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. Note that (unlike private
533 symbols) linker_private symbols are subject to coalescing by the linker:
534 weak symbols get merged and redefinitions are rejected. However, unlike
535 normal strong symbols, they are removed by the linker from the final
536 linked image (executable or dynamic library).</dd>
538 <dt><tt><b><a name="linkage_internal">internal</a></b></tt>: </dt>
539 <dd>Similar to private, but the value shows as a local symbol
540 (<tt>STB_LOCAL</tt> in the case of ELF) in the object file. This
541 corresponds to the notion of the '<tt>static</tt>' keyword in C.</dd>
543 <dt><tt><b><a name="linkage_available_externally">available_externally</a></b></tt>: </dt>
544 <dd>Globals with "<tt>available_externally</tt>" linkage are never emitted
545 into the object file corresponding to the LLVM module. They exist to
546 allow inlining and other optimizations to take place given knowledge of
547 the definition of the global, which is known to be somewhere outside the
548 module. Globals with <tt>available_externally</tt> linkage are allowed to
549 be discarded at will, and are otherwise the same as <tt>linkonce_odr</tt>.
550 This linkage type is only allowed on definitions, not declarations.</dd>
552 <dt><tt><b><a name="linkage_linkonce">linkonce</a></b></tt>: </dt>
553 <dd>Globals with "<tt>linkonce</tt>" linkage are merged with other globals of
554 the same name when linkage occurs. This is typically used to implement
555 inline functions, templates, or other code which must be generated in each
556 translation unit that uses it. Unreferenced <tt>linkonce</tt> globals are
557 allowed to be discarded.</dd>
559 <dt><tt><b><a name="linkage_weak">weak</a></b></tt>: </dt>
560 <dd>"<tt>weak</tt>" linkage has the same merging semantics as
561 <tt>linkonce</tt> linkage, except that unreferenced globals with
562 <tt>weak</tt> linkage may not be discarded. This is used for globals that
563 are declared "weak" in C source code.</dd>
565 <dt><tt><b><a name="linkage_common">common</a></b></tt>: </dt>
566 <dd>"<tt>common</tt>" linkage is most similar to "<tt>weak</tt>" linkage, but
567 they are used for tentative definitions in C, such as "<tt>int X;</tt>" at
569 Symbols with "<tt>common</tt>" linkage are merged in the same way as
570 <tt>weak symbols</tt>, and they may not be deleted if unreferenced.
571 <tt>common</tt> symbols may not have an explicit section,
572 must have a zero initializer, and may not be marked '<a
573 href="#globalvars"><tt>constant</tt></a>'. Functions and aliases may not
574 have common linkage.</dd>
577 <dt><tt><b><a name="linkage_appending">appending</a></b></tt>: </dt>
578 <dd>"<tt>appending</tt>" linkage may only be applied to global variables of
579 pointer to array type. When two global variables with appending linkage
580 are linked together, the two global arrays are appended together. This is
581 the LLVM, typesafe, equivalent of having the system linker append together
582 "sections" with identical names when .o files are linked.</dd>
584 <dt><tt><b><a name="linkage_externweak">extern_weak</a></b></tt>: </dt>
585 <dd>The semantics of this linkage follow the ELF object file model: the symbol
586 is weak until linked, if not linked, the symbol becomes null instead of
587 being an undefined reference.</dd>
589 <dt><tt><b><a name="linkage_linkonce">linkonce_odr</a></b></tt>: </dt>
590 <dt><tt><b><a name="linkage_weak">weak_odr</a></b></tt>: </dt>
591 <dd>Some languages allow differing globals to be merged, such as two functions
592 with different semantics. Other languages, such as <tt>C++</tt>, ensure
593 that only equivalent globals are ever merged (the "one definition rule" -
594 "ODR"). Such languages can use the <tt>linkonce_odr</tt>
595 and <tt>weak_odr</tt> linkage types to indicate that the global will only
596 be merged with equivalent globals. These linkage types are otherwise the
597 same as their non-<tt>odr</tt> versions.</dd>
599 <dt><tt><b><a name="linkage_external">externally visible</a></b></tt>:</dt>
600 <dd>If none of the above identifiers are used, the global is externally
601 visible, meaning that it participates in linkage and can be used to
602 resolve external symbol references.</dd>
605 <p>The next two types of linkage are targeted for Microsoft Windows platform
606 only. They are designed to support importing (exporting) symbols from (to)
607 DLLs (Dynamic Link Libraries).</p>
610 <dt><tt><b><a name="linkage_dllimport">dllimport</a></b></tt>: </dt>
611 <dd>"<tt>dllimport</tt>" linkage causes the compiler to reference a function
612 or variable via a global pointer to a pointer that is set up by the DLL
613 exporting the symbol. On Microsoft Windows targets, the pointer name is
614 formed by combining <code>__imp_</code> and the function or variable
617 <dt><tt><b><a name="linkage_dllexport">dllexport</a></b></tt>: </dt>
618 <dd>"<tt>dllexport</tt>" linkage causes the compiler to provide a global
619 pointer to a pointer in a DLL, so that it can be referenced with the
620 <tt>dllimport</tt> attribute. On Microsoft Windows targets, the pointer
621 name is formed by combining <code>__imp_</code> and the function or
625 <p>For example, since the "<tt>.LC0</tt>" variable is defined to be internal, if
626 another module defined a "<tt>.LC0</tt>" variable and was linked with this
627 one, one of the two would be renamed, preventing a collision. Since
628 "<tt>main</tt>" and "<tt>puts</tt>" are external (i.e., lacking any linkage
629 declarations), they are accessible outside of the current module.</p>
631 <p>It is illegal for a function <i>declaration</i> to have any linkage type
632 other than "externally visible", <tt>dllimport</tt>
633 or <tt>extern_weak</tt>.</p>
635 <p>Aliases can have only <tt>external</tt>, <tt>internal</tt>, <tt>weak</tt>
636 or <tt>weak_odr</tt> linkages.</p>
640 <!-- ======================================================================= -->
641 <div class="doc_subsection">
642 <a name="callingconv">Calling Conventions</a>
645 <div class="doc_text">
647 <p>LLVM <a href="#functionstructure">functions</a>, <a href="#i_call">calls</a>
648 and <a href="#i_invoke">invokes</a> can all have an optional calling
649 convention specified for the call. The calling convention of any pair of
650 dynamic caller/callee must match, or the behavior of the program is
651 undefined. The following calling conventions are supported by LLVM, and more
652 may be added in the future:</p>
655 <dt><b>"<tt>ccc</tt>" - The C calling convention</b>:</dt>
656 <dd>This calling convention (the default if no other calling convention is
657 specified) matches the target C calling conventions. This calling
658 convention supports varargs function calls and tolerates some mismatch in
659 the declared prototype and implemented declaration of the function (as
662 <dt><b>"<tt>fastcc</tt>" - The fast calling convention</b>:</dt>
663 <dd>This calling convention attempts to make calls as fast as possible
664 (e.g. by passing things in registers). This calling convention allows the
665 target to use whatever tricks it wants to produce fast code for the
666 target, without having to conform to an externally specified ABI
667 (Application Binary Interface). Implementations of this convention should
668 allow arbitrary <a href="CodeGenerator.html#tailcallopt">tail call
669 optimization</a> to be supported. This calling convention does not
670 support varargs and requires the prototype of all callees to exactly match
671 the prototype of the function definition.</dd>
673 <dt><b>"<tt>coldcc</tt>" - The cold calling convention</b>:</dt>
674 <dd>This calling convention attempts to make code in the caller as efficient
675 as possible under the assumption that the call is not commonly executed.
676 As such, these calls often preserve all registers so that the call does
677 not break any live ranges in the caller side. This calling convention
678 does not support varargs and requires the prototype of all callees to
679 exactly match the prototype of the function definition.</dd>
681 <dt><b>"<tt>cc <<em>n</em>></tt>" - Numbered convention</b>:</dt>
682 <dd>Any calling convention may be specified by number, allowing
683 target-specific calling conventions to be used. Target specific calling
684 conventions start at 64.</dd>
687 <p>More calling conventions can be added/defined on an as-needed basis, to
688 support Pascal conventions or any other well-known target-independent
693 <!-- ======================================================================= -->
694 <div class="doc_subsection">
695 <a name="visibility">Visibility Styles</a>
698 <div class="doc_text">
700 <p>All Global Variables and Functions have one of the following visibility
704 <dt><b>"<tt>default</tt>" - Default style</b>:</dt>
705 <dd>On targets that use the ELF object file format, default visibility means
706 that the declaration is visible to other modules and, in shared libraries,
707 means that the declared entity may be overridden. On Darwin, default
708 visibility means that the declaration is visible to other modules. Default
709 visibility corresponds to "external linkage" in the language.</dd>
711 <dt><b>"<tt>hidden</tt>" - Hidden style</b>:</dt>
712 <dd>Two declarations of an object with hidden visibility refer to the same
713 object if they are in the same shared object. Usually, hidden visibility
714 indicates that the symbol will not be placed into the dynamic symbol
715 table, so no other module (executable or shared library) can reference it
718 <dt><b>"<tt>protected</tt>" - Protected style</b>:</dt>
719 <dd>On ELF, protected visibility indicates that the symbol will be placed in
720 the dynamic symbol table, but that references within the defining module
721 will bind to the local symbol. That is, the symbol cannot be overridden by
727 <!-- ======================================================================= -->
728 <div class="doc_subsection">
729 <a name="namedtypes">Named Types</a>
732 <div class="doc_text">
734 <p>LLVM IR allows you to specify name aliases for certain types. This can make
735 it easier to read the IR and make the IR more condensed (particularly when
736 recursive types are involved). An example of a name specification is:</p>
738 <div class="doc_code">
740 %mytype = type { %mytype*, i32 }
744 <p>You may give a name to any <a href="#typesystem">type</a> except
745 "<a href="t_void">void</a>". Type name aliases may be used anywhere a type
746 is expected with the syntax "%mytype".</p>
748 <p>Note that type names are aliases for the structural type that they indicate,
749 and that you can therefore specify multiple names for the same type. This
750 often leads to confusing behavior when dumping out a .ll file. Since LLVM IR
751 uses structural typing, the name is not part of the type. When printing out
752 LLVM IR, the printer will pick <em>one name</em> to render all types of a
753 particular shape. This means that if you have code where two different
754 source types end up having the same LLVM type, that the dumper will sometimes
755 print the "wrong" or unexpected type. This is an important design point and
756 isn't going to change.</p>
760 <!-- ======================================================================= -->
761 <div class="doc_subsection">
762 <a name="globalvars">Global Variables</a>
765 <div class="doc_text">
767 <p>Global variables define regions of memory allocated at compilation time
768 instead of run-time. Global variables may optionally be initialized, may
769 have an explicit section to be placed in, and may have an optional explicit
770 alignment specified. A variable may be defined as "thread_local", which
771 means that it will not be shared by threads (each thread will have a
772 separated copy of the variable). A variable may be defined as a global
773 "constant," which indicates that the contents of the variable
774 will <b>never</b> be modified (enabling better optimization, allowing the
775 global data to be placed in the read-only section of an executable, etc).
776 Note that variables that need runtime initialization cannot be marked
777 "constant" as there is a store to the variable.</p>
779 <p>LLVM explicitly allows <em>declarations</em> of global variables to be marked
780 constant, even if the final definition of the global is not. This capability
781 can be used to enable slightly better optimization of the program, but
782 requires the language definition to guarantee that optimizations based on the
783 'constantness' are valid for the translation units that do not include the
786 <p>As SSA values, global variables define pointer values that are in scope
787 (i.e. they dominate) all basic blocks in the program. Global variables
788 always define a pointer to their "content" type because they describe a
789 region of memory, and all memory objects in LLVM are accessed through
792 <p>A global variable may be declared to reside in a target-specific numbered
793 address space. For targets that support them, address spaces may affect how
794 optimizations are performed and/or what target instructions are used to
795 access the variable. The default address space is zero. The address space
796 qualifier must precede any other attributes.</p>
798 <p>LLVM allows an explicit section to be specified for globals. If the target
799 supports it, it will emit globals to the section specified.</p>
801 <p>An explicit alignment may be specified for a global. If not present, or if
802 the alignment is set to zero, the alignment of the global is set by the
803 target to whatever it feels convenient. If an explicit alignment is
804 specified, the global is forced to have at least that much alignment. All
805 alignments must be a power of 2.</p>
807 <p>For example, the following defines a global in a numbered address space with
808 an initializer, section, and alignment:</p>
810 <div class="doc_code">
812 @G = addrspace(5) constant float 1.0, section "foo", align 4
819 <!-- ======================================================================= -->
820 <div class="doc_subsection">
821 <a name="functionstructure">Functions</a>
824 <div class="doc_text">
826 <p>LLVM function definitions consist of the "<tt>define</tt>" keyord, an
827 optional <a href="#linkage">linkage type</a>, an optional
828 <a href="#visibility">visibility style</a>, an optional
829 <a href="#callingconv">calling convention</a>, a return type, an optional
830 <a href="#paramattrs">parameter attribute</a> for the return type, a function
831 name, a (possibly empty) argument list (each with optional
832 <a href="#paramattrs">parameter attributes</a>), optional
833 <a href="#fnattrs">function attributes</a>, an optional section, an optional
834 alignment, an optional <a href="#gc">garbage collector name</a>, an opening
835 curly brace, a list of basic blocks, and a closing curly brace.</p>
837 <p>LLVM function declarations consist of the "<tt>declare</tt>" keyword, an
838 optional <a href="#linkage">linkage type</a>, an optional
839 <a href="#visibility">visibility style</a>, an optional
840 <a href="#callingconv">calling convention</a>, a return type, an optional
841 <a href="#paramattrs">parameter attribute</a> for the return type, a function
842 name, a possibly empty list of arguments, an optional alignment, and an
843 optional <a href="#gc">garbage collector name</a>.</p>
845 <p>A function definition contains a list of basic blocks, forming the CFG
846 (Control Flow Graph) for the function. Each basic block may optionally start
847 with a label (giving the basic block a symbol table entry), contains a list
848 of instructions, and ends with a <a href="#terminators">terminator</a>
849 instruction (such as a branch or function return).</p>
851 <p>The first basic block in a function is special in two ways: it is immediately
852 executed on entrance to the function, and it is not allowed to have
853 predecessor basic blocks (i.e. there can not be any branches to the entry
854 block of a function). Because the block can have no predecessors, it also
855 cannot have any <a href="#i_phi">PHI nodes</a>.</p>
857 <p>LLVM allows an explicit section to be specified for functions. If the target
858 supports it, it will emit functions to the section specified.</p>
860 <p>An explicit alignment may be specified for a function. If not present, or if
861 the alignment is set to zero, the alignment of the function is set by the
862 target to whatever it feels convenient. If an explicit alignment is
863 specified, the function is forced to have at least that much alignment. All
864 alignments must be a power of 2.</p>
867 <div class="doc_code">
869 define [<a href="#linkage">linkage</a>] [<a href="#visibility">visibility</a>]
870 [<a href="#callingconv">cconv</a>] [<a href="#paramattrs">ret attrs</a>]
871 <ResultType> @<FunctionName> ([argument list])
872 [<a href="#fnattrs">fn Attrs</a>] [section "name"] [align N]
873 [<a href="#gc">gc</a>] { ... }
879 <!-- ======================================================================= -->
880 <div class="doc_subsection">
881 <a name="aliasstructure">Aliases</a>
884 <div class="doc_text">
886 <p>Aliases act as "second name" for the aliasee value (which can be either
887 function, global variable, another alias or bitcast of global value). Aliases
888 may have an optional <a href="#linkage">linkage type</a>, and an
889 optional <a href="#visibility">visibility style</a>.</p>
892 <div class="doc_code">
894 @<Name> = alias [Linkage] [Visibility] <AliaseeTy> @<Aliasee>
900 <!-- ======================================================================= -->
901 <div class="doc_subsection"><a name="paramattrs">Parameter Attributes</a></div>
903 <div class="doc_text">
905 <p>The return type and each parameter of a function type may have a set of
906 <i>parameter attributes</i> associated with them. Parameter attributes are
907 used to communicate additional information about the result or parameters of
908 a function. Parameter attributes are considered to be part of the function,
909 not of the function type, so functions with different parameter attributes
910 can have the same function type.</p>
912 <p>Parameter attributes are simple keywords that follow the type specified. If
913 multiple parameter attributes are needed, they are space separated. For
916 <div class="doc_code">
918 declare i32 @printf(i8* noalias nocapture, ...)
919 declare i32 @atoi(i8 zeroext)
920 declare signext i8 @returns_signed_char()
924 <p>Note that any attributes for the function result (<tt>nounwind</tt>,
925 <tt>readonly</tt>) come immediately after the argument list.</p>
927 <p>Currently, only the following parameter attributes are defined:</p>
930 <dt><tt>zeroext</tt></dt>
931 <dd>This indicates to the code generator that the parameter or return value
932 should be zero-extended to a 32-bit value by the caller (for a parameter)
933 or the callee (for a return value).</dd>
935 <dt><tt>signext</tt></dt>
936 <dd>This indicates to the code generator that the parameter or return value
937 should be sign-extended to a 32-bit value by the caller (for a parameter)
938 or the callee (for a return value).</dd>
940 <dt><tt>inreg</tt></dt>
941 <dd>This indicates that this parameter or return value should be treated in a
942 special target-dependent fashion during while emitting code for a function
943 call or return (usually, by putting it in a register as opposed to memory,
944 though some targets use it to distinguish between two different kinds of
945 registers). Use of this attribute is target-specific.</dd>
947 <dt><tt><a name="byval">byval</a></tt></dt>
948 <dd>This indicates that the pointer parameter should really be passed by value
949 to the function. The attribute implies that a hidden copy of the pointee
950 is made between the caller and the callee, so the callee is unable to
951 modify the value in the callee. This attribute is only valid on LLVM
952 pointer arguments. It is generally used to pass structs and arrays by
953 value, but is also valid on pointers to scalars. The copy is considered
954 to belong to the caller not the callee (for example,
955 <tt><a href="#readonly">readonly</a></tt> functions should not write to
956 <tt>byval</tt> parameters). This is not a valid attribute for return
957 values. The byval attribute also supports specifying an alignment with
958 the align attribute. This has a target-specific effect on the code
959 generator that usually indicates a desired alignment for the synthesized
962 <dt><tt>sret</tt></dt>
963 <dd>This indicates that the pointer parameter specifies the address of a
964 structure that is the return value of the function in the source program.
965 This pointer must be guaranteed by the caller to be valid: loads and
966 stores to the structure may be assumed by the callee to not to trap. This
967 may only be applied to the first parameter. This is not a valid attribute
968 for return values. </dd>
970 <dt><tt>noalias</tt></dt>
971 <dd>This indicates that the pointer does not alias any global or any other
972 parameter. The caller is responsible for ensuring that this is the
973 case. On a function return value, <tt>noalias</tt> additionally indicates
974 that the pointer does not alias any other pointers visible to the
975 caller. For further details, please see the discussion of the NoAlias
977 <a href="http://llvm.org/docs/AliasAnalysis.html#MustMayNo">alias
980 <dt><tt>nocapture</tt></dt>
981 <dd>This indicates that the callee does not make any copies of the pointer
982 that outlive the callee itself. This is not a valid attribute for return
985 <dt><tt>nest</tt></dt>
986 <dd>This indicates that the pointer parameter can be excised using the
987 <a href="#int_trampoline">trampoline intrinsics</a>. This is not a valid
988 attribute for return values.</dd>
993 <!-- ======================================================================= -->
994 <div class="doc_subsection">
995 <a name="gc">Garbage Collector Names</a>
998 <div class="doc_text">
1000 <p>Each function may specify a garbage collector name, which is simply a
1003 <div class="doc_code">
1005 define void @f() gc "name" { ...
1009 <p>The compiler declares the supported values of <i>name</i>. Specifying a
1010 collector which will cause the compiler to alter its output in order to
1011 support the named garbage collection algorithm.</p>
1015 <!-- ======================================================================= -->
1016 <div class="doc_subsection">
1017 <a name="fnattrs">Function Attributes</a>
1020 <div class="doc_text">
1022 <p>Function attributes are set to communicate additional information about a
1023 function. Function attributes are considered to be part of the function, not
1024 of the function type, so functions with different parameter attributes can
1025 have the same function type.</p>
1027 <p>Function attributes are simple keywords that follow the type specified. If
1028 multiple attributes are needed, they are space separated. For example:</p>
1030 <div class="doc_code">
1032 define void @f() noinline { ... }
1033 define void @f() alwaysinline { ... }
1034 define void @f() alwaysinline optsize { ... }
1035 define void @f() optsize
1040 <dt><tt>alwaysinline</tt></dt>
1041 <dd>This attribute indicates that the inliner should attempt to inline this
1042 function into callers whenever possible, ignoring any active inlining size
1043 threshold for this caller.</dd>
1045 <dt><tt>noinline</tt></dt>
1046 <dd>This attribute indicates that the inliner should never inline this
1047 function in any situation. This attribute may not be used together with
1048 the <tt>alwaysinline</tt> attribute.</dd>
1050 <dt><tt>optsize</tt></dt>
1051 <dd>This attribute suggests that optimization passes and code generator passes
1052 make choices that keep the code size of this function low, and otherwise
1053 do optimizations specifically to reduce code size.</dd>
1055 <dt><tt>noreturn</tt></dt>
1056 <dd>This function attribute indicates that the function never returns
1057 normally. This produces undefined behavior at runtime if the function
1058 ever does dynamically return.</dd>
1060 <dt><tt>nounwind</tt></dt>
1061 <dd>This function attribute indicates that the function never returns with an
1062 unwind or exceptional control flow. If the function does unwind, its
1063 runtime behavior is undefined.</dd>
1065 <dt><tt>readnone</tt></dt>
1066 <dd>This attribute indicates that the function computes its result (or decides
1067 to unwind an exception) based strictly on its arguments, without
1068 dereferencing any pointer arguments or otherwise accessing any mutable
1069 state (e.g. memory, control registers, etc) visible to caller functions.
1070 It does not write through any pointer arguments
1071 (including <tt><a href="#byval">byval</a></tt> arguments) and never
1072 changes any state visible to callers. This means that it cannot unwind
1073 exceptions by calling the <tt>C++</tt> exception throwing methods, but
1074 could use the <tt>unwind</tt> instruction.</dd>
1076 <dt><tt><a name="readonly">readonly</a></tt></dt>
1077 <dd>This attribute indicates that the function does not write through any
1078 pointer arguments (including <tt><a href="#byval">byval</a></tt>
1079 arguments) or otherwise modify any state (e.g. memory, control registers,
1080 etc) visible to caller functions. It may dereference pointer arguments
1081 and read state that may be set in the caller. A readonly function always
1082 returns the same value (or unwinds an exception identically) when called
1083 with the same set of arguments and global state. It cannot unwind an
1084 exception by calling the <tt>C++</tt> exception throwing methods, but may
1085 use the <tt>unwind</tt> instruction.</dd>
1087 <dt><tt><a name="ssp">ssp</a></tt></dt>
1088 <dd>This attribute indicates that the function should emit a stack smashing
1089 protector. It is in the form of a "canary"—a random value placed on
1090 the stack before the local variables that's checked upon return from the
1091 function to see if it has been overwritten. A heuristic is used to
1092 determine if a function needs stack protectors or not.<br>
1094 If a function that has an <tt>ssp</tt> attribute is inlined into a
1095 function that doesn't have an <tt>ssp</tt> attribute, then the resulting
1096 function will have an <tt>ssp</tt> attribute.</dd>
1098 <dt><tt>sspreq</tt></dt>
1099 <dd>This attribute indicates that the function should <em>always</em> emit a
1100 stack smashing protector. This overrides
1101 the <tt><a href="#ssp">ssp</a></tt> function attribute.<br>
1103 If a function that has an <tt>sspreq</tt> attribute is inlined into a
1104 function that doesn't have an <tt>sspreq</tt> attribute or which has
1105 an <tt>ssp</tt> attribute, then the resulting function will have
1106 an <tt>sspreq</tt> attribute.</dd>
1108 <dt><tt>noredzone</tt></dt>
1109 <dd>This attribute indicates that the code generator should not use a red
1110 zone, even if the target-specific ABI normally permits it.</dd>
1112 <dt><tt>noimplicitfloat</tt></dt>
1113 <dd>This attributes disables implicit floating point instructions.</dd>
1115 <dt><tt>naked</tt></dt>
1116 <dd>This attribute disables prologue / epilogue emission for the function.
1117 This can have very system-specific consequences.</dd>
1122 <!-- ======================================================================= -->
1123 <div class="doc_subsection">
1124 <a name="moduleasm">Module-Level Inline Assembly</a>
1127 <div class="doc_text">
1129 <p>Modules may contain "module-level inline asm" blocks, which corresponds to
1130 the GCC "file scope inline asm" blocks. These blocks are internally
1131 concatenated by LLVM and treated as a single unit, but may be separated in
1132 the <tt>.ll</tt> file if desired. The syntax is very simple:</p>
1134 <div class="doc_code">
1136 module asm "inline asm code goes here"
1137 module asm "more can go here"
1141 <p>The strings can contain any character by escaping non-printable characters.
1142 The escape sequence used is simply "\xx" where "xx" is the two digit hex code
1145 <p>The inline asm code is simply printed to the machine code .s file when
1146 assembly code is generated.</p>
1150 <!-- ======================================================================= -->
1151 <div class="doc_subsection">
1152 <a name="datalayout">Data Layout</a>
1155 <div class="doc_text">
1157 <p>A module may specify a target specific data layout string that specifies how
1158 data is to be laid out in memory. The syntax for the data layout is
1161 <div class="doc_code">
1163 target datalayout = "<i>layout specification</i>"
1167 <p>The <i>layout specification</i> consists of a list of specifications
1168 separated by the minus sign character ('-'). Each specification starts with
1169 a letter and may include other information after the letter to define some
1170 aspect of the data layout. The specifications accepted are as follows:</p>
1174 <dd>Specifies that the target lays out data in big-endian form. That is, the
1175 bits with the most significance have the lowest address location.</dd>
1178 <dd>Specifies that the target lays out data in little-endian form. That is,
1179 the bits with the least significance have the lowest address
1182 <dt><tt>p:<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1183 <dd>This specifies the <i>size</i> of a pointer and its <i>abi</i> and
1184 <i>preferred</i> alignments. All sizes are in bits. Specifying
1185 the <i>pref</i> alignment is optional. If omitted, the
1186 preceding <tt>:</tt> should be omitted too.</dd>
1188 <dt><tt>i<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1189 <dd>This specifies the alignment for an integer type of a given bit
1190 <i>size</i>. The value of <i>size</i> must be in the range [1,2^23).</dd>
1192 <dt><tt>v<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1193 <dd>This specifies the alignment for a vector type of a given bit
1196 <dt><tt>f<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1197 <dd>This specifies the alignment for a floating point type of a given bit
1198 <i>size</i>. The value of <i>size</i> must be either 32 (float) or 64
1201 <dt><tt>a<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1202 <dd>This specifies the alignment for an aggregate type of a given bit
1205 <dt><tt>s<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1206 <dd>This specifies the alignment for a stack object of a given bit
1210 <p>When constructing the data layout for a given target, LLVM starts with a
1211 default set of specifications which are then (possibly) overriden by the
1212 specifications in the <tt>datalayout</tt> keyword. The default specifications
1213 are given in this list:</p>
1216 <li><tt>E</tt> - big endian</li>
1217 <li><tt>p:32:64:64</tt> - 32-bit pointers with 64-bit alignment</li>
1218 <li><tt>i1:8:8</tt> - i1 is 8-bit (byte) aligned</li>
1219 <li><tt>i8:8:8</tt> - i8 is 8-bit (byte) aligned</li>
1220 <li><tt>i16:16:16</tt> - i16 is 16-bit aligned</li>
1221 <li><tt>i32:32:32</tt> - i32 is 32-bit aligned</li>
1222 <li><tt>i64:32:64</tt> - i64 has ABI alignment of 32-bits but preferred
1223 alignment of 64-bits</li>
1224 <li><tt>f32:32:32</tt> - float is 32-bit aligned</li>
1225 <li><tt>f64:64:64</tt> - double is 64-bit aligned</li>
1226 <li><tt>v64:64:64</tt> - 64-bit vector is 64-bit aligned</li>
1227 <li><tt>v128:128:128</tt> - 128-bit vector is 128-bit aligned</li>
1228 <li><tt>a0:0:1</tt> - aggregates are 8-bit aligned</li>
1229 <li><tt>s0:64:64</tt> - stack objects are 64-bit aligned</li>
1232 <p>When LLVM is determining the alignment for a given type, it uses the
1233 following rules:</p>
1236 <li>If the type sought is an exact match for one of the specifications, that
1237 specification is used.</li>
1239 <li>If no match is found, and the type sought is an integer type, then the
1240 smallest integer type that is larger than the bitwidth of the sought type
1241 is used. If none of the specifications are larger than the bitwidth then
1242 the the largest integer type is used. For example, given the default
1243 specifications above, the i7 type will use the alignment of i8 (next
1244 largest) while both i65 and i256 will use the alignment of i64 (largest
1247 <li>If no match is found, and the type sought is a vector type, then the
1248 largest vector type that is smaller than the sought vector type will be
1249 used as a fall back. This happens because <128 x double> can be
1250 implemented in terms of 64 <2 x double>, for example.</li>
1255 <!-- ======================================================================= -->
1256 <div class="doc_subsection">
1257 <a name="pointeraliasing">Pointer Aliasing Rules</a>
1260 <div class="doc_text">
1262 <p>Any memory access must be done through a pointer value associated
1263 with an address range of the memory access, otherwise the behavior
1264 is undefined. Pointer values are associated with address ranges
1265 according to the following rules:</p>
1268 <li>A pointer value formed from a
1269 <tt><a href="#i_getelementptr">getelementptr</a></tt> instruction
1270 is associated with the addresses associated with the first operand
1271 of the <tt>getelementptr</tt>.</li>
1272 <li>An address of a global variable is associated with the address
1273 range of the variable's storage.</li>
1274 <li>The result value of an allocation instruction is associated with
1275 the address range of the allocated storage.</li>
1276 <li>A null pointer in the default address-space is associated with
1278 <li>A pointer value formed by an
1279 <tt><a href="#i_inttoptr">inttoptr</a></tt> is associated with all
1280 address ranges of all pointer values that contribute (directly or
1281 indirectly) to the computation of the pointer's value.</li>
1282 <li>The result value of a
1283 <tt><a href="#i_bitcast">bitcast</a></tt> is associated with all
1284 addresses associated with the operand of the <tt>bitcast</tt>.</li>
1285 <li>An integer constant other than zero or a pointer value returned
1286 from a function not defined within LLVM may be associated with address
1287 ranges allocated through mechanisms other than those provided by
1288 LLVM. Such ranges shall not overlap with any ranges of addresses
1289 allocated by mechanisms provided by LLVM.</li>
1292 <p>LLVM IR does not associate types with memory. The result type of a
1293 <tt><a href="#i_load">load</a></tt> merely indicates the size and
1294 alignment of the memory from which to load, as well as the
1295 interpretation of the value. The first operand of a
1296 <tt><a href="#i_store">store</a></tt> similarly only indicates the size
1297 and alignment of the store.</p>
1299 <p>Consequently, type-based alias analysis, aka TBAA, aka
1300 <tt>-fstrict-aliasing</tt>, is not applicable to general unadorned
1301 LLVM IR. <a href="#metadata">Metadata</a> may be used to encode
1302 additional information which specialized optimization passes may use
1303 to implement type-based alias analysis.</p>
1307 <!-- *********************************************************************** -->
1308 <div class="doc_section"> <a name="typesystem">Type System</a> </div>
1309 <!-- *********************************************************************** -->
1311 <div class="doc_text">
1313 <p>The LLVM type system is one of the most important features of the
1314 intermediate representation. Being typed enables a number of optimizations
1315 to be performed on the intermediate representation directly, without having
1316 to do extra analyses on the side before the transformation. A strong type
1317 system makes it easier to read the generated code and enables novel analyses
1318 and transformations that are not feasible to perform on normal three address
1319 code representations.</p>
1323 <!-- ======================================================================= -->
1324 <div class="doc_subsection"> <a name="t_classifications">Type
1325 Classifications</a> </div>
1327 <div class="doc_text">
1329 <p>The types fall into a few useful classifications:</p>
1331 <table border="1" cellspacing="0" cellpadding="4">
1333 <tr><th>Classification</th><th>Types</th></tr>
1335 <td><a href="#t_integer">integer</a></td>
1336 <td><tt>i1, i2, i3, ... i8, ... i16, ... i32, ... i64, ... </tt></td>
1339 <td><a href="#t_floating">floating point</a></td>
1340 <td><tt>float, double, x86_fp80, fp128, ppc_fp128</tt></td>
1343 <td><a name="t_firstclass">first class</a></td>
1344 <td><a href="#t_integer">integer</a>,
1345 <a href="#t_floating">floating point</a>,
1346 <a href="#t_pointer">pointer</a>,
1347 <a href="#t_vector">vector</a>,
1348 <a href="#t_struct">structure</a>,
1349 <a href="#t_array">array</a>,
1350 <a href="#t_label">label</a>,
1351 <a href="#t_metadata">metadata</a>.
1355 <td><a href="#t_primitive">primitive</a></td>
1356 <td><a href="#t_label">label</a>,
1357 <a href="#t_void">void</a>,
1358 <a href="#t_floating">floating point</a>,
1359 <a href="#t_metadata">metadata</a>.</td>
1362 <td><a href="#t_derived">derived</a></td>
1363 <td><a href="#t_integer">integer</a>,
1364 <a href="#t_array">array</a>,
1365 <a href="#t_function">function</a>,
1366 <a href="#t_pointer">pointer</a>,
1367 <a href="#t_struct">structure</a>,
1368 <a href="#t_pstruct">packed structure</a>,
1369 <a href="#t_vector">vector</a>,
1370 <a href="#t_opaque">opaque</a>.
1376 <p>The <a href="#t_firstclass">first class</a> types are perhaps the most
1377 important. Values of these types are the only ones which can be produced by
1378 instructions, passed as arguments, or used as operands to instructions.</p>
1382 <!-- ======================================================================= -->
1383 <div class="doc_subsection"> <a name="t_primitive">Primitive Types</a> </div>
1385 <div class="doc_text">
1387 <p>The primitive types are the fundamental building blocks of the LLVM
1392 <!-- _______________________________________________________________________ -->
1393 <div class="doc_subsubsection"> <a name="t_floating">Floating Point Types</a> </div>
1395 <div class="doc_text">
1399 <tr><th>Type</th><th>Description</th></tr>
1400 <tr><td><tt>float</tt></td><td>32-bit floating point value</td></tr>
1401 <tr><td><tt>double</tt></td><td>64-bit floating point value</td></tr>
1402 <tr><td><tt>fp128</tt></td><td>128-bit floating point value (112-bit mantissa)</td></tr>
1403 <tr><td><tt>x86_fp80</tt></td><td>80-bit floating point value (X87)</td></tr>
1404 <tr><td><tt>ppc_fp128</tt></td><td>128-bit floating point value (two 64-bits)</td></tr>
1410 <!-- _______________________________________________________________________ -->
1411 <div class="doc_subsubsection"> <a name="t_void">Void Type</a> </div>
1413 <div class="doc_text">
1416 <p>The void type does not represent any value and has no size.</p>
1425 <!-- _______________________________________________________________________ -->
1426 <div class="doc_subsubsection"> <a name="t_label">Label Type</a> </div>
1428 <div class="doc_text">
1431 <p>The label type represents code labels.</p>
1440 <!-- _______________________________________________________________________ -->
1441 <div class="doc_subsubsection"> <a name="t_metadata">Metadata Type</a> </div>
1443 <div class="doc_text">
1446 <p>The metadata type represents embedded metadata. The only derived type that
1447 may contain metadata is <tt>metadata*</tt> or a function type that returns or
1448 takes metadata typed parameters, but not pointer to metadata types.</p>
1458 <!-- ======================================================================= -->
1459 <div class="doc_subsection"> <a name="t_derived">Derived Types</a> </div>
1461 <div class="doc_text">
1463 <p>The real power in LLVM comes from the derived types in the system. This is
1464 what allows a programmer to represent arrays, functions, pointers, and other
1465 useful types. Note that these derived types may be recursive: For example,
1466 it is possible to have a two dimensional array.</p>
1470 <!-- _______________________________________________________________________ -->
1471 <div class="doc_subsubsection"> <a name="t_integer">Integer Type</a> </div>
1473 <div class="doc_text">
1476 <p>The integer type is a very simple derived type that simply specifies an
1477 arbitrary bit width for the integer type desired. Any bit width from 1 bit to
1478 2^23-1 (about 8 million) can be specified.</p>
1485 <p>The number of bits the integer will occupy is specified by the <tt>N</tt>
1489 <table class="layout">
1491 <td class="left"><tt>i1</tt></td>
1492 <td class="left">a single-bit integer.</td>
1495 <td class="left"><tt>i32</tt></td>
1496 <td class="left">a 32-bit integer.</td>
1499 <td class="left"><tt>i1942652</tt></td>
1500 <td class="left">a really big integer of over 1 million bits.</td>
1504 <p>Note that the code generator does not yet support large integer types to be
1505 used as function return types. The specific limit on how large a return type
1506 the code generator can currently handle is target-dependent; currently it's
1507 often 64 bits for 32-bit targets and 128 bits for 64-bit targets.</p>
1511 <!-- _______________________________________________________________________ -->
1512 <div class="doc_subsubsection"> <a name="t_array">Array Type</a> </div>
1514 <div class="doc_text">
1517 <p>The array type is a very simple derived type that arranges elements
1518 sequentially in memory. The array type requires a size (number of elements)
1519 and an underlying data type.</p>
1523 [<# elements> x <elementtype>]
1526 <p>The number of elements is a constant integer value; <tt>elementtype</tt> may
1527 be any type with a size.</p>
1530 <table class="layout">
1532 <td class="left"><tt>[40 x i32]</tt></td>
1533 <td class="left">Array of 40 32-bit integer values.</td>
1536 <td class="left"><tt>[41 x i32]</tt></td>
1537 <td class="left">Array of 41 32-bit integer values.</td>
1540 <td class="left"><tt>[4 x i8]</tt></td>
1541 <td class="left">Array of 4 8-bit integer values.</td>
1544 <p>Here are some examples of multidimensional arrays:</p>
1545 <table class="layout">
1547 <td class="left"><tt>[3 x [4 x i32]]</tt></td>
1548 <td class="left">3x4 array of 32-bit integer values.</td>
1551 <td class="left"><tt>[12 x [10 x float]]</tt></td>
1552 <td class="left">12x10 array of single precision floating point values.</td>
1555 <td class="left"><tt>[2 x [3 x [4 x i16]]]</tt></td>
1556 <td class="left">2x3x4 array of 16-bit integer values.</td>
1560 <p>Note that 'variable sized arrays' can be implemented in LLVM with a zero
1561 length array. Normally, accesses past the end of an array are undefined in
1562 LLVM (e.g. it is illegal to access the 5th element of a 3 element array). As
1563 a special case, however, zero length arrays are recognized to be variable
1564 length. This allows implementation of 'pascal style arrays' with the LLVM
1565 type "<tt>{ i32, [0 x float]}</tt>", for example.</p>
1567 <p>Note that the code generator does not yet support large aggregate types to be
1568 used as function return types. The specific limit on how large an aggregate
1569 return type the code generator can currently handle is target-dependent, and
1570 also dependent on the aggregate element types.</p>
1574 <!-- _______________________________________________________________________ -->
1575 <div class="doc_subsubsection"> <a name="t_function">Function Type</a> </div>
1577 <div class="doc_text">
1580 <p>The function type can be thought of as a function signature. It consists of
1581 a return type and a list of formal parameter types. The return type of a
1582 function type is a scalar type, a void type, or a struct type. If the return
1583 type is a struct type then all struct elements must be of first class types,
1584 and the struct must have at least one element.</p>
1588 <returntype list> (<parameter list>)
1591 <p>...where '<tt><parameter list></tt>' is a comma-separated list of type
1592 specifiers. Optionally, the parameter list may include a type <tt>...</tt>,
1593 which indicates that the function takes a variable number of arguments.
1594 Variable argument functions can access their arguments with
1595 the <a href="#int_varargs">variable argument handling intrinsic</a>
1596 functions. '<tt><returntype list></tt>' is a comma-separated list of
1597 <a href="#t_firstclass">first class</a> type specifiers.</p>
1600 <table class="layout">
1602 <td class="left"><tt>i32 (i32)</tt></td>
1603 <td class="left">function taking an <tt>i32</tt>, returning an <tt>i32</tt>
1605 </tr><tr class="layout">
1606 <td class="left"><tt>float (i16 signext, i32 *) *
1608 <td class="left"><a href="#t_pointer">Pointer</a> to a function that takes
1609 an <tt>i16</tt> that should be sign extended and a
1610 <a href="#t_pointer">pointer</a> to <tt>i32</tt>, returning
1613 </tr><tr class="layout">
1614 <td class="left"><tt>i32 (i8*, ...)</tt></td>
1615 <td class="left">A vararg function that takes at least one
1616 <a href="#t_pointer">pointer</a> to <tt>i8 </tt> (char in C),
1617 which returns an integer. This is the signature for <tt>printf</tt> in
1620 </tr><tr class="layout">
1621 <td class="left"><tt>{i32, i32} (i32)</tt></td>
1622 <td class="left">A function taking an <tt>i32</tt>, returning two
1623 <tt>i32</tt> values as an aggregate of type <tt>{ i32, i32 }</tt>
1630 <!-- _______________________________________________________________________ -->
1631 <div class="doc_subsubsection"> <a name="t_struct">Structure Type</a> </div>
1633 <div class="doc_text">
1636 <p>The structure type is used to represent a collection of data members together
1637 in memory. The packing of the field types is defined to match the ABI of the
1638 underlying processor. The elements of a structure may be any type that has a
1641 <p>Structures are accessed using '<tt><a href="#i_load">load</a></tt> and
1642 '<tt><a href="#i_store">store</a></tt>' by getting a pointer to a field with
1643 the '<tt><a href="#i_getelementptr">getelementptr</a></tt>' instruction.</p>
1647 { <type list> }
1651 <table class="layout">
1653 <td class="left"><tt>{ i32, i32, i32 }</tt></td>
1654 <td class="left">A triple of three <tt>i32</tt> values</td>
1655 </tr><tr class="layout">
1656 <td class="left"><tt>{ float, i32 (i32) * }</tt></td>
1657 <td class="left">A pair, where the first element is a <tt>float</tt> and the
1658 second element is a <a href="#t_pointer">pointer</a> to a
1659 <a href="#t_function">function</a> that takes an <tt>i32</tt>, returning
1660 an <tt>i32</tt>.</td>
1664 <p>Note that the code generator does not yet support large aggregate types to be
1665 used as function return types. The specific limit on how large an aggregate
1666 return type the code generator can currently handle is target-dependent, and
1667 also dependent on the aggregate element types.</p>
1671 <!-- _______________________________________________________________________ -->
1672 <div class="doc_subsubsection"> <a name="t_pstruct">Packed Structure Type</a>
1675 <div class="doc_text">
1678 <p>The packed structure type is used to represent a collection of data members
1679 together in memory. There is no padding between fields. Further, the
1680 alignment of a packed structure is 1 byte. The elements of a packed
1681 structure may be any type that has a size.</p>
1683 <p>Structures are accessed using '<tt><a href="#i_load">load</a></tt> and
1684 '<tt><a href="#i_store">store</a></tt>' by getting a pointer to a field with
1685 the '<tt><a href="#i_getelementptr">getelementptr</a></tt>' instruction.</p>
1689 < { <type list> } >
1693 <table class="layout">
1695 <td class="left"><tt>< { i32, i32, i32 } ></tt></td>
1696 <td class="left">A triple of three <tt>i32</tt> values</td>
1697 </tr><tr class="layout">
1699 <tt>< { float, i32 (i32)* } ></tt></td>
1700 <td class="left">A pair, where the first element is a <tt>float</tt> and the
1701 second element is a <a href="#t_pointer">pointer</a> to a
1702 <a href="#t_function">function</a> that takes an <tt>i32</tt>, returning
1703 an <tt>i32</tt>.</td>
1709 <!-- _______________________________________________________________________ -->
1710 <div class="doc_subsubsection"> <a name="t_pointer">Pointer Type</a> </div>
1712 <div class="doc_text">
1715 <p>As in many languages, the pointer type represents a pointer or reference to
1716 another object, which must live in memory. Pointer types may have an optional
1717 address space attribute defining the target-specific numbered address space
1718 where the pointed-to object resides. The default address space is zero.</p>
1720 <p>Note that LLVM does not permit pointers to void (<tt>void*</tt>) nor does it
1721 permit pointers to labels (<tt>label*</tt>). Use <tt>i8*</tt> instead.</p>
1729 <table class="layout">
1731 <td class="left"><tt>[4 x i32]*</tt></td>
1732 <td class="left">A <a href="#t_pointer">pointer</a> to <a
1733 href="#t_array">array</a> of four <tt>i32</tt> values.</td>
1736 <td class="left"><tt>i32 (i32 *) *</tt></td>
1737 <td class="left"> A <a href="#t_pointer">pointer</a> to a <a
1738 href="#t_function">function</a> that takes an <tt>i32*</tt>, returning an
1742 <td class="left"><tt>i32 addrspace(5)*</tt></td>
1743 <td class="left">A <a href="#t_pointer">pointer</a> to an <tt>i32</tt> value
1744 that resides in address space #5.</td>
1750 <!-- _______________________________________________________________________ -->
1751 <div class="doc_subsubsection"> <a name="t_vector">Vector Type</a> </div>
1753 <div class="doc_text">
1756 <p>A vector type is a simple derived type that represents a vector of elements.
1757 Vector types are used when multiple primitive data are operated in parallel
1758 using a single instruction (SIMD). A vector type requires a size (number of
1759 elements) and an underlying primitive data type. Vectors must have a power
1760 of two length (1, 2, 4, 8, 16 ...). Vector types are considered
1761 <a href="#t_firstclass">first class</a>.</p>
1765 < <# elements> x <elementtype> >
1768 <p>The number of elements is a constant integer value; elementtype may be any
1769 integer or floating point type.</p>
1772 <table class="layout">
1774 <td class="left"><tt><4 x i32></tt></td>
1775 <td class="left">Vector of 4 32-bit integer values.</td>
1778 <td class="left"><tt><8 x float></tt></td>
1779 <td class="left">Vector of 8 32-bit floating-point values.</td>
1782 <td class="left"><tt><2 x i64></tt></td>
1783 <td class="left">Vector of 2 64-bit integer values.</td>
1787 <p>Note that the code generator does not yet support large vector types to be
1788 used as function return types. The specific limit on how large a vector
1789 return type codegen can currently handle is target-dependent; currently it's
1790 often a few times longer than a hardware vector register.</p>
1794 <!-- _______________________________________________________________________ -->
1795 <div class="doc_subsubsection"> <a name="t_opaque">Opaque Type</a> </div>
1796 <div class="doc_text">
1799 <p>Opaque types are used to represent unknown types in the system. This
1800 corresponds (for example) to the C notion of a forward declared structure
1801 type. In LLVM, opaque types can eventually be resolved to any type (not just
1802 a structure type).</p>
1810 <table class="layout">
1812 <td class="left"><tt>opaque</tt></td>
1813 <td class="left">An opaque type.</td>
1819 <!-- ======================================================================= -->
1820 <div class="doc_subsection">
1821 <a name="t_uprefs">Type Up-references</a>
1824 <div class="doc_text">
1827 <p>An "up reference" allows you to refer to a lexically enclosing type without
1828 requiring it to have a name. For instance, a structure declaration may
1829 contain a pointer to any of the types it is lexically a member of. Example
1830 of up references (with their equivalent as named type declarations)
1834 { \2 * } %x = type { %x* }
1835 { \2 }* %y = type { %y }*
1839 <p>An up reference is needed by the asmprinter for printing out cyclic types
1840 when there is no declared name for a type in the cycle. Because the
1841 asmprinter does not want to print out an infinite type string, it needs a
1842 syntax to handle recursive types that have no names (all names are optional
1850 <p>The level is the count of the lexical type that is being referred to.</p>
1853 <table class="layout">
1855 <td class="left"><tt>\1*</tt></td>
1856 <td class="left">Self-referential pointer.</td>
1859 <td class="left"><tt>{ { \3*, i8 }, i32 }</tt></td>
1860 <td class="left">Recursive structure where the upref refers to the out-most
1867 <!-- *********************************************************************** -->
1868 <div class="doc_section"> <a name="constants">Constants</a> </div>
1869 <!-- *********************************************************************** -->
1871 <div class="doc_text">
1873 <p>LLVM has several different basic types of constants. This section describes
1874 them all and their syntax.</p>
1878 <!-- ======================================================================= -->
1879 <div class="doc_subsection"><a name="simpleconstants">Simple Constants</a></div>
1881 <div class="doc_text">
1884 <dt><b>Boolean constants</b></dt>
1885 <dd>The two strings '<tt>true</tt>' and '<tt>false</tt>' are both valid
1886 constants of the <tt><a href="#t_primitive">i1</a></tt> type.</dd>
1888 <dt><b>Integer constants</b></dt>
1889 <dd>Standard integers (such as '4') are constants of
1890 the <a href="#t_integer">integer</a> type. Negative numbers may be used
1891 with integer types.</dd>
1893 <dt><b>Floating point constants</b></dt>
1894 <dd>Floating point constants use standard decimal notation (e.g. 123.421),
1895 exponential notation (e.g. 1.23421e+2), or a more precise hexadecimal
1896 notation (see below). The assembler requires the exact decimal value of a
1897 floating-point constant. For example, the assembler accepts 1.25 but
1898 rejects 1.3 because 1.3 is a repeating decimal in binary. Floating point
1899 constants must have a <a href="#t_floating">floating point</a> type. </dd>
1901 <dt><b>Null pointer constants</b></dt>
1902 <dd>The identifier '<tt>null</tt>' is recognized as a null pointer constant
1903 and must be of <a href="#t_pointer">pointer type</a>.</dd>
1906 <p>The one non-intuitive notation for constants is the hexadecimal form of
1907 floating point constants. For example, the form '<tt>double
1908 0x432ff973cafa8000</tt>' is equivalent to (but harder to read than)
1909 '<tt>double 4.5e+15</tt>'. The only time hexadecimal floating point
1910 constants are required (and the only time that they are generated by the
1911 disassembler) is when a floating point constant must be emitted but it cannot
1912 be represented as a decimal floating point number in a reasonable number of
1913 digits. For example, NaN's, infinities, and other special values are
1914 represented in their IEEE hexadecimal format so that assembly and disassembly
1915 do not cause any bits to change in the constants.</p>
1917 <p>When using the hexadecimal form, constants of types float and double are
1918 represented using the 16-digit form shown above (which matches the IEEE754
1919 representation for double); float values must, however, be exactly
1920 representable as IEE754 single precision. Hexadecimal format is always used
1921 for long double, and there are three forms of long double. The 80-bit format
1922 used by x86 is represented as <tt>0xK</tt> followed by 20 hexadecimal digits.
1923 The 128-bit format used by PowerPC (two adjacent doubles) is represented
1924 by <tt>0xM</tt> followed by 32 hexadecimal digits. The IEEE 128-bit format
1925 is represented by <tt>0xL</tt> followed by 32 hexadecimal digits; no
1926 currently supported target uses this format. Long doubles will only work if
1927 they match the long double format on your target. All hexadecimal formats
1928 are big-endian (sign bit at the left).</p>
1932 <!-- ======================================================================= -->
1933 <div class="doc_subsection">
1934 <a name="aggregateconstants"></a> <!-- old anchor -->
1935 <a name="complexconstants">Complex Constants</a>
1938 <div class="doc_text">
1940 <p>Complex constants are a (potentially recursive) combination of simple
1941 constants and smaller complex constants.</p>
1944 <dt><b>Structure constants</b></dt>
1945 <dd>Structure constants are represented with notation similar to structure
1946 type definitions (a comma separated list of elements, surrounded by braces
1947 (<tt>{}</tt>)). For example: "<tt>{ i32 4, float 17.0, i32* @G }</tt>",
1948 where "<tt>@G</tt>" is declared as "<tt>@G = external global i32</tt>".
1949 Structure constants must have <a href="#t_struct">structure type</a>, and
1950 the number and types of elements must match those specified by the
1953 <dt><b>Array constants</b></dt>
1954 <dd>Array constants are represented with notation similar to array type
1955 definitions (a comma separated list of elements, surrounded by square
1956 brackets (<tt>[]</tt>)). For example: "<tt>[ i32 42, i32 11, i32 74
1957 ]</tt>". Array constants must have <a href="#t_array">array type</a>, and
1958 the number and types of elements must match those specified by the
1961 <dt><b>Vector constants</b></dt>
1962 <dd>Vector constants are represented with notation similar to vector type
1963 definitions (a comma separated list of elements, surrounded by
1964 less-than/greater-than's (<tt><></tt>)). For example: "<tt>< i32
1965 42, i32 11, i32 74, i32 100 ></tt>". Vector constants must
1966 have <a href="#t_vector">vector type</a>, and the number and types of
1967 elements must match those specified by the type.</dd>
1969 <dt><b>Zero initialization</b></dt>
1970 <dd>The string '<tt>zeroinitializer</tt>' can be used to zero initialize a
1971 value to zero of <em>any</em> type, including scalar and aggregate types.
1972 This is often used to avoid having to print large zero initializers
1973 (e.g. for large arrays) and is always exactly equivalent to using explicit
1974 zero initializers.</dd>
1976 <dt><b>Metadata node</b></dt>
1977 <dd>A metadata node is a structure-like constant with
1978 <a href="#t_metadata">metadata type</a>. For example: "<tt>metadata !{
1979 i32 0, metadata !"test" }</tt>". Unlike other constants that are meant to
1980 be interpreted as part of the instruction stream, metadata is a place to
1981 attach additional information such as debug info.</dd>
1986 <!-- ======================================================================= -->
1987 <div class="doc_subsection">
1988 <a name="globalconstants">Global Variable and Function Addresses</a>
1991 <div class="doc_text">
1993 <p>The addresses of <a href="#globalvars">global variables</a>
1994 and <a href="#functionstructure">functions</a> are always implicitly valid
1995 (link-time) constants. These constants are explicitly referenced when
1996 the <a href="#identifiers">identifier for the global</a> is used and always
1997 have <a href="#t_pointer">pointer</a> type. For example, the following is a
1998 legal LLVM file:</p>
2000 <div class="doc_code">
2004 @Z = global [2 x i32*] [ i32* @X, i32* @Y ]
2010 <!-- ======================================================================= -->
2011 <div class="doc_subsection"><a name="undefvalues">Undefined Values</a></div>
2012 <div class="doc_text">
2014 <p>The string '<tt>undef</tt>' is recognized as a type-less constant that has no
2015 specific value. Undefined values may be of any type and be used anywhere a
2016 constant is permitted.</p>
2018 <p>Undefined values indicate to the compiler that the program is well defined no
2019 matter what value is used, giving the compiler more freedom to optimize.</p>
2023 <!-- ======================================================================= -->
2024 <div class="doc_subsection"><a name="constantexprs">Constant Expressions</a>
2027 <div class="doc_text">
2029 <p>Constant expressions are used to allow expressions involving other constants
2030 to be used as constants. Constant expressions may be of
2031 any <a href="#t_firstclass">first class</a> type and may involve any LLVM
2032 operation that does not have side effects (e.g. load and call are not
2033 supported). The following is the syntax for constant expressions:</p>
2036 <dt><b><tt>trunc ( CST to TYPE )</tt></b></dt>
2037 <dd>Truncate a constant to another type. The bit size of CST must be larger
2038 than the bit size of TYPE. Both types must be integers.</dd>
2040 <dt><b><tt>zext ( CST to TYPE )</tt></b></dt>
2041 <dd>Zero extend a constant to another type. The bit size of CST must be
2042 smaller or equal to the bit size of TYPE. Both types must be
2045 <dt><b><tt>sext ( CST to TYPE )</tt></b></dt>
2046 <dd>Sign extend a constant to another type. The bit size of CST must be
2047 smaller or equal to the bit size of TYPE. Both types must be
2050 <dt><b><tt>fptrunc ( CST to TYPE )</tt></b></dt>
2051 <dd>Truncate a floating point constant to another floating point type. The
2052 size of CST must be larger than the size of TYPE. Both types must be
2053 floating point.</dd>
2055 <dt><b><tt>fpext ( CST to TYPE )</tt></b></dt>
2056 <dd>Floating point extend a constant to another type. The size of CST must be
2057 smaller or equal to the size of TYPE. Both types must be floating
2060 <dt><b><tt>fptoui ( CST to TYPE )</tt></b></dt>
2061 <dd>Convert a floating point constant to the corresponding unsigned integer
2062 constant. TYPE must be a scalar or vector integer type. CST must be of
2063 scalar or vector floating point type. Both CST and TYPE must be scalars,
2064 or vectors of the same number of elements. If the value won't fit in the
2065 integer type, the results are undefined.</dd>
2067 <dt><b><tt>fptosi ( CST to TYPE )</tt></b></dt>
2068 <dd>Convert a floating point constant to the corresponding signed integer
2069 constant. TYPE must be a scalar or vector integer type. CST must be of
2070 scalar or vector floating point type. Both CST and TYPE must be scalars,
2071 or vectors of the same number of elements. If the value won't fit in the
2072 integer type, the results are undefined.</dd>
2074 <dt><b><tt>uitofp ( CST to TYPE )</tt></b></dt>
2075 <dd>Convert an unsigned integer constant to the corresponding floating point
2076 constant. TYPE must be a scalar or vector floating point type. CST must be
2077 of scalar or vector integer type. Both CST and TYPE must be scalars, or
2078 vectors of the same number of elements. If the value won't fit in the
2079 floating point type, the results are undefined.</dd>
2081 <dt><b><tt>sitofp ( CST to TYPE )</tt></b></dt>
2082 <dd>Convert a signed integer constant to the corresponding floating point
2083 constant. TYPE must be a scalar or vector floating point type. CST must be
2084 of scalar or vector integer type. Both CST and TYPE must be scalars, or
2085 vectors of the same number of elements. If the value won't fit in the
2086 floating point type, the results are undefined.</dd>
2088 <dt><b><tt>ptrtoint ( CST to TYPE )</tt></b></dt>
2089 <dd>Convert a pointer typed constant to the corresponding integer constant
2090 <tt>TYPE</tt> must be an integer type. <tt>CST</tt> must be of pointer
2091 type. The <tt>CST</tt> value is zero extended, truncated, or unchanged to
2092 make it fit in <tt>TYPE</tt>.</dd>
2094 <dt><b><tt>inttoptr ( CST to TYPE )</tt></b></dt>
2095 <dd>Convert a integer constant to a pointer constant. TYPE must be a pointer
2096 type. CST must be of integer type. The CST value is zero extended,
2097 truncated, or unchanged to make it fit in a pointer size. This one is
2098 <i>really</i> dangerous!</dd>
2100 <dt><b><tt>bitcast ( CST to TYPE )</tt></b></dt>
2101 <dd>Convert a constant, CST, to another TYPE. The constraints of the operands
2102 are the same as those for the <a href="#i_bitcast">bitcast
2103 instruction</a>.</dd>
2105 <dt><b><tt>getelementptr ( CSTPTR, IDX0, IDX1, ... )</tt></b></dt>
2106 <dt><b><tt>getelementptr inbounds ( CSTPTR, IDX0, IDX1, ... )</tt></b></dt>
2107 <dd>Perform the <a href="#i_getelementptr">getelementptr operation</a> on
2108 constants. As with the <a href="#i_getelementptr">getelementptr</a>
2109 instruction, the index list may have zero or more indexes, which are
2110 required to make sense for the type of "CSTPTR".</dd>
2112 <dt><b><tt>select ( COND, VAL1, VAL2 )</tt></b></dt>
2113 <dd>Perform the <a href="#i_select">select operation</a> on constants.</dd>
2115 <dt><b><tt>icmp COND ( VAL1, VAL2 )</tt></b></dt>
2116 <dd>Performs the <a href="#i_icmp">icmp operation</a> on constants.</dd>
2118 <dt><b><tt>fcmp COND ( VAL1, VAL2 )</tt></b></dt>
2119 <dd>Performs the <a href="#i_fcmp">fcmp operation</a> on constants.</dd>
2121 <dt><b><tt>extractelement ( VAL, IDX )</tt></b></dt>
2122 <dd>Perform the <a href="#i_extractelement">extractelement operation</a> on
2125 <dt><b><tt>insertelement ( VAL, ELT, IDX )</tt></b></dt>
2126 <dd>Perform the <a href="#i_insertelement">insertelement operation</a> on
2129 <dt><b><tt>shufflevector ( VEC1, VEC2, IDXMASK )</tt></b></dt>
2130 <dd>Perform the <a href="#i_shufflevector">shufflevector operation</a> on
2133 <dt><b><tt>OPCODE ( LHS, RHS )</tt></b></dt>
2134 <dd>Perform the specified operation of the LHS and RHS constants. OPCODE may
2135 be any of the <a href="#binaryops">binary</a>
2136 or <a href="#bitwiseops">bitwise binary</a> operations. The constraints
2137 on operands are the same as those for the corresponding instruction
2138 (e.g. no bitwise operations on floating point values are allowed).</dd>
2143 <!-- ======================================================================= -->
2144 <div class="doc_subsection"><a name="metadata">Embedded Metadata</a>
2147 <div class="doc_text">
2149 <p>Embedded metadata provides a way to attach arbitrary data to the instruction
2150 stream without affecting the behaviour of the program. There are two
2151 metadata primitives, strings and nodes. All metadata has the
2152 <tt>metadata</tt> type and is identified in syntax by a preceding exclamation
2153 point ('<tt>!</tt>').</p>
2155 <p>A metadata string is a string surrounded by double quotes. It can contain
2156 any character by escaping non-printable characters with "\xx" where "xx" is
2157 the two digit hex code. For example: "<tt>!"test\00"</tt>".</p>
2159 <p>Metadata nodes are represented with notation similar to structure constants
2160 (a comma separated list of elements, surrounded by braces and preceeded by an
2161 exclamation point). For example: "<tt>!{ metadata !"test\00", i32
2164 <p>A metadata node will attempt to track changes to the values it holds. In the
2165 event that a value is deleted, it will be replaced with a typeless
2166 "<tt>null</tt>", such as "<tt>metadata !{null, i32 10}</tt>".</p>
2168 <p>Optimizations may rely on metadata to provide additional information about
2169 the program that isn't available in the instructions, or that isn't easily
2170 computable. Similarly, the code generator may expect a certain metadata
2171 format to be used to express debugging information.</p>
2175 <!-- *********************************************************************** -->
2176 <div class="doc_section"> <a name="othervalues">Other Values</a> </div>
2177 <!-- *********************************************************************** -->
2179 <!-- ======================================================================= -->
2180 <div class="doc_subsection">
2181 <a name="inlineasm">Inline Assembler Expressions</a>
2184 <div class="doc_text">
2186 <p>LLVM supports inline assembler expressions (as opposed
2187 to <a href="#moduleasm"> Module-Level Inline Assembly</a>) through the use of
2188 a special value. This value represents the inline assembler as a string
2189 (containing the instructions to emit), a list of operand constraints (stored
2190 as a string), and a flag that indicates whether or not the inline asm
2191 expression has side effects. An example inline assembler expression is:</p>
2193 <div class="doc_code">
2195 i32 (i32) asm "bswap $0", "=r,r"
2199 <p>Inline assembler expressions may <b>only</b> be used as the callee operand of
2200 a <a href="#i_call"><tt>call</tt> instruction</a>. Thus, typically we
2203 <div class="doc_code">
2205 %X = call i32 asm "<a href="#int_bswap">bswap</a> $0", "=r,r"(i32 %Y)
2209 <p>Inline asms with side effects not visible in the constraint list must be
2210 marked as having side effects. This is done through the use of the
2211 '<tt>sideeffect</tt>' keyword, like so:</p>
2213 <div class="doc_code">
2215 call void asm sideeffect "eieio", ""()
2219 <p>TODO: The format of the asm and constraints string still need to be
2220 documented here. Constraints on what can be done (e.g. duplication, moving,
2221 etc need to be documented). This is probably best done by reference to
2222 another document that covers inline asm from a holistic perspective.</p>
2227 <!-- *********************************************************************** -->
2228 <div class="doc_section">
2229 <a name="intrinsic_globals">Intrinsic Global Variables</a>
2231 <!-- *********************************************************************** -->
2233 <p>LLVM has a number of "magic" global variables that contain data that affect
2234 code generation or other IR semantics. These are documented here. All globals
2235 of this sort should have a section specified as "<tt>llvm.metadata</tt>". This
2236 section and all globals that start with "<tt>llvm.</tt>" are reserved for use
2239 <!-- ======================================================================= -->
2240 <div class="doc_subsection">
2241 <a name="intg_used">The '<tt>llvm.used</tt>' Global Variable</a>
2244 <div class="doc_text">
2246 <p>The <tt>@llvm.used</tt> global is an array with i8* element type which has <a
2247 href="#linkage_appending">appending linkage</a>. This array contains a list of
2248 pointers to global variables and functions which may optionally have a pointer
2249 cast formed of bitcast or getelementptr. For example, a legal use of it is:</p>
2255 @llvm.used = appending global [2 x i8*] [
2257 i8* bitcast (i32* @Y to i8*)
2258 ], section "llvm.metadata"
2261 <p>If a global variable appears in the <tt>@llvm.used</tt> list, then the
2262 compiler, assembler, and linker are required to treat the symbol as if there is
2263 a reference to the global that it cannot see. For example, if a variable has
2264 internal linkage and no references other than that from the <tt>@llvm.used</tt>
2265 list, it cannot be deleted. This is commonly used to represent references from
2266 inline asms and other things the compiler cannot "see", and corresponds to
2267 "attribute((used))" in GNU C.</p>
2269 <p>On some targets, the code generator must emit a directive to the assembler or
2270 object file to prevent the assembler and linker from molesting the symbol.</p>
2274 <!-- ======================================================================= -->
2275 <div class="doc_subsection">
2276 <a name="intg_compiler_used">The '<tt>llvm.compiler.used</tt>' Global Variable</a>
2279 <div class="doc_text">
2281 <p>The <tt>@llvm.compiler.used</tt> directive is the same as the
2282 <tt>@llvm.used</tt> directive, except that it only prevents the compiler from
2283 touching the symbol. On targets that support it, this allows an intelligent
2284 linker to optimize references to the symbol without being impeded as it would be
2285 by <tt>@llvm.used</tt>.</p>
2287 <p>This is a rare construct that should only be used in rare circumstances, and
2288 should not be exposed to source languages.</p>
2292 <!-- ======================================================================= -->
2293 <div class="doc_subsection">
2294 <a name="intg_global_ctors">The '<tt>llvm.global_ctors</tt>' Global Variable</a>
2297 <div class="doc_text">
2299 <p>TODO: Describe this.</p>
2303 <!-- ======================================================================= -->
2304 <div class="doc_subsection">
2305 <a name="intg_global_dtors">The '<tt>llvm.global_dtors</tt>' Global Variable</a>
2308 <div class="doc_text">
2310 <p>TODO: Describe this.</p>
2315 <!-- *********************************************************************** -->
2316 <div class="doc_section"> <a name="instref">Instruction Reference</a> </div>
2317 <!-- *********************************************************************** -->
2319 <div class="doc_text">
2321 <p>The LLVM instruction set consists of several different classifications of
2322 instructions: <a href="#terminators">terminator
2323 instructions</a>, <a href="#binaryops">binary instructions</a>,
2324 <a href="#bitwiseops">bitwise binary instructions</a>,
2325 <a href="#memoryops">memory instructions</a>, and
2326 <a href="#otherops">other instructions</a>.</p>
2330 <!-- ======================================================================= -->
2331 <div class="doc_subsection"> <a name="terminators">Terminator
2332 Instructions</a> </div>
2334 <div class="doc_text">
2336 <p>As mentioned <a href="#functionstructure">previously</a>, every basic block
2337 in a program ends with a "Terminator" instruction, which indicates which
2338 block should be executed after the current block is finished. These
2339 terminator instructions typically yield a '<tt>void</tt>' value: they produce
2340 control flow, not values (the one exception being the
2341 '<a href="#i_invoke"><tt>invoke</tt></a>' instruction).</p>
2343 <p>There are six different terminator instructions: the
2344 '<a href="#i_ret"><tt>ret</tt></a>' instruction, the
2345 '<a href="#i_br"><tt>br</tt></a>' instruction, the
2346 '<a href="#i_switch"><tt>switch</tt></a>' instruction, the
2347 '<a href="#i_invoke"><tt>invoke</tt></a>' instruction, the
2348 '<a href="#i_unwind"><tt>unwind</tt></a>' instruction, and the
2349 '<a href="#i_unreachable"><tt>unreachable</tt></a>' instruction.</p>
2353 <!-- _______________________________________________________________________ -->
2354 <div class="doc_subsubsection"> <a name="i_ret">'<tt>ret</tt>'
2355 Instruction</a> </div>
2357 <div class="doc_text">
2361 ret <type> <value> <i>; Return a value from a non-void function</i>
2362 ret void <i>; Return from void function</i>
2366 <p>The '<tt>ret</tt>' instruction is used to return control flow (and optionally
2367 a value) from a function back to the caller.</p>
2369 <p>There are two forms of the '<tt>ret</tt>' instruction: one that returns a
2370 value and then causes control flow, and one that just causes control flow to
2374 <p>The '<tt>ret</tt>' instruction optionally accepts a single argument, the
2375 return value. The type of the return value must be a
2376 '<a href="#t_firstclass">first class</a>' type.</p>
2378 <p>A function is not <a href="#wellformed">well formed</a> if it it has a
2379 non-void return type and contains a '<tt>ret</tt>' instruction with no return
2380 value or a return value with a type that does not match its type, or if it
2381 has a void return type and contains a '<tt>ret</tt>' instruction with a
2385 <p>When the '<tt>ret</tt>' instruction is executed, control flow returns back to
2386 the calling function's context. If the caller is a
2387 "<a href="#i_call"><tt>call</tt></a>" instruction, execution continues at the
2388 instruction after the call. If the caller was an
2389 "<a href="#i_invoke"><tt>invoke</tt></a>" instruction, execution continues at
2390 the beginning of the "normal" destination block. If the instruction returns
2391 a value, that value shall set the call or invoke instruction's return
2396 ret i32 5 <i>; Return an integer value of 5</i>
2397 ret void <i>; Return from a void function</i>
2398 ret { i32, i8 } { i32 4, i8 2 } <i>; Return a struct of values 4 and 2</i>
2401 <p>Note that the code generator does not yet fully support large
2402 return values. The specific sizes that are currently supported are
2403 dependent on the target. For integers, on 32-bit targets the limit
2404 is often 64 bits, and on 64-bit targets the limit is often 128 bits.
2405 For aggregate types, the current limits are dependent on the element
2406 types; for example targets are often limited to 2 total integer
2407 elements and 2 total floating-point elements.</p>
2410 <!-- _______________________________________________________________________ -->
2411 <div class="doc_subsubsection"> <a name="i_br">'<tt>br</tt>' Instruction</a> </div>
2413 <div class="doc_text">
2417 br i1 <cond>, label <iftrue>, label <iffalse><br> br label <dest> <i>; Unconditional branch</i>
2421 <p>The '<tt>br</tt>' instruction is used to cause control flow to transfer to a
2422 different basic block in the current function. There are two forms of this
2423 instruction, corresponding to a conditional branch and an unconditional
2427 <p>The conditional branch form of the '<tt>br</tt>' instruction takes a single
2428 '<tt>i1</tt>' value and two '<tt>label</tt>' values. The unconditional form
2429 of the '<tt>br</tt>' instruction takes a single '<tt>label</tt>' value as a
2433 <p>Upon execution of a conditional '<tt>br</tt>' instruction, the '<tt>i1</tt>'
2434 argument is evaluated. If the value is <tt>true</tt>, control flows to the
2435 '<tt>iftrue</tt>' <tt>label</tt> argument. If "cond" is <tt>false</tt>,
2436 control flows to the '<tt>iffalse</tt>' <tt>label</tt> argument.</p>
2441 %cond = <a href="#i_icmp">icmp</a> eq i32 %a, %b
2442 br i1 %cond, label %IfEqual, label %IfUnequal
2444 <a href="#i_ret">ret</a> i32 1
2446 <a href="#i_ret">ret</a> i32 0
2451 <!-- _______________________________________________________________________ -->
2452 <div class="doc_subsubsection">
2453 <a name="i_switch">'<tt>switch</tt>' Instruction</a>
2456 <div class="doc_text">
2460 switch <intty> <value>, label <defaultdest> [ <intty> <val>, label <dest> ... ]
2464 <p>The '<tt>switch</tt>' instruction is used to transfer control flow to one of
2465 several different places. It is a generalization of the '<tt>br</tt>'
2466 instruction, allowing a branch to occur to one of many possible
2470 <p>The '<tt>switch</tt>' instruction uses three parameters: an integer
2471 comparison value '<tt>value</tt>', a default '<tt>label</tt>' destination,
2472 and an array of pairs of comparison value constants and '<tt>label</tt>'s.
2473 The table is not allowed to contain duplicate constant entries.</p>
2476 <p>The <tt>switch</tt> instruction specifies a table of values and
2477 destinations. When the '<tt>switch</tt>' instruction is executed, this table
2478 is searched for the given value. If the value is found, control flow is
2479 transfered to the corresponding destination; otherwise, control flow is
2480 transfered to the default destination.</p>
2482 <h5>Implementation:</h5>
2483 <p>Depending on properties of the target machine and the particular
2484 <tt>switch</tt> instruction, this instruction may be code generated in
2485 different ways. For example, it could be generated as a series of chained
2486 conditional branches or with a lookup table.</p>
2490 <i>; Emulate a conditional br instruction</i>
2491 %Val = <a href="#i_zext">zext</a> i1 %value to i32
2492 switch i32 %Val, label %truedest [ i32 0, label %falsedest ]
2494 <i>; Emulate an unconditional br instruction</i>
2495 switch i32 0, label %dest [ ]
2497 <i>; Implement a jump table:</i>
2498 switch i32 %val, label %otherwise [ i32 0, label %onzero
2500 i32 2, label %ontwo ]
2505 <!-- _______________________________________________________________________ -->
2506 <div class="doc_subsubsection">
2507 <a name="i_invoke">'<tt>invoke</tt>' Instruction</a>
2510 <div class="doc_text">
2514 <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>]
2515 to label <normal label> unwind label <exception label>
2519 <p>The '<tt>invoke</tt>' instruction causes control to transfer to a specified
2520 function, with the possibility of control flow transfer to either the
2521 '<tt>normal</tt>' label or the '<tt>exception</tt>' label. If the callee
2522 function returns with the "<tt><a href="#i_ret">ret</a></tt>" instruction,
2523 control flow will return to the "normal" label. If the callee (or any
2524 indirect callees) returns with the "<a href="#i_unwind"><tt>unwind</tt></a>"
2525 instruction, control is interrupted and continued at the dynamically nearest
2526 "exception" label.</p>
2529 <p>This instruction requires several arguments:</p>
2532 <li>The optional "cconv" marker indicates which <a href="#callingconv">calling
2533 convention</a> the call should use. If none is specified, the call
2534 defaults to using C calling conventions.</li>
2536 <li>The optional <a href="#paramattrs">Parameter Attributes</a> list for
2537 return values. Only '<tt>zeroext</tt>', '<tt>signext</tt>', and
2538 '<tt>inreg</tt>' attributes are valid here.</li>
2540 <li>'<tt>ptr to function ty</tt>': shall be the signature of the pointer to
2541 function value being invoked. In most cases, this is a direct function
2542 invocation, but indirect <tt>invoke</tt>s are just as possible, branching
2543 off an arbitrary pointer to function value.</li>
2545 <li>'<tt>function ptr val</tt>': An LLVM value containing a pointer to a
2546 function to be invoked. </li>
2548 <li>'<tt>function args</tt>': argument list whose types match the function
2549 signature argument types. If the function signature indicates the
2550 function accepts a variable number of arguments, the extra arguments can
2553 <li>'<tt>normal label</tt>': the label reached when the called function
2554 executes a '<tt><a href="#i_ret">ret</a></tt>' instruction. </li>
2556 <li>'<tt>exception label</tt>': the label reached when a callee returns with
2557 the <a href="#i_unwind"><tt>unwind</tt></a> instruction. </li>
2559 <li>The optional <a href="#fnattrs">function attributes</a> list. Only
2560 '<tt>noreturn</tt>', '<tt>nounwind</tt>', '<tt>readonly</tt>' and
2561 '<tt>readnone</tt>' attributes are valid here.</li>
2565 <p>This instruction is designed to operate as a standard
2566 '<tt><a href="#i_call">call</a></tt>' instruction in most regards. The
2567 primary difference is that it establishes an association with a label, which
2568 is used by the runtime library to unwind the stack.</p>
2570 <p>This instruction is used in languages with destructors to ensure that proper
2571 cleanup is performed in the case of either a <tt>longjmp</tt> or a thrown
2572 exception. Additionally, this is important for implementation of
2573 '<tt>catch</tt>' clauses in high-level languages that support them.</p>
2575 <p>For the purposes of the SSA form, the definition of the value returned by the
2576 '<tt>invoke</tt>' instruction is deemed to occur on the edge from the current
2577 block to the "normal" label. If the callee unwinds then no return value is
2582 %retval = invoke i32 @Test(i32 15) to label %Continue
2583 unwind label %TestCleanup <i>; {i32}:retval set</i>
2584 %retval = invoke <a href="#callingconv">coldcc</a> i32 %Testfnptr(i32 15) to label %Continue
2585 unwind label %TestCleanup <i>; {i32}:retval set</i>
2590 <!-- _______________________________________________________________________ -->
2592 <div class="doc_subsubsection"> <a name="i_unwind">'<tt>unwind</tt>'
2593 Instruction</a> </div>
2595 <div class="doc_text">
2603 <p>The '<tt>unwind</tt>' instruction unwinds the stack, continuing control flow
2604 at the first callee in the dynamic call stack which used
2605 an <a href="#i_invoke"><tt>invoke</tt></a> instruction to perform the call.
2606 This is primarily used to implement exception handling.</p>
2609 <p>The '<tt>unwind</tt>' instruction causes execution of the current function to
2610 immediately halt. The dynamic call stack is then searched for the
2611 first <a href="#i_invoke"><tt>invoke</tt></a> instruction on the call stack.
2612 Once found, execution continues at the "exceptional" destination block
2613 specified by the <tt>invoke</tt> instruction. If there is no <tt>invoke</tt>
2614 instruction in the dynamic call chain, undefined behavior results.</p>
2618 <!-- _______________________________________________________________________ -->
2620 <div class="doc_subsubsection"> <a name="i_unreachable">'<tt>unreachable</tt>'
2621 Instruction</a> </div>
2623 <div class="doc_text">
2631 <p>The '<tt>unreachable</tt>' instruction has no defined semantics. This
2632 instruction is used to inform the optimizer that a particular portion of the
2633 code is not reachable. This can be used to indicate that the code after a
2634 no-return function cannot be reached, and other facts.</p>
2637 <p>The '<tt>unreachable</tt>' instruction has no defined semantics.</p>
2641 <!-- ======================================================================= -->
2642 <div class="doc_subsection"> <a name="binaryops">Binary Operations</a> </div>
2644 <div class="doc_text">
2646 <p>Binary operators are used to do most of the computation in a program. They
2647 require two operands of the same type, execute an operation on them, and
2648 produce a single value. The operands might represent multiple data, as is
2649 the case with the <a href="#t_vector">vector</a> data type. The result value
2650 has the same type as its operands.</p>
2652 <p>There are several different binary operators:</p>
2656 <!-- _______________________________________________________________________ -->
2657 <div class="doc_subsubsection">
2658 <a name="i_add">'<tt>add</tt>' Instruction</a>
2661 <div class="doc_text">
2665 <result> = add <ty> <op1>, <op2> <i>; yields {ty}:result</i>
2666 <result> = nuw add <ty> <op1>, <op2> <i>; yields {ty}:result</i>
2667 <result> = nsw add <ty> <op1>, <op2> <i>; yields {ty}:result</i>
2668 <result> = nuw nsw add <ty> <op1>, <op2> <i>; yields {ty}:result</i>
2672 <p>The '<tt>add</tt>' instruction returns the sum of its two operands.</p>
2675 <p>The two arguments to the '<tt>add</tt>' instruction must
2676 be <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of
2677 integer values. Both arguments must have identical types.</p>
2680 <p>The value produced is the integer sum of the two operands.</p>
2682 <p>If the sum has unsigned overflow, the result returned is the mathematical
2683 result modulo 2<sup>n</sup>, where n is the bit width of the result.</p>
2685 <p>Because LLVM integers use a two's complement representation, this instruction
2686 is appropriate for both signed and unsigned integers.</p>
2688 <p><tt>nuw</tt> and <tt>nsw</tt> stand for "No Unsigned Wrap"
2689 and "No Signed Wrap", respectively. If the <tt>nuw</tt> and/or
2690 <tt>nsw</tt> keywords are present, the result value of the <tt>add</tt>
2691 is undefined if unsigned and/or signed overflow, respectively, occurs.</p>
2695 <result> = add i32 4, %var <i>; yields {i32}:result = 4 + %var</i>
2700 <!-- _______________________________________________________________________ -->
2701 <div class="doc_subsubsection">
2702 <a name="i_fadd">'<tt>fadd</tt>' Instruction</a>
2705 <div class="doc_text">
2709 <result> = fadd <ty> <op1>, <op2> <i>; yields {ty}:result</i>
2713 <p>The '<tt>fadd</tt>' instruction returns the sum of its two operands.</p>
2716 <p>The two arguments to the '<tt>fadd</tt>' instruction must be
2717 <a href="#t_floating">floating point</a> or <a href="#t_vector">vector</a> of
2718 floating point values. Both arguments must have identical types.</p>
2721 <p>The value produced is the floating point sum of the two operands.</p>
2725 <result> = fadd float 4.0, %var <i>; yields {float}:result = 4.0 + %var</i>
2730 <!-- _______________________________________________________________________ -->
2731 <div class="doc_subsubsection">
2732 <a name="i_sub">'<tt>sub</tt>' Instruction</a>
2735 <div class="doc_text">
2739 <result> = sub <ty> <op1>, <op2> <i>; yields {ty}:result</i>
2740 <result> = nuw sub <ty> <op1>, <op2> <i>; yields {ty}:result</i>
2741 <result> = nsw sub <ty> <op1>, <op2> <i>; yields {ty}:result</i>
2742 <result> = nuw nsw sub <ty> <op1>, <op2> <i>; yields {ty}:result</i>
2746 <p>The '<tt>sub</tt>' instruction returns the difference of its two
2749 <p>Note that the '<tt>sub</tt>' instruction is used to represent the
2750 '<tt>neg</tt>' instruction present in most other intermediate
2751 representations.</p>
2754 <p>The two arguments to the '<tt>sub</tt>' instruction must
2755 be <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of
2756 integer values. Both arguments must have identical types.</p>
2759 <p>The value produced is the integer difference of the two operands.</p>
2761 <p>If the difference has unsigned overflow, the result returned is the
2762 mathematical result modulo 2<sup>n</sup>, where n is the bit width of the
2765 <p>Because LLVM integers use a two's complement representation, this instruction
2766 is appropriate for both signed and unsigned integers.</p>
2768 <p><tt>nuw</tt> and <tt>nsw</tt> stand for "No Unsigned Wrap"
2769 and "No Signed Wrap", respectively. If the <tt>nuw</tt> and/or
2770 <tt>nsw</tt> keywords are present, the result value of the <tt>sub</tt>
2771 is undefined if unsigned and/or signed overflow, respectively, occurs.</p>
2775 <result> = sub i32 4, %var <i>; yields {i32}:result = 4 - %var</i>
2776 <result> = sub i32 0, %val <i>; yields {i32}:result = -%var</i>
2781 <!-- _______________________________________________________________________ -->
2782 <div class="doc_subsubsection">
2783 <a name="i_fsub">'<tt>fsub</tt>' Instruction</a>
2786 <div class="doc_text">
2790 <result> = fsub <ty> <op1>, <op2> <i>; yields {ty}:result</i>
2794 <p>The '<tt>fsub</tt>' instruction returns the difference of its two
2797 <p>Note that the '<tt>fsub</tt>' instruction is used to represent the
2798 '<tt>fneg</tt>' instruction present in most other intermediate
2799 representations.</p>
2802 <p>The two arguments to the '<tt>fsub</tt>' instruction must be
2803 <a href="#t_floating">floating point</a> or <a href="#t_vector">vector</a> of
2804 floating point values. Both arguments must have identical types.</p>
2807 <p>The value produced is the floating point difference of the two operands.</p>
2811 <result> = fsub float 4.0, %var <i>; yields {float}:result = 4.0 - %var</i>
2812 <result> = fsub float -0.0, %val <i>; yields {float}:result = -%var</i>
2817 <!-- _______________________________________________________________________ -->
2818 <div class="doc_subsubsection">
2819 <a name="i_mul">'<tt>mul</tt>' Instruction</a>
2822 <div class="doc_text">
2826 <result> = mul <ty> <op1>, <op2> <i>; yields {ty}:result</i>
2827 <result> = nuw mul <ty> <op1>, <op2> <i>; yields {ty}:result</i>
2828 <result> = nsw mul <ty> <op1>, <op2> <i>; yields {ty}:result</i>
2829 <result> = nuw nsw mul <ty> <op1>, <op2> <i>; yields {ty}:result</i>
2833 <p>The '<tt>mul</tt>' instruction returns the product of its two operands.</p>
2836 <p>The two arguments to the '<tt>mul</tt>' instruction must
2837 be <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of
2838 integer values. Both arguments must have identical types.</p>
2841 <p>The value produced is the integer product of the two operands.</p>
2843 <p>If the result of the multiplication has unsigned overflow, the result
2844 returned is the mathematical result modulo 2<sup>n</sup>, where n is the bit
2845 width of the result.</p>
2847 <p>Because LLVM integers use a two's complement representation, and the result
2848 is the same width as the operands, this instruction returns the correct
2849 result for both signed and unsigned integers. If a full product
2850 (e.g. <tt>i32</tt>x<tt>i32</tt>-><tt>i64</tt>) is needed, the operands should
2851 be sign-extended or zero-extended as appropriate to the width of the full
2854 <p><tt>nuw</tt> and <tt>nsw</tt> stand for "No Unsigned Wrap"
2855 and "No Signed Wrap", respectively. If the <tt>nuw</tt> and/or
2856 <tt>nsw</tt> keywords are present, the result value of the <tt>mul</tt>
2857 is undefined if unsigned and/or signed overflow, respectively, occurs.</p>
2861 <result> = mul i32 4, %var <i>; yields {i32}:result = 4 * %var</i>
2866 <!-- _______________________________________________________________________ -->
2867 <div class="doc_subsubsection">
2868 <a name="i_fmul">'<tt>fmul</tt>' Instruction</a>
2871 <div class="doc_text">
2875 <result> = fmul <ty> <op1>, <op2> <i>; yields {ty}:result</i>
2879 <p>The '<tt>fmul</tt>' instruction returns the product of its two operands.</p>
2882 <p>The two arguments to the '<tt>fmul</tt>' instruction must be
2883 <a href="#t_floating">floating point</a> or <a href="#t_vector">vector</a> of
2884 floating point values. Both arguments must have identical types.</p>
2887 <p>The value produced is the floating point product of the two operands.</p>
2891 <result> = fmul float 4.0, %var <i>; yields {float}:result = 4.0 * %var</i>
2896 <!-- _______________________________________________________________________ -->
2897 <div class="doc_subsubsection"> <a name="i_udiv">'<tt>udiv</tt>' Instruction
2900 <div class="doc_text">
2904 <result> = udiv <ty> <op1>, <op2> <i>; yields {ty}:result</i>
2908 <p>The '<tt>udiv</tt>' instruction returns the quotient of its two operands.</p>
2911 <p>The two arguments to the '<tt>udiv</tt>' instruction must be
2912 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
2913 values. Both arguments must have identical types.</p>
2916 <p>The value produced is the unsigned integer quotient of the two operands.</p>
2918 <p>Note that unsigned integer division and signed integer division are distinct
2919 operations; for signed integer division, use '<tt>sdiv</tt>'.</p>
2921 <p>Division by zero leads to undefined behavior.</p>
2925 <result> = udiv i32 4, %var <i>; yields {i32}:result = 4 / %var</i>
2930 <!-- _______________________________________________________________________ -->
2931 <div class="doc_subsubsection"> <a name="i_sdiv">'<tt>sdiv</tt>' Instruction
2934 <div class="doc_text">
2938 <result> = sdiv <ty> <op1>, <op2> <i>; yields {ty}:result</i>
2939 <result> = exact sdiv <ty> <op1>, <op2> <i>; yields {ty}:result</i>
2943 <p>The '<tt>sdiv</tt>' instruction returns the quotient of its two operands.</p>
2946 <p>The two arguments to the '<tt>sdiv</tt>' instruction must be
2947 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
2948 values. Both arguments must have identical types.</p>
2951 <p>The value produced is the signed integer quotient of the two operands rounded
2954 <p>Note that signed integer division and unsigned integer division are distinct
2955 operations; for unsigned integer division, use '<tt>udiv</tt>'.</p>
2957 <p>Division by zero leads to undefined behavior. Overflow also leads to
2958 undefined behavior; this is a rare case, but can occur, for example, by doing
2959 a 32-bit division of -2147483648 by -1.</p>
2961 <p>If the <tt>exact</tt> keyword is present, the result value of the
2962 <tt>sdiv</tt> is undefined if the result would be rounded or if overflow
2967 <result> = sdiv i32 4, %var <i>; yields {i32}:result = 4 / %var</i>
2972 <!-- _______________________________________________________________________ -->
2973 <div class="doc_subsubsection"> <a name="i_fdiv">'<tt>fdiv</tt>'
2974 Instruction</a> </div>
2976 <div class="doc_text">
2980 <result> = fdiv <ty> <op1>, <op2> <i>; yields {ty}:result</i>
2984 <p>The '<tt>fdiv</tt>' instruction returns the quotient of its two operands.</p>
2987 <p>The two arguments to the '<tt>fdiv</tt>' instruction must be
2988 <a href="#t_floating">floating point</a> or <a href="#t_vector">vector</a> of
2989 floating point values. Both arguments must have identical types.</p>
2992 <p>The value produced is the floating point quotient of the two operands.</p>
2996 <result> = fdiv float 4.0, %var <i>; yields {float}:result = 4.0 / %var</i>
3001 <!-- _______________________________________________________________________ -->
3002 <div class="doc_subsubsection"> <a name="i_urem">'<tt>urem</tt>' Instruction</a>
3005 <div class="doc_text">
3009 <result> = urem <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3013 <p>The '<tt>urem</tt>' instruction returns the remainder from the unsigned
3014 division of its two arguments.</p>
3017 <p>The two arguments to the '<tt>urem</tt>' instruction must be
3018 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
3019 values. Both arguments must have identical types.</p>
3022 <p>This instruction returns the unsigned integer <i>remainder</i> of a division.
3023 This instruction always performs an unsigned division to get the
3026 <p>Note that unsigned integer remainder and signed integer remainder are
3027 distinct operations; for signed integer remainder, use '<tt>srem</tt>'.</p>
3029 <p>Taking the remainder of a division by zero leads to undefined behavior.</p>
3033 <result> = urem i32 4, %var <i>; yields {i32}:result = 4 % %var</i>
3038 <!-- _______________________________________________________________________ -->
3039 <div class="doc_subsubsection">
3040 <a name="i_srem">'<tt>srem</tt>' Instruction</a>
3043 <div class="doc_text">
3047 <result> = srem <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3051 <p>The '<tt>srem</tt>' instruction returns the remainder from the signed
3052 division of its two operands. This instruction can also take
3053 <a href="#t_vector">vector</a> versions of the values in which case the
3054 elements must be integers.</p>
3057 <p>The two arguments to the '<tt>srem</tt>' instruction must be
3058 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
3059 values. Both arguments must have identical types.</p>
3062 <p>This instruction returns the <i>remainder</i> of a division (where the result
3063 has the same sign as the dividend, <tt>op1</tt>), not the <i>modulo</i>
3064 operator (where the result has the same sign as the divisor, <tt>op2</tt>) of
3065 a value. For more information about the difference,
3066 see <a href="http://mathforum.org/dr.math/problems/anne.4.28.99.html">The
3067 Math Forum</a>. For a table of how this is implemented in various languages,
3068 please see <a href="http://en.wikipedia.org/wiki/Modulo_operation">
3069 Wikipedia: modulo operation</a>.</p>
3071 <p>Note that signed integer remainder and unsigned integer remainder are
3072 distinct operations; for unsigned integer remainder, use '<tt>urem</tt>'.</p>
3074 <p>Taking the remainder of a division by zero leads to undefined behavior.
3075 Overflow also leads to undefined behavior; this is a rare case, but can
3076 occur, for example, by taking the remainder of a 32-bit division of
3077 -2147483648 by -1. (The remainder doesn't actually overflow, but this rule
3078 lets srem be implemented using instructions that return both the result of
3079 the division and the remainder.)</p>
3083 <result> = srem i32 4, %var <i>; yields {i32}:result = 4 % %var</i>
3088 <!-- _______________________________________________________________________ -->
3089 <div class="doc_subsubsection">
3090 <a name="i_frem">'<tt>frem</tt>' Instruction</a> </div>
3092 <div class="doc_text">
3096 <result> = frem <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3100 <p>The '<tt>frem</tt>' instruction returns the remainder from the division of
3101 its two operands.</p>
3104 <p>The two arguments to the '<tt>frem</tt>' instruction must be
3105 <a href="#t_floating">floating point</a> or <a href="#t_vector">vector</a> of
3106 floating point values. Both arguments must have identical types.</p>
3109 <p>This instruction returns the <i>remainder</i> of a division. The remainder
3110 has the same sign as the dividend.</p>
3114 <result> = frem float 4.0, %var <i>; yields {float}:result = 4.0 % %var</i>
3119 <!-- ======================================================================= -->
3120 <div class="doc_subsection"> <a name="bitwiseops">Bitwise Binary
3121 Operations</a> </div>
3123 <div class="doc_text">
3125 <p>Bitwise binary operators are used to do various forms of bit-twiddling in a
3126 program. They are generally very efficient instructions and can commonly be
3127 strength reduced from other instructions. They require two operands of the
3128 same type, execute an operation on them, and produce a single value. The
3129 resulting value is the same type as its operands.</p>
3133 <!-- _______________________________________________________________________ -->
3134 <div class="doc_subsubsection"> <a name="i_shl">'<tt>shl</tt>'
3135 Instruction</a> </div>
3137 <div class="doc_text">
3141 <result> = shl <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3145 <p>The '<tt>shl</tt>' instruction returns the first operand shifted to the left
3146 a specified number of bits.</p>
3149 <p>Both arguments to the '<tt>shl</tt>' instruction must be the
3150 same <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of
3151 integer type. '<tt>op2</tt>' is treated as an unsigned value.</p>
3154 <p>The value produced is <tt>op1</tt> * 2<sup><tt>op2</tt></sup> mod
3155 2<sup>n</sup>, where <tt>n</tt> is the width of the result. If <tt>op2</tt>
3156 is (statically or dynamically) negative or equal to or larger than the number
3157 of bits in <tt>op1</tt>, the result is undefined. If the arguments are
3158 vectors, each vector element of <tt>op1</tt> is shifted by the corresponding
3159 shift amount in <tt>op2</tt>.</p>
3163 <result> = shl i32 4, %var <i>; yields {i32}: 4 << %var</i>
3164 <result> = shl i32 4, 2 <i>; yields {i32}: 16</i>
3165 <result> = shl i32 1, 10 <i>; yields {i32}: 1024</i>
3166 <result> = shl i32 1, 32 <i>; undefined</i>
3167 <result> = shl <2 x i32> < i32 1, i32 1>, < i32 1, i32 2> <i>; yields: result=<2 x i32> < i32 2, i32 4></i>
3172 <!-- _______________________________________________________________________ -->
3173 <div class="doc_subsubsection"> <a name="i_lshr">'<tt>lshr</tt>'
3174 Instruction</a> </div>
3176 <div class="doc_text">
3180 <result> = lshr <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3184 <p>The '<tt>lshr</tt>' instruction (logical shift right) returns the first
3185 operand shifted to the right a specified number of bits with zero fill.</p>
3188 <p>Both arguments to the '<tt>lshr</tt>' instruction must be the same
3189 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
3190 type. '<tt>op2</tt>' is treated as an unsigned value.</p>
3193 <p>This instruction always performs a logical shift right operation. The most
3194 significant bits of the result will be filled with zero bits after the shift.
3195 If <tt>op2</tt> is (statically or dynamically) equal to or larger than the
3196 number of bits in <tt>op1</tt>, the result is undefined. If the arguments are
3197 vectors, each vector element of <tt>op1</tt> is shifted by the corresponding
3198 shift amount in <tt>op2</tt>.</p>
3202 <result> = lshr i32 4, 1 <i>; yields {i32}:result = 2</i>
3203 <result> = lshr i32 4, 2 <i>; yields {i32}:result = 1</i>
3204 <result> = lshr i8 4, 3 <i>; yields {i8}:result = 0</i>
3205 <result> = lshr i8 -2, 1 <i>; yields {i8}:result = 0x7FFFFFFF </i>
3206 <result> = lshr i32 1, 32 <i>; undefined</i>
3207 <result> = lshr <2 x i32> < i32 -2, i32 4>, < i32 1, i32 2> <i>; yields: result=<2 x i32> < i32 0x7FFFFFFF, i32 1></i>
3212 <!-- _______________________________________________________________________ -->
3213 <div class="doc_subsubsection"> <a name="i_ashr">'<tt>ashr</tt>'
3214 Instruction</a> </div>
3215 <div class="doc_text">
3219 <result> = ashr <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3223 <p>The '<tt>ashr</tt>' instruction (arithmetic shift right) returns the first
3224 operand shifted to the right a specified number of bits with sign
3228 <p>Both arguments to the '<tt>ashr</tt>' instruction must be the same
3229 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
3230 type. '<tt>op2</tt>' is treated as an unsigned value.</p>
3233 <p>This instruction always performs an arithmetic shift right operation, The
3234 most significant bits of the result will be filled with the sign bit
3235 of <tt>op1</tt>. If <tt>op2</tt> is (statically or dynamically) equal to or
3236 larger than the number of bits in <tt>op1</tt>, the result is undefined. If
3237 the arguments are vectors, each vector element of <tt>op1</tt> is shifted by
3238 the corresponding shift amount in <tt>op2</tt>.</p>
3242 <result> = ashr i32 4, 1 <i>; yields {i32}:result = 2</i>
3243 <result> = ashr i32 4, 2 <i>; yields {i32}:result = 1</i>
3244 <result> = ashr i8 4, 3 <i>; yields {i8}:result = 0</i>
3245 <result> = ashr i8 -2, 1 <i>; yields {i8}:result = -1</i>
3246 <result> = ashr i32 1, 32 <i>; undefined</i>
3247 <result> = ashr <2 x i32> < i32 -2, i32 4>, < i32 1, i32 3> <i>; yields: result=<2 x i32> < i32 -1, i32 0></i>
3252 <!-- _______________________________________________________________________ -->
3253 <div class="doc_subsubsection"> <a name="i_and">'<tt>and</tt>'
3254 Instruction</a> </div>
3256 <div class="doc_text">
3260 <result> = and <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3264 <p>The '<tt>and</tt>' instruction returns the bitwise logical and of its two
3268 <p>The two arguments to the '<tt>and</tt>' instruction must be
3269 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
3270 values. Both arguments must have identical types.</p>
3273 <p>The truth table used for the '<tt>and</tt>' instruction is:</p>
3275 <table border="1" cellspacing="0" cellpadding="4">
3307 <result> = and i32 4, %var <i>; yields {i32}:result = 4 & %var</i>
3308 <result> = and i32 15, 40 <i>; yields {i32}:result = 8</i>
3309 <result> = and i32 4, 8 <i>; yields {i32}:result = 0</i>
3312 <!-- _______________________________________________________________________ -->
3313 <div class="doc_subsubsection"> <a name="i_or">'<tt>or</tt>' Instruction</a> </div>
3315 <div class="doc_text">
3319 <result> = or <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3323 <p>The '<tt>or</tt>' instruction returns the bitwise logical inclusive or of its
3327 <p>The two arguments to the '<tt>or</tt>' instruction must be
3328 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
3329 values. Both arguments must have identical types.</p>
3332 <p>The truth table used for the '<tt>or</tt>' instruction is:</p>
3334 <table border="1" cellspacing="0" cellpadding="4">
3366 <result> = or i32 4, %var <i>; yields {i32}:result = 4 | %var</i>
3367 <result> = or i32 15, 40 <i>; yields {i32}:result = 47</i>
3368 <result> = or i32 4, 8 <i>; yields {i32}:result = 12</i>
3373 <!-- _______________________________________________________________________ -->
3374 <div class="doc_subsubsection"> <a name="i_xor">'<tt>xor</tt>'
3375 Instruction</a> </div>
3377 <div class="doc_text">
3381 <result> = xor <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3385 <p>The '<tt>xor</tt>' instruction returns the bitwise logical exclusive or of
3386 its two operands. The <tt>xor</tt> is used to implement the "one's
3387 complement" operation, which is the "~" operator in C.</p>
3390 <p>The two arguments to the '<tt>xor</tt>' instruction must be
3391 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
3392 values. Both arguments must have identical types.</p>
3395 <p>The truth table used for the '<tt>xor</tt>' instruction is:</p>
3397 <table border="1" cellspacing="0" cellpadding="4">
3429 <result> = xor i32 4, %var <i>; yields {i32}:result = 4 ^ %var</i>
3430 <result> = xor i32 15, 40 <i>; yields {i32}:result = 39</i>
3431 <result> = xor i32 4, 8 <i>; yields {i32}:result = 12</i>
3432 <result> = xor i32 %V, -1 <i>; yields {i32}:result = ~%V</i>
3437 <!-- ======================================================================= -->
3438 <div class="doc_subsection">
3439 <a name="vectorops">Vector Operations</a>
3442 <div class="doc_text">
3444 <p>LLVM supports several instructions to represent vector operations in a
3445 target-independent manner. These instructions cover the element-access and
3446 vector-specific operations needed to process vectors effectively. While LLVM
3447 does directly support these vector operations, many sophisticated algorithms
3448 will want to use target-specific intrinsics to take full advantage of a
3449 specific target.</p>
3453 <!-- _______________________________________________________________________ -->
3454 <div class="doc_subsubsection">
3455 <a name="i_extractelement">'<tt>extractelement</tt>' Instruction</a>
3458 <div class="doc_text">
3462 <result> = extractelement <n x <ty>> <val>, i32 <idx> <i>; yields <ty></i>
3466 <p>The '<tt>extractelement</tt>' instruction extracts a single scalar element
3467 from a vector at a specified index.</p>
3471 <p>The first operand of an '<tt>extractelement</tt>' instruction is a value
3472 of <a href="#t_vector">vector</a> type. The second operand is an index
3473 indicating the position from which to extract the element. The index may be
3477 <p>The result is a scalar of the same type as the element type of
3478 <tt>val</tt>. Its value is the value at position <tt>idx</tt> of
3479 <tt>val</tt>. If <tt>idx</tt> exceeds the length of <tt>val</tt>, the
3480 results are undefined.</p>
3484 %result = extractelement <4 x i32> %vec, i32 0 <i>; yields i32</i>
3489 <!-- _______________________________________________________________________ -->
3490 <div class="doc_subsubsection">
3491 <a name="i_insertelement">'<tt>insertelement</tt>' Instruction</a>
3494 <div class="doc_text">
3498 <result> = insertelement <n x <ty>> <val>, <ty> <elt>, i32 <idx> <i>; yields <n x <ty>></i>
3502 <p>The '<tt>insertelement</tt>' instruction inserts a scalar element into a
3503 vector at a specified index.</p>
3506 <p>The first operand of an '<tt>insertelement</tt>' instruction is a value
3507 of <a href="#t_vector">vector</a> type. The second operand is a scalar value
3508 whose type must equal the element type of the first operand. The third
3509 operand is an index indicating the position at which to insert the value.
3510 The index may be a variable.</p>
3513 <p>The result is a vector of the same type as <tt>val</tt>. Its element values
3514 are those of <tt>val</tt> except at position <tt>idx</tt>, where it gets the
3515 value <tt>elt</tt>. If <tt>idx</tt> exceeds the length of <tt>val</tt>, the
3516 results are undefined.</p>
3520 %result = insertelement <4 x i32> %vec, i32 1, i32 0 <i>; yields <4 x i32></i>
3525 <!-- _______________________________________________________________________ -->
3526 <div class="doc_subsubsection">
3527 <a name="i_shufflevector">'<tt>shufflevector</tt>' Instruction</a>
3530 <div class="doc_text">
3534 <result> = shufflevector <n x <ty>> <v1>, <n x <ty>> <v2>, <m x i32> <mask> <i>; yields <m x <ty>></i>
3538 <p>The '<tt>shufflevector</tt>' instruction constructs a permutation of elements
3539 from two input vectors, returning a vector with the same element type as the
3540 input and length that is the same as the shuffle mask.</p>
3543 <p>The first two operands of a '<tt>shufflevector</tt>' instruction are vectors
3544 with types that match each other. The third argument is a shuffle mask whose
3545 element type is always 'i32'. The result of the instruction is a vector
3546 whose length is the same as the shuffle mask and whose element type is the
3547 same as the element type of the first two operands.</p>
3549 <p>The shuffle mask operand is required to be a constant vector with either
3550 constant integer or undef values.</p>
3553 <p>The elements of the two input vectors are numbered from left to right across
3554 both of the vectors. The shuffle mask operand specifies, for each element of
3555 the result vector, which element of the two input vectors the result element
3556 gets. The element selector may be undef (meaning "don't care") and the
3557 second operand may be undef if performing a shuffle from only one vector.</p>
3561 %result = shufflevector <4 x i32> %v1, <4 x i32> %v2,
3562 <4 x i32> <i32 0, i32 4, i32 1, i32 5> <i>; yields <4 x i32></i>
3563 %result = shufflevector <4 x i32> %v1, <4 x i32> undef,
3564 <4 x i32> <i32 0, i32 1, i32 2, i32 3> <i>; yields <4 x i32></i> - Identity shuffle.
3565 %result = shufflevector <8 x i32> %v1, <8 x i32> undef,
3566 <4 x i32> <i32 0, i32 1, i32 2, i32 3> <i>; yields <4 x i32></i>
3567 %result = shufflevector <4 x i32> %v1, <4 x i32> %v2,
3568 <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>
3573 <!-- ======================================================================= -->
3574 <div class="doc_subsection">
3575 <a name="aggregateops">Aggregate Operations</a>
3578 <div class="doc_text">
3580 <p>LLVM supports several instructions for working with aggregate values.</p>
3584 <!-- _______________________________________________________________________ -->
3585 <div class="doc_subsubsection">
3586 <a name="i_extractvalue">'<tt>extractvalue</tt>' Instruction</a>
3589 <div class="doc_text">
3593 <result> = extractvalue <aggregate type> <val>, <idx>{, <idx>}*
3597 <p>The '<tt>extractvalue</tt>' instruction extracts the value of a struct field
3598 or array element from an aggregate value.</p>
3601 <p>The first operand of an '<tt>extractvalue</tt>' instruction is a value
3602 of <a href="#t_struct">struct</a> or <a href="#t_array">array</a> type. The
3603 operands are constant indices to specify which value to extract in a similar
3604 manner as indices in a
3605 '<tt><a href="#i_getelementptr">getelementptr</a></tt>' instruction.</p>
3608 <p>The result is the value at the position in the aggregate specified by the
3613 %result = extractvalue {i32, float} %agg, 0 <i>; yields i32</i>
3618 <!-- _______________________________________________________________________ -->
3619 <div class="doc_subsubsection">
3620 <a name="i_insertvalue">'<tt>insertvalue</tt>' Instruction</a>
3623 <div class="doc_text">
3627 <result> = insertvalue <aggregate type> <val>, <ty> <val>, <idx> <i>; yields <n x <ty>></i>
3631 <p>The '<tt>insertvalue</tt>' instruction inserts a value into a struct field or
3632 array element in an aggregate.</p>
3636 <p>The first operand of an '<tt>insertvalue</tt>' instruction is a value
3637 of <a href="#t_struct">struct</a> or <a href="#t_array">array</a> type. The
3638 second operand is a first-class value to insert. The following operands are
3639 constant indices indicating the position at which to insert the value in a
3640 similar manner as indices in a
3641 '<tt><a href="#i_getelementptr">getelementptr</a></tt>' instruction. The
3642 value to insert must have the same type as the value identified by the
3646 <p>The result is an aggregate of the same type as <tt>val</tt>. Its value is
3647 that of <tt>val</tt> except that the value at the position specified by the
3648 indices is that of <tt>elt</tt>.</p>
3652 %result = insertvalue {i32, float} %agg, i32 1, 0 <i>; yields {i32, float}</i>
3658 <!-- ======================================================================= -->
3659 <div class="doc_subsection">
3660 <a name="memoryops">Memory Access and Addressing Operations</a>
3663 <div class="doc_text">
3665 <p>A key design point of an SSA-based representation is how it represents
3666 memory. In LLVM, no memory locations are in SSA form, which makes things
3667 very simple. This section describes how to read, write, allocate, and free
3672 <!-- _______________________________________________________________________ -->
3673 <div class="doc_subsubsection">
3674 <a name="i_malloc">'<tt>malloc</tt>' Instruction</a>
3677 <div class="doc_text">
3681 <result> = malloc <type>[, i32 <NumElements>][, align <alignment>] <i>; yields {type*}:result</i>
3685 <p>The '<tt>malloc</tt>' instruction allocates memory from the system heap and
3686 returns a pointer to it. The object is always allocated in the generic
3687 address space (address space zero).</p>
3690 <p>The '<tt>malloc</tt>' instruction allocates
3691 <tt>sizeof(<type>)*NumElements</tt> bytes of memory from the operating
3692 system and returns a pointer of the appropriate type to the program. If
3693 "NumElements" is specified, it is the number of elements allocated, otherwise
3694 "NumElements" is defaulted to be one. If a constant alignment is specified,
3695 the value result of the allocation is guaranteed to be aligned to at least
3696 that boundary. If not specified, or if zero, the target can choose to align
3697 the allocation on any convenient boundary compatible with the type.</p>
3699 <p>'<tt>type</tt>' must be a sized type.</p>
3702 <p>Memory is allocated using the system "<tt>malloc</tt>" function, and a
3703 pointer is returned. The result of a zero byte allocation is undefined. The
3704 result is null if there is insufficient memory available.</p>
3708 %array = malloc [4 x i8] <i>; yields {[%4 x i8]*}:array</i>
3710 %size = <a href="#i_add">add</a> i32 2, 2 <i>; yields {i32}:size = i32 4</i>
3711 %array1 = malloc i8, i32 4 <i>; yields {i8*}:array1</i>
3712 %array2 = malloc [12 x i8], i32 %size <i>; yields {[12 x i8]*}:array2</i>
3713 %array3 = malloc i32, i32 4, align 1024 <i>; yields {i32*}:array3</i>
3714 %array4 = malloc i32, align 1024 <i>; yields {i32*}:array4</i>
3717 <p>Note that the code generator does not yet respect the alignment value.</p>
3721 <!-- _______________________________________________________________________ -->
3722 <div class="doc_subsubsection">
3723 <a name="i_free">'<tt>free</tt>' Instruction</a>
3726 <div class="doc_text">
3730 free <type> <value> <i>; yields {void}</i>
3734 <p>The '<tt>free</tt>' instruction returns memory back to the unused memory heap
3735 to be reallocated in the future.</p>
3738 <p>'<tt>value</tt>' shall be a pointer value that points to a value that was
3739 allocated with the '<tt><a href="#i_malloc">malloc</a></tt>' instruction.</p>
3742 <p>Access to the memory pointed to by the pointer is no longer defined after
3743 this instruction executes. If the pointer is null, the operation is a
3748 %array = <a href="#i_malloc">malloc</a> [4 x i8] <i>; yields {[4 x i8]*}:array</i>
3749 free [4 x i8]* %array
3754 <!-- _______________________________________________________________________ -->
3755 <div class="doc_subsubsection">
3756 <a name="i_alloca">'<tt>alloca</tt>' Instruction</a>
3759 <div class="doc_text">
3763 <result> = alloca <type>[, i32 <NumElements>][, align <alignment>] <i>; yields {type*}:result</i>
3767 <p>The '<tt>alloca</tt>' instruction allocates memory on the stack frame of the
3768 currently executing function, to be automatically released when this function
3769 returns to its caller. The object is always allocated in the generic address
3770 space (address space zero).</p>
3773 <p>The '<tt>alloca</tt>' instruction
3774 allocates <tt>sizeof(<type>)*NumElements</tt> bytes of memory on the
3775 runtime stack, returning a pointer of the appropriate type to the program.
3776 If "NumElements" is specified, it is the number of elements allocated,
3777 otherwise "NumElements" is defaulted to be one. If a constant alignment is
3778 specified, the value result of the allocation is guaranteed to be aligned to
3779 at least that boundary. If not specified, or if zero, the target can choose
3780 to align the allocation on any convenient boundary compatible with the
3783 <p>'<tt>type</tt>' may be any sized type.</p>
3786 <p>Memory is allocated; a pointer is returned. The operation is undefined if
3787 there is insufficient stack space for the allocation. '<tt>alloca</tt>'d
3788 memory is automatically released when the function returns. The
3789 '<tt>alloca</tt>' instruction is commonly used to represent automatic
3790 variables that must have an address available. When the function returns
3791 (either with the <tt><a href="#i_ret">ret</a></tt>
3792 or <tt><a href="#i_unwind">unwind</a></tt> instructions), the memory is
3793 reclaimed. Allocating zero bytes is legal, but the result is undefined.</p>
3797 %ptr = alloca i32 <i>; yields {i32*}:ptr</i>
3798 %ptr = alloca i32, i32 4 <i>; yields {i32*}:ptr</i>
3799 %ptr = alloca i32, i32 4, align 1024 <i>; yields {i32*}:ptr</i>
3800 %ptr = alloca i32, align 1024 <i>; yields {i32*}:ptr</i>
3805 <!-- _______________________________________________________________________ -->
3806 <div class="doc_subsubsection"> <a name="i_load">'<tt>load</tt>'
3807 Instruction</a> </div>
3809 <div class="doc_text">
3813 <result> = load <ty>* <pointer>[, align <alignment>]
3814 <result> = volatile load <ty>* <pointer>[, align <alignment>]
3818 <p>The '<tt>load</tt>' instruction is used to read from memory.</p>
3821 <p>The argument to the '<tt>load</tt>' instruction specifies the memory address
3822 from which to load. The pointer must point to
3823 a <a href="#t_firstclass">first class</a> type. If the <tt>load</tt> is
3824 marked as <tt>volatile</tt>, then the optimizer is not allowed to modify the
3825 number or order of execution of this <tt>load</tt> with other
3826 volatile <tt>load</tt> and <tt><a href="#i_store">store</a></tt>
3829 <p>The optional constant "align" argument specifies the alignment of the
3830 operation (that is, the alignment of the memory address). A value of 0 or an
3831 omitted "align" argument means that the operation has the preferential
3832 alignment for the target. It is the responsibility of the code emitter to
3833 ensure that the alignment information is correct. Overestimating the
3834 alignment results in an undefined behavior. Underestimating the alignment may
3835 produce less efficient code. An alignment of 1 is always safe.</p>
3838 <p>The location of memory pointed to is loaded. If the value being loaded is of
3839 scalar type then the number of bytes read does not exceed the minimum number
3840 of bytes needed to hold all bits of the type. For example, loading an
3841 <tt>i24</tt> reads at most three bytes. When loading a value of a type like
3842 <tt>i20</tt> with a size that is not an integral number of bytes, the result
3843 is undefined if the value was not originally written using a store of the
3848 %ptr = <a href="#i_alloca">alloca</a> i32 <i>; yields {i32*}:ptr</i>
3849 <a href="#i_store">store</a> i32 3, i32* %ptr <i>; yields {void}</i>
3850 %val = load i32* %ptr <i>; yields {i32}:val = i32 3</i>
3855 <!-- _______________________________________________________________________ -->
3856 <div class="doc_subsubsection"> <a name="i_store">'<tt>store</tt>'
3857 Instruction</a> </div>
3859 <div class="doc_text">
3863 store <ty> <value>, <ty>* <pointer>[, align <alignment>] <i>; yields {void}</i>
3864 volatile store <ty> <value>, <ty>* <pointer>[, align <alignment>] <i>; yields {void}</i>
3868 <p>The '<tt>store</tt>' instruction is used to write to memory.</p>
3871 <p>There are two arguments to the '<tt>store</tt>' instruction: a value to store
3872 and an address at which to store it. The type of the
3873 '<tt><pointer></tt>' operand must be a pointer to
3874 the <a href="#t_firstclass">first class</a> type of the
3875 '<tt><value></tt>' operand. If the <tt>store</tt> is marked
3876 as <tt>volatile</tt>, then the optimizer is not allowed to modify the number
3877 or order of execution of this <tt>store</tt> with other
3878 volatile <tt>load</tt> and <tt><a href="#i_store">store</a></tt>
3881 <p>The optional constant "align" argument specifies the alignment of the
3882 operation (that is, the alignment of the memory address). A value of 0 or an
3883 omitted "align" argument means that the operation has the preferential
3884 alignment for the target. It is the responsibility of the code emitter to
3885 ensure that the alignment information is correct. Overestimating the
3886 alignment results in an undefined behavior. Underestimating the alignment may
3887 produce less efficient code. An alignment of 1 is always safe.</p>
3890 <p>The contents of memory are updated to contain '<tt><value></tt>' at the
3891 location specified by the '<tt><pointer></tt>' operand. If
3892 '<tt><value></tt>' is of scalar type then the number of bytes written
3893 does not exceed the minimum number of bytes needed to hold all bits of the
3894 type. For example, storing an <tt>i24</tt> writes at most three bytes. When
3895 writing a value of a type like <tt>i20</tt> with a size that is not an
3896 integral number of bytes, it is unspecified what happens to the extra bits
3897 that do not belong to the type, but they will typically be overwritten.</p>
3901 %ptr = <a href="#i_alloca">alloca</a> i32 <i>; yields {i32*}:ptr</i>
3902 store i32 3, i32* %ptr <i>; yields {void}</i>
3903 %val = <a href="#i_load">load</a> i32* %ptr <i>; yields {i32}:val = i32 3</i>
3908 <!-- _______________________________________________________________________ -->
3909 <div class="doc_subsubsection">
3910 <a name="i_getelementptr">'<tt>getelementptr</tt>' Instruction</a>
3913 <div class="doc_text">
3917 <result> = getelementptr <pty>* <ptrval>{, <ty> <idx>}*
3918 <result> = getelementptr inbounds <pty>* <ptrval>{, <ty> <idx>}*
3922 <p>The '<tt>getelementptr</tt>' instruction is used to get the address of a
3923 subelement of an aggregate data structure. It performs address calculation
3924 only and does not access memory.</p>
3927 <p>The first argument is always a pointer, and forms the basis of the
3928 calculation. The remaining arguments are indices that indicate which of the
3929 elements of the aggregate object are indexed. The interpretation of each
3930 index is dependent on the type being indexed into. The first index always
3931 indexes the pointer value given as the first argument, the second index
3932 indexes a value of the type pointed to (not necessarily the value directly
3933 pointed to, since the first index can be non-zero), etc. The first type
3934 indexed into must be a pointer value, subsequent types can be arrays, vectors
3935 and structs. Note that subsequent types being indexed into can never be
3936 pointers, since that would require loading the pointer before continuing
3939 <p>The type of each index argument depends on the type it is indexing into.
3940 When indexing into a (optionally packed) structure, only <tt>i32</tt> integer
3941 <b>constants</b> are allowed. When indexing into an array, pointer or
3942 vector, integers of any width are allowed, and they are not required to be
3945 <p>For example, let's consider a C code fragment and how it gets compiled to
3948 <div class="doc_code">
3961 int *foo(struct ST *s) {
3962 return &s[1].Z.B[5][13];
3967 <p>The LLVM code generated by the GCC frontend is:</p>
3969 <div class="doc_code">
3971 %RT = <a href="#namedtypes">type</a> { i8 , [10 x [20 x i32]], i8 }
3972 %ST = <a href="#namedtypes">type</a> { i32, double, %RT }
3974 define i32* @foo(%ST* %s) {
3976 %reg = getelementptr %ST* %s, i32 1, i32 2, i32 1, i32 5, i32 13
3983 <p>In the example above, the first index is indexing into the '<tt>%ST*</tt>'
3984 type, which is a pointer, yielding a '<tt>%ST</tt>' = '<tt>{ i32, double, %RT
3985 }</tt>' type, a structure. The second index indexes into the third element
3986 of the structure, yielding a '<tt>%RT</tt>' = '<tt>{ i8 , [10 x [20 x i32]],
3987 i8 }</tt>' type, another structure. The third index indexes into the second
3988 element of the structure, yielding a '<tt>[10 x [20 x i32]]</tt>' type, an
3989 array. The two dimensions of the array are subscripted into, yielding an
3990 '<tt>i32</tt>' type. The '<tt>getelementptr</tt>' instruction returns a
3991 pointer to this element, thus computing a value of '<tt>i32*</tt>' type.</p>
3993 <p>Note that it is perfectly legal to index partially through a structure,
3994 returning a pointer to an inner element. Because of this, the LLVM code for
3995 the given testcase is equivalent to:</p>
3998 define i32* @foo(%ST* %s) {
3999 %t1 = getelementptr %ST* %s, i32 1 <i>; yields %ST*:%t1</i>
4000 %t2 = getelementptr %ST* %t1, i32 0, i32 2 <i>; yields %RT*:%t2</i>
4001 %t3 = getelementptr %RT* %t2, i32 0, i32 1 <i>; yields [10 x [20 x i32]]*:%t3</i>
4002 %t4 = getelementptr [10 x [20 x i32]]* %t3, i32 0, i32 5 <i>; yields [20 x i32]*:%t4</i>
4003 %t5 = getelementptr [20 x i32]* %t4, i32 0, i32 13 <i>; yields i32*:%t5</i>
4008 <p>If the <tt>inbounds</tt> keyword is present, the result value of the
4009 <tt>getelementptr</tt> is undefined if the base pointer is not an
4010 <i>in bounds</i> address of an allocated object, or if any of the addresses
4011 that would be formed by successive addition of the offsets implied by the
4012 indices to the base address with infinitely precise arithmetic are not an
4013 <i>in bounds</i> address of that allocated object.
4014 The <i>in bounds</i> addresses for an allocated object are all the addresses
4015 that point into the object, plus the address one byte past the end.</p>
4017 <p>If the <tt>inbounds</tt> keyword is not present, the offsets are added to
4018 the base address with silently-wrapping two's complement arithmetic, and
4019 the result value of the <tt>getelementptr</tt> may be outside the object
4020 pointed to by the base pointer. The result value may not necessarily be
4021 used to access memory though, even if it happens to point into allocated
4022 storage. See the <a href="#pointeraliasing">Pointer Aliasing Rules</a>
4023 section for more information.</p>
4025 <p>The getelementptr instruction is often confusing. For some more insight into
4026 how it works, see <a href="GetElementPtr.html">the getelementptr FAQ</a>.</p>
4030 <i>; yields [12 x i8]*:aptr</i>
4031 %aptr = getelementptr {i32, [12 x i8]}* %saptr, i64 0, i32 1
4032 <i>; yields i8*:vptr</i>
4033 %vptr = getelementptr {i32, <2 x i8>}* %svptr, i64 0, i32 1, i32 1
4034 <i>; yields i8*:eptr</i>
4035 %eptr = getelementptr [12 x i8]* %aptr, i64 0, i32 1
4036 <i>; yields i32*:iptr</i>
4037 %iptr = getelementptr [10 x i32]* @arr, i16 0, i16 0
4042 <!-- ======================================================================= -->
4043 <div class="doc_subsection"> <a name="convertops">Conversion Operations</a>
4046 <div class="doc_text">
4048 <p>The instructions in this category are the conversion instructions (casting)
4049 which all take a single operand and a type. They perform various bit
4050 conversions on the operand.</p>
4054 <!-- _______________________________________________________________________ -->
4055 <div class="doc_subsubsection">
4056 <a name="i_trunc">'<tt>trunc .. to</tt>' Instruction</a>
4058 <div class="doc_text">
4062 <result> = trunc <ty> <value> to <ty2> <i>; yields ty2</i>
4066 <p>The '<tt>trunc</tt>' instruction truncates its operand to the
4067 type <tt>ty2</tt>.</p>
4070 <p>The '<tt>trunc</tt>' instruction takes a <tt>value</tt> to trunc, which must
4071 be an <a href="#t_integer">integer</a> type, and a type that specifies the
4072 size and type of the result, which must be
4073 an <a href="#t_integer">integer</a> type. The bit size of <tt>value</tt> must
4074 be larger than the bit size of <tt>ty2</tt>. Equal sized types are not
4078 <p>The '<tt>trunc</tt>' instruction truncates the high order bits
4079 in <tt>value</tt> and converts the remaining bits to <tt>ty2</tt>. Since the
4080 source size must be larger than the destination size, <tt>trunc</tt> cannot
4081 be a <i>no-op cast</i>. It will always truncate bits.</p>
4085 %X = trunc i32 257 to i8 <i>; yields i8:1</i>
4086 %Y = trunc i32 123 to i1 <i>; yields i1:true</i>
4087 %Y = trunc i32 122 to i1 <i>; yields i1:false</i>
4092 <!-- _______________________________________________________________________ -->
4093 <div class="doc_subsubsection">
4094 <a name="i_zext">'<tt>zext .. to</tt>' Instruction</a>
4096 <div class="doc_text">
4100 <result> = zext <ty> <value> to <ty2> <i>; yields ty2</i>
4104 <p>The '<tt>zext</tt>' instruction zero extends its operand to type
4109 <p>The '<tt>zext</tt>' instruction takes a value to cast, which must be of
4110 <a href="#t_integer">integer</a> type, and a type to cast it to, which must
4111 also be of <a href="#t_integer">integer</a> type. The bit size of the
4112 <tt>value</tt> must be smaller than the bit size of the destination type,
4116 <p>The <tt>zext</tt> fills the high order bits of the <tt>value</tt> with zero
4117 bits until it reaches the size of the destination type, <tt>ty2</tt>.</p>
4119 <p>When zero extending from i1, the result will always be either 0 or 1.</p>
4123 %X = zext i32 257 to i64 <i>; yields i64:257</i>
4124 %Y = zext i1 true to i32 <i>; yields i32:1</i>
4129 <!-- _______________________________________________________________________ -->
4130 <div class="doc_subsubsection">
4131 <a name="i_sext">'<tt>sext .. to</tt>' Instruction</a>
4133 <div class="doc_text">
4137 <result> = sext <ty> <value> to <ty2> <i>; yields ty2</i>
4141 <p>The '<tt>sext</tt>' sign extends <tt>value</tt> to the type <tt>ty2</tt>.</p>
4144 <p>The '<tt>sext</tt>' instruction takes a value to cast, which must be of
4145 <a href="#t_integer">integer</a> type, and a type to cast it to, which must
4146 also be of <a href="#t_integer">integer</a> type. The bit size of the
4147 <tt>value</tt> must be smaller than the bit size of the destination type,
4151 <p>The '<tt>sext</tt>' instruction performs a sign extension by copying the sign
4152 bit (highest order bit) of the <tt>value</tt> until it reaches the bit size
4153 of the type <tt>ty2</tt>.</p>
4155 <p>When sign extending from i1, the extension always results in -1 or 0.</p>
4159 %X = sext i8 -1 to i16 <i>; yields i16 :65535</i>
4160 %Y = sext i1 true to i32 <i>; yields i32:-1</i>
4165 <!-- _______________________________________________________________________ -->
4166 <div class="doc_subsubsection">
4167 <a name="i_fptrunc">'<tt>fptrunc .. to</tt>' Instruction</a>
4170 <div class="doc_text">
4174 <result> = fptrunc <ty> <value> to <ty2> <i>; yields ty2</i>
4178 <p>The '<tt>fptrunc</tt>' instruction truncates <tt>value</tt> to type
4182 <p>The '<tt>fptrunc</tt>' instruction takes a <a href="#t_floating">floating
4183 point</a> value to cast and a <a href="#t_floating">floating point</a> type
4184 to cast it to. The size of <tt>value</tt> must be larger than the size of
4185 <tt>ty2</tt>. This implies that <tt>fptrunc</tt> cannot be used to make a
4186 <i>no-op cast</i>.</p>
4189 <p>The '<tt>fptrunc</tt>' instruction truncates a <tt>value</tt> from a larger
4190 <a href="#t_floating">floating point</a> type to a smaller
4191 <a href="#t_floating">floating point</a> type. If the value cannot fit
4192 within the destination type, <tt>ty2</tt>, then the results are
4197 %X = fptrunc double 123.0 to float <i>; yields float:123.0</i>
4198 %Y = fptrunc double 1.0E+300 to float <i>; yields undefined</i>
4203 <!-- _______________________________________________________________________ -->
4204 <div class="doc_subsubsection">
4205 <a name="i_fpext">'<tt>fpext .. to</tt>' Instruction</a>
4207 <div class="doc_text">
4211 <result> = fpext <ty> <value> to <ty2> <i>; yields ty2</i>
4215 <p>The '<tt>fpext</tt>' extends a floating point <tt>value</tt> to a larger
4216 floating point value.</p>
4219 <p>The '<tt>fpext</tt>' instruction takes a
4220 <a href="#t_floating">floating point</a> <tt>value</tt> to cast, and
4221 a <a href="#t_floating">floating point</a> type to cast it to. The source
4222 type must be smaller than the destination type.</p>
4225 <p>The '<tt>fpext</tt>' instruction extends the <tt>value</tt> from a smaller
4226 <a href="#t_floating">floating point</a> type to a larger
4227 <a href="#t_floating">floating point</a> type. The <tt>fpext</tt> cannot be
4228 used to make a <i>no-op cast</i> because it always changes bits. Use
4229 <tt>bitcast</tt> to make a <i>no-op cast</i> for a floating point cast.</p>
4233 %X = fpext float 3.1415 to double <i>; yields double:3.1415</i>
4234 %Y = fpext float 1.0 to float <i>; yields float:1.0 (no-op)</i>
4239 <!-- _______________________________________________________________________ -->
4240 <div class="doc_subsubsection">
4241 <a name="i_fptoui">'<tt>fptoui .. to</tt>' Instruction</a>
4243 <div class="doc_text">
4247 <result> = fptoui <ty> <value> to <ty2> <i>; yields ty2</i>
4251 <p>The '<tt>fptoui</tt>' converts a floating point <tt>value</tt> to its
4252 unsigned integer equivalent of type <tt>ty2</tt>.</p>
4255 <p>The '<tt>fptoui</tt>' instruction takes a value to cast, which must be a
4256 scalar or vector <a href="#t_floating">floating point</a> value, and a type
4257 to cast it to <tt>ty2</tt>, which must be an <a href="#t_integer">integer</a>
4258 type. If <tt>ty</tt> is a vector floating point type, <tt>ty2</tt> must be a
4259 vector integer type with the same number of elements as <tt>ty</tt></p>
4262 <p>The '<tt>fptoui</tt>' instruction converts its
4263 <a href="#t_floating">floating point</a> operand into the nearest (rounding
4264 towards zero) unsigned integer value. If the value cannot fit
4265 in <tt>ty2</tt>, the results are undefined.</p>
4269 %X = fptoui double 123.0 to i32 <i>; yields i32:123</i>
4270 %Y = fptoui float 1.0E+300 to i1 <i>; yields undefined:1</i>
4271 %X = fptoui float 1.04E+17 to i8 <i>; yields undefined:1</i>
4276 <!-- _______________________________________________________________________ -->
4277 <div class="doc_subsubsection">
4278 <a name="i_fptosi">'<tt>fptosi .. to</tt>' Instruction</a>
4280 <div class="doc_text">
4284 <result> = fptosi <ty> <value> to <ty2> <i>; yields ty2</i>
4288 <p>The '<tt>fptosi</tt>' instruction converts
4289 <a href="#t_floating">floating point</a> <tt>value</tt> to
4290 type <tt>ty2</tt>.</p>
4293 <p>The '<tt>fptosi</tt>' instruction takes a value to cast, which must be a
4294 scalar or vector <a href="#t_floating">floating point</a> value, and a type
4295 to cast it to <tt>ty2</tt>, which must be an <a href="#t_integer">integer</a>
4296 type. If <tt>ty</tt> is a vector floating point type, <tt>ty2</tt> must be a
4297 vector integer type with the same number of elements as <tt>ty</tt></p>
4300 <p>The '<tt>fptosi</tt>' instruction converts its
4301 <a href="#t_floating">floating point</a> operand into the nearest (rounding
4302 towards zero) signed integer value. If the value cannot fit in <tt>ty2</tt>,
4303 the results are undefined.</p>
4307 %X = fptosi double -123.0 to i32 <i>; yields i32:-123</i>
4308 %Y = fptosi float 1.0E-247 to i1 <i>; yields undefined:1</i>
4309 %X = fptosi float 1.04E+17 to i8 <i>; yields undefined:1</i>
4314 <!-- _______________________________________________________________________ -->
4315 <div class="doc_subsubsection">
4316 <a name="i_uitofp">'<tt>uitofp .. to</tt>' Instruction</a>
4318 <div class="doc_text">
4322 <result> = uitofp <ty> <value> to <ty2> <i>; yields ty2</i>
4326 <p>The '<tt>uitofp</tt>' instruction regards <tt>value</tt> as an unsigned
4327 integer and converts that value to the <tt>ty2</tt> type.</p>
4330 <p>The '<tt>uitofp</tt>' instruction takes a value to cast, which must be a
4331 scalar or vector <a href="#t_integer">integer</a> value, and a type to cast
4332 it to <tt>ty2</tt>, which must be an <a href="#t_floating">floating point</a>
4333 type. If <tt>ty</tt> is a vector integer type, <tt>ty2</tt> must be a vector
4334 floating point type with the same number of elements as <tt>ty</tt></p>
4337 <p>The '<tt>uitofp</tt>' instruction interprets its operand as an unsigned
4338 integer quantity and converts it to the corresponding floating point
4339 value. If the value cannot fit in the floating point value, the results are
4344 %X = uitofp i32 257 to float <i>; yields float:257.0</i>
4345 %Y = uitofp i8 -1 to double <i>; yields double:255.0</i>
4350 <!-- _______________________________________________________________________ -->
4351 <div class="doc_subsubsection">
4352 <a name="i_sitofp">'<tt>sitofp .. to</tt>' Instruction</a>
4354 <div class="doc_text">
4358 <result> = sitofp <ty> <value> to <ty2> <i>; yields ty2</i>
4362 <p>The '<tt>sitofp</tt>' instruction regards <tt>value</tt> as a signed integer
4363 and converts that value to the <tt>ty2</tt> type.</p>
4366 <p>The '<tt>sitofp</tt>' instruction takes a value to cast, which must be a
4367 scalar or vector <a href="#t_integer">integer</a> value, and a type to cast
4368 it to <tt>ty2</tt>, which must be an <a href="#t_floating">floating point</a>
4369 type. If <tt>ty</tt> is a vector integer type, <tt>ty2</tt> must be a vector
4370 floating point type with the same number of elements as <tt>ty</tt></p>
4373 <p>The '<tt>sitofp</tt>' instruction interprets its operand as a signed integer
4374 quantity and converts it to the corresponding floating point value. If the
4375 value cannot fit in the floating point value, the results are undefined.</p>
4379 %X = sitofp i32 257 to float <i>; yields float:257.0</i>
4380 %Y = sitofp i8 -1 to double <i>; yields double:-1.0</i>
4385 <!-- _______________________________________________________________________ -->
4386 <div class="doc_subsubsection">
4387 <a name="i_ptrtoint">'<tt>ptrtoint .. to</tt>' Instruction</a>
4389 <div class="doc_text">
4393 <result> = ptrtoint <ty> <value> to <ty2> <i>; yields ty2</i>
4397 <p>The '<tt>ptrtoint</tt>' instruction converts the pointer <tt>value</tt> to
4398 the integer type <tt>ty2</tt>.</p>
4401 <p>The '<tt>ptrtoint</tt>' instruction takes a <tt>value</tt> to cast, which
4402 must be a <a href="#t_pointer">pointer</a> value, and a type to cast it to
4403 <tt>ty2</tt>, which must be an <a href="#t_integer">integer</a> type.</p>
4406 <p>The '<tt>ptrtoint</tt>' instruction converts <tt>value</tt> to integer type
4407 <tt>ty2</tt> by interpreting the pointer value as an integer and either
4408 truncating or zero extending that value to the size of the integer type. If
4409 <tt>value</tt> is smaller than <tt>ty2</tt> then a zero extension is done. If
4410 <tt>value</tt> is larger than <tt>ty2</tt> then a truncation is done. If they
4411 are the same size, then nothing is done (<i>no-op cast</i>) other than a type
4416 %X = ptrtoint i32* %X to i8 <i>; yields truncation on 32-bit architecture</i>
4417 %Y = ptrtoint i32* %x to i64 <i>; yields zero extension on 32-bit architecture</i>
4422 <!-- _______________________________________________________________________ -->
4423 <div class="doc_subsubsection">
4424 <a name="i_inttoptr">'<tt>inttoptr .. to</tt>' Instruction</a>
4426 <div class="doc_text">
4430 <result> = inttoptr <ty> <value> to <ty2> <i>; yields ty2</i>
4434 <p>The '<tt>inttoptr</tt>' instruction converts an integer <tt>value</tt> to a
4435 pointer type, <tt>ty2</tt>.</p>
4438 <p>The '<tt>inttoptr</tt>' instruction takes an <a href="#t_integer">integer</a>
4439 value to cast, and a type to cast it to, which must be a
4440 <a href="#t_pointer">pointer</a> type.</p>
4443 <p>The '<tt>inttoptr</tt>' instruction converts <tt>value</tt> to type
4444 <tt>ty2</tt> by applying either a zero extension or a truncation depending on
4445 the size of the integer <tt>value</tt>. If <tt>value</tt> is larger than the
4446 size of a pointer then a truncation is done. If <tt>value</tt> is smaller
4447 than the size of a pointer then a zero extension is done. If they are the
4448 same size, nothing is done (<i>no-op cast</i>).</p>
4452 %X = inttoptr i32 255 to i32* <i>; yields zero extension on 64-bit architecture</i>
4453 %X = inttoptr i32 255 to i32* <i>; yields no-op on 32-bit architecture</i>
4454 %Y = inttoptr i64 0 to i32* <i>; yields truncation on 32-bit architecture</i>
4459 <!-- _______________________________________________________________________ -->
4460 <div class="doc_subsubsection">
4461 <a name="i_bitcast">'<tt>bitcast .. to</tt>' Instruction</a>
4463 <div class="doc_text">
4467 <result> = bitcast <ty> <value> to <ty2> <i>; yields ty2</i>
4471 <p>The '<tt>bitcast</tt>' instruction converts <tt>value</tt> to type
4472 <tt>ty2</tt> without changing any bits.</p>
4475 <p>The '<tt>bitcast</tt>' instruction takes a value to cast, which must be a
4476 non-aggregate first class value, and a type to cast it to, which must also be
4477 a non-aggregate <a href="#t_firstclass">first class</a> type. The bit sizes
4478 of <tt>value</tt> and the destination type, <tt>ty2</tt>, must be
4479 identical. If the source type is a pointer, the destination type must also be
4480 a pointer. This instruction supports bitwise conversion of vectors to
4481 integers and to vectors of other types (as long as they have the same
4485 <p>The '<tt>bitcast</tt>' instruction converts <tt>value</tt> to type
4486 <tt>ty2</tt>. It is always a <i>no-op cast</i> because no bits change with
4487 this conversion. The conversion is done as if the <tt>value</tt> had been
4488 stored to memory and read back as type <tt>ty2</tt>. Pointer types may only
4489 be converted to other pointer types with this instruction. To convert
4490 pointers to other types, use the <a href="#i_inttoptr">inttoptr</a> or
4491 <a href="#i_ptrtoint">ptrtoint</a> instructions first.</p>
4495 %X = bitcast i8 255 to i8 <i>; yields i8 :-1</i>
4496 %Y = bitcast i32* %x to sint* <i>; yields sint*:%x</i>
4497 %Z = bitcast <2 x int> %V to i64; <i>; yields i64: %V</i>
4502 <!-- ======================================================================= -->
4503 <div class="doc_subsection"> <a name="otherops">Other Operations</a> </div>
4505 <div class="doc_text">
4507 <p>The instructions in this category are the "miscellaneous" instructions, which
4508 defy better classification.</p>
4512 <!-- _______________________________________________________________________ -->
4513 <div class="doc_subsubsection"><a name="i_icmp">'<tt>icmp</tt>' Instruction</a>
4516 <div class="doc_text">
4520 <result> = icmp <cond> <ty> <op1>, <op2> <i>; yields {i1} or {<N x i1>}:result</i>
4524 <p>The '<tt>icmp</tt>' instruction returns a boolean value or a vector of
4525 boolean values based on comparison of its two integer, integer vector, or
4526 pointer operands.</p>
4529 <p>The '<tt>icmp</tt>' instruction takes three operands. The first operand is
4530 the condition code indicating the kind of comparison to perform. It is not a
4531 value, just a keyword. The possible condition code are:</p>
4534 <li><tt>eq</tt>: equal</li>
4535 <li><tt>ne</tt>: not equal </li>
4536 <li><tt>ugt</tt>: unsigned greater than</li>
4537 <li><tt>uge</tt>: unsigned greater or equal</li>
4538 <li><tt>ult</tt>: unsigned less than</li>
4539 <li><tt>ule</tt>: unsigned less or equal</li>
4540 <li><tt>sgt</tt>: signed greater than</li>
4541 <li><tt>sge</tt>: signed greater or equal</li>
4542 <li><tt>slt</tt>: signed less than</li>
4543 <li><tt>sle</tt>: signed less or equal</li>
4546 <p>The remaining two arguments must be <a href="#t_integer">integer</a> or
4547 <a href="#t_pointer">pointer</a> or integer <a href="#t_vector">vector</a>
4548 typed. They must also be identical types.</p>
4551 <p>The '<tt>icmp</tt>' compares <tt>op1</tt> and <tt>op2</tt> according to the
4552 condition code given as <tt>cond</tt>. The comparison performed always yields
4553 either an <a href="#t_primitive"><tt>i1</tt></a> or vector of <tt>i1</tt>
4554 result, as follows:</p>
4557 <li><tt>eq</tt>: yields <tt>true</tt> if the operands are equal,
4558 <tt>false</tt> otherwise. No sign interpretation is necessary or
4561 <li><tt>ne</tt>: yields <tt>true</tt> if the operands are unequal,
4562 <tt>false</tt> otherwise. No sign interpretation is necessary or
4565 <li><tt>ugt</tt>: interprets the operands as unsigned values and yields
4566 <tt>true</tt> if <tt>op1</tt> is greater than <tt>op2</tt>.</li>
4568 <li><tt>uge</tt>: interprets the operands as unsigned values and yields
4569 <tt>true</tt> if <tt>op1</tt> is greater than or equal
4570 to <tt>op2</tt>.</li>
4572 <li><tt>ult</tt>: interprets the operands as unsigned values and yields
4573 <tt>true</tt> if <tt>op1</tt> is less than <tt>op2</tt>.</li>
4575 <li><tt>ule</tt>: interprets the operands as unsigned values and yields
4576 <tt>true</tt> if <tt>op1</tt> is less than or equal to <tt>op2</tt>.</li>
4578 <li><tt>sgt</tt>: interprets the operands as signed values and yields
4579 <tt>true</tt> if <tt>op1</tt> is greater than <tt>op2</tt>.</li>
4581 <li><tt>sge</tt>: interprets the operands as signed values and yields
4582 <tt>true</tt> if <tt>op1</tt> is greater than or equal
4583 to <tt>op2</tt>.</li>
4585 <li><tt>slt</tt>: interprets the operands as signed values and yields
4586 <tt>true</tt> if <tt>op1</tt> is less than <tt>op2</tt>.</li>
4588 <li><tt>sle</tt>: interprets the operands as signed values and yields
4589 <tt>true</tt> if <tt>op1</tt> is less than or equal to <tt>op2</tt>.</li>
4592 <p>If the operands are <a href="#t_pointer">pointer</a> typed, the pointer
4593 values are compared as if they were integers.</p>
4595 <p>If the operands are integer vectors, then they are compared element by
4596 element. The result is an <tt>i1</tt> vector with the same number of elements
4597 as the values being compared. Otherwise, the result is an <tt>i1</tt>.</p>
4601 <result> = icmp eq i32 4, 5 <i>; yields: result=false</i>
4602 <result> = icmp ne float* %X, %X <i>; yields: result=false</i>
4603 <result> = icmp ult i16 4, 5 <i>; yields: result=true</i>
4604 <result> = icmp sgt i16 4, 5 <i>; yields: result=false</i>
4605 <result> = icmp ule i16 -4, 5 <i>; yields: result=false</i>
4606 <result> = icmp sge i16 4, 5 <i>; yields: result=false</i>
4609 <p>Note that the code generator does not yet support vector types with
4610 the <tt>icmp</tt> instruction.</p>
4614 <!-- _______________________________________________________________________ -->
4615 <div class="doc_subsubsection"><a name="i_fcmp">'<tt>fcmp</tt>' Instruction</a>
4618 <div class="doc_text">
4622 <result> = fcmp <cond> <ty> <op1>, <op2> <i>; yields {i1} or {<N x i1>}:result</i>
4626 <p>The '<tt>fcmp</tt>' instruction returns a boolean value or vector of boolean
4627 values based on comparison of its operands.</p>
4629 <p>If the operands are floating point scalars, then the result type is a boolean
4630 (<a href="#t_primitive"><tt>i1</tt></a>).</p>
4632 <p>If the operands are floating point vectors, then the result type is a vector
4633 of boolean with the same number of elements as the operands being
4637 <p>The '<tt>fcmp</tt>' instruction takes three operands. The first operand is
4638 the condition code indicating the kind of comparison to perform. It is not a
4639 value, just a keyword. The possible condition code are:</p>
4642 <li><tt>false</tt>: no comparison, always returns false</li>
4643 <li><tt>oeq</tt>: ordered and equal</li>
4644 <li><tt>ogt</tt>: ordered and greater than </li>
4645 <li><tt>oge</tt>: ordered and greater than or equal</li>
4646 <li><tt>olt</tt>: ordered and less than </li>
4647 <li><tt>ole</tt>: ordered and less than or equal</li>
4648 <li><tt>one</tt>: ordered and not equal</li>
4649 <li><tt>ord</tt>: ordered (no nans)</li>
4650 <li><tt>ueq</tt>: unordered or equal</li>
4651 <li><tt>ugt</tt>: unordered or greater than </li>
4652 <li><tt>uge</tt>: unordered or greater than or equal</li>
4653 <li><tt>ult</tt>: unordered or less than </li>
4654 <li><tt>ule</tt>: unordered or less than or equal</li>
4655 <li><tt>une</tt>: unordered or not equal</li>
4656 <li><tt>uno</tt>: unordered (either nans)</li>
4657 <li><tt>true</tt>: no comparison, always returns true</li>
4660 <p><i>Ordered</i> means that neither operand is a QNAN while
4661 <i>unordered</i> means that either operand may be a QNAN.</p>
4663 <p>Each of <tt>val1</tt> and <tt>val2</tt> arguments must be either
4664 a <a href="#t_floating">floating point</a> type or
4665 a <a href="#t_vector">vector</a> of floating point type. They must have
4666 identical types.</p>
4669 <p>The '<tt>fcmp</tt>' instruction compares <tt>op1</tt> and <tt>op2</tt>
4670 according to the condition code given as <tt>cond</tt>. If the operands are
4671 vectors, then the vectors are compared element by element. Each comparison
4672 performed always yields an <a href="#t_primitive">i1</a> result, as
4676 <li><tt>false</tt>: always yields <tt>false</tt>, regardless of operands.</li>
4678 <li><tt>oeq</tt>: yields <tt>true</tt> if both operands are not a QNAN and
4679 <tt>op1</tt> is equal to <tt>op2</tt>.</li>
4681 <li><tt>ogt</tt>: yields <tt>true</tt> if both operands are not a QNAN and
4682 <tt>op1</tt> is greather than <tt>op2</tt>.</li>
4684 <li><tt>oge</tt>: yields <tt>true</tt> if both operands are not a QNAN and
4685 <tt>op1</tt> is greater than or equal to <tt>op2</tt>.</li>
4687 <li><tt>olt</tt>: yields <tt>true</tt> if both operands are not a QNAN and
4688 <tt>op1</tt> is less than <tt>op2</tt>.</li>
4690 <li><tt>ole</tt>: yields <tt>true</tt> if both operands are not a QNAN and
4691 <tt>op1</tt> is less than or equal to <tt>op2</tt>.</li>
4693 <li><tt>one</tt>: yields <tt>true</tt> if both operands are not a QNAN and
4694 <tt>op1</tt> is not equal to <tt>op2</tt>.</li>
4696 <li><tt>ord</tt>: yields <tt>true</tt> if both operands are not a QNAN.</li>
4698 <li><tt>ueq</tt>: yields <tt>true</tt> if either operand is a QNAN or
4699 <tt>op1</tt> is equal to <tt>op2</tt>.</li>
4701 <li><tt>ugt</tt>: yields <tt>true</tt> if either operand is a QNAN or
4702 <tt>op1</tt> is greater than <tt>op2</tt>.</li>
4704 <li><tt>uge</tt>: yields <tt>true</tt> if either operand is a QNAN or
4705 <tt>op1</tt> is greater than or equal to <tt>op2</tt>.</li>
4707 <li><tt>ult</tt>: yields <tt>true</tt> if either operand is a QNAN or
4708 <tt>op1</tt> is less than <tt>op2</tt>.</li>
4710 <li><tt>ule</tt>: yields <tt>true</tt> if either operand is a QNAN or
4711 <tt>op1</tt> is less than or equal to <tt>op2</tt>.</li>
4713 <li><tt>une</tt>: yields <tt>true</tt> if either operand is a QNAN or
4714 <tt>op1</tt> is not equal to <tt>op2</tt>.</li>
4716 <li><tt>uno</tt>: yields <tt>true</tt> if either operand is a QNAN.</li>
4718 <li><tt>true</tt>: always yields <tt>true</tt>, regardless of operands.</li>
4723 <result> = fcmp oeq float 4.0, 5.0 <i>; yields: result=false</i>
4724 <result> = fcmp one float 4.0, 5.0 <i>; yields: result=true</i>
4725 <result> = fcmp olt float 4.0, 5.0 <i>; yields: result=true</i>
4726 <result> = fcmp ueq double 1.0, 2.0 <i>; yields: result=false</i>
4729 <p>Note that the code generator does not yet support vector types with
4730 the <tt>fcmp</tt> instruction.</p>
4734 <!-- _______________________________________________________________________ -->
4735 <div class="doc_subsubsection">
4736 <a name="i_phi">'<tt>phi</tt>' Instruction</a>
4739 <div class="doc_text">
4743 <result> = phi <ty> [ <val0>, <label0>], ...
4747 <p>The '<tt>phi</tt>' instruction is used to implement the φ node in the
4748 SSA graph representing the function.</p>
4751 <p>The type of the incoming values is specified with the first type field. After
4752 this, the '<tt>phi</tt>' instruction takes a list of pairs as arguments, with
4753 one pair for each predecessor basic block of the current block. Only values
4754 of <a href="#t_firstclass">first class</a> type may be used as the value
4755 arguments to the PHI node. Only labels may be used as the label
4758 <p>There must be no non-phi instructions between the start of a basic block and
4759 the PHI instructions: i.e. PHI instructions must be first in a basic
4762 <p>For the purposes of the SSA form, the use of each incoming value is deemed to
4763 occur on the edge from the corresponding predecessor block to the current
4764 block (but after any definition of an '<tt>invoke</tt>' instruction's return
4765 value on the same edge).</p>
4768 <p>At runtime, the '<tt>phi</tt>' instruction logically takes on the value
4769 specified by the pair corresponding to the predecessor basic block that
4770 executed just prior to the current block.</p>
4774 Loop: ; Infinite loop that counts from 0 on up...
4775 %indvar = phi i32 [ 0, %LoopHeader ], [ %nextindvar, %Loop ]
4776 %nextindvar = add i32 %indvar, 1
4782 <!-- _______________________________________________________________________ -->
4783 <div class="doc_subsubsection">
4784 <a name="i_select">'<tt>select</tt>' Instruction</a>
4787 <div class="doc_text">
4791 <result> = select <i>selty</i> <cond>, <ty> <val1>, <ty> <val2> <i>; yields ty</i>
4793 <i>selty</i> is either i1 or {<N x i1>}
4797 <p>The '<tt>select</tt>' instruction is used to choose one value based on a
4798 condition, without branching.</p>
4802 <p>The '<tt>select</tt>' instruction requires an 'i1' value or a vector of 'i1'
4803 values indicating the condition, and two values of the
4804 same <a href="#t_firstclass">first class</a> type. If the val1/val2 are
4805 vectors and the condition is a scalar, then entire vectors are selected, not
4806 individual elements.</p>
4809 <p>If the condition is an i1 and it evaluates to 1, the instruction returns the
4810 first value argument; otherwise, it returns the second value argument.</p>
4812 <p>If the condition is a vector of i1, then the value arguments must be vectors
4813 of the same size, and the selection is done element by element.</p>
4817 %X = select i1 true, i8 17, i8 42 <i>; yields i8:17</i>
4820 <p>Note that the code generator does not yet support conditions
4821 with vector type.</p>
4825 <!-- _______________________________________________________________________ -->
4826 <div class="doc_subsubsection">
4827 <a name="i_call">'<tt>call</tt>' Instruction</a>
4830 <div class="doc_text">
4834 <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>]
4838 <p>The '<tt>call</tt>' instruction represents a simple function call.</p>
4841 <p>This instruction requires several arguments:</p>
4844 <li>The optional "tail" marker indicates whether the callee function accesses
4845 any allocas or varargs in the caller. If the "tail" marker is present,
4846 the function call is eligible for tail call optimization. Note that calls
4847 may be marked "tail" even if they do not occur before
4848 a <a href="#i_ret"><tt>ret</tt></a> instruction.</li>
4850 <li>The optional "cconv" marker indicates which <a href="#callingconv">calling
4851 convention</a> the call should use. If none is specified, the call
4852 defaults to using C calling conventions.</li>
4854 <li>The optional <a href="#paramattrs">Parameter Attributes</a> list for
4855 return values. Only '<tt>zeroext</tt>', '<tt>signext</tt>', and
4856 '<tt>inreg</tt>' attributes are valid here.</li>
4858 <li>'<tt>ty</tt>': the type of the call instruction itself which is also the
4859 type of the return value. Functions that return no value are marked
4860 <tt><a href="#t_void">void</a></tt>.</li>
4862 <li>'<tt>fnty</tt>': shall be the signature of the pointer to function value
4863 being invoked. The argument types must match the types implied by this
4864 signature. This type can be omitted if the function is not varargs and if
4865 the function type does not return a pointer to a function.</li>
4867 <li>'<tt>fnptrval</tt>': An LLVM value containing a pointer to a function to
4868 be invoked. In most cases, this is a direct function invocation, but
4869 indirect <tt>call</tt>s are just as possible, calling an arbitrary pointer
4870 to function value.</li>
4872 <li>'<tt>function args</tt>': argument list whose types match the function
4873 signature argument types. All arguments must be of
4874 <a href="#t_firstclass">first class</a> type. If the function signature
4875 indicates the function accepts a variable number of arguments, the extra
4876 arguments can be specified.</li>
4878 <li>The optional <a href="#fnattrs">function attributes</a> list. Only
4879 '<tt>noreturn</tt>', '<tt>nounwind</tt>', '<tt>readonly</tt>' and
4880 '<tt>readnone</tt>' attributes are valid here.</li>
4884 <p>The '<tt>call</tt>' instruction is used to cause control flow to transfer to
4885 a specified function, with its incoming arguments bound to the specified
4886 values. Upon a '<tt><a href="#i_ret">ret</a></tt>' instruction in the called
4887 function, control flow continues with the instruction after the function
4888 call, and the return value of the function is bound to the result
4893 %retval = call i32 @test(i32 %argc)
4894 call i32 (i8 *, ...)* @printf(i8 * %msg, i32 12, i8 42) <i>; yields i32</i>
4895 %X = tail call i32 @foo() <i>; yields i32</i>
4896 %Y = tail call <a href="#callingconv">fastcc</a> i32 @foo() <i>; yields i32</i>
4897 call void %foo(i8 97 signext)
4899 %struct.A = type { i32, i8 }
4900 %r = call %struct.A @foo() <i>; yields { 32, i8 }</i>
4901 %gr = extractvalue %struct.A %r, 0 <i>; yields i32</i>
4902 %gr1 = extractvalue %struct.A %r, 1 <i>; yields i8</i>
4903 %Z = call void @foo() noreturn <i>; indicates that %foo never returns normally</i>
4904 %ZZ = call zeroext i32 @bar() <i>; Return value is %zero extended</i>
4909 <!-- _______________________________________________________________________ -->
4910 <div class="doc_subsubsection">
4911 <a name="i_va_arg">'<tt>va_arg</tt>' Instruction</a>
4914 <div class="doc_text">
4918 <resultval> = va_arg <va_list*> <arglist>, <argty>
4922 <p>The '<tt>va_arg</tt>' instruction is used to access arguments passed through
4923 the "variable argument" area of a function call. It is used to implement the
4924 <tt>va_arg</tt> macro in C.</p>
4927 <p>This instruction takes a <tt>va_list*</tt> value and the type of the
4928 argument. It returns a value of the specified argument type and increments
4929 the <tt>va_list</tt> to point to the next argument. The actual type
4930 of <tt>va_list</tt> is target specific.</p>
4933 <p>The '<tt>va_arg</tt>' instruction loads an argument of the specified type
4934 from the specified <tt>va_list</tt> and causes the <tt>va_list</tt> to point
4935 to the next argument. For more information, see the variable argument
4936 handling <a href="#int_varargs">Intrinsic Functions</a>.</p>
4938 <p>It is legal for this instruction to be called in a function which does not
4939 take a variable number of arguments, for example, the <tt>vfprintf</tt>
4942 <p><tt>va_arg</tt> is an LLVM instruction instead of
4943 an <a href="#intrinsics">intrinsic function</a> because it takes a type as an
4947 <p>See the <a href="#int_varargs">variable argument processing</a> section.</p>
4949 <p>Note that the code generator does not yet fully support va_arg on many
4950 targets. Also, it does not currently support va_arg with aggregate types on
4955 <!-- *********************************************************************** -->
4956 <div class="doc_section"> <a name="intrinsics">Intrinsic Functions</a> </div>
4957 <!-- *********************************************************************** -->
4959 <div class="doc_text">
4961 <p>LLVM supports the notion of an "intrinsic function". These functions have
4962 well known names and semantics and are required to follow certain
4963 restrictions. Overall, these intrinsics represent an extension mechanism for
4964 the LLVM language that does not require changing all of the transformations
4965 in LLVM when adding to the language (or the bitcode reader/writer, the
4966 parser, etc...).</p>
4968 <p>Intrinsic function names must all start with an "<tt>llvm.</tt>" prefix. This
4969 prefix is reserved in LLVM for intrinsic names; thus, function names may not
4970 begin with this prefix. Intrinsic functions must always be external
4971 functions: you cannot define the body of intrinsic functions. Intrinsic
4972 functions may only be used in call or invoke instructions: it is illegal to
4973 take the address of an intrinsic function. Additionally, because intrinsic
4974 functions are part of the LLVM language, it is required if any are added that
4975 they be documented here.</p>
4977 <p>Some intrinsic functions can be overloaded, i.e., the intrinsic represents a
4978 family of functions that perform the same operation but on different data
4979 types. Because LLVM can represent over 8 million different integer types,
4980 overloading is used commonly to allow an intrinsic function to operate on any
4981 integer type. One or more of the argument types or the result type can be
4982 overloaded to accept any integer type. Argument types may also be defined as
4983 exactly matching a previous argument's type or the result type. This allows
4984 an intrinsic function which accepts multiple arguments, but needs all of them
4985 to be of the same type, to only be overloaded with respect to a single
4986 argument or the result.</p>
4988 <p>Overloaded intrinsics will have the names of its overloaded argument types
4989 encoded into its function name, each preceded by a period. Only those types
4990 which are overloaded result in a name suffix. Arguments whose type is matched
4991 against another type do not. For example, the <tt>llvm.ctpop</tt> function
4992 can take an integer of any width and returns an integer of exactly the same
4993 integer width. This leads to a family of functions such as
4994 <tt>i8 @llvm.ctpop.i8(i8 %val)</tt> and <tt>i29 @llvm.ctpop.i29(i29
4995 %val)</tt>. Only one type, the return type, is overloaded, and only one type
4996 suffix is required. Because the argument's type is matched against the return
4997 type, it does not require its own name suffix.</p>
4999 <p>To learn how to add an intrinsic function, please see the
5000 <a href="ExtendingLLVM.html">Extending LLVM Guide</a>.</p>
5004 <!-- ======================================================================= -->
5005 <div class="doc_subsection">
5006 <a name="int_varargs">Variable Argument Handling Intrinsics</a>
5009 <div class="doc_text">
5011 <p>Variable argument support is defined in LLVM with
5012 the <a href="#i_va_arg"><tt>va_arg</tt></a> instruction and these three
5013 intrinsic functions. These functions are related to the similarly named
5014 macros defined in the <tt><stdarg.h></tt> header file.</p>
5016 <p>All of these functions operate on arguments that use a target-specific value
5017 type "<tt>va_list</tt>". The LLVM assembly language reference manual does
5018 not define what this type is, so all transformations should be prepared to
5019 handle these functions regardless of the type used.</p>
5021 <p>This example shows how the <a href="#i_va_arg"><tt>va_arg</tt></a>
5022 instruction and the variable argument handling intrinsic functions are
5025 <div class="doc_code">
5027 define i32 @test(i32 %X, ...) {
5028 ; Initialize variable argument processing
5030 %ap2 = bitcast i8** %ap to i8*
5031 call void @llvm.va_start(i8* %ap2)
5033 ; Read a single integer argument
5034 %tmp = va_arg i8** %ap, i32
5036 ; Demonstrate usage of llvm.va_copy and llvm.va_end
5038 %aq2 = bitcast i8** %aq to i8*
5039 call void @llvm.va_copy(i8* %aq2, i8* %ap2)
5040 call void @llvm.va_end(i8* %aq2)
5042 ; Stop processing of arguments.
5043 call void @llvm.va_end(i8* %ap2)
5047 declare void @llvm.va_start(i8*)
5048 declare void @llvm.va_copy(i8*, i8*)
5049 declare void @llvm.va_end(i8*)
5055 <!-- _______________________________________________________________________ -->
5056 <div class="doc_subsubsection">
5057 <a name="int_va_start">'<tt>llvm.va_start</tt>' Intrinsic</a>
5061 <div class="doc_text">
5065 declare void %llvm.va_start(i8* <arglist>)
5069 <p>The '<tt>llvm.va_start</tt>' intrinsic initializes <tt>*<arglist></tt>
5070 for subsequent use by <tt><a href="#i_va_arg">va_arg</a></tt>.</p>
5073 <p>The argument is a pointer to a <tt>va_list</tt> element to initialize.</p>
5076 <p>The '<tt>llvm.va_start</tt>' intrinsic works just like the <tt>va_start</tt>
5077 macro available in C. In a target-dependent way, it initializes
5078 the <tt>va_list</tt> element to which the argument points, so that the next
5079 call to <tt>va_arg</tt> will produce the first variable argument passed to
5080 the function. Unlike the C <tt>va_start</tt> macro, this intrinsic does not
5081 need to know the last argument of the function as the compiler can figure
5086 <!-- _______________________________________________________________________ -->
5087 <div class="doc_subsubsection">
5088 <a name="int_va_end">'<tt>llvm.va_end</tt>' Intrinsic</a>
5091 <div class="doc_text">
5095 declare void @llvm.va_end(i8* <arglist>)
5099 <p>The '<tt>llvm.va_end</tt>' intrinsic destroys <tt>*<arglist></tt>,
5100 which has been initialized previously
5101 with <tt><a href="#int_va_start">llvm.va_start</a></tt>
5102 or <tt><a href="#i_va_copy">llvm.va_copy</a></tt>.</p>
5105 <p>The argument is a pointer to a <tt>va_list</tt> to destroy.</p>
5108 <p>The '<tt>llvm.va_end</tt>' intrinsic works just like the <tt>va_end</tt>
5109 macro available in C. In a target-dependent way, it destroys
5110 the <tt>va_list</tt> element to which the argument points. Calls
5111 to <a href="#int_va_start"><tt>llvm.va_start</tt></a>
5112 and <a href="#int_va_copy"> <tt>llvm.va_copy</tt></a> must be matched exactly
5113 with calls to <tt>llvm.va_end</tt>.</p>
5117 <!-- _______________________________________________________________________ -->
5118 <div class="doc_subsubsection">
5119 <a name="int_va_copy">'<tt>llvm.va_copy</tt>' Intrinsic</a>
5122 <div class="doc_text">
5126 declare void @llvm.va_copy(i8* <destarglist>, i8* <srcarglist>)
5130 <p>The '<tt>llvm.va_copy</tt>' intrinsic copies the current argument position
5131 from the source argument list to the destination argument list.</p>
5134 <p>The first argument is a pointer to a <tt>va_list</tt> element to initialize.
5135 The second argument is a pointer to a <tt>va_list</tt> element to copy
5139 <p>The '<tt>llvm.va_copy</tt>' intrinsic works just like the <tt>va_copy</tt>
5140 macro available in C. In a target-dependent way, it copies the
5141 source <tt>va_list</tt> element into the destination <tt>va_list</tt>
5142 element. This intrinsic is necessary because
5143 the <tt><a href="#int_va_start"> llvm.va_start</a></tt> intrinsic may be
5144 arbitrarily complex and require, for example, memory allocation.</p>
5148 <!-- ======================================================================= -->
5149 <div class="doc_subsection">
5150 <a name="int_gc">Accurate Garbage Collection Intrinsics</a>
5153 <div class="doc_text">
5155 <p>LLVM support for <a href="GarbageCollection.html">Accurate Garbage
5156 Collection</a> (GC) requires the implementation and generation of these
5157 intrinsics. These intrinsics allow identification of <a href="#int_gcroot">GC
5158 roots on the stack</a>, as well as garbage collector implementations that
5159 require <a href="#int_gcread">read</a> and <a href="#int_gcwrite">write</a>
5160 barriers. Front-ends for type-safe garbage collected languages should generate
5161 these intrinsics to make use of the LLVM garbage collectors. For more details,
5162 see <a href="GarbageCollection.html">Accurate Garbage Collection with
5165 <p>The garbage collection intrinsics only operate on objects in the generic
5166 address space (address space zero).</p>
5170 <!-- _______________________________________________________________________ -->
5171 <div class="doc_subsubsection">
5172 <a name="int_gcroot">'<tt>llvm.gcroot</tt>' Intrinsic</a>
5175 <div class="doc_text">
5179 declare void @llvm.gcroot(i8** %ptrloc, i8* %metadata)
5183 <p>The '<tt>llvm.gcroot</tt>' intrinsic declares the existence of a GC root to
5184 the code generator, and allows some metadata to be associated with it.</p>
5187 <p>The first argument specifies the address of a stack object that contains the
5188 root pointer. The second pointer (which must be either a constant or a
5189 global value address) contains the meta-data to be associated with the
5193 <p>At runtime, a call to this intrinsic stores a null pointer into the "ptrloc"
5194 location. At compile-time, the code generator generates information to allow
5195 the runtime to find the pointer at GC safe points. The '<tt>llvm.gcroot</tt>'
5196 intrinsic may only be used in a function which <a href="#gc">specifies a GC
5201 <!-- _______________________________________________________________________ -->
5202 <div class="doc_subsubsection">
5203 <a name="int_gcread">'<tt>llvm.gcread</tt>' Intrinsic</a>
5206 <div class="doc_text">
5210 declare i8* @llvm.gcread(i8* %ObjPtr, i8** %Ptr)
5214 <p>The '<tt>llvm.gcread</tt>' intrinsic identifies reads of references from heap
5215 locations, allowing garbage collector implementations that require read
5219 <p>The second argument is the address to read from, which should be an address
5220 allocated from the garbage collector. The first object is a pointer to the
5221 start of the referenced object, if needed by the language runtime (otherwise
5225 <p>The '<tt>llvm.gcread</tt>' intrinsic has the same semantics as a load
5226 instruction, but may be replaced with substantially more complex code by the
5227 garbage collector runtime, as needed. The '<tt>llvm.gcread</tt>' intrinsic
5228 may only be used in a function which <a href="#gc">specifies a GC
5233 <!-- _______________________________________________________________________ -->
5234 <div class="doc_subsubsection">
5235 <a name="int_gcwrite">'<tt>llvm.gcwrite</tt>' Intrinsic</a>
5238 <div class="doc_text">
5242 declare void @llvm.gcwrite(i8* %P1, i8* %Obj, i8** %P2)
5246 <p>The '<tt>llvm.gcwrite</tt>' intrinsic identifies writes of references to heap
5247 locations, allowing garbage collector implementations that require write
5248 barriers (such as generational or reference counting collectors).</p>
5251 <p>The first argument is the reference to store, the second is the start of the
5252 object to store it to, and the third is the address of the field of Obj to
5253 store to. If the runtime does not require a pointer to the object, Obj may
5257 <p>The '<tt>llvm.gcwrite</tt>' intrinsic has the same semantics as a store
5258 instruction, but may be replaced with substantially more complex code by the
5259 garbage collector runtime, as needed. The '<tt>llvm.gcwrite</tt>' intrinsic
5260 may only be used in a function which <a href="#gc">specifies a GC
5265 <!-- ======================================================================= -->
5266 <div class="doc_subsection">
5267 <a name="int_codegen">Code Generator Intrinsics</a>
5270 <div class="doc_text">
5272 <p>These intrinsics are provided by LLVM to expose special features that may
5273 only be implemented with code generator support.</p>
5277 <!-- _______________________________________________________________________ -->
5278 <div class="doc_subsubsection">
5279 <a name="int_returnaddress">'<tt>llvm.returnaddress</tt>' Intrinsic</a>
5282 <div class="doc_text">
5286 declare i8 *@llvm.returnaddress(i32 <level>)
5290 <p>The '<tt>llvm.returnaddress</tt>' intrinsic attempts to compute a
5291 target-specific value indicating the return address of the current function
5292 or one of its callers.</p>
5295 <p>The argument to this intrinsic indicates which function to return the address
5296 for. Zero indicates the calling function, one indicates its caller, etc.
5297 The argument is <b>required</b> to be a constant integer value.</p>
5300 <p>The '<tt>llvm.returnaddress</tt>' intrinsic either returns a pointer
5301 indicating the return address of the specified call frame, or zero if it
5302 cannot be identified. The value returned by this intrinsic is likely to be
5303 incorrect or 0 for arguments other than zero, so it should only be used for
5304 debugging purposes.</p>
5306 <p>Note that calling this intrinsic does not prevent function inlining or other
5307 aggressive transformations, so the value returned may not be that of the
5308 obvious source-language caller.</p>
5312 <!-- _______________________________________________________________________ -->
5313 <div class="doc_subsubsection">
5314 <a name="int_frameaddress">'<tt>llvm.frameaddress</tt>' Intrinsic</a>
5317 <div class="doc_text">
5321 declare i8 *@llvm.frameaddress(i32 <level>)
5325 <p>The '<tt>llvm.frameaddress</tt>' intrinsic attempts to return the
5326 target-specific frame pointer value for the specified stack frame.</p>
5329 <p>The argument to this intrinsic indicates which function to return the frame
5330 pointer for. Zero indicates the calling function, one indicates its caller,
5331 etc. The argument is <b>required</b> to be a constant integer value.</p>
5334 <p>The '<tt>llvm.frameaddress</tt>' intrinsic either returns a pointer
5335 indicating the frame address of the specified call frame, or zero if it
5336 cannot be identified. The value returned by this intrinsic is likely to be
5337 incorrect or 0 for arguments other than zero, so it should only be used for
5338 debugging purposes.</p>
5340 <p>Note that calling this intrinsic does not prevent function inlining or other
5341 aggressive transformations, so the value returned may not be that of the
5342 obvious source-language caller.</p>
5346 <!-- _______________________________________________________________________ -->
5347 <div class="doc_subsubsection">
5348 <a name="int_stacksave">'<tt>llvm.stacksave</tt>' Intrinsic</a>
5351 <div class="doc_text">
5355 declare i8 *@llvm.stacksave()
5359 <p>The '<tt>llvm.stacksave</tt>' intrinsic is used to remember the current state
5360 of the function stack, for use
5361 with <a href="#int_stackrestore"> <tt>llvm.stackrestore</tt></a>. This is
5362 useful for implementing language features like scoped automatic variable
5363 sized arrays in C99.</p>
5366 <p>This intrinsic returns a opaque pointer value that can be passed
5367 to <a href="#int_stackrestore"><tt>llvm.stackrestore</tt></a>. When
5368 an <tt>llvm.stackrestore</tt> intrinsic is executed with a value saved
5369 from <tt>llvm.stacksave</tt>, it effectively restores the state of the stack
5370 to the state it was in when the <tt>llvm.stacksave</tt> intrinsic executed.
5371 In practice, this pops any <a href="#i_alloca">alloca</a> blocks from the
5372 stack that were allocated after the <tt>llvm.stacksave</tt> was executed.</p>
5376 <!-- _______________________________________________________________________ -->
5377 <div class="doc_subsubsection">
5378 <a name="int_stackrestore">'<tt>llvm.stackrestore</tt>' Intrinsic</a>
5381 <div class="doc_text">
5385 declare void @llvm.stackrestore(i8 * %ptr)
5389 <p>The '<tt>llvm.stackrestore</tt>' intrinsic is used to restore the state of
5390 the function stack to the state it was in when the
5391 corresponding <a href="#int_stacksave"><tt>llvm.stacksave</tt></a> intrinsic
5392 executed. This is useful for implementing language features like scoped
5393 automatic variable sized arrays in C99.</p>
5396 <p>See the description
5397 for <a href="#int_stacksave"><tt>llvm.stacksave</tt></a>.</p>
5401 <!-- _______________________________________________________________________ -->
5402 <div class="doc_subsubsection">
5403 <a name="int_prefetch">'<tt>llvm.prefetch</tt>' Intrinsic</a>
5406 <div class="doc_text">
5410 declare void @llvm.prefetch(i8* <address>, i32 <rw>, i32 <locality>)
5414 <p>The '<tt>llvm.prefetch</tt>' intrinsic is a hint to the code generator to
5415 insert a prefetch instruction if supported; otherwise, it is a noop.
5416 Prefetches have no effect on the behavior of the program but can change its
5417 performance characteristics.</p>
5420 <p><tt>address</tt> is the address to be prefetched, <tt>rw</tt> is the
5421 specifier determining if the fetch should be for a read (0) or write (1),
5422 and <tt>locality</tt> is a temporal locality specifier ranging from (0) - no
5423 locality, to (3) - extremely local keep in cache. The <tt>rw</tt>
5424 and <tt>locality</tt> arguments must be constant integers.</p>
5427 <p>This intrinsic does not modify the behavior of the program. In particular,
5428 prefetches cannot trap and do not produce a value. On targets that support
5429 this intrinsic, the prefetch can provide hints to the processor cache for
5430 better performance.</p>
5434 <!-- _______________________________________________________________________ -->
5435 <div class="doc_subsubsection">
5436 <a name="int_pcmarker">'<tt>llvm.pcmarker</tt>' Intrinsic</a>
5439 <div class="doc_text">
5443 declare void @llvm.pcmarker(i32 <id>)
5447 <p>The '<tt>llvm.pcmarker</tt>' intrinsic is a method to export a Program
5448 Counter (PC) in a region of code to simulators and other tools. The method
5449 is target specific, but it is expected that the marker will use exported
5450 symbols to transmit the PC of the marker. The marker makes no guarantees
5451 that it will remain with any specific instruction after optimizations. It is
5452 possible that the presence of a marker will inhibit optimizations. The
5453 intended use is to be inserted after optimizations to allow correlations of
5454 simulation runs.</p>
5457 <p><tt>id</tt> is a numerical id identifying the marker.</p>
5460 <p>This intrinsic does not modify the behavior of the program. Backends that do
5461 not support this intrinisic may ignore it.</p>
5465 <!-- _______________________________________________________________________ -->
5466 <div class="doc_subsubsection">
5467 <a name="int_readcyclecounter">'<tt>llvm.readcyclecounter</tt>' Intrinsic</a>
5470 <div class="doc_text">
5474 declare i64 @llvm.readcyclecounter( )
5478 <p>The '<tt>llvm.readcyclecounter</tt>' intrinsic provides access to the cycle
5479 counter register (or similar low latency, high accuracy clocks) on those
5480 targets that support it. On X86, it should map to RDTSC. On Alpha, it
5481 should map to RPCC. As the backing counters overflow quickly (on the order
5482 of 9 seconds on alpha), this should only be used for small timings.</p>
5485 <p>When directly supported, reading the cycle counter should not modify any
5486 memory. Implementations are allowed to either return a application specific
5487 value or a system wide value. On backends without support, this is lowered
5488 to a constant 0.</p>
5492 <!-- ======================================================================= -->
5493 <div class="doc_subsection">
5494 <a name="int_libc">Standard C Library Intrinsics</a>
5497 <div class="doc_text">
5499 <p>LLVM provides intrinsics for a few important standard C library functions.
5500 These intrinsics allow source-language front-ends to pass information about
5501 the alignment of the pointer arguments to the code generator, providing
5502 opportunity for more efficient code generation.</p>
5506 <!-- _______________________________________________________________________ -->
5507 <div class="doc_subsubsection">
5508 <a name="int_memcpy">'<tt>llvm.memcpy</tt>' Intrinsic</a>
5511 <div class="doc_text">
5514 <p>This is an overloaded intrinsic. You can use <tt>llvm.memcpy</tt> on any
5515 integer bit width. Not all targets support all bit widths however.</p>
5518 declare void @llvm.memcpy.i8(i8 * <dest>, i8 * <src>,
5519 i8 <len>, i32 <align>)
5520 declare void @llvm.memcpy.i16(i8 * <dest>, i8 * <src>,
5521 i16 <len>, i32 <align>)
5522 declare void @llvm.memcpy.i32(i8 * <dest>, i8 * <src>,
5523 i32 <len>, i32 <align>)
5524 declare void @llvm.memcpy.i64(i8 * <dest>, i8 * <src>,
5525 i64 <len>, i32 <align>)
5529 <p>The '<tt>llvm.memcpy.*</tt>' intrinsics copy a block of memory from the
5530 source location to the destination location.</p>
5532 <p>Note that, unlike the standard libc function, the <tt>llvm.memcpy.*</tt>
5533 intrinsics do not return a value, and takes an extra alignment argument.</p>
5536 <p>The first argument is a pointer to the destination, the second is a pointer
5537 to the source. The third argument is an integer argument specifying the
5538 number of bytes to copy, and the fourth argument is the alignment of the
5539 source and destination locations.</p>
5541 <p>If the call to this intrinisic has an alignment value that is not 0 or 1,
5542 then the caller guarantees that both the source and destination pointers are
5543 aligned to that boundary.</p>
5546 <p>The '<tt>llvm.memcpy.*</tt>' intrinsics copy a block of memory from the
5547 source location to the destination location, which are not allowed to
5548 overlap. It copies "len" bytes of memory over. If the argument is known to
5549 be aligned to some boundary, this can be specified as the fourth argument,
5550 otherwise it should be set to 0 or 1.</p>
5554 <!-- _______________________________________________________________________ -->
5555 <div class="doc_subsubsection">
5556 <a name="int_memmove">'<tt>llvm.memmove</tt>' Intrinsic</a>
5559 <div class="doc_text">
5562 <p>This is an overloaded intrinsic. You can use llvm.memmove on any integer bit
5563 width. Not all targets support all bit widths however.</p>
5566 declare void @llvm.memmove.i8(i8 * <dest>, i8 * <src>,
5567 i8 <len>, i32 <align>)
5568 declare void @llvm.memmove.i16(i8 * <dest>, i8 * <src>,
5569 i16 <len>, i32 <align>)
5570 declare void @llvm.memmove.i32(i8 * <dest>, i8 * <src>,
5571 i32 <len>, i32 <align>)
5572 declare void @llvm.memmove.i64(i8 * <dest>, i8 * <src>,
5573 i64 <len>, i32 <align>)
5577 <p>The '<tt>llvm.memmove.*</tt>' intrinsics move a block of memory from the
5578 source location to the destination location. It is similar to the
5579 '<tt>llvm.memcpy</tt>' intrinsic but allows the two memory locations to
5582 <p>Note that, unlike the standard libc function, the <tt>llvm.memmove.*</tt>
5583 intrinsics do not return a value, and takes an extra alignment argument.</p>
5586 <p>The first argument is a pointer to the destination, the second is a pointer
5587 to the source. The third argument is an integer argument specifying the
5588 number of bytes to copy, and the fourth argument is the alignment of the
5589 source and destination locations.</p>
5591 <p>If the call to this intrinisic has an alignment value that is not 0 or 1,
5592 then the caller guarantees that the source and destination pointers are
5593 aligned to that boundary.</p>
5596 <p>The '<tt>llvm.memmove.*</tt>' intrinsics copy a block of memory from the
5597 source location to the destination location, which may overlap. It copies
5598 "len" bytes of memory over. If the argument is known to be aligned to some
5599 boundary, this can be specified as the fourth argument, otherwise it should
5600 be set to 0 or 1.</p>
5604 <!-- _______________________________________________________________________ -->
5605 <div class="doc_subsubsection">
5606 <a name="int_memset">'<tt>llvm.memset.*</tt>' Intrinsics</a>
5609 <div class="doc_text">
5612 <p>This is an overloaded intrinsic. You can use llvm.memset on any integer bit
5613 width. Not all targets support all bit widths however.</p>
5616 declare void @llvm.memset.i8(i8 * <dest>, i8 <val>,
5617 i8 <len>, i32 <align>)
5618 declare void @llvm.memset.i16(i8 * <dest>, i8 <val>,
5619 i16 <len>, i32 <align>)
5620 declare void @llvm.memset.i32(i8 * <dest>, i8 <val>,
5621 i32 <len>, i32 <align>)
5622 declare void @llvm.memset.i64(i8 * <dest>, i8 <val>,
5623 i64 <len>, i32 <align>)
5627 <p>The '<tt>llvm.memset.*</tt>' intrinsics fill a block of memory with a
5628 particular byte value.</p>
5630 <p>Note that, unlike the standard libc function, the <tt>llvm.memset</tt>
5631 intrinsic does not return a value, and takes an extra alignment argument.</p>
5634 <p>The first argument is a pointer to the destination to fill, the second is the
5635 byte value to fill it with, the third argument is an integer argument
5636 specifying the number of bytes to fill, and the fourth argument is the known
5637 alignment of destination location.</p>
5639 <p>If the call to this intrinisic has an alignment value that is not 0 or 1,
5640 then the caller guarantees that the destination pointer is aligned to that
5644 <p>The '<tt>llvm.memset.*</tt>' intrinsics fill "len" bytes of memory starting
5645 at the destination location. If the argument is known to be aligned to some
5646 boundary, this can be specified as the fourth argument, otherwise it should
5647 be set to 0 or 1.</p>
5651 <!-- _______________________________________________________________________ -->
5652 <div class="doc_subsubsection">
5653 <a name="int_sqrt">'<tt>llvm.sqrt.*</tt>' Intrinsic</a>
5656 <div class="doc_text">
5659 <p>This is an overloaded intrinsic. You can use <tt>llvm.sqrt</tt> on any
5660 floating point or vector of floating point type. Not all targets support all
5664 declare float @llvm.sqrt.f32(float %Val)
5665 declare double @llvm.sqrt.f64(double %Val)
5666 declare x86_fp80 @llvm.sqrt.f80(x86_fp80 %Val)
5667 declare fp128 @llvm.sqrt.f128(fp128 %Val)
5668 declare ppc_fp128 @llvm.sqrt.ppcf128(ppc_fp128 %Val)
5672 <p>The '<tt>llvm.sqrt</tt>' intrinsics return the sqrt of the specified operand,
5673 returning the same value as the libm '<tt>sqrt</tt>' functions would.
5674 Unlike <tt>sqrt</tt> in libm, however, <tt>llvm.sqrt</tt> has undefined
5675 behavior for negative numbers other than -0.0 (which allows for better
5676 optimization, because there is no need to worry about errno being
5677 set). <tt>llvm.sqrt(-0.0)</tt> is defined to return -0.0 like IEEE sqrt.</p>
5680 <p>The argument and return value are floating point numbers of the same
5684 <p>This function returns the sqrt of the specified operand if it is a
5685 nonnegative floating point number.</p>
5689 <!-- _______________________________________________________________________ -->
5690 <div class="doc_subsubsection">
5691 <a name="int_powi">'<tt>llvm.powi.*</tt>' Intrinsic</a>
5694 <div class="doc_text">
5697 <p>This is an overloaded intrinsic. You can use <tt>llvm.powi</tt> on any
5698 floating point or vector of floating point type. Not all targets support all
5702 declare float @llvm.powi.f32(float %Val, i32 %power)
5703 declare double @llvm.powi.f64(double %Val, i32 %power)
5704 declare x86_fp80 @llvm.powi.f80(x86_fp80 %Val, i32 %power)
5705 declare fp128 @llvm.powi.f128(fp128 %Val, i32 %power)
5706 declare ppc_fp128 @llvm.powi.ppcf128(ppc_fp128 %Val, i32 %power)
5710 <p>The '<tt>llvm.powi.*</tt>' intrinsics return the first operand raised to the
5711 specified (positive or negative) power. The order of evaluation of
5712 multiplications is not defined. When a vector of floating point type is
5713 used, the second argument remains a scalar integer value.</p>
5716 <p>The second argument is an integer power, and the first is a value to raise to
5720 <p>This function returns the first value raised to the second power with an
5721 unspecified sequence of rounding operations.</p>
5725 <!-- _______________________________________________________________________ -->
5726 <div class="doc_subsubsection">
5727 <a name="int_sin">'<tt>llvm.sin.*</tt>' Intrinsic</a>
5730 <div class="doc_text">
5733 <p>This is an overloaded intrinsic. You can use <tt>llvm.sin</tt> on any
5734 floating point or vector of floating point type. Not all targets support all
5738 declare float @llvm.sin.f32(float %Val)
5739 declare double @llvm.sin.f64(double %Val)
5740 declare x86_fp80 @llvm.sin.f80(x86_fp80 %Val)
5741 declare fp128 @llvm.sin.f128(fp128 %Val)
5742 declare ppc_fp128 @llvm.sin.ppcf128(ppc_fp128 %Val)
5746 <p>The '<tt>llvm.sin.*</tt>' intrinsics return the sine of the operand.</p>
5749 <p>The argument and return value are floating point numbers of the same
5753 <p>This function returns the sine of the specified operand, returning the same
5754 values as the libm <tt>sin</tt> functions would, and handles error conditions
5755 in the same way.</p>
5759 <!-- _______________________________________________________________________ -->
5760 <div class="doc_subsubsection">
5761 <a name="int_cos">'<tt>llvm.cos.*</tt>' Intrinsic</a>
5764 <div class="doc_text">
5767 <p>This is an overloaded intrinsic. You can use <tt>llvm.cos</tt> on any
5768 floating point or vector of floating point type. Not all targets support all
5772 declare float @llvm.cos.f32(float %Val)
5773 declare double @llvm.cos.f64(double %Val)
5774 declare x86_fp80 @llvm.cos.f80(x86_fp80 %Val)
5775 declare fp128 @llvm.cos.f128(fp128 %Val)
5776 declare ppc_fp128 @llvm.cos.ppcf128(ppc_fp128 %Val)
5780 <p>The '<tt>llvm.cos.*</tt>' intrinsics return the cosine of the operand.</p>
5783 <p>The argument and return value are floating point numbers of the same
5787 <p>This function returns the cosine of the specified operand, returning the same
5788 values as the libm <tt>cos</tt> functions would, and handles error conditions
5789 in the same way.</p>
5793 <!-- _______________________________________________________________________ -->
5794 <div class="doc_subsubsection">
5795 <a name="int_pow">'<tt>llvm.pow.*</tt>' Intrinsic</a>
5798 <div class="doc_text">
5801 <p>This is an overloaded intrinsic. You can use <tt>llvm.pow</tt> on any
5802 floating point or vector of floating point type. Not all targets support all
5806 declare float @llvm.pow.f32(float %Val, float %Power)
5807 declare double @llvm.pow.f64(double %Val, double %Power)
5808 declare x86_fp80 @llvm.pow.f80(x86_fp80 %Val, x86_fp80 %Power)
5809 declare fp128 @llvm.pow.f128(fp128 %Val, fp128 %Power)
5810 declare ppc_fp128 @llvm.pow.ppcf128(ppc_fp128 %Val, ppc_fp128 Power)
5814 <p>The '<tt>llvm.pow.*</tt>' intrinsics return the first operand raised to the
5815 specified (positive or negative) power.</p>
5818 <p>The second argument is a floating point power, and the first is a value to
5819 raise to that power.</p>
5822 <p>This function returns the first value raised to the second power, returning
5823 the same values as the libm <tt>pow</tt> functions would, and handles error
5824 conditions in the same way.</p>
5828 <!-- ======================================================================= -->
5829 <div class="doc_subsection">
5830 <a name="int_manip">Bit Manipulation Intrinsics</a>
5833 <div class="doc_text">
5835 <p>LLVM provides intrinsics for a few important bit manipulation operations.
5836 These allow efficient code generation for some algorithms.</p>
5840 <!-- _______________________________________________________________________ -->
5841 <div class="doc_subsubsection">
5842 <a name="int_bswap">'<tt>llvm.bswap.*</tt>' Intrinsics</a>
5845 <div class="doc_text">
5848 <p>This is an overloaded intrinsic function. You can use bswap on any integer
5849 type that is an even number of bytes (i.e. BitWidth % 16 == 0).</p>
5852 declare i16 @llvm.bswap.i16(i16 <id>)
5853 declare i32 @llvm.bswap.i32(i32 <id>)
5854 declare i64 @llvm.bswap.i64(i64 <id>)
5858 <p>The '<tt>llvm.bswap</tt>' family of intrinsics is used to byte swap integer
5859 values with an even number of bytes (positive multiple of 16 bits). These
5860 are useful for performing operations on data that is not in the target's
5861 native byte order.</p>
5864 <p>The <tt>llvm.bswap.i16</tt> intrinsic returns an i16 value that has the high
5865 and low byte of the input i16 swapped. Similarly,
5866 the <tt>llvm.bswap.i32</tt> intrinsic returns an i32 value that has the four
5867 bytes of the input i32 swapped, so that if the input bytes are numbered 0, 1,
5868 2, 3 then the returned i32 will have its bytes in 3, 2, 1, 0 order.
5869 The <tt>llvm.bswap.i48</tt>, <tt>llvm.bswap.i64</tt> and other intrinsics
5870 extend this concept to additional even-byte lengths (6 bytes, 8 bytes and
5871 more, respectively).</p>
5875 <!-- _______________________________________________________________________ -->
5876 <div class="doc_subsubsection">
5877 <a name="int_ctpop">'<tt>llvm.ctpop.*</tt>' Intrinsic</a>
5880 <div class="doc_text">
5883 <p>This is an overloaded intrinsic. You can use llvm.ctpop on any integer bit
5884 width. Not all targets support all bit widths however.</p>
5887 declare i8 @llvm.ctpop.i8(i8 <src>)
5888 declare i16 @llvm.ctpop.i16(i16 <src>)
5889 declare i32 @llvm.ctpop.i32(i32 <src>)
5890 declare i64 @llvm.ctpop.i64(i64 <src>)
5891 declare i256 @llvm.ctpop.i256(i256 <src>)
5895 <p>The '<tt>llvm.ctpop</tt>' family of intrinsics counts the number of bits set
5899 <p>The only argument is the value to be counted. The argument may be of any
5900 integer type. The return type must match the argument type.</p>
5903 <p>The '<tt>llvm.ctpop</tt>' intrinsic counts the 1's in a variable.</p>
5907 <!-- _______________________________________________________________________ -->
5908 <div class="doc_subsubsection">
5909 <a name="int_ctlz">'<tt>llvm.ctlz.*</tt>' Intrinsic</a>
5912 <div class="doc_text">
5915 <p>This is an overloaded intrinsic. You can use <tt>llvm.ctlz</tt> on any
5916 integer bit width. Not all targets support all bit widths however.</p>
5919 declare i8 @llvm.ctlz.i8 (i8 <src>)
5920 declare i16 @llvm.ctlz.i16(i16 <src>)
5921 declare i32 @llvm.ctlz.i32(i32 <src>)
5922 declare i64 @llvm.ctlz.i64(i64 <src>)
5923 declare i256 @llvm.ctlz.i256(i256 <src>)
5927 <p>The '<tt>llvm.ctlz</tt>' family of intrinsic functions counts the number of
5928 leading zeros in a variable.</p>
5931 <p>The only argument is the value to be counted. The argument may be of any
5932 integer type. The return type must match the argument type.</p>
5935 <p>The '<tt>llvm.ctlz</tt>' intrinsic counts the leading (most significant)
5936 zeros in a variable. If the src == 0 then the result is the size in bits of
5937 the type of src. For example, <tt>llvm.ctlz(i32 2) = 30</tt>.</p>
5941 <!-- _______________________________________________________________________ -->
5942 <div class="doc_subsubsection">
5943 <a name="int_cttz">'<tt>llvm.cttz.*</tt>' Intrinsic</a>
5946 <div class="doc_text">
5949 <p>This is an overloaded intrinsic. You can use <tt>llvm.cttz</tt> on any
5950 integer bit width. Not all targets support all bit widths however.</p>
5953 declare i8 @llvm.cttz.i8 (i8 <src>)
5954 declare i16 @llvm.cttz.i16(i16 <src>)
5955 declare i32 @llvm.cttz.i32(i32 <src>)
5956 declare i64 @llvm.cttz.i64(i64 <src>)
5957 declare i256 @llvm.cttz.i256(i256 <src>)
5961 <p>The '<tt>llvm.cttz</tt>' family of intrinsic functions counts the number of
5965 <p>The only argument is the value to be counted. The argument may be of any
5966 integer type. The return type must match the argument type.</p>
5969 <p>The '<tt>llvm.cttz</tt>' intrinsic counts the trailing (least significant)
5970 zeros in a variable. If the src == 0 then the result is the size in bits of
5971 the type of src. For example, <tt>llvm.cttz(2) = 1</tt>.</p>
5975 <!-- ======================================================================= -->
5976 <div class="doc_subsection">
5977 <a name="int_overflow">Arithmetic with Overflow Intrinsics</a>
5980 <div class="doc_text">
5982 <p>LLVM provides intrinsics for some arithmetic with overflow operations.</p>
5986 <!-- _______________________________________________________________________ -->
5987 <div class="doc_subsubsection">
5988 <a name="int_sadd_overflow">'<tt>llvm.sadd.with.overflow.*</tt>' Intrinsics</a>
5991 <div class="doc_text">
5994 <p>This is an overloaded intrinsic. You can use <tt>llvm.sadd.with.overflow</tt>
5995 on any integer bit width.</p>
5998 declare {i16, i1} @llvm.sadd.with.overflow.i16(i16 %a, i16 %b)
5999 declare {i32, i1} @llvm.sadd.with.overflow.i32(i32 %a, i32 %b)
6000 declare {i64, i1} @llvm.sadd.with.overflow.i64(i64 %a, i64 %b)
6004 <p>The '<tt>llvm.sadd.with.overflow</tt>' family of intrinsic functions perform
6005 a signed addition of the two arguments, and indicate whether an overflow
6006 occurred during the signed summation.</p>
6009 <p>The arguments (%a and %b) and the first element of the result structure may
6010 be of integer types of any bit width, but they must have the same bit
6011 width. The second element of the result structure must be of
6012 type <tt>i1</tt>. <tt>%a</tt> and <tt>%b</tt> are the two values that will
6013 undergo signed addition.</p>
6016 <p>The '<tt>llvm.sadd.with.overflow</tt>' family of intrinsic functions perform
6017 a signed addition of the two variables. They return a structure — the
6018 first element of which is the signed summation, and the second element of
6019 which is a bit specifying if the signed summation resulted in an
6024 %res = call {i32, i1} @llvm.sadd.with.overflow.i32(i32 %a, i32 %b)
6025 %sum = extractvalue {i32, i1} %res, 0
6026 %obit = extractvalue {i32, i1} %res, 1
6027 br i1 %obit, label %overflow, label %normal
6032 <!-- _______________________________________________________________________ -->
6033 <div class="doc_subsubsection">
6034 <a name="int_uadd_overflow">'<tt>llvm.uadd.with.overflow.*</tt>' Intrinsics</a>
6037 <div class="doc_text">
6040 <p>This is an overloaded intrinsic. You can use <tt>llvm.uadd.with.overflow</tt>
6041 on any integer bit width.</p>
6044 declare {i16, i1} @llvm.uadd.with.overflow.i16(i16 %a, i16 %b)
6045 declare {i32, i1} @llvm.uadd.with.overflow.i32(i32 %a, i32 %b)
6046 declare {i64, i1} @llvm.uadd.with.overflow.i64(i64 %a, i64 %b)
6050 <p>The '<tt>llvm.uadd.with.overflow</tt>' family of intrinsic functions perform
6051 an unsigned addition of the two arguments, and indicate whether a carry
6052 occurred during the unsigned summation.</p>
6055 <p>The arguments (%a and %b) and the first element of the result structure may
6056 be of integer types of any bit width, but they must have the same bit
6057 width. The second element of the result structure must be of
6058 type <tt>i1</tt>. <tt>%a</tt> and <tt>%b</tt> are the two values that will
6059 undergo unsigned addition.</p>
6062 <p>The '<tt>llvm.uadd.with.overflow</tt>' family of intrinsic functions perform
6063 an unsigned addition of the two arguments. They return a structure —
6064 the first element of which is the sum, and the second element of which is a
6065 bit specifying if the unsigned summation resulted in a carry.</p>
6069 %res = call {i32, i1} @llvm.uadd.with.overflow.i32(i32 %a, i32 %b)
6070 %sum = extractvalue {i32, i1} %res, 0
6071 %obit = extractvalue {i32, i1} %res, 1
6072 br i1 %obit, label %carry, label %normal
6077 <!-- _______________________________________________________________________ -->
6078 <div class="doc_subsubsection">
6079 <a name="int_ssub_overflow">'<tt>llvm.ssub.with.overflow.*</tt>' Intrinsics</a>
6082 <div class="doc_text">
6085 <p>This is an overloaded intrinsic. You can use <tt>llvm.ssub.with.overflow</tt>
6086 on any integer bit width.</p>
6089 declare {i16, i1} @llvm.ssub.with.overflow.i16(i16 %a, i16 %b)
6090 declare {i32, i1} @llvm.ssub.with.overflow.i32(i32 %a, i32 %b)
6091 declare {i64, i1} @llvm.ssub.with.overflow.i64(i64 %a, i64 %b)
6095 <p>The '<tt>llvm.ssub.with.overflow</tt>' family of intrinsic functions perform
6096 a signed subtraction of the two arguments, and indicate whether an overflow
6097 occurred during the signed subtraction.</p>
6100 <p>The arguments (%a and %b) and the first element of the result structure may
6101 be of integer types of any bit width, but they must have the same bit
6102 width. The second element of the result structure must be of
6103 type <tt>i1</tt>. <tt>%a</tt> and <tt>%b</tt> are the two values that will
6104 undergo signed subtraction.</p>
6107 <p>The '<tt>llvm.ssub.with.overflow</tt>' family of intrinsic functions perform
6108 a signed subtraction of the two arguments. They return a structure —
6109 the first element of which is the subtraction, and the second element of
6110 which is a bit specifying if the signed subtraction resulted in an
6115 %res = call {i32, i1} @llvm.ssub.with.overflow.i32(i32 %a, i32 %b)
6116 %sum = extractvalue {i32, i1} %res, 0
6117 %obit = extractvalue {i32, i1} %res, 1
6118 br i1 %obit, label %overflow, label %normal
6123 <!-- _______________________________________________________________________ -->
6124 <div class="doc_subsubsection">
6125 <a name="int_usub_overflow">'<tt>llvm.usub.with.overflow.*</tt>' Intrinsics</a>
6128 <div class="doc_text">
6131 <p>This is an overloaded intrinsic. You can use <tt>llvm.usub.with.overflow</tt>
6132 on any integer bit width.</p>
6135 declare {i16, i1} @llvm.usub.with.overflow.i16(i16 %a, i16 %b)
6136 declare {i32, i1} @llvm.usub.with.overflow.i32(i32 %a, i32 %b)
6137 declare {i64, i1} @llvm.usub.with.overflow.i64(i64 %a, i64 %b)
6141 <p>The '<tt>llvm.usub.with.overflow</tt>' family of intrinsic functions perform
6142 an unsigned subtraction of the two arguments, and indicate whether an
6143 overflow occurred during the unsigned subtraction.</p>
6146 <p>The arguments (%a and %b) and the first element of the result structure may
6147 be of integer types of any bit width, but they must have the same bit
6148 width. The second element of the result structure must be of
6149 type <tt>i1</tt>. <tt>%a</tt> and <tt>%b</tt> are the two values that will
6150 undergo unsigned subtraction.</p>
6153 <p>The '<tt>llvm.usub.with.overflow</tt>' family of intrinsic functions perform
6154 an unsigned subtraction of the two arguments. They return a structure —
6155 the first element of which is the subtraction, and the second element of
6156 which is a bit specifying if the unsigned subtraction resulted in an
6161 %res = call {i32, i1} @llvm.usub.with.overflow.i32(i32 %a, i32 %b)
6162 %sum = extractvalue {i32, i1} %res, 0
6163 %obit = extractvalue {i32, i1} %res, 1
6164 br i1 %obit, label %overflow, label %normal
6169 <!-- _______________________________________________________________________ -->
6170 <div class="doc_subsubsection">
6171 <a name="int_smul_overflow">'<tt>llvm.smul.with.overflow.*</tt>' Intrinsics</a>
6174 <div class="doc_text">
6177 <p>This is an overloaded intrinsic. You can use <tt>llvm.smul.with.overflow</tt>
6178 on any integer bit width.</p>
6181 declare {i16, i1} @llvm.smul.with.overflow.i16(i16 %a, i16 %b)
6182 declare {i32, i1} @llvm.smul.with.overflow.i32(i32 %a, i32 %b)
6183 declare {i64, i1} @llvm.smul.with.overflow.i64(i64 %a, i64 %b)
6188 <p>The '<tt>llvm.smul.with.overflow</tt>' family of intrinsic functions perform
6189 a signed multiplication of the two arguments, and indicate whether an
6190 overflow occurred during the signed multiplication.</p>
6193 <p>The arguments (%a and %b) and the first element of the result structure may
6194 be of integer types of any bit width, but they must have the same bit
6195 width. The second element of the result structure must be of
6196 type <tt>i1</tt>. <tt>%a</tt> and <tt>%b</tt> are the two values that will
6197 undergo signed multiplication.</p>
6200 <p>The '<tt>llvm.smul.with.overflow</tt>' family of intrinsic functions perform
6201 a signed multiplication of the two arguments. They return a structure —
6202 the first element of which is the multiplication, and the second element of
6203 which is a bit specifying if the signed multiplication resulted in an
6208 %res = call {i32, i1} @llvm.smul.with.overflow.i32(i32 %a, i32 %b)
6209 %sum = extractvalue {i32, i1} %res, 0
6210 %obit = extractvalue {i32, i1} %res, 1
6211 br i1 %obit, label %overflow, label %normal
6216 <!-- _______________________________________________________________________ -->
6217 <div class="doc_subsubsection">
6218 <a name="int_umul_overflow">'<tt>llvm.umul.with.overflow.*</tt>' Intrinsics</a>
6221 <div class="doc_text">
6224 <p>This is an overloaded intrinsic. You can use <tt>llvm.umul.with.overflow</tt>
6225 on any integer bit width.</p>
6228 declare {i16, i1} @llvm.umul.with.overflow.i16(i16 %a, i16 %b)
6229 declare {i32, i1} @llvm.umul.with.overflow.i32(i32 %a, i32 %b)
6230 declare {i64, i1} @llvm.umul.with.overflow.i64(i64 %a, i64 %b)
6234 <p>The '<tt>llvm.umul.with.overflow</tt>' family of intrinsic functions perform
6235 a unsigned multiplication of the two arguments, and indicate whether an
6236 overflow occurred during the unsigned multiplication.</p>
6239 <p>The arguments (%a and %b) and the first element of the result structure may
6240 be of integer types of any bit width, but they must have the same bit
6241 width. The second element of the result structure must be of
6242 type <tt>i1</tt>. <tt>%a</tt> and <tt>%b</tt> are the two values that will
6243 undergo unsigned multiplication.</p>
6246 <p>The '<tt>llvm.umul.with.overflow</tt>' family of intrinsic functions perform
6247 an unsigned multiplication of the two arguments. They return a structure
6248 — the first element of which is the multiplication, and the second
6249 element of which is a bit specifying if the unsigned multiplication resulted
6254 %res = call {i32, i1} @llvm.umul.with.overflow.i32(i32 %a, i32 %b)
6255 %sum = extractvalue {i32, i1} %res, 0
6256 %obit = extractvalue {i32, i1} %res, 1
6257 br i1 %obit, label %overflow, label %normal
6262 <!-- ======================================================================= -->
6263 <div class="doc_subsection">
6264 <a name="int_debugger">Debugger Intrinsics</a>
6267 <div class="doc_text">
6269 <p>The LLVM debugger intrinsics (which all start with <tt>llvm.dbg.</tt>
6270 prefix), are described in
6271 the <a href="SourceLevelDebugging.html#format_common_intrinsics">LLVM Source
6272 Level Debugging</a> document.</p>
6276 <!-- ======================================================================= -->
6277 <div class="doc_subsection">
6278 <a name="int_eh">Exception Handling Intrinsics</a>
6281 <div class="doc_text">
6283 <p>The LLVM exception handling intrinsics (which all start with
6284 <tt>llvm.eh.</tt> prefix), are described in
6285 the <a href="ExceptionHandling.html#format_common_intrinsics">LLVM Exception
6286 Handling</a> document.</p>
6290 <!-- ======================================================================= -->
6291 <div class="doc_subsection">
6292 <a name="int_trampoline">Trampoline Intrinsic</a>
6295 <div class="doc_text">
6297 <p>This intrinsic makes it possible to excise one parameter, marked with
6298 the <tt>nest</tt> attribute, from a function. The result is a callable
6299 function pointer lacking the nest parameter - the caller does not need to
6300 provide a value for it. Instead, the value to use is stored in advance in a
6301 "trampoline", a block of memory usually allocated on the stack, which also
6302 contains code to splice the nest value into the argument list. This is used
6303 to implement the GCC nested function address extension.</p>
6305 <p>For example, if the function is
6306 <tt>i32 f(i8* nest %c, i32 %x, i32 %y)</tt> then the resulting function
6307 pointer has signature <tt>i32 (i32, i32)*</tt>. It can be created as
6310 <div class="doc_code">
6312 %tramp = alloca [10 x i8], align 4 ; size and alignment only correct for X86
6313 %tramp1 = getelementptr [10 x i8]* %tramp, i32 0, i32 0
6314 %p = call i8* @llvm.init.trampoline( i8* %tramp1, i8* bitcast (i32 (i8* nest , i32, i32)* @f to i8*), i8* %nval )
6315 %fp = bitcast i8* %p to i32 (i32, i32)*
6319 <p>The call <tt>%val = call i32 %fp( i32 %x, i32 %y )</tt> is then equivalent
6320 to <tt>%val = call i32 %f( i8* %nval, i32 %x, i32 %y )</tt>.</p>
6324 <!-- _______________________________________________________________________ -->
6325 <div class="doc_subsubsection">
6326 <a name="int_it">'<tt>llvm.init.trampoline</tt>' Intrinsic</a>
6329 <div class="doc_text">
6333 declare i8* @llvm.init.trampoline(i8* <tramp>, i8* <func>, i8* <nval>)
6337 <p>This fills the memory pointed to by <tt>tramp</tt> with code and returns a
6338 function pointer suitable for executing it.</p>
6341 <p>The <tt>llvm.init.trampoline</tt> intrinsic takes three arguments, all
6342 pointers. The <tt>tramp</tt> argument must point to a sufficiently large and
6343 sufficiently aligned block of memory; this memory is written to by the
6344 intrinsic. Note that the size and the alignment are target-specific - LLVM
6345 currently provides no portable way of determining them, so a front-end that
6346 generates this intrinsic needs to have some target-specific knowledge.
6347 The <tt>func</tt> argument must hold a function bitcast to
6348 an <tt>i8*</tt>.</p>
6351 <p>The block of memory pointed to by <tt>tramp</tt> is filled with target
6352 dependent code, turning it into a function. A pointer to this function is
6353 returned, but needs to be bitcast to an <a href="#int_trampoline">appropriate
6354 function pointer type</a> before being called. The new function's signature
6355 is the same as that of <tt>func</tt> with any arguments marked with
6356 the <tt>nest</tt> attribute removed. At most one such <tt>nest</tt> argument
6357 is allowed, and it must be of pointer type. Calling the new function is
6358 equivalent to calling <tt>func</tt> with the same argument list, but
6359 with <tt>nval</tt> used for the missing <tt>nest</tt> argument. If, after
6360 calling <tt>llvm.init.trampoline</tt>, the memory pointed to
6361 by <tt>tramp</tt> is modified, then the effect of any later call to the
6362 returned function pointer is undefined.</p>
6366 <!-- ======================================================================= -->
6367 <div class="doc_subsection">
6368 <a name="int_atomics">Atomic Operations and Synchronization Intrinsics</a>
6371 <div class="doc_text">
6373 <p>These intrinsic functions expand the "universal IR" of LLVM to represent
6374 hardware constructs for atomic operations and memory synchronization. This
6375 provides an interface to the hardware, not an interface to the programmer. It
6376 is aimed at a low enough level to allow any programming models or APIs
6377 (Application Programming Interfaces) which need atomic behaviors to map
6378 cleanly onto it. It is also modeled primarily on hardware behavior. Just as
6379 hardware provides a "universal IR" for source languages, it also provides a
6380 starting point for developing a "universal" atomic operation and
6381 synchronization IR.</p>
6383 <p>These do <em>not</em> form an API such as high-level threading libraries,
6384 software transaction memory systems, atomic primitives, and intrinsic
6385 functions as found in BSD, GNU libc, atomic_ops, APR, and other system and
6386 application libraries. The hardware interface provided by LLVM should allow
6387 a clean implementation of all of these APIs and parallel programming models.
6388 No one model or paradigm should be selected above others unless the hardware
6389 itself ubiquitously does so.</p>
6393 <!-- _______________________________________________________________________ -->
6394 <div class="doc_subsubsection">
6395 <a name="int_memory_barrier">'<tt>llvm.memory.barrier</tt>' Intrinsic</a>
6397 <div class="doc_text">
6400 declare void @llvm.memory.barrier( i1 <ll>, i1 <ls>, i1 <sl>, i1 <ss>, i1 <device> )
6404 <p>The <tt>llvm.memory.barrier</tt> intrinsic guarantees ordering between
6405 specific pairs of memory access types.</p>
6408 <p>The <tt>llvm.memory.barrier</tt> intrinsic requires five boolean arguments.
6409 The first four arguments enables a specific barrier as listed below. The
6410 fith argument specifies that the barrier applies to io or device or uncached
6414 <li><tt>ll</tt>: load-load barrier</li>
6415 <li><tt>ls</tt>: load-store barrier</li>
6416 <li><tt>sl</tt>: store-load barrier</li>
6417 <li><tt>ss</tt>: store-store barrier</li>
6418 <li><tt>device</tt>: barrier applies to device and uncached memory also.</li>
6422 <p>This intrinsic causes the system to enforce some ordering constraints upon
6423 the loads and stores of the program. This barrier does not
6424 indicate <em>when</em> any events will occur, it only enforces
6425 an <em>order</em> in which they occur. For any of the specified pairs of load
6426 and store operations (f.ex. load-load, or store-load), all of the first
6427 operations preceding the barrier will complete before any of the second
6428 operations succeeding the barrier begin. Specifically the semantics for each
6429 pairing is as follows:</p>
6432 <li><tt>ll</tt>: All loads before the barrier must complete before any load
6433 after the barrier begins.</li>
6434 <li><tt>ls</tt>: All loads before the barrier must complete before any
6435 store after the barrier begins.</li>
6436 <li><tt>ss</tt>: All stores before the barrier must complete before any
6437 store after the barrier begins.</li>
6438 <li><tt>sl</tt>: All stores before the barrier must complete before any
6439 load after the barrier begins.</li>
6442 <p>These semantics are applied with a logical "and" behavior when more than one
6443 is enabled in a single memory barrier intrinsic.</p>
6445 <p>Backends may implement stronger barriers than those requested when they do
6446 not support as fine grained a barrier as requested. Some architectures do
6447 not need all types of barriers and on such architectures, these become
6455 %result1 = load i32* %ptr <i>; yields {i32}:result1 = 4</i>
6456 call void @llvm.memory.barrier( i1 false, i1 true, i1 false, i1 false )
6457 <i>; guarantee the above finishes</i>
6458 store i32 8, %ptr <i>; before this begins</i>
6463 <!-- _______________________________________________________________________ -->
6464 <div class="doc_subsubsection">
6465 <a name="int_atomic_cmp_swap">'<tt>llvm.atomic.cmp.swap.*</tt>' Intrinsic</a>
6468 <div class="doc_text">
6471 <p>This is an overloaded intrinsic. You can use <tt>llvm.atomic.cmp.swap</tt> on
6472 any integer bit width and for different address spaces. Not all targets
6473 support all bit widths however.</p>
6476 declare i8 @llvm.atomic.cmp.swap.i8.p0i8( i8* <ptr>, i8 <cmp>, i8 <val> )
6477 declare i16 @llvm.atomic.cmp.swap.i16.p0i16( i16* <ptr>, i16 <cmp>, i16 <val> )
6478 declare i32 @llvm.atomic.cmp.swap.i32.p0i32( i32* <ptr>, i32 <cmp>, i32 <val> )
6479 declare i64 @llvm.atomic.cmp.swap.i64.p0i64( i64* <ptr>, i64 <cmp>, i64 <val> )
6483 <p>This loads a value in memory and compares it to a given value. If they are
6484 equal, it stores a new value into the memory.</p>
6487 <p>The <tt>llvm.atomic.cmp.swap</tt> intrinsic takes three arguments. The result
6488 as well as both <tt>cmp</tt> and <tt>val</tt> must be integer values with the
6489 same bit width. The <tt>ptr</tt> argument must be a pointer to a value of
6490 this integer type. While any bit width integer may be used, targets may only
6491 lower representations they support in hardware.</p>
6494 <p>This entire intrinsic must be executed atomically. It first loads the value
6495 in memory pointed to by <tt>ptr</tt> and compares it with the
6496 value <tt>cmp</tt>. If they are equal, <tt>val</tt> is stored into the
6497 memory. The loaded value is yielded in all cases. This provides the
6498 equivalent of an atomic compare-and-swap operation within the SSA
6506 %val1 = add i32 4, 4
6507 %result1 = call i32 @llvm.atomic.cmp.swap.i32.p0i32( i32* %ptr, i32 4, %val1 )
6508 <i>; yields {i32}:result1 = 4</i>
6509 %stored1 = icmp eq i32 %result1, 4 <i>; yields {i1}:stored1 = true</i>
6510 %memval1 = load i32* %ptr <i>; yields {i32}:memval1 = 8</i>
6512 %val2 = add i32 1, 1
6513 %result2 = call i32 @llvm.atomic.cmp.swap.i32.p0i32( i32* %ptr, i32 5, %val2 )
6514 <i>; yields {i32}:result2 = 8</i>
6515 %stored2 = icmp eq i32 %result2, 5 <i>; yields {i1}:stored2 = false</i>
6517 %memval2 = load i32* %ptr <i>; yields {i32}:memval2 = 8</i>
6522 <!-- _______________________________________________________________________ -->
6523 <div class="doc_subsubsection">
6524 <a name="int_atomic_swap">'<tt>llvm.atomic.swap.*</tt>' Intrinsic</a>
6526 <div class="doc_text">
6529 <p>This is an overloaded intrinsic. You can use <tt>llvm.atomic.swap</tt> on any
6530 integer bit width. Not all targets support all bit widths however.</p>
6533 declare i8 @llvm.atomic.swap.i8.p0i8( i8* <ptr>, i8 <val> )
6534 declare i16 @llvm.atomic.swap.i16.p0i16( i16* <ptr>, i16 <val> )
6535 declare i32 @llvm.atomic.swap.i32.p0i32( i32* <ptr>, i32 <val> )
6536 declare i64 @llvm.atomic.swap.i64.p0i64( i64* <ptr>, i64 <val> )
6540 <p>This intrinsic loads the value stored in memory at <tt>ptr</tt> and yields
6541 the value from memory. It then stores the value in <tt>val</tt> in the memory
6542 at <tt>ptr</tt>.</p>
6545 <p>The <tt>llvm.atomic.swap</tt> intrinsic takes two arguments. Both
6546 the <tt>val</tt> argument and the result must be integers of the same bit
6547 width. The first argument, <tt>ptr</tt>, must be a pointer to a value of this
6548 integer type. The targets may only lower integer representations they
6552 <p>This intrinsic loads the value pointed to by <tt>ptr</tt>, yields it, and
6553 stores <tt>val</tt> back into <tt>ptr</tt> atomically. This provides the
6554 equivalent of an atomic swap operation within the SSA framework.</p>
6561 %val1 = add i32 4, 4
6562 %result1 = call i32 @llvm.atomic.swap.i32.p0i32( i32* %ptr, i32 %val1 )
6563 <i>; yields {i32}:result1 = 4</i>
6564 %stored1 = icmp eq i32 %result1, 4 <i>; yields {i1}:stored1 = true</i>
6565 %memval1 = load i32* %ptr <i>; yields {i32}:memval1 = 8</i>
6567 %val2 = add i32 1, 1
6568 %result2 = call i32 @llvm.atomic.swap.i32.p0i32( i32* %ptr, i32 %val2 )
6569 <i>; yields {i32}:result2 = 8</i>
6571 %stored2 = icmp eq i32 %result2, 8 <i>; yields {i1}:stored2 = true</i>
6572 %memval2 = load i32* %ptr <i>; yields {i32}:memval2 = 2</i>
6577 <!-- _______________________________________________________________________ -->
6578 <div class="doc_subsubsection">
6579 <a name="int_atomic_load_add">'<tt>llvm.atomic.load.add.*</tt>' Intrinsic</a>
6583 <div class="doc_text">
6586 <p>This is an overloaded intrinsic. You can use <tt>llvm.atomic.load.add</tt> on
6587 any integer bit width. Not all targets support all bit widths however.</p>
6590 declare i8 @llvm.atomic.load.add.i8..p0i8( i8* <ptr>, i8 <delta> )
6591 declare i16 @llvm.atomic.load.add.i16..p0i16( i16* <ptr>, i16 <delta> )
6592 declare i32 @llvm.atomic.load.add.i32..p0i32( i32* <ptr>, i32 <delta> )
6593 declare i64 @llvm.atomic.load.add.i64..p0i64( i64* <ptr>, i64 <delta> )
6597 <p>This intrinsic adds <tt>delta</tt> to the value stored in memory
6598 at <tt>ptr</tt>. It yields the original value at <tt>ptr</tt>.</p>
6601 <p>The intrinsic takes two arguments, the first a pointer to an integer value
6602 and the second an integer value. The result is also an integer value. These
6603 integer types can have any bit width, but they must all have the same bit
6604 width. The targets may only lower integer representations they support.</p>
6607 <p>This intrinsic does a series of operations atomically. It first loads the
6608 value stored at <tt>ptr</tt>. It then adds <tt>delta</tt>, stores the result
6609 to <tt>ptr</tt>. It yields the original value stored at <tt>ptr</tt>.</p>
6615 %result1 = call i32 @llvm.atomic.load.add.i32.p0i32( i32* %ptr, i32 4 )
6616 <i>; yields {i32}:result1 = 4</i>
6617 %result2 = call i32 @llvm.atomic.load.add.i32.p0i32( i32* %ptr, i32 2 )
6618 <i>; yields {i32}:result2 = 8</i>
6619 %result3 = call i32 @llvm.atomic.load.add.i32.p0i32( i32* %ptr, i32 5 )
6620 <i>; yields {i32}:result3 = 10</i>
6621 %memval1 = load i32* %ptr <i>; yields {i32}:memval1 = 15</i>
6626 <!-- _______________________________________________________________________ -->
6627 <div class="doc_subsubsection">
6628 <a name="int_atomic_load_sub">'<tt>llvm.atomic.load.sub.*</tt>' Intrinsic</a>
6632 <div class="doc_text">
6635 <p>This is an overloaded intrinsic. You can use <tt>llvm.atomic.load.sub</tt> on
6636 any integer bit width and for different address spaces. Not all targets
6637 support all bit widths however.</p>
6640 declare i8 @llvm.atomic.load.sub.i8.p0i32( i8* <ptr>, i8 <delta> )
6641 declare i16 @llvm.atomic.load.sub.i16.p0i32( i16* <ptr>, i16 <delta> )
6642 declare i32 @llvm.atomic.load.sub.i32.p0i32( i32* <ptr>, i32 <delta> )
6643 declare i64 @llvm.atomic.load.sub.i64.p0i32( i64* <ptr>, i64 <delta> )
6647 <p>This intrinsic subtracts <tt>delta</tt> to the value stored in memory at
6648 <tt>ptr</tt>. It yields the original value at <tt>ptr</tt>.</p>
6651 <p>The intrinsic takes two arguments, the first a pointer to an integer value
6652 and the second an integer value. The result is also an integer value. These
6653 integer types can have any bit width, but they must all have the same bit
6654 width. The targets may only lower integer representations they support.</p>
6657 <p>This intrinsic does a series of operations atomically. It first loads the
6658 value stored at <tt>ptr</tt>. It then subtracts <tt>delta</tt>, stores the
6659 result to <tt>ptr</tt>. It yields the original value stored
6660 at <tt>ptr</tt>.</p>
6666 %result1 = call i32 @llvm.atomic.load.sub.i32.p0i32( i32* %ptr, i32 4 )
6667 <i>; yields {i32}:result1 = 8</i>
6668 %result2 = call i32 @llvm.atomic.load.sub.i32.p0i32( i32* %ptr, i32 2 )
6669 <i>; yields {i32}:result2 = 4</i>
6670 %result3 = call i32 @llvm.atomic.load.sub.i32.p0i32( i32* %ptr, i32 5 )
6671 <i>; yields {i32}:result3 = 2</i>
6672 %memval1 = load i32* %ptr <i>; yields {i32}:memval1 = -3</i>
6677 <!-- _______________________________________________________________________ -->
6678 <div class="doc_subsubsection">
6679 <a name="int_atomic_load_and">'<tt>llvm.atomic.load.and.*</tt>' Intrinsic</a><br>
6680 <a name="int_atomic_load_nand">'<tt>llvm.atomic.load.nand.*</tt>' Intrinsic</a><br>
6681 <a name="int_atomic_load_or">'<tt>llvm.atomic.load.or.*</tt>' Intrinsic</a><br>
6682 <a name="int_atomic_load_xor">'<tt>llvm.atomic.load.xor.*</tt>' Intrinsic</a><br>
6685 <div class="doc_text">
6688 <p>These are overloaded intrinsics. You can
6689 use <tt>llvm.atomic.load_and</tt>, <tt>llvm.atomic.load_nand</tt>,
6690 <tt>llvm.atomic.load_or</tt>, and <tt>llvm.atomic.load_xor</tt> on any integer
6691 bit width and for different address spaces. Not all targets support all bit
6695 declare i8 @llvm.atomic.load.and.i8.p0i8( i8* <ptr>, i8 <delta> )
6696 declare i16 @llvm.atomic.load.and.i16.p0i16( i16* <ptr>, i16 <delta> )
6697 declare i32 @llvm.atomic.load.and.i32.p0i32( i32* <ptr>, i32 <delta> )
6698 declare i64 @llvm.atomic.load.and.i64.p0i64( i64* <ptr>, i64 <delta> )
6702 declare i8 @llvm.atomic.load.or.i8.p0i8( i8* <ptr>, i8 <delta> )
6703 declare i16 @llvm.atomic.load.or.i16.p0i16( i16* <ptr>, i16 <delta> )
6704 declare i32 @llvm.atomic.load.or.i32.p0i32( i32* <ptr>, i32 <delta> )
6705 declare i64 @llvm.atomic.load.or.i64.p0i64( i64* <ptr>, i64 <delta> )
6709 declare i8 @llvm.atomic.load.nand.i8.p0i32( i8* <ptr>, i8 <delta> )
6710 declare i16 @llvm.atomic.load.nand.i16.p0i32( i16* <ptr>, i16 <delta> )
6711 declare i32 @llvm.atomic.load.nand.i32.p0i32( i32* <ptr>, i32 <delta> )
6712 declare i64 @llvm.atomic.load.nand.i64.p0i32( i64* <ptr>, i64 <delta> )
6716 declare i8 @llvm.atomic.load.xor.i8.p0i32( i8* <ptr>, i8 <delta> )
6717 declare i16 @llvm.atomic.load.xor.i16.p0i32( i16* <ptr>, i16 <delta> )
6718 declare i32 @llvm.atomic.load.xor.i32.p0i32( i32* <ptr>, i32 <delta> )
6719 declare i64 @llvm.atomic.load.xor.i64.p0i32( i64* <ptr>, i64 <delta> )
6723 <p>These intrinsics bitwise the operation (and, nand, or, xor) <tt>delta</tt> to
6724 the value stored in memory at <tt>ptr</tt>. It yields the original value
6725 at <tt>ptr</tt>.</p>
6728 <p>These intrinsics take two arguments, the first a pointer to an integer value
6729 and the second an integer value. The result is also an integer value. These
6730 integer types can have any bit width, but they must all have the same bit
6731 width. The targets may only lower integer representations they support.</p>
6734 <p>These intrinsics does a series of operations atomically. They first load the
6735 value stored at <tt>ptr</tt>. They then do the bitwise
6736 operation <tt>delta</tt>, store the result to <tt>ptr</tt>. They yield the
6737 original value stored at <tt>ptr</tt>.</p>
6742 store i32 0x0F0F, %ptr
6743 %result0 = call i32 @llvm.atomic.load.nand.i32.p0i32( i32* %ptr, i32 0xFF )
6744 <i>; yields {i32}:result0 = 0x0F0F</i>
6745 %result1 = call i32 @llvm.atomic.load.and.i32.p0i32( i32* %ptr, i32 0xFF )
6746 <i>; yields {i32}:result1 = 0xFFFFFFF0</i>
6747 %result2 = call i32 @llvm.atomic.load.or.i32.p0i32( i32* %ptr, i32 0F )
6748 <i>; yields {i32}:result2 = 0xF0</i>
6749 %result3 = call i32 @llvm.atomic.load.xor.i32.p0i32( i32* %ptr, i32 0F )
6750 <i>; yields {i32}:result3 = FF</i>
6751 %memval1 = load i32* %ptr <i>; yields {i32}:memval1 = F0</i>
6756 <!-- _______________________________________________________________________ -->
6757 <div class="doc_subsubsection">
6758 <a name="int_atomic_load_max">'<tt>llvm.atomic.load.max.*</tt>' Intrinsic</a><br>
6759 <a name="int_atomic_load_min">'<tt>llvm.atomic.load.min.*</tt>' Intrinsic</a><br>
6760 <a name="int_atomic_load_umax">'<tt>llvm.atomic.load.umax.*</tt>' Intrinsic</a><br>
6761 <a name="int_atomic_load_umin">'<tt>llvm.atomic.load.umin.*</tt>' Intrinsic</a><br>
6764 <div class="doc_text">
6767 <p>These are overloaded intrinsics. You can use <tt>llvm.atomic.load_max</tt>,
6768 <tt>llvm.atomic.load_min</tt>, <tt>llvm.atomic.load_umax</tt>, and
6769 <tt>llvm.atomic.load_umin</tt> on any integer bit width and for different
6770 address spaces. Not all targets support all bit widths however.</p>
6773 declare i8 @llvm.atomic.load.max.i8.p0i8( i8* <ptr>, i8 <delta> )
6774 declare i16 @llvm.atomic.load.max.i16.p0i16( i16* <ptr>, i16 <delta> )
6775 declare i32 @llvm.atomic.load.max.i32.p0i32( i32* <ptr>, i32 <delta> )
6776 declare i64 @llvm.atomic.load.max.i64.p0i64( i64* <ptr>, i64 <delta> )
6780 declare i8 @llvm.atomic.load.min.i8.p0i8( i8* <ptr>, i8 <delta> )
6781 declare i16 @llvm.atomic.load.min.i16.p0i16( i16* <ptr>, i16 <delta> )
6782 declare i32 @llvm.atomic.load.min.i32..p0i32( i32* <ptr>, i32 <delta> )
6783 declare i64 @llvm.atomic.load.min.i64..p0i64( i64* <ptr>, i64 <delta> )
6787 declare i8 @llvm.atomic.load.umax.i8.p0i8( i8* <ptr>, i8 <delta> )
6788 declare i16 @llvm.atomic.load.umax.i16.p0i16( i16* <ptr>, i16 <delta> )
6789 declare i32 @llvm.atomic.load.umax.i32.p0i32( i32* <ptr>, i32 <delta> )
6790 declare i64 @llvm.atomic.load.umax.i64.p0i64( i64* <ptr>, i64 <delta> )
6794 declare i8 @llvm.atomic.load.umin.i8..p0i8( i8* <ptr>, i8 <delta> )
6795 declare i16 @llvm.atomic.load.umin.i16.p0i16( i16* <ptr>, i16 <delta> )
6796 declare i32 @llvm.atomic.load.umin.i32..p0i32( i32* <ptr>, i32 <delta> )
6797 declare i64 @llvm.atomic.load.umin.i64..p0i64( i64* <ptr>, i64 <delta> )
6801 <p>These intrinsics takes the signed or unsigned minimum or maximum of
6802 <tt>delta</tt> and the value stored in memory at <tt>ptr</tt>. It yields the
6803 original value at <tt>ptr</tt>.</p>
6806 <p>These intrinsics take two arguments, the first a pointer to an integer value
6807 and the second an integer value. The result is also an integer value. These
6808 integer types can have any bit width, but they must all have the same bit
6809 width. The targets may only lower integer representations they support.</p>
6812 <p>These intrinsics does a series of operations atomically. They first load the
6813 value stored at <tt>ptr</tt>. They then do the signed or unsigned min or
6814 max <tt>delta</tt> and the value, store the result to <tt>ptr</tt>. They
6815 yield the original value stored at <tt>ptr</tt>.</p>
6821 %result0 = call i32 @llvm.atomic.load.min.i32.p0i32( i32* %ptr, i32 -2 )
6822 <i>; yields {i32}:result0 = 7</i>
6823 %result1 = call i32 @llvm.atomic.load.max.i32.p0i32( i32* %ptr, i32 8 )
6824 <i>; yields {i32}:result1 = -2</i>
6825 %result2 = call i32 @llvm.atomic.load.umin.i32.p0i32( i32* %ptr, i32 10 )
6826 <i>; yields {i32}:result2 = 8</i>
6827 %result3 = call i32 @llvm.atomic.load.umax.i32.p0i32( i32* %ptr, i32 30 )
6828 <i>; yields {i32}:result3 = 8</i>
6829 %memval1 = load i32* %ptr <i>; yields {i32}:memval1 = 30</i>
6834 <!-- ======================================================================= -->
6835 <div class="doc_subsection">
6836 <a name="int_general">General Intrinsics</a>
6839 <div class="doc_text">
6841 <p>This class of intrinsics is designed to be generic and has no specific
6846 <!-- _______________________________________________________________________ -->
6847 <div class="doc_subsubsection">
6848 <a name="int_var_annotation">'<tt>llvm.var.annotation</tt>' Intrinsic</a>
6851 <div class="doc_text">
6855 declare void @llvm.var.annotation(i8* <val>, i8* <str>, i8* <str>, i32 <int> )
6859 <p>The '<tt>llvm.var.annotation</tt>' intrinsic.</p>
6862 <p>The first argument is a pointer to a value, the second is a pointer to a
6863 global string, the third is a pointer to a global string which is the source
6864 file name, and the last argument is the line number.</p>
6867 <p>This intrinsic allows annotation of local variables with arbitrary strings.
6868 This can be useful for special purpose optimizations that want to look for
6869 these annotations. These have no other defined use, they are ignored by code
6870 generation and optimization.</p>
6874 <!-- _______________________________________________________________________ -->
6875 <div class="doc_subsubsection">
6876 <a name="int_annotation">'<tt>llvm.annotation.*</tt>' Intrinsic</a>
6879 <div class="doc_text">
6882 <p>This is an overloaded intrinsic. You can use '<tt>llvm.annotation</tt>' on
6883 any integer bit width.</p>
6886 declare i8 @llvm.annotation.i8(i8 <val>, i8* <str>, i8* <str>, i32 <int> )
6887 declare i16 @llvm.annotation.i16(i16 <val>, i8* <str>, i8* <str>, i32 <int> )
6888 declare i32 @llvm.annotation.i32(i32 <val>, i8* <str>, i8* <str>, i32 <int> )
6889 declare i64 @llvm.annotation.i64(i64 <val>, i8* <str>, i8* <str>, i32 <int> )
6890 declare i256 @llvm.annotation.i256(i256 <val>, i8* <str>, i8* <str>, i32 <int> )
6894 <p>The '<tt>llvm.annotation</tt>' intrinsic.</p>
6897 <p>The first argument is an integer value (result of some expression), the
6898 second is a pointer to a global string, the third is a pointer to a global
6899 string which is the source file name, and the last argument is the line
6900 number. It returns the value of the first argument.</p>
6903 <p>This intrinsic allows annotations to be put on arbitrary expressions with
6904 arbitrary strings. This can be useful for special purpose optimizations that
6905 want to look for these annotations. These have no other defined use, they
6906 are ignored by code generation and optimization.</p>
6910 <!-- _______________________________________________________________________ -->
6911 <div class="doc_subsubsection">
6912 <a name="int_trap">'<tt>llvm.trap</tt>' Intrinsic</a>
6915 <div class="doc_text">
6919 declare void @llvm.trap()
6923 <p>The '<tt>llvm.trap</tt>' intrinsic.</p>
6929 <p>This intrinsics is lowered to the target dependent trap instruction. If the
6930 target does not have a trap instruction, this intrinsic will be lowered to
6931 the call of the <tt>abort()</tt> function.</p>
6935 <!-- _______________________________________________________________________ -->
6936 <div class="doc_subsubsection">
6937 <a name="int_stackprotector">'<tt>llvm.stackprotector</tt>' Intrinsic</a>
6940 <div class="doc_text">
6944 declare void @llvm.stackprotector( i8* <guard>, i8** <slot> )
6948 <p>The <tt>llvm.stackprotector</tt> intrinsic takes the <tt>guard</tt> and
6949 stores it onto the stack at <tt>slot</tt>. The stack slot is adjusted to
6950 ensure that it is placed on the stack before local variables.</p>
6953 <p>The <tt>llvm.stackprotector</tt> intrinsic requires two pointer
6954 arguments. The first argument is the value loaded from the stack
6955 guard <tt>@__stack_chk_guard</tt>. The second variable is an <tt>alloca</tt>
6956 that has enough space to hold the value of the guard.</p>
6959 <p>This intrinsic causes the prologue/epilogue inserter to force the position of
6960 the <tt>AllocaInst</tt> stack slot to be before local variables on the
6961 stack. This is to ensure that if a local variable on the stack is
6962 overwritten, it will destroy the value of the guard. When the function exits,
6963 the guard on the stack is checked against the original guard. If they're
6964 different, then the program aborts by calling the <tt>__stack_chk_fail()</tt>
6969 <!-- *********************************************************************** -->
6972 <a href="http://jigsaw.w3.org/css-validator/check/referer"><img
6973 src="http://jigsaw.w3.org/css-validator/images/vcss-blue" alt="Valid CSS"></a>
6974 <a href="http://validator.w3.org/check/referer"><img
6975 src="http://www.w3.org/Icons/valid-html401-blue" alt="Valid HTML 4.01"></a>
6977 <a href="mailto:sabre@nondot.org">Chris Lattner</a><br>
6978 <a href="http://llvm.org">The LLVM Compiler Infrastructure</a><br>
6979 Last modified: $Date$