at the
3887 location specified by the '<tt><pointer></tt>' operand. If
3888 '<tt><value></tt>' is of scalar type then the number of bytes written
3889 does not exceed the minimum number of bytes needed to hold all bits of the
3890 type. For example, storing an <tt>i24</tt> writes at most three bytes. When
3891 writing a value of a type like <tt>i20</tt> with a size that is not an
3892 integral number of bytes, it is unspecified what happens to the extra bits
3893 that do not belong to the type, but they will typically be overwritten.</p>
3897 %ptr = <a href="#i_alloca">alloca</a> i32 <i>; yields {i32*}:ptr</i>
3898 store i32 3, i32* %ptr <i>; yields {void}</i>
3899 %val = <a href="#i_load">load</a> i32* %ptr <i>; yields {i32}:val = i32 3</i>
3904 <!-- _______________________________________________________________________ -->
3905 <div class="doc_subsubsection">
3906 <a name="i_getelementptr">'<tt>getelementptr</tt>' Instruction</a>
3909 <div class="doc_text">
3913 <result> = getelementptr <pty>* <ptrval>{, <ty> <idx>}*
3914 <result> = getelementptr inbounds <pty>* <ptrval>{, <ty> <idx>}*
3918 <p>The '<tt>getelementptr</tt>' instruction is used to get the address of a
3919 subelement of an aggregate data structure. It performs address calculation
3920 only and does not access memory.</p>
3923 <p>The first argument is always a pointer, and forms the basis of the
3924 calculation. The remaining arguments are indices that indicate which of the
3925 elements of the aggregate object are indexed. The interpretation of each
3926 index is dependent on the type being indexed into. The first index always
3927 indexes the pointer value given as the first argument, the second index
3928 indexes a value of the type pointed to (not necessarily the value directly
3929 pointed to, since the first index can be non-zero), etc. The first type
3930 indexed into must be a pointer value, subsequent types can be arrays, vectors
3931 and structs. Note that subsequent types being indexed into can never be
3932 pointers, since that would require loading the pointer before continuing
3935 <p>The type of each index argument depends on the type it is indexing into.
3936 When indexing into a (optionally packed) structure, only <tt>i32</tt> integer
3937 <b>constants</b> are allowed. When indexing into an array, pointer or
3938 vector, integers of any width are allowed, and they are not required to be
3941 <p>For example, let's consider a C code fragment and how it gets compiled to
3944 <div class="doc_code">
3957 int *foo(struct ST *s) {
3958 return &s[1].Z.B[5][13];
3963 <p>The LLVM code generated by the GCC frontend is:</p>
3965 <div class="doc_code">
3967 %RT = <a href="#namedtypes">type</a> { i8 , [10 x [20 x i32]], i8 }
3968 %ST = <a href="#namedtypes">type</a> { i32, double, %RT }
3970 define i32* @foo(%ST* %s) {
3972 %reg = getelementptr %ST* %s, i32 1, i32 2, i32 1, i32 5, i32 13
3979 <p>In the example above, the first index is indexing into the '<tt>%ST*</tt>'
3980 type, which is a pointer, yielding a '<tt>%ST</tt>' = '<tt>{ i32, double, %RT
3981 }</tt>' type, a structure. The second index indexes into the third element
3982 of the structure, yielding a '<tt>%RT</tt>' = '<tt>{ i8 , [10 x [20 x i32]],
3983 i8 }</tt>' type, another structure. The third index indexes into the second
3984 element of the structure, yielding a '<tt>[10 x [20 x i32]]</tt>' type, an
3985 array. The two dimensions of the array are subscripted into, yielding an
3986 '<tt>i32</tt>' type. The '<tt>getelementptr</tt>' instruction returns a
3987 pointer to this element, thus computing a value of '<tt>i32*</tt>' type.</p>
3989 <p>Note that it is perfectly legal to index partially through a structure,
3990 returning a pointer to an inner element. Because of this, the LLVM code for
3991 the given testcase is equivalent to:</p>
3994 define i32* @foo(%ST* %s) {
3995 %t1 = getelementptr %ST* %s, i32 1 <i>; yields %ST*:%t1</i>
3996 %t2 = getelementptr %ST* %t1, i32 0, i32 2 <i>; yields %RT*:%t2</i>
3997 %t3 = getelementptr %RT* %t2, i32 0, i32 1 <i>; yields [10 x [20 x i32]]*:%t3</i>
3998 %t4 = getelementptr [10 x [20 x i32]]* %t3, i32 0, i32 5 <i>; yields [20 x i32]*:%t4</i>
3999 %t5 = getelementptr [20 x i32]* %t4, i32 0, i32 13 <i>; yields i32*:%t5</i>
4004 <p>If the <tt>inbounds</tt> keyword is present, the result value of the
4005 <tt>getelementptr</tt> is undefined if the base pointer is not an
4006 <i>in bounds</i> address of an allocated object, or if any of the addresses
4007 formed by successive addition of the offsets implied by the indices to
4008 the base address are not an <i>in bounds</i> address of that allocated
4010 The <i>in bounds</i> addresses for an allocated object are all the addresses
4011 that point into the object, plus the address one past the end.</p>
4013 <p>If the <tt>inbounds</tt> keyword is not present, the offsets are added to
4014 the base address with silently-wrapping two's complement arithmetic, and
4015 the result value of the <tt>getelementptr</tt> may be outside the object
4016 pointed to by the base pointer. The result value may not necessarily be
4017 used to access memory though, even if it happens to point into allocated
4018 storage. See the <a href="#pointeraliasing">Pointer Aliasing Rules</a>
4019 section for more information.</p>
4021 <p>The getelementptr instruction is often confusing. For some more insight into
4022 how it works, see <a href="GetElementPtr.html">the getelementptr FAQ</a>.</p>
4026 <i>; yields [12 x i8]*:aptr</i>
4027 %aptr = getelementptr {i32, [12 x i8]}* %saptr, i64 0, i32 1
4028 <i>; yields i8*:vptr</i>
4029 %vptr = getelementptr {i32, <2 x i8>}* %svptr, i64 0, i32 1, i32 1
4030 <i>; yields i8*:eptr</i>
4031 %eptr = getelementptr [12 x i8]* %aptr, i64 0, i32 1
4032 <i>; yields i32*:iptr</i>
4033 %iptr = getelementptr [10 x i32]* @arr, i16 0, i16 0
4038 <!-- ======================================================================= -->
4039 <div class="doc_subsection"> <a name="convertops">Conversion Operations</a>
4042 <div class="doc_text">
4044 <p>The instructions in this category are the conversion instructions (casting)
4045 which all take a single operand and a type. They perform various bit
4046 conversions on the operand.</p>
4050 <!-- _______________________________________________________________________ -->
4051 <div class="doc_subsubsection">
4052 <a name="i_trunc">'<tt>trunc .. to</tt>' Instruction</a>
4054 <div class="doc_text">
4058 <result> = trunc <ty> <value> to <ty2> <i>; yields ty2</i>
4062 <p>The '<tt>trunc</tt>' instruction truncates its operand to the
4063 type <tt>ty2</tt>.</p>
4066 <p>The '<tt>trunc</tt>' instruction takes a <tt>value</tt> to trunc, which must
4067 be an <a href="#t_integer">integer</a> type, and a type that specifies the
4068 size and type of the result, which must be
4069 an <a href="#t_integer">integer</a> type. The bit size of <tt>value</tt> must
4070 be larger than the bit size of <tt>ty2</tt>. Equal sized types are not
4074 <p>The '<tt>trunc</tt>' instruction truncates the high order bits
4075 in <tt>value</tt> and converts the remaining bits to <tt>ty2</tt>. Since the
4076 source size must be larger than the destination size, <tt>trunc</tt> cannot
4077 be a <i>no-op cast</i>. It will always truncate bits.</p>
4081 %X = trunc i32 257 to i8 <i>; yields i8:1</i>
4082 %Y = trunc i32 123 to i1 <i>; yields i1:true</i>
4083 %Y = trunc i32 122 to i1 <i>; yields i1:false</i>
4088 <!-- _______________________________________________________________________ -->
4089 <div class="doc_subsubsection">
4090 <a name="i_zext">'<tt>zext .. to</tt>' Instruction</a>
4092 <div class="doc_text">
4096 <result> = zext <ty> <value> to <ty2> <i>; yields ty2</i>
4100 <p>The '<tt>zext</tt>' instruction zero extends its operand to type
4105 <p>The '<tt>zext</tt>' instruction takes a value to cast, which must be of
4106 <a href="#t_integer">integer</a> type, and a type to cast it to, which must
4107 also be of <a href="#t_integer">integer</a> type. The bit size of the
4108 <tt>value</tt> must be smaller than the bit size of the destination type,
4112 <p>The <tt>zext</tt> fills the high order bits of the <tt>value</tt> with zero
4113 bits until it reaches the size of the destination type, <tt>ty2</tt>.</p>
4115 <p>When zero extending from i1, the result will always be either 0 or 1.</p>
4119 %X = zext i32 257 to i64 <i>; yields i64:257</i>
4120 %Y = zext i1 true to i32 <i>; yields i32:1</i>
4125 <!-- _______________________________________________________________________ -->
4126 <div class="doc_subsubsection">
4127 <a name="i_sext">'<tt>sext .. to</tt>' Instruction</a>
4129 <div class="doc_text">
4133 <result> = sext <ty> <value> to <ty2> <i>; yields ty2</i>
4137 <p>The '<tt>sext</tt>' sign extends <tt>value</tt> to the type <tt>ty2</tt>.</p>
4140 <p>The '<tt>sext</tt>' instruction takes a value to cast, which must be of
4141 <a href="#t_integer">integer</a> type, and a type to cast it to, which must
4142 also be of <a href="#t_integer">integer</a> type. The bit size of the
4143 <tt>value</tt> must be smaller than the bit size of the destination type,
4147 <p>The '<tt>sext</tt>' instruction performs a sign extension by copying the sign
4148 bit (highest order bit) of the <tt>value</tt> until it reaches the bit size
4149 of the type <tt>ty2</tt>.</p>
4151 <p>When sign extending from i1, the extension always results in -1 or 0.</p>
4155 %X = sext i8 -1 to i16 <i>; yields i16 :65535</i>
4156 %Y = sext i1 true to i32 <i>; yields i32:-1</i>
4161 <!-- _______________________________________________________________________ -->
4162 <div class="doc_subsubsection">
4163 <a name="i_fptrunc">'<tt>fptrunc .. to</tt>' Instruction</a>
4166 <div class="doc_text">
4170 <result> = fptrunc <ty> <value> to <ty2> <i>; yields ty2</i>
4174 <p>The '<tt>fptrunc</tt>' instruction truncates <tt>value</tt> to type
4178 <p>The '<tt>fptrunc</tt>' instruction takes a <a href="#t_floating">floating
4179 point</a> value to cast and a <a href="#t_floating">floating point</a> type
4180 to cast it to. The size of <tt>value</tt> must be larger than the size of
4181 <tt>ty2</tt>. This implies that <tt>fptrunc</tt> cannot be used to make a
4182 <i>no-op cast</i>.</p>
4185 <p>The '<tt>fptrunc</tt>' instruction truncates a <tt>value</tt> from a larger
4186 <a href="#t_floating">floating point</a> type to a smaller
4187 <a href="#t_floating">floating point</a> type. If the value cannot fit
4188 within the destination type, <tt>ty2</tt>, then the results are
4193 %X = fptrunc double 123.0 to float <i>; yields float:123.0</i>
4194 %Y = fptrunc double 1.0E+300 to float <i>; yields undefined</i>
4199 <!-- _______________________________________________________________________ -->
4200 <div class="doc_subsubsection">
4201 <a name="i_fpext">'<tt>fpext .. to</tt>' Instruction</a>
4203 <div class="doc_text">
4207 <result> = fpext <ty> <value> to <ty2> <i>; yields ty2</i>
4211 <p>The '<tt>fpext</tt>' extends a floating point <tt>value</tt> to a larger
4212 floating point value.</p>
4215 <p>The '<tt>fpext</tt>' instruction takes a
4216 <a href="#t_floating">floating point</a> <tt>value</tt> to cast, and
4217 a <a href="#t_floating">floating point</a> type to cast it to. The source
4218 type must be smaller than the destination type.</p>
4221 <p>The '<tt>fpext</tt>' instruction extends the <tt>value</tt> from a smaller
4222 <a href="#t_floating">floating point</a> type to a larger
4223 <a href="#t_floating">floating point</a> type. The <tt>fpext</tt> cannot be
4224 used to make a <i>no-op cast</i> because it always changes bits. Use
4225 <tt>bitcast</tt> to make a <i>no-op cast</i> for a floating point cast.</p>
4229 %X = fpext float 3.1415 to double <i>; yields double:3.1415</i>
4230 %Y = fpext float 1.0 to float <i>; yields float:1.0 (no-op)</i>
4235 <!-- _______________________________________________________________________ -->
4236 <div class="doc_subsubsection">
4237 <a name="i_fptoui">'<tt>fptoui .. to</tt>' Instruction</a>
4239 <div class="doc_text">
4243 <result> = fptoui <ty> <value> to <ty2> <i>; yields ty2</i>
4247 <p>The '<tt>fptoui</tt>' converts a floating point <tt>value</tt> to its
4248 unsigned integer equivalent of type <tt>ty2</tt>.</p>
4251 <p>The '<tt>fptoui</tt>' instruction takes a value to cast, which must be a
4252 scalar or vector <a href="#t_floating">floating point</a> value, and a type
4253 to cast it to <tt>ty2</tt>, which must be an <a href="#t_integer">integer</a>
4254 type. If <tt>ty</tt> is a vector floating point type, <tt>ty2</tt> must be a
4255 vector integer type with the same number of elements as <tt>ty</tt></p>
4258 <p>The '<tt>fptoui</tt>' instruction converts its
4259 <a href="#t_floating">floating point</a> operand into the nearest (rounding
4260 towards zero) unsigned integer value. If the value cannot fit
4261 in <tt>ty2</tt>, the results are undefined.</p>
4265 %X = fptoui double 123.0 to i32 <i>; yields i32:123</i>
4266 %Y = fptoui float 1.0E+300 to i1 <i>; yields undefined:1</i>
4267 %X = fptoui float 1.04E+17 to i8 <i>; yields undefined:1</i>
4272 <!-- _______________________________________________________________________ -->
4273 <div class="doc_subsubsection">
4274 <a name="i_fptosi">'<tt>fptosi .. to</tt>' Instruction</a>
4276 <div class="doc_text">
4280 <result> = fptosi <ty> <value> to <ty2> <i>; yields ty2</i>
4284 <p>The '<tt>fptosi</tt>' instruction converts
4285 <a href="#t_floating">floating point</a> <tt>value</tt> to
4286 type <tt>ty2</tt>.</p>
4289 <p>The '<tt>fptosi</tt>' instruction takes a value to cast, which must be a
4290 scalar or vector <a href="#t_floating">floating point</a> value, and a type
4291 to cast it to <tt>ty2</tt>, which must be an <a href="#t_integer">integer</a>
4292 type. If <tt>ty</tt> is a vector floating point type, <tt>ty2</tt> must be a
4293 vector integer type with the same number of elements as <tt>ty</tt></p>
4296 <p>The '<tt>fptosi</tt>' instruction converts its
4297 <a href="#t_floating">floating point</a> operand into the nearest (rounding
4298 towards zero) signed integer value. If the value cannot fit in <tt>ty2</tt>,
4299 the results are undefined.</p>
4303 %X = fptosi double -123.0 to i32 <i>; yields i32:-123</i>
4304 %Y = fptosi float 1.0E-247 to i1 <i>; yields undefined:1</i>
4305 %X = fptosi float 1.04E+17 to i8 <i>; yields undefined:1</i>
4310 <!-- _______________________________________________________________________ -->
4311 <div class="doc_subsubsection">
4312 <a name="i_uitofp">'<tt>uitofp .. to</tt>' Instruction</a>
4314 <div class="doc_text">
4318 <result> = uitofp <ty> <value> to <ty2> <i>; yields ty2</i>
4322 <p>The '<tt>uitofp</tt>' instruction regards <tt>value</tt> as an unsigned
4323 integer and converts that value to the <tt>ty2</tt> type.</p>
4326 <p>The '<tt>uitofp</tt>' instruction takes a value to cast, which must be a
4327 scalar or vector <a href="#t_integer">integer</a> value, and a type to cast
4328 it to <tt>ty2</tt>, which must be an <a href="#t_floating">floating point</a>
4329 type. If <tt>ty</tt> is a vector integer type, <tt>ty2</tt> must be a vector
4330 floating point type with the same number of elements as <tt>ty</tt></p>
4333 <p>The '<tt>uitofp</tt>' instruction interprets its operand as an unsigned
4334 integer quantity and converts it to the corresponding floating point
4335 value. If the value cannot fit in the floating point value, the results are
4340 %X = uitofp i32 257 to float <i>; yields float:257.0</i>
4341 %Y = uitofp i8 -1 to double <i>; yields double:255.0</i>
4346 <!-- _______________________________________________________________________ -->
4347 <div class="doc_subsubsection">
4348 <a name="i_sitofp">'<tt>sitofp .. to</tt>' Instruction</a>
4350 <div class="doc_text">
4354 <result> = sitofp <ty> <value> to <ty2> <i>; yields ty2</i>
4358 <p>The '<tt>sitofp</tt>' instruction regards <tt>value</tt> as a signed integer
4359 and converts that value to the <tt>ty2</tt> type.</p>
4362 <p>The '<tt>sitofp</tt>' instruction takes a value to cast, which must be a
4363 scalar or vector <a href="#t_integer">integer</a> value, and a type to cast
4364 it to <tt>ty2</tt>, which must be an <a href="#t_floating">floating point</a>
4365 type. If <tt>ty</tt> is a vector integer type, <tt>ty2</tt> must be a vector
4366 floating point type with the same number of elements as <tt>ty</tt></p>
4369 <p>The '<tt>sitofp</tt>' instruction interprets its operand as a signed integer
4370 quantity and converts it to the corresponding floating point value. If the
4371 value cannot fit in the floating point value, the results are undefined.</p>
4375 %X = sitofp i32 257 to float <i>; yields float:257.0</i>
4376 %Y = sitofp i8 -1 to double <i>; yields double:-1.0</i>
4381 <!-- _______________________________________________________________________ -->
4382 <div class="doc_subsubsection">
4383 <a name="i_ptrtoint">'<tt>ptrtoint .. to</tt>' Instruction</a>
4385 <div class="doc_text">
4389 <result> = ptrtoint <ty> <value> to <ty2> <i>; yields ty2</i>
4393 <p>The '<tt>ptrtoint</tt>' instruction converts the pointer <tt>value</tt> to
4394 the integer type <tt>ty2</tt>.</p>
4397 <p>The '<tt>ptrtoint</tt>' instruction takes a <tt>value</tt> to cast, which
4398 must be a <a href="#t_pointer">pointer</a> value, and a type to cast it to
4399 <tt>ty2</tt>, which must be an <a href="#t_integer">integer</a> type.</p>
4402 <p>The '<tt>ptrtoint</tt>' instruction converts <tt>value</tt> to integer type
4403 <tt>ty2</tt> by interpreting the pointer value as an integer and either
4404 truncating or zero extending that value to the size of the integer type. If
4405 <tt>value</tt> is smaller than <tt>ty2</tt> then a zero extension is done. If
4406 <tt>value</tt> is larger than <tt>ty2</tt> then a truncation is done. If they
4407 are the same size, then nothing is done (<i>no-op cast</i>) other than a type
4412 %X = ptrtoint i32* %X to i8 <i>; yields truncation on 32-bit architecture</i>
4413 %Y = ptrtoint i32* %x to i64 <i>; yields zero extension on 32-bit architecture</i>
4418 <!-- _______________________________________________________________________ -->
4419 <div class="doc_subsubsection">
4420 <a name="i_inttoptr">'<tt>inttoptr .. to</tt>' Instruction</a>
4422 <div class="doc_text">
4426 <result> = inttoptr <ty> <value> to <ty2> <i>; yields ty2</i>
4430 <p>The '<tt>inttoptr</tt>' instruction converts an integer <tt>value</tt> to a
4431 pointer type, <tt>ty2</tt>.</p>
4434 <p>The '<tt>inttoptr</tt>' instruction takes an <a href="#t_integer">integer</a>
4435 value to cast, and a type to cast it to, which must be a
4436 <a href="#t_pointer">pointer</a> type.</p>
4439 <p>The '<tt>inttoptr</tt>' instruction converts <tt>value</tt> to type
4440 <tt>ty2</tt> by applying either a zero extension or a truncation depending on
4441 the size of the integer <tt>value</tt>. If <tt>value</tt> is larger than the
4442 size of a pointer then a truncation is done. If <tt>value</tt> is smaller
4443 than the size of a pointer then a zero extension is done. If they are the
4444 same size, nothing is done (<i>no-op cast</i>).</p>
4448 %X = inttoptr i32 255 to i32* <i>; yields zero extension on 64-bit architecture</i>
4449 %X = inttoptr i32 255 to i32* <i>; yields no-op on 32-bit architecture</i>
4450 %Y = inttoptr i64 0 to i32* <i>; yields truncation on 32-bit architecture</i>
4455 <!-- _______________________________________________________________________ -->
4456 <div class="doc_subsubsection">
4457 <a name="i_bitcast">'<tt>bitcast .. to</tt>' Instruction</a>
4459 <div class="doc_text">
4463 <result> = bitcast <ty> <value> to <ty2> <i>; yields ty2</i>
4467 <p>The '<tt>bitcast</tt>' instruction converts <tt>value</tt> to type
4468 <tt>ty2</tt> without changing any bits.</p>
4471 <p>The '<tt>bitcast</tt>' instruction takes a value to cast, which must be a
4472 non-aggregate first class value, and a type to cast it to, which must also be
4473 a non-aggregate <a href="#t_firstclass">first class</a> type. The bit sizes
4474 of <tt>value</tt> and the destination type, <tt>ty2</tt>, must be
4475 identical. If the source type is a pointer, the destination type must also be
4476 a pointer. This instruction supports bitwise conversion of vectors to
4477 integers and to vectors of other types (as long as they have the same
4481 <p>The '<tt>bitcast</tt>' instruction converts <tt>value</tt> to type
4482 <tt>ty2</tt>. It is always a <i>no-op cast</i> because no bits change with
4483 this conversion. The conversion is done as if the <tt>value</tt> had been
4484 stored to memory and read back as type <tt>ty2</tt>. Pointer types may only
4485 be converted to other pointer types with this instruction. To convert
4486 pointers to other types, use the <a href="#i_inttoptr">inttoptr</a> or
4487 <a href="#i_ptrtoint">ptrtoint</a> instructions first.</p>
4491 %X = bitcast i8 255 to i8 <i>; yields i8 :-1</i>
4492 %Y = bitcast i32* %x to sint* <i>; yields sint*:%x</i>
4493 %Z = bitcast <2 x int> %V to i64; <i>; yields i64: %V</i>
4498 <!-- ======================================================================= -->
4499 <div class="doc_subsection"> <a name="otherops">Other Operations</a> </div>
4501 <div class="doc_text">
4503 <p>The instructions in this category are the "miscellaneous" instructions, which
4504 defy better classification.</p>
4508 <!-- _______________________________________________________________________ -->
4509 <div class="doc_subsubsection"><a name="i_icmp">'<tt>icmp</tt>' Instruction</a>
4512 <div class="doc_text">
4516 <result> = icmp <cond> <ty> <op1>, <op2> <i>; yields {i1} or {<N x i1>}:result</i>
4520 <p>The '<tt>icmp</tt>' instruction returns a boolean value or a vector of
4521 boolean values based on comparison of its two integer, integer vector, or
4522 pointer operands.</p>
4525 <p>The '<tt>icmp</tt>' instruction takes three operands. The first operand is
4526 the condition code indicating the kind of comparison to perform. It is not a
4527 value, just a keyword. The possible condition code are:</p>
4530 <li><tt>eq</tt>: equal</li>
4531 <li><tt>ne</tt>: not equal </li>
4532 <li><tt>ugt</tt>: unsigned greater than</li>
4533 <li><tt>uge</tt>: unsigned greater or equal</li>
4534 <li><tt>ult</tt>: unsigned less than</li>
4535 <li><tt>ule</tt>: unsigned less or equal</li>
4536 <li><tt>sgt</tt>: signed greater than</li>
4537 <li><tt>sge</tt>: signed greater or equal</li>
4538 <li><tt>slt</tt>: signed less than</li>
4539 <li><tt>sle</tt>: signed less or equal</li>
4542 <p>The remaining two arguments must be <a href="#t_integer">integer</a> or
4543 <a href="#t_pointer">pointer</a> or integer <a href="#t_vector">vector</a>
4544 typed. They must also be identical types.</p>
4547 <p>The '<tt>icmp</tt>' compares <tt>op1</tt> and <tt>op2</tt> according to the
4548 condition code given as <tt>cond</tt>. The comparison performed always yields
4549 either an <a href="#t_primitive"><tt>i1</tt></a> or vector of <tt>i1</tt>
4550 result, as follows:</p>
4553 <li><tt>eq</tt>: yields <tt>true</tt> if the operands are equal,
4554 <tt>false</tt> otherwise. No sign interpretation is necessary or
4557 <li><tt>ne</tt>: yields <tt>true</tt> if the operands are unequal,
4558 <tt>false</tt> otherwise. No sign interpretation is necessary or
4561 <li><tt>ugt</tt>: interprets the operands as unsigned values and yields
4562 <tt>true</tt> if <tt>op1</tt> is greater than <tt>op2</tt>.</li>
4564 <li><tt>uge</tt>: interprets the operands as unsigned values and yields
4565 <tt>true</tt> if <tt>op1</tt> is greater than or equal
4566 to <tt>op2</tt>.</li>
4568 <li><tt>ult</tt>: interprets the operands as unsigned values and yields
4569 <tt>true</tt> if <tt>op1</tt> is less than <tt>op2</tt>.</li>
4571 <li><tt>ule</tt>: interprets the operands as unsigned values and yields
4572 <tt>true</tt> if <tt>op1</tt> is less than or equal to <tt>op2</tt>.</li>
4574 <li><tt>sgt</tt>: interprets the operands as signed values and yields
4575 <tt>true</tt> if <tt>op1</tt> is greater than <tt>op2</tt>.</li>
4577 <li><tt>sge</tt>: interprets the operands as signed values and yields
4578 <tt>true</tt> if <tt>op1</tt> is greater than or equal
4579 to <tt>op2</tt>.</li>
4581 <li><tt>slt</tt>: interprets the operands as signed values and yields
4582 <tt>true</tt> if <tt>op1</tt> is less than <tt>op2</tt>.</li>
4584 <li><tt>sle</tt>: interprets the operands as signed values and yields
4585 <tt>true</tt> if <tt>op1</tt> is less than or equal to <tt>op2</tt>.</li>
4588 <p>If the operands are <a href="#t_pointer">pointer</a> typed, the pointer
4589 values are compared as if they were integers.</p>
4591 <p>If the operands are integer vectors, then they are compared element by
4592 element. The result is an <tt>i1</tt> vector with the same number of elements
4593 as the values being compared. Otherwise, the result is an <tt>i1</tt>.</p>
4597 <result> = icmp eq i32 4, 5 <i>; yields: result=false</i>
4598 <result> = icmp ne float* %X, %X <i>; yields: result=false</i>
4599 <result> = icmp ult i16 4, 5 <i>; yields: result=true</i>
4600 <result> = icmp sgt i16 4, 5 <i>; yields: result=false</i>
4601 <result> = icmp ule i16 -4, 5 <i>; yields: result=false</i>
4602 <result> = icmp sge i16 4, 5 <i>; yields: result=false</i>
4605 <p>Note that the code generator does not yet support vector types with
4606 the <tt>icmp</tt> instruction.</p>
4610 <!-- _______________________________________________________________________ -->
4611 <div class="doc_subsubsection"><a name="i_fcmp">'<tt>fcmp</tt>' Instruction</a>
4614 <div class="doc_text">
4618 <result> = fcmp <cond> <ty> <op1>, <op2> <i>; yields {i1} or {<N x i1>}:result</i>
4622 <p>The '<tt>fcmp</tt>' instruction returns a boolean value or vector of boolean
4623 values based on comparison of its operands.</p>
4625 <p>If the operands are floating point scalars, then the result type is a boolean
4626 (<a href="#t_primitive"><tt>i1</tt></a>).</p>
4628 <p>If the operands are floating point vectors, then the result type is a vector
4629 of boolean with the same number of elements as the operands being
4633 <p>The '<tt>fcmp</tt>' instruction takes three operands. The first operand is
4634 the condition code indicating the kind of comparison to perform. It is not a
4635 value, just a keyword. The possible condition code are:</p>
4638 <li><tt>false</tt>: no comparison, always returns false</li>
4639 <li><tt>oeq</tt>: ordered and equal</li>
4640 <li><tt>ogt</tt>: ordered and greater than </li>
4641 <li><tt>oge</tt>: ordered and greater than or equal</li>
4642 <li><tt>olt</tt>: ordered and less than </li>
4643 <li><tt>ole</tt>: ordered and less than or equal</li>
4644 <li><tt>one</tt>: ordered and not equal</li>
4645 <li><tt>ord</tt>: ordered (no nans)</li>
4646 <li><tt>ueq</tt>: unordered or equal</li>
4647 <li><tt>ugt</tt>: unordered or greater than </li>
4648 <li><tt>uge</tt>: unordered or greater than or equal</li>
4649 <li><tt>ult</tt>: unordered or less than </li>
4650 <li><tt>ule</tt>: unordered or less than or equal</li>
4651 <li><tt>une</tt>: unordered or not equal</li>
4652 <li><tt>uno</tt>: unordered (either nans)</li>
4653 <li><tt>true</tt>: no comparison, always returns true</li>
4656 <p><i>Ordered</i> means that neither operand is a QNAN while
4657 <i>unordered</i> means that either operand may be a QNAN.</p>
4659 <p>Each of <tt>val1</tt> and <tt>val2</tt> arguments must be either
4660 a <a href="#t_floating">floating point</a> type or
4661 a <a href="#t_vector">vector</a> of floating point type. They must have
4662 identical types.</p>
4665 <p>The '<tt>fcmp</tt>' instruction compares <tt>op1</tt> and <tt>op2</tt>
4666 according to the condition code given as <tt>cond</tt>. If the operands are
4667 vectors, then the vectors are compared element by element. Each comparison
4668 performed always yields an <a href="#t_primitive">i1</a> result, as
4672 <li><tt>false</tt>: always yields <tt>false</tt>, regardless of operands.</li>
4674 <li><tt>oeq</tt>: yields <tt>true</tt> if both operands are not a QNAN and
4675 <tt>op1</tt> is equal to <tt>op2</tt>.</li>
4677 <li><tt>ogt</tt>: yields <tt>true</tt> if both operands are not a QNAN and
4678 <tt>op1</tt> is greather than <tt>op2</tt>.</li>
4680 <li><tt>oge</tt>: yields <tt>true</tt> if both operands are not a QNAN and
4681 <tt>op1</tt> is greater than or equal to <tt>op2</tt>.</li>
4683 <li><tt>olt</tt>: yields <tt>true</tt> if both operands are not a QNAN and
4684 <tt>op1</tt> is less than <tt>op2</tt>.</li>
4686 <li><tt>ole</tt>: yields <tt>true</tt> if both operands are not a QNAN and
4687 <tt>op1</tt> is less than or equal to <tt>op2</tt>.</li>
4689 <li><tt>one</tt>: yields <tt>true</tt> if both operands are not a QNAN and
4690 <tt>op1</tt> is not equal to <tt>op2</tt>.</li>
4692 <li><tt>ord</tt>: yields <tt>true</tt> if both operands are not a QNAN.</li>
4694 <li><tt>ueq</tt>: yields <tt>true</tt> if either operand is a QNAN or
4695 <tt>op1</tt> is equal to <tt>op2</tt>.</li>
4697 <li><tt>ugt</tt>: yields <tt>true</tt> if either operand is a QNAN or
4698 <tt>op1</tt> is greater than <tt>op2</tt>.</li>
4700 <li><tt>uge</tt>: yields <tt>true</tt> if either operand is a QNAN or
4701 <tt>op1</tt> is greater than or equal to <tt>op2</tt>.</li>
4703 <li><tt>ult</tt>: yields <tt>true</tt> if either operand is a QNAN or
4704 <tt>op1</tt> is less than <tt>op2</tt>.</li>
4706 <li><tt>ule</tt>: yields <tt>true</tt> if either operand is a QNAN or
4707 <tt>op1</tt> is less than or equal to <tt>op2</tt>.</li>
4709 <li><tt>une</tt>: yields <tt>true</tt> if either operand is a QNAN or
4710 <tt>op1</tt> is not equal to <tt>op2</tt>.</li>
4712 <li><tt>uno</tt>: yields <tt>true</tt> if either operand is a QNAN.</li>
4714 <li><tt>true</tt>: always yields <tt>true</tt>, regardless of operands.</li>
4719 <result> = fcmp oeq float 4.0, 5.0 <i>; yields: result=false</i>
4720 <result> = fcmp one float 4.0, 5.0 <i>; yields: result=true</i>
4721 <result> = fcmp olt float 4.0, 5.0 <i>; yields: result=true</i>
4722 <result> = fcmp ueq double 1.0, 2.0 <i>; yields: result=false</i>
4725 <p>Note that the code generator does not yet support vector types with
4726 the <tt>fcmp</tt> instruction.</p>
4730 <!-- _______________________________________________________________________ -->
4731 <div class="doc_subsubsection">
4732 <a name="i_phi">'<tt>phi</tt>' Instruction</a>
4735 <div class="doc_text">
4739 <result> = phi <ty> [ <val0>, <label0>], ...
4743 <p>The '<tt>phi</tt>' instruction is used to implement the φ node in the
4744 SSA graph representing the function.</p>
4747 <p>The type of the incoming values is specified with the first type field. After
4748 this, the '<tt>phi</tt>' instruction takes a list of pairs as arguments, with
4749 one pair for each predecessor basic block of the current block. Only values
4750 of <a href="#t_firstclass">first class</a> type may be used as the value
4751 arguments to the PHI node. Only labels may be used as the label
4754 <p>There must be no non-phi instructions between the start of a basic block and
4755 the PHI instructions: i.e. PHI instructions must be first in a basic
4758 <p>For the purposes of the SSA form, the use of each incoming value is deemed to
4759 occur on the edge from the corresponding predecessor block to the current
4760 block (but after any definition of an '<tt>invoke</tt>' instruction's return
4761 value on the same edge).</p>
4764 <p>At runtime, the '<tt>phi</tt>' instruction logically takes on the value
4765 specified by the pair corresponding to the predecessor basic block that
4766 executed just prior to the current block.</p>
4770 Loop: ; Infinite loop that counts from 0 on up...
4771 %indvar = phi i32 [ 0, %LoopHeader ], [ %nextindvar, %Loop ]
4772 %nextindvar = add i32 %indvar, 1
4778 <!-- _______________________________________________________________________ -->
4779 <div class="doc_subsubsection">
4780 <a name="i_select">'<tt>select</tt>' Instruction</a>
4783 <div class="doc_text">
4787 <result> = select <i>selty</i> <cond>, <ty> <val1>, <ty> <val2> <i>; yields ty</i>
4789 <i>selty</i> is either i1 or {<N x i1>}
4793 <p>The '<tt>select</tt>' instruction is used to choose one value based on a
4794 condition, without branching.</p>
4798 <p>The '<tt>select</tt>' instruction requires an 'i1' value or a vector of 'i1'
4799 values indicating the condition, and two values of the
4800 same <a href="#t_firstclass">first class</a> type. If the val1/val2 are
4801 vectors and the condition is a scalar, then entire vectors are selected, not
4802 individual elements.</p>
4805 <p>If the condition is an i1 and it evaluates to 1, the instruction returns the
4806 first value argument; otherwise, it returns the second value argument.</p>
4808 <p>If the condition is a vector of i1, then the value arguments must be vectors
4809 of the same size, and the selection is done element by element.</p>
4813 %X = select i1 true, i8 17, i8 42 <i>; yields i8:17</i>
4816 <p>Note that the code generator does not yet support conditions
4817 with vector type.</p>
4821 <!-- _______________________________________________________________________ -->
4822 <div class="doc_subsubsection">
4823 <a name="i_call">'<tt>call</tt>' Instruction</a>
4826 <div class="doc_text">
4830 <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>]
4834 <p>The '<tt>call</tt>' instruction represents a simple function call.</p>
4837 <p>This instruction requires several arguments:</p>
4840 <li>The optional "tail" marker indicates whether the callee function accesses
4841 any allocas or varargs in the caller. If the "tail" marker is present,
4842 the function call is eligible for tail call optimization. Note that calls
4843 may be marked "tail" even if they do not occur before
4844 a <a href="#i_ret"><tt>ret</tt></a> instruction.</li>
4846 <li>The optional "cconv" marker indicates which <a href="#callingconv">calling
4847 convention</a> the call should use. If none is specified, the call
4848 defaults to using C calling conventions.</li>
4850 <li>The optional <a href="#paramattrs">Parameter Attributes</a> list for
4851 return values. Only '<tt>zeroext</tt>', '<tt>signext</tt>', and
4852 '<tt>inreg</tt>' attributes are valid here.</li>
4854 <li>'<tt>ty</tt>': the type of the call instruction itself which is also the
4855 type of the return value. Functions that return no value are marked
4856 <tt><a href="#t_void">void</a></tt>.</li>
4858 <li>'<tt>fnty</tt>': shall be the signature of the pointer to function value
4859 being invoked. The argument types must match the types implied by this
4860 signature. This type can be omitted if the function is not varargs and if
4861 the function type does not return a pointer to a function.</li>
4863 <li>'<tt>fnptrval</tt>': An LLVM value containing a pointer to a function to
4864 be invoked. In most cases, this is a direct function invocation, but
4865 indirect <tt>call</tt>s are just as possible, calling an arbitrary pointer
4866 to function value.</li>
4868 <li>'<tt>function args</tt>': argument list whose types match the function
4869 signature argument types. All arguments must be of
4870 <a href="#t_firstclass">first class</a> type. If the function signature
4871 indicates the function accepts a variable number of arguments, the extra
4872 arguments can be specified.</li>
4874 <li>The optional <a href="#fnattrs">function attributes</a> list. Only
4875 '<tt>noreturn</tt>', '<tt>nounwind</tt>', '<tt>readonly</tt>' and
4876 '<tt>readnone</tt>' attributes are valid here.</li>
4880 <p>The '<tt>call</tt>' instruction is used to cause control flow to transfer to
4881 a specified function, with its incoming arguments bound to the specified
4882 values. Upon a '<tt><a href="#i_ret">ret</a></tt>' instruction in the called
4883 function, control flow continues with the instruction after the function
4884 call, and the return value of the function is bound to the result
4889 %retval = call i32 @test(i32 %argc)
4890 call i32 (i8 *, ...)* @printf(i8 * %msg, i32 12, i8 42) <i>; yields i32</i>
4891 %X = tail call i32 @foo() <i>; yields i32</i>
4892 %Y = tail call <a href="#callingconv">fastcc</a> i32 @foo() <i>; yields i32</i>
4893 call void %foo(i8 97 signext)
4895 %struct.A = type { i32, i8 }
4896 %r = call %struct.A @foo() <i>; yields { 32, i8 }</i>
4897 %gr = extractvalue %struct.A %r, 0 <i>; yields i32</i>
4898 %gr1 = extractvalue %struct.A %r, 1 <i>; yields i8</i>
4899 %Z = call void @foo() noreturn <i>; indicates that %foo never returns normally</i>
4900 %ZZ = call zeroext i32 @bar() <i>; Return value is %zero extended</i>
4905 <!-- _______________________________________________________________________ -->
4906 <div class="doc_subsubsection">
4907 <a name="i_va_arg">'<tt>va_arg</tt>' Instruction</a>
4910 <div class="doc_text">
4914 <resultval> = va_arg <va_list*> <arglist>, <argty>
4918 <p>The '<tt>va_arg</tt>' instruction is used to access arguments passed through
4919 the "variable argument" area of a function call. It is used to implement the
4920 <tt>va_arg</tt> macro in C.</p>
4923 <p>This instruction takes a <tt>va_list*</tt> value and the type of the
4924 argument. It returns a value of the specified argument type and increments
4925 the <tt>va_list</tt> to point to the next argument. The actual type
4926 of <tt>va_list</tt> is target specific.</p>
4929 <p>The '<tt>va_arg</tt>' instruction loads an argument of the specified type
4930 from the specified <tt>va_list</tt> and causes the <tt>va_list</tt> to point
4931 to the next argument. For more information, see the variable argument
4932 handling <a href="#int_varargs">Intrinsic Functions</a>.</p>
4934 <p>It is legal for this instruction to be called in a function which does not
4935 take a variable number of arguments, for example, the <tt>vfprintf</tt>
4938 <p><tt>va_arg</tt> is an LLVM instruction instead of
4939 an <a href="#intrinsics">intrinsic function</a> because it takes a type as an
4943 <p>See the <a href="#int_varargs">variable argument processing</a> section.</p>
4945 <p>Note that the code generator does not yet fully support va_arg on many
4946 targets. Also, it does not currently support va_arg with aggregate types on
4951 <!-- *********************************************************************** -->
4952 <div class="doc_section"> <a name="intrinsics">Intrinsic Functions</a> </div>
4953 <!-- *********************************************************************** -->
4955 <div class="doc_text">
4957 <p>LLVM supports the notion of an "intrinsic function". These functions have
4958 well known names and semantics and are required to follow certain
4959 restrictions. Overall, these intrinsics represent an extension mechanism for
4960 the LLVM language that does not require changing all of the transformations
4961 in LLVM when adding to the language (or the bitcode reader/writer, the
4962 parser, etc...).</p>
4964 <p>Intrinsic function names must all start with an "<tt>llvm.</tt>" prefix. This
4965 prefix is reserved in LLVM for intrinsic names; thus, function names may not
4966 begin with this prefix. Intrinsic functions must always be external
4967 functions: you cannot define the body of intrinsic functions. Intrinsic
4968 functions may only be used in call or invoke instructions: it is illegal to
4969 take the address of an intrinsic function. Additionally, because intrinsic
4970 functions are part of the LLVM language, it is required if any are added that
4971 they be documented here.</p>
4973 <p>Some intrinsic functions can be overloaded, i.e., the intrinsic represents a
4974 family of functions that perform the same operation but on different data
4975 types. Because LLVM can represent over 8 million different integer types,
4976 overloading is used commonly to allow an intrinsic function to operate on any
4977 integer type. One or more of the argument types or the result type can be
4978 overloaded to accept any integer type. Argument types may also be defined as
4979 exactly matching a previous argument's type or the result type. This allows
4980 an intrinsic function which accepts multiple arguments, but needs all of them
4981 to be of the same type, to only be overloaded with respect to a single
4982 argument or the result.</p>
4984 <p>Overloaded intrinsics will have the names of its overloaded argument types
4985 encoded into its function name, each preceded by a period. Only those types
4986 which are overloaded result in a name suffix. Arguments whose type is matched
4987 against another type do not. For example, the <tt>llvm.ctpop</tt> function
4988 can take an integer of any width and returns an integer of exactly the same
4989 integer width. This leads to a family of functions such as
4990 <tt>i8 @llvm.ctpop.i8(i8 %val)</tt> and <tt>i29 @llvm.ctpop.i29(i29
4991 %val)</tt>. Only one type, the return type, is overloaded, and only one type
4992 suffix is required. Because the argument's type is matched against the return
4993 type, it does not require its own name suffix.</p>
4995 <p>To learn how to add an intrinsic function, please see the
4996 <a href="ExtendingLLVM.html">Extending LLVM Guide</a>.</p>
5000 <!-- ======================================================================= -->
5001 <div class="doc_subsection">
5002 <a name="int_varargs">Variable Argument Handling Intrinsics</a>
5005 <div class="doc_text">
5007 <p>Variable argument support is defined in LLVM with
5008 the <a href="#i_va_arg"><tt>va_arg</tt></a> instruction and these three
5009 intrinsic functions. These functions are related to the similarly named
5010 macros defined in the <tt><stdarg.h></tt> header file.</p>
5012 <p>All of these functions operate on arguments that use a target-specific value
5013 type "<tt>va_list</tt>". The LLVM assembly language reference manual does
5014 not define what this type is, so all transformations should be prepared to
5015 handle these functions regardless of the type used.</p>
5017 <p>This example shows how the <a href="#i_va_arg"><tt>va_arg</tt></a>
5018 instruction and the variable argument handling intrinsic functions are
5021 <div class="doc_code">
5023 define i32 @test(i32 %X, ...) {
5024 ; Initialize variable argument processing
5026 %ap2 = bitcast i8** %ap to i8*
5027 call void @llvm.va_start(i8* %ap2)
5029 ; Read a single integer argument
5030 %tmp = va_arg i8** %ap, i32
5032 ; Demonstrate usage of llvm.va_copy and llvm.va_end
5034 %aq2 = bitcast i8** %aq to i8*
5035 call void @llvm.va_copy(i8* %aq2, i8* %ap2)
5036 call void @llvm.va_end(i8* %aq2)
5038 ; Stop processing of arguments.
5039 call void @llvm.va_end(i8* %ap2)
5043 declare void @llvm.va_start(i8*)
5044 declare void @llvm.va_copy(i8*, i8*)
5045 declare void @llvm.va_end(i8*)
5051 <!-- _______________________________________________________________________ -->
5052 <div class="doc_subsubsection">
5053 <a name="int_va_start">'<tt>llvm.va_start</tt>' Intrinsic</a>
5057 <div class="doc_text">
5061 declare void %llvm.va_start(i8* <arglist>)
5065 <p>The '<tt>llvm.va_start</tt>' intrinsic initializes <tt>*<arglist></tt>
5066 for subsequent use by <tt><a href="#i_va_arg">va_arg</a></tt>.</p>
5069 <p>The argument is a pointer to a <tt>va_list</tt> element to initialize.</p>
5072 <p>The '<tt>llvm.va_start</tt>' intrinsic works just like the <tt>va_start</tt>
5073 macro available in C. In a target-dependent way, it initializes
5074 the <tt>va_list</tt> element to which the argument points, so that the next
5075 call to <tt>va_arg</tt> will produce the first variable argument passed to
5076 the function. Unlike the C <tt>va_start</tt> macro, this intrinsic does not
5077 need to know the last argument of the function as the compiler can figure
5082 <!-- _______________________________________________________________________ -->
5083 <div class="doc_subsubsection">
5084 <a name="int_va_end">'<tt>llvm.va_end</tt>' Intrinsic</a>
5087 <div class="doc_text">
5091 declare void @llvm.va_end(i8* <arglist>)
5095 <p>The '<tt>llvm.va_end</tt>' intrinsic destroys <tt>*<arglist></tt>,
5096 which has been initialized previously
5097 with <tt><a href="#int_va_start">llvm.va_start</a></tt>
5098 or <tt><a href="#i_va_copy">llvm.va_copy</a></tt>.</p>
5101 <p>The argument is a pointer to a <tt>va_list</tt> to destroy.</p>
5104 <p>The '<tt>llvm.va_end</tt>' intrinsic works just like the <tt>va_end</tt>
5105 macro available in C. In a target-dependent way, it destroys
5106 the <tt>va_list</tt> element to which the argument points. Calls
5107 to <a href="#int_va_start"><tt>llvm.va_start</tt></a>
5108 and <a href="#int_va_copy"> <tt>llvm.va_copy</tt></a> must be matched exactly
5109 with calls to <tt>llvm.va_end</tt>.</p>
5113 <!-- _______________________________________________________________________ -->
5114 <div class="doc_subsubsection">
5115 <a name="int_va_copy">'<tt>llvm.va_copy</tt>' Intrinsic</a>
5118 <div class="doc_text">
5122 declare void @llvm.va_copy(i8* <destarglist>, i8* <srcarglist>)
5126 <p>The '<tt>llvm.va_copy</tt>' intrinsic copies the current argument position
5127 from the source argument list to the destination argument list.</p>
5130 <p>The first argument is a pointer to a <tt>va_list</tt> element to initialize.
5131 The second argument is a pointer to a <tt>va_list</tt> element to copy
5135 <p>The '<tt>llvm.va_copy</tt>' intrinsic works just like the <tt>va_copy</tt>
5136 macro available in C. In a target-dependent way, it copies the
5137 source <tt>va_list</tt> element into the destination <tt>va_list</tt>
5138 element. This intrinsic is necessary because
5139 the <tt><a href="#int_va_start"> llvm.va_start</a></tt> intrinsic may be
5140 arbitrarily complex and require, for example, memory allocation.</p>
5144 <!-- ======================================================================= -->
5145 <div class="doc_subsection">
5146 <a name="int_gc">Accurate Garbage Collection Intrinsics</a>
5149 <div class="doc_text">
5151 <p>LLVM support for <a href="GarbageCollection.html">Accurate Garbage
5152 Collection</a> (GC) requires the implementation and generation of these
5153 intrinsics. These intrinsics allow identification of <a href="#int_gcroot">GC
5154 roots on the stack</a>, as well as garbage collector implementations that
5155 require <a href="#int_gcread">read</a> and <a href="#int_gcwrite">write</a>
5156 barriers. Front-ends for type-safe garbage collected languages should generate
5157 these intrinsics to make use of the LLVM garbage collectors. For more details,
5158 see <a href="GarbageCollection.html">Accurate Garbage Collection with
5161 <p>The garbage collection intrinsics only operate on objects in the generic
5162 address space (address space zero).</p>
5166 <!-- _______________________________________________________________________ -->
5167 <div class="doc_subsubsection">
5168 <a name="int_gcroot">'<tt>llvm.gcroot</tt>' Intrinsic</a>
5171 <div class="doc_text">
5175 declare void @llvm.gcroot(i8** %ptrloc, i8* %metadata)
5179 <p>The '<tt>llvm.gcroot</tt>' intrinsic declares the existence of a GC root to
5180 the code generator, and allows some metadata to be associated with it.</p>
5183 <p>The first argument specifies the address of a stack object that contains the
5184 root pointer. The second pointer (which must be either a constant or a
5185 global value address) contains the meta-data to be associated with the
5189 <p>At runtime, a call to this intrinsic stores a null pointer into the "ptrloc"
5190 location. At compile-time, the code generator generates information to allow
5191 the runtime to find the pointer at GC safe points. The '<tt>llvm.gcroot</tt>'
5192 intrinsic may only be used in a function which <a href="#gc">specifies a GC
5197 <!-- _______________________________________________________________________ -->
5198 <div class="doc_subsubsection">
5199 <a name="int_gcread">'<tt>llvm.gcread</tt>' Intrinsic</a>
5202 <div class="doc_text">
5206 declare i8* @llvm.gcread(i8* %ObjPtr, i8** %Ptr)
5210 <p>The '<tt>llvm.gcread</tt>' intrinsic identifies reads of references from heap
5211 locations, allowing garbage collector implementations that require read
5215 <p>The second argument is the address to read from, which should be an address
5216 allocated from the garbage collector. The first object is a pointer to the
5217 start of the referenced object, if needed by the language runtime (otherwise
5221 <p>The '<tt>llvm.gcread</tt>' intrinsic has the same semantics as a load
5222 instruction, but may be replaced with substantially more complex code by the
5223 garbage collector runtime, as needed. The '<tt>llvm.gcread</tt>' intrinsic
5224 may only be used in a function which <a href="#gc">specifies a GC
5229 <!-- _______________________________________________________________________ -->
5230 <div class="doc_subsubsection">
5231 <a name="int_gcwrite">'<tt>llvm.gcwrite</tt>' Intrinsic</a>
5234 <div class="doc_text">
5238 declare void @llvm.gcwrite(i8* %P1, i8* %Obj, i8** %P2)
5242 <p>The '<tt>llvm.gcwrite</tt>' intrinsic identifies writes of references to heap
5243 locations, allowing garbage collector implementations that require write
5244 barriers (such as generational or reference counting collectors).</p>
5247 <p>The first argument is the reference to store, the second is the start of the
5248 object to store it to, and the third is the address of the field of Obj to
5249 store to. If the runtime does not require a pointer to the object, Obj may
5253 <p>The '<tt>llvm.gcwrite</tt>' intrinsic has the same semantics as a store
5254 instruction, but may be replaced with substantially more complex code by the
5255 garbage collector runtime, as needed. The '<tt>llvm.gcwrite</tt>' intrinsic
5256 may only be used in a function which <a href="#gc">specifies a GC
5261 <!-- ======================================================================= -->
5262 <div class="doc_subsection">
5263 <a name="int_codegen">Code Generator Intrinsics</a>
5266 <div class="doc_text">
5268 <p>These intrinsics are provided by LLVM to expose special features that may
5269 only be implemented with code generator support.</p>
5273 <!-- _______________________________________________________________________ -->
5274 <div class="doc_subsubsection">
5275 <a name="int_returnaddress">'<tt>llvm.returnaddress</tt>' Intrinsic</a>
5278 <div class="doc_text">
5282 declare i8 *@llvm.returnaddress(i32 <level>)
5286 <p>The '<tt>llvm.returnaddress</tt>' intrinsic attempts to compute a
5287 target-specific value indicating the return address of the current function
5288 or one of its callers.</p>
5291 <p>The argument to this intrinsic indicates which function to return the address
5292 for. Zero indicates the calling function, one indicates its caller, etc.
5293 The argument is <b>required</b> to be a constant integer value.</p>
5296 <p>The '<tt>llvm.returnaddress</tt>' intrinsic either returns a pointer
5297 indicating the return address of the specified call frame, or zero if it
5298 cannot be identified. The value returned by this intrinsic is likely to be
5299 incorrect or 0 for arguments other than zero, so it should only be used for
5300 debugging purposes.</p>
5302 <p>Note that calling this intrinsic does not prevent function inlining or other
5303 aggressive transformations, so the value returned may not be that of the
5304 obvious source-language caller.</p>
5308 <!-- _______________________________________________________________________ -->
5309 <div class="doc_subsubsection">
5310 <a name="int_frameaddress">'<tt>llvm.frameaddress</tt>' Intrinsic</a>
5313 <div class="doc_text">
5317 declare i8 *@llvm.frameaddress(i32 <level>)
5321 <p>The '<tt>llvm.frameaddress</tt>' intrinsic attempts to return the
5322 target-specific frame pointer value for the specified stack frame.</p>
5325 <p>The argument to this intrinsic indicates which function to return the frame
5326 pointer for. Zero indicates the calling function, one indicates its caller,
5327 etc. The argument is <b>required</b> to be a constant integer value.</p>
5330 <p>The '<tt>llvm.frameaddress</tt>' intrinsic either returns a pointer
5331 indicating the frame address of the specified call frame, or zero if it
5332 cannot be identified. The value returned by this intrinsic is likely to be
5333 incorrect or 0 for arguments other than zero, so it should only be used for
5334 debugging purposes.</p>
5336 <p>Note that calling this intrinsic does not prevent function inlining or other
5337 aggressive transformations, so the value returned may not be that of the
5338 obvious source-language caller.</p>
5342 <!-- _______________________________________________________________________ -->
5343 <div class="doc_subsubsection">
5344 <a name="int_stacksave">'<tt>llvm.stacksave</tt>' Intrinsic</a>
5347 <div class="doc_text">
5351 declare i8 *@llvm.stacksave()
5355 <p>The '<tt>llvm.stacksave</tt>' intrinsic is used to remember the current state
5356 of the function stack, for use
5357 with <a href="#int_stackrestore"> <tt>llvm.stackrestore</tt></a>. This is
5358 useful for implementing language features like scoped automatic variable
5359 sized arrays in C99.</p>
5362 <p>This intrinsic returns a opaque pointer value that can be passed
5363 to <a href="#int_stackrestore"><tt>llvm.stackrestore</tt></a>. When
5364 an <tt>llvm.stackrestore</tt> intrinsic is executed with a value saved
5365 from <tt>llvm.stacksave</tt>, it effectively restores the state of the stack
5366 to the state it was in when the <tt>llvm.stacksave</tt> intrinsic executed.
5367 In practice, this pops any <a href="#i_alloca">alloca</a> blocks from the
5368 stack that were allocated after the <tt>llvm.stacksave</tt> was executed.</p>
5372 <!-- _______________________________________________________________________ -->
5373 <div class="doc_subsubsection">
5374 <a name="int_stackrestore">'<tt>llvm.stackrestore</tt>' Intrinsic</a>
5377 <div class="doc_text">
5381 declare void @llvm.stackrestore(i8 * %ptr)
5385 <p>The '<tt>llvm.stackrestore</tt>' intrinsic is used to restore the state of
5386 the function stack to the state it was in when the
5387 corresponding <a href="#int_stacksave"><tt>llvm.stacksave</tt></a> intrinsic
5388 executed. This is useful for implementing language features like scoped
5389 automatic variable sized arrays in C99.</p>
5392 <p>See the description
5393 for <a href="#int_stacksave"><tt>llvm.stacksave</tt></a>.</p>
5397 <!-- _______________________________________________________________________ -->
5398 <div class="doc_subsubsection">
5399 <a name="int_prefetch">'<tt>llvm.prefetch</tt>' Intrinsic</a>
5402 <div class="doc_text">
5406 declare void @llvm.prefetch(i8* <address>, i32 <rw>, i32 <locality>)
5410 <p>The '<tt>llvm.prefetch</tt>' intrinsic is a hint to the code generator to
5411 insert a prefetch instruction if supported; otherwise, it is a noop.
5412 Prefetches have no effect on the behavior of the program but can change its
5413 performance characteristics.</p>
5416 <p><tt>address</tt> is the address to be prefetched, <tt>rw</tt> is the
5417 specifier determining if the fetch should be for a read (0) or write (1),
5418 and <tt>locality</tt> is a temporal locality specifier ranging from (0) - no
5419 locality, to (3) - extremely local keep in cache. The <tt>rw</tt>
5420 and <tt>locality</tt> arguments must be constant integers.</p>
5423 <p>This intrinsic does not modify the behavior of the program. In particular,
5424 prefetches cannot trap and do not produce a value. On targets that support
5425 this intrinsic, the prefetch can provide hints to the processor cache for
5426 better performance.</p>
5430 <!-- _______________________________________________________________________ -->
5431 <div class="doc_subsubsection">
5432 <a name="int_pcmarker">'<tt>llvm.pcmarker</tt>' Intrinsic</a>
5435 <div class="doc_text">
5439 declare void @llvm.pcmarker(i32 <id>)
5443 <p>The '<tt>llvm.pcmarker</tt>' intrinsic is a method to export a Program
5444 Counter (PC) in a region of code to simulators and other tools. The method
5445 is target specific, but it is expected that the marker will use exported
5446 symbols to transmit the PC of the marker. The marker makes no guarantees
5447 that it will remain with any specific instruction after optimizations. It is
5448 possible that the presence of a marker will inhibit optimizations. The
5449 intended use is to be inserted after optimizations to allow correlations of
5450 simulation runs.</p>
5453 <p><tt>id</tt> is a numerical id identifying the marker.</p>
5456 <p>This intrinsic does not modify the behavior of the program. Backends that do
5457 not support this intrinisic may ignore it.</p>
5461 <!-- _______________________________________________________________________ -->
5462 <div class="doc_subsubsection">
5463 <a name="int_readcyclecounter">'<tt>llvm.readcyclecounter</tt>' Intrinsic</a>
5466 <div class="doc_text">
5470 declare i64 @llvm.readcyclecounter( )
5474 <p>The '<tt>llvm.readcyclecounter</tt>' intrinsic provides access to the cycle
5475 counter register (or similar low latency, high accuracy clocks) on those
5476 targets that support it. On X86, it should map to RDTSC. On Alpha, it
5477 should map to RPCC. As the backing counters overflow quickly (on the order
5478 of 9 seconds on alpha), this should only be used for small timings.</p>
5481 <p>When directly supported, reading the cycle counter should not modify any
5482 memory. Implementations are allowed to either return a application specific
5483 value or a system wide value. On backends without support, this is lowered
5484 to a constant 0.</p>
5488 <!-- ======================================================================= -->
5489 <div class="doc_subsection">
5490 <a name="int_libc">Standard C Library Intrinsics</a>
5493 <div class="doc_text">
5495 <p>LLVM provides intrinsics for a few important standard C library functions.
5496 These intrinsics allow source-language front-ends to pass information about
5497 the alignment of the pointer arguments to the code generator, providing
5498 opportunity for more efficient code generation.</p>
5502 <!-- _______________________________________________________________________ -->
5503 <div class="doc_subsubsection">
5504 <a name="int_memcpy">'<tt>llvm.memcpy</tt>' Intrinsic</a>
5507 <div class="doc_text">
5510 <p>This is an overloaded intrinsic. You can use <tt>llvm.memcpy</tt> on any
5511 integer bit width. Not all targets support all bit widths however.</p>
5514 declare void @llvm.memcpy.i8(i8 * <dest>, i8 * <src>,
5515 i8 <len>, i32 <align>)
5516 declare void @llvm.memcpy.i16(i8 * <dest>, i8 * <src>,
5517 i16 <len>, i32 <align>)
5518 declare void @llvm.memcpy.i32(i8 * <dest>, i8 * <src>,
5519 i32 <len>, i32 <align>)
5520 declare void @llvm.memcpy.i64(i8 * <dest>, i8 * <src>,
5521 i64 <len>, i32 <align>)
5525 <p>The '<tt>llvm.memcpy.*</tt>' intrinsics copy a block of memory from the
5526 source location to the destination location.</p>
5528 <p>Note that, unlike the standard libc function, the <tt>llvm.memcpy.*</tt>
5529 intrinsics do not return a value, and takes an extra alignment argument.</p>
5532 <p>The first argument is a pointer to the destination, the second is a pointer
5533 to the source. The third argument is an integer argument specifying the
5534 number of bytes to copy, and the fourth argument is the alignment of the
5535 source and destination locations.</p>
5537 <p>If the call to this intrinisic has an alignment value that is not 0 or 1,
5538 then the caller guarantees that both the source and destination pointers are
5539 aligned to that boundary.</p>
5542 <p>The '<tt>llvm.memcpy.*</tt>' intrinsics copy a block of memory from the
5543 source location to the destination location, which are not allowed to
5544 overlap. It copies "len" bytes of memory over. If the argument is known to
5545 be aligned to some boundary, this can be specified as the fourth argument,
5546 otherwise it should be set to 0 or 1.</p>
5550 <!-- _______________________________________________________________________ -->
5551 <div class="doc_subsubsection">
5552 <a name="int_memmove">'<tt>llvm.memmove</tt>' Intrinsic</a>
5555 <div class="doc_text">
5558 <p>This is an overloaded intrinsic. You can use llvm.memmove on any integer bit
5559 width. Not all targets support all bit widths however.</p>
5562 declare void @llvm.memmove.i8(i8 * <dest>, i8 * <src>,
5563 i8 <len>, i32 <align>)
5564 declare void @llvm.memmove.i16(i8 * <dest>, i8 * <src>,
5565 i16 <len>, i32 <align>)
5566 declare void @llvm.memmove.i32(i8 * <dest>, i8 * <src>,
5567 i32 <len>, i32 <align>)
5568 declare void @llvm.memmove.i64(i8 * <dest>, i8 * <src>,
5569 i64 <len>, i32 <align>)
5573 <p>The '<tt>llvm.memmove.*</tt>' intrinsics move a block of memory from the
5574 source location to the destination location. It is similar to the
5575 '<tt>llvm.memcpy</tt>' intrinsic but allows the two memory locations to
5578 <p>Note that, unlike the standard libc function, the <tt>llvm.memmove.*</tt>
5579 intrinsics do not return a value, and takes an extra alignment argument.</p>
5582 <p>The first argument is a pointer to the destination, the second is a pointer
5583 to the source. The third argument is an integer argument specifying the
5584 number of bytes to copy, and the fourth argument is the alignment of the
5585 source and destination locations.</p>
5587 <p>If the call to this intrinisic has an alignment value that is not 0 or 1,
5588 then the caller guarantees that the source and destination pointers are
5589 aligned to that boundary.</p>
5592 <p>The '<tt>llvm.memmove.*</tt>' intrinsics copy a block of memory from the
5593 source location to the destination location, which may overlap. It copies
5594 "len" bytes of memory over. If the argument is known to be aligned to some
5595 boundary, this can be specified as the fourth argument, otherwise it should
5596 be set to 0 or 1.</p>
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>
5612 declare void @llvm.memset.i8(i8 * <dest>, i8 <val>,
5613 i8 <len>, i32 <align>)
5614 declare void @llvm.memset.i16(i8 * <dest>, i8 <val>,
5615 i16 <len>, i32 <align>)
5616 declare void @llvm.memset.i32(i8 * <dest>, i8 <val>,
5617 i32 <len>, i32 <align>)
5618 declare void @llvm.memset.i64(i8 * <dest>, i8 <val>,
5619 i64 <len>, i32 <align>)
5623 <p>The '<tt>llvm.memset.*</tt>' intrinsics fill a block of memory with a
5624 particular byte value.</p>
5626 <p>Note that, unlike the standard libc function, the <tt>llvm.memset</tt>
5627 intrinsic does not return a value, and takes an extra alignment argument.</p>
5630 <p>The first argument is a pointer to the destination to fill, the second is the
5631 byte value to fill it with, the third argument is an integer argument
5632 specifying the number of bytes to fill, and the fourth argument is the known
5633 alignment of destination location.</p>
5635 <p>If the call to this intrinisic has an alignment value that is not 0 or 1,
5636 then the caller guarantees that the destination pointer is aligned to that
5640 <p>The '<tt>llvm.memset.*</tt>' intrinsics fill "len" bytes of memory starting
5641 at the destination location. If the argument is known to be aligned to some
5642 boundary, this can be specified as the fourth argument, otherwise it should
5643 be set to 0 or 1.</p>
5647 <!-- _______________________________________________________________________ -->
5648 <div class="doc_subsubsection">
5649 <a name="int_sqrt">'<tt>llvm.sqrt.*</tt>' Intrinsic</a>
5652 <div class="doc_text">
5655 <p>This is an overloaded intrinsic. You can use <tt>llvm.sqrt</tt> on any
5656 floating point or vector of floating point type. Not all targets support all
5660 declare float @llvm.sqrt.f32(float %Val)
5661 declare double @llvm.sqrt.f64(double %Val)
5662 declare x86_fp80 @llvm.sqrt.f80(x86_fp80 %Val)
5663 declare fp128 @llvm.sqrt.f128(fp128 %Val)
5664 declare ppc_fp128 @llvm.sqrt.ppcf128(ppc_fp128 %Val)
5668 <p>The '<tt>llvm.sqrt</tt>' intrinsics return the sqrt of the specified operand,
5669 returning the same value as the libm '<tt>sqrt</tt>' functions would.
5670 Unlike <tt>sqrt</tt> in libm, however, <tt>llvm.sqrt</tt> has undefined
5671 behavior for negative numbers other than -0.0 (which allows for better
5672 optimization, because there is no need to worry about errno being
5673 set). <tt>llvm.sqrt(-0.0)</tt> is defined to return -0.0 like IEEE sqrt.</p>
5676 <p>The argument and return value are floating point numbers of the same
5680 <p>This function returns the sqrt of the specified operand if it is a
5681 nonnegative floating point number.</p>
5685 <!-- _______________________________________________________________________ -->
5686 <div class="doc_subsubsection">
5687 <a name="int_powi">'<tt>llvm.powi.*</tt>' Intrinsic</a>
5690 <div class="doc_text">
5693 <p>This is an overloaded intrinsic. You can use <tt>llvm.powi</tt> on any
5694 floating point or vector of floating point type. Not all targets support all
5698 declare float @llvm.powi.f32(float %Val, i32 %power)
5699 declare double @llvm.powi.f64(double %Val, i32 %power)
5700 declare x86_fp80 @llvm.powi.f80(x86_fp80 %Val, i32 %power)
5701 declare fp128 @llvm.powi.f128(fp128 %Val, i32 %power)
5702 declare ppc_fp128 @llvm.powi.ppcf128(ppc_fp128 %Val, i32 %power)
5706 <p>The '<tt>llvm.powi.*</tt>' intrinsics return the first operand raised to the
5707 specified (positive or negative) power. The order of evaluation of
5708 multiplications is not defined. When a vector of floating point type is
5709 used, the second argument remains a scalar integer value.</p>
5712 <p>The second argument is an integer power, and the first is a value to raise to
5716 <p>This function returns the first value raised to the second power with an
5717 unspecified sequence of rounding operations.</p>
5721 <!-- _______________________________________________________________________ -->
5722 <div class="doc_subsubsection">
5723 <a name="int_sin">'<tt>llvm.sin.*</tt>' Intrinsic</a>
5726 <div class="doc_text">
5729 <p>This is an overloaded intrinsic. You can use <tt>llvm.sin</tt> on any
5730 floating point or vector of floating point type. Not all targets support all
5734 declare float @llvm.sin.f32(float %Val)
5735 declare double @llvm.sin.f64(double %Val)
5736 declare x86_fp80 @llvm.sin.f80(x86_fp80 %Val)
5737 declare fp128 @llvm.sin.f128(fp128 %Val)
5738 declare ppc_fp128 @llvm.sin.ppcf128(ppc_fp128 %Val)
5742 <p>The '<tt>llvm.sin.*</tt>' intrinsics return the sine of the operand.</p>
5745 <p>The argument and return value are floating point numbers of the same
5749 <p>This function returns the sine of the specified operand, returning the same
5750 values as the libm <tt>sin</tt> functions would, and handles error conditions
5751 in the same way.</p>
5755 <!-- _______________________________________________________________________ -->
5756 <div class="doc_subsubsection">
5757 <a name="int_cos">'<tt>llvm.cos.*</tt>' Intrinsic</a>
5760 <div class="doc_text">
5763 <p>This is an overloaded intrinsic. You can use <tt>llvm.cos</tt> on any
5764 floating point or vector of floating point type. Not all targets support all
5768 declare float @llvm.cos.f32(float %Val)
5769 declare double @llvm.cos.f64(double %Val)
5770 declare x86_fp80 @llvm.cos.f80(x86_fp80 %Val)
5771 declare fp128 @llvm.cos.f128(fp128 %Val)
5772 declare ppc_fp128 @llvm.cos.ppcf128(ppc_fp128 %Val)
5776 <p>The '<tt>llvm.cos.*</tt>' intrinsics return the cosine of the operand.</p>
5779 <p>The argument and return value are floating point numbers of the same
5783 <p>This function returns the cosine of the specified operand, returning the same
5784 values as the libm <tt>cos</tt> functions would, and handles error conditions
5785 in the same way.</p>
5789 <!-- _______________________________________________________________________ -->
5790 <div class="doc_subsubsection">
5791 <a name="int_pow">'<tt>llvm.pow.*</tt>' Intrinsic</a>
5794 <div class="doc_text">
5797 <p>This is an overloaded intrinsic. You can use <tt>llvm.pow</tt> on any
5798 floating point or vector of floating point type. Not all targets support all
5802 declare float @llvm.pow.f32(float %Val, float %Power)
5803 declare double @llvm.pow.f64(double %Val, double %Power)
5804 declare x86_fp80 @llvm.pow.f80(x86_fp80 %Val, x86_fp80 %Power)
5805 declare fp128 @llvm.pow.f128(fp128 %Val, fp128 %Power)
5806 declare ppc_fp128 @llvm.pow.ppcf128(ppc_fp128 %Val, ppc_fp128 Power)
5810 <p>The '<tt>llvm.pow.*</tt>' intrinsics return the first operand raised to the
5811 specified (positive or negative) power.</p>
5814 <p>The second argument is a floating point power, and the first is a value to
5815 raise to that power.</p>
5818 <p>This function returns the first value raised to the second power, returning
5819 the same values as the libm <tt>pow</tt> functions would, and handles error
5820 conditions in the same way.</p>
5824 <!-- ======================================================================= -->
5825 <div class="doc_subsection">
5826 <a name="int_manip">Bit Manipulation Intrinsics</a>
5829 <div class="doc_text">
5831 <p>LLVM provides intrinsics for a few important bit manipulation operations.
5832 These allow efficient code generation for some algorithms.</p>
5836 <!-- _______________________________________________________________________ -->
5837 <div class="doc_subsubsection">
5838 <a name="int_bswap">'<tt>llvm.bswap.*</tt>' Intrinsics</a>
5841 <div class="doc_text">
5844 <p>This is an overloaded intrinsic function. You can use bswap on any integer
5845 type that is an even number of bytes (i.e. BitWidth % 16 == 0).</p>
5848 declare i16 @llvm.bswap.i16(i16 <id>)
5849 declare i32 @llvm.bswap.i32(i32 <id>)
5850 declare i64 @llvm.bswap.i64(i64 <id>)
5854 <p>The '<tt>llvm.bswap</tt>' family of intrinsics is used to byte swap integer
5855 values with an even number of bytes (positive multiple of 16 bits). These
5856 are useful for performing operations on data that is not in the target's
5857 native byte order.</p>
5860 <p>The <tt>llvm.bswap.i16</tt> intrinsic returns an i16 value that has the high
5861 and low byte of the input i16 swapped. Similarly,
5862 the <tt>llvm.bswap.i32</tt> intrinsic returns an i32 value that has the four
5863 bytes of the input i32 swapped, so that if the input bytes are numbered 0, 1,
5864 2, 3 then the returned i32 will have its bytes in 3, 2, 1, 0 order.
5865 The <tt>llvm.bswap.i48</tt>, <tt>llvm.bswap.i64</tt> and other intrinsics
5866 extend this concept to additional even-byte lengths (6 bytes, 8 bytes and
5867 more, respectively).</p>
5871 <!-- _______________________________________________________________________ -->
5872 <div class="doc_subsubsection">
5873 <a name="int_ctpop">'<tt>llvm.ctpop.*</tt>' Intrinsic</a>
5876 <div class="doc_text">
5879 <p>This is an overloaded intrinsic. You can use llvm.ctpop on any integer bit
5880 width. Not all targets support all bit widths however.</p>
5883 declare i8 @llvm.ctpop.i8(i8 <src>)
5884 declare i16 @llvm.ctpop.i16(i16 <src>)
5885 declare i32 @llvm.ctpop.i32(i32 <src>)
5886 declare i64 @llvm.ctpop.i64(i64 <src>)
5887 declare i256 @llvm.ctpop.i256(i256 <src>)
5891 <p>The '<tt>llvm.ctpop</tt>' family of intrinsics counts the number of bits set
5895 <p>The only argument is the value to be counted. The argument may be of any
5896 integer type. The return type must match the argument type.</p>
5899 <p>The '<tt>llvm.ctpop</tt>' intrinsic counts the 1's in a variable.</p>
5903 <!-- _______________________________________________________________________ -->
5904 <div class="doc_subsubsection">
5905 <a name="int_ctlz">'<tt>llvm.ctlz.*</tt>' Intrinsic</a>
5908 <div class="doc_text">
5911 <p>This is an overloaded intrinsic. You can use <tt>llvm.ctlz</tt> on any
5912 integer bit width. Not all targets support all bit widths however.</p>
5915 declare i8 @llvm.ctlz.i8 (i8 <src>)
5916 declare i16 @llvm.ctlz.i16(i16 <src>)
5917 declare i32 @llvm.ctlz.i32(i32 <src>)
5918 declare i64 @llvm.ctlz.i64(i64 <src>)
5919 declare i256 @llvm.ctlz.i256(i256 <src>)
5923 <p>The '<tt>llvm.ctlz</tt>' family of intrinsic functions counts the number of
5924 leading zeros in a variable.</p>
5927 <p>The only argument is the value to be counted. The argument may be of any
5928 integer type. The return type must match the argument type.</p>
5931 <p>The '<tt>llvm.ctlz</tt>' intrinsic counts the leading (most significant)
5932 zeros in a variable. If the src == 0 then the result is the size in bits of
5933 the type of src. For example, <tt>llvm.ctlz(i32 2) = 30</tt>.</p>
5937 <!-- _______________________________________________________________________ -->
5938 <div class="doc_subsubsection">
5939 <a name="int_cttz">'<tt>llvm.cttz.*</tt>' Intrinsic</a>
5942 <div class="doc_text">
5945 <p>This is an overloaded intrinsic. You can use <tt>llvm.cttz</tt> on any
5946 integer bit width. Not all targets support all bit widths however.</p>
5949 declare i8 @llvm.cttz.i8 (i8 <src>)
5950 declare i16 @llvm.cttz.i16(i16 <src>)
5951 declare i32 @llvm.cttz.i32(i32 <src>)
5952 declare i64 @llvm.cttz.i64(i64 <src>)
5953 declare i256 @llvm.cttz.i256(i256 <src>)
5957 <p>The '<tt>llvm.cttz</tt>' family of intrinsic functions counts the number of
5961 <p>The only argument is the value to be counted. The argument may be of any
5962 integer type. The return type must match the argument type.</p>
5965 <p>The '<tt>llvm.cttz</tt>' intrinsic counts the trailing (least significant)
5966 zeros in a variable. If the src == 0 then the result is the size in bits of
5967 the type of src. For example, <tt>llvm.cttz(2) = 1</tt>.</p>
5971 <!-- ======================================================================= -->
5972 <div class="doc_subsection">
5973 <a name="int_overflow">Arithmetic with Overflow Intrinsics</a>
5976 <div class="doc_text">
5978 <p>LLVM provides intrinsics for some arithmetic with overflow operations.</p>
5982 <!-- _______________________________________________________________________ -->
5983 <div class="doc_subsubsection">
5984 <a name="int_sadd_overflow">'<tt>llvm.sadd.with.overflow.*</tt>' Intrinsics</a>
5987 <div class="doc_text">
5990 <p>This is an overloaded intrinsic. You can use <tt>llvm.sadd.with.overflow</tt>
5991 on any integer bit width.</p>
5994 declare {i16, i1} @llvm.sadd.with.overflow.i16(i16 %a, i16 %b)
5995 declare {i32, i1} @llvm.sadd.with.overflow.i32(i32 %a, i32 %b)
5996 declare {i64, i1} @llvm.sadd.with.overflow.i64(i64 %a, i64 %b)
6000 <p>The '<tt>llvm.sadd.with.overflow</tt>' family of intrinsic functions perform
6001 a signed addition of the two arguments, and indicate whether an overflow
6002 occurred during the signed summation.</p>
6005 <p>The arguments (%a and %b) and the first element of the result structure may
6006 be of integer types of any bit width, but they must have the same bit
6007 width. The second element of the result structure must be of
6008 type <tt>i1</tt>. <tt>%a</tt> and <tt>%b</tt> are the two values that will
6009 undergo signed addition.</p>
6012 <p>The '<tt>llvm.sadd.with.overflow</tt>' family of intrinsic functions perform
6013 a signed addition of the two variables. They return a structure — the
6014 first element of which is the signed summation, and the second element of
6015 which is a bit specifying if the signed summation resulted in an
6020 %res = call {i32, i1} @llvm.sadd.with.overflow.i32(i32 %a, i32 %b)
6021 %sum = extractvalue {i32, i1} %res, 0
6022 %obit = extractvalue {i32, i1} %res, 1
6023 br i1 %obit, label %overflow, label %normal
6028 <!-- _______________________________________________________________________ -->
6029 <div class="doc_subsubsection">
6030 <a name="int_uadd_overflow">'<tt>llvm.uadd.with.overflow.*</tt>' Intrinsics</a>
6033 <div class="doc_text">
6036 <p>This is an overloaded intrinsic. You can use <tt>llvm.uadd.with.overflow</tt>
6037 on any integer bit width.</p>
6040 declare {i16, i1} @llvm.uadd.with.overflow.i16(i16 %a, i16 %b)
6041 declare {i32, i1} @llvm.uadd.with.overflow.i32(i32 %a, i32 %b)
6042 declare {i64, i1} @llvm.uadd.with.overflow.i64(i64 %a, i64 %b)
6046 <p>The '<tt>llvm.uadd.with.overflow</tt>' family of intrinsic functions perform
6047 an unsigned addition of the two arguments, and indicate whether a carry
6048 occurred during the unsigned summation.</p>
6051 <p>The arguments (%a and %b) and the first element of the result structure may
6052 be of integer types of any bit width, but they must have the same bit
6053 width. The second element of the result structure must be of
6054 type <tt>i1</tt>. <tt>%a</tt> and <tt>%b</tt> are the two values that will
6055 undergo unsigned addition.</p>
6058 <p>The '<tt>llvm.uadd.with.overflow</tt>' family of intrinsic functions perform
6059 an unsigned addition of the two arguments. They return a structure —
6060 the first element of which is the sum, and the second element of which is a
6061 bit specifying if the unsigned summation resulted in a carry.</p>
6065 %res = call {i32, i1} @llvm.uadd.with.overflow.i32(i32 %a, i32 %b)
6066 %sum = extractvalue {i32, i1} %res, 0
6067 %obit = extractvalue {i32, i1} %res, 1
6068 br i1 %obit, label %carry, label %normal
6073 <!-- _______________________________________________________________________ -->
6074 <div class="doc_subsubsection">
6075 <a name="int_ssub_overflow">'<tt>llvm.ssub.with.overflow.*</tt>' Intrinsics</a>
6078 <div class="doc_text">
6081 <p>This is an overloaded intrinsic. You can use <tt>llvm.ssub.with.overflow</tt>
6082 on any integer bit width.</p>
6085 declare {i16, i1} @llvm.ssub.with.overflow.i16(i16 %a, i16 %b)
6086 declare {i32, i1} @llvm.ssub.with.overflow.i32(i32 %a, i32 %b)
6087 declare {i64, i1} @llvm.ssub.with.overflow.i64(i64 %a, i64 %b)
6091 <p>The '<tt>llvm.ssub.with.overflow</tt>' family of intrinsic functions perform
6092 a signed subtraction of the two arguments, and indicate whether an overflow
6093 occurred during the signed subtraction.</p>
6096 <p>The arguments (%a and %b) and the first element of the result structure may
6097 be of integer types of any bit width, but they must have the same bit
6098 width. The second element of the result structure must be of
6099 type <tt>i1</tt>. <tt>%a</tt> and <tt>%b</tt> are the two values that will
6100 undergo signed subtraction.</p>
6103 <p>The '<tt>llvm.ssub.with.overflow</tt>' family of intrinsic functions perform
6104 a signed subtraction of the two arguments. They return a structure —
6105 the first element of which is the subtraction, and the second element of
6106 which is a bit specifying if the signed subtraction resulted in an
6111 %res = call {i32, i1} @llvm.ssub.with.overflow.i32(i32 %a, i32 %b)
6112 %sum = extractvalue {i32, i1} %res, 0
6113 %obit = extractvalue {i32, i1} %res, 1
6114 br i1 %obit, label %overflow, label %normal
6119 <!-- _______________________________________________________________________ -->
6120 <div class="doc_subsubsection">
6121 <a name="int_usub_overflow">'<tt>llvm.usub.with.overflow.*</tt>' Intrinsics</a>
6124 <div class="doc_text">
6127 <p>This is an overloaded intrinsic. You can use <tt>llvm.usub.with.overflow</tt>
6128 on any integer bit width.</p>
6131 declare {i16, i1} @llvm.usub.with.overflow.i16(i16 %a, i16 %b)
6132 declare {i32, i1} @llvm.usub.with.overflow.i32(i32 %a, i32 %b)
6133 declare {i64, i1} @llvm.usub.with.overflow.i64(i64 %a, i64 %b)
6137 <p>The '<tt>llvm.usub.with.overflow</tt>' family of intrinsic functions perform
6138 an unsigned subtraction of the two arguments, and indicate whether an
6139 overflow occurred during the unsigned subtraction.</p>
6142 <p>The arguments (%a and %b) and the first element of the result structure may
6143 be of integer types of any bit width, but they must have the same bit
6144 width. The second element of the result structure must be of
6145 type <tt>i1</tt>. <tt>%a</tt> and <tt>%b</tt> are the two values that will
6146 undergo unsigned subtraction.</p>
6149 <p>The '<tt>llvm.usub.with.overflow</tt>' family of intrinsic functions perform
6150 an unsigned subtraction of the two arguments. They return a structure —
6151 the first element of which is the subtraction, and the second element of
6152 which is a bit specifying if the unsigned subtraction resulted in an
6157 %res = call {i32, i1} @llvm.usub.with.overflow.i32(i32 %a, i32 %b)
6158 %sum = extractvalue {i32, i1} %res, 0
6159 %obit = extractvalue {i32, i1} %res, 1
6160 br i1 %obit, label %overflow, label %normal
6165 <!-- _______________________________________________________________________ -->
6166 <div class="doc_subsubsection">
6167 <a name="int_smul_overflow">'<tt>llvm.smul.with.overflow.*</tt>' Intrinsics</a>
6170 <div class="doc_text">
6173 <p>This is an overloaded intrinsic. You can use <tt>llvm.smul.with.overflow</tt>
6174 on any integer bit width.</p>
6177 declare {i16, i1} @llvm.smul.with.overflow.i16(i16 %a, i16 %b)
6178 declare {i32, i1} @llvm.smul.with.overflow.i32(i32 %a, i32 %b)
6179 declare {i64, i1} @llvm.smul.with.overflow.i64(i64 %a, i64 %b)
6184 <p>The '<tt>llvm.smul.with.overflow</tt>' family of intrinsic functions perform
6185 a signed multiplication of the two arguments, and indicate whether an
6186 overflow occurred during the signed multiplication.</p>
6189 <p>The arguments (%a and %b) and the first element of the result structure may
6190 be of integer types of any bit width, but they must have the same bit
6191 width. The second element of the result structure must be of
6192 type <tt>i1</tt>. <tt>%a</tt> and <tt>%b</tt> are the two values that will
6193 undergo signed multiplication.</p>
6196 <p>The '<tt>llvm.smul.with.overflow</tt>' family of intrinsic functions perform
6197 a signed multiplication of the two arguments. They return a structure —
6198 the first element of which is the multiplication, and the second element of
6199 which is a bit specifying if the signed multiplication resulted in an
6204 %res = call {i32, i1} @llvm.smul.with.overflow.i32(i32 %a, i32 %b)
6205 %sum = extractvalue {i32, i1} %res, 0
6206 %obit = extractvalue {i32, i1} %res, 1
6207 br i1 %obit, label %overflow, label %normal
6212 <!-- _______________________________________________________________________ -->
6213 <div class="doc_subsubsection">
6214 <a name="int_umul_overflow">'<tt>llvm.umul.with.overflow.*</tt>' Intrinsics</a>
6217 <div class="doc_text">
6220 <p>This is an overloaded intrinsic. You can use <tt>llvm.umul.with.overflow</tt>
6221 on any integer bit width.</p>
6224 declare {i16, i1} @llvm.umul.with.overflow.i16(i16 %a, i16 %b)
6225 declare {i32, i1} @llvm.umul.with.overflow.i32(i32 %a, i32 %b)
6226 declare {i64, i1} @llvm.umul.with.overflow.i64(i64 %a, i64 %b)
6230 <p>The '<tt>llvm.umul.with.overflow</tt>' family of intrinsic functions perform
6231 a unsigned multiplication of the two arguments, and indicate whether an
6232 overflow occurred during the unsigned multiplication.</p>
6235 <p>The arguments (%a and %b) and the first element of the result structure may
6236 be of integer types of any bit width, but they must have the same bit
6237 width. The second element of the result structure must be of
6238 type <tt>i1</tt>. <tt>%a</tt> and <tt>%b</tt> are the two values that will
6239 undergo unsigned multiplication.</p>
6242 <p>The '<tt>llvm.umul.with.overflow</tt>' family of intrinsic functions perform
6243 an unsigned multiplication of the two arguments. They return a structure
6244 — the first element of which is the multiplication, and the second
6245 element of which is a bit specifying if the unsigned multiplication resulted
6250 %res = call {i32, i1} @llvm.umul.with.overflow.i32(i32 %a, i32 %b)
6251 %sum = extractvalue {i32, i1} %res, 0
6252 %obit = extractvalue {i32, i1} %res, 1
6253 br i1 %obit, label %overflow, label %normal
6258 <!-- ======================================================================= -->
6259 <div class="doc_subsection">
6260 <a name="int_debugger">Debugger Intrinsics</a>
6263 <div class="doc_text">
6265 <p>The LLVM debugger intrinsics (which all start with <tt>llvm.dbg.</tt>
6266 prefix), are described in
6267 the <a href="SourceLevelDebugging.html#format_common_intrinsics">LLVM Source
6268 Level Debugging</a> document.</p>
6272 <!-- ======================================================================= -->
6273 <div class="doc_subsection">
6274 <a name="int_eh">Exception Handling Intrinsics</a>
6277 <div class="doc_text">
6279 <p>The LLVM exception handling intrinsics (which all start with
6280 <tt>llvm.eh.</tt> prefix), are described in
6281 the <a href="ExceptionHandling.html#format_common_intrinsics">LLVM Exception
6282 Handling</a> document.</p>
6286 <!-- ======================================================================= -->
6287 <div class="doc_subsection">
6288 <a name="int_trampoline">Trampoline Intrinsic</a>
6291 <div class="doc_text">
6293 <p>This intrinsic makes it possible to excise one parameter, marked with
6294 the <tt>nest</tt> attribute, from a function. The result is a callable
6295 function pointer lacking the nest parameter - the caller does not need to
6296 provide a value for it. Instead, the value to use is stored in advance in a
6297 "trampoline", a block of memory usually allocated on the stack, which also
6298 contains code to splice the nest value into the argument list. This is used
6299 to implement the GCC nested function address extension.</p>
6301 <p>For example, if the function is
6302 <tt>i32 f(i8* nest %c, i32 %x, i32 %y)</tt> then the resulting function
6303 pointer has signature <tt>i32 (i32, i32)*</tt>. It can be created as
6306 <div class="doc_code">
6308 %tramp = alloca [10 x i8], align 4 ; size and alignment only correct for X86
6309 %tramp1 = getelementptr [10 x i8]* %tramp, i32 0, i32 0
6310 %p = call i8* @llvm.init.trampoline( i8* %tramp1, i8* bitcast (i32 (i8* nest , i32, i32)* @f to i8*), i8* %nval )
6311 %fp = bitcast i8* %p to i32 (i32, i32)*
6315 <p>The call <tt>%val = call i32 %fp( i32 %x, i32 %y )</tt> is then equivalent
6316 to <tt>%val = call i32 %f( i8* %nval, i32 %x, i32 %y )</tt>.</p>
6320 <!-- _______________________________________________________________________ -->
6321 <div class="doc_subsubsection">
6322 <a name="int_it">'<tt>llvm.init.trampoline</tt>' Intrinsic</a>
6325 <div class="doc_text">
6329 declare i8* @llvm.init.trampoline(i8* <tramp>, i8* <func>, i8* <nval>)
6333 <p>This fills the memory pointed to by <tt>tramp</tt> with code and returns a
6334 function pointer suitable for executing it.</p>
6337 <p>The <tt>llvm.init.trampoline</tt> intrinsic takes three arguments, all
6338 pointers. The <tt>tramp</tt> argument must point to a sufficiently large and
6339 sufficiently aligned block of memory; this memory is written to by the
6340 intrinsic. Note that the size and the alignment are target-specific - LLVM
6341 currently provides no portable way of determining them, so a front-end that
6342 generates this intrinsic needs to have some target-specific knowledge.
6343 The <tt>func</tt> argument must hold a function bitcast to
6344 an <tt>i8*</tt>.</p>
6347 <p>The block of memory pointed to by <tt>tramp</tt> is filled with target
6348 dependent code, turning it into a function. A pointer to this function is
6349 returned, but needs to be bitcast to an <a href="#int_trampoline">appropriate
6350 function pointer type</a> before being called. The new function's signature
6351 is the same as that of <tt>func</tt> with any arguments marked with
6352 the <tt>nest</tt> attribute removed. At most one such <tt>nest</tt> argument
6353 is allowed, and it must be of pointer type. Calling the new function is
6354 equivalent to calling <tt>func</tt> with the same argument list, but
6355 with <tt>nval</tt> used for the missing <tt>nest</tt> argument. If, after
6356 calling <tt>llvm.init.trampoline</tt>, the memory pointed to
6357 by <tt>tramp</tt> is modified, then the effect of any later call to the
6358 returned function pointer is undefined.</p>
6362 <!-- ======================================================================= -->
6363 <div class="doc_subsection">
6364 <a name="int_atomics">Atomic Operations and Synchronization Intrinsics</a>
6367 <div class="doc_text">
6369 <p>These intrinsic functions expand the "universal IR" of LLVM to represent
6370 hardware constructs for atomic operations and memory synchronization. This
6371 provides an interface to the hardware, not an interface to the programmer. It
6372 is aimed at a low enough level to allow any programming models or APIs
6373 (Application Programming Interfaces) which need atomic behaviors to map
6374 cleanly onto it. It is also modeled primarily on hardware behavior. Just as
6375 hardware provides a "universal IR" for source languages, it also provides a
6376 starting point for developing a "universal" atomic operation and
6377 synchronization IR.</p>
6379 <p>These do <em>not</em> form an API such as high-level threading libraries,
6380 software transaction memory systems, atomic primitives, and intrinsic
6381 functions as found in BSD, GNU libc, atomic_ops, APR, and other system and
6382 application libraries. The hardware interface provided by LLVM should allow
6383 a clean implementation of all of these APIs and parallel programming models.
6384 No one model or paradigm should be selected above others unless the hardware
6385 itself ubiquitously does so.</p>
6389 <!-- _______________________________________________________________________ -->
6390 <div class="doc_subsubsection">
6391 <a name="int_memory_barrier">'<tt>llvm.memory.barrier</tt>' Intrinsic</a>
6393 <div class="doc_text">
6396 declare void @llvm.memory.barrier( i1 <ll>, i1 <ls>, i1 <sl>, i1 <ss>, i1 <device> )
6400 <p>The <tt>llvm.memory.barrier</tt> intrinsic guarantees ordering between
6401 specific pairs of memory access types.</p>
6404 <p>The <tt>llvm.memory.barrier</tt> intrinsic requires five boolean arguments.
6405 The first four arguments enables a specific barrier as listed below. The
6406 fith argument specifies that the barrier applies to io or device or uncached
6410 <li><tt>ll</tt>: load-load barrier</li>
6411 <li><tt>ls</tt>: load-store barrier</li>
6412 <li><tt>sl</tt>: store-load barrier</li>
6413 <li><tt>ss</tt>: store-store barrier</li>
6414 <li><tt>device</tt>: barrier applies to device and uncached memory also.</li>
6418 <p>This intrinsic causes the system to enforce some ordering constraints upon
6419 the loads and stores of the program. This barrier does not
6420 indicate <em>when</em> any events will occur, it only enforces
6421 an <em>order</em> in which they occur. For any of the specified pairs of load
6422 and store operations (f.ex. load-load, or store-load), all of the first
6423 operations preceding the barrier will complete before any of the second
6424 operations succeeding the barrier begin. Specifically the semantics for each
6425 pairing is as follows:</p>
6428 <li><tt>ll</tt>: All loads before the barrier must complete before any load
6429 after the barrier begins.</li>
6430 <li><tt>ls</tt>: All loads before the barrier must complete before any
6431 store after the barrier begins.</li>
6432 <li><tt>ss</tt>: All stores before the barrier must complete before any
6433 store after the barrier begins.</li>
6434 <li><tt>sl</tt>: All stores before the barrier must complete before any
6435 load after the barrier begins.</li>
6438 <p>These semantics are applied with a logical "and" behavior when more than one
6439 is enabled in a single memory barrier intrinsic.</p>
6441 <p>Backends may implement stronger barriers than those requested when they do
6442 not support as fine grained a barrier as requested. Some architectures do
6443 not need all types of barriers and on such architectures, these become
6451 %result1 = load i32* %ptr <i>; yields {i32}:result1 = 4</i>
6452 call void @llvm.memory.barrier( i1 false, i1 true, i1 false, i1 false )
6453 <i>; guarantee the above finishes</i>
6454 store i32 8, %ptr <i>; before this begins</i>
6459 <!-- _______________________________________________________________________ -->
6460 <div class="doc_subsubsection">
6461 <a name="int_atomic_cmp_swap">'<tt>llvm.atomic.cmp.swap.*</tt>' Intrinsic</a>
6464 <div class="doc_text">
6467 <p>This is an overloaded intrinsic. You can use <tt>llvm.atomic.cmp.swap</tt> on
6468 any integer bit width and for different address spaces. Not all targets
6469 support all bit widths however.</p>
6472 declare i8 @llvm.atomic.cmp.swap.i8.p0i8( i8* <ptr>, i8 <cmp>, i8 <val> )
6473 declare i16 @llvm.atomic.cmp.swap.i16.p0i16( i16* <ptr>, i16 <cmp>, i16 <val> )
6474 declare i32 @llvm.atomic.cmp.swap.i32.p0i32( i32* <ptr>, i32 <cmp>, i32 <val> )
6475 declare i64 @llvm.atomic.cmp.swap.i64.p0i64( i64* <ptr>, i64 <cmp>, i64 <val> )
6479 <p>This loads a value in memory and compares it to a given value. If they are
6480 equal, it stores a new value into the memory.</p>
6483 <p>The <tt>llvm.atomic.cmp.swap</tt> intrinsic takes three arguments. The result
6484 as well as both <tt>cmp</tt> and <tt>val</tt> must be integer values with the
6485 same bit width. The <tt>ptr</tt> argument must be a pointer to a value of
6486 this integer type. While any bit width integer may be used, targets may only
6487 lower representations they support in hardware.</p>
6490 <p>This entire intrinsic must be executed atomically. It first loads the value
6491 in memory pointed to by <tt>ptr</tt> and compares it with the
6492 value <tt>cmp</tt>. If they are equal, <tt>val</tt> is stored into the
6493 memory. The loaded value is yielded in all cases. This provides the
6494 equivalent of an atomic compare-and-swap operation within the SSA
6502 %val1 = add i32 4, 4
6503 %result1 = call i32 @llvm.atomic.cmp.swap.i32.p0i32( i32* %ptr, i32 4, %val1 )
6504 <i>; yields {i32}:result1 = 4</i>
6505 %stored1 = icmp eq i32 %result1, 4 <i>; yields {i1}:stored1 = true</i>
6506 %memval1 = load i32* %ptr <i>; yields {i32}:memval1 = 8</i>
6508 %val2 = add i32 1, 1
6509 %result2 = call i32 @llvm.atomic.cmp.swap.i32.p0i32( i32* %ptr, i32 5, %val2 )
6510 <i>; yields {i32}:result2 = 8</i>
6511 %stored2 = icmp eq i32 %result2, 5 <i>; yields {i1}:stored2 = false</i>
6513 %memval2 = load i32* %ptr <i>; yields {i32}:memval2 = 8</i>
6518 <!-- _______________________________________________________________________ -->
6519 <div class="doc_subsubsection">
6520 <a name="int_atomic_swap">'<tt>llvm.atomic.swap.*</tt>' Intrinsic</a>
6522 <div class="doc_text">
6525 <p>This is an overloaded intrinsic. You can use <tt>llvm.atomic.swap</tt> on any
6526 integer bit width. Not all targets support all bit widths however.</p>
6529 declare i8 @llvm.atomic.swap.i8.p0i8( i8* <ptr>, i8 <val> )
6530 declare i16 @llvm.atomic.swap.i16.p0i16( i16* <ptr>, i16 <val> )
6531 declare i32 @llvm.atomic.swap.i32.p0i32( i32* <ptr>, i32 <val> )
6532 declare i64 @llvm.atomic.swap.i64.p0i64( i64* <ptr>, i64 <val> )
6536 <p>This intrinsic loads the value stored in memory at <tt>ptr</tt> and yields
6537 the value from memory. It then stores the value in <tt>val</tt> in the memory
6538 at <tt>ptr</tt>.</p>
6541 <p>The <tt>llvm.atomic.swap</tt> intrinsic takes two arguments. Both
6542 the <tt>val</tt> argument and the result must be integers of the same bit
6543 width. The first argument, <tt>ptr</tt>, must be a pointer to a value of this
6544 integer type. The targets may only lower integer representations they
6548 <p>This intrinsic loads the value pointed to by <tt>ptr</tt>, yields it, and
6549 stores <tt>val</tt> back into <tt>ptr</tt> atomically. This provides the
6550 equivalent of an atomic swap operation within the SSA framework.</p>
6557 %val1 = add i32 4, 4
6558 %result1 = call i32 @llvm.atomic.swap.i32.p0i32( i32* %ptr, i32 %val1 )
6559 <i>; yields {i32}:result1 = 4</i>
6560 %stored1 = icmp eq i32 %result1, 4 <i>; yields {i1}:stored1 = true</i>
6561 %memval1 = load i32* %ptr <i>; yields {i32}:memval1 = 8</i>
6563 %val2 = add i32 1, 1
6564 %result2 = call i32 @llvm.atomic.swap.i32.p0i32( i32* %ptr, i32 %val2 )
6565 <i>; yields {i32}:result2 = 8</i>
6567 %stored2 = icmp eq i32 %result2, 8 <i>; yields {i1}:stored2 = true</i>
6568 %memval2 = load i32* %ptr <i>; yields {i32}:memval2 = 2</i>
6573 <!-- _______________________________________________________________________ -->
6574 <div class="doc_subsubsection">
6575 <a name="int_atomic_load_add">'<tt>llvm.atomic.load.add.*</tt>' Intrinsic</a>
6579 <div class="doc_text">
6582 <p>This is an overloaded intrinsic. You can use <tt>llvm.atomic.load.add</tt> on
6583 any integer bit width. Not all targets support all bit widths however.</p>
6586 declare i8 @llvm.atomic.load.add.i8..p0i8( i8* <ptr>, i8 <delta> )
6587 declare i16 @llvm.atomic.load.add.i16..p0i16( i16* <ptr>, i16 <delta> )
6588 declare i32 @llvm.atomic.load.add.i32..p0i32( i32* <ptr>, i32 <delta> )
6589 declare i64 @llvm.atomic.load.add.i64..p0i64( i64* <ptr>, i64 <delta> )
6593 <p>This intrinsic adds <tt>delta</tt> to the value stored in memory
6594 at <tt>ptr</tt>. It yields the original value at <tt>ptr</tt>.</p>
6597 <p>The intrinsic takes two arguments, the first a pointer to an integer value
6598 and the second an integer value. The result is also an integer value. These
6599 integer types can have any bit width, but they must all have the same bit
6600 width. The targets may only lower integer representations they support.</p>
6603 <p>This intrinsic does a series of operations atomically. It first loads the
6604 value stored at <tt>ptr</tt>. It then adds <tt>delta</tt>, stores the result
6605 to <tt>ptr</tt>. It yields the original value stored at <tt>ptr</tt>.</p>
6611 %result1 = call i32 @llvm.atomic.load.add.i32.p0i32( i32* %ptr, i32 4 )
6612 <i>; yields {i32}:result1 = 4</i>
6613 %result2 = call i32 @llvm.atomic.load.add.i32.p0i32( i32* %ptr, i32 2 )
6614 <i>; yields {i32}:result2 = 8</i>
6615 %result3 = call i32 @llvm.atomic.load.add.i32.p0i32( i32* %ptr, i32 5 )
6616 <i>; yields {i32}:result3 = 10</i>
6617 %memval1 = load i32* %ptr <i>; yields {i32}:memval1 = 15</i>
6622 <!-- _______________________________________________________________________ -->
6623 <div class="doc_subsubsection">
6624 <a name="int_atomic_load_sub">'<tt>llvm.atomic.load.sub.*</tt>' Intrinsic</a>
6628 <div class="doc_text">
6631 <p>This is an overloaded intrinsic. You can use <tt>llvm.atomic.load.sub</tt> on
6632 any integer bit width and for different address spaces. Not all targets
6633 support all bit widths however.</p>
6636 declare i8 @llvm.atomic.load.sub.i8.p0i32( i8* <ptr>, i8 <delta> )
6637 declare i16 @llvm.atomic.load.sub.i16.p0i32( i16* <ptr>, i16 <delta> )
6638 declare i32 @llvm.atomic.load.sub.i32.p0i32( i32* <ptr>, i32 <delta> )
6639 declare i64 @llvm.atomic.load.sub.i64.p0i32( i64* <ptr>, i64 <delta> )
6643 <p>This intrinsic subtracts <tt>delta</tt> to the value stored in memory at
6644 <tt>ptr</tt>. It yields the original value at <tt>ptr</tt>.</p>
6647 <p>The intrinsic takes two arguments, the first a pointer to an integer value
6648 and the second an integer value. The result is also an integer value. These
6649 integer types can have any bit width, but they must all have the same bit
6650 width. The targets may only lower integer representations they support.</p>
6653 <p>This intrinsic does a series of operations atomically. It first loads the
6654 value stored at <tt>ptr</tt>. It then subtracts <tt>delta</tt>, stores the
6655 result to <tt>ptr</tt>. It yields the original value stored
6656 at <tt>ptr</tt>.</p>
6662 %result1 = call i32 @llvm.atomic.load.sub.i32.p0i32( i32* %ptr, i32 4 )
6663 <i>; yields {i32}:result1 = 8</i>
6664 %result2 = call i32 @llvm.atomic.load.sub.i32.p0i32( i32* %ptr, i32 2 )
6665 <i>; yields {i32}:result2 = 4</i>
6666 %result3 = call i32 @llvm.atomic.load.sub.i32.p0i32( i32* %ptr, i32 5 )
6667 <i>; yields {i32}:result3 = 2</i>
6668 %memval1 = load i32* %ptr <i>; yields {i32}:memval1 = -3</i>
6673 <!-- _______________________________________________________________________ -->
6674 <div class="doc_subsubsection">
6675 <a name="int_atomic_load_and">'<tt>llvm.atomic.load.and.*</tt>' Intrinsic</a><br>
6676 <a name="int_atomic_load_nand">'<tt>llvm.atomic.load.nand.*</tt>' Intrinsic</a><br>
6677 <a name="int_atomic_load_or">'<tt>llvm.atomic.load.or.*</tt>' Intrinsic</a><br>
6678 <a name="int_atomic_load_xor">'<tt>llvm.atomic.load.xor.*</tt>' Intrinsic</a><br>
6681 <div class="doc_text">
6684 <p>These are overloaded intrinsics. You can
6685 use <tt>llvm.atomic.load_and</tt>, <tt>llvm.atomic.load_nand</tt>,
6686 <tt>llvm.atomic.load_or</tt>, and <tt>llvm.atomic.load_xor</tt> on any integer
6687 bit width and for different address spaces. Not all targets support all bit
6691 declare i8 @llvm.atomic.load.and.i8.p0i8( i8* <ptr>, i8 <delta> )
6692 declare i16 @llvm.atomic.load.and.i16.p0i16( i16* <ptr>, i16 <delta> )
6693 declare i32 @llvm.atomic.load.and.i32.p0i32( i32* <ptr>, i32 <delta> )
6694 declare i64 @llvm.atomic.load.and.i64.p0i64( i64* <ptr>, i64 <delta> )
6698 declare i8 @llvm.atomic.load.or.i8.p0i8( i8* <ptr>, i8 <delta> )
6699 declare i16 @llvm.atomic.load.or.i16.p0i16( i16* <ptr>, i16 <delta> )
6700 declare i32 @llvm.atomic.load.or.i32.p0i32( i32* <ptr>, i32 <delta> )
6701 declare i64 @llvm.atomic.load.or.i64.p0i64( i64* <ptr>, i64 <delta> )
6705 declare i8 @llvm.atomic.load.nand.i8.p0i32( i8* <ptr>, i8 <delta> )
6706 declare i16 @llvm.atomic.load.nand.i16.p0i32( i16* <ptr>, i16 <delta> )
6707 declare i32 @llvm.atomic.load.nand.i32.p0i32( i32* <ptr>, i32 <delta> )
6708 declare i64 @llvm.atomic.load.nand.i64.p0i32( i64* <ptr>, i64 <delta> )
6712 declare i8 @llvm.atomic.load.xor.i8.p0i32( i8* <ptr>, i8 <delta> )
6713 declare i16 @llvm.atomic.load.xor.i16.p0i32( i16* <ptr>, i16 <delta> )
6714 declare i32 @llvm.atomic.load.xor.i32.p0i32( i32* <ptr>, i32 <delta> )
6715 declare i64 @llvm.atomic.load.xor.i64.p0i32( i64* <ptr>, i64 <delta> )
6719 <p>These intrinsics bitwise the operation (and, nand, or, xor) <tt>delta</tt> to
6720 the value stored in memory at <tt>ptr</tt>. It yields the original value
6721 at <tt>ptr</tt>.</p>
6724 <p>These intrinsics take two arguments, the first a pointer to an integer value
6725 and the second an integer value. The result is also an integer value. These
6726 integer types can have any bit width, but they must all have the same bit
6727 width. The targets may only lower integer representations they support.</p>
6730 <p>These intrinsics does a series of operations atomically. They first load the
6731 value stored at <tt>ptr</tt>. They then do the bitwise
6732 operation <tt>delta</tt>, store the result to <tt>ptr</tt>. They yield the
6733 original value stored at <tt>ptr</tt>.</p>
6738 store i32 0x0F0F, %ptr
6739 %result0 = call i32 @llvm.atomic.load.nand.i32.p0i32( i32* %ptr, i32 0xFF )
6740 <i>; yields {i32}:result0 = 0x0F0F</i>
6741 %result1 = call i32 @llvm.atomic.load.and.i32.p0i32( i32* %ptr, i32 0xFF )
6742 <i>; yields {i32}:result1 = 0xFFFFFFF0</i>
6743 %result2 = call i32 @llvm.atomic.load.or.i32.p0i32( i32* %ptr, i32 0F )
6744 <i>; yields {i32}:result2 = 0xF0</i>
6745 %result3 = call i32 @llvm.atomic.load.xor.i32.p0i32( i32* %ptr, i32 0F )
6746 <i>; yields {i32}:result3 = FF</i>
6747 %memval1 = load i32* %ptr <i>; yields {i32}:memval1 = F0</i>
6752 <!-- _______________________________________________________________________ -->
6753 <div class="doc_subsubsection">
6754 <a name="int_atomic_load_max">'<tt>llvm.atomic.load.max.*</tt>' Intrinsic</a><br>
6755 <a name="int_atomic_load_min">'<tt>llvm.atomic.load.min.*</tt>' Intrinsic</a><br>
6756 <a name="int_atomic_load_umax">'<tt>llvm.atomic.load.umax.*</tt>' Intrinsic</a><br>
6757 <a name="int_atomic_load_umin">'<tt>llvm.atomic.load.umin.*</tt>' Intrinsic</a><br>
6760 <div class="doc_text">
6763 <p>These are overloaded intrinsics. You can use <tt>llvm.atomic.load_max</tt>,
6764 <tt>llvm.atomic.load_min</tt>, <tt>llvm.atomic.load_umax</tt>, and
6765 <tt>llvm.atomic.load_umin</tt> on any integer bit width and for different
6766 address spaces. Not all targets support all bit widths however.</p>
6769 declare i8 @llvm.atomic.load.max.i8.p0i8( i8* <ptr>, i8 <delta> )
6770 declare i16 @llvm.atomic.load.max.i16.p0i16( i16* <ptr>, i16 <delta> )
6771 declare i32 @llvm.atomic.load.max.i32.p0i32( i32* <ptr>, i32 <delta> )
6772 declare i64 @llvm.atomic.load.max.i64.p0i64( i64* <ptr>, i64 <delta> )
6776 declare i8 @llvm.atomic.load.min.i8.p0i8( i8* <ptr>, i8 <delta> )
6777 declare i16 @llvm.atomic.load.min.i16.p0i16( i16* <ptr>, i16 <delta> )
6778 declare i32 @llvm.atomic.load.min.i32..p0i32( i32* <ptr>, i32 <delta> )
6779 declare i64 @llvm.atomic.load.min.i64..p0i64( i64* <ptr>, i64 <delta> )
6783 declare i8 @llvm.atomic.load.umax.i8.p0i8( i8* <ptr>, i8 <delta> )
6784 declare i16 @llvm.atomic.load.umax.i16.p0i16( i16* <ptr>, i16 <delta> )
6785 declare i32 @llvm.atomic.load.umax.i32.p0i32( i32* <ptr>, i32 <delta> )
6786 declare i64 @llvm.atomic.load.umax.i64.p0i64( i64* <ptr>, i64 <delta> )
6790 declare i8 @llvm.atomic.load.umin.i8..p0i8( i8* <ptr>, i8 <delta> )
6791 declare i16 @llvm.atomic.load.umin.i16.p0i16( i16* <ptr>, i16 <delta> )
6792 declare i32 @llvm.atomic.load.umin.i32..p0i32( i32* <ptr>, i32 <delta> )
6793 declare i64 @llvm.atomic.load.umin.i64..p0i64( i64* <ptr>, i64 <delta> )
6797 <p>These intrinsics takes the signed or unsigned minimum or maximum of
6798 <tt>delta</tt> and the value stored in memory at <tt>ptr</tt>. It yields the
6799 original value at <tt>ptr</tt>.</p>
6802 <p>These intrinsics take two arguments, the first a pointer to an integer value
6803 and the second an integer value. The result is also an integer value. These
6804 integer types can have any bit width, but they must all have the same bit
6805 width. The targets may only lower integer representations they support.</p>
6808 <p>These intrinsics does a series of operations atomically. They first load the
6809 value stored at <tt>ptr</tt>. They then do the signed or unsigned min or
6810 max <tt>delta</tt> and the value, store the result to <tt>ptr</tt>. They
6811 yield the original value stored at <tt>ptr</tt>.</p>
6817 %result0 = call i32 @llvm.atomic.load.min.i32.p0i32( i32* %ptr, i32 -2 )
6818 <i>; yields {i32}:result0 = 7</i>
6819 %result1 = call i32 @llvm.atomic.load.max.i32.p0i32( i32* %ptr, i32 8 )
6820 <i>; yields {i32}:result1 = -2</i>
6821 %result2 = call i32 @llvm.atomic.load.umin.i32.p0i32( i32* %ptr, i32 10 )
6822 <i>; yields {i32}:result2 = 8</i>
6823 %result3 = call i32 @llvm.atomic.load.umax.i32.p0i32( i32* %ptr, i32 30 )
6824 <i>; yields {i32}:result3 = 8</i>
6825 %memval1 = load i32* %ptr <i>; yields {i32}:memval1 = 30</i>
6830 <!-- ======================================================================= -->
6831 <div class="doc_subsection">
6832 <a name="int_general">General Intrinsics</a>
6835 <div class="doc_text">
6837 <p>This class of intrinsics is designed to be generic and has no specific
6842 <!-- _______________________________________________________________________ -->
6843 <div class="doc_subsubsection">
6844 <a name="int_var_annotation">'<tt>llvm.var.annotation</tt>' Intrinsic</a>
6847 <div class="doc_text">
6851 declare void @llvm.var.annotation(i8* <val>, i8* <str>, i8* <str>, i32 <int> )
6855 <p>The '<tt>llvm.var.annotation</tt>' intrinsic.</p>
6858 <p>The first argument is a pointer to a value, the second is a pointer to a
6859 global string, the third is a pointer to a global string which is the source
6860 file name, and the last argument is the line number.</p>
6863 <p>This intrinsic allows annotation of local variables with arbitrary strings.
6864 This can be useful for special purpose optimizations that want to look for
6865 these annotations. These have no other defined use, they are ignored by code
6866 generation and optimization.</p>
6870 <!-- _______________________________________________________________________ -->
6871 <div class="doc_subsubsection">
6872 <a name="int_annotation">'<tt>llvm.annotation.*</tt>' Intrinsic</a>
6875 <div class="doc_text">
6878 <p>This is an overloaded intrinsic. You can use '<tt>llvm.annotation</tt>' on
6879 any integer bit width.</p>
6882 declare i8 @llvm.annotation.i8(i8 <val>, i8* <str>, i8* <str>, i32 <int> )
6883 declare i16 @llvm.annotation.i16(i16 <val>, i8* <str>, i8* <str>, i32 <int> )
6884 declare i32 @llvm.annotation.i32(i32 <val>, i8* <str>, i8* <str>, i32 <int> )
6885 declare i64 @llvm.annotation.i64(i64 <val>, i8* <str>, i8* <str>, i32 <int> )
6886 declare i256 @llvm.annotation.i256(i256 <val>, i8* <str>, i8* <str>, i32 <int> )
6890 <p>The '<tt>llvm.annotation</tt>' intrinsic.</p>
6893 <p>The first argument is an integer value (result of some expression), the
6894 second is a pointer to a global string, the third is a pointer to a global
6895 string which is the source file name, and the last argument is the line
6896 number. It returns the value of the first argument.</p>
6899 <p>This intrinsic allows annotations to be put on arbitrary expressions with
6900 arbitrary strings. This can be useful for special purpose optimizations that
6901 want to look for these annotations. These have no other defined use, they
6902 are ignored by code generation and optimization.</p>
6906 <!-- _______________________________________________________________________ -->
6907 <div class="doc_subsubsection">
6908 <a name="int_trap">'<tt>llvm.trap</tt>' Intrinsic</a>
6911 <div class="doc_text">
6915 declare void @llvm.trap()
6919 <p>The '<tt>llvm.trap</tt>' intrinsic.</p>
6925 <p>This intrinsics is lowered to the target dependent trap instruction. If the
6926 target does not have a trap instruction, this intrinsic will be lowered to
6927 the call of the <tt>abort()</tt> function.</p>
6931 <!-- _______________________________________________________________________ -->
6932 <div class="doc_subsubsection">
6933 <a name="int_stackprotector">'<tt>llvm.stackprotector</tt>' Intrinsic</a>
6936 <div class="doc_text">
6940 declare void @llvm.stackprotector( i8* <guard>, i8** <slot> )
6944 <p>The <tt>llvm.stackprotector</tt> intrinsic takes the <tt>guard</tt> and
6945 stores it onto the stack at <tt>slot</tt>. The stack slot is adjusted to
6946 ensure that it is placed on the stack before local variables.</p>
6949 <p>The <tt>llvm.stackprotector</tt> intrinsic requires two pointer
6950 arguments. The first argument is the value loaded from the stack
6951 guard <tt>@__stack_chk_guard</tt>. The second variable is an <tt>alloca</tt>
6952 that has enough space to hold the value of the guard.</p>
6955 <p>This intrinsic causes the prologue/epilogue inserter to force the position of
6956 the <tt>AllocaInst</tt> stack slot to be before local variables on the
6957 stack. This is to ensure that if a local variable on the stack is
6958 overwritten, it will destroy the value of the guard. When the function exits,
6959 the guard on the stack is checked against the original guard. If they're
6960 different, then the program aborts by calling the <tt>__stack_chk_fail()</tt>
6965 <!-- *********************************************************************** -->
6968 <a href="http://jigsaw.w3.org/css-validator/check/referer"><img
6969 src="http://jigsaw.w3.org/css-validator/images/vcss-blue" alt="Valid CSS"></a>
6970 <a href="http://validator.w3.org/check/referer"><img
6971 src="http://www.w3.org/Icons/valid-html401-blue" alt="Valid HTML 4.01"></a>
6973 <a href="mailto:sabre@nondot.org">Chris Lattner</a><br>
6974 <a href="http://llvm.org">The LLVM Compiler Infrastructure</a><br>
6975 Last modified: $Date$