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_linker_private_weak">'<tt>linker_private_weak</tt>' Linkage</a></li>
28 <li><a href="#linkage_linker_private_weak_def_auto">'<tt>linker_private_weak_def_auto</tt>' Linkage</a></li>
29 <li><a href="#linkage_internal">'<tt>internal</tt>' Linkage</a></li>
30 <li><a href="#linkage_available_externally">'<tt>available_externally</tt>' Linkage</a></li>
31 <li><a href="#linkage_linkonce">'<tt>linkonce</tt>' Linkage</a></li>
32 <li><a href="#linkage_common">'<tt>common</tt>' Linkage</a></li>
33 <li><a href="#linkage_weak">'<tt>weak</tt>' Linkage</a></li>
34 <li><a href="#linkage_appending">'<tt>appending</tt>' Linkage</a></li>
35 <li><a href="#linkage_externweak">'<tt>extern_weak</tt>' Linkage</a></li>
36 <li><a href="#linkage_linkonce_odr">'<tt>linkonce_odr</tt>' Linkage</a></li>
37 <li><a href="#linkage_weak">'<tt>weak_odr</tt>' Linkage</a></li>
38 <li><a href="#linkage_external">'<tt>externally visible</tt>' Linkage</a></li>
39 <li><a href="#linkage_dllimport">'<tt>dllimport</tt>' Linkage</a></li>
40 <li><a href="#linkage_dllexport">'<tt>dllexport</tt>' Linkage</a></li>
43 <li><a href="#callingconv">Calling Conventions</a></li>
44 <li><a href="#namedtypes">Named Types</a></li>
45 <li><a href="#globalvars">Global Variables</a></li>
46 <li><a href="#functionstructure">Functions</a></li>
47 <li><a href="#aliasstructure">Aliases</a></li>
48 <li><a href="#namedmetadatastructure">Named Metadata</a></li>
49 <li><a href="#paramattrs">Parameter Attributes</a></li>
50 <li><a href="#fnattrs">Function Attributes</a></li>
51 <li><a href="#gc">Garbage Collector Names</a></li>
52 <li><a href="#moduleasm">Module-Level Inline Assembly</a></li>
53 <li><a href="#datalayout">Data Layout</a></li>
54 <li><a href="#pointeraliasing">Pointer Aliasing Rules</a></li>
55 <li><a href="#volatile">Volatile Memory Accesses</a></li>
58 <li><a href="#typesystem">Type System</a>
60 <li><a href="#t_classifications">Type Classifications</a></li>
61 <li><a href="#t_primitive">Primitive Types</a>
63 <li><a href="#t_integer">Integer Type</a></li>
64 <li><a href="#t_floating">Floating Point Types</a></li>
65 <li><a href="#t_x86mmx">X86mmx Type</a></li>
66 <li><a href="#t_void">Void Type</a></li>
67 <li><a href="#t_label">Label Type</a></li>
68 <li><a href="#t_metadata">Metadata Type</a></li>
71 <li><a href="#t_derived">Derived Types</a>
73 <li><a href="#t_aggregate">Aggregate Types</a>
75 <li><a href="#t_array">Array Type</a></li>
76 <li><a href="#t_struct">Structure Type</a></li>
77 <li><a href="#t_pstruct">Packed Structure Type</a></li>
78 <li><a href="#t_vector">Vector Type</a></li>
81 <li><a href="#t_function">Function Type</a></li>
82 <li><a href="#t_pointer">Pointer Type</a></li>
83 <li><a href="#t_opaque">Opaque Type</a></li>
86 <li><a href="#t_uprefs">Type Up-references</a></li>
89 <li><a href="#constants">Constants</a>
91 <li><a href="#simpleconstants">Simple Constants</a></li>
92 <li><a href="#complexconstants">Complex Constants</a></li>
93 <li><a href="#globalconstants">Global Variable and Function Addresses</a></li>
94 <li><a href="#undefvalues">Undefined Values</a></li>
95 <li><a href="#trapvalues">Trap Values</a></li>
96 <li><a href="#blockaddress">Addresses of Basic Blocks</a></li>
97 <li><a href="#constantexprs">Constant Expressions</a></li>
100 <li><a href="#othervalues">Other Values</a>
102 <li><a href="#inlineasm">Inline Assembler Expressions</a></li>
103 <li><a href="#metadata">Metadata Nodes and Metadata Strings</a></li>
106 <li><a href="#intrinsic_globals">Intrinsic Global Variables</a>
108 <li><a href="#intg_used">The '<tt>llvm.used</tt>' Global Variable</a></li>
109 <li><a href="#intg_compiler_used">The '<tt>llvm.compiler.used</tt>'
110 Global Variable</a></li>
111 <li><a href="#intg_global_ctors">The '<tt>llvm.global_ctors</tt>'
112 Global Variable</a></li>
113 <li><a href="#intg_global_dtors">The '<tt>llvm.global_dtors</tt>'
114 Global Variable</a></li>
117 <li><a href="#instref">Instruction Reference</a>
119 <li><a href="#terminators">Terminator Instructions</a>
121 <li><a href="#i_ret">'<tt>ret</tt>' Instruction</a></li>
122 <li><a href="#i_br">'<tt>br</tt>' Instruction</a></li>
123 <li><a href="#i_switch">'<tt>switch</tt>' Instruction</a></li>
124 <li><a href="#i_indirectbr">'<tt>indirectbr</tt>' Instruction</a></li>
125 <li><a href="#i_invoke">'<tt>invoke</tt>' Instruction</a></li>
126 <li><a href="#i_unwind">'<tt>unwind</tt>' Instruction</a></li>
127 <li><a href="#i_unreachable">'<tt>unreachable</tt>' Instruction</a></li>
130 <li><a href="#binaryops">Binary Operations</a>
132 <li><a href="#i_add">'<tt>add</tt>' Instruction</a></li>
133 <li><a href="#i_fadd">'<tt>fadd</tt>' Instruction</a></li>
134 <li><a href="#i_sub">'<tt>sub</tt>' Instruction</a></li>
135 <li><a href="#i_fsub">'<tt>fsub</tt>' Instruction</a></li>
136 <li><a href="#i_mul">'<tt>mul</tt>' Instruction</a></li>
137 <li><a href="#i_fmul">'<tt>fmul</tt>' Instruction</a></li>
138 <li><a href="#i_udiv">'<tt>udiv</tt>' Instruction</a></li>
139 <li><a href="#i_sdiv">'<tt>sdiv</tt>' Instruction</a></li>
140 <li><a href="#i_fdiv">'<tt>fdiv</tt>' Instruction</a></li>
141 <li><a href="#i_urem">'<tt>urem</tt>' Instruction</a></li>
142 <li><a href="#i_srem">'<tt>srem</tt>' Instruction</a></li>
143 <li><a href="#i_frem">'<tt>frem</tt>' Instruction</a></li>
146 <li><a href="#bitwiseops">Bitwise Binary Operations</a>
148 <li><a href="#i_shl">'<tt>shl</tt>' Instruction</a></li>
149 <li><a href="#i_lshr">'<tt>lshr</tt>' Instruction</a></li>
150 <li><a href="#i_ashr">'<tt>ashr</tt>' Instruction</a></li>
151 <li><a href="#i_and">'<tt>and</tt>' Instruction</a></li>
152 <li><a href="#i_or">'<tt>or</tt>' Instruction</a></li>
153 <li><a href="#i_xor">'<tt>xor</tt>' Instruction</a></li>
156 <li><a href="#vectorops">Vector Operations</a>
158 <li><a href="#i_extractelement">'<tt>extractelement</tt>' Instruction</a></li>
159 <li><a href="#i_insertelement">'<tt>insertelement</tt>' Instruction</a></li>
160 <li><a href="#i_shufflevector">'<tt>shufflevector</tt>' Instruction</a></li>
163 <li><a href="#aggregateops">Aggregate Operations</a>
165 <li><a href="#i_extractvalue">'<tt>extractvalue</tt>' Instruction</a></li>
166 <li><a href="#i_insertvalue">'<tt>insertvalue</tt>' Instruction</a></li>
169 <li><a href="#memoryops">Memory Access and Addressing Operations</a>
171 <li><a href="#i_alloca">'<tt>alloca</tt>' Instruction</a></li>
172 <li><a href="#i_load">'<tt>load</tt>' Instruction</a></li>
173 <li><a href="#i_store">'<tt>store</tt>' Instruction</a></li>
174 <li><a href="#i_getelementptr">'<tt>getelementptr</tt>' Instruction</a></li>
177 <li><a href="#convertops">Conversion Operations</a>
179 <li><a href="#i_trunc">'<tt>trunc .. to</tt>' Instruction</a></li>
180 <li><a href="#i_zext">'<tt>zext .. to</tt>' Instruction</a></li>
181 <li><a href="#i_sext">'<tt>sext .. to</tt>' Instruction</a></li>
182 <li><a href="#i_fptrunc">'<tt>fptrunc .. to</tt>' Instruction</a></li>
183 <li><a href="#i_fpext">'<tt>fpext .. to</tt>' Instruction</a></li>
184 <li><a href="#i_fptoui">'<tt>fptoui .. to</tt>' Instruction</a></li>
185 <li><a href="#i_fptosi">'<tt>fptosi .. to</tt>' Instruction</a></li>
186 <li><a href="#i_uitofp">'<tt>uitofp .. to</tt>' Instruction</a></li>
187 <li><a href="#i_sitofp">'<tt>sitofp .. to</tt>' Instruction</a></li>
188 <li><a href="#i_ptrtoint">'<tt>ptrtoint .. to</tt>' Instruction</a></li>
189 <li><a href="#i_inttoptr">'<tt>inttoptr .. to</tt>' Instruction</a></li>
190 <li><a href="#i_bitcast">'<tt>bitcast .. to</tt>' Instruction</a></li>
193 <li><a href="#otherops">Other Operations</a>
195 <li><a href="#i_icmp">'<tt>icmp</tt>' Instruction</a></li>
196 <li><a href="#i_fcmp">'<tt>fcmp</tt>' Instruction</a></li>
197 <li><a href="#i_phi">'<tt>phi</tt>' Instruction</a></li>
198 <li><a href="#i_select">'<tt>select</tt>' Instruction</a></li>
199 <li><a href="#i_call">'<tt>call</tt>' Instruction</a></li>
200 <li><a href="#i_va_arg">'<tt>va_arg</tt>' Instruction</a></li>
205 <li><a href="#intrinsics">Intrinsic Functions</a>
207 <li><a href="#int_varargs">Variable Argument Handling Intrinsics</a>
209 <li><a href="#int_va_start">'<tt>llvm.va_start</tt>' Intrinsic</a></li>
210 <li><a href="#int_va_end">'<tt>llvm.va_end</tt>' Intrinsic</a></li>
211 <li><a href="#int_va_copy">'<tt>llvm.va_copy</tt>' Intrinsic</a></li>
214 <li><a href="#int_gc">Accurate Garbage Collection Intrinsics</a>
216 <li><a href="#int_gcroot">'<tt>llvm.gcroot</tt>' Intrinsic</a></li>
217 <li><a href="#int_gcread">'<tt>llvm.gcread</tt>' Intrinsic</a></li>
218 <li><a href="#int_gcwrite">'<tt>llvm.gcwrite</tt>' Intrinsic</a></li>
221 <li><a href="#int_codegen">Code Generator Intrinsics</a>
223 <li><a href="#int_returnaddress">'<tt>llvm.returnaddress</tt>' Intrinsic</a></li>
224 <li><a href="#int_frameaddress">'<tt>llvm.frameaddress</tt>' Intrinsic</a></li>
225 <li><a href="#int_stacksave">'<tt>llvm.stacksave</tt>' Intrinsic</a></li>
226 <li><a href="#int_stackrestore">'<tt>llvm.stackrestore</tt>' Intrinsic</a></li>
227 <li><a href="#int_prefetch">'<tt>llvm.prefetch</tt>' Intrinsic</a></li>
228 <li><a href="#int_pcmarker">'<tt>llvm.pcmarker</tt>' Intrinsic</a></li>
229 <li><a href="#int_readcyclecounter">'<tt>llvm.readcyclecounter</tt>' Intrinsic</a></li>
232 <li><a href="#int_libc">Standard C Library Intrinsics</a>
234 <li><a href="#int_memcpy">'<tt>llvm.memcpy.*</tt>' Intrinsic</a></li>
235 <li><a href="#int_memmove">'<tt>llvm.memmove.*</tt>' Intrinsic</a></li>
236 <li><a href="#int_memset">'<tt>llvm.memset.*</tt>' Intrinsic</a></li>
237 <li><a href="#int_sqrt">'<tt>llvm.sqrt.*</tt>' Intrinsic</a></li>
238 <li><a href="#int_powi">'<tt>llvm.powi.*</tt>' Intrinsic</a></li>
239 <li><a href="#int_sin">'<tt>llvm.sin.*</tt>' Intrinsic</a></li>
240 <li><a href="#int_cos">'<tt>llvm.cos.*</tt>' Intrinsic</a></li>
241 <li><a href="#int_pow">'<tt>llvm.pow.*</tt>' Intrinsic</a></li>
244 <li><a href="#int_manip">Bit Manipulation Intrinsics</a>
246 <li><a href="#int_bswap">'<tt>llvm.bswap.*</tt>' Intrinsics</a></li>
247 <li><a href="#int_ctpop">'<tt>llvm.ctpop.*</tt>' Intrinsic </a></li>
248 <li><a href="#int_ctlz">'<tt>llvm.ctlz.*</tt>' Intrinsic </a></li>
249 <li><a href="#int_cttz">'<tt>llvm.cttz.*</tt>' Intrinsic </a></li>
252 <li><a href="#int_overflow">Arithmetic with Overflow Intrinsics</a>
254 <li><a href="#int_sadd_overflow">'<tt>llvm.sadd.with.overflow.*</tt> Intrinsics</a></li>
255 <li><a href="#int_uadd_overflow">'<tt>llvm.uadd.with.overflow.*</tt> Intrinsics</a></li>
256 <li><a href="#int_ssub_overflow">'<tt>llvm.ssub.with.overflow.*</tt> Intrinsics</a></li>
257 <li><a href="#int_usub_overflow">'<tt>llvm.usub.with.overflow.*</tt> Intrinsics</a></li>
258 <li><a href="#int_smul_overflow">'<tt>llvm.smul.with.overflow.*</tt> Intrinsics</a></li>
259 <li><a href="#int_umul_overflow">'<tt>llvm.umul.with.overflow.*</tt> Intrinsics</a></li>
262 <li><a href="#int_fp16">Half Precision Floating Point Intrinsics</a>
264 <li><a href="#int_convert_to_fp16">'<tt>llvm.convert.to.fp16</tt>' Intrinsic</a></li>
265 <li><a href="#int_convert_from_fp16">'<tt>llvm.convert.from.fp16</tt>' Intrinsic</a></li>
268 <li><a href="#int_debugger">Debugger intrinsics</a></li>
269 <li><a href="#int_eh">Exception Handling intrinsics</a></li>
270 <li><a href="#int_trampoline">Trampoline Intrinsic</a>
272 <li><a href="#int_it">'<tt>llvm.init.trampoline</tt>' Intrinsic</a></li>
275 <li><a href="#int_atomics">Atomic intrinsics</a>
277 <li><a href="#int_memory_barrier"><tt>llvm.memory_barrier</tt></a></li>
278 <li><a href="#int_atomic_cmp_swap"><tt>llvm.atomic.cmp.swap</tt></a></li>
279 <li><a href="#int_atomic_swap"><tt>llvm.atomic.swap</tt></a></li>
280 <li><a href="#int_atomic_load_add"><tt>llvm.atomic.load.add</tt></a></li>
281 <li><a href="#int_atomic_load_sub"><tt>llvm.atomic.load.sub</tt></a></li>
282 <li><a href="#int_atomic_load_and"><tt>llvm.atomic.load.and</tt></a></li>
283 <li><a href="#int_atomic_load_nand"><tt>llvm.atomic.load.nand</tt></a></li>
284 <li><a href="#int_atomic_load_or"><tt>llvm.atomic.load.or</tt></a></li>
285 <li><a href="#int_atomic_load_xor"><tt>llvm.atomic.load.xor</tt></a></li>
286 <li><a href="#int_atomic_load_max"><tt>llvm.atomic.load.max</tt></a></li>
287 <li><a href="#int_atomic_load_min"><tt>llvm.atomic.load.min</tt></a></li>
288 <li><a href="#int_atomic_load_umax"><tt>llvm.atomic.load.umax</tt></a></li>
289 <li><a href="#int_atomic_load_umin"><tt>llvm.atomic.load.umin</tt></a></li>
292 <li><a href="#int_memorymarkers">Memory Use Markers</a>
294 <li><a href="#int_lifetime_start"><tt>llvm.lifetime.start</tt></a></li>
295 <li><a href="#int_lifetime_end"><tt>llvm.lifetime.end</tt></a></li>
296 <li><a href="#int_invariant_start"><tt>llvm.invariant.start</tt></a></li>
297 <li><a href="#int_invariant_end"><tt>llvm.invariant.end</tt></a></li>
300 <li><a href="#int_general">General intrinsics</a>
302 <li><a href="#int_var_annotation">
303 '<tt>llvm.var.annotation</tt>' Intrinsic</a></li>
304 <li><a href="#int_annotation">
305 '<tt>llvm.annotation.*</tt>' Intrinsic</a></li>
306 <li><a href="#int_trap">
307 '<tt>llvm.trap</tt>' Intrinsic</a></li>
308 <li><a href="#int_stackprotector">
309 '<tt>llvm.stackprotector</tt>' Intrinsic</a></li>
310 <li><a href="#int_objectsize">
311 '<tt>llvm.objectsize</tt>' Intrinsic</a></li>
318 <div class="doc_author">
319 <p>Written by <a href="mailto:sabre@nondot.org">Chris Lattner</a>
320 and <a href="mailto:vadve@cs.uiuc.edu">Vikram Adve</a></p>
323 <!-- *********************************************************************** -->
324 <div class="doc_section"> <a name="abstract">Abstract </a></div>
325 <!-- *********************************************************************** -->
327 <div class="doc_text">
329 <p>This document is a reference manual for the LLVM assembly language. LLVM is
330 a Static Single Assignment (SSA) based representation that provides type
331 safety, low-level operations, flexibility, and the capability of representing
332 'all' high-level languages cleanly. It is the common code representation
333 used throughout all phases of the LLVM compilation strategy.</p>
337 <!-- *********************************************************************** -->
338 <div class="doc_section"> <a name="introduction">Introduction</a> </div>
339 <!-- *********************************************************************** -->
341 <div class="doc_text">
343 <p>The LLVM code representation is designed to be used in three different forms:
344 as an in-memory compiler IR, as an on-disk bitcode representation (suitable
345 for fast loading by a Just-In-Time compiler), and as a human readable
346 assembly language representation. This allows LLVM to provide a powerful
347 intermediate representation for efficient compiler transformations and
348 analysis, while providing a natural means to debug and visualize the
349 transformations. The three different forms of LLVM are all equivalent. This
350 document describes the human readable representation and notation.</p>
352 <p>The LLVM representation aims to be light-weight and low-level while being
353 expressive, typed, and extensible at the same time. It aims to be a
354 "universal IR" of sorts, by being at a low enough level that high-level ideas
355 may be cleanly mapped to it (similar to how microprocessors are "universal
356 IR's", allowing many source languages to be mapped to them). By providing
357 type information, LLVM can be used as the target of optimizations: for
358 example, through pointer analysis, it can be proven that a C automatic
359 variable is never accessed outside of the current function, allowing it to
360 be promoted to a simple SSA value instead of a memory location.</p>
364 <!-- _______________________________________________________________________ -->
365 <div class="doc_subsubsection"> <a name="wellformed">Well-Formedness</a> </div>
367 <div class="doc_text">
369 <p>It is important to note that this document describes 'well formed' LLVM
370 assembly language. There is a difference between what the parser accepts and
371 what is considered 'well formed'. For example, the following instruction is
372 syntactically okay, but not well formed:</p>
374 <pre class="doc_code">
375 %x = <a href="#i_add">add</a> i32 1, %x
378 <p>because the definition of <tt>%x</tt> does not dominate all of its uses. The
379 LLVM infrastructure provides a verification pass that may be used to verify
380 that an LLVM module is well formed. This pass is automatically run by the
381 parser after parsing input assembly and by the optimizer before it outputs
382 bitcode. The violations pointed out by the verifier pass indicate bugs in
383 transformation passes or input to the parser.</p>
387 <!-- Describe the typesetting conventions here. -->
389 <!-- *********************************************************************** -->
390 <div class="doc_section"> <a name="identifiers">Identifiers</a> </div>
391 <!-- *********************************************************************** -->
393 <div class="doc_text">
395 <p>LLVM identifiers come in two basic types: global and local. Global
396 identifiers (functions, global variables) begin with the <tt>'@'</tt>
397 character. Local identifiers (register names, types) begin with
398 the <tt>'%'</tt> character. Additionally, there are three different formats
399 for identifiers, for different purposes:</p>
402 <li>Named values are represented as a string of characters with their prefix.
403 For example, <tt>%foo</tt>, <tt>@DivisionByZero</tt>,
404 <tt>%a.really.long.identifier</tt>. The actual regular expression used is
405 '<tt>[%@][a-zA-Z$._][a-zA-Z$._0-9]*</tt>'. Identifiers which require
406 other characters in their names can be surrounded with quotes. Special
407 characters may be escaped using <tt>"\xx"</tt> where <tt>xx</tt> is the
408 ASCII code for the character in hexadecimal. In this way, any character
409 can be used in a name value, even quotes themselves.</li>
411 <li>Unnamed values are represented as an unsigned numeric value with their
412 prefix. For example, <tt>%12</tt>, <tt>@2</tt>, <tt>%44</tt>.</li>
414 <li>Constants, which are described in a <a href="#constants">section about
415 constants</a>, below.</li>
418 <p>LLVM requires that values start with a prefix for two reasons: Compilers
419 don't need to worry about name clashes with reserved words, and the set of
420 reserved words may be expanded in the future without penalty. Additionally,
421 unnamed identifiers allow a compiler to quickly come up with a temporary
422 variable without having to avoid symbol table conflicts.</p>
424 <p>Reserved words in LLVM are very similar to reserved words in other
425 languages. There are keywords for different opcodes
426 ('<tt><a href="#i_add">add</a></tt>',
427 '<tt><a href="#i_bitcast">bitcast</a></tt>',
428 '<tt><a href="#i_ret">ret</a></tt>', etc...), for primitive type names
429 ('<tt><a href="#t_void">void</a></tt>',
430 '<tt><a href="#t_primitive">i32</a></tt>', etc...), and others. These
431 reserved words cannot conflict with variable names, because none of them
432 start with a prefix character (<tt>'%'</tt> or <tt>'@'</tt>).</p>
434 <p>Here is an example of LLVM code to multiply the integer variable
435 '<tt>%X</tt>' by 8:</p>
439 <pre class="doc_code">
440 %result = <a href="#i_mul">mul</a> i32 %X, 8
443 <p>After strength reduction:</p>
445 <pre class="doc_code">
446 %result = <a href="#i_shl">shl</a> i32 %X, i8 3
449 <p>And the hard way:</p>
451 <pre class="doc_code">
452 %0 = <a href="#i_add">add</a> i32 %X, %X <i>; yields {i32}:%0</i>
453 %1 = <a href="#i_add">add</a> i32 %0, %0 <i>; yields {i32}:%1</i>
454 %result = <a href="#i_add">add</a> i32 %1, %1
457 <p>This last way of multiplying <tt>%X</tt> by 8 illustrates several important
458 lexical features of LLVM:</p>
461 <li>Comments are delimited with a '<tt>;</tt>' and go until the end of
464 <li>Unnamed temporaries are created when the result of a computation is not
465 assigned to a named value.</li>
467 <li>Unnamed temporaries are numbered sequentially</li>
470 <p>It also shows a convention that we follow in this document. When
471 demonstrating instructions, we will follow an instruction with a comment that
472 defines the type and name of value produced. Comments are shown in italic
477 <!-- *********************************************************************** -->
478 <div class="doc_section"> <a name="highlevel">High Level Structure</a> </div>
479 <!-- *********************************************************************** -->
481 <!-- ======================================================================= -->
482 <div class="doc_subsection"> <a name="modulestructure">Module Structure</a>
485 <div class="doc_text">
487 <p>LLVM programs are composed of "Module"s, each of which is a translation unit
488 of the input programs. Each module consists of functions, global variables,
489 and symbol table entries. Modules may be combined together with the LLVM
490 linker, which merges function (and global variable) definitions, resolves
491 forward declarations, and merges symbol table entries. Here is an example of
492 the "hello world" module:</p>
494 <pre class="doc_code">
495 <i>; Declare the string constant as a global constant.</i>
496 <a href="#identifiers">@.LC0</a> = <a href="#linkage_internal">internal</a> <a href="#globalvars">constant</a> <a href="#t_array">[13 x i8]</a> c"hello world\0A\00" <i>; [13 x i8]*</i>
498 <i>; External declaration of the puts function</i>
499 <a href="#functionstructure">declare</a> i32 @puts(i8*) <i>; i32 (i8*)* </i>
501 <i>; Definition of main function</i>
502 define i32 @main() { <i>; i32()* </i>
503 <i>; Convert [13 x i8]* to i8 *...</i>
504 %cast210 = <a href="#i_getelementptr">getelementptr</a> [13 x i8]* @.LC0, i64 0, i64 0 <i>; i8*</i>
506 <i>; Call puts function to write out the string to stdout.</i>
507 <a href="#i_call">call</a> i32 @puts(i8* %cast210) <i>; i32</i>
508 <a href="#i_ret">ret</a> i32 0
511 <i>; Named metadata</i>
512 !1 = metadata !{i32 41}
516 <p>This example is made up of a <a href="#globalvars">global variable</a> named
517 "<tt>.LC0</tt>", an external declaration of the "<tt>puts</tt>" function,
518 a <a href="#functionstructure">function definition</a> for
519 "<tt>main</tt>" and <a href="#namedmetadatastructure">named metadata</a>
522 <p>In general, a module is made up of a list of global values, where both
523 functions and global variables are global values. Global values are
524 represented by a pointer to a memory location (in this case, a pointer to an
525 array of char, and a pointer to a function), and have one of the
526 following <a href="#linkage">linkage types</a>.</p>
530 <!-- ======================================================================= -->
531 <div class="doc_subsection">
532 <a name="linkage">Linkage Types</a>
535 <div class="doc_text">
537 <p>All Global Variables and Functions have one of the following types of
541 <dt><tt><b><a name="linkage_private">private</a></b></tt></dt>
542 <dd>Global values with "<tt>private</tt>" linkage are only directly accessible
543 by objects in the current module. In particular, linking code into a
544 module with an private global value may cause the private to be renamed as
545 necessary to avoid collisions. Because the symbol is private to the
546 module, all references can be updated. This doesn't show up in any symbol
547 table in the object file.</dd>
549 <dt><tt><b><a name="linkage_linker_private">linker_private</a></b></tt></dt>
550 <dd>Similar to <tt>private</tt>, but the symbol is passed through the
551 assembler and evaluated by the linker. Unlike normal strong symbols, they
552 are removed by the linker from the final linked image (executable or
553 dynamic library).</dd>
555 <dt><tt><b><a name="linkage_linker_private_weak">linker_private_weak</a></b></tt></dt>
556 <dd>Similar to "<tt>linker_private</tt>", but the symbol is weak. Note that
557 <tt>linker_private_weak</tt> symbols are subject to coalescing by the
558 linker. The symbols are removed by the linker from the final linked image
559 (executable or dynamic library).</dd>
561 <dt><tt><b><a name="linkage_linker_private_weak_def_auto">linker_private_weak_def_auto</a></b></tt></dt>
562 <dd>Similar to "<tt>linker_private_weak</tt>", but it's known that the address
563 of the object is not taken. For instance, functions that had an inline
564 definition, but the compiler decided not to inline it. Note,
565 unlike <tt>linker_private</tt> and <tt>linker_private_weak</tt>,
566 <tt>linker_private_weak_def_auto</tt> may have only <tt>default</tt>
567 visibility. The symbols are removed by the linker from the final linked
568 image (executable or dynamic library).</dd>
570 <dt><tt><b><a name="linkage_internal">internal</a></b></tt></dt>
571 <dd>Similar to private, but the value shows as a local symbol
572 (<tt>STB_LOCAL</tt> in the case of ELF) in the object file. This
573 corresponds to the notion of the '<tt>static</tt>' keyword in C.</dd>
575 <dt><tt><b><a name="linkage_available_externally">available_externally</a></b></tt></dt>
576 <dd>Globals with "<tt>available_externally</tt>" linkage are never emitted
577 into the object file corresponding to the LLVM module. They exist to
578 allow inlining and other optimizations to take place given knowledge of
579 the definition of the global, which is known to be somewhere outside the
580 module. Globals with <tt>available_externally</tt> linkage are allowed to
581 be discarded at will, and are otherwise the same as <tt>linkonce_odr</tt>.
582 This linkage type is only allowed on definitions, not declarations.</dd>
584 <dt><tt><b><a name="linkage_linkonce">linkonce</a></b></tt></dt>
585 <dd>Globals with "<tt>linkonce</tt>" linkage are merged with other globals of
586 the same name when linkage occurs. This can be used to implement
587 some forms of inline functions, templates, or other code which must be
588 generated in each translation unit that uses it, but where the body may
589 be overridden with a more definitive definition later. Unreferenced
590 <tt>linkonce</tt> globals are allowed to be discarded. Note that
591 <tt>linkonce</tt> linkage does not actually allow the optimizer to
592 inline the body of this function into callers because it doesn't know if
593 this definition of the function is the definitive definition within the
594 program or whether it will be overridden by a stronger definition.
595 To enable inlining and other optimizations, use "<tt>linkonce_odr</tt>"
598 <dt><tt><b><a name="linkage_weak">weak</a></b></tt></dt>
599 <dd>"<tt>weak</tt>" linkage has the same merging semantics as
600 <tt>linkonce</tt> linkage, except that unreferenced globals with
601 <tt>weak</tt> linkage may not be discarded. This is used for globals that
602 are declared "weak" in C source code.</dd>
604 <dt><tt><b><a name="linkage_common">common</a></b></tt></dt>
605 <dd>"<tt>common</tt>" linkage is most similar to "<tt>weak</tt>" linkage, but
606 they are used for tentative definitions in C, such as "<tt>int X;</tt>" at
608 Symbols with "<tt>common</tt>" linkage are merged in the same way as
609 <tt>weak symbols</tt>, and they may not be deleted if unreferenced.
610 <tt>common</tt> symbols may not have an explicit section,
611 must have a zero initializer, and may not be marked '<a
612 href="#globalvars"><tt>constant</tt></a>'. Functions and aliases may not
613 have common linkage.</dd>
616 <dt><tt><b><a name="linkage_appending">appending</a></b></tt></dt>
617 <dd>"<tt>appending</tt>" linkage may only be applied to global variables of
618 pointer to array type. When two global variables with appending linkage
619 are linked together, the two global arrays are appended together. This is
620 the LLVM, typesafe, equivalent of having the system linker append together
621 "sections" with identical names when .o files are linked.</dd>
623 <dt><tt><b><a name="linkage_externweak">extern_weak</a></b></tt></dt>
624 <dd>The semantics of this linkage follow the ELF object file model: the symbol
625 is weak until linked, if not linked, the symbol becomes null instead of
626 being an undefined reference.</dd>
628 <dt><tt><b><a name="linkage_linkonce_odr">linkonce_odr</a></b></tt></dt>
629 <dt><tt><b><a name="linkage_weak_odr">weak_odr</a></b></tt></dt>
630 <dd>Some languages allow differing globals to be merged, such as two functions
631 with different semantics. Other languages, such as <tt>C++</tt>, ensure
632 that only equivalent globals are ever merged (the "one definition rule"
633 — "ODR"). Such languages can use the <tt>linkonce_odr</tt>
634 and <tt>weak_odr</tt> linkage types to indicate that the global will only
635 be merged with equivalent globals. These linkage types are otherwise the
636 same as their non-<tt>odr</tt> versions.</dd>
638 <dt><tt><b><a name="linkage_external">externally visible</a></b></tt>:</dt>
639 <dd>If none of the above identifiers are used, the global is externally
640 visible, meaning that it participates in linkage and can be used to
641 resolve external symbol references.</dd>
644 <p>The next two types of linkage are targeted for Microsoft Windows platform
645 only. They are designed to support importing (exporting) symbols from (to)
646 DLLs (Dynamic Link Libraries).</p>
649 <dt><tt><b><a name="linkage_dllimport">dllimport</a></b></tt></dt>
650 <dd>"<tt>dllimport</tt>" linkage causes the compiler to reference a function
651 or variable via a global pointer to a pointer that is set up by the DLL
652 exporting the symbol. On Microsoft Windows targets, the pointer name is
653 formed by combining <code>__imp_</code> and the function or variable
656 <dt><tt><b><a name="linkage_dllexport">dllexport</a></b></tt></dt>
657 <dd>"<tt>dllexport</tt>" linkage causes the compiler to provide a global
658 pointer to a pointer in a DLL, so that it can be referenced with the
659 <tt>dllimport</tt> attribute. On Microsoft Windows targets, the pointer
660 name is formed by combining <code>__imp_</code> and the function or
664 <p>For example, since the "<tt>.LC0</tt>" variable is defined to be internal, if
665 another module defined a "<tt>.LC0</tt>" variable and was linked with this
666 one, one of the two would be renamed, preventing a collision. Since
667 "<tt>main</tt>" and "<tt>puts</tt>" are external (i.e., lacking any linkage
668 declarations), they are accessible outside of the current module.</p>
670 <p>It is illegal for a function <i>declaration</i> to have any linkage type
671 other than "externally visible", <tt>dllimport</tt>
672 or <tt>extern_weak</tt>.</p>
674 <p>Aliases can have only <tt>external</tt>, <tt>internal</tt>, <tt>weak</tt>
675 or <tt>weak_odr</tt> linkages.</p>
679 <!-- ======================================================================= -->
680 <div class="doc_subsection">
681 <a name="callingconv">Calling Conventions</a>
684 <div class="doc_text">
686 <p>LLVM <a href="#functionstructure">functions</a>, <a href="#i_call">calls</a>
687 and <a href="#i_invoke">invokes</a> can all have an optional calling
688 convention specified for the call. The calling convention of any pair of
689 dynamic caller/callee must match, or the behavior of the program is
690 undefined. The following calling conventions are supported by LLVM, and more
691 may be added in the future:</p>
694 <dt><b>"<tt>ccc</tt>" - The C calling convention</b>:</dt>
695 <dd>This calling convention (the default if no other calling convention is
696 specified) matches the target C calling conventions. This calling
697 convention supports varargs function calls and tolerates some mismatch in
698 the declared prototype and implemented declaration of the function (as
701 <dt><b>"<tt>fastcc</tt>" - The fast calling convention</b>:</dt>
702 <dd>This calling convention attempts to make calls as fast as possible
703 (e.g. by passing things in registers). This calling convention allows the
704 target to use whatever tricks it wants to produce fast code for the
705 target, without having to conform to an externally specified ABI
706 (Application Binary Interface).
707 <a href="CodeGenerator.html#tailcallopt">Tail calls can only be optimized
708 when this or the GHC convention is used.</a> This calling convention
709 does not support varargs and requires the prototype of all callees to
710 exactly match the prototype of the function definition.</dd>
712 <dt><b>"<tt>coldcc</tt>" - The cold calling convention</b>:</dt>
713 <dd>This calling convention attempts to make code in the caller as efficient
714 as possible under the assumption that the call is not commonly executed.
715 As such, these calls often preserve all registers so that the call does
716 not break any live ranges in the caller side. This calling convention
717 does not support varargs and requires the prototype of all callees to
718 exactly match the prototype of the function definition.</dd>
720 <dt><b>"<tt>cc <em>10</em></tt>" - GHC convention</b>:</dt>
721 <dd>This calling convention has been implemented specifically for use by the
722 <a href="http://www.haskell.org/ghc">Glasgow Haskell Compiler (GHC)</a>.
723 It passes everything in registers, going to extremes to achieve this by
724 disabling callee save registers. This calling convention should not be
725 used lightly but only for specific situations such as an alternative to
726 the <em>register pinning</em> performance technique often used when
727 implementing functional programming languages.At the moment only X86
728 supports this convention and it has the following limitations:
730 <li>On <em>X86-32</em> only supports up to 4 bit type parameters. No
731 floating point types are supported.</li>
732 <li>On <em>X86-64</em> only supports up to 10 bit type parameters and
733 6 floating point parameters.</li>
735 This calling convention supports
736 <a href="CodeGenerator.html#tailcallopt">tail call optimization</a> but
737 requires both the caller and callee are using it.
740 <dt><b>"<tt>cc <<em>n</em>></tt>" - Numbered convention</b>:</dt>
741 <dd>Any calling convention may be specified by number, allowing
742 target-specific calling conventions to be used. Target specific calling
743 conventions start at 64.</dd>
746 <p>More calling conventions can be added/defined on an as-needed basis, to
747 support Pascal conventions or any other well-known target-independent
752 <!-- ======================================================================= -->
753 <div class="doc_subsection">
754 <a name="visibility">Visibility Styles</a>
757 <div class="doc_text">
759 <p>All Global Variables and Functions have one of the following visibility
763 <dt><b>"<tt>default</tt>" - Default style</b>:</dt>
764 <dd>On targets that use the ELF object file format, default visibility means
765 that the declaration is visible to other modules and, in shared libraries,
766 means that the declared entity may be overridden. On Darwin, default
767 visibility means that the declaration is visible to other modules. Default
768 visibility corresponds to "external linkage" in the language.</dd>
770 <dt><b>"<tt>hidden</tt>" - Hidden style</b>:</dt>
771 <dd>Two declarations of an object with hidden visibility refer to the same
772 object if they are in the same shared object. Usually, hidden visibility
773 indicates that the symbol will not be placed into the dynamic symbol
774 table, so no other module (executable or shared library) can reference it
777 <dt><b>"<tt>protected</tt>" - Protected style</b>:</dt>
778 <dd>On ELF, protected visibility indicates that the symbol will be placed in
779 the dynamic symbol table, but that references within the defining module
780 will bind to the local symbol. That is, the symbol cannot be overridden by
786 <!-- ======================================================================= -->
787 <div class="doc_subsection">
788 <a name="namedtypes">Named Types</a>
791 <div class="doc_text">
793 <p>LLVM IR allows you to specify name aliases for certain types. This can make
794 it easier to read the IR and make the IR more condensed (particularly when
795 recursive types are involved). An example of a name specification is:</p>
797 <pre class="doc_code">
798 %mytype = type { %mytype*, i32 }
801 <p>You may give a name to any <a href="#typesystem">type</a> except
802 "<a href="#t_void">void</a>". Type name aliases may be used anywhere a type
803 is expected with the syntax "%mytype".</p>
805 <p>Note that type names are aliases for the structural type that they indicate,
806 and that you can therefore specify multiple names for the same type. This
807 often leads to confusing behavior when dumping out a .ll file. Since LLVM IR
808 uses structural typing, the name is not part of the type. When printing out
809 LLVM IR, the printer will pick <em>one name</em> to render all types of a
810 particular shape. This means that if you have code where two different
811 source types end up having the same LLVM type, that the dumper will sometimes
812 print the "wrong" or unexpected type. This is an important design point and
813 isn't going to change.</p>
817 <!-- ======================================================================= -->
818 <div class="doc_subsection">
819 <a name="globalvars">Global Variables</a>
822 <div class="doc_text">
824 <p>Global variables define regions of memory allocated at compilation time
825 instead of run-time. Global variables may optionally be initialized, may
826 have an explicit section to be placed in, and may have an optional explicit
827 alignment specified. A variable may be defined as "thread_local", which
828 means that it will not be shared by threads (each thread will have a
829 separated copy of the variable). A variable may be defined as a global
830 "constant," which indicates that the contents of the variable
831 will <b>never</b> be modified (enabling better optimization, allowing the
832 global data to be placed in the read-only section of an executable, etc).
833 Note that variables that need runtime initialization cannot be marked
834 "constant" as there is a store to the variable.</p>
836 <p>LLVM explicitly allows <em>declarations</em> of global variables to be marked
837 constant, even if the final definition of the global is not. This capability
838 can be used to enable slightly better optimization of the program, but
839 requires the language definition to guarantee that optimizations based on the
840 'constantness' are valid for the translation units that do not include the
843 <p>As SSA values, global variables define pointer values that are in scope
844 (i.e. they dominate) all basic blocks in the program. Global variables
845 always define a pointer to their "content" type because they describe a
846 region of memory, and all memory objects in LLVM are accessed through
849 <p>A global variable may be declared to reside in a target-specific numbered
850 address space. For targets that support them, address spaces may affect how
851 optimizations are performed and/or what target instructions are used to
852 access the variable. The default address space is zero. The address space
853 qualifier must precede any other attributes.</p>
855 <p>LLVM allows an explicit section to be specified for globals. If the target
856 supports it, it will emit globals to the section specified.</p>
858 <p>An explicit alignment may be specified for a global, which must be a power
859 of 2. If not present, or if the alignment is set to zero, the alignment of
860 the global is set by the target to whatever it feels convenient. If an
861 explicit alignment is specified, the global is forced to have exactly that
862 alignment. Targets and optimizers are not allowed to over-align the global
863 if the global has an assigned section. In this case, the extra alignment
864 could be observable: for example, code could assume that the globals are
865 densely packed in their section and try to iterate over them as an array,
866 alignment padding would break this iteration.</p>
868 <p>For example, the following defines a global in a numbered address space with
869 an initializer, section, and alignment:</p>
871 <pre class="doc_code">
872 @G = addrspace(5) constant float 1.0, section "foo", align 4
878 <!-- ======================================================================= -->
879 <div class="doc_subsection">
880 <a name="functionstructure">Functions</a>
883 <div class="doc_text">
885 <p>LLVM function definitions consist of the "<tt>define</tt>" keyword, an
886 optional <a href="#linkage">linkage type</a>, an optional
887 <a href="#visibility">visibility style</a>, an optional
888 <a href="#callingconv">calling convention</a>, a return type, an optional
889 <a href="#paramattrs">parameter attribute</a> for the return type, a function
890 name, a (possibly empty) argument list (each with optional
891 <a href="#paramattrs">parameter attributes</a>), optional
892 <a href="#fnattrs">function attributes</a>, an optional section, an optional
893 alignment, an optional <a href="#gc">garbage collector name</a>, an opening
894 curly brace, a list of basic blocks, and a closing curly brace.</p>
896 <p>LLVM function declarations consist of the "<tt>declare</tt>" keyword, an
897 optional <a href="#linkage">linkage type</a>, an optional
898 <a href="#visibility">visibility style</a>, an optional
899 <a href="#callingconv">calling convention</a>, a return type, an optional
900 <a href="#paramattrs">parameter attribute</a> for the return type, a function
901 name, a possibly empty list of arguments, an optional alignment, and an
902 optional <a href="#gc">garbage collector name</a>.</p>
904 <p>A function definition contains a list of basic blocks, forming the CFG
905 (Control Flow Graph) for the function. Each basic block may optionally start
906 with a label (giving the basic block a symbol table entry), contains a list
907 of instructions, and ends with a <a href="#terminators">terminator</a>
908 instruction (such as a branch or function return).</p>
910 <p>The first basic block in a function is special in two ways: it is immediately
911 executed on entrance to the function, and it is not allowed to have
912 predecessor basic blocks (i.e. there can not be any branches to the entry
913 block of a function). Because the block can have no predecessors, it also
914 cannot have any <a href="#i_phi">PHI nodes</a>.</p>
916 <p>LLVM allows an explicit section to be specified for functions. If the target
917 supports it, it will emit functions to the section specified.</p>
919 <p>An explicit alignment may be specified for a function. If not present, or if
920 the alignment is set to zero, the alignment of the function is set by the
921 target to whatever it feels convenient. If an explicit alignment is
922 specified, the function is forced to have at least that much alignment. All
923 alignments must be a power of 2.</p>
926 <pre class="doc_code">
927 define [<a href="#linkage">linkage</a>] [<a href="#visibility">visibility</a>]
928 [<a href="#callingconv">cconv</a>] [<a href="#paramattrs">ret attrs</a>]
929 <ResultType> @<FunctionName> ([argument list])
930 [<a href="#fnattrs">fn Attrs</a>] [section "name"] [align N]
931 [<a href="#gc">gc</a>] { ... }
936 <!-- ======================================================================= -->
937 <div class="doc_subsection">
938 <a name="aliasstructure">Aliases</a>
941 <div class="doc_text">
943 <p>Aliases act as "second name" for the aliasee value (which can be either
944 function, global variable, another alias or bitcast of global value). Aliases
945 may have an optional <a href="#linkage">linkage type</a>, and an
946 optional <a href="#visibility">visibility style</a>.</p>
949 <pre class="doc_code">
950 @<Name> = alias [Linkage] [Visibility] <AliaseeTy> @<Aliasee>
955 <!-- ======================================================================= -->
956 <div class="doc_subsection">
957 <a name="namedmetadatastructure">Named Metadata</a>
960 <div class="doc_text">
962 <p>Named metadata is a collection of metadata. <a href="#metadata">Metadata
963 nodes</a> (but not metadata strings) are the only valid operands for
964 a named metadata.</p>
967 <pre class="doc_code">
968 ; Some unnamed metadata nodes, which are referenced by the named metadata.
969 !0 = metadata !{metadata !"zero"}
970 !1 = metadata !{metadata !"one"}
971 !2 = metadata !{metadata !"two"}
973 !name = !{!0, !1, !2}
978 <!-- ======================================================================= -->
979 <div class="doc_subsection"><a name="paramattrs">Parameter Attributes</a></div>
981 <div class="doc_text">
983 <p>The return type and each parameter of a function type may have a set of
984 <i>parameter attributes</i> associated with them. Parameter attributes are
985 used to communicate additional information about the result or parameters of
986 a function. Parameter attributes are considered to be part of the function,
987 not of the function type, so functions with different parameter attributes
988 can have the same function type.</p>
990 <p>Parameter attributes are simple keywords that follow the type specified. If
991 multiple parameter attributes are needed, they are space separated. For
994 <pre class="doc_code">
995 declare i32 @printf(i8* noalias nocapture, ...)
996 declare i32 @atoi(i8 zeroext)
997 declare signext i8 @returns_signed_char()
1000 <p>Note that any attributes for the function result (<tt>nounwind</tt>,
1001 <tt>readonly</tt>) come immediately after the argument list.</p>
1003 <p>Currently, only the following parameter attributes are defined:</p>
1006 <dt><tt><b>zeroext</b></tt></dt>
1007 <dd>This indicates to the code generator that the parameter or return value
1008 should be zero-extended to a 32-bit value by the caller (for a parameter)
1009 or the callee (for a return value).</dd>
1011 <dt><tt><b>signext</b></tt></dt>
1012 <dd>This indicates to the code generator that the parameter or return value
1013 should be sign-extended to a 32-bit value by the caller (for a parameter)
1014 or the callee (for a return value).</dd>
1016 <dt><tt><b>inreg</b></tt></dt>
1017 <dd>This indicates that this parameter or return value should be treated in a
1018 special target-dependent fashion during while emitting code for a function
1019 call or return (usually, by putting it in a register as opposed to memory,
1020 though some targets use it to distinguish between two different kinds of
1021 registers). Use of this attribute is target-specific.</dd>
1023 <dt><tt><b><a name="byval">byval</a></b></tt></dt>
1024 <dd>This indicates that the pointer parameter should really be passed by value
1025 to the function. The attribute implies that a hidden copy of the pointee
1026 is made between the caller and the callee, so the callee is unable to
1027 modify the value in the callee. This attribute is only valid on LLVM
1028 pointer arguments. It is generally used to pass structs and arrays by
1029 value, but is also valid on pointers to scalars. The copy is considered
1030 to belong to the caller not the callee (for example,
1031 <tt><a href="#readonly">readonly</a></tt> functions should not write to
1032 <tt>byval</tt> parameters). This is not a valid attribute for return
1033 values. The byval attribute also supports specifying an alignment with
1034 the align attribute. This has a target-specific effect on the code
1035 generator that usually indicates a desired alignment for the synthesized
1038 <dt><tt><b><a name="sret">sret</a></b></tt></dt>
1039 <dd>This indicates that the pointer parameter specifies the address of a
1040 structure that is the return value of the function in the source program.
1041 This pointer must be guaranteed by the caller to be valid: loads and
1042 stores to the structure may be assumed by the callee to not to trap. This
1043 may only be applied to the first parameter. This is not a valid attribute
1044 for return values. </dd>
1046 <dt><tt><b><a name="noalias">noalias</a></b></tt></dt>
1047 <dd>This indicates that pointer values
1048 <a href="#pointeraliasing"><i>based</i></a> on the argument or return
1049 value do not alias pointer values which are not <i>based</i> on it,
1050 ignoring certain "irrelevant" dependencies.
1051 For a call to the parent function, dependencies between memory
1052 references from before or after the call and from those during the call
1053 are "irrelevant" to the <tt>noalias</tt> keyword for the arguments and
1054 return value used in that call.
1055 The caller shares the responsibility with the callee for ensuring that
1056 these requirements are met.
1057 For further details, please see the discussion of the NoAlias response in
1058 <a href="AliasAnalysis.html#MustMayNo">alias analysis</a>.<br>
1060 Note that this definition of <tt>noalias</tt> is intentionally
1061 similar to the definition of <tt>restrict</tt> in C99 for function
1062 arguments, though it is slightly weaker.
1064 For function return values, C99's <tt>restrict</tt> is not meaningful,
1065 while LLVM's <tt>noalias</tt> is.
1068 <dt><tt><b><a name="nocapture">nocapture</a></b></tt></dt>
1069 <dd>This indicates that the callee does not make any copies of the pointer
1070 that outlive the callee itself. This is not a valid attribute for return
1073 <dt><tt><b><a name="nest">nest</a></b></tt></dt>
1074 <dd>This indicates that the pointer parameter can be excised using the
1075 <a href="#int_trampoline">trampoline intrinsics</a>. This is not a valid
1076 attribute for return values.</dd>
1081 <!-- ======================================================================= -->
1082 <div class="doc_subsection">
1083 <a name="gc">Garbage Collector Names</a>
1086 <div class="doc_text">
1088 <p>Each function may specify a garbage collector name, which is simply a
1091 <pre class="doc_code">
1092 define void @f() gc "name" { ... }
1095 <p>The compiler declares the supported values of <i>name</i>. Specifying a
1096 collector which will cause the compiler to alter its output in order to
1097 support the named garbage collection algorithm.</p>
1101 <!-- ======================================================================= -->
1102 <div class="doc_subsection">
1103 <a name="fnattrs">Function Attributes</a>
1106 <div class="doc_text">
1108 <p>Function attributes are set to communicate additional information about a
1109 function. Function attributes are considered to be part of the function, not
1110 of the function type, so functions with different parameter attributes can
1111 have the same function type.</p>
1113 <p>Function attributes are simple keywords that follow the type specified. If
1114 multiple attributes are needed, they are space separated. For example:</p>
1116 <pre class="doc_code">
1117 define void @f() noinline { ... }
1118 define void @f() alwaysinline { ... }
1119 define void @f() alwaysinline optsize { ... }
1120 define void @f() optsize { ... }
1124 <dt><tt><b>alignstack(<<em>n</em>>)</b></tt></dt>
1125 <dd>This attribute indicates that, when emitting the prologue and epilogue,
1126 the backend should forcibly align the stack pointer. Specify the
1127 desired alignment, which must be a power of two, in parentheses.
1129 <dt><tt><b>alwaysinline</b></tt></dt>
1130 <dd>This attribute indicates that the inliner should attempt to inline this
1131 function into callers whenever possible, ignoring any active inlining size
1132 threshold for this caller.</dd>
1134 <dt><tt><b>inlinehint</b></tt></dt>
1135 <dd>This attribute indicates that the source code contained a hint that inlining
1136 this function is desirable (such as the "inline" keyword in C/C++). It
1137 is just a hint; it imposes no requirements on the inliner.</dd>
1139 <dt><tt><b>naked</b></tt></dt>
1140 <dd>This attribute disables prologue / epilogue emission for the function.
1141 This can have very system-specific consequences.</dd>
1143 <dt><tt><b>noimplicitfloat</b></tt></dt>
1144 <dd>This attributes disables implicit floating point instructions.</dd>
1146 <dt><tt><b>noinline</b></tt></dt>
1147 <dd>This attribute indicates that the inliner should never inline this
1148 function in any situation. This attribute may not be used together with
1149 the <tt>alwaysinline</tt> attribute.</dd>
1151 <dt><tt><b>noredzone</b></tt></dt>
1152 <dd>This attribute indicates that the code generator should not use a red
1153 zone, even if the target-specific ABI normally permits it.</dd>
1155 <dt><tt><b>noreturn</b></tt></dt>
1156 <dd>This function attribute indicates that the function never returns
1157 normally. This produces undefined behavior at runtime if the function
1158 ever does dynamically return.</dd>
1160 <dt><tt><b>nounwind</b></tt></dt>
1161 <dd>This function attribute indicates that the function never returns with an
1162 unwind or exceptional control flow. If the function does unwind, its
1163 runtime behavior is undefined.</dd>
1165 <dt><tt><b>optsize</b></tt></dt>
1166 <dd>This attribute suggests that optimization passes and code generator passes
1167 make choices that keep the code size of this function low, and otherwise
1168 do optimizations specifically to reduce code size.</dd>
1170 <dt><tt><b>readnone</b></tt></dt>
1171 <dd>This attribute indicates that the function computes its result (or decides
1172 to unwind an exception) based strictly on its arguments, without
1173 dereferencing any pointer arguments or otherwise accessing any mutable
1174 state (e.g. memory, control registers, etc) visible to caller functions.
1175 It does not write through any pointer arguments
1176 (including <tt><a href="#byval">byval</a></tt> arguments) and never
1177 changes any state visible to callers. This means that it cannot unwind
1178 exceptions by calling the <tt>C++</tt> exception throwing methods, but
1179 could use the <tt>unwind</tt> instruction.</dd>
1181 <dt><tt><b><a name="readonly">readonly</a></b></tt></dt>
1182 <dd>This attribute indicates that the function does not write through any
1183 pointer arguments (including <tt><a href="#byval">byval</a></tt>
1184 arguments) or otherwise modify any state (e.g. memory, control registers,
1185 etc) visible to caller functions. It may dereference pointer arguments
1186 and read state that may be set in the caller. A readonly function always
1187 returns the same value (or unwinds an exception identically) when called
1188 with the same set of arguments and global state. It cannot unwind an
1189 exception by calling the <tt>C++</tt> exception throwing methods, but may
1190 use the <tt>unwind</tt> instruction.</dd>
1192 <dt><tt><b><a name="ssp">ssp</a></b></tt></dt>
1193 <dd>This attribute indicates that the function should emit a stack smashing
1194 protector. It is in the form of a "canary"—a random value placed on
1195 the stack before the local variables that's checked upon return from the
1196 function to see if it has been overwritten. A heuristic is used to
1197 determine if a function needs stack protectors or not.<br>
1199 If a function that has an <tt>ssp</tt> attribute is inlined into a
1200 function that doesn't have an <tt>ssp</tt> attribute, then the resulting
1201 function will have an <tt>ssp</tt> attribute.</dd>
1203 <dt><tt><b>sspreq</b></tt></dt>
1204 <dd>This attribute indicates that the function should <em>always</em> emit a
1205 stack smashing protector. This overrides
1206 the <tt><a href="#ssp">ssp</a></tt> function attribute.<br>
1208 If a function that has an <tt>sspreq</tt> attribute is inlined into a
1209 function that doesn't have an <tt>sspreq</tt> attribute or which has
1210 an <tt>ssp</tt> attribute, then the resulting function will have
1211 an <tt>sspreq</tt> attribute.</dd>
1216 <!-- ======================================================================= -->
1217 <div class="doc_subsection">
1218 <a name="moduleasm">Module-Level Inline Assembly</a>
1221 <div class="doc_text">
1223 <p>Modules may contain "module-level inline asm" blocks, which corresponds to
1224 the GCC "file scope inline asm" blocks. These blocks are internally
1225 concatenated by LLVM and treated as a single unit, but may be separated in
1226 the <tt>.ll</tt> file if desired. The syntax is very simple:</p>
1228 <pre class="doc_code">
1229 module asm "inline asm code goes here"
1230 module asm "more can go here"
1233 <p>The strings can contain any character by escaping non-printable characters.
1234 The escape sequence used is simply "\xx" where "xx" is the two digit hex code
1237 <p>The inline asm code is simply printed to the machine code .s file when
1238 assembly code is generated.</p>
1242 <!-- ======================================================================= -->
1243 <div class="doc_subsection">
1244 <a name="datalayout">Data Layout</a>
1247 <div class="doc_text">
1249 <p>A module may specify a target specific data layout string that specifies how
1250 data is to be laid out in memory. The syntax for the data layout is
1253 <pre class="doc_code">
1254 target datalayout = "<i>layout specification</i>"
1257 <p>The <i>layout specification</i> consists of a list of specifications
1258 separated by the minus sign character ('-'). Each specification starts with
1259 a letter and may include other information after the letter to define some
1260 aspect of the data layout. The specifications accepted are as follows:</p>
1264 <dd>Specifies that the target lays out data in big-endian form. That is, the
1265 bits with the most significance have the lowest address location.</dd>
1268 <dd>Specifies that the target lays out data in little-endian form. That is,
1269 the bits with the least significance have the lowest address
1272 <dt><tt>p:<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1273 <dd>This specifies the <i>size</i> of a pointer and its <i>abi</i> and
1274 <i>preferred</i> alignments. All sizes are in bits. Specifying
1275 the <i>pref</i> alignment is optional. If omitted, the
1276 preceding <tt>:</tt> should be omitted too.</dd>
1278 <dt><tt>i<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1279 <dd>This specifies the alignment for an integer type of a given bit
1280 <i>size</i>. The value of <i>size</i> must be in the range [1,2^23).</dd>
1282 <dt><tt>v<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1283 <dd>This specifies the alignment for a vector type of a given bit
1286 <dt><tt>f<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1287 <dd>This specifies the alignment for a floating point type of a given bit
1288 <i>size</i>. Only values of <i>size</i> that are supported by the target
1289 will work. 32 (float) and 64 (double) are supported on all targets;
1290 80 or 128 (different flavors of long double) are also supported on some
1293 <dt><tt>a<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1294 <dd>This specifies the alignment for an aggregate type of a given bit
1297 <dt><tt>s<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1298 <dd>This specifies the alignment for a stack object of a given bit
1301 <dt><tt>n<i>size1</i>:<i>size2</i>:<i>size3</i>...</tt></dt>
1302 <dd>This specifies a set of native integer widths for the target CPU
1303 in bits. For example, it might contain "n32" for 32-bit PowerPC,
1304 "n32:64" for PowerPC 64, or "n8:16:32:64" for X86-64. Elements of
1305 this set are considered to support most general arithmetic
1306 operations efficiently.</dd>
1309 <p>When constructing the data layout for a given target, LLVM starts with a
1310 default set of specifications which are then (possibly) overridden by the
1311 specifications in the <tt>datalayout</tt> keyword. The default specifications
1312 are given in this list:</p>
1315 <li><tt>E</tt> - big endian</li>
1316 <li><tt>p:64:64:64</tt> - 64-bit pointers with 64-bit alignment</li>
1317 <li><tt>i1:8:8</tt> - i1 is 8-bit (byte) aligned</li>
1318 <li><tt>i8:8:8</tt> - i8 is 8-bit (byte) aligned</li>
1319 <li><tt>i16:16:16</tt> - i16 is 16-bit aligned</li>
1320 <li><tt>i32:32:32</tt> - i32 is 32-bit aligned</li>
1321 <li><tt>i64:32:64</tt> - i64 has ABI alignment of 32-bits but preferred
1322 alignment of 64-bits</li>
1323 <li><tt>f32:32:32</tt> - float is 32-bit aligned</li>
1324 <li><tt>f64:64:64</tt> - double is 64-bit aligned</li>
1325 <li><tt>v64:64:64</tt> - 64-bit vector is 64-bit aligned</li>
1326 <li><tt>v128:128:128</tt> - 128-bit vector is 128-bit aligned</li>
1327 <li><tt>a0:0:1</tt> - aggregates are 8-bit aligned</li>
1328 <li><tt>s0:64:64</tt> - stack objects are 64-bit aligned</li>
1331 <p>When LLVM is determining the alignment for a given type, it uses the
1332 following rules:</p>
1335 <li>If the type sought is an exact match for one of the specifications, that
1336 specification is used.</li>
1338 <li>If no match is found, and the type sought is an integer type, then the
1339 smallest integer type that is larger than the bitwidth of the sought type
1340 is used. If none of the specifications are larger than the bitwidth then
1341 the the largest integer type is used. For example, given the default
1342 specifications above, the i7 type will use the alignment of i8 (next
1343 largest) while both i65 and i256 will use the alignment of i64 (largest
1346 <li>If no match is found, and the type sought is a vector type, then the
1347 largest vector type that is smaller than the sought vector type will be
1348 used as a fall back. This happens because <128 x double> can be
1349 implemented in terms of 64 <2 x double>, for example.</li>
1354 <!-- ======================================================================= -->
1355 <div class="doc_subsection">
1356 <a name="pointeraliasing">Pointer Aliasing Rules</a>
1359 <div class="doc_text">
1361 <p>Any memory access must be done through a pointer value associated
1362 with an address range of the memory access, otherwise the behavior
1363 is undefined. Pointer values are associated with address ranges
1364 according to the following rules:</p>
1367 <li>A pointer value is associated with the addresses associated with
1368 any value it is <i>based</i> on.
1369 <li>An address of a global variable is associated with the address
1370 range of the variable's storage.</li>
1371 <li>The result value of an allocation instruction is associated with
1372 the address range of the allocated storage.</li>
1373 <li>A null pointer in the default address-space is associated with
1375 <li>An integer constant other than zero or a pointer value returned
1376 from a function not defined within LLVM may be associated with address
1377 ranges allocated through mechanisms other than those provided by
1378 LLVM. Such ranges shall not overlap with any ranges of addresses
1379 allocated by mechanisms provided by LLVM.</li>
1382 <p>A pointer value is <i>based</i> on another pointer value according
1383 to the following rules:</p>
1386 <li>A pointer value formed from a
1387 <tt><a href="#i_getelementptr">getelementptr</a></tt> operation
1388 is <i>based</i> on the first operand of the <tt>getelementptr</tt>.</li>
1389 <li>The result value of a
1390 <tt><a href="#i_bitcast">bitcast</a></tt> is <i>based</i> on the operand
1391 of the <tt>bitcast</tt>.</li>
1392 <li>A pointer value formed by an
1393 <tt><a href="#i_inttoptr">inttoptr</a></tt> is <i>based</i> on all
1394 pointer values that contribute (directly or indirectly) to the
1395 computation of the pointer's value.</li>
1396 <li>The "<i>based</i> on" relationship is transitive.</li>
1399 <p>Note that this definition of <i>"based"</i> is intentionally
1400 similar to the definition of <i>"based"</i> in C99, though it is
1401 slightly weaker.</p>
1403 <p>LLVM IR does not associate types with memory. The result type of a
1404 <tt><a href="#i_load">load</a></tt> merely indicates the size and
1405 alignment of the memory from which to load, as well as the
1406 interpretation of the value. The first operand type of a
1407 <tt><a href="#i_store">store</a></tt> similarly only indicates the size
1408 and alignment of the store.</p>
1410 <p>Consequently, type-based alias analysis, aka TBAA, aka
1411 <tt>-fstrict-aliasing</tt>, is not applicable to general unadorned
1412 LLVM IR. <a href="#metadata">Metadata</a> may be used to encode
1413 additional information which specialized optimization passes may use
1414 to implement type-based alias analysis.</p>
1418 <!-- ======================================================================= -->
1419 <div class="doc_subsection">
1420 <a name="volatile">Volatile Memory Accesses</a>
1423 <div class="doc_text">
1425 <p>Certain memory accesses, such as <a href="#i_load"><tt>load</tt></a>s, <a
1426 href="#i_store"><tt>store</tt></a>s, and <a
1427 href="#int_memcpy"><tt>llvm.memcpy</tt></a>s may be marked <tt>volatile</tt>.
1428 The optimizers must not change the number of volatile operations or change their
1429 order of execution relative to other volatile operations. The optimizers
1430 <i>may</i> change the order of volatile operations relative to non-volatile
1431 operations. This is not Java's "volatile" and has no cross-thread
1432 synchronization behavior.</p>
1436 <!-- *********************************************************************** -->
1437 <div class="doc_section"> <a name="typesystem">Type System</a> </div>
1438 <!-- *********************************************************************** -->
1440 <div class="doc_text">
1442 <p>The LLVM type system is one of the most important features of the
1443 intermediate representation. Being typed enables a number of optimizations
1444 to be performed on the intermediate representation directly, without having
1445 to do extra analyses on the side before the transformation. A strong type
1446 system makes it easier to read the generated code and enables novel analyses
1447 and transformations that are not feasible to perform on normal three address
1448 code representations.</p>
1452 <!-- ======================================================================= -->
1453 <div class="doc_subsection"> <a name="t_classifications">Type
1454 Classifications</a> </div>
1456 <div class="doc_text">
1458 <p>The types fall into a few useful classifications:</p>
1460 <table border="1" cellspacing="0" cellpadding="4">
1462 <tr><th>Classification</th><th>Types</th></tr>
1464 <td><a href="#t_integer">integer</a></td>
1465 <td><tt>i1, i2, i3, ... i8, ... i16, ... i32, ... i64, ... </tt></td>
1468 <td><a href="#t_floating">floating point</a></td>
1469 <td><tt>float, double, x86_fp80, fp128, ppc_fp128</tt></td>
1472 <td><a name="t_firstclass">first class</a></td>
1473 <td><a href="#t_integer">integer</a>,
1474 <a href="#t_floating">floating point</a>,
1475 <a href="#t_pointer">pointer</a>,
1476 <a href="#t_vector">vector</a>,
1477 <a href="#t_struct">structure</a>,
1478 <a href="#t_array">array</a>,
1479 <a href="#t_label">label</a>,
1480 <a href="#t_metadata">metadata</a>.
1484 <td><a href="#t_primitive">primitive</a></td>
1485 <td><a href="#t_label">label</a>,
1486 <a href="#t_void">void</a>,
1487 <a href="#t_floating">floating point</a>,
1488 <a href="#t_x86mmx">x86mmx</a>,
1489 <a href="#t_metadata">metadata</a>.</td>
1492 <td><a href="#t_derived">derived</a></td>
1493 <td><a href="#t_array">array</a>,
1494 <a href="#t_function">function</a>,
1495 <a href="#t_pointer">pointer</a>,
1496 <a href="#t_struct">structure</a>,
1497 <a href="#t_pstruct">packed structure</a>,
1498 <a href="#t_vector">vector</a>,
1499 <a href="#t_opaque">opaque</a>.
1505 <p>The <a href="#t_firstclass">first class</a> types are perhaps the most
1506 important. Values of these types are the only ones which can be produced by
1511 <!-- ======================================================================= -->
1512 <div class="doc_subsection"> <a name="t_primitive">Primitive Types</a> </div>
1514 <div class="doc_text">
1516 <p>The primitive types are the fundamental building blocks of the LLVM
1521 <!-- _______________________________________________________________________ -->
1522 <div class="doc_subsubsection"> <a name="t_integer">Integer Type</a> </div>
1524 <div class="doc_text">
1527 <p>The integer type is a very simple type that simply specifies an arbitrary
1528 bit width for the integer type desired. Any bit width from 1 bit to
1529 2<sup>23</sup>-1 (about 8 million) can be specified.</p>
1536 <p>The number of bits the integer will occupy is specified by the <tt>N</tt>
1540 <table class="layout">
1542 <td class="left"><tt>i1</tt></td>
1543 <td class="left">a single-bit integer.</td>
1546 <td class="left"><tt>i32</tt></td>
1547 <td class="left">a 32-bit integer.</td>
1550 <td class="left"><tt>i1942652</tt></td>
1551 <td class="left">a really big integer of over 1 million bits.</td>
1557 <!-- _______________________________________________________________________ -->
1558 <div class="doc_subsubsection"> <a name="t_floating">Floating Point Types</a> </div>
1560 <div class="doc_text">
1564 <tr><th>Type</th><th>Description</th></tr>
1565 <tr><td><tt>float</tt></td><td>32-bit floating point value</td></tr>
1566 <tr><td><tt>double</tt></td><td>64-bit floating point value</td></tr>
1567 <tr><td><tt>fp128</tt></td><td>128-bit floating point value (112-bit mantissa)</td></tr>
1568 <tr><td><tt>x86_fp80</tt></td><td>80-bit floating point value (X87)</td></tr>
1569 <tr><td><tt>ppc_fp128</tt></td><td>128-bit floating point value (two 64-bits)</td></tr>
1575 <!-- _______________________________________________________________________ -->
1576 <div class="doc_subsubsection"> <a name="t_x86mmx">X86mmx Type</a> </div>
1578 <div class="doc_text">
1581 <p>The x86mmx type represents a value held in an MMX register on an x86 machine. The operations allowed on it are quite limited: parameters and return values, load and store, and bitcast. User-specified MMX instructions are represented as intrinsic or asm calls with arguments and/or results of this type. There are no arrays, vectors or constants of this type.</p>
1590 <!-- _______________________________________________________________________ -->
1591 <div class="doc_subsubsection"> <a name="t_void">Void Type</a> </div>
1593 <div class="doc_text">
1596 <p>The void type does not represent any value and has no size.</p>
1605 <!-- _______________________________________________________________________ -->
1606 <div class="doc_subsubsection"> <a name="t_label">Label Type</a> </div>
1608 <div class="doc_text">
1611 <p>The label type represents code labels.</p>
1620 <!-- _______________________________________________________________________ -->
1621 <div class="doc_subsubsection"> <a name="t_metadata">Metadata Type</a> </div>
1623 <div class="doc_text">
1626 <p>The metadata type represents embedded metadata. No derived types may be
1627 created from metadata except for <a href="#t_function">function</a>
1638 <!-- ======================================================================= -->
1639 <div class="doc_subsection"> <a name="t_derived">Derived Types</a> </div>
1641 <div class="doc_text">
1643 <p>The real power in LLVM comes from the derived types in the system. This is
1644 what allows a programmer to represent arrays, functions, pointers, and other
1645 useful types. Each of these types contain one or more element types which
1646 may be a primitive type, or another derived type. For example, it is
1647 possible to have a two dimensional array, using an array as the element type
1648 of another array.</p>
1653 <!-- _______________________________________________________________________ -->
1654 <div class="doc_subsubsection"> <a name="t_aggregate">Aggregate Types</a> </div>
1656 <div class="doc_text">
1658 <p>Aggregate Types are a subset of derived types that can contain multiple
1659 member types. <a href="#t_array">Arrays</a>,
1660 <a href="#t_struct">structs</a>, and <a href="#t_vector">vectors</a> are
1661 aggregate types.</p>
1665 <!-- _______________________________________________________________________ -->
1666 <div class="doc_subsubsection"> <a name="t_array">Array Type</a> </div>
1668 <div class="doc_text">
1671 <p>The array type is a very simple derived type that arranges elements
1672 sequentially in memory. The array type requires a size (number of elements)
1673 and an underlying data type.</p>
1677 [<# elements> x <elementtype>]
1680 <p>The number of elements is a constant integer value; <tt>elementtype</tt> may
1681 be any type with a size.</p>
1684 <table class="layout">
1686 <td class="left"><tt>[40 x i32]</tt></td>
1687 <td class="left">Array of 40 32-bit integer values.</td>
1690 <td class="left"><tt>[41 x i32]</tt></td>
1691 <td class="left">Array of 41 32-bit integer values.</td>
1694 <td class="left"><tt>[4 x i8]</tt></td>
1695 <td class="left">Array of 4 8-bit integer values.</td>
1698 <p>Here are some examples of multidimensional arrays:</p>
1699 <table class="layout">
1701 <td class="left"><tt>[3 x [4 x i32]]</tt></td>
1702 <td class="left">3x4 array of 32-bit integer values.</td>
1705 <td class="left"><tt>[12 x [10 x float]]</tt></td>
1706 <td class="left">12x10 array of single precision floating point values.</td>
1709 <td class="left"><tt>[2 x [3 x [4 x i16]]]</tt></td>
1710 <td class="left">2x3x4 array of 16-bit integer values.</td>
1714 <p>There is no restriction on indexing beyond the end of the array implied by
1715 a static type (though there are restrictions on indexing beyond the bounds
1716 of an allocated object in some cases). This means that single-dimension
1717 'variable sized array' addressing can be implemented in LLVM with a zero
1718 length array type. An implementation of 'pascal style arrays' in LLVM could
1719 use the type "<tt>{ i32, [0 x float]}</tt>", for example.</p>
1723 <!-- _______________________________________________________________________ -->
1724 <div class="doc_subsubsection"> <a name="t_function">Function Type</a> </div>
1726 <div class="doc_text">
1729 <p>The function type can be thought of as a function signature. It consists of
1730 a return type and a list of formal parameter types. The return type of a
1731 function type is a first class type or a void type.</p>
1735 <returntype> (<parameter list>)
1738 <p>...where '<tt><parameter list></tt>' is a comma-separated list of type
1739 specifiers. Optionally, the parameter list may include a type <tt>...</tt>,
1740 which indicates that the function takes a variable number of arguments.
1741 Variable argument functions can access their arguments with
1742 the <a href="#int_varargs">variable argument handling intrinsic</a>
1743 functions. '<tt><returntype></tt>' is any type except
1744 <a href="#t_label">label</a>.</p>
1747 <table class="layout">
1749 <td class="left"><tt>i32 (i32)</tt></td>
1750 <td class="left">function taking an <tt>i32</tt>, returning an <tt>i32</tt>
1752 </tr><tr class="layout">
1753 <td class="left"><tt>float (i16, i32 *) *
1755 <td class="left"><a href="#t_pointer">Pointer</a> to a function that takes
1756 an <tt>i16</tt> and a <a href="#t_pointer">pointer</a> to <tt>i32</tt>,
1757 returning <tt>float</tt>.
1759 </tr><tr class="layout">
1760 <td class="left"><tt>i32 (i8*, ...)</tt></td>
1761 <td class="left">A vararg function that takes at least one
1762 <a href="#t_pointer">pointer</a> to <tt>i8 </tt> (char in C),
1763 which returns an integer. This is the signature for <tt>printf</tt> in
1766 </tr><tr class="layout">
1767 <td class="left"><tt>{i32, i32} (i32)</tt></td>
1768 <td class="left">A function taking an <tt>i32</tt>, returning a
1769 <a href="#t_struct">structure</a> containing two <tt>i32</tt> values
1776 <!-- _______________________________________________________________________ -->
1777 <div class="doc_subsubsection"> <a name="t_struct">Structure Type</a> </div>
1779 <div class="doc_text">
1782 <p>The structure type is used to represent a collection of data members together
1783 in memory. The packing of the field types is defined to match the ABI of the
1784 underlying processor. The elements of a structure may be any type that has a
1787 <p>Structures in memory are accessed using '<tt><a href="#i_load">load</a></tt>'
1788 and '<tt><a href="#i_store">store</a></tt>' by getting a pointer to a field
1789 with the '<tt><a href="#i_getelementptr">getelementptr</a></tt>' instruction.
1790 Structures in registers are accessed using the
1791 '<tt><a href="#i_extractvalue">extractvalue</a></tt>' and
1792 '<tt><a href="#i_insertvalue">insertvalue</a></tt>' instructions.</p>
1795 { <type list> }
1799 <table class="layout">
1801 <td class="left"><tt>{ i32, i32, i32 }</tt></td>
1802 <td class="left">A triple of three <tt>i32</tt> values</td>
1803 </tr><tr class="layout">
1804 <td class="left"><tt>{ float, i32 (i32) * }</tt></td>
1805 <td class="left">A pair, where the first element is a <tt>float</tt> and the
1806 second element is a <a href="#t_pointer">pointer</a> to a
1807 <a href="#t_function">function</a> that takes an <tt>i32</tt>, returning
1808 an <tt>i32</tt>.</td>
1814 <!-- _______________________________________________________________________ -->
1815 <div class="doc_subsubsection"> <a name="t_pstruct">Packed Structure Type</a>
1818 <div class="doc_text">
1821 <p>The packed structure type is used to represent a collection of data members
1822 together in memory. There is no padding between fields. Further, the
1823 alignment of a packed structure is 1 byte. The elements of a packed
1824 structure may be any type that has a size.</p>
1826 <p>Structures are accessed using '<tt><a href="#i_load">load</a></tt> and
1827 '<tt><a href="#i_store">store</a></tt>' by getting a pointer to a field with
1828 the '<tt><a href="#i_getelementptr">getelementptr</a></tt>' instruction.</p>
1832 < { <type list> } >
1836 <table class="layout">
1838 <td class="left"><tt>< { i32, i32, i32 } ></tt></td>
1839 <td class="left">A triple of three <tt>i32</tt> values</td>
1840 </tr><tr class="layout">
1842 <tt>< { float, i32 (i32)* } ></tt></td>
1843 <td class="left">A pair, where the first element is a <tt>float</tt> and the
1844 second element is a <a href="#t_pointer">pointer</a> to a
1845 <a href="#t_function">function</a> that takes an <tt>i32</tt>, returning
1846 an <tt>i32</tt>.</td>
1852 <!-- _______________________________________________________________________ -->
1853 <div class="doc_subsubsection"> <a name="t_pointer">Pointer Type</a> </div>
1855 <div class="doc_text">
1858 <p>The pointer type is used to specify memory locations.
1859 Pointers are commonly used to reference objects in memory.</p>
1861 <p>Pointer types may have an optional address space attribute defining the
1862 numbered address space where the pointed-to object resides. The default
1863 address space is number zero. The semantics of non-zero address
1864 spaces are target-specific.</p>
1866 <p>Note that LLVM does not permit pointers to void (<tt>void*</tt>) nor does it
1867 permit pointers to labels (<tt>label*</tt>). Use <tt>i8*</tt> instead.</p>
1875 <table class="layout">
1877 <td class="left"><tt>[4 x i32]*</tt></td>
1878 <td class="left">A <a href="#t_pointer">pointer</a> to <a
1879 href="#t_array">array</a> of four <tt>i32</tt> values.</td>
1882 <td class="left"><tt>i32 (i32*) *</tt></td>
1883 <td class="left"> A <a href="#t_pointer">pointer</a> to a <a
1884 href="#t_function">function</a> that takes an <tt>i32*</tt>, returning an
1888 <td class="left"><tt>i32 addrspace(5)*</tt></td>
1889 <td class="left">A <a href="#t_pointer">pointer</a> to an <tt>i32</tt> value
1890 that resides in address space #5.</td>
1896 <!-- _______________________________________________________________________ -->
1897 <div class="doc_subsubsection"> <a name="t_vector">Vector Type</a> </div>
1899 <div class="doc_text">
1902 <p>A vector type is a simple derived type that represents a vector of elements.
1903 Vector types are used when multiple primitive data are operated in parallel
1904 using a single instruction (SIMD). A vector type requires a size (number of
1905 elements) and an underlying primitive data type. Vector types are considered
1906 <a href="#t_firstclass">first class</a>.</p>
1910 < <# elements> x <elementtype> >
1913 <p>The number of elements is a constant integer value; elementtype may be any
1914 integer or floating point type.</p>
1917 <table class="layout">
1919 <td class="left"><tt><4 x i32></tt></td>
1920 <td class="left">Vector of 4 32-bit integer values.</td>
1923 <td class="left"><tt><8 x float></tt></td>
1924 <td class="left">Vector of 8 32-bit floating-point values.</td>
1927 <td class="left"><tt><2 x i64></tt></td>
1928 <td class="left">Vector of 2 64-bit integer values.</td>
1934 <!-- _______________________________________________________________________ -->
1935 <div class="doc_subsubsection"> <a name="t_opaque">Opaque Type</a> </div>
1936 <div class="doc_text">
1939 <p>Opaque types are used to represent unknown types in the system. This
1940 corresponds (for example) to the C notion of a forward declared structure
1941 type. In LLVM, opaque types can eventually be resolved to any type (not just
1942 a structure type).</p>
1950 <table class="layout">
1952 <td class="left"><tt>opaque</tt></td>
1953 <td class="left">An opaque type.</td>
1959 <!-- ======================================================================= -->
1960 <div class="doc_subsection">
1961 <a name="t_uprefs">Type Up-references</a>
1964 <div class="doc_text">
1967 <p>An "up reference" allows you to refer to a lexically enclosing type without
1968 requiring it to have a name. For instance, a structure declaration may
1969 contain a pointer to any of the types it is lexically a member of. Example
1970 of up references (with their equivalent as named type declarations)
1974 { \2 * } %x = type { %x* }
1975 { \2 }* %y = type { %y }*
1979 <p>An up reference is needed by the asmprinter for printing out cyclic types
1980 when there is no declared name for a type in the cycle. Because the
1981 asmprinter does not want to print out an infinite type string, it needs a
1982 syntax to handle recursive types that have no names (all names are optional
1990 <p>The level is the count of the lexical type that is being referred to.</p>
1993 <table class="layout">
1995 <td class="left"><tt>\1*</tt></td>
1996 <td class="left">Self-referential pointer.</td>
1999 <td class="left"><tt>{ { \3*, i8 }, i32 }</tt></td>
2000 <td class="left">Recursive structure where the upref refers to the out-most
2007 <!-- *********************************************************************** -->
2008 <div class="doc_section"> <a name="constants">Constants</a> </div>
2009 <!-- *********************************************************************** -->
2011 <div class="doc_text">
2013 <p>LLVM has several different basic types of constants. This section describes
2014 them all and their syntax.</p>
2018 <!-- ======================================================================= -->
2019 <div class="doc_subsection"><a name="simpleconstants">Simple Constants</a></div>
2021 <div class="doc_text">
2024 <dt><b>Boolean constants</b></dt>
2025 <dd>The two strings '<tt>true</tt>' and '<tt>false</tt>' are both valid
2026 constants of the <tt><a href="#t_integer">i1</a></tt> type.</dd>
2028 <dt><b>Integer constants</b></dt>
2029 <dd>Standard integers (such as '4') are constants of
2030 the <a href="#t_integer">integer</a> type. Negative numbers may be used
2031 with integer types.</dd>
2033 <dt><b>Floating point constants</b></dt>
2034 <dd>Floating point constants use standard decimal notation (e.g. 123.421),
2035 exponential notation (e.g. 1.23421e+2), or a more precise hexadecimal
2036 notation (see below). The assembler requires the exact decimal value of a
2037 floating-point constant. For example, the assembler accepts 1.25 but
2038 rejects 1.3 because 1.3 is a repeating decimal in binary. Floating point
2039 constants must have a <a href="#t_floating">floating point</a> type. </dd>
2041 <dt><b>Null pointer constants</b></dt>
2042 <dd>The identifier '<tt>null</tt>' is recognized as a null pointer constant
2043 and must be of <a href="#t_pointer">pointer type</a>.</dd>
2046 <p>The one non-intuitive notation for constants is the hexadecimal form of
2047 floating point constants. For example, the form '<tt>double
2048 0x432ff973cafa8000</tt>' is equivalent to (but harder to read than)
2049 '<tt>double 4.5e+15</tt>'. The only time hexadecimal floating point
2050 constants are required (and the only time that they are generated by the
2051 disassembler) is when a floating point constant must be emitted but it cannot
2052 be represented as a decimal floating point number in a reasonable number of
2053 digits. For example, NaN's, infinities, and other special values are
2054 represented in their IEEE hexadecimal format so that assembly and disassembly
2055 do not cause any bits to change in the constants.</p>
2057 <p>When using the hexadecimal form, constants of types float and double are
2058 represented using the 16-digit form shown above (which matches the IEEE754
2059 representation for double); float values must, however, be exactly
2060 representable as IEE754 single precision. Hexadecimal format is always used
2061 for long double, and there are three forms of long double. The 80-bit format
2062 used by x86 is represented as <tt>0xK</tt> followed by 20 hexadecimal digits.
2063 The 128-bit format used by PowerPC (two adjacent doubles) is represented
2064 by <tt>0xM</tt> followed by 32 hexadecimal digits. The IEEE 128-bit format
2065 is represented by <tt>0xL</tt> followed by 32 hexadecimal digits; no
2066 currently supported target uses this format. Long doubles will only work if
2067 they match the long double format on your target. All hexadecimal formats
2068 are big-endian (sign bit at the left).</p>
2070 <p>There are no constants of type x86mmx.</p>
2073 <!-- ======================================================================= -->
2074 <div class="doc_subsection">
2075 <a name="aggregateconstants"></a> <!-- old anchor -->
2076 <a name="complexconstants">Complex Constants</a>
2079 <div class="doc_text">
2081 <p>Complex constants are a (potentially recursive) combination of simple
2082 constants and smaller complex constants.</p>
2085 <dt><b>Structure constants</b></dt>
2086 <dd>Structure constants are represented with notation similar to structure
2087 type definitions (a comma separated list of elements, surrounded by braces
2088 (<tt>{}</tt>)). For example: "<tt>{ i32 4, float 17.0, i32* @G }</tt>",
2089 where "<tt>@G</tt>" is declared as "<tt>@G = external global i32</tt>".
2090 Structure constants must have <a href="#t_struct">structure type</a>, and
2091 the number and types of elements must match those specified by the
2094 <dt><b>Array constants</b></dt>
2095 <dd>Array constants are represented with notation similar to array type
2096 definitions (a comma separated list of elements, surrounded by square
2097 brackets (<tt>[]</tt>)). For example: "<tt>[ i32 42, i32 11, i32 74
2098 ]</tt>". Array constants must have <a href="#t_array">array type</a>, and
2099 the number and types of elements must match those specified by the
2102 <dt><b>Vector constants</b></dt>
2103 <dd>Vector constants are represented with notation similar to vector type
2104 definitions (a comma separated list of elements, surrounded by
2105 less-than/greater-than's (<tt><></tt>)). For example: "<tt>< i32
2106 42, i32 11, i32 74, i32 100 ></tt>". Vector constants must
2107 have <a href="#t_vector">vector type</a>, and the number and types of
2108 elements must match those specified by the type.</dd>
2110 <dt><b>Zero initialization</b></dt>
2111 <dd>The string '<tt>zeroinitializer</tt>' can be used to zero initialize a
2112 value to zero of <em>any</em> type, including scalar and
2113 <a href="#t_aggregate">aggregate</a> types.
2114 This is often used to avoid having to print large zero initializers
2115 (e.g. for large arrays) and is always exactly equivalent to using explicit
2116 zero initializers.</dd>
2118 <dt><b>Metadata node</b></dt>
2119 <dd>A metadata node is a structure-like constant with
2120 <a href="#t_metadata">metadata type</a>. For example: "<tt>metadata !{
2121 i32 0, metadata !"test" }</tt>". Unlike other constants that are meant to
2122 be interpreted as part of the instruction stream, metadata is a place to
2123 attach additional information such as debug info.</dd>
2128 <!-- ======================================================================= -->
2129 <div class="doc_subsection">
2130 <a name="globalconstants">Global Variable and Function Addresses</a>
2133 <div class="doc_text">
2135 <p>The addresses of <a href="#globalvars">global variables</a>
2136 and <a href="#functionstructure">functions</a> are always implicitly valid
2137 (link-time) constants. These constants are explicitly referenced when
2138 the <a href="#identifiers">identifier for the global</a> is used and always
2139 have <a href="#t_pointer">pointer</a> type. For example, the following is a
2140 legal LLVM file:</p>
2142 <pre class="doc_code">
2145 @Z = global [2 x i32*] [ i32* @X, i32* @Y ]
2150 <!-- ======================================================================= -->
2151 <div class="doc_subsection"><a name="undefvalues">Undefined Values</a></div>
2152 <div class="doc_text">
2154 <p>The string '<tt>undef</tt>' can be used anywhere a constant is expected, and
2155 indicates that the user of the value may receive an unspecified bit-pattern.
2156 Undefined values may be of any type (other than label or void) and be used
2157 anywhere a constant is permitted.</p>
2159 <p>Undefined values are useful because they indicate to the compiler that the
2160 program is well defined no matter what value is used. This gives the
2161 compiler more freedom to optimize. Here are some examples of (potentially
2162 surprising) transformations that are valid (in pseudo IR):</p>
2165 <pre class="doc_code">
2175 <p>This is safe because all of the output bits are affected by the undef bits.
2176 Any output bit can have a zero or one depending on the input bits.</p>
2178 <pre class="doc_code">
2189 <p>These logical operations have bits that are not always affected by the input.
2190 For example, if "%X" has a zero bit, then the output of the 'and' operation will
2191 always be a zero, no matter what the corresponding bit from the undef is. As
2192 such, it is unsafe to optimize or assume that the result of the and is undef.
2193 However, it is safe to assume that all bits of the undef could be 0, and
2194 optimize the and to 0. Likewise, it is safe to assume that all the bits of
2195 the undef operand to the or could be set, allowing the or to be folded to
2198 <pre class="doc_code">
2199 %A = select undef, %X, %Y
2200 %B = select undef, 42, %Y
2201 %C = select %X, %Y, undef
2212 <p>This set of examples show that undefined select (and conditional branch)
2213 conditions can go "either way" but they have to come from one of the two
2214 operands. In the %A example, if %X and %Y were both known to have a clear low
2215 bit, then %A would have to have a cleared low bit. However, in the %C example,
2216 the optimizer is allowed to assume that the undef operand could be the same as
2217 %Y, allowing the whole select to be eliminated.</p>
2220 <pre class="doc_code">
2221 %A = xor undef, undef
2239 <p>This example points out that two undef operands are not necessarily the same.
2240 This can be surprising to people (and also matches C semantics) where they
2241 assume that "X^X" is always zero, even if X is undef. This isn't true for a
2242 number of reasons, but the short answer is that an undef "variable" can
2243 arbitrarily change its value over its "live range". This is true because the
2244 "variable" doesn't actually <em>have a live range</em>. Instead, the value is
2245 logically read from arbitrary registers that happen to be around when needed,
2246 so the value is not necessarily consistent over time. In fact, %A and %C need
2247 to have the same semantics or the core LLVM "replace all uses with" concept
2250 <pre class="doc_code">
2258 <p>These examples show the crucial difference between an <em>undefined
2259 value</em> and <em>undefined behavior</em>. An undefined value (like undef) is
2260 allowed to have an arbitrary bit-pattern. This means that the %A operation
2261 can be constant folded to undef because the undef could be an SNaN, and fdiv is
2262 not (currently) defined on SNaN's. However, in the second example, we can make
2263 a more aggressive assumption: because the undef is allowed to be an arbitrary
2264 value, we are allowed to assume that it could be zero. Since a divide by zero
2265 has <em>undefined behavior</em>, we are allowed to assume that the operation
2266 does not execute at all. This allows us to delete the divide and all code after
2267 it: since the undefined operation "can't happen", the optimizer can assume that
2268 it occurs in dead code.
2271 <pre class="doc_code">
2272 a: store undef -> %X
2273 b: store %X -> undef
2279 <p>These examples reiterate the fdiv example: a store "of" an undefined value
2280 can be assumed to not have any effect: we can assume that the value is
2281 overwritten with bits that happen to match what was already there. However, a
2282 store "to" an undefined location could clobber arbitrary memory, therefore, it
2283 has undefined behavior.</p>
2287 <!-- ======================================================================= -->
2288 <div class="doc_subsection"><a name="trapvalues">Trap Values</a></div>
2289 <div class="doc_text">
2291 <p>Trap values are similar to <a href="#undefvalues">undef values</a>, however
2292 instead of representing an unspecified bit pattern, they represent the
2293 fact that an instruction or constant expression which cannot evoke side
2294 effects has nevertheless detected a condition which results in undefined
2297 <p>There is currently no way of representing a trap value in the IR; they
2298 only exist when produced by operations such as
2299 <a href="#i_add"><tt>add</tt></a> with the <tt>nsw</tt> flag.</p>
2301 <p>Trap value behavior is defined in terms of value <i>dependence</i>:</p>
2304 <li>Values other than <a href="#i_phi"><tt>phi</tt></a> nodes depend on
2305 their operands.</li>
2307 <li><a href="#i_phi"><tt>Phi</tt></a> nodes depend on the operand corresponding
2308 to their dynamic predecessor basic block.</li>
2310 <li>Function arguments depend on the corresponding actual argument values in
2311 the dynamic callers of their functions.</li>
2313 <li><a href="#i_call"><tt>Call</tt></a> instructions depend on the
2314 <a href="#i_ret"><tt>ret</tt></a> instructions that dynamically transfer
2315 control back to them.</li>
2317 <li><a href="#i_invoke"><tt>Invoke</tt></a> instructions depend on the
2318 <a href="#i_ret"><tt>ret</tt></a>, <a href="#i_unwind"><tt>unwind</tt></a>,
2319 or exception-throwing call instructions that dynamically transfer control
2322 <li>Non-volatile loads and stores depend on the most recent stores to all of the
2323 referenced memory addresses, following the order in the IR
2324 (including loads and stores implied by intrinsics such as
2325 <a href="#int_memcpy"><tt>@llvm.memcpy</tt></a>.)</li>
2327 <!-- TODO: In the case of multiple threads, this only applies if the store
2328 "happens-before" the load or store. -->
2330 <!-- TODO: floating-point exception state -->
2332 <li>An instruction with externally visible side effects depends on the most
2333 recent preceding instruction with externally visible side effects, following
2334 the order in the IR. (This includes
2335 <a href="#volatile">volatile operations</a>.)</li>
2337 <li>An instruction <i>control-depends</i> on a
2338 <a href="#terminators">terminator instruction</a>
2339 if the terminator instruction has multiple successors and the instruction
2340 is always executed when control transfers to one of the successors, and
2341 may not be executed when control is transfered to another.</li>
2343 <li>Dependence is transitive.</li>
2347 <p>Whenever a trap value is generated, all values which depend on it evaluate
2348 to trap. If they have side effects, the evoke their side effects as if each
2349 operand with a trap value were undef. If they have externally-visible side
2350 effects, the behavior is undefined.</p>
2352 <p>Here are some examples:</p>
2354 <pre class="doc_code">
2356 %trap = sub nuw i32 0, 1 ; Results in a trap value.
2357 %still_trap = and i32 %trap, 0 ; Whereas (and i32 undef, 0) would return 0.
2358 %trap_yet_again = getelementptr i32* @h, i32 %still_trap
2359 store i32 0, i32* %trap_yet_again ; undefined behavior
2361 store i32 %trap, i32* @g ; Trap value conceptually stored to memory.
2362 %trap2 = load i32* @g ; Returns a trap value, not just undef.
2364 volatile store i32 %trap, i32* @g ; External observation; undefined behavior.
2366 %narrowaddr = bitcast i32* @g to i16*
2367 %wideaddr = bitcast i32* @g to i64*
2368 %trap3 = load 16* %narrowaddr ; Returns a trap value.
2369 %trap4 = load i64* %widaddr ; Returns a trap value.
2371 %cmp = icmp i32 slt %trap, 0 ; Returns a trap value.
2372 %br i1 %cmp, %true, %end ; Branch to either destination.
2375 volatile store i32 0, i32* @g ; This is control-dependent on %cmp, so
2376 ; it has undefined behavior.
2380 %p = phi i32 [ 0, %entry ], [ 1, %true ]
2381 ; Both edges into this PHI are
2382 ; control-dependent on %cmp, so this
2383 ; always results in a trap value.
2385 volatile store i32 0, i32* @g ; %end is control-equivalent to %entry
2386 ; so this is defined (ignoring earlier
2387 ; undefined behavior in this example).
2392 <!-- ======================================================================= -->
2393 <div class="doc_subsection"><a name="blockaddress">Addresses of Basic
2395 <div class="doc_text">
2397 <p><b><tt>blockaddress(@function, %block)</tt></b></p>
2399 <p>The '<tt>blockaddress</tt>' constant computes the address of the specified
2400 basic block in the specified function, and always has an i8* type. Taking
2401 the address of the entry block is illegal.</p>
2403 <p>This value only has defined behavior when used as an operand to the
2404 '<a href="#i_indirectbr"><tt>indirectbr</tt></a>' instruction or for comparisons
2405 against null. Pointer equality tests between labels addresses is undefined
2406 behavior - though, again, comparison against null is ok, and no label is
2407 equal to the null pointer. This may also be passed around as an opaque
2408 pointer sized value as long as the bits are not inspected. This allows
2409 <tt>ptrtoint</tt> and arithmetic to be performed on these values so long as
2410 the original value is reconstituted before the <tt>indirectbr</tt>.</p>
2412 <p>Finally, some targets may provide defined semantics when
2413 using the value as the operand to an inline assembly, but that is target
2420 <!-- ======================================================================= -->
2421 <div class="doc_subsection"><a name="constantexprs">Constant Expressions</a>
2424 <div class="doc_text">
2426 <p>Constant expressions are used to allow expressions involving other constants
2427 to be used as constants. Constant expressions may be of
2428 any <a href="#t_firstclass">first class</a> type and may involve any LLVM
2429 operation that does not have side effects (e.g. load and call are not
2430 supported). The following is the syntax for constant expressions:</p>
2433 <dt><b><tt>trunc (CST to TYPE)</tt></b></dt>
2434 <dd>Truncate a constant to another type. The bit size of CST must be larger
2435 than the bit size of TYPE. Both types must be integers.</dd>
2437 <dt><b><tt>zext (CST to TYPE)</tt></b></dt>
2438 <dd>Zero extend a constant to another type. The bit size of CST must be
2439 smaller than the bit size of TYPE. Both types must be integers.</dd>
2441 <dt><b><tt>sext (CST to TYPE)</tt></b></dt>
2442 <dd>Sign extend a constant to another type. The bit size of CST must be
2443 smaller than the bit size of TYPE. Both types must be integers.</dd>
2445 <dt><b><tt>fptrunc (CST to TYPE)</tt></b></dt>
2446 <dd>Truncate a floating point constant to another floating point type. The
2447 size of CST must be larger than the size of TYPE. Both types must be
2448 floating point.</dd>
2450 <dt><b><tt>fpext (CST to TYPE)</tt></b></dt>
2451 <dd>Floating point extend a constant to another type. The size of CST must be
2452 smaller or equal to the size of TYPE. Both types must be floating
2455 <dt><b><tt>fptoui (CST to TYPE)</tt></b></dt>
2456 <dd>Convert a floating point constant to the corresponding unsigned integer
2457 constant. TYPE must be a scalar or vector integer type. CST must be of
2458 scalar or vector floating point type. Both CST and TYPE must be scalars,
2459 or vectors of the same number of elements. If the value won't fit in the
2460 integer type, the results are undefined.</dd>
2462 <dt><b><tt>fptosi (CST to TYPE)</tt></b></dt>
2463 <dd>Convert a floating point constant to the corresponding signed integer
2464 constant. TYPE must be a scalar or vector integer type. CST must be of
2465 scalar or vector floating point type. Both CST and TYPE must be scalars,
2466 or vectors of the same number of elements. If the value won't fit in the
2467 integer type, the results are undefined.</dd>
2469 <dt><b><tt>uitofp (CST to TYPE)</tt></b></dt>
2470 <dd>Convert an unsigned integer constant to the corresponding floating point
2471 constant. TYPE must be a scalar or vector floating point type. CST must be
2472 of scalar or vector integer type. Both CST and TYPE must be scalars, or
2473 vectors of the same number of elements. If the value won't fit in the
2474 floating point type, the results are undefined.</dd>
2476 <dt><b><tt>sitofp (CST to TYPE)</tt></b></dt>
2477 <dd>Convert a signed integer constant to the corresponding floating point
2478 constant. TYPE must be a scalar or vector floating point type. CST must be
2479 of scalar or vector integer type. Both CST and TYPE must be scalars, or
2480 vectors of the same number of elements. If the value won't fit in the
2481 floating point type, the results are undefined.</dd>
2483 <dt><b><tt>ptrtoint (CST to TYPE)</tt></b></dt>
2484 <dd>Convert a pointer typed constant to the corresponding integer constant
2485 <tt>TYPE</tt> must be an integer type. <tt>CST</tt> must be of pointer
2486 type. The <tt>CST</tt> value is zero extended, truncated, or unchanged to
2487 make it fit in <tt>TYPE</tt>.</dd>
2489 <dt><b><tt>inttoptr (CST to TYPE)</tt></b></dt>
2490 <dd>Convert a integer constant to a pointer constant. TYPE must be a pointer
2491 type. CST must be of integer type. The CST value is zero extended,
2492 truncated, or unchanged to make it fit in a pointer size. This one is
2493 <i>really</i> dangerous!</dd>
2495 <dt><b><tt>bitcast (CST to TYPE)</tt></b></dt>
2496 <dd>Convert a constant, CST, to another TYPE. The constraints of the operands
2497 are the same as those for the <a href="#i_bitcast">bitcast
2498 instruction</a>.</dd>
2500 <dt><b><tt>getelementptr (CSTPTR, IDX0, IDX1, ...)</tt></b></dt>
2501 <dt><b><tt>getelementptr inbounds (CSTPTR, IDX0, IDX1, ...)</tt></b></dt>
2502 <dd>Perform the <a href="#i_getelementptr">getelementptr operation</a> on
2503 constants. As with the <a href="#i_getelementptr">getelementptr</a>
2504 instruction, the index list may have zero or more indexes, which are
2505 required to make sense for the type of "CSTPTR".</dd>
2507 <dt><b><tt>select (COND, VAL1, VAL2)</tt></b></dt>
2508 <dd>Perform the <a href="#i_select">select operation</a> on constants.</dd>
2510 <dt><b><tt>icmp COND (VAL1, VAL2)</tt></b></dt>
2511 <dd>Performs the <a href="#i_icmp">icmp operation</a> on constants.</dd>
2513 <dt><b><tt>fcmp COND (VAL1, VAL2)</tt></b></dt>
2514 <dd>Performs the <a href="#i_fcmp">fcmp operation</a> on constants.</dd>
2516 <dt><b><tt>extractelement (VAL, IDX)</tt></b></dt>
2517 <dd>Perform the <a href="#i_extractelement">extractelement operation</a> on
2520 <dt><b><tt>insertelement (VAL, ELT, IDX)</tt></b></dt>
2521 <dd>Perform the <a href="#i_insertelement">insertelement operation</a> on
2524 <dt><b><tt>shufflevector (VEC1, VEC2, IDXMASK)</tt></b></dt>
2525 <dd>Perform the <a href="#i_shufflevector">shufflevector operation</a> on
2528 <dt><b><tt>extractvalue (VAL, IDX0, IDX1, ...)</tt></b></dt>
2529 <dd>Perform the <a href="#i_extractvalue">extractvalue operation</a> on
2530 constants. The index list is interpreted in a similar manner as indices in
2531 a '<a href="#i_getelementptr">getelementptr</a>' operation. At least one
2532 index value must be specified.</dd>
2534 <dt><b><tt>insertvalue (VAL, ELT, IDX0, IDX1, ...)</tt></b></dt>
2535 <dd>Perform the <a href="#i_insertvalue">insertvalue operation</a> on
2536 constants. The index list is interpreted in a similar manner as indices in
2537 a '<a href="#i_getelementptr">getelementptr</a>' operation. At least one
2538 index value must be specified.</dd>
2540 <dt><b><tt>OPCODE (LHS, RHS)</tt></b></dt>
2541 <dd>Perform the specified operation of the LHS and RHS constants. OPCODE may
2542 be any of the <a href="#binaryops">binary</a>
2543 or <a href="#bitwiseops">bitwise binary</a> operations. The constraints
2544 on operands are the same as those for the corresponding instruction
2545 (e.g. no bitwise operations on floating point values are allowed).</dd>
2550 <!-- *********************************************************************** -->
2551 <div class="doc_section"> <a name="othervalues">Other Values</a> </div>
2552 <!-- *********************************************************************** -->
2554 <!-- ======================================================================= -->
2555 <div class="doc_subsection">
2556 <a name="inlineasm">Inline Assembler Expressions</a>
2559 <div class="doc_text">
2561 <p>LLVM supports inline assembler expressions (as opposed
2562 to <a href="#moduleasm"> Module-Level Inline Assembly</a>) through the use of
2563 a special value. This value represents the inline assembler as a string
2564 (containing the instructions to emit), a list of operand constraints (stored
2565 as a string), a flag that indicates whether or not the inline asm
2566 expression has side effects, and a flag indicating whether the function
2567 containing the asm needs to align its stack conservatively. An example
2568 inline assembler expression is:</p>
2570 <pre class="doc_code">
2571 i32 (i32) asm "bswap $0", "=r,r"
2574 <p>Inline assembler expressions may <b>only</b> be used as the callee operand of
2575 a <a href="#i_call"><tt>call</tt> instruction</a>. Thus, typically we
2578 <pre class="doc_code">
2579 %X = call i32 asm "<a href="#int_bswap">bswap</a> $0", "=r,r"(i32 %Y)
2582 <p>Inline asms with side effects not visible in the constraint list must be
2583 marked as having side effects. This is done through the use of the
2584 '<tt>sideeffect</tt>' keyword, like so:</p>
2586 <pre class="doc_code">
2587 call void asm sideeffect "eieio", ""()
2590 <p>In some cases inline asms will contain code that will not work unless the
2591 stack is aligned in some way, such as calls or SSE instructions on x86,
2592 yet will not contain code that does that alignment within the asm.
2593 The compiler should make conservative assumptions about what the asm might
2594 contain and should generate its usual stack alignment code in the prologue
2595 if the '<tt>alignstack</tt>' keyword is present:</p>
2597 <pre class="doc_code">
2598 call void asm alignstack "eieio", ""()
2601 <p>If both keywords appear the '<tt>sideeffect</tt>' keyword must come
2604 <p>TODO: The format of the asm and constraints string still need to be
2605 documented here. Constraints on what can be done (e.g. duplication, moving,
2606 etc need to be documented). This is probably best done by reference to
2607 another document that covers inline asm from a holistic perspective.</p>
2610 <div class="doc_subsubsection">
2611 <a name="inlineasm_md">Inline Asm Metadata</a>
2614 <div class="doc_text">
2616 <p>The call instructions that wrap inline asm nodes may have a "!srcloc" MDNode
2617 attached to it that contains a constant integer. If present, the code
2618 generator will use the integer as the location cookie value when report
2619 errors through the LLVMContext error reporting mechanisms. This allows a
2620 front-end to correlate backend errors that occur with inline asm back to the
2621 source code that produced it. For example:</p>
2623 <pre class="doc_code">
2624 call void asm sideeffect "something bad", ""()<b>, !srcloc !42</b>
2626 !42 = !{ i32 1234567 }
2629 <p>It is up to the front-end to make sense of the magic numbers it places in the
2634 <!-- ======================================================================= -->
2635 <div class="doc_subsection"><a name="metadata">Metadata Nodes and Metadata
2639 <div class="doc_text">
2641 <p>LLVM IR allows metadata to be attached to instructions in the program that
2642 can convey extra information about the code to the optimizers and code
2643 generator. One example application of metadata is source-level debug
2644 information. There are two metadata primitives: strings and nodes. All
2645 metadata has the <tt>metadata</tt> type and is identified in syntax by a
2646 preceding exclamation point ('<tt>!</tt>').</p>
2648 <p>A metadata string is a string surrounded by double quotes. It can contain
2649 any character by escaping non-printable characters with "\xx" where "xx" is
2650 the two digit hex code. For example: "<tt>!"test\00"</tt>".</p>
2652 <p>Metadata nodes are represented with notation similar to structure constants
2653 (a comma separated list of elements, surrounded by braces and preceded by an
2654 exclamation point). For example: "<tt>!{ metadata !"test\00", i32
2655 10}</tt>". Metadata nodes can have any values as their operand.</p>
2657 <p>A <a href="#namedmetadatastructure">named metadata</a> is a collection of
2658 metadata nodes, which can be looked up in the module symbol table. For
2659 example: "<tt>!foo = metadata !{!4, !3}</tt>".
2661 <p>Metadata can be used as function arguments. Here <tt>llvm.dbg.value</tt>
2662 function is using two metadata arguments.</p>
2664 <pre class="doc_code">
2665 call void @llvm.dbg.value(metadata !24, i64 0, metadata !25)
2668 <p>Metadata can be attached with an instruction. Here metadata <tt>!21</tt> is
2669 attached with <tt>add</tt> instruction using <tt>!dbg</tt> identifier.</p>
2671 <pre class="doc_code">
2672 %indvar.next = add i64 %indvar, 1, !dbg !21
2677 <!-- *********************************************************************** -->
2678 <div class="doc_section">
2679 <a name="intrinsic_globals">Intrinsic Global Variables</a>
2681 <!-- *********************************************************************** -->
2683 <p>LLVM has a number of "magic" global variables that contain data that affect
2684 code generation or other IR semantics. These are documented here. All globals
2685 of this sort should have a section specified as "<tt>llvm.metadata</tt>". This
2686 section and all globals that start with "<tt>llvm.</tt>" are reserved for use
2689 <!-- ======================================================================= -->
2690 <div class="doc_subsection">
2691 <a name="intg_used">The '<tt>llvm.used</tt>' Global Variable</a>
2694 <div class="doc_text">
2696 <p>The <tt>@llvm.used</tt> global is an array with i8* element type which has <a
2697 href="#linkage_appending">appending linkage</a>. This array contains a list of
2698 pointers to global variables and functions which may optionally have a pointer
2699 cast formed of bitcast or getelementptr. For example, a legal use of it is:</p>
2705 @llvm.used = appending global [2 x i8*] [
2707 i8* bitcast (i32* @Y to i8*)
2708 ], section "llvm.metadata"
2711 <p>If a global variable appears in the <tt>@llvm.used</tt> list, then the
2712 compiler, assembler, and linker are required to treat the symbol as if there is
2713 a reference to the global that it cannot see. For example, if a variable has
2714 internal linkage and no references other than that from the <tt>@llvm.used</tt>
2715 list, it cannot be deleted. This is commonly used to represent references from
2716 inline asms and other things the compiler cannot "see", and corresponds to
2717 "attribute((used))" in GNU C.</p>
2719 <p>On some targets, the code generator must emit a directive to the assembler or
2720 object file to prevent the assembler and linker from molesting the symbol.</p>
2724 <!-- ======================================================================= -->
2725 <div class="doc_subsection">
2726 <a name="intg_compiler_used">The '<tt>llvm.compiler.used</tt>' Global Variable</a>
2729 <div class="doc_text">
2731 <p>The <tt>@llvm.compiler.used</tt> directive is the same as the
2732 <tt>@llvm.used</tt> directive, except that it only prevents the compiler from
2733 touching the symbol. On targets that support it, this allows an intelligent
2734 linker to optimize references to the symbol without being impeded as it would be
2735 by <tt>@llvm.used</tt>.</p>
2737 <p>This is a rare construct that should only be used in rare circumstances, and
2738 should not be exposed to source languages.</p>
2742 <!-- ======================================================================= -->
2743 <div class="doc_subsection">
2744 <a name="intg_global_ctors">The '<tt>llvm.global_ctors</tt>' Global Variable</a>
2747 <div class="doc_text">
2749 %0 = type { i32, void ()* }
2750 @llvm.global_ctors = appending global [1 x %0] [%0 { i32 65535, void ()* @ctor }]
2752 <p>The <tt>@llvm.global_ctors</tt> array contains a list of constructor functions and associated priorities. The functions referenced by this array will be called in ascending order of priority (i.e. lowest first) when the module is loaded. The order of functions with the same priority is not defined.
2757 <!-- ======================================================================= -->
2758 <div class="doc_subsection">
2759 <a name="intg_global_dtors">The '<tt>llvm.global_dtors</tt>' Global Variable</a>
2762 <div class="doc_text">
2764 %0 = type { i32, void ()* }
2765 @llvm.global_dtors = appending global [1 x %0] [%0 { i32 65535, void ()* @dtor }]
2768 <p>The <tt>@llvm.global_dtors</tt> array contains a list of destructor functions and associated priorities. The functions referenced by this array will be called in descending order of priority (i.e. highest first) when the module is loaded. The order of functions with the same priority is not defined.
2774 <!-- *********************************************************************** -->
2775 <div class="doc_section"> <a name="instref">Instruction Reference</a> </div>
2776 <!-- *********************************************************************** -->
2778 <div class="doc_text">
2780 <p>The LLVM instruction set consists of several different classifications of
2781 instructions: <a href="#terminators">terminator
2782 instructions</a>, <a href="#binaryops">binary instructions</a>,
2783 <a href="#bitwiseops">bitwise binary instructions</a>,
2784 <a href="#memoryops">memory instructions</a>, and
2785 <a href="#otherops">other instructions</a>.</p>
2789 <!-- ======================================================================= -->
2790 <div class="doc_subsection"> <a name="terminators">Terminator
2791 Instructions</a> </div>
2793 <div class="doc_text">
2795 <p>As mentioned <a href="#functionstructure">previously</a>, every basic block
2796 in a program ends with a "Terminator" instruction, which indicates which
2797 block should be executed after the current block is finished. These
2798 terminator instructions typically yield a '<tt>void</tt>' value: they produce
2799 control flow, not values (the one exception being the
2800 '<a href="#i_invoke"><tt>invoke</tt></a>' instruction).</p>
2802 <p>There are seven different terminator instructions: the
2803 '<a href="#i_ret"><tt>ret</tt></a>' instruction, the
2804 '<a href="#i_br"><tt>br</tt></a>' instruction, the
2805 '<a href="#i_switch"><tt>switch</tt></a>' instruction, the
2806 '<a href="#i_indirectbr">'<tt>indirectbr</tt></a>' Instruction, the
2807 '<a href="#i_invoke"><tt>invoke</tt></a>' instruction, the
2808 '<a href="#i_unwind"><tt>unwind</tt></a>' instruction, and the
2809 '<a href="#i_unreachable"><tt>unreachable</tt></a>' instruction.</p>
2813 <!-- _______________________________________________________________________ -->
2814 <div class="doc_subsubsection"> <a name="i_ret">'<tt>ret</tt>'
2815 Instruction</a> </div>
2817 <div class="doc_text">
2821 ret <type> <value> <i>; Return a value from a non-void function</i>
2822 ret void <i>; Return from void function</i>
2826 <p>The '<tt>ret</tt>' instruction is used to return control flow (and optionally
2827 a value) from a function back to the caller.</p>
2829 <p>There are two forms of the '<tt>ret</tt>' instruction: one that returns a
2830 value and then causes control flow, and one that just causes control flow to
2834 <p>The '<tt>ret</tt>' instruction optionally accepts a single argument, the
2835 return value. The type of the return value must be a
2836 '<a href="#t_firstclass">first class</a>' type.</p>
2838 <p>A function is not <a href="#wellformed">well formed</a> if it it has a
2839 non-void return type and contains a '<tt>ret</tt>' instruction with no return
2840 value or a return value with a type that does not match its type, or if it
2841 has a void return type and contains a '<tt>ret</tt>' instruction with a
2845 <p>When the '<tt>ret</tt>' instruction is executed, control flow returns back to
2846 the calling function's context. If the caller is a
2847 "<a href="#i_call"><tt>call</tt></a>" instruction, execution continues at the
2848 instruction after the call. If the caller was an
2849 "<a href="#i_invoke"><tt>invoke</tt></a>" instruction, execution continues at
2850 the beginning of the "normal" destination block. If the instruction returns
2851 a value, that value shall set the call or invoke instruction's return
2856 ret i32 5 <i>; Return an integer value of 5</i>
2857 ret void <i>; Return from a void function</i>
2858 ret { i32, i8 } { i32 4, i8 2 } <i>; Return a struct of values 4 and 2</i>
2862 <!-- _______________________________________________________________________ -->
2863 <div class="doc_subsubsection"> <a name="i_br">'<tt>br</tt>' Instruction</a> </div>
2865 <div class="doc_text">
2869 br i1 <cond>, label <iftrue>, label <iffalse><br> br label <dest> <i>; Unconditional branch</i>
2873 <p>The '<tt>br</tt>' instruction is used to cause control flow to transfer to a
2874 different basic block in the current function. There are two forms of this
2875 instruction, corresponding to a conditional branch and an unconditional
2879 <p>The conditional branch form of the '<tt>br</tt>' instruction takes a single
2880 '<tt>i1</tt>' value and two '<tt>label</tt>' values. The unconditional form
2881 of the '<tt>br</tt>' instruction takes a single '<tt>label</tt>' value as a
2885 <p>Upon execution of a conditional '<tt>br</tt>' instruction, the '<tt>i1</tt>'
2886 argument is evaluated. If the value is <tt>true</tt>, control flows to the
2887 '<tt>iftrue</tt>' <tt>label</tt> argument. If "cond" is <tt>false</tt>,
2888 control flows to the '<tt>iffalse</tt>' <tt>label</tt> argument.</p>
2893 %cond = <a href="#i_icmp">icmp</a> eq i32 %a, %b
2894 br i1 %cond, label %IfEqual, label %IfUnequal
2896 <a href="#i_ret">ret</a> i32 1
2898 <a href="#i_ret">ret</a> i32 0
2903 <!-- _______________________________________________________________________ -->
2904 <div class="doc_subsubsection">
2905 <a name="i_switch">'<tt>switch</tt>' Instruction</a>
2908 <div class="doc_text">
2912 switch <intty> <value>, label <defaultdest> [ <intty> <val>, label <dest> ... ]
2916 <p>The '<tt>switch</tt>' instruction is used to transfer control flow to one of
2917 several different places. It is a generalization of the '<tt>br</tt>'
2918 instruction, allowing a branch to occur to one of many possible
2922 <p>The '<tt>switch</tt>' instruction uses three parameters: an integer
2923 comparison value '<tt>value</tt>', a default '<tt>label</tt>' destination,
2924 and an array of pairs of comparison value constants and '<tt>label</tt>'s.
2925 The table is not allowed to contain duplicate constant entries.</p>
2928 <p>The <tt>switch</tt> instruction specifies a table of values and
2929 destinations. When the '<tt>switch</tt>' instruction is executed, this table
2930 is searched for the given value. If the value is found, control flow is
2931 transferred to the corresponding destination; otherwise, control flow is
2932 transferred to the default destination.</p>
2934 <h5>Implementation:</h5>
2935 <p>Depending on properties of the target machine and the particular
2936 <tt>switch</tt> instruction, this instruction may be code generated in
2937 different ways. For example, it could be generated as a series of chained
2938 conditional branches or with a lookup table.</p>
2942 <i>; Emulate a conditional br instruction</i>
2943 %Val = <a href="#i_zext">zext</a> i1 %value to i32
2944 switch i32 %Val, label %truedest [ i32 0, label %falsedest ]
2946 <i>; Emulate an unconditional br instruction</i>
2947 switch i32 0, label %dest [ ]
2949 <i>; Implement a jump table:</i>
2950 switch i32 %val, label %otherwise [ i32 0, label %onzero
2952 i32 2, label %ontwo ]
2958 <!-- _______________________________________________________________________ -->
2959 <div class="doc_subsubsection">
2960 <a name="i_indirectbr">'<tt>indirectbr</tt>' Instruction</a>
2963 <div class="doc_text">
2967 indirectbr <somety>* <address>, [ label <dest1>, label <dest2>, ... ]
2972 <p>The '<tt>indirectbr</tt>' instruction implements an indirect branch to a label
2973 within the current function, whose address is specified by
2974 "<tt>address</tt>". Address must be derived from a <a
2975 href="#blockaddress">blockaddress</a> constant.</p>
2979 <p>The '<tt>address</tt>' argument is the address of the label to jump to. The
2980 rest of the arguments indicate the full set of possible destinations that the
2981 address may point to. Blocks are allowed to occur multiple times in the
2982 destination list, though this isn't particularly useful.</p>
2984 <p>This destination list is required so that dataflow analysis has an accurate
2985 understanding of the CFG.</p>
2989 <p>Control transfers to the block specified in the address argument. All
2990 possible destination blocks must be listed in the label list, otherwise this
2991 instruction has undefined behavior. This implies that jumps to labels
2992 defined in other functions have undefined behavior as well.</p>
2994 <h5>Implementation:</h5>
2996 <p>This is typically implemented with a jump through a register.</p>
3000 indirectbr i8* %Addr, [ label %bb1, label %bb2, label %bb3 ]
3006 <!-- _______________________________________________________________________ -->
3007 <div class="doc_subsubsection">
3008 <a name="i_invoke">'<tt>invoke</tt>' Instruction</a>
3011 <div class="doc_text">
3015 <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>]
3016 to label <normal label> unwind label <exception label>
3020 <p>The '<tt>invoke</tt>' instruction causes control to transfer to a specified
3021 function, with the possibility of control flow transfer to either the
3022 '<tt>normal</tt>' label or the '<tt>exception</tt>' label. If the callee
3023 function returns with the "<tt><a href="#i_ret">ret</a></tt>" instruction,
3024 control flow will return to the "normal" label. If the callee (or any
3025 indirect callees) returns with the "<a href="#i_unwind"><tt>unwind</tt></a>"
3026 instruction, control is interrupted and continued at the dynamically nearest
3027 "exception" label.</p>
3030 <p>This instruction requires several arguments:</p>
3033 <li>The optional "cconv" marker indicates which <a href="#callingconv">calling
3034 convention</a> the call should use. If none is specified, the call
3035 defaults to using C calling conventions.</li>
3037 <li>The optional <a href="#paramattrs">Parameter Attributes</a> list for
3038 return values. Only '<tt>zeroext</tt>', '<tt>signext</tt>', and
3039 '<tt>inreg</tt>' attributes are valid here.</li>
3041 <li>'<tt>ptr to function ty</tt>': shall be the signature of the pointer to
3042 function value being invoked. In most cases, this is a direct function
3043 invocation, but indirect <tt>invoke</tt>s are just as possible, branching
3044 off an arbitrary pointer to function value.</li>
3046 <li>'<tt>function ptr val</tt>': An LLVM value containing a pointer to a
3047 function to be invoked. </li>
3049 <li>'<tt>function args</tt>': argument list whose types match the function
3050 signature argument types and parameter attributes. All arguments must be
3051 of <a href="#t_firstclass">first class</a> type. If the function
3052 signature indicates the function accepts a variable number of arguments,
3053 the extra arguments can be specified.</li>
3055 <li>'<tt>normal label</tt>': the label reached when the called function
3056 executes a '<tt><a href="#i_ret">ret</a></tt>' instruction. </li>
3058 <li>'<tt>exception label</tt>': the label reached when a callee returns with
3059 the <a href="#i_unwind"><tt>unwind</tt></a> instruction. </li>
3061 <li>The optional <a href="#fnattrs">function attributes</a> list. Only
3062 '<tt>noreturn</tt>', '<tt>nounwind</tt>', '<tt>readonly</tt>' and
3063 '<tt>readnone</tt>' attributes are valid here.</li>
3067 <p>This instruction is designed to operate as a standard
3068 '<tt><a href="#i_call">call</a></tt>' instruction in most regards. The
3069 primary difference is that it establishes an association with a label, which
3070 is used by the runtime library to unwind the stack.</p>
3072 <p>This instruction is used in languages with destructors to ensure that proper
3073 cleanup is performed in the case of either a <tt>longjmp</tt> or a thrown
3074 exception. Additionally, this is important for implementation of
3075 '<tt>catch</tt>' clauses in high-level languages that support them.</p>
3077 <p>For the purposes of the SSA form, the definition of the value returned by the
3078 '<tt>invoke</tt>' instruction is deemed to occur on the edge from the current
3079 block to the "normal" label. If the callee unwinds then no return value is
3082 <p>Note that the code generator does not yet completely support unwind, and
3083 that the invoke/unwind semantics are likely to change in future versions.</p>
3087 %retval = invoke i32 @Test(i32 15) to label %Continue
3088 unwind label %TestCleanup <i>; {i32}:retval set</i>
3089 %retval = invoke <a href="#callingconv">coldcc</a> i32 %Testfnptr(i32 15) to label %Continue
3090 unwind label %TestCleanup <i>; {i32}:retval set</i>
3095 <!-- _______________________________________________________________________ -->
3097 <div class="doc_subsubsection"> <a name="i_unwind">'<tt>unwind</tt>'
3098 Instruction</a> </div>
3100 <div class="doc_text">
3108 <p>The '<tt>unwind</tt>' instruction unwinds the stack, continuing control flow
3109 at the first callee in the dynamic call stack which used
3110 an <a href="#i_invoke"><tt>invoke</tt></a> instruction to perform the call.
3111 This is primarily used to implement exception handling.</p>
3114 <p>The '<tt>unwind</tt>' instruction causes execution of the current function to
3115 immediately halt. The dynamic call stack is then searched for the
3116 first <a href="#i_invoke"><tt>invoke</tt></a> instruction on the call stack.
3117 Once found, execution continues at the "exceptional" destination block
3118 specified by the <tt>invoke</tt> instruction. If there is no <tt>invoke</tt>
3119 instruction in the dynamic call chain, undefined behavior results.</p>
3121 <p>Note that the code generator does not yet completely support unwind, and
3122 that the invoke/unwind semantics are likely to change in future versions.</p>
3126 <!-- _______________________________________________________________________ -->
3128 <div class="doc_subsubsection"> <a name="i_unreachable">'<tt>unreachable</tt>'
3129 Instruction</a> </div>
3131 <div class="doc_text">
3139 <p>The '<tt>unreachable</tt>' instruction has no defined semantics. This
3140 instruction is used to inform the optimizer that a particular portion of the
3141 code is not reachable. This can be used to indicate that the code after a
3142 no-return function cannot be reached, and other facts.</p>
3145 <p>The '<tt>unreachable</tt>' instruction has no defined semantics.</p>
3149 <!-- ======================================================================= -->
3150 <div class="doc_subsection"> <a name="binaryops">Binary Operations</a> </div>
3152 <div class="doc_text">
3154 <p>Binary operators are used to do most of the computation in a program. They
3155 require two operands of the same type, execute an operation on them, and
3156 produce a single value. The operands might represent multiple data, as is
3157 the case with the <a href="#t_vector">vector</a> data type. The result value
3158 has the same type as its operands.</p>
3160 <p>There are several different binary operators:</p>
3164 <!-- _______________________________________________________________________ -->
3165 <div class="doc_subsubsection">
3166 <a name="i_add">'<tt>add</tt>' Instruction</a>
3169 <div class="doc_text">
3173 <result> = add <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3174 <result> = add nuw <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3175 <result> = add nsw <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3176 <result> = add nuw nsw <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3180 <p>The '<tt>add</tt>' instruction returns the sum of its two operands.</p>
3183 <p>The two arguments to the '<tt>add</tt>' instruction must
3184 be <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of
3185 integer values. Both arguments must have identical types.</p>
3188 <p>The value produced is the integer sum of the two operands.</p>
3190 <p>If the sum has unsigned overflow, the result returned is the mathematical
3191 result modulo 2<sup>n</sup>, where n is the bit width of the result.</p>
3193 <p>Because LLVM integers use a two's complement representation, this instruction
3194 is appropriate for both signed and unsigned integers.</p>
3196 <p><tt>nuw</tt> and <tt>nsw</tt> stand for "No Unsigned Wrap"
3197 and "No Signed Wrap", respectively. If the <tt>nuw</tt> and/or
3198 <tt>nsw</tt> keywords are present, the result value of the <tt>add</tt>
3199 is a <a href="#trapvalues">trap value</a> if unsigned and/or signed overflow,
3200 respectively, occurs.</p>
3204 <result> = add i32 4, %var <i>; yields {i32}:result = 4 + %var</i>
3209 <!-- _______________________________________________________________________ -->
3210 <div class="doc_subsubsection">
3211 <a name="i_fadd">'<tt>fadd</tt>' Instruction</a>
3214 <div class="doc_text">
3218 <result> = fadd <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3222 <p>The '<tt>fadd</tt>' instruction returns the sum of its two operands.</p>
3225 <p>The two arguments to the '<tt>fadd</tt>' instruction must be
3226 <a href="#t_floating">floating point</a> or <a href="#t_vector">vector</a> of
3227 floating point values. Both arguments must have identical types.</p>
3230 <p>The value produced is the floating point sum of the two operands.</p>
3234 <result> = fadd float 4.0, %var <i>; yields {float}:result = 4.0 + %var</i>
3239 <!-- _______________________________________________________________________ -->
3240 <div class="doc_subsubsection">
3241 <a name="i_sub">'<tt>sub</tt>' Instruction</a>
3244 <div class="doc_text">
3248 <result> = sub <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3249 <result> = sub nuw <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3250 <result> = sub nsw <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3251 <result> = sub nuw nsw <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3255 <p>The '<tt>sub</tt>' instruction returns the difference of its two
3258 <p>Note that the '<tt>sub</tt>' instruction is used to represent the
3259 '<tt>neg</tt>' instruction present in most other intermediate
3260 representations.</p>
3263 <p>The two arguments to the '<tt>sub</tt>' instruction must
3264 be <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of
3265 integer values. Both arguments must have identical types.</p>
3268 <p>The value produced is the integer difference of the two operands.</p>
3270 <p>If the difference has unsigned overflow, the result returned is the
3271 mathematical result modulo 2<sup>n</sup>, where n is the bit width of the
3274 <p>Because LLVM integers use a two's complement representation, this instruction
3275 is appropriate for both signed and unsigned integers.</p>
3277 <p><tt>nuw</tt> and <tt>nsw</tt> stand for "No Unsigned Wrap"
3278 and "No Signed Wrap", respectively. If the <tt>nuw</tt> and/or
3279 <tt>nsw</tt> keywords are present, the result value of the <tt>sub</tt>
3280 is a <a href="#trapvalues">trap value</a> if unsigned and/or signed overflow,
3281 respectively, occurs.</p>
3285 <result> = sub i32 4, %var <i>; yields {i32}:result = 4 - %var</i>
3286 <result> = sub i32 0, %val <i>; yields {i32}:result = -%var</i>
3291 <!-- _______________________________________________________________________ -->
3292 <div class="doc_subsubsection">
3293 <a name="i_fsub">'<tt>fsub</tt>' Instruction</a>
3296 <div class="doc_text">
3300 <result> = fsub <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3304 <p>The '<tt>fsub</tt>' instruction returns the difference of its two
3307 <p>Note that the '<tt>fsub</tt>' instruction is used to represent the
3308 '<tt>fneg</tt>' instruction present in most other intermediate
3309 representations.</p>
3312 <p>The two arguments to the '<tt>fsub</tt>' instruction must be
3313 <a href="#t_floating">floating point</a> or <a href="#t_vector">vector</a> of
3314 floating point values. Both arguments must have identical types.</p>
3317 <p>The value produced is the floating point difference of the two operands.</p>
3321 <result> = fsub float 4.0, %var <i>; yields {float}:result = 4.0 - %var</i>
3322 <result> = fsub float -0.0, %val <i>; yields {float}:result = -%var</i>
3327 <!-- _______________________________________________________________________ -->
3328 <div class="doc_subsubsection">
3329 <a name="i_mul">'<tt>mul</tt>' Instruction</a>
3332 <div class="doc_text">
3336 <result> = mul <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3337 <result> = mul nuw <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3338 <result> = mul nsw <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3339 <result> = mul nuw nsw <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3343 <p>The '<tt>mul</tt>' instruction returns the product of its two operands.</p>
3346 <p>The two arguments to the '<tt>mul</tt>' instruction must
3347 be <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of
3348 integer values. Both arguments must have identical types.</p>
3351 <p>The value produced is the integer product of the two operands.</p>
3353 <p>If the result of the multiplication has unsigned overflow, the result
3354 returned is the mathematical result modulo 2<sup>n</sup>, where n is the bit
3355 width of the result.</p>
3357 <p>Because LLVM integers use a two's complement representation, and the result
3358 is the same width as the operands, this instruction returns the correct
3359 result for both signed and unsigned integers. If a full product
3360 (e.g. <tt>i32</tt>x<tt>i32</tt>-><tt>i64</tt>) is needed, the operands should
3361 be sign-extended or zero-extended as appropriate to the width of the full
3364 <p><tt>nuw</tt> and <tt>nsw</tt> stand for "No Unsigned Wrap"
3365 and "No Signed Wrap", respectively. If the <tt>nuw</tt> and/or
3366 <tt>nsw</tt> keywords are present, the result value of the <tt>mul</tt>
3367 is a <a href="#trapvalues">trap value</a> if unsigned and/or signed overflow,
3368 respectively, occurs.</p>
3372 <result> = mul i32 4, %var <i>; yields {i32}:result = 4 * %var</i>
3377 <!-- _______________________________________________________________________ -->
3378 <div class="doc_subsubsection">
3379 <a name="i_fmul">'<tt>fmul</tt>' Instruction</a>
3382 <div class="doc_text">
3386 <result> = fmul <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3390 <p>The '<tt>fmul</tt>' instruction returns the product of its two operands.</p>
3393 <p>The two arguments to the '<tt>fmul</tt>' instruction must be
3394 <a href="#t_floating">floating point</a> or <a href="#t_vector">vector</a> of
3395 floating point values. Both arguments must have identical types.</p>
3398 <p>The value produced is the floating point product of the two operands.</p>
3402 <result> = fmul float 4.0, %var <i>; yields {float}:result = 4.0 * %var</i>
3407 <!-- _______________________________________________________________________ -->
3408 <div class="doc_subsubsection"> <a name="i_udiv">'<tt>udiv</tt>' Instruction
3411 <div class="doc_text">
3415 <result> = udiv <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3419 <p>The '<tt>udiv</tt>' instruction returns the quotient of its two operands.</p>
3422 <p>The two arguments to the '<tt>udiv</tt>' instruction must be
3423 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
3424 values. Both arguments must have identical types.</p>
3427 <p>The value produced is the unsigned integer quotient of the two operands.</p>
3429 <p>Note that unsigned integer division and signed integer division are distinct
3430 operations; for signed integer division, use '<tt>sdiv</tt>'.</p>
3432 <p>Division by zero leads to undefined behavior.</p>
3436 <result> = udiv i32 4, %var <i>; yields {i32}:result = 4 / %var</i>
3441 <!-- _______________________________________________________________________ -->
3442 <div class="doc_subsubsection"> <a name="i_sdiv">'<tt>sdiv</tt>' Instruction
3445 <div class="doc_text">
3449 <result> = sdiv <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3450 <result> = sdiv exact <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3454 <p>The '<tt>sdiv</tt>' instruction returns the quotient of its two operands.</p>
3457 <p>The two arguments to the '<tt>sdiv</tt>' instruction must be
3458 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
3459 values. Both arguments must have identical types.</p>
3462 <p>The value produced is the signed integer quotient of the two operands rounded
3465 <p>Note that signed integer division and unsigned integer division are distinct
3466 operations; for unsigned integer division, use '<tt>udiv</tt>'.</p>
3468 <p>Division by zero leads to undefined behavior. Overflow also leads to
3469 undefined behavior; this is a rare case, but can occur, for example, by doing
3470 a 32-bit division of -2147483648 by -1.</p>
3472 <p>If the <tt>exact</tt> keyword is present, the result value of the
3473 <tt>sdiv</tt> is a <a href="#trapvalues">trap value</a> if the result would
3478 <result> = sdiv i32 4, %var <i>; yields {i32}:result = 4 / %var</i>
3483 <!-- _______________________________________________________________________ -->
3484 <div class="doc_subsubsection"> <a name="i_fdiv">'<tt>fdiv</tt>'
3485 Instruction</a> </div>
3487 <div class="doc_text">
3491 <result> = fdiv <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3495 <p>The '<tt>fdiv</tt>' instruction returns the quotient of its two operands.</p>
3498 <p>The two arguments to the '<tt>fdiv</tt>' instruction must be
3499 <a href="#t_floating">floating point</a> or <a href="#t_vector">vector</a> of
3500 floating point values. Both arguments must have identical types.</p>
3503 <p>The value produced is the floating point quotient of the two operands.</p>
3507 <result> = fdiv float 4.0, %var <i>; yields {float}:result = 4.0 / %var</i>
3512 <!-- _______________________________________________________________________ -->
3513 <div class="doc_subsubsection"> <a name="i_urem">'<tt>urem</tt>' Instruction</a>
3516 <div class="doc_text">
3520 <result> = urem <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3524 <p>The '<tt>urem</tt>' instruction returns the remainder from the unsigned
3525 division of its two arguments.</p>
3528 <p>The two arguments to the '<tt>urem</tt>' instruction must be
3529 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
3530 values. Both arguments must have identical types.</p>
3533 <p>This instruction returns the unsigned integer <i>remainder</i> of a division.
3534 This instruction always performs an unsigned division to get the
3537 <p>Note that unsigned integer remainder and signed integer remainder are
3538 distinct operations; for signed integer remainder, use '<tt>srem</tt>'.</p>
3540 <p>Taking the remainder of a division by zero leads to undefined behavior.</p>
3544 <result> = urem i32 4, %var <i>; yields {i32}:result = 4 % %var</i>
3549 <!-- _______________________________________________________________________ -->
3550 <div class="doc_subsubsection">
3551 <a name="i_srem">'<tt>srem</tt>' Instruction</a>
3554 <div class="doc_text">
3558 <result> = srem <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3562 <p>The '<tt>srem</tt>' instruction returns the remainder from the signed
3563 division of its two operands. This instruction can also take
3564 <a href="#t_vector">vector</a> versions of the values in which case the
3565 elements must be integers.</p>
3568 <p>The two arguments to the '<tt>srem</tt>' instruction must be
3569 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
3570 values. Both arguments must have identical types.</p>
3573 <p>This instruction returns the <i>remainder</i> of a division (where the result
3574 has the same sign as the dividend, <tt>op1</tt>), not the <i>modulo</i>
3575 operator (where the result has the same sign as the divisor, <tt>op2</tt>) of
3576 a value. For more information about the difference,
3577 see <a href="http://mathforum.org/dr.math/problems/anne.4.28.99.html">The
3578 Math Forum</a>. For a table of how this is implemented in various languages,
3579 please see <a href="http://en.wikipedia.org/wiki/Modulo_operation">
3580 Wikipedia: modulo operation</a>.</p>
3582 <p>Note that signed integer remainder and unsigned integer remainder are
3583 distinct operations; for unsigned integer remainder, use '<tt>urem</tt>'.</p>
3585 <p>Taking the remainder of a division by zero leads to undefined behavior.
3586 Overflow also leads to undefined behavior; this is a rare case, but can
3587 occur, for example, by taking the remainder of a 32-bit division of
3588 -2147483648 by -1. (The remainder doesn't actually overflow, but this rule
3589 lets srem be implemented using instructions that return both the result of
3590 the division and the remainder.)</p>
3594 <result> = srem i32 4, %var <i>; yields {i32}:result = 4 % %var</i>
3599 <!-- _______________________________________________________________________ -->
3600 <div class="doc_subsubsection">
3601 <a name="i_frem">'<tt>frem</tt>' Instruction</a> </div>
3603 <div class="doc_text">
3607 <result> = frem <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3611 <p>The '<tt>frem</tt>' instruction returns the remainder from the division of
3612 its two operands.</p>
3615 <p>The two arguments to the '<tt>frem</tt>' instruction must be
3616 <a href="#t_floating">floating point</a> or <a href="#t_vector">vector</a> of
3617 floating point values. Both arguments must have identical types.</p>
3620 <p>This instruction returns the <i>remainder</i> of a division. The remainder
3621 has the same sign as the dividend.</p>
3625 <result> = frem float 4.0, %var <i>; yields {float}:result = 4.0 % %var</i>
3630 <!-- ======================================================================= -->
3631 <div class="doc_subsection"> <a name="bitwiseops">Bitwise Binary
3632 Operations</a> </div>
3634 <div class="doc_text">
3636 <p>Bitwise binary operators are used to do various forms of bit-twiddling in a
3637 program. They are generally very efficient instructions and can commonly be
3638 strength reduced from other instructions. They require two operands of the
3639 same type, execute an operation on them, and produce a single value. The
3640 resulting value is the same type as its operands.</p>
3644 <!-- _______________________________________________________________________ -->
3645 <div class="doc_subsubsection"> <a name="i_shl">'<tt>shl</tt>'
3646 Instruction</a> </div>
3648 <div class="doc_text">
3652 <result> = shl <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3656 <p>The '<tt>shl</tt>' instruction returns the first operand shifted to the left
3657 a specified number of bits.</p>
3660 <p>Both arguments to the '<tt>shl</tt>' instruction must be the
3661 same <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of
3662 integer type. '<tt>op2</tt>' is treated as an unsigned value.</p>
3665 <p>The value produced is <tt>op1</tt> * 2<sup><tt>op2</tt></sup> mod
3666 2<sup>n</sup>, where <tt>n</tt> is the width of the result. If <tt>op2</tt>
3667 is (statically or dynamically) negative or equal to or larger than the number
3668 of bits in <tt>op1</tt>, the result is undefined. If the arguments are
3669 vectors, each vector element of <tt>op1</tt> is shifted by the corresponding
3670 shift amount in <tt>op2</tt>.</p>
3674 <result> = shl i32 4, %var <i>; yields {i32}: 4 << %var</i>
3675 <result> = shl i32 4, 2 <i>; yields {i32}: 16</i>
3676 <result> = shl i32 1, 10 <i>; yields {i32}: 1024</i>
3677 <result> = shl i32 1, 32 <i>; undefined</i>
3678 <result> = shl <2 x i32> < i32 1, i32 1>, < i32 1, i32 2> <i>; yields: result=<2 x i32> < i32 2, i32 4></i>
3683 <!-- _______________________________________________________________________ -->
3684 <div class="doc_subsubsection"> <a name="i_lshr">'<tt>lshr</tt>'
3685 Instruction</a> </div>
3687 <div class="doc_text">
3691 <result> = lshr <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3695 <p>The '<tt>lshr</tt>' instruction (logical shift right) returns the first
3696 operand shifted to the right a specified number of bits with zero fill.</p>
3699 <p>Both arguments to the '<tt>lshr</tt>' instruction must be the same
3700 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
3701 type. '<tt>op2</tt>' is treated as an unsigned value.</p>
3704 <p>This instruction always performs a logical shift right operation. The most
3705 significant bits of the result will be filled with zero bits after the shift.
3706 If <tt>op2</tt> is (statically or dynamically) equal to or larger than the
3707 number of bits in <tt>op1</tt>, the result is undefined. If the arguments are
3708 vectors, each vector element of <tt>op1</tt> is shifted by the corresponding
3709 shift amount in <tt>op2</tt>.</p>
3713 <result> = lshr i32 4, 1 <i>; yields {i32}:result = 2</i>
3714 <result> = lshr i32 4, 2 <i>; yields {i32}:result = 1</i>
3715 <result> = lshr i8 4, 3 <i>; yields {i8}:result = 0</i>
3716 <result> = lshr i8 -2, 1 <i>; yields {i8}:result = 0x7FFFFFFF </i>
3717 <result> = lshr i32 1, 32 <i>; undefined</i>
3718 <result> = lshr <2 x i32> < i32 -2, i32 4>, < i32 1, i32 2> <i>; yields: result=<2 x i32> < i32 0x7FFFFFFF, i32 1></i>
3723 <!-- _______________________________________________________________________ -->
3724 <div class="doc_subsubsection"> <a name="i_ashr">'<tt>ashr</tt>'
3725 Instruction</a> </div>
3726 <div class="doc_text">
3730 <result> = ashr <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3734 <p>The '<tt>ashr</tt>' instruction (arithmetic shift right) returns the first
3735 operand shifted to the right a specified number of bits with sign
3739 <p>Both arguments to the '<tt>ashr</tt>' instruction must be the same
3740 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
3741 type. '<tt>op2</tt>' is treated as an unsigned value.</p>
3744 <p>This instruction always performs an arithmetic shift right operation, The
3745 most significant bits of the result will be filled with the sign bit
3746 of <tt>op1</tt>. If <tt>op2</tt> is (statically or dynamically) equal to or
3747 larger than the number of bits in <tt>op1</tt>, the result is undefined. If
3748 the arguments are vectors, each vector element of <tt>op1</tt> is shifted by
3749 the corresponding shift amount in <tt>op2</tt>.</p>
3753 <result> = ashr i32 4, 1 <i>; yields {i32}:result = 2</i>
3754 <result> = ashr i32 4, 2 <i>; yields {i32}:result = 1</i>
3755 <result> = ashr i8 4, 3 <i>; yields {i8}:result = 0</i>
3756 <result> = ashr i8 -2, 1 <i>; yields {i8}:result = -1</i>
3757 <result> = ashr i32 1, 32 <i>; undefined</i>
3758 <result> = ashr <2 x i32> < i32 -2, i32 4>, < i32 1, i32 3> <i>; yields: result=<2 x i32> < i32 -1, i32 0></i>
3763 <!-- _______________________________________________________________________ -->
3764 <div class="doc_subsubsection"> <a name="i_and">'<tt>and</tt>'
3765 Instruction</a> </div>
3767 <div class="doc_text">
3771 <result> = and <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3775 <p>The '<tt>and</tt>' instruction returns the bitwise logical and of its two
3779 <p>The two arguments to the '<tt>and</tt>' instruction must be
3780 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
3781 values. Both arguments must have identical types.</p>
3784 <p>The truth table used for the '<tt>and</tt>' instruction is:</p>
3786 <table border="1" cellspacing="0" cellpadding="4">
3818 <result> = and i32 4, %var <i>; yields {i32}:result = 4 & %var</i>
3819 <result> = and i32 15, 40 <i>; yields {i32}:result = 8</i>
3820 <result> = and i32 4, 8 <i>; yields {i32}:result = 0</i>
3823 <!-- _______________________________________________________________________ -->
3824 <div class="doc_subsubsection"> <a name="i_or">'<tt>or</tt>' Instruction</a> </div>
3826 <div class="doc_text">
3830 <result> = or <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3834 <p>The '<tt>or</tt>' instruction returns the bitwise logical inclusive or of its
3838 <p>The two arguments to the '<tt>or</tt>' instruction must be
3839 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
3840 values. Both arguments must have identical types.</p>
3843 <p>The truth table used for the '<tt>or</tt>' instruction is:</p>
3845 <table border="1" cellspacing="0" cellpadding="4">
3877 <result> = or i32 4, %var <i>; yields {i32}:result = 4 | %var</i>
3878 <result> = or i32 15, 40 <i>; yields {i32}:result = 47</i>
3879 <result> = or i32 4, 8 <i>; yields {i32}:result = 12</i>
3884 <!-- _______________________________________________________________________ -->
3885 <div class="doc_subsubsection"> <a name="i_xor">'<tt>xor</tt>'
3886 Instruction</a> </div>
3888 <div class="doc_text">
3892 <result> = xor <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3896 <p>The '<tt>xor</tt>' instruction returns the bitwise logical exclusive or of
3897 its two operands. The <tt>xor</tt> is used to implement the "one's
3898 complement" operation, which is the "~" operator in C.</p>
3901 <p>The two arguments to the '<tt>xor</tt>' instruction must be
3902 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
3903 values. Both arguments must have identical types.</p>
3906 <p>The truth table used for the '<tt>xor</tt>' instruction is:</p>
3908 <table border="1" cellspacing="0" cellpadding="4">
3940 <result> = xor i32 4, %var <i>; yields {i32}:result = 4 ^ %var</i>
3941 <result> = xor i32 15, 40 <i>; yields {i32}:result = 39</i>
3942 <result> = xor i32 4, 8 <i>; yields {i32}:result = 12</i>
3943 <result> = xor i32 %V, -1 <i>; yields {i32}:result = ~%V</i>
3948 <!-- ======================================================================= -->
3949 <div class="doc_subsection">
3950 <a name="vectorops">Vector Operations</a>
3953 <div class="doc_text">
3955 <p>LLVM supports several instructions to represent vector operations in a
3956 target-independent manner. These instructions cover the element-access and
3957 vector-specific operations needed to process vectors effectively. While LLVM
3958 does directly support these vector operations, many sophisticated algorithms
3959 will want to use target-specific intrinsics to take full advantage of a
3960 specific target.</p>
3964 <!-- _______________________________________________________________________ -->
3965 <div class="doc_subsubsection">
3966 <a name="i_extractelement">'<tt>extractelement</tt>' Instruction</a>
3969 <div class="doc_text">
3973 <result> = extractelement <n x <ty>> <val>, i32 <idx> <i>; yields <ty></i>
3977 <p>The '<tt>extractelement</tt>' instruction extracts a single scalar element
3978 from a vector at a specified index.</p>
3982 <p>The first operand of an '<tt>extractelement</tt>' instruction is a value
3983 of <a href="#t_vector">vector</a> type. The second operand is an index
3984 indicating the position from which to extract the element. The index may be
3988 <p>The result is a scalar of the same type as the element type of
3989 <tt>val</tt>. Its value is the value at position <tt>idx</tt> of
3990 <tt>val</tt>. If <tt>idx</tt> exceeds the length of <tt>val</tt>, the
3991 results are undefined.</p>
3995 <result> = extractelement <4 x i32> %vec, i32 0 <i>; yields i32</i>
4000 <!-- _______________________________________________________________________ -->
4001 <div class="doc_subsubsection">
4002 <a name="i_insertelement">'<tt>insertelement</tt>' Instruction</a>
4005 <div class="doc_text">
4009 <result> = insertelement <n x <ty>> <val>, <ty> <elt>, i32 <idx> <i>; yields <n x <ty>></i>
4013 <p>The '<tt>insertelement</tt>' instruction inserts a scalar element into a
4014 vector at a specified index.</p>
4017 <p>The first operand of an '<tt>insertelement</tt>' instruction is a value
4018 of <a href="#t_vector">vector</a> type. The second operand is a scalar value
4019 whose type must equal the element type of the first operand. The third
4020 operand is an index indicating the position at which to insert the value.
4021 The index may be a variable.</p>
4024 <p>The result is a vector of the same type as <tt>val</tt>. Its element values
4025 are those of <tt>val</tt> except at position <tt>idx</tt>, where it gets the
4026 value <tt>elt</tt>. If <tt>idx</tt> exceeds the length of <tt>val</tt>, the
4027 results are undefined.</p>
4031 <result> = insertelement <4 x i32> %vec, i32 1, i32 0 <i>; yields <4 x i32></i>
4036 <!-- _______________________________________________________________________ -->
4037 <div class="doc_subsubsection">
4038 <a name="i_shufflevector">'<tt>shufflevector</tt>' Instruction</a>
4041 <div class="doc_text">
4045 <result> = shufflevector <n x <ty>> <v1>, <n x <ty>> <v2>, <m x i32> <mask> <i>; yields <m x <ty>></i>
4049 <p>The '<tt>shufflevector</tt>' instruction constructs a permutation of elements
4050 from two input vectors, returning a vector with the same element type as the
4051 input and length that is the same as the shuffle mask.</p>
4054 <p>The first two operands of a '<tt>shufflevector</tt>' instruction are vectors
4055 with types that match each other. The third argument is a shuffle mask whose
4056 element type is always 'i32'. The result of the instruction is a vector
4057 whose length is the same as the shuffle mask and whose element type is the
4058 same as the element type of the first two operands.</p>
4060 <p>The shuffle mask operand is required to be a constant vector with either
4061 constant integer or undef values.</p>
4064 <p>The elements of the two input vectors are numbered from left to right across
4065 both of the vectors. The shuffle mask operand specifies, for each element of
4066 the result vector, which element of the two input vectors the result element
4067 gets. The element selector may be undef (meaning "don't care") and the
4068 second operand may be undef if performing a shuffle from only one vector.</p>
4072 <result> = shufflevector <4 x i32> %v1, <4 x i32> %v2,
4073 <4 x i32> <i32 0, i32 4, i32 1, i32 5> <i>; yields <4 x i32></i>
4074 <result> = shufflevector <4 x i32> %v1, <4 x i32> undef,
4075 <4 x i32> <i32 0, i32 1, i32 2, i32 3> <i>; yields <4 x i32></i> - Identity shuffle.
4076 <result> = shufflevector <8 x i32> %v1, <8 x i32> undef,
4077 <4 x i32> <i32 0, i32 1, i32 2, i32 3> <i>; yields <4 x i32></i>
4078 <result> = shufflevector <4 x i32> %v1, <4 x i32> %v2,
4079 <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>
4084 <!-- ======================================================================= -->
4085 <div class="doc_subsection">
4086 <a name="aggregateops">Aggregate Operations</a>
4089 <div class="doc_text">
4091 <p>LLVM supports several instructions for working with
4092 <a href="#t_aggregate">aggregate</a> values.</p>
4096 <!-- _______________________________________________________________________ -->
4097 <div class="doc_subsubsection">
4098 <a name="i_extractvalue">'<tt>extractvalue</tt>' Instruction</a>
4101 <div class="doc_text">
4105 <result> = extractvalue <aggregate type> <val>, <idx>{, <idx>}*
4109 <p>The '<tt>extractvalue</tt>' instruction extracts the value of a member field
4110 from an <a href="#t_aggregate">aggregate</a> value.</p>
4113 <p>The first operand of an '<tt>extractvalue</tt>' instruction is a value
4114 of <a href="#t_struct">struct</a> or
4115 <a href="#t_array">array</a> type. The operands are constant indices to
4116 specify which value to extract in a similar manner as indices in a
4117 '<tt><a href="#i_getelementptr">getelementptr</a></tt>' instruction.</p>
4120 <p>The result is the value at the position in the aggregate specified by the
4125 <result> = extractvalue {i32, float} %agg, 0 <i>; yields i32</i>
4130 <!-- _______________________________________________________________________ -->
4131 <div class="doc_subsubsection">
4132 <a name="i_insertvalue">'<tt>insertvalue</tt>' Instruction</a>
4135 <div class="doc_text">
4139 <result> = insertvalue <aggregate type> <val>, <ty> <elt>, <idx> <i>; yields <aggregate type></i>
4143 <p>The '<tt>insertvalue</tt>' instruction inserts a value into a member field
4144 in an <a href="#t_aggregate">aggregate</a> value.</p>
4147 <p>The first operand of an '<tt>insertvalue</tt>' instruction is a value
4148 of <a href="#t_struct">struct</a> or
4149 <a href="#t_array">array</a> type. The second operand is a first-class
4150 value to insert. The following operands are constant indices indicating
4151 the position at which to insert the value in a similar manner as indices in a
4152 '<tt><a href="#i_getelementptr">getelementptr</a></tt>' instruction. The
4153 value to insert must have the same type as the value identified by the
4157 <p>The result is an aggregate of the same type as <tt>val</tt>. Its value is
4158 that of <tt>val</tt> except that the value at the position specified by the
4159 indices is that of <tt>elt</tt>.</p>
4163 %agg1 = insertvalue {i32, float} undef, i32 1, 0 <i>; yields {i32 1, float undef}</i>
4164 %agg2 = insertvalue {i32, float} %agg1, float %val, 1 <i>; yields {i32 1, float %val}</i>
4170 <!-- ======================================================================= -->
4171 <div class="doc_subsection">
4172 <a name="memoryops">Memory Access and Addressing Operations</a>
4175 <div class="doc_text">
4177 <p>A key design point of an SSA-based representation is how it represents
4178 memory. In LLVM, no memory locations are in SSA form, which makes things
4179 very simple. This section describes how to read, write, and allocate
4184 <!-- _______________________________________________________________________ -->
4185 <div class="doc_subsubsection">
4186 <a name="i_alloca">'<tt>alloca</tt>' Instruction</a>
4189 <div class="doc_text">
4193 <result> = alloca <type>[, <ty> <NumElements>][, align <alignment>] <i>; yields {type*}:result</i>
4197 <p>The '<tt>alloca</tt>' instruction allocates memory on the stack frame of the
4198 currently executing function, to be automatically released when this function
4199 returns to its caller. The object is always allocated in the generic address
4200 space (address space zero).</p>
4203 <p>The '<tt>alloca</tt>' instruction
4204 allocates <tt>sizeof(<type>)*NumElements</tt> bytes of memory on the
4205 runtime stack, returning a pointer of the appropriate type to the program.
4206 If "NumElements" is specified, it is the number of elements allocated,
4207 otherwise "NumElements" is defaulted to be one. If a constant alignment is
4208 specified, the value result of the allocation is guaranteed to be aligned to
4209 at least that boundary. If not specified, or if zero, the target can choose
4210 to align the allocation on any convenient boundary compatible with the
4213 <p>'<tt>type</tt>' may be any sized type.</p>
4216 <p>Memory is allocated; a pointer is returned. The operation is undefined if
4217 there is insufficient stack space for the allocation. '<tt>alloca</tt>'d
4218 memory is automatically released when the function returns. The
4219 '<tt>alloca</tt>' instruction is commonly used to represent automatic
4220 variables that must have an address available. When the function returns
4221 (either with the <tt><a href="#i_ret">ret</a></tt>
4222 or <tt><a href="#i_unwind">unwind</a></tt> instructions), the memory is
4223 reclaimed. Allocating zero bytes is legal, but the result is undefined.</p>
4227 %ptr = alloca i32 <i>; yields {i32*}:ptr</i>
4228 %ptr = alloca i32, i32 4 <i>; yields {i32*}:ptr</i>
4229 %ptr = alloca i32, i32 4, align 1024 <i>; yields {i32*}:ptr</i>
4230 %ptr = alloca i32, align 1024 <i>; yields {i32*}:ptr</i>
4235 <!-- _______________________________________________________________________ -->
4236 <div class="doc_subsubsection"> <a name="i_load">'<tt>load</tt>'
4237 Instruction</a> </div>
4239 <div class="doc_text">
4243 <result> = load <ty>* <pointer>[, align <alignment>][, !nontemporal !<index>]
4244 <result> = volatile load <ty>* <pointer>[, align <alignment>][, !nontemporal !<index>]
4245 !<index> = !{ i32 1 }
4249 <p>The '<tt>load</tt>' instruction is used to read from memory.</p>
4252 <p>The argument to the '<tt>load</tt>' instruction specifies the memory address
4253 from which to load. The pointer must point to
4254 a <a href="#t_firstclass">first class</a> type. If the <tt>load</tt> is
4255 marked as <tt>volatile</tt>, then the optimizer is not allowed to modify the
4256 number or order of execution of this <tt>load</tt> with other <a
4257 href="#volatile">volatile operations</a>.</p>
4259 <p>The optional constant <tt>align</tt> argument specifies the alignment of the
4260 operation (that is, the alignment of the memory address). A value of 0 or an
4261 omitted <tt>align</tt> argument means that the operation has the preferential
4262 alignment for the target. It is the responsibility of the code emitter to
4263 ensure that the alignment information is correct. Overestimating the
4264 alignment results in undefined behavior. Underestimating the alignment may
4265 produce less efficient code. An alignment of 1 is always safe.</p>
4267 <p>The optional <tt>!nontemporal</tt> metadata must reference a single
4268 metatadata name <index> corresponding to a metadata node with
4269 one <tt>i32</tt> entry of value 1. The existence of
4270 the <tt>!nontemporal</tt> metatadata on the instruction tells the optimizer
4271 and code generator that this load is not expected to be reused in the cache.
4272 The code generator may select special instructions to save cache bandwidth,
4273 such as the <tt>MOVNT</tt> instruction on x86.</p>
4276 <p>The location of memory pointed to is loaded. If the value being loaded is of
4277 scalar type then the number of bytes read does not exceed the minimum number
4278 of bytes needed to hold all bits of the type. For example, loading an
4279 <tt>i24</tt> reads at most three bytes. When loading a value of a type like
4280 <tt>i20</tt> with a size that is not an integral number of bytes, the result
4281 is undefined if the value was not originally written using a store of the
4286 %ptr = <a href="#i_alloca">alloca</a> i32 <i>; yields {i32*}:ptr</i>
4287 <a href="#i_store">store</a> i32 3, i32* %ptr <i>; yields {void}</i>
4288 %val = load i32* %ptr <i>; yields {i32}:val = i32 3</i>
4293 <!-- _______________________________________________________________________ -->
4294 <div class="doc_subsubsection"> <a name="i_store">'<tt>store</tt>'
4295 Instruction</a> </div>
4297 <div class="doc_text">
4301 store <ty> <value>, <ty>* <pointer>[, align <alignment>][, !nontemporal !<index>] <i>; yields {void}</i>
4302 volatile store <ty> <value>, <ty>* <pointer>[, align <alignment>][, !nontemporal !<index>] <i>; yields {void}</i>
4306 <p>The '<tt>store</tt>' instruction is used to write to memory.</p>
4309 <p>There are two arguments to the '<tt>store</tt>' instruction: a value to store
4310 and an address at which to store it. The type of the
4311 '<tt><pointer></tt>' operand must be a pointer to
4312 the <a href="#t_firstclass">first class</a> type of the
4313 '<tt><value></tt>' operand. If the <tt>store</tt> is marked as
4314 <tt>volatile</tt>, then the optimizer is not allowed to modify the number or
4315 order of execution of this <tt>store</tt> with other <a
4316 href="#volatile">volatile operations</a>.</p>
4318 <p>The optional constant "align" argument specifies the alignment of the
4319 operation (that is, the alignment of the memory address). A value of 0 or an
4320 omitted "align" argument means that the operation has the preferential
4321 alignment for the target. It is the responsibility of the code emitter to
4322 ensure that the alignment information is correct. Overestimating the
4323 alignment results in an undefined behavior. Underestimating the alignment may
4324 produce less efficient code. An alignment of 1 is always safe.</p>
4326 <p>The optional !nontemporal metadata must reference a single metatadata
4327 name <index> corresponding to a metadata node with one i32 entry of
4328 value 1. The existence of the !nontemporal metatadata on the
4329 instruction tells the optimizer and code generator that this load is
4330 not expected to be reused in the cache. The code generator may
4331 select special instructions to save cache bandwidth, such as the
4332 MOVNT instruction on x86.</p>
4336 <p>The contents of memory are updated to contain '<tt><value></tt>' at the
4337 location specified by the '<tt><pointer></tt>' operand. If
4338 '<tt><value></tt>' is of scalar type then the number of bytes written
4339 does not exceed the minimum number of bytes needed to hold all bits of the
4340 type. For example, storing an <tt>i24</tt> writes at most three bytes. When
4341 writing a value of a type like <tt>i20</tt> with a size that is not an
4342 integral number of bytes, it is unspecified what happens to the extra bits
4343 that do not belong to the type, but they will typically be overwritten.</p>
4347 %ptr = <a href="#i_alloca">alloca</a> i32 <i>; yields {i32*}:ptr</i>
4348 store i32 3, i32* %ptr <i>; yields {void}</i>
4349 %val = <a href="#i_load">load</a> i32* %ptr <i>; yields {i32}:val = i32 3</i>
4354 <!-- _______________________________________________________________________ -->
4355 <div class="doc_subsubsection">
4356 <a name="i_getelementptr">'<tt>getelementptr</tt>' Instruction</a>
4359 <div class="doc_text">
4363 <result> = getelementptr <pty>* <ptrval>{, <ty> <idx>}*
4364 <result> = getelementptr inbounds <pty>* <ptrval>{, <ty> <idx>}*
4368 <p>The '<tt>getelementptr</tt>' instruction is used to get the address of a
4369 subelement of an <a href="#t_aggregate">aggregate</a> data structure.
4370 It performs address calculation only and does not access memory.</p>
4373 <p>The first argument is always a pointer, and forms the basis of the
4374 calculation. The remaining arguments are indices that indicate which of the
4375 elements of the aggregate object are indexed. The interpretation of each
4376 index is dependent on the type being indexed into. The first index always
4377 indexes the pointer value given as the first argument, the second index
4378 indexes a value of the type pointed to (not necessarily the value directly
4379 pointed to, since the first index can be non-zero), etc. The first type
4380 indexed into must be a pointer value, subsequent types can be arrays,
4381 vectors, and structs. Note that subsequent types being indexed into
4382 can never be pointers, since that would require loading the pointer before
4383 continuing calculation.</p>
4385 <p>The type of each index argument depends on the type it is indexing into.
4386 When indexing into a (optionally packed) structure, only <tt>i32</tt>
4387 integer <b>constants</b> are allowed. When indexing into an array, pointer
4388 or vector, integers of any width are allowed, and they are not required to be
4391 <p>For example, let's consider a C code fragment and how it gets compiled to
4394 <pre class="doc_code">
4406 int *foo(struct ST *s) {
4407 return &s[1].Z.B[5][13];
4411 <p>The LLVM code generated by the GCC frontend is:</p>
4413 <pre class="doc_code">
4414 %RT = <a href="#namedtypes">type</a> { i8 , [10 x [20 x i32]], i8 }
4415 %ST = <a href="#namedtypes">type</a> { i32, double, %RT }
4417 define i32* @foo(%ST* %s) {
4419 %reg = getelementptr %ST* %s, i32 1, i32 2, i32 1, i32 5, i32 13
4425 <p>In the example above, the first index is indexing into the '<tt>%ST*</tt>'
4426 type, which is a pointer, yielding a '<tt>%ST</tt>' = '<tt>{ i32, double, %RT
4427 }</tt>' type, a structure. The second index indexes into the third element
4428 of the structure, yielding a '<tt>%RT</tt>' = '<tt>{ i8 , [10 x [20 x i32]],
4429 i8 }</tt>' type, another structure. The third index indexes into the second
4430 element of the structure, yielding a '<tt>[10 x [20 x i32]]</tt>' type, an
4431 array. The two dimensions of the array are subscripted into, yielding an
4432 '<tt>i32</tt>' type. The '<tt>getelementptr</tt>' instruction returns a
4433 pointer to this element, thus computing a value of '<tt>i32*</tt>' type.</p>
4435 <p>Note that it is perfectly legal to index partially through a structure,
4436 returning a pointer to an inner element. Because of this, the LLVM code for
4437 the given testcase is equivalent to:</p>
4440 define i32* @foo(%ST* %s) {
4441 %t1 = getelementptr %ST* %s, i32 1 <i>; yields %ST*:%t1</i>
4442 %t2 = getelementptr %ST* %t1, i32 0, i32 2 <i>; yields %RT*:%t2</i>
4443 %t3 = getelementptr %RT* %t2, i32 0, i32 1 <i>; yields [10 x [20 x i32]]*:%t3</i>
4444 %t4 = getelementptr [10 x [20 x i32]]* %t3, i32 0, i32 5 <i>; yields [20 x i32]*:%t4</i>
4445 %t5 = getelementptr [20 x i32]* %t4, i32 0, i32 13 <i>; yields i32*:%t5</i>
4450 <p>If the <tt>inbounds</tt> keyword is present, the result value of the
4451 <tt>getelementptr</tt> is a <a href="#trapvalues">trap value</a> if the
4452 base pointer is not an <i>in bounds</i> address of an allocated object,
4453 or if any of the addresses that would be formed by successive addition of
4454 the offsets implied by the indices to the base address with infinitely
4455 precise arithmetic are not an <i>in bounds</i> address of that allocated
4456 object. The <i>in bounds</i> addresses for an allocated object are all
4457 the addresses that point into the object, plus the address one byte past
4460 <p>If the <tt>inbounds</tt> keyword is not present, the offsets are added to
4461 the base address with silently-wrapping two's complement arithmetic, and
4462 the result value of the <tt>getelementptr</tt> may be outside the object
4463 pointed to by the base pointer. The result value may not necessarily be
4464 used to access memory though, even if it happens to point into allocated
4465 storage. See the <a href="#pointeraliasing">Pointer Aliasing Rules</a>
4466 section for more information.</p>
4468 <p>The getelementptr instruction is often confusing. For some more insight into
4469 how it works, see <a href="GetElementPtr.html">the getelementptr FAQ</a>.</p>
4473 <i>; yields [12 x i8]*:aptr</i>
4474 %aptr = getelementptr {i32, [12 x i8]}* %saptr, i64 0, i32 1
4475 <i>; yields i8*:vptr</i>
4476 %vptr = getelementptr {i32, <2 x i8>}* %svptr, i64 0, i32 1, i32 1
4477 <i>; yields i8*:eptr</i>
4478 %eptr = getelementptr [12 x i8]* %aptr, i64 0, i32 1
4479 <i>; yields i32*:iptr</i>
4480 %iptr = getelementptr [10 x i32]* @arr, i16 0, i16 0
4485 <!-- ======================================================================= -->
4486 <div class="doc_subsection"> <a name="convertops">Conversion Operations</a>
4489 <div class="doc_text">
4491 <p>The instructions in this category are the conversion instructions (casting)
4492 which all take a single operand and a type. They perform various bit
4493 conversions on the operand.</p>
4497 <!-- _______________________________________________________________________ -->
4498 <div class="doc_subsubsection">
4499 <a name="i_trunc">'<tt>trunc .. to</tt>' Instruction</a>
4501 <div class="doc_text">
4505 <result> = trunc <ty> <value> to <ty2> <i>; yields ty2</i>
4509 <p>The '<tt>trunc</tt>' instruction truncates its operand to the
4510 type <tt>ty2</tt>.</p>
4513 <p>The '<tt>trunc</tt>' instruction takes a <tt>value</tt> to trunc, which must
4514 be an <a href="#t_integer">integer</a> type, and a type that specifies the
4515 size and type of the result, which must be
4516 an <a href="#t_integer">integer</a> type. The bit size of <tt>value</tt> must
4517 be larger than the bit size of <tt>ty2</tt>. Equal sized types are not
4521 <p>The '<tt>trunc</tt>' instruction truncates the high order bits
4522 in <tt>value</tt> and converts the remaining bits to <tt>ty2</tt>. Since the
4523 source size must be larger than the destination size, <tt>trunc</tt> cannot
4524 be a <i>no-op cast</i>. It will always truncate bits.</p>
4528 %X = trunc i32 257 to i8 <i>; yields i8:1</i>
4529 %Y = trunc i32 123 to i1 <i>; yields i1:true</i>
4530 %Z = trunc i32 122 to i1 <i>; yields i1:false</i>
4535 <!-- _______________________________________________________________________ -->
4536 <div class="doc_subsubsection">
4537 <a name="i_zext">'<tt>zext .. to</tt>' Instruction</a>
4539 <div class="doc_text">
4543 <result> = zext <ty> <value> to <ty2> <i>; yields ty2</i>
4547 <p>The '<tt>zext</tt>' instruction zero extends its operand to type
4552 <p>The '<tt>zext</tt>' instruction takes a value to cast, which must be of
4553 <a href="#t_integer">integer</a> type, and a type to cast it to, which must
4554 also be of <a href="#t_integer">integer</a> type. The bit size of the
4555 <tt>value</tt> must be smaller than the bit size of the destination type,
4559 <p>The <tt>zext</tt> fills the high order bits of the <tt>value</tt> with zero
4560 bits until it reaches the size of the destination type, <tt>ty2</tt>.</p>
4562 <p>When zero extending from i1, the result will always be either 0 or 1.</p>
4566 %X = zext i32 257 to i64 <i>; yields i64:257</i>
4567 %Y = zext i1 true to i32 <i>; yields i32:1</i>
4572 <!-- _______________________________________________________________________ -->
4573 <div class="doc_subsubsection">
4574 <a name="i_sext">'<tt>sext .. to</tt>' Instruction</a>
4576 <div class="doc_text">
4580 <result> = sext <ty> <value> to <ty2> <i>; yields ty2</i>
4584 <p>The '<tt>sext</tt>' sign extends <tt>value</tt> to the type <tt>ty2</tt>.</p>
4587 <p>The '<tt>sext</tt>' instruction takes a value to cast, which must be of
4588 <a href="#t_integer">integer</a> type, and a type to cast it to, which must
4589 also be of <a href="#t_integer">integer</a> type. The bit size of the
4590 <tt>value</tt> must be smaller than the bit size of the destination type,
4594 <p>The '<tt>sext</tt>' instruction performs a sign extension by copying the sign
4595 bit (highest order bit) of the <tt>value</tt> until it reaches the bit size
4596 of the type <tt>ty2</tt>.</p>
4598 <p>When sign extending from i1, the extension always results in -1 or 0.</p>
4602 %X = sext i8 -1 to i16 <i>; yields i16 :65535</i>
4603 %Y = sext i1 true to i32 <i>; yields i32:-1</i>
4608 <!-- _______________________________________________________________________ -->
4609 <div class="doc_subsubsection">
4610 <a name="i_fptrunc">'<tt>fptrunc .. to</tt>' Instruction</a>
4613 <div class="doc_text">
4617 <result> = fptrunc <ty> <value> to <ty2> <i>; yields ty2</i>
4621 <p>The '<tt>fptrunc</tt>' instruction truncates <tt>value</tt> to type
4625 <p>The '<tt>fptrunc</tt>' instruction takes a <a href="#t_floating">floating
4626 point</a> value to cast and a <a href="#t_floating">floating point</a> type
4627 to cast it to. The size of <tt>value</tt> must be larger than the size of
4628 <tt>ty2</tt>. This implies that <tt>fptrunc</tt> cannot be used to make a
4629 <i>no-op cast</i>.</p>
4632 <p>The '<tt>fptrunc</tt>' instruction truncates a <tt>value</tt> from a larger
4633 <a href="#t_floating">floating point</a> type to a smaller
4634 <a href="#t_floating">floating point</a> type. If the value cannot fit
4635 within the destination type, <tt>ty2</tt>, then the results are
4640 %X = fptrunc double 123.0 to float <i>; yields float:123.0</i>
4641 %Y = fptrunc double 1.0E+300 to float <i>; yields undefined</i>
4646 <!-- _______________________________________________________________________ -->
4647 <div class="doc_subsubsection">
4648 <a name="i_fpext">'<tt>fpext .. to</tt>' Instruction</a>
4650 <div class="doc_text">
4654 <result> = fpext <ty> <value> to <ty2> <i>; yields ty2</i>
4658 <p>The '<tt>fpext</tt>' extends a floating point <tt>value</tt> to a larger
4659 floating point value.</p>
4662 <p>The '<tt>fpext</tt>' instruction takes a
4663 <a href="#t_floating">floating point</a> <tt>value</tt> to cast, and
4664 a <a href="#t_floating">floating point</a> type to cast it to. The source
4665 type must be smaller than the destination type.</p>
4668 <p>The '<tt>fpext</tt>' instruction extends the <tt>value</tt> from a smaller
4669 <a href="#t_floating">floating point</a> type to a larger
4670 <a href="#t_floating">floating point</a> type. The <tt>fpext</tt> cannot be
4671 used to make a <i>no-op cast</i> because it always changes bits. Use
4672 <tt>bitcast</tt> to make a <i>no-op cast</i> for a floating point cast.</p>
4676 %X = fpext float 3.1415 to double <i>; yields double:3.1415</i>
4677 %Y = fpext float 1.0 to float <i>; yields float:1.0 (no-op)</i>
4682 <!-- _______________________________________________________________________ -->
4683 <div class="doc_subsubsection">
4684 <a name="i_fptoui">'<tt>fptoui .. to</tt>' Instruction</a>
4686 <div class="doc_text">
4690 <result> = fptoui <ty> <value> to <ty2> <i>; yields ty2</i>
4694 <p>The '<tt>fptoui</tt>' converts a floating point <tt>value</tt> to its
4695 unsigned integer equivalent of type <tt>ty2</tt>.</p>
4698 <p>The '<tt>fptoui</tt>' instruction takes a value to cast, which must be a
4699 scalar or vector <a href="#t_floating">floating point</a> value, and a type
4700 to cast it to <tt>ty2</tt>, which must be an <a href="#t_integer">integer</a>
4701 type. If <tt>ty</tt> is a vector floating point type, <tt>ty2</tt> must be a
4702 vector integer type with the same number of elements as <tt>ty</tt></p>
4705 <p>The '<tt>fptoui</tt>' instruction converts its
4706 <a href="#t_floating">floating point</a> operand into the nearest (rounding
4707 towards zero) unsigned integer value. If the value cannot fit
4708 in <tt>ty2</tt>, the results are undefined.</p>
4712 %X = fptoui double 123.0 to i32 <i>; yields i32:123</i>
4713 %Y = fptoui float 1.0E+300 to i1 <i>; yields undefined:1</i>
4714 %Z = fptoui float 1.04E+17 to i8 <i>; yields undefined:1</i>
4719 <!-- _______________________________________________________________________ -->
4720 <div class="doc_subsubsection">
4721 <a name="i_fptosi">'<tt>fptosi .. to</tt>' Instruction</a>
4723 <div class="doc_text">
4727 <result> = fptosi <ty> <value> to <ty2> <i>; yields ty2</i>
4731 <p>The '<tt>fptosi</tt>' instruction converts
4732 <a href="#t_floating">floating point</a> <tt>value</tt> to
4733 type <tt>ty2</tt>.</p>
4736 <p>The '<tt>fptosi</tt>' instruction takes a value to cast, which must be a
4737 scalar or vector <a href="#t_floating">floating point</a> value, and a type
4738 to cast it to <tt>ty2</tt>, which must be an <a href="#t_integer">integer</a>
4739 type. If <tt>ty</tt> is a vector floating point type, <tt>ty2</tt> must be a
4740 vector integer type with the same number of elements as <tt>ty</tt></p>
4743 <p>The '<tt>fptosi</tt>' instruction converts its
4744 <a href="#t_floating">floating point</a> operand into the nearest (rounding
4745 towards zero) signed integer value. If the value cannot fit in <tt>ty2</tt>,
4746 the results are undefined.</p>
4750 %X = fptosi double -123.0 to i32 <i>; yields i32:-123</i>
4751 %Y = fptosi float 1.0E-247 to i1 <i>; yields undefined:1</i>
4752 %Z = fptosi float 1.04E+17 to i8 <i>; yields undefined:1</i>
4757 <!-- _______________________________________________________________________ -->
4758 <div class="doc_subsubsection">
4759 <a name="i_uitofp">'<tt>uitofp .. to</tt>' Instruction</a>
4761 <div class="doc_text">
4765 <result> = uitofp <ty> <value> to <ty2> <i>; yields ty2</i>
4769 <p>The '<tt>uitofp</tt>' instruction regards <tt>value</tt> as an unsigned
4770 integer and converts that value to the <tt>ty2</tt> type.</p>
4773 <p>The '<tt>uitofp</tt>' instruction takes a value to cast, which must be a
4774 scalar or vector <a href="#t_integer">integer</a> value, and a type to cast
4775 it to <tt>ty2</tt>, which must be an <a href="#t_floating">floating point</a>
4776 type. If <tt>ty</tt> is a vector integer type, <tt>ty2</tt> must be a vector
4777 floating point type with the same number of elements as <tt>ty</tt></p>
4780 <p>The '<tt>uitofp</tt>' instruction interprets its operand as an unsigned
4781 integer quantity and converts it to the corresponding floating point
4782 value. If the value cannot fit in the floating point value, the results are
4787 %X = uitofp i32 257 to float <i>; yields float:257.0</i>
4788 %Y = uitofp i8 -1 to double <i>; yields double:255.0</i>
4793 <!-- _______________________________________________________________________ -->
4794 <div class="doc_subsubsection">
4795 <a name="i_sitofp">'<tt>sitofp .. to</tt>' Instruction</a>
4797 <div class="doc_text">
4801 <result> = sitofp <ty> <value> to <ty2> <i>; yields ty2</i>
4805 <p>The '<tt>sitofp</tt>' instruction regards <tt>value</tt> as a signed integer
4806 and converts that value to the <tt>ty2</tt> type.</p>
4809 <p>The '<tt>sitofp</tt>' instruction takes a value to cast, which must be a
4810 scalar or vector <a href="#t_integer">integer</a> value, and a type to cast
4811 it to <tt>ty2</tt>, which must be an <a href="#t_floating">floating point</a>
4812 type. If <tt>ty</tt> is a vector integer type, <tt>ty2</tt> must be a vector
4813 floating point type with the same number of elements as <tt>ty</tt></p>
4816 <p>The '<tt>sitofp</tt>' instruction interprets its operand as a signed integer
4817 quantity and converts it to the corresponding floating point value. If the
4818 value cannot fit in the floating point value, the results are undefined.</p>
4822 %X = sitofp i32 257 to float <i>; yields float:257.0</i>
4823 %Y = sitofp i8 -1 to double <i>; yields double:-1.0</i>
4828 <!-- _______________________________________________________________________ -->
4829 <div class="doc_subsubsection">
4830 <a name="i_ptrtoint">'<tt>ptrtoint .. to</tt>' Instruction</a>
4832 <div class="doc_text">
4836 <result> = ptrtoint <ty> <value> to <ty2> <i>; yields ty2</i>
4840 <p>The '<tt>ptrtoint</tt>' instruction converts the pointer <tt>value</tt> to
4841 the integer type <tt>ty2</tt>.</p>
4844 <p>The '<tt>ptrtoint</tt>' instruction takes a <tt>value</tt> to cast, which
4845 must be a <a href="#t_pointer">pointer</a> value, and a type to cast it to
4846 <tt>ty2</tt>, which must be an <a href="#t_integer">integer</a> type.</p>
4849 <p>The '<tt>ptrtoint</tt>' instruction converts <tt>value</tt> to integer type
4850 <tt>ty2</tt> by interpreting the pointer value as an integer and either
4851 truncating or zero extending that value to the size of the integer type. If
4852 <tt>value</tt> is smaller than <tt>ty2</tt> then a zero extension is done. If
4853 <tt>value</tt> is larger than <tt>ty2</tt> then a truncation is done. If they
4854 are the same size, then nothing is done (<i>no-op cast</i>) other than a type
4859 %X = ptrtoint i32* %X to i8 <i>; yields truncation on 32-bit architecture</i>
4860 %Y = ptrtoint i32* %x to i64 <i>; yields zero extension on 32-bit architecture</i>
4865 <!-- _______________________________________________________________________ -->
4866 <div class="doc_subsubsection">
4867 <a name="i_inttoptr">'<tt>inttoptr .. to</tt>' Instruction</a>
4869 <div class="doc_text">
4873 <result> = inttoptr <ty> <value> to <ty2> <i>; yields ty2</i>
4877 <p>The '<tt>inttoptr</tt>' instruction converts an integer <tt>value</tt> to a
4878 pointer type, <tt>ty2</tt>.</p>
4881 <p>The '<tt>inttoptr</tt>' instruction takes an <a href="#t_integer">integer</a>
4882 value to cast, and a type to cast it to, which must be a
4883 <a href="#t_pointer">pointer</a> type.</p>
4886 <p>The '<tt>inttoptr</tt>' instruction converts <tt>value</tt> to type
4887 <tt>ty2</tt> by applying either a zero extension or a truncation depending on
4888 the size of the integer <tt>value</tt>. If <tt>value</tt> is larger than the
4889 size of a pointer then a truncation is done. If <tt>value</tt> is smaller
4890 than the size of a pointer then a zero extension is done. If they are the
4891 same size, nothing is done (<i>no-op cast</i>).</p>
4895 %X = inttoptr i32 255 to i32* <i>; yields zero extension on 64-bit architecture</i>
4896 %Y = inttoptr i32 255 to i32* <i>; yields no-op on 32-bit architecture</i>
4897 %Z = inttoptr i64 0 to i32* <i>; yields truncation on 32-bit architecture</i>
4902 <!-- _______________________________________________________________________ -->
4903 <div class="doc_subsubsection">
4904 <a name="i_bitcast">'<tt>bitcast .. to</tt>' Instruction</a>
4906 <div class="doc_text">
4910 <result> = bitcast <ty> <value> to <ty2> <i>; yields ty2</i>
4914 <p>The '<tt>bitcast</tt>' instruction converts <tt>value</tt> to type
4915 <tt>ty2</tt> without changing any bits.</p>
4918 <p>The '<tt>bitcast</tt>' instruction takes a value to cast, which must be a
4919 non-aggregate first class value, and a type to cast it to, which must also be
4920 a non-aggregate <a href="#t_firstclass">first class</a> type. The bit sizes
4921 of <tt>value</tt> and the destination type, <tt>ty2</tt>, must be
4922 identical. If the source type is a pointer, the destination type must also be
4923 a pointer. This instruction supports bitwise conversion of vectors to
4924 integers and to vectors of other types (as long as they have the same
4928 <p>The '<tt>bitcast</tt>' instruction converts <tt>value</tt> to type
4929 <tt>ty2</tt>. It is always a <i>no-op cast</i> because no bits change with
4930 this conversion. The conversion is done as if the <tt>value</tt> had been
4931 stored to memory and read back as type <tt>ty2</tt>. Pointer types may only
4932 be converted to other pointer types with this instruction. To convert
4933 pointers to other types, use the <a href="#i_inttoptr">inttoptr</a> or
4934 <a href="#i_ptrtoint">ptrtoint</a> instructions first.</p>
4938 %X = bitcast i8 255 to i8 <i>; yields i8 :-1</i>
4939 %Y = bitcast i32* %x to sint* <i>; yields sint*:%x</i>
4940 %Z = bitcast <2 x int> %V to i64; <i>; yields i64: %V</i>
4945 <!-- ======================================================================= -->
4946 <div class="doc_subsection"> <a name="otherops">Other Operations</a> </div>
4948 <div class="doc_text">
4950 <p>The instructions in this category are the "miscellaneous" instructions, which
4951 defy better classification.</p>
4955 <!-- _______________________________________________________________________ -->
4956 <div class="doc_subsubsection"><a name="i_icmp">'<tt>icmp</tt>' Instruction</a>
4959 <div class="doc_text">
4963 <result> = icmp <cond> <ty> <op1>, <op2> <i>; yields {i1} or {<N x i1>}:result</i>
4967 <p>The '<tt>icmp</tt>' instruction returns a boolean value or a vector of
4968 boolean values based on comparison of its two integer, integer vector, or
4969 pointer operands.</p>
4972 <p>The '<tt>icmp</tt>' instruction takes three operands. The first operand is
4973 the condition code indicating the kind of comparison to perform. It is not a
4974 value, just a keyword. The possible condition code are:</p>
4977 <li><tt>eq</tt>: equal</li>
4978 <li><tt>ne</tt>: not equal </li>
4979 <li><tt>ugt</tt>: unsigned greater than</li>
4980 <li><tt>uge</tt>: unsigned greater or equal</li>
4981 <li><tt>ult</tt>: unsigned less than</li>
4982 <li><tt>ule</tt>: unsigned less or equal</li>
4983 <li><tt>sgt</tt>: signed greater than</li>
4984 <li><tt>sge</tt>: signed greater or equal</li>
4985 <li><tt>slt</tt>: signed less than</li>
4986 <li><tt>sle</tt>: signed less or equal</li>
4989 <p>The remaining two arguments must be <a href="#t_integer">integer</a> or
4990 <a href="#t_pointer">pointer</a> or integer <a href="#t_vector">vector</a>
4991 typed. They must also be identical types.</p>
4994 <p>The '<tt>icmp</tt>' compares <tt>op1</tt> and <tt>op2</tt> according to the
4995 condition code given as <tt>cond</tt>. The comparison performed always yields
4996 either an <a href="#t_integer"><tt>i1</tt></a> or vector of <tt>i1</tt>
4997 result, as follows:</p>
5000 <li><tt>eq</tt>: yields <tt>true</tt> if the operands are equal,
5001 <tt>false</tt> otherwise. No sign interpretation is necessary or
5004 <li><tt>ne</tt>: yields <tt>true</tt> if the operands are unequal,
5005 <tt>false</tt> otherwise. No sign interpretation is necessary or
5008 <li><tt>ugt</tt>: interprets the operands as unsigned values and yields
5009 <tt>true</tt> if <tt>op1</tt> is greater than <tt>op2</tt>.</li>
5011 <li><tt>uge</tt>: interprets the operands as unsigned values and yields
5012 <tt>true</tt> if <tt>op1</tt> is greater than or equal
5013 to <tt>op2</tt>.</li>
5015 <li><tt>ult</tt>: interprets the operands as unsigned values and yields
5016 <tt>true</tt> if <tt>op1</tt> is less than <tt>op2</tt>.</li>
5018 <li><tt>ule</tt>: interprets the operands as unsigned values and yields
5019 <tt>true</tt> if <tt>op1</tt> is less than or equal to <tt>op2</tt>.</li>
5021 <li><tt>sgt</tt>: interprets the operands as signed values and yields
5022 <tt>true</tt> if <tt>op1</tt> is greater than <tt>op2</tt>.</li>
5024 <li><tt>sge</tt>: interprets the operands as signed values and yields
5025 <tt>true</tt> if <tt>op1</tt> is greater than or equal
5026 to <tt>op2</tt>.</li>
5028 <li><tt>slt</tt>: interprets the operands as signed values and yields
5029 <tt>true</tt> if <tt>op1</tt> is less than <tt>op2</tt>.</li>
5031 <li><tt>sle</tt>: interprets the operands as signed values and yields
5032 <tt>true</tt> if <tt>op1</tt> is less than or equal to <tt>op2</tt>.</li>
5035 <p>If the operands are <a href="#t_pointer">pointer</a> typed, the pointer
5036 values are compared as if they were integers.</p>
5038 <p>If the operands are integer vectors, then they are compared element by
5039 element. The result is an <tt>i1</tt> vector with the same number of elements
5040 as the values being compared. Otherwise, the result is an <tt>i1</tt>.</p>
5044 <result> = icmp eq i32 4, 5 <i>; yields: result=false</i>
5045 <result> = icmp ne float* %X, %X <i>; yields: result=false</i>
5046 <result> = icmp ult i16 4, 5 <i>; yields: result=true</i>
5047 <result> = icmp sgt i16 4, 5 <i>; yields: result=false</i>
5048 <result> = icmp ule i16 -4, 5 <i>; yields: result=false</i>
5049 <result> = icmp sge i16 4, 5 <i>; yields: result=false</i>
5052 <p>Note that the code generator does not yet support vector types with
5053 the <tt>icmp</tt> instruction.</p>
5057 <!-- _______________________________________________________________________ -->
5058 <div class="doc_subsubsection"><a name="i_fcmp">'<tt>fcmp</tt>' Instruction</a>
5061 <div class="doc_text">
5065 <result> = fcmp <cond> <ty> <op1>, <op2> <i>; yields {i1} or {<N x i1>}:result</i>
5069 <p>The '<tt>fcmp</tt>' instruction returns a boolean value or vector of boolean
5070 values based on comparison of its operands.</p>
5072 <p>If the operands are floating point scalars, then the result type is a boolean
5073 (<a href="#t_integer"><tt>i1</tt></a>).</p>
5075 <p>If the operands are floating point vectors, then the result type is a vector
5076 of boolean with the same number of elements as the operands being
5080 <p>The '<tt>fcmp</tt>' instruction takes three operands. The first operand is
5081 the condition code indicating the kind of comparison to perform. It is not a
5082 value, just a keyword. The possible condition code are:</p>
5085 <li><tt>false</tt>: no comparison, always returns false</li>
5086 <li><tt>oeq</tt>: ordered and equal</li>
5087 <li><tt>ogt</tt>: ordered and greater than </li>
5088 <li><tt>oge</tt>: ordered and greater than or equal</li>
5089 <li><tt>olt</tt>: ordered and less than </li>
5090 <li><tt>ole</tt>: ordered and less than or equal</li>
5091 <li><tt>one</tt>: ordered and not equal</li>
5092 <li><tt>ord</tt>: ordered (no nans)</li>
5093 <li><tt>ueq</tt>: unordered or equal</li>
5094 <li><tt>ugt</tt>: unordered or greater than </li>
5095 <li><tt>uge</tt>: unordered or greater than or equal</li>
5096 <li><tt>ult</tt>: unordered or less than </li>
5097 <li><tt>ule</tt>: unordered or less than or equal</li>
5098 <li><tt>une</tt>: unordered or not equal</li>
5099 <li><tt>uno</tt>: unordered (either nans)</li>
5100 <li><tt>true</tt>: no comparison, always returns true</li>
5103 <p><i>Ordered</i> means that neither operand is a QNAN while
5104 <i>unordered</i> means that either operand may be a QNAN.</p>
5106 <p>Each of <tt>val1</tt> and <tt>val2</tt> arguments must be either
5107 a <a href="#t_floating">floating point</a> type or
5108 a <a href="#t_vector">vector</a> of floating point type. They must have
5109 identical types.</p>
5112 <p>The '<tt>fcmp</tt>' instruction compares <tt>op1</tt> and <tt>op2</tt>
5113 according to the condition code given as <tt>cond</tt>. If the operands are
5114 vectors, then the vectors are compared element by element. Each comparison
5115 performed always yields an <a href="#t_integer">i1</a> result, as
5119 <li><tt>false</tt>: always yields <tt>false</tt>, regardless of operands.</li>
5121 <li><tt>oeq</tt>: yields <tt>true</tt> if both operands are not a QNAN and
5122 <tt>op1</tt> is equal to <tt>op2</tt>.</li>
5124 <li><tt>ogt</tt>: yields <tt>true</tt> if both operands are not a QNAN and
5125 <tt>op1</tt> is greater than <tt>op2</tt>.</li>
5127 <li><tt>oge</tt>: yields <tt>true</tt> if both operands are not a QNAN and
5128 <tt>op1</tt> is greater than or equal to <tt>op2</tt>.</li>
5130 <li><tt>olt</tt>: yields <tt>true</tt> if both operands are not a QNAN and
5131 <tt>op1</tt> is less than <tt>op2</tt>.</li>
5133 <li><tt>ole</tt>: yields <tt>true</tt> if both operands are not a QNAN and
5134 <tt>op1</tt> is less than or equal to <tt>op2</tt>.</li>
5136 <li><tt>one</tt>: yields <tt>true</tt> if both operands are not a QNAN and
5137 <tt>op1</tt> is not equal to <tt>op2</tt>.</li>
5139 <li><tt>ord</tt>: yields <tt>true</tt> if both operands are not a QNAN.</li>
5141 <li><tt>ueq</tt>: yields <tt>true</tt> if either operand is a QNAN or
5142 <tt>op1</tt> is equal to <tt>op2</tt>.</li>
5144 <li><tt>ugt</tt>: yields <tt>true</tt> if either operand is a QNAN or
5145 <tt>op1</tt> is greater than <tt>op2</tt>.</li>
5147 <li><tt>uge</tt>: yields <tt>true</tt> if either operand is a QNAN or
5148 <tt>op1</tt> is greater than or equal to <tt>op2</tt>.</li>
5150 <li><tt>ult</tt>: yields <tt>true</tt> if either operand is a QNAN or
5151 <tt>op1</tt> is less than <tt>op2</tt>.</li>
5153 <li><tt>ule</tt>: yields <tt>true</tt> if either operand is a QNAN or
5154 <tt>op1</tt> is less than or equal to <tt>op2</tt>.</li>
5156 <li><tt>une</tt>: yields <tt>true</tt> if either operand is a QNAN or
5157 <tt>op1</tt> is not equal to <tt>op2</tt>.</li>
5159 <li><tt>uno</tt>: yields <tt>true</tt> if either operand is a QNAN.</li>
5161 <li><tt>true</tt>: always yields <tt>true</tt>, regardless of operands.</li>
5166 <result> = fcmp oeq float 4.0, 5.0 <i>; yields: result=false</i>
5167 <result> = fcmp one float 4.0, 5.0 <i>; yields: result=true</i>
5168 <result> = fcmp olt float 4.0, 5.0 <i>; yields: result=true</i>
5169 <result> = fcmp ueq double 1.0, 2.0 <i>; yields: result=false</i>
5172 <p>Note that the code generator does not yet support vector types with
5173 the <tt>fcmp</tt> instruction.</p>
5177 <!-- _______________________________________________________________________ -->
5178 <div class="doc_subsubsection">
5179 <a name="i_phi">'<tt>phi</tt>' Instruction</a>
5182 <div class="doc_text">
5186 <result> = phi <ty> [ <val0>, <label0>], ...
5190 <p>The '<tt>phi</tt>' instruction is used to implement the φ node in the
5191 SSA graph representing the function.</p>
5194 <p>The type of the incoming values is specified with the first type field. After
5195 this, the '<tt>phi</tt>' instruction takes a list of pairs as arguments, with
5196 one pair for each predecessor basic block of the current block. Only values
5197 of <a href="#t_firstclass">first class</a> type may be used as the value
5198 arguments to the PHI node. Only labels may be used as the label
5201 <p>There must be no non-phi instructions between the start of a basic block and
5202 the PHI instructions: i.e. PHI instructions must be first in a basic
5205 <p>For the purposes of the SSA form, the use of each incoming value is deemed to
5206 occur on the edge from the corresponding predecessor block to the current
5207 block (but after any definition of an '<tt>invoke</tt>' instruction's return
5208 value on the same edge).</p>
5211 <p>At runtime, the '<tt>phi</tt>' instruction logically takes on the value
5212 specified by the pair corresponding to the predecessor basic block that
5213 executed just prior to the current block.</p>
5217 Loop: ; Infinite loop that counts from 0 on up...
5218 %indvar = phi i32 [ 0, %LoopHeader ], [ %nextindvar, %Loop ]
5219 %nextindvar = add i32 %indvar, 1
5225 <!-- _______________________________________________________________________ -->
5226 <div class="doc_subsubsection">
5227 <a name="i_select">'<tt>select</tt>' Instruction</a>
5230 <div class="doc_text">
5234 <result> = select <i>selty</i> <cond>, <ty> <val1>, <ty> <val2> <i>; yields ty</i>
5236 <i>selty</i> is either i1 or {<N x i1>}
5240 <p>The '<tt>select</tt>' instruction is used to choose one value based on a
5241 condition, without branching.</p>
5245 <p>The '<tt>select</tt>' instruction requires an 'i1' value or a vector of 'i1'
5246 values indicating the condition, and two values of the
5247 same <a href="#t_firstclass">first class</a> type. If the val1/val2 are
5248 vectors and the condition is a scalar, then entire vectors are selected, not
5249 individual elements.</p>
5252 <p>If the condition is an i1 and it evaluates to 1, the instruction returns the
5253 first value argument; otherwise, it returns the second value argument.</p>
5255 <p>If the condition is a vector of i1, then the value arguments must be vectors
5256 of the same size, and the selection is done element by element.</p>
5260 %X = select i1 true, i8 17, i8 42 <i>; yields i8:17</i>
5263 <p>Note that the code generator does not yet support conditions
5264 with vector type.</p>
5268 <!-- _______________________________________________________________________ -->
5269 <div class="doc_subsubsection">
5270 <a name="i_call">'<tt>call</tt>' Instruction</a>
5273 <div class="doc_text">
5277 <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>]
5281 <p>The '<tt>call</tt>' instruction represents a simple function call.</p>
5284 <p>This instruction requires several arguments:</p>
5287 <li>The optional "tail" marker indicates that the callee function does not
5288 access any allocas or varargs in the caller. Note that calls may be
5289 marked "tail" even if they do not occur before
5290 a <a href="#i_ret"><tt>ret</tt></a> instruction. If the "tail" marker is
5291 present, the function call is eligible for tail call optimization,
5292 but <a href="CodeGenerator.html#tailcallopt">might not in fact be
5293 optimized into a jump</a>. The code generator may optimize calls marked
5294 "tail" with either 1) automatic <a href="CodeGenerator.html#sibcallopt">
5295 sibling call optimization</a> when the caller and callee have
5296 matching signatures, or 2) forced tail call optimization when the
5297 following extra requirements are met:
5299 <li>Caller and callee both have the calling
5300 convention <tt>fastcc</tt>.</li>
5301 <li>The call is in tail position (ret immediately follows call and ret
5302 uses value of call or is void).</li>
5303 <li>Option <tt>-tailcallopt</tt> is enabled,
5304 or <code>llvm::GuaranteedTailCallOpt</code> is <code>true</code>.</li>
5305 <li><a href="CodeGenerator.html#tailcallopt">Platform specific
5306 constraints are met.</a></li>
5310 <li>The optional "cconv" marker indicates which <a href="#callingconv">calling
5311 convention</a> the call should use. If none is specified, the call
5312 defaults to using C calling conventions. The calling convention of the
5313 call must match the calling convention of the target function, or else the
5314 behavior is undefined.</li>
5316 <li>The optional <a href="#paramattrs">Parameter Attributes</a> list for
5317 return values. Only '<tt>zeroext</tt>', '<tt>signext</tt>', and
5318 '<tt>inreg</tt>' attributes are valid here.</li>
5320 <li>'<tt>ty</tt>': the type of the call instruction itself which is also the
5321 type of the return value. Functions that return no value are marked
5322 <tt><a href="#t_void">void</a></tt>.</li>
5324 <li>'<tt>fnty</tt>': shall be the signature of the pointer to function value
5325 being invoked. The argument types must match the types implied by this
5326 signature. This type can be omitted if the function is not varargs and if
5327 the function type does not return a pointer to a function.</li>
5329 <li>'<tt>fnptrval</tt>': An LLVM value containing a pointer to a function to
5330 be invoked. In most cases, this is a direct function invocation, but
5331 indirect <tt>call</tt>s are just as possible, calling an arbitrary pointer
5332 to function value.</li>
5334 <li>'<tt>function args</tt>': argument list whose types match the function
5335 signature argument types and parameter attributes. All arguments must be
5336 of <a href="#t_firstclass">first class</a> type. If the function
5337 signature indicates the function accepts a variable number of arguments,
5338 the extra arguments can be specified.</li>
5340 <li>The optional <a href="#fnattrs">function attributes</a> list. Only
5341 '<tt>noreturn</tt>', '<tt>nounwind</tt>', '<tt>readonly</tt>' and
5342 '<tt>readnone</tt>' attributes are valid here.</li>
5346 <p>The '<tt>call</tt>' instruction is used to cause control flow to transfer to
5347 a specified function, with its incoming arguments bound to the specified
5348 values. Upon a '<tt><a href="#i_ret">ret</a></tt>' instruction in the called
5349 function, control flow continues with the instruction after the function
5350 call, and the return value of the function is bound to the result
5355 %retval = call i32 @test(i32 %argc)
5356 call i32 (i8*, ...)* @printf(i8* %msg, i32 12, i8 42) <i>; yields i32</i>
5357 %X = tail call i32 @foo() <i>; yields i32</i>
5358 %Y = tail call <a href="#callingconv">fastcc</a> i32 @foo() <i>; yields i32</i>
5359 call void %foo(i8 97 signext)
5361 %struct.A = type { i32, i8 }
5362 %r = call %struct.A @foo() <i>; yields { 32, i8 }</i>
5363 %gr = extractvalue %struct.A %r, 0 <i>; yields i32</i>
5364 %gr1 = extractvalue %struct.A %r, 1 <i>; yields i8</i>
5365 %Z = call void @foo() noreturn <i>; indicates that %foo never returns normally</i>
5366 %ZZ = call zeroext i32 @bar() <i>; Return value is %zero extended</i>
5369 <p>llvm treats calls to some functions with names and arguments that match the
5370 standard C99 library as being the C99 library functions, and may perform
5371 optimizations or generate code for them under that assumption. This is
5372 something we'd like to change in the future to provide better support for
5373 freestanding environments and non-C-based languages.</p>
5377 <!-- _______________________________________________________________________ -->
5378 <div class="doc_subsubsection">
5379 <a name="i_va_arg">'<tt>va_arg</tt>' Instruction</a>
5382 <div class="doc_text">
5386 <resultval> = va_arg <va_list*> <arglist>, <argty>
5390 <p>The '<tt>va_arg</tt>' instruction is used to access arguments passed through
5391 the "variable argument" area of a function call. It is used to implement the
5392 <tt>va_arg</tt> macro in C.</p>
5395 <p>This instruction takes a <tt>va_list*</tt> value and the type of the
5396 argument. It returns a value of the specified argument type and increments
5397 the <tt>va_list</tt> to point to the next argument. The actual type
5398 of <tt>va_list</tt> is target specific.</p>
5401 <p>The '<tt>va_arg</tt>' instruction loads an argument of the specified type
5402 from the specified <tt>va_list</tt> and causes the <tt>va_list</tt> to point
5403 to the next argument. For more information, see the variable argument
5404 handling <a href="#int_varargs">Intrinsic Functions</a>.</p>
5406 <p>It is legal for this instruction to be called in a function which does not
5407 take a variable number of arguments, for example, the <tt>vfprintf</tt>
5410 <p><tt>va_arg</tt> is an LLVM instruction instead of
5411 an <a href="#intrinsics">intrinsic function</a> because it takes a type as an
5415 <p>See the <a href="#int_varargs">variable argument processing</a> section.</p>
5417 <p>Note that the code generator does not yet fully support va_arg on many
5418 targets. Also, it does not currently support va_arg with aggregate types on
5423 <!-- *********************************************************************** -->
5424 <div class="doc_section"> <a name="intrinsics">Intrinsic Functions</a> </div>
5425 <!-- *********************************************************************** -->
5427 <div class="doc_text">
5429 <p>LLVM supports the notion of an "intrinsic function". These functions have
5430 well known names and semantics and are required to follow certain
5431 restrictions. Overall, these intrinsics represent an extension mechanism for
5432 the LLVM language that does not require changing all of the transformations
5433 in LLVM when adding to the language (or the bitcode reader/writer, the
5434 parser, etc...).</p>
5436 <p>Intrinsic function names must all start with an "<tt>llvm.</tt>" prefix. This
5437 prefix is reserved in LLVM for intrinsic names; thus, function names may not
5438 begin with this prefix. Intrinsic functions must always be external
5439 functions: you cannot define the body of intrinsic functions. Intrinsic
5440 functions may only be used in call or invoke instructions: it is illegal to
5441 take the address of an intrinsic function. Additionally, because intrinsic
5442 functions are part of the LLVM language, it is required if any are added that
5443 they be documented here.</p>
5445 <p>Some intrinsic functions can be overloaded, i.e., the intrinsic represents a
5446 family of functions that perform the same operation but on different data
5447 types. Because LLVM can represent over 8 million different integer types,
5448 overloading is used commonly to allow an intrinsic function to operate on any
5449 integer type. One or more of the argument types or the result type can be
5450 overloaded to accept any integer type. Argument types may also be defined as
5451 exactly matching a previous argument's type or the result type. This allows
5452 an intrinsic function which accepts multiple arguments, but needs all of them
5453 to be of the same type, to only be overloaded with respect to a single
5454 argument or the result.</p>
5456 <p>Overloaded intrinsics will have the names of its overloaded argument types
5457 encoded into its function name, each preceded by a period. Only those types
5458 which are overloaded result in a name suffix. Arguments whose type is matched
5459 against another type do not. For example, the <tt>llvm.ctpop</tt> function
5460 can take an integer of any width and returns an integer of exactly the same
5461 integer width. This leads to a family of functions such as
5462 <tt>i8 @llvm.ctpop.i8(i8 %val)</tt> and <tt>i29 @llvm.ctpop.i29(i29
5463 %val)</tt>. Only one type, the return type, is overloaded, and only one type
5464 suffix is required. Because the argument's type is matched against the return
5465 type, it does not require its own name suffix.</p>
5467 <p>To learn how to add an intrinsic function, please see the
5468 <a href="ExtendingLLVM.html">Extending LLVM Guide</a>.</p>
5472 <!-- ======================================================================= -->
5473 <div class="doc_subsection">
5474 <a name="int_varargs">Variable Argument Handling Intrinsics</a>
5477 <div class="doc_text">
5479 <p>Variable argument support is defined in LLVM with
5480 the <a href="#i_va_arg"><tt>va_arg</tt></a> instruction and these three
5481 intrinsic functions. These functions are related to the similarly named
5482 macros defined in the <tt><stdarg.h></tt> header file.</p>
5484 <p>All of these functions operate on arguments that use a target-specific value
5485 type "<tt>va_list</tt>". The LLVM assembly language reference manual does
5486 not define what this type is, so all transformations should be prepared to
5487 handle these functions regardless of the type used.</p>
5489 <p>This example shows how the <a href="#i_va_arg"><tt>va_arg</tt></a>
5490 instruction and the variable argument handling intrinsic functions are
5493 <pre class="doc_code">
5494 define i32 @test(i32 %X, ...) {
5495 ; Initialize variable argument processing
5497 %ap2 = bitcast i8** %ap to i8*
5498 call void @llvm.va_start(i8* %ap2)
5500 ; Read a single integer argument
5501 %tmp = va_arg i8** %ap, i32
5503 ; Demonstrate usage of llvm.va_copy and llvm.va_end
5505 %aq2 = bitcast i8** %aq to i8*
5506 call void @llvm.va_copy(i8* %aq2, i8* %ap2)
5507 call void @llvm.va_end(i8* %aq2)
5509 ; Stop processing of arguments.
5510 call void @llvm.va_end(i8* %ap2)
5514 declare void @llvm.va_start(i8*)
5515 declare void @llvm.va_copy(i8*, i8*)
5516 declare void @llvm.va_end(i8*)
5521 <!-- _______________________________________________________________________ -->
5522 <div class="doc_subsubsection">
5523 <a name="int_va_start">'<tt>llvm.va_start</tt>' Intrinsic</a>
5527 <div class="doc_text">
5531 declare void %llvm.va_start(i8* <arglist>)
5535 <p>The '<tt>llvm.va_start</tt>' intrinsic initializes <tt>*<arglist></tt>
5536 for subsequent use by <tt><a href="#i_va_arg">va_arg</a></tt>.</p>
5539 <p>The argument is a pointer to a <tt>va_list</tt> element to initialize.</p>
5542 <p>The '<tt>llvm.va_start</tt>' intrinsic works just like the <tt>va_start</tt>
5543 macro available in C. In a target-dependent way, it initializes
5544 the <tt>va_list</tt> element to which the argument points, so that the next
5545 call to <tt>va_arg</tt> will produce the first variable argument passed to
5546 the function. Unlike the C <tt>va_start</tt> macro, this intrinsic does not
5547 need to know the last argument of the function as the compiler can figure
5552 <!-- _______________________________________________________________________ -->
5553 <div class="doc_subsubsection">
5554 <a name="int_va_end">'<tt>llvm.va_end</tt>' Intrinsic</a>
5557 <div class="doc_text">
5561 declare void @llvm.va_end(i8* <arglist>)
5565 <p>The '<tt>llvm.va_end</tt>' intrinsic destroys <tt>*<arglist></tt>,
5566 which has been initialized previously
5567 with <tt><a href="#int_va_start">llvm.va_start</a></tt>
5568 or <tt><a href="#i_va_copy">llvm.va_copy</a></tt>.</p>
5571 <p>The argument is a pointer to a <tt>va_list</tt> to destroy.</p>
5574 <p>The '<tt>llvm.va_end</tt>' intrinsic works just like the <tt>va_end</tt>
5575 macro available in C. In a target-dependent way, it destroys
5576 the <tt>va_list</tt> element to which the argument points. Calls
5577 to <a href="#int_va_start"><tt>llvm.va_start</tt></a>
5578 and <a href="#int_va_copy"> <tt>llvm.va_copy</tt></a> must be matched exactly
5579 with calls to <tt>llvm.va_end</tt>.</p>
5583 <!-- _______________________________________________________________________ -->
5584 <div class="doc_subsubsection">
5585 <a name="int_va_copy">'<tt>llvm.va_copy</tt>' Intrinsic</a>
5588 <div class="doc_text">
5592 declare void @llvm.va_copy(i8* <destarglist>, i8* <srcarglist>)
5596 <p>The '<tt>llvm.va_copy</tt>' intrinsic copies the current argument position
5597 from the source argument list to the destination argument list.</p>
5600 <p>The first argument is a pointer to a <tt>va_list</tt> element to initialize.
5601 The second argument is a pointer to a <tt>va_list</tt> element to copy
5605 <p>The '<tt>llvm.va_copy</tt>' intrinsic works just like the <tt>va_copy</tt>
5606 macro available in C. In a target-dependent way, it copies the
5607 source <tt>va_list</tt> element into the destination <tt>va_list</tt>
5608 element. This intrinsic is necessary because
5609 the <tt><a href="#int_va_start"> llvm.va_start</a></tt> intrinsic may be
5610 arbitrarily complex and require, for example, memory allocation.</p>
5614 <!-- ======================================================================= -->
5615 <div class="doc_subsection">
5616 <a name="int_gc">Accurate Garbage Collection Intrinsics</a>
5619 <div class="doc_text">
5621 <p>LLVM support for <a href="GarbageCollection.html">Accurate Garbage
5622 Collection</a> (GC) requires the implementation and generation of these
5623 intrinsics. These intrinsics allow identification of <a href="#int_gcroot">GC
5624 roots on the stack</a>, as well as garbage collector implementations that
5625 require <a href="#int_gcread">read</a> and <a href="#int_gcwrite">write</a>
5626 barriers. Front-ends for type-safe garbage collected languages should generate
5627 these intrinsics to make use of the LLVM garbage collectors. For more details,
5628 see <a href="GarbageCollection.html">Accurate Garbage Collection with
5631 <p>The garbage collection intrinsics only operate on objects in the generic
5632 address space (address space zero).</p>
5636 <!-- _______________________________________________________________________ -->
5637 <div class="doc_subsubsection">
5638 <a name="int_gcroot">'<tt>llvm.gcroot</tt>' Intrinsic</a>
5641 <div class="doc_text">
5645 declare void @llvm.gcroot(i8** %ptrloc, i8* %metadata)
5649 <p>The '<tt>llvm.gcroot</tt>' intrinsic declares the existence of a GC root to
5650 the code generator, and allows some metadata to be associated with it.</p>
5653 <p>The first argument specifies the address of a stack object that contains the
5654 root pointer. The second pointer (which must be either a constant or a
5655 global value address) contains the meta-data to be associated with the
5659 <p>At runtime, a call to this intrinsic stores a null pointer into the "ptrloc"
5660 location. At compile-time, the code generator generates information to allow
5661 the runtime to find the pointer at GC safe points. The '<tt>llvm.gcroot</tt>'
5662 intrinsic may only be used in a function which <a href="#gc">specifies a GC
5667 <!-- _______________________________________________________________________ -->
5668 <div class="doc_subsubsection">
5669 <a name="int_gcread">'<tt>llvm.gcread</tt>' Intrinsic</a>
5672 <div class="doc_text">
5676 declare i8* @llvm.gcread(i8* %ObjPtr, i8** %Ptr)
5680 <p>The '<tt>llvm.gcread</tt>' intrinsic identifies reads of references from heap
5681 locations, allowing garbage collector implementations that require read
5685 <p>The second argument is the address to read from, which should be an address
5686 allocated from the garbage collector. The first object is a pointer to the
5687 start of the referenced object, if needed by the language runtime (otherwise
5691 <p>The '<tt>llvm.gcread</tt>' intrinsic has the same semantics as a load
5692 instruction, but may be replaced with substantially more complex code by the
5693 garbage collector runtime, as needed. The '<tt>llvm.gcread</tt>' intrinsic
5694 may only be used in a function which <a href="#gc">specifies a GC
5699 <!-- _______________________________________________________________________ -->
5700 <div class="doc_subsubsection">
5701 <a name="int_gcwrite">'<tt>llvm.gcwrite</tt>' Intrinsic</a>
5704 <div class="doc_text">
5708 declare void @llvm.gcwrite(i8* %P1, i8* %Obj, i8** %P2)
5712 <p>The '<tt>llvm.gcwrite</tt>' intrinsic identifies writes of references to heap
5713 locations, allowing garbage collector implementations that require write
5714 barriers (such as generational or reference counting collectors).</p>
5717 <p>The first argument is the reference to store, the second is the start of the
5718 object to store it to, and the third is the address of the field of Obj to
5719 store to. If the runtime does not require a pointer to the object, Obj may
5723 <p>The '<tt>llvm.gcwrite</tt>' intrinsic has the same semantics as a store
5724 instruction, but may be replaced with substantially more complex code by the
5725 garbage collector runtime, as needed. The '<tt>llvm.gcwrite</tt>' intrinsic
5726 may only be used in a function which <a href="#gc">specifies a GC
5731 <!-- ======================================================================= -->
5732 <div class="doc_subsection">
5733 <a name="int_codegen">Code Generator Intrinsics</a>
5736 <div class="doc_text">
5738 <p>These intrinsics are provided by LLVM to expose special features that may
5739 only be implemented with code generator support.</p>
5743 <!-- _______________________________________________________________________ -->
5744 <div class="doc_subsubsection">
5745 <a name="int_returnaddress">'<tt>llvm.returnaddress</tt>' Intrinsic</a>
5748 <div class="doc_text">
5752 declare i8 *@llvm.returnaddress(i32 <level>)
5756 <p>The '<tt>llvm.returnaddress</tt>' intrinsic attempts to compute a
5757 target-specific value indicating the return address of the current function
5758 or one of its callers.</p>
5761 <p>The argument to this intrinsic indicates which function to return the address
5762 for. Zero indicates the calling function, one indicates its caller, etc.
5763 The argument is <b>required</b> to be a constant integer value.</p>
5766 <p>The '<tt>llvm.returnaddress</tt>' intrinsic either returns a pointer
5767 indicating the return address of the specified call frame, or zero if it
5768 cannot be identified. The value returned by this intrinsic is likely to be
5769 incorrect or 0 for arguments other than zero, so it should only be used for
5770 debugging purposes.</p>
5772 <p>Note that calling this intrinsic does not prevent function inlining or other
5773 aggressive transformations, so the value returned may not be that of the
5774 obvious source-language caller.</p>
5778 <!-- _______________________________________________________________________ -->
5779 <div class="doc_subsubsection">
5780 <a name="int_frameaddress">'<tt>llvm.frameaddress</tt>' Intrinsic</a>
5783 <div class="doc_text">
5787 declare i8* @llvm.frameaddress(i32 <level>)
5791 <p>The '<tt>llvm.frameaddress</tt>' intrinsic attempts to return the
5792 target-specific frame pointer value for the specified stack frame.</p>
5795 <p>The argument to this intrinsic indicates which function to return the frame
5796 pointer for. Zero indicates the calling function, one indicates its caller,
5797 etc. The argument is <b>required</b> to be a constant integer value.</p>
5800 <p>The '<tt>llvm.frameaddress</tt>' intrinsic either returns a pointer
5801 indicating the frame address of the specified call frame, or zero if it
5802 cannot be identified. The value returned by this intrinsic is likely to be
5803 incorrect or 0 for arguments other than zero, so it should only be used for
5804 debugging purposes.</p>
5806 <p>Note that calling this intrinsic does not prevent function inlining or other
5807 aggressive transformations, so the value returned may not be that of the
5808 obvious source-language caller.</p>
5812 <!-- _______________________________________________________________________ -->
5813 <div class="doc_subsubsection">
5814 <a name="int_stacksave">'<tt>llvm.stacksave</tt>' Intrinsic</a>
5817 <div class="doc_text">
5821 declare i8* @llvm.stacksave()
5825 <p>The '<tt>llvm.stacksave</tt>' intrinsic is used to remember the current state
5826 of the function stack, for use
5827 with <a href="#int_stackrestore"> <tt>llvm.stackrestore</tt></a>. This is
5828 useful for implementing language features like scoped automatic variable
5829 sized arrays in C99.</p>
5832 <p>This intrinsic returns a opaque pointer value that can be passed
5833 to <a href="#int_stackrestore"><tt>llvm.stackrestore</tt></a>. When
5834 an <tt>llvm.stackrestore</tt> intrinsic is executed with a value saved
5835 from <tt>llvm.stacksave</tt>, it effectively restores the state of the stack
5836 to the state it was in when the <tt>llvm.stacksave</tt> intrinsic executed.
5837 In practice, this pops any <a href="#i_alloca">alloca</a> blocks from the
5838 stack that were allocated after the <tt>llvm.stacksave</tt> was executed.</p>
5842 <!-- _______________________________________________________________________ -->
5843 <div class="doc_subsubsection">
5844 <a name="int_stackrestore">'<tt>llvm.stackrestore</tt>' Intrinsic</a>
5847 <div class="doc_text">
5851 declare void @llvm.stackrestore(i8* %ptr)
5855 <p>The '<tt>llvm.stackrestore</tt>' intrinsic is used to restore the state of
5856 the function stack to the state it was in when the
5857 corresponding <a href="#int_stacksave"><tt>llvm.stacksave</tt></a> intrinsic
5858 executed. This is useful for implementing language features like scoped
5859 automatic variable sized arrays in C99.</p>
5862 <p>See the description
5863 for <a href="#int_stacksave"><tt>llvm.stacksave</tt></a>.</p>
5867 <!-- _______________________________________________________________________ -->
5868 <div class="doc_subsubsection">
5869 <a name="int_prefetch">'<tt>llvm.prefetch</tt>' Intrinsic</a>
5872 <div class="doc_text">
5876 declare void @llvm.prefetch(i8* <address>, i32 <rw>, i32 <locality>)
5880 <p>The '<tt>llvm.prefetch</tt>' intrinsic is a hint to the code generator to
5881 insert a prefetch instruction if supported; otherwise, it is a noop.
5882 Prefetches have no effect on the behavior of the program but can change its
5883 performance characteristics.</p>
5886 <p><tt>address</tt> is the address to be prefetched, <tt>rw</tt> is the
5887 specifier determining if the fetch should be for a read (0) or write (1),
5888 and <tt>locality</tt> is a temporal locality specifier ranging from (0) - no
5889 locality, to (3) - extremely local keep in cache. The <tt>rw</tt>
5890 and <tt>locality</tt> arguments must be constant integers.</p>
5893 <p>This intrinsic does not modify the behavior of the program. In particular,
5894 prefetches cannot trap and do not produce a value. On targets that support
5895 this intrinsic, the prefetch can provide hints to the processor cache for
5896 better performance.</p>
5900 <!-- _______________________________________________________________________ -->
5901 <div class="doc_subsubsection">
5902 <a name="int_pcmarker">'<tt>llvm.pcmarker</tt>' Intrinsic</a>
5905 <div class="doc_text">
5909 declare void @llvm.pcmarker(i32 <id>)
5913 <p>The '<tt>llvm.pcmarker</tt>' intrinsic is a method to export a Program
5914 Counter (PC) in a region of code to simulators and other tools. The method
5915 is target specific, but it is expected that the marker will use exported
5916 symbols to transmit the PC of the marker. The marker makes no guarantees
5917 that it will remain with any specific instruction after optimizations. It is
5918 possible that the presence of a marker will inhibit optimizations. The
5919 intended use is to be inserted after optimizations to allow correlations of
5920 simulation runs.</p>
5923 <p><tt>id</tt> is a numerical id identifying the marker.</p>
5926 <p>This intrinsic does not modify the behavior of the program. Backends that do
5927 not support this intrinsic may ignore it.</p>
5931 <!-- _______________________________________________________________________ -->
5932 <div class="doc_subsubsection">
5933 <a name="int_readcyclecounter">'<tt>llvm.readcyclecounter</tt>' Intrinsic</a>
5936 <div class="doc_text">
5940 declare i64 @llvm.readcyclecounter()
5944 <p>The '<tt>llvm.readcyclecounter</tt>' intrinsic provides access to the cycle
5945 counter register (or similar low latency, high accuracy clocks) on those
5946 targets that support it. On X86, it should map to RDTSC. On Alpha, it
5947 should map to RPCC. As the backing counters overflow quickly (on the order
5948 of 9 seconds on alpha), this should only be used for small timings.</p>
5951 <p>When directly supported, reading the cycle counter should not modify any
5952 memory. Implementations are allowed to either return a application specific
5953 value or a system wide value. On backends without support, this is lowered
5954 to a constant 0.</p>
5958 <!-- ======================================================================= -->
5959 <div class="doc_subsection">
5960 <a name="int_libc">Standard C Library Intrinsics</a>
5963 <div class="doc_text">
5965 <p>LLVM provides intrinsics for a few important standard C library functions.
5966 These intrinsics allow source-language front-ends to pass information about
5967 the alignment of the pointer arguments to the code generator, providing
5968 opportunity for more efficient code generation.</p>
5972 <!-- _______________________________________________________________________ -->
5973 <div class="doc_subsubsection">
5974 <a name="int_memcpy">'<tt>llvm.memcpy</tt>' Intrinsic</a>
5977 <div class="doc_text">
5980 <p>This is an overloaded intrinsic. You can use <tt>llvm.memcpy</tt> on any
5981 integer bit width and for different address spaces. Not all targets support
5982 all bit widths however.</p>
5985 declare void @llvm.memcpy.p0i8.p0i8.i32(i8* <dest>, i8* <src>,
5986 i32 <len>, i32 <align>, i1 <isvolatile>)
5987 declare void @llvm.memcpy.p0i8.p0i8.i64(i8* <dest>, i8* <src>,
5988 i64 <len>, i32 <align>, i1 <isvolatile>)
5992 <p>The '<tt>llvm.memcpy.*</tt>' intrinsics copy a block of memory from the
5993 source location to the destination location.</p>
5995 <p>Note that, unlike the standard libc function, the <tt>llvm.memcpy.*</tt>
5996 intrinsics do not return a value, takes extra alignment/isvolatile arguments
5997 and the pointers can be in specified address spaces.</p>
6001 <p>The first argument is a pointer to the destination, the second is a pointer
6002 to the source. The third argument is an integer argument specifying the
6003 number of bytes to copy, the fourth argument is the alignment of the
6004 source and destination locations, and the fifth is a boolean indicating a
6005 volatile access.</p>
6007 <p>If the call to this intrinsic has an alignment value that is not 0 or 1,
6008 then the caller guarantees that both the source and destination pointers are
6009 aligned to that boundary.</p>
6011 <p>If the <tt>isvolatile</tt> parameter is <tt>true</tt>, the
6012 <tt>llvm.memcpy</tt> call is a <a href="#volatile">volatile operation</a>.
6013 The detailed access behavior is not very cleanly specified and it is unwise
6014 to depend on it.</p>
6018 <p>The '<tt>llvm.memcpy.*</tt>' intrinsics copy a block of memory from the
6019 source location to the destination location, which are not allowed to
6020 overlap. It copies "len" bytes of memory over. If the argument is known to
6021 be aligned to some boundary, this can be specified as the fourth argument,
6022 otherwise it should be set to 0 or 1.</p>
6026 <!-- _______________________________________________________________________ -->
6027 <div class="doc_subsubsection">
6028 <a name="int_memmove">'<tt>llvm.memmove</tt>' Intrinsic</a>
6031 <div class="doc_text">
6034 <p>This is an overloaded intrinsic. You can use llvm.memmove on any integer bit
6035 width and for different address space. Not all targets support all bit
6039 declare void @llvm.memmove.p0i8.p0i8.i32(i8* <dest>, i8* <src>,
6040 i32 <len>, i32 <align>, i1 <isvolatile>)
6041 declare void @llvm.memmove.p0i8.p0i8.i64(i8* <dest>, i8* <src>,
6042 i64 <len>, i32 <align>, i1 <isvolatile>)
6046 <p>The '<tt>llvm.memmove.*</tt>' intrinsics move a block of memory from the
6047 source location to the destination location. It is similar to the
6048 '<tt>llvm.memcpy</tt>' intrinsic but allows the two memory locations to
6051 <p>Note that, unlike the standard libc function, the <tt>llvm.memmove.*</tt>
6052 intrinsics do not return a value, takes extra alignment/isvolatile arguments
6053 and the pointers can be in specified address spaces.</p>
6057 <p>The first argument is a pointer to the destination, the second is a pointer
6058 to the source. The third argument is an integer argument specifying the
6059 number of bytes to copy, the fourth argument is the alignment of the
6060 source and destination locations, and the fifth is a boolean indicating a
6061 volatile access.</p>
6063 <p>If the call to this intrinsic has an alignment value that is not 0 or 1,
6064 then the caller guarantees that the source and destination pointers are
6065 aligned to that boundary.</p>
6067 <p>If the <tt>isvolatile</tt> parameter is <tt>true</tt>, the
6068 <tt>llvm.memmove</tt> call is a <a href="#volatile">volatile operation</a>.
6069 The detailed access behavior is not very cleanly specified and it is unwise
6070 to depend on it.</p>
6074 <p>The '<tt>llvm.memmove.*</tt>' intrinsics copy a block of memory from the
6075 source location to the destination location, which may overlap. It copies
6076 "len" bytes of memory over. If the argument is known to be aligned to some
6077 boundary, this can be specified as the fourth argument, otherwise it should
6078 be set to 0 or 1.</p>
6082 <!-- _______________________________________________________________________ -->
6083 <div class="doc_subsubsection">
6084 <a name="int_memset">'<tt>llvm.memset.*</tt>' Intrinsics</a>
6087 <div class="doc_text">
6090 <p>This is an overloaded intrinsic. You can use llvm.memset on any integer bit
6091 width and for different address spaces. However, not all targets support all
6095 declare void @llvm.memset.p0i8.i32(i8* <dest>, i8 <val>,
6096 i32 <len>, i32 <align>, i1 <isvolatile>)
6097 declare void @llvm.memset.p0i8.i64(i8* <dest>, i8 <val>,
6098 i64 <len>, i32 <align>, i1 <isvolatile>)
6102 <p>The '<tt>llvm.memset.*</tt>' intrinsics fill a block of memory with a
6103 particular byte value.</p>
6105 <p>Note that, unlike the standard libc function, the <tt>llvm.memset</tt>
6106 intrinsic does not return a value and takes extra alignment/volatile
6107 arguments. Also, the destination can be in an arbitrary address space.</p>
6110 <p>The first argument is a pointer to the destination to fill, the second is the
6111 byte value with which to fill it, the third argument is an integer argument
6112 specifying the number of bytes to fill, and the fourth argument is the known
6113 alignment of the destination location.</p>
6115 <p>If the call to this intrinsic has an alignment value that is not 0 or 1,
6116 then the caller guarantees that the destination pointer is aligned to that
6119 <p>If the <tt>isvolatile</tt> parameter is <tt>true</tt>, the
6120 <tt>llvm.memset</tt> call is a <a href="#volatile">volatile operation</a>.
6121 The detailed access behavior is not very cleanly specified and it is unwise
6122 to depend on it.</p>
6125 <p>The '<tt>llvm.memset.*</tt>' intrinsics fill "len" bytes of memory starting
6126 at the destination location. If the argument is known to be aligned to some
6127 boundary, this can be specified as the fourth argument, otherwise it should
6128 be set to 0 or 1.</p>
6132 <!-- _______________________________________________________________________ -->
6133 <div class="doc_subsubsection">
6134 <a name="int_sqrt">'<tt>llvm.sqrt.*</tt>' Intrinsic</a>
6137 <div class="doc_text">
6140 <p>This is an overloaded intrinsic. You can use <tt>llvm.sqrt</tt> on any
6141 floating point or vector of floating point type. Not all targets support all
6145 declare float @llvm.sqrt.f32(float %Val)
6146 declare double @llvm.sqrt.f64(double %Val)
6147 declare x86_fp80 @llvm.sqrt.f80(x86_fp80 %Val)
6148 declare fp128 @llvm.sqrt.f128(fp128 %Val)
6149 declare ppc_fp128 @llvm.sqrt.ppcf128(ppc_fp128 %Val)
6153 <p>The '<tt>llvm.sqrt</tt>' intrinsics return the sqrt of the specified operand,
6154 returning the same value as the libm '<tt>sqrt</tt>' functions would.
6155 Unlike <tt>sqrt</tt> in libm, however, <tt>llvm.sqrt</tt> has undefined
6156 behavior for negative numbers other than -0.0 (which allows for better
6157 optimization, because there is no need to worry about errno being
6158 set). <tt>llvm.sqrt(-0.0)</tt> is defined to return -0.0 like IEEE sqrt.</p>
6161 <p>The argument and return value are floating point numbers of the same
6165 <p>This function returns the sqrt of the specified operand if it is a
6166 nonnegative floating point number.</p>
6170 <!-- _______________________________________________________________________ -->
6171 <div class="doc_subsubsection">
6172 <a name="int_powi">'<tt>llvm.powi.*</tt>' Intrinsic</a>
6175 <div class="doc_text">
6178 <p>This is an overloaded intrinsic. You can use <tt>llvm.powi</tt> on any
6179 floating point or vector of floating point type. Not all targets support all
6183 declare float @llvm.powi.f32(float %Val, i32 %power)
6184 declare double @llvm.powi.f64(double %Val, i32 %power)
6185 declare x86_fp80 @llvm.powi.f80(x86_fp80 %Val, i32 %power)
6186 declare fp128 @llvm.powi.f128(fp128 %Val, i32 %power)
6187 declare ppc_fp128 @llvm.powi.ppcf128(ppc_fp128 %Val, i32 %power)
6191 <p>The '<tt>llvm.powi.*</tt>' intrinsics return the first operand raised to the
6192 specified (positive or negative) power. The order of evaluation of
6193 multiplications is not defined. When a vector of floating point type is
6194 used, the second argument remains a scalar integer value.</p>
6197 <p>The second argument is an integer power, and the first is a value to raise to
6201 <p>This function returns the first value raised to the second power with an
6202 unspecified sequence of rounding operations.</p>
6206 <!-- _______________________________________________________________________ -->
6207 <div class="doc_subsubsection">
6208 <a name="int_sin">'<tt>llvm.sin.*</tt>' Intrinsic</a>
6211 <div class="doc_text">
6214 <p>This is an overloaded intrinsic. You can use <tt>llvm.sin</tt> on any
6215 floating point or vector of floating point type. Not all targets support all
6219 declare float @llvm.sin.f32(float %Val)
6220 declare double @llvm.sin.f64(double %Val)
6221 declare x86_fp80 @llvm.sin.f80(x86_fp80 %Val)
6222 declare fp128 @llvm.sin.f128(fp128 %Val)
6223 declare ppc_fp128 @llvm.sin.ppcf128(ppc_fp128 %Val)
6227 <p>The '<tt>llvm.sin.*</tt>' intrinsics return the sine of the operand.</p>
6230 <p>The argument and return value are floating point numbers of the same
6234 <p>This function returns the sine of the specified operand, returning the same
6235 values as the libm <tt>sin</tt> functions would, and handles error conditions
6236 in the same way.</p>
6240 <!-- _______________________________________________________________________ -->
6241 <div class="doc_subsubsection">
6242 <a name="int_cos">'<tt>llvm.cos.*</tt>' Intrinsic</a>
6245 <div class="doc_text">
6248 <p>This is an overloaded intrinsic. You can use <tt>llvm.cos</tt> on any
6249 floating point or vector of floating point type. Not all targets support all
6253 declare float @llvm.cos.f32(float %Val)
6254 declare double @llvm.cos.f64(double %Val)
6255 declare x86_fp80 @llvm.cos.f80(x86_fp80 %Val)
6256 declare fp128 @llvm.cos.f128(fp128 %Val)
6257 declare ppc_fp128 @llvm.cos.ppcf128(ppc_fp128 %Val)
6261 <p>The '<tt>llvm.cos.*</tt>' intrinsics return the cosine of the operand.</p>
6264 <p>The argument and return value are floating point numbers of the same
6268 <p>This function returns the cosine of the specified operand, returning the same
6269 values as the libm <tt>cos</tt> functions would, and handles error conditions
6270 in the same way.</p>
6274 <!-- _______________________________________________________________________ -->
6275 <div class="doc_subsubsection">
6276 <a name="int_pow">'<tt>llvm.pow.*</tt>' Intrinsic</a>
6279 <div class="doc_text">
6282 <p>This is an overloaded intrinsic. You can use <tt>llvm.pow</tt> on any
6283 floating point or vector of floating point type. Not all targets support all
6287 declare float @llvm.pow.f32(float %Val, float %Power)
6288 declare double @llvm.pow.f64(double %Val, double %Power)
6289 declare x86_fp80 @llvm.pow.f80(x86_fp80 %Val, x86_fp80 %Power)
6290 declare fp128 @llvm.pow.f128(fp128 %Val, fp128 %Power)
6291 declare ppc_fp128 @llvm.pow.ppcf128(ppc_fp128 %Val, ppc_fp128 Power)
6295 <p>The '<tt>llvm.pow.*</tt>' intrinsics return the first operand raised to the
6296 specified (positive or negative) power.</p>
6299 <p>The second argument is a floating point power, and the first is a value to
6300 raise to that power.</p>
6303 <p>This function returns the first value raised to the second power, returning
6304 the same values as the libm <tt>pow</tt> functions would, and handles error
6305 conditions in the same way.</p>
6309 <!-- ======================================================================= -->
6310 <div class="doc_subsection">
6311 <a name="int_manip">Bit Manipulation Intrinsics</a>
6314 <div class="doc_text">
6316 <p>LLVM provides intrinsics for a few important bit manipulation operations.
6317 These allow efficient code generation for some algorithms.</p>
6321 <!-- _______________________________________________________________________ -->
6322 <div class="doc_subsubsection">
6323 <a name="int_bswap">'<tt>llvm.bswap.*</tt>' Intrinsics</a>
6326 <div class="doc_text">
6329 <p>This is an overloaded intrinsic function. You can use bswap on any integer
6330 type that is an even number of bytes (i.e. BitWidth % 16 == 0).</p>
6333 declare i16 @llvm.bswap.i16(i16 <id>)
6334 declare i32 @llvm.bswap.i32(i32 <id>)
6335 declare i64 @llvm.bswap.i64(i64 <id>)
6339 <p>The '<tt>llvm.bswap</tt>' family of intrinsics is used to byte swap integer
6340 values with an even number of bytes (positive multiple of 16 bits). These
6341 are useful for performing operations on data that is not in the target's
6342 native byte order.</p>
6345 <p>The <tt>llvm.bswap.i16</tt> intrinsic returns an i16 value that has the high
6346 and low byte of the input i16 swapped. Similarly,
6347 the <tt>llvm.bswap.i32</tt> intrinsic returns an i32 value that has the four
6348 bytes of the input i32 swapped, so that if the input bytes are numbered 0, 1,
6349 2, 3 then the returned i32 will have its bytes in 3, 2, 1, 0 order.
6350 The <tt>llvm.bswap.i48</tt>, <tt>llvm.bswap.i64</tt> and other intrinsics
6351 extend this concept to additional even-byte lengths (6 bytes, 8 bytes and
6352 more, respectively).</p>
6356 <!-- _______________________________________________________________________ -->
6357 <div class="doc_subsubsection">
6358 <a name="int_ctpop">'<tt>llvm.ctpop.*</tt>' Intrinsic</a>
6361 <div class="doc_text">
6364 <p>This is an overloaded intrinsic. You can use llvm.ctpop on any integer bit
6365 width. Not all targets support all bit widths however.</p>
6368 declare i8 @llvm.ctpop.i8(i8 <src>)
6369 declare i16 @llvm.ctpop.i16(i16 <src>)
6370 declare i32 @llvm.ctpop.i32(i32 <src>)
6371 declare i64 @llvm.ctpop.i64(i64 <src>)
6372 declare i256 @llvm.ctpop.i256(i256 <src>)
6376 <p>The '<tt>llvm.ctpop</tt>' family of intrinsics counts the number of bits set
6380 <p>The only argument is the value to be counted. The argument may be of any
6381 integer type. The return type must match the argument type.</p>
6384 <p>The '<tt>llvm.ctpop</tt>' intrinsic counts the 1's in a variable.</p>
6388 <!-- _______________________________________________________________________ -->
6389 <div class="doc_subsubsection">
6390 <a name="int_ctlz">'<tt>llvm.ctlz.*</tt>' Intrinsic</a>
6393 <div class="doc_text">
6396 <p>This is an overloaded intrinsic. You can use <tt>llvm.ctlz</tt> on any
6397 integer bit width. Not all targets support all bit widths however.</p>
6400 declare i8 @llvm.ctlz.i8 (i8 <src>)
6401 declare i16 @llvm.ctlz.i16(i16 <src>)
6402 declare i32 @llvm.ctlz.i32(i32 <src>)
6403 declare i64 @llvm.ctlz.i64(i64 <src>)
6404 declare i256 @llvm.ctlz.i256(i256 <src>)
6408 <p>The '<tt>llvm.ctlz</tt>' family of intrinsic functions counts the number of
6409 leading zeros in a variable.</p>
6412 <p>The only argument is the value to be counted. The argument may be of any
6413 integer type. The return type must match the argument type.</p>
6416 <p>The '<tt>llvm.ctlz</tt>' intrinsic counts the leading (most significant)
6417 zeros in a variable. If the src == 0 then the result is the size in bits of
6418 the type of src. For example, <tt>llvm.ctlz(i32 2) = 30</tt>.</p>
6422 <!-- _______________________________________________________________________ -->
6423 <div class="doc_subsubsection">
6424 <a name="int_cttz">'<tt>llvm.cttz.*</tt>' Intrinsic</a>
6427 <div class="doc_text">
6430 <p>This is an overloaded intrinsic. You can use <tt>llvm.cttz</tt> on any
6431 integer bit width. Not all targets support all bit widths however.</p>
6434 declare i8 @llvm.cttz.i8 (i8 <src>)
6435 declare i16 @llvm.cttz.i16(i16 <src>)
6436 declare i32 @llvm.cttz.i32(i32 <src>)
6437 declare i64 @llvm.cttz.i64(i64 <src>)
6438 declare i256 @llvm.cttz.i256(i256 <src>)
6442 <p>The '<tt>llvm.cttz</tt>' family of intrinsic functions counts the number of
6446 <p>The only argument is the value to be counted. The argument may be of any
6447 integer type. The return type must match the argument type.</p>
6450 <p>The '<tt>llvm.cttz</tt>' intrinsic counts the trailing (least significant)
6451 zeros in a variable. If the src == 0 then the result is the size in bits of
6452 the type of src. For example, <tt>llvm.cttz(2) = 1</tt>.</p>
6456 <!-- ======================================================================= -->
6457 <div class="doc_subsection">
6458 <a name="int_overflow">Arithmetic with Overflow Intrinsics</a>
6461 <div class="doc_text">
6463 <p>LLVM provides intrinsics for some arithmetic with overflow operations.</p>
6467 <!-- _______________________________________________________________________ -->
6468 <div class="doc_subsubsection">
6469 <a name="int_sadd_overflow">'<tt>llvm.sadd.with.overflow.*</tt>' Intrinsics</a>
6472 <div class="doc_text">
6475 <p>This is an overloaded intrinsic. You can use <tt>llvm.sadd.with.overflow</tt>
6476 on any integer bit width.</p>
6479 declare {i16, i1} @llvm.sadd.with.overflow.i16(i16 %a, i16 %b)
6480 declare {i32, i1} @llvm.sadd.with.overflow.i32(i32 %a, i32 %b)
6481 declare {i64, i1} @llvm.sadd.with.overflow.i64(i64 %a, i64 %b)
6485 <p>The '<tt>llvm.sadd.with.overflow</tt>' family of intrinsic functions perform
6486 a signed addition of the two arguments, and indicate whether an overflow
6487 occurred during the signed summation.</p>
6490 <p>The arguments (%a and %b) and the first element of the result structure may
6491 be of integer types of any bit width, but they must have the same bit
6492 width. The second element of the result structure must be of
6493 type <tt>i1</tt>. <tt>%a</tt> and <tt>%b</tt> are the two values that will
6494 undergo signed addition.</p>
6497 <p>The '<tt>llvm.sadd.with.overflow</tt>' family of intrinsic functions perform
6498 a signed addition of the two variables. They return a structure — the
6499 first element of which is the signed summation, and the second element of
6500 which is a bit specifying if the signed summation resulted in an
6505 %res = call {i32, i1} @llvm.sadd.with.overflow.i32(i32 %a, i32 %b)
6506 %sum = extractvalue {i32, i1} %res, 0
6507 %obit = extractvalue {i32, i1} %res, 1
6508 br i1 %obit, label %overflow, label %normal
6513 <!-- _______________________________________________________________________ -->
6514 <div class="doc_subsubsection">
6515 <a name="int_uadd_overflow">'<tt>llvm.uadd.with.overflow.*</tt>' Intrinsics</a>
6518 <div class="doc_text">
6521 <p>This is an overloaded intrinsic. You can use <tt>llvm.uadd.with.overflow</tt>
6522 on any integer bit width.</p>
6525 declare {i16, i1} @llvm.uadd.with.overflow.i16(i16 %a, i16 %b)
6526 declare {i32, i1} @llvm.uadd.with.overflow.i32(i32 %a, i32 %b)
6527 declare {i64, i1} @llvm.uadd.with.overflow.i64(i64 %a, i64 %b)
6531 <p>The '<tt>llvm.uadd.with.overflow</tt>' family of intrinsic functions perform
6532 an unsigned addition of the two arguments, and indicate whether a carry
6533 occurred during the unsigned summation.</p>
6536 <p>The arguments (%a and %b) and the first element of the result structure may
6537 be of integer types of any bit width, but they must have the same bit
6538 width. The second element of the result structure must be of
6539 type <tt>i1</tt>. <tt>%a</tt> and <tt>%b</tt> are the two values that will
6540 undergo unsigned addition.</p>
6543 <p>The '<tt>llvm.uadd.with.overflow</tt>' family of intrinsic functions perform
6544 an unsigned addition of the two arguments. They return a structure —
6545 the first element of which is the sum, and the second element of which is a
6546 bit specifying if the unsigned summation resulted in a carry.</p>
6550 %res = call {i32, i1} @llvm.uadd.with.overflow.i32(i32 %a, i32 %b)
6551 %sum = extractvalue {i32, i1} %res, 0
6552 %obit = extractvalue {i32, i1} %res, 1
6553 br i1 %obit, label %carry, label %normal
6558 <!-- _______________________________________________________________________ -->
6559 <div class="doc_subsubsection">
6560 <a name="int_ssub_overflow">'<tt>llvm.ssub.with.overflow.*</tt>' Intrinsics</a>
6563 <div class="doc_text">
6566 <p>This is an overloaded intrinsic. You can use <tt>llvm.ssub.with.overflow</tt>
6567 on any integer bit width.</p>
6570 declare {i16, i1} @llvm.ssub.with.overflow.i16(i16 %a, i16 %b)
6571 declare {i32, i1} @llvm.ssub.with.overflow.i32(i32 %a, i32 %b)
6572 declare {i64, i1} @llvm.ssub.with.overflow.i64(i64 %a, i64 %b)
6576 <p>The '<tt>llvm.ssub.with.overflow</tt>' family of intrinsic functions perform
6577 a signed subtraction of the two arguments, and indicate whether an overflow
6578 occurred during the signed subtraction.</p>
6581 <p>The arguments (%a and %b) and the first element of the result structure may
6582 be of integer types of any bit width, but they must have the same bit
6583 width. The second element of the result structure must be of
6584 type <tt>i1</tt>. <tt>%a</tt> and <tt>%b</tt> are the two values that will
6585 undergo signed subtraction.</p>
6588 <p>The '<tt>llvm.ssub.with.overflow</tt>' family of intrinsic functions perform
6589 a signed subtraction of the two arguments. They return a structure —
6590 the first element of which is the subtraction, and the second element of
6591 which is a bit specifying if the signed subtraction resulted in an
6596 %res = call {i32, i1} @llvm.ssub.with.overflow.i32(i32 %a, i32 %b)
6597 %sum = extractvalue {i32, i1} %res, 0
6598 %obit = extractvalue {i32, i1} %res, 1
6599 br i1 %obit, label %overflow, label %normal
6604 <!-- _______________________________________________________________________ -->
6605 <div class="doc_subsubsection">
6606 <a name="int_usub_overflow">'<tt>llvm.usub.with.overflow.*</tt>' Intrinsics</a>
6609 <div class="doc_text">
6612 <p>This is an overloaded intrinsic. You can use <tt>llvm.usub.with.overflow</tt>
6613 on any integer bit width.</p>
6616 declare {i16, i1} @llvm.usub.with.overflow.i16(i16 %a, i16 %b)
6617 declare {i32, i1} @llvm.usub.with.overflow.i32(i32 %a, i32 %b)
6618 declare {i64, i1} @llvm.usub.with.overflow.i64(i64 %a, i64 %b)
6622 <p>The '<tt>llvm.usub.with.overflow</tt>' family of intrinsic functions perform
6623 an unsigned subtraction of the two arguments, and indicate whether an
6624 overflow occurred during the unsigned subtraction.</p>
6627 <p>The arguments (%a and %b) and the first element of the result structure may
6628 be of integer types of any bit width, but they must have the same bit
6629 width. The second element of the result structure must be of
6630 type <tt>i1</tt>. <tt>%a</tt> and <tt>%b</tt> are the two values that will
6631 undergo unsigned subtraction.</p>
6634 <p>The '<tt>llvm.usub.with.overflow</tt>' family of intrinsic functions perform
6635 an unsigned subtraction of the two arguments. They return a structure —
6636 the first element of which is the subtraction, and the second element of
6637 which is a bit specifying if the unsigned subtraction resulted in an
6642 %res = call {i32, i1} @llvm.usub.with.overflow.i32(i32 %a, i32 %b)
6643 %sum = extractvalue {i32, i1} %res, 0
6644 %obit = extractvalue {i32, i1} %res, 1
6645 br i1 %obit, label %overflow, label %normal
6650 <!-- _______________________________________________________________________ -->
6651 <div class="doc_subsubsection">
6652 <a name="int_smul_overflow">'<tt>llvm.smul.with.overflow.*</tt>' Intrinsics</a>
6655 <div class="doc_text">
6658 <p>This is an overloaded intrinsic. You can use <tt>llvm.smul.with.overflow</tt>
6659 on any integer bit width.</p>
6662 declare {i16, i1} @llvm.smul.with.overflow.i16(i16 %a, i16 %b)
6663 declare {i32, i1} @llvm.smul.with.overflow.i32(i32 %a, i32 %b)
6664 declare {i64, i1} @llvm.smul.with.overflow.i64(i64 %a, i64 %b)
6669 <p>The '<tt>llvm.smul.with.overflow</tt>' family of intrinsic functions perform
6670 a signed multiplication of the two arguments, and indicate whether an
6671 overflow occurred during the signed multiplication.</p>
6674 <p>The arguments (%a and %b) and the first element of the result structure may
6675 be of integer types of any bit width, but they must have the same bit
6676 width. The second element of the result structure must be of
6677 type <tt>i1</tt>. <tt>%a</tt> and <tt>%b</tt> are the two values that will
6678 undergo signed multiplication.</p>
6681 <p>The '<tt>llvm.smul.with.overflow</tt>' family of intrinsic functions perform
6682 a signed multiplication of the two arguments. They return a structure —
6683 the first element of which is the multiplication, and the second element of
6684 which is a bit specifying if the signed multiplication resulted in an
6689 %res = call {i32, i1} @llvm.smul.with.overflow.i32(i32 %a, i32 %b)
6690 %sum = extractvalue {i32, i1} %res, 0
6691 %obit = extractvalue {i32, i1} %res, 1
6692 br i1 %obit, label %overflow, label %normal
6697 <!-- _______________________________________________________________________ -->
6698 <div class="doc_subsubsection">
6699 <a name="int_umul_overflow">'<tt>llvm.umul.with.overflow.*</tt>' Intrinsics</a>
6702 <div class="doc_text">
6705 <p>This is an overloaded intrinsic. You can use <tt>llvm.umul.with.overflow</tt>
6706 on any integer bit width.</p>
6709 declare {i16, i1} @llvm.umul.with.overflow.i16(i16 %a, i16 %b)
6710 declare {i32, i1} @llvm.umul.with.overflow.i32(i32 %a, i32 %b)
6711 declare {i64, i1} @llvm.umul.with.overflow.i64(i64 %a, i64 %b)
6715 <p>The '<tt>llvm.umul.with.overflow</tt>' family of intrinsic functions perform
6716 a unsigned multiplication of the two arguments, and indicate whether an
6717 overflow occurred during the unsigned multiplication.</p>
6720 <p>The arguments (%a and %b) and the first element of the result structure may
6721 be of integer types of any bit width, but they must have the same bit
6722 width. The second element of the result structure must be of
6723 type <tt>i1</tt>. <tt>%a</tt> and <tt>%b</tt> are the two values that will
6724 undergo unsigned multiplication.</p>
6727 <p>The '<tt>llvm.umul.with.overflow</tt>' family of intrinsic functions perform
6728 an unsigned multiplication of the two arguments. They return a structure
6729 — the first element of which is the multiplication, and the second
6730 element of which is a bit specifying if the unsigned multiplication resulted
6735 %res = call {i32, i1} @llvm.umul.with.overflow.i32(i32 %a, i32 %b)
6736 %sum = extractvalue {i32, i1} %res, 0
6737 %obit = extractvalue {i32, i1} %res, 1
6738 br i1 %obit, label %overflow, label %normal
6743 <!-- ======================================================================= -->
6744 <div class="doc_subsection">
6745 <a name="int_fp16">Half Precision Floating Point Intrinsics</a>
6748 <div class="doc_text">
6750 <p>Half precision floating point is a storage-only format. This means that it is
6751 a dense encoding (in memory) but does not support computation in the
6754 <p>This means that code must first load the half-precision floating point
6755 value as an i16, then convert it to float with <a
6756 href="#int_convert_from_fp16"><tt>llvm.convert.from.fp16</tt></a>.
6757 Computation can then be performed on the float value (including extending to
6758 double etc). To store the value back to memory, it is first converted to
6759 float if needed, then converted to i16 with
6760 <a href="#int_convert_to_fp16"><tt>llvm.convert.to.fp16</tt></a>, then
6761 storing as an i16 value.</p>
6764 <!-- _______________________________________________________________________ -->
6765 <div class="doc_subsubsection">
6766 <a name="int_convert_to_fp16">'<tt>llvm.convert.to.fp16</tt>' Intrinsic</a>
6769 <div class="doc_text">
6773 declare i16 @llvm.convert.to.fp16(f32 %a)
6777 <p>The '<tt>llvm.convert.to.fp16</tt>' intrinsic function performs
6778 a conversion from single precision floating point format to half precision
6779 floating point format.</p>
6782 <p>The intrinsic function contains single argument - the value to be
6786 <p>The '<tt>llvm.convert.to.fp16</tt>' intrinsic function performs
6787 a conversion from single precision floating point format to half precision
6788 floating point format. The return value is an <tt>i16</tt> which
6789 contains the converted number.</p>
6793 %res = call i16 @llvm.convert.to.fp16(f32 %a)
6794 store i16 %res, i16* @x, align 2
6799 <!-- _______________________________________________________________________ -->
6800 <div class="doc_subsubsection">
6801 <a name="int_convert_from_fp16">'<tt>llvm.convert.from.fp16</tt>' Intrinsic</a>
6804 <div class="doc_text">
6808 declare f32 @llvm.convert.from.fp16(i16 %a)
6812 <p>The '<tt>llvm.convert.from.fp16</tt>' intrinsic function performs
6813 a conversion from half precision floating point format to single precision
6814 floating point format.</p>
6817 <p>The intrinsic function contains single argument - the value to be
6821 <p>The '<tt>llvm.convert.from.fp16</tt>' intrinsic function performs a
6822 conversion from half single precision floating point format to single
6823 precision floating point format. The input half-float value is represented by
6824 an <tt>i16</tt> value.</p>
6828 %a = load i16* @x, align 2
6829 %res = call f32 @llvm.convert.from.fp16(i16 %a)
6834 <!-- ======================================================================= -->
6835 <div class="doc_subsection">
6836 <a name="int_debugger">Debugger Intrinsics</a>
6839 <div class="doc_text">
6841 <p>The LLVM debugger intrinsics (which all start with <tt>llvm.dbg.</tt>
6842 prefix), are described in
6843 the <a href="SourceLevelDebugging.html#format_common_intrinsics">LLVM Source
6844 Level Debugging</a> document.</p>
6848 <!-- ======================================================================= -->
6849 <div class="doc_subsection">
6850 <a name="int_eh">Exception Handling Intrinsics</a>
6853 <div class="doc_text">
6855 <p>The LLVM exception handling intrinsics (which all start with
6856 <tt>llvm.eh.</tt> prefix), are described in
6857 the <a href="ExceptionHandling.html#format_common_intrinsics">LLVM Exception
6858 Handling</a> document.</p>
6862 <!-- ======================================================================= -->
6863 <div class="doc_subsection">
6864 <a name="int_trampoline">Trampoline Intrinsic</a>
6867 <div class="doc_text">
6869 <p>This intrinsic makes it possible to excise one parameter, marked with
6870 the <a href="#nest"><tt>nest</tt></a> attribute, from a function.
6871 The result is a callable
6872 function pointer lacking the nest parameter - the caller does not need to
6873 provide a value for it. Instead, the value to use is stored in advance in a
6874 "trampoline", a block of memory usually allocated on the stack, which also
6875 contains code to splice the nest value into the argument list. This is used
6876 to implement the GCC nested function address extension.</p>
6878 <p>For example, if the function is
6879 <tt>i32 f(i8* nest %c, i32 %x, i32 %y)</tt> then the resulting function
6880 pointer has signature <tt>i32 (i32, i32)*</tt>. It can be created as
6883 <pre class="doc_code">
6884 %tramp = alloca [10 x i8], align 4 ; size and alignment only correct for X86
6885 %tramp1 = getelementptr [10 x i8]* %tramp, i32 0, i32 0
6886 %p = call i8* @llvm.init.trampoline(i8* %tramp1, i8* bitcast (i32 (i8* nest , i32, i32)* @f to i8*), i8* %nval)
6887 %fp = bitcast i8* %p to i32 (i32, i32)*
6890 <p>The call <tt>%val = call i32 %fp(i32 %x, i32 %y)</tt> is then equivalent
6891 to <tt>%val = call i32 %f(i8* %nval, i32 %x, i32 %y)</tt>.</p>
6895 <!-- _______________________________________________________________________ -->
6896 <div class="doc_subsubsection">
6897 <a name="int_it">'<tt>llvm.init.trampoline</tt>' Intrinsic</a>
6900 <div class="doc_text">
6904 declare i8* @llvm.init.trampoline(i8* <tramp>, i8* <func>, i8* <nval>)
6908 <p>This fills the memory pointed to by <tt>tramp</tt> with code and returns a
6909 function pointer suitable for executing it.</p>
6912 <p>The <tt>llvm.init.trampoline</tt> intrinsic takes three arguments, all
6913 pointers. The <tt>tramp</tt> argument must point to a sufficiently large and
6914 sufficiently aligned block of memory; this memory is written to by the
6915 intrinsic. Note that the size and the alignment are target-specific - LLVM
6916 currently provides no portable way of determining them, so a front-end that
6917 generates this intrinsic needs to have some target-specific knowledge.
6918 The <tt>func</tt> argument must hold a function bitcast to
6919 an <tt>i8*</tt>.</p>
6922 <p>The block of memory pointed to by <tt>tramp</tt> is filled with target
6923 dependent code, turning it into a function. A pointer to this function is
6924 returned, but needs to be bitcast to an <a href="#int_trampoline">appropriate
6925 function pointer type</a> before being called. The new function's signature
6926 is the same as that of <tt>func</tt> with any arguments marked with
6927 the <tt>nest</tt> attribute removed. At most one such <tt>nest</tt> argument
6928 is allowed, and it must be of pointer type. Calling the new function is
6929 equivalent to calling <tt>func</tt> with the same argument list, but
6930 with <tt>nval</tt> used for the missing <tt>nest</tt> argument. If, after
6931 calling <tt>llvm.init.trampoline</tt>, the memory pointed to
6932 by <tt>tramp</tt> is modified, then the effect of any later call to the
6933 returned function pointer is undefined.</p>
6937 <!-- ======================================================================= -->
6938 <div class="doc_subsection">
6939 <a name="int_atomics">Atomic Operations and Synchronization Intrinsics</a>
6942 <div class="doc_text">
6944 <p>These intrinsic functions expand the "universal IR" of LLVM to represent
6945 hardware constructs for atomic operations and memory synchronization. This
6946 provides an interface to the hardware, not an interface to the programmer. It
6947 is aimed at a low enough level to allow any programming models or APIs
6948 (Application Programming Interfaces) which need atomic behaviors to map
6949 cleanly onto it. It is also modeled primarily on hardware behavior. Just as
6950 hardware provides a "universal IR" for source languages, it also provides a
6951 starting point for developing a "universal" atomic operation and
6952 synchronization IR.</p>
6954 <p>These do <em>not</em> form an API such as high-level threading libraries,
6955 software transaction memory systems, atomic primitives, and intrinsic
6956 functions as found in BSD, GNU libc, atomic_ops, APR, and other system and
6957 application libraries. The hardware interface provided by LLVM should allow
6958 a clean implementation of all of these APIs and parallel programming models.
6959 No one model or paradigm should be selected above others unless the hardware
6960 itself ubiquitously does so.</p>
6964 <!-- _______________________________________________________________________ -->
6965 <div class="doc_subsubsection">
6966 <a name="int_memory_barrier">'<tt>llvm.memory.barrier</tt>' Intrinsic</a>
6968 <div class="doc_text">
6971 declare void @llvm.memory.barrier(i1 <ll>, i1 <ls>, i1 <sl>, i1 <ss>, i1 <device>)
6975 <p>The <tt>llvm.memory.barrier</tt> intrinsic guarantees ordering between
6976 specific pairs of memory access types.</p>
6979 <p>The <tt>llvm.memory.barrier</tt> intrinsic requires five boolean arguments.
6980 The first four arguments enables a specific barrier as listed below. The
6981 fifth argument specifies that the barrier applies to io or device or uncached
6985 <li><tt>ll</tt>: load-load barrier</li>
6986 <li><tt>ls</tt>: load-store barrier</li>
6987 <li><tt>sl</tt>: store-load barrier</li>
6988 <li><tt>ss</tt>: store-store barrier</li>
6989 <li><tt>device</tt>: barrier applies to device and uncached memory also.</li>
6993 <p>This intrinsic causes the system to enforce some ordering constraints upon
6994 the loads and stores of the program. This barrier does not
6995 indicate <em>when</em> any events will occur, it only enforces
6996 an <em>order</em> in which they occur. For any of the specified pairs of load
6997 and store operations (f.ex. load-load, or store-load), all of the first
6998 operations preceding the barrier will complete before any of the second
6999 operations succeeding the barrier begin. Specifically the semantics for each
7000 pairing is as follows:</p>
7003 <li><tt>ll</tt>: All loads before the barrier must complete before any load
7004 after the barrier begins.</li>
7005 <li><tt>ls</tt>: All loads before the barrier must complete before any
7006 store after the barrier begins.</li>
7007 <li><tt>ss</tt>: All stores before the barrier must complete before any
7008 store after the barrier begins.</li>
7009 <li><tt>sl</tt>: All stores before the barrier must complete before any
7010 load after the barrier begins.</li>
7013 <p>These semantics are applied with a logical "and" behavior when more than one
7014 is enabled in a single memory barrier intrinsic.</p>
7016 <p>Backends may implement stronger barriers than those requested when they do
7017 not support as fine grained a barrier as requested. Some architectures do
7018 not need all types of barriers and on such architectures, these become
7023 %mallocP = tail call i8* @malloc(i32 ptrtoint (i32* getelementptr (i32* null, i32 1) to i32))
7024 %ptr = bitcast i8* %mallocP to i32*
7027 %result1 = load i32* %ptr <i>; yields {i32}:result1 = 4</i>
7028 call void @llvm.memory.barrier(i1 false, i1 true, i1 false, i1 false)
7029 <i>; guarantee the above finishes</i>
7030 store i32 8, %ptr <i>; before this begins</i>
7035 <!-- _______________________________________________________________________ -->
7036 <div class="doc_subsubsection">
7037 <a name="int_atomic_cmp_swap">'<tt>llvm.atomic.cmp.swap.*</tt>' Intrinsic</a>
7040 <div class="doc_text">
7043 <p>This is an overloaded intrinsic. You can use <tt>llvm.atomic.cmp.swap</tt> on
7044 any integer bit width and for different address spaces. Not all targets
7045 support all bit widths however.</p>
7048 declare i8 @llvm.atomic.cmp.swap.i8.p0i8(i8* <ptr>, i8 <cmp>, i8 <val>)
7049 declare i16 @llvm.atomic.cmp.swap.i16.p0i16(i16* <ptr>, i16 <cmp>, i16 <val>)
7050 declare i32 @llvm.atomic.cmp.swap.i32.p0i32(i32* <ptr>, i32 <cmp>, i32 <val>)
7051 declare i64 @llvm.atomic.cmp.swap.i64.p0i64(i64* <ptr>, i64 <cmp>, i64 <val>)
7055 <p>This loads a value in memory and compares it to a given value. If they are
7056 equal, it stores a new value into the memory.</p>
7059 <p>The <tt>llvm.atomic.cmp.swap</tt> intrinsic takes three arguments. The result
7060 as well as both <tt>cmp</tt> and <tt>val</tt> must be integer values with the
7061 same bit width. The <tt>ptr</tt> argument must be a pointer to a value of
7062 this integer type. While any bit width integer may be used, targets may only
7063 lower representations they support in hardware.</p>
7066 <p>This entire intrinsic must be executed atomically. It first loads the value
7067 in memory pointed to by <tt>ptr</tt> and compares it with the
7068 value <tt>cmp</tt>. If they are equal, <tt>val</tt> is stored into the
7069 memory. The loaded value is yielded in all cases. This provides the
7070 equivalent of an atomic compare-and-swap operation within the SSA
7075 %mallocP = tail call i8* @malloc(i32 ptrtoint (i32* getelementptr (i32* null, i32 1) to i32))
7076 %ptr = bitcast i8* %mallocP to i32*
7079 %val1 = add i32 4, 4
7080 %result1 = call i32 @llvm.atomic.cmp.swap.i32.p0i32(i32* %ptr, i32 4, %val1)
7081 <i>; yields {i32}:result1 = 4</i>
7082 %stored1 = icmp eq i32 %result1, 4 <i>; yields {i1}:stored1 = true</i>
7083 %memval1 = load i32* %ptr <i>; yields {i32}:memval1 = 8</i>
7085 %val2 = add i32 1, 1
7086 %result2 = call i32 @llvm.atomic.cmp.swap.i32.p0i32(i32* %ptr, i32 5, %val2)
7087 <i>; yields {i32}:result2 = 8</i>
7088 %stored2 = icmp eq i32 %result2, 5 <i>; yields {i1}:stored2 = false</i>
7090 %memval2 = load i32* %ptr <i>; yields {i32}:memval2 = 8</i>
7095 <!-- _______________________________________________________________________ -->
7096 <div class="doc_subsubsection">
7097 <a name="int_atomic_swap">'<tt>llvm.atomic.swap.*</tt>' Intrinsic</a>
7099 <div class="doc_text">
7102 <p>This is an overloaded intrinsic. You can use <tt>llvm.atomic.swap</tt> on any
7103 integer bit width. Not all targets support all bit widths however.</p>
7106 declare i8 @llvm.atomic.swap.i8.p0i8(i8* <ptr>, i8 <val>)
7107 declare i16 @llvm.atomic.swap.i16.p0i16(i16* <ptr>, i16 <val>)
7108 declare i32 @llvm.atomic.swap.i32.p0i32(i32* <ptr>, i32 <val>)
7109 declare i64 @llvm.atomic.swap.i64.p0i64(i64* <ptr>, i64 <val>)
7113 <p>This intrinsic loads the value stored in memory at <tt>ptr</tt> and yields
7114 the value from memory. It then stores the value in <tt>val</tt> in the memory
7115 at <tt>ptr</tt>.</p>
7118 <p>The <tt>llvm.atomic.swap</tt> intrinsic takes two arguments. Both
7119 the <tt>val</tt> argument and the result must be integers of the same bit
7120 width. The first argument, <tt>ptr</tt>, must be a pointer to a value of this
7121 integer type. The targets may only lower integer representations they
7125 <p>This intrinsic loads the value pointed to by <tt>ptr</tt>, yields it, and
7126 stores <tt>val</tt> back into <tt>ptr</tt> atomically. This provides the
7127 equivalent of an atomic swap operation within the SSA framework.</p>
7131 %mallocP = tail call i8* @malloc(i32 ptrtoint (i32* getelementptr (i32* null, i32 1) to i32))
7132 %ptr = bitcast i8* %mallocP to i32*
7135 %val1 = add i32 4, 4
7136 %result1 = call i32 @llvm.atomic.swap.i32.p0i32(i32* %ptr, i32 %val1)
7137 <i>; yields {i32}:result1 = 4</i>
7138 %stored1 = icmp eq i32 %result1, 4 <i>; yields {i1}:stored1 = true</i>
7139 %memval1 = load i32* %ptr <i>; yields {i32}:memval1 = 8</i>
7141 %val2 = add i32 1, 1
7142 %result2 = call i32 @llvm.atomic.swap.i32.p0i32(i32* %ptr, i32 %val2)
7143 <i>; yields {i32}:result2 = 8</i>
7145 %stored2 = icmp eq i32 %result2, 8 <i>; yields {i1}:stored2 = true</i>
7146 %memval2 = load i32* %ptr <i>; yields {i32}:memval2 = 2</i>
7151 <!-- _______________________________________________________________________ -->
7152 <div class="doc_subsubsection">
7153 <a name="int_atomic_load_add">'<tt>llvm.atomic.load.add.*</tt>' Intrinsic</a>
7157 <div class="doc_text">
7160 <p>This is an overloaded intrinsic. You can use <tt>llvm.atomic.load.add</tt> on
7161 any integer bit width. Not all targets support all bit widths however.</p>
7164 declare i8 @llvm.atomic.load.add.i8.p0i8(i8* <ptr>, i8 <delta>)
7165 declare i16 @llvm.atomic.load.add.i16.p0i16(i16* <ptr>, i16 <delta>)
7166 declare i32 @llvm.atomic.load.add.i32.p0i32(i32* <ptr>, i32 <delta>)
7167 declare i64 @llvm.atomic.load.add.i64.p0i64(i64* <ptr>, i64 <delta>)
7171 <p>This intrinsic adds <tt>delta</tt> to the value stored in memory
7172 at <tt>ptr</tt>. It yields the original value at <tt>ptr</tt>.</p>
7175 <p>The intrinsic takes two arguments, the first a pointer to an integer value
7176 and the second an integer value. The result is also an integer value. These
7177 integer types can have any bit width, but they must all have the same bit
7178 width. The targets may only lower integer representations they support.</p>
7181 <p>This intrinsic does a series of operations atomically. It first loads the
7182 value stored at <tt>ptr</tt>. It then adds <tt>delta</tt>, stores the result
7183 to <tt>ptr</tt>. It yields the original value stored at <tt>ptr</tt>.</p>
7187 %mallocP = tail call i8* @malloc(i32 ptrtoint (i32* getelementptr (i32* null, i32 1) to i32))
7188 %ptr = bitcast i8* %mallocP to i32*
7190 %result1 = call i32 @llvm.atomic.load.add.i32.p0i32(i32* %ptr, i32 4)
7191 <i>; yields {i32}:result1 = 4</i>
7192 %result2 = call i32 @llvm.atomic.load.add.i32.p0i32(i32* %ptr, i32 2)
7193 <i>; yields {i32}:result2 = 8</i>
7194 %result3 = call i32 @llvm.atomic.load.add.i32.p0i32(i32* %ptr, i32 5)
7195 <i>; yields {i32}:result3 = 10</i>
7196 %memval1 = load i32* %ptr <i>; yields {i32}:memval1 = 15</i>
7201 <!-- _______________________________________________________________________ -->
7202 <div class="doc_subsubsection">
7203 <a name="int_atomic_load_sub">'<tt>llvm.atomic.load.sub.*</tt>' Intrinsic</a>
7207 <div class="doc_text">
7210 <p>This is an overloaded intrinsic. You can use <tt>llvm.atomic.load.sub</tt> on
7211 any integer bit width and for different address spaces. Not all targets
7212 support all bit widths however.</p>
7215 declare i8 @llvm.atomic.load.sub.i8.p0i32(i8* <ptr>, i8 <delta>)
7216 declare i16 @llvm.atomic.load.sub.i16.p0i32(i16* <ptr>, i16 <delta>)
7217 declare i32 @llvm.atomic.load.sub.i32.p0i32(i32* <ptr>, i32 <delta>)
7218 declare i64 @llvm.atomic.load.sub.i64.p0i32(i64* <ptr>, i64 <delta>)
7222 <p>This intrinsic subtracts <tt>delta</tt> to the value stored in memory at
7223 <tt>ptr</tt>. It yields the original value at <tt>ptr</tt>.</p>
7226 <p>The intrinsic takes two arguments, the first a pointer to an integer value
7227 and the second an integer value. The result is also an integer value. These
7228 integer types can have any bit width, but they must all have the same bit
7229 width. The targets may only lower integer representations they support.</p>
7232 <p>This intrinsic does a series of operations atomically. It first loads the
7233 value stored at <tt>ptr</tt>. It then subtracts <tt>delta</tt>, stores the
7234 result to <tt>ptr</tt>. It yields the original value stored
7235 at <tt>ptr</tt>.</p>
7239 %mallocP = tail call i8* @malloc(i32 ptrtoint (i32* getelementptr (i32* null, i32 1) to i32))
7240 %ptr = bitcast i8* %mallocP to i32*
7242 %result1 = call i32 @llvm.atomic.load.sub.i32.p0i32(i32* %ptr, i32 4)
7243 <i>; yields {i32}:result1 = 8</i>
7244 %result2 = call i32 @llvm.atomic.load.sub.i32.p0i32(i32* %ptr, i32 2)
7245 <i>; yields {i32}:result2 = 4</i>
7246 %result3 = call i32 @llvm.atomic.load.sub.i32.p0i32(i32* %ptr, i32 5)
7247 <i>; yields {i32}:result3 = 2</i>
7248 %memval1 = load i32* %ptr <i>; yields {i32}:memval1 = -3</i>
7253 <!-- _______________________________________________________________________ -->
7254 <div class="doc_subsubsection">
7255 <a name="int_atomic_load_and">'<tt>llvm.atomic.load.and.*</tt>' Intrinsic</a><br>
7256 <a name="int_atomic_load_nand">'<tt>llvm.atomic.load.nand.*</tt>' Intrinsic</a><br>
7257 <a name="int_atomic_load_or">'<tt>llvm.atomic.load.or.*</tt>' Intrinsic</a><br>
7258 <a name="int_atomic_load_xor">'<tt>llvm.atomic.load.xor.*</tt>' Intrinsic</a><br>
7261 <div class="doc_text">
7264 <p>These are overloaded intrinsics. You can
7265 use <tt>llvm.atomic.load_and</tt>, <tt>llvm.atomic.load_nand</tt>,
7266 <tt>llvm.atomic.load_or</tt>, and <tt>llvm.atomic.load_xor</tt> on any integer
7267 bit width and for different address spaces. Not all targets support all bit
7271 declare i8 @llvm.atomic.load.and.i8.p0i8(i8* <ptr>, i8 <delta>)
7272 declare i16 @llvm.atomic.load.and.i16.p0i16(i16* <ptr>, i16 <delta>)
7273 declare i32 @llvm.atomic.load.and.i32.p0i32(i32* <ptr>, i32 <delta>)
7274 declare i64 @llvm.atomic.load.and.i64.p0i64(i64* <ptr>, i64 <delta>)
7278 declare i8 @llvm.atomic.load.or.i8.p0i8(i8* <ptr>, i8 <delta>)
7279 declare i16 @llvm.atomic.load.or.i16.p0i16(i16* <ptr>, i16 <delta>)
7280 declare i32 @llvm.atomic.load.or.i32.p0i32(i32* <ptr>, i32 <delta>)
7281 declare i64 @llvm.atomic.load.or.i64.p0i64(i64* <ptr>, i64 <delta>)
7285 declare i8 @llvm.atomic.load.nand.i8.p0i32(i8* <ptr>, i8 <delta>)
7286 declare i16 @llvm.atomic.load.nand.i16.p0i32(i16* <ptr>, i16 <delta>)
7287 declare i32 @llvm.atomic.load.nand.i32.p0i32(i32* <ptr>, i32 <delta>)
7288 declare i64 @llvm.atomic.load.nand.i64.p0i32(i64* <ptr>, i64 <delta>)
7292 declare i8 @llvm.atomic.load.xor.i8.p0i32(i8* <ptr>, i8 <delta>)
7293 declare i16 @llvm.atomic.load.xor.i16.p0i32(i16* <ptr>, i16 <delta>)
7294 declare i32 @llvm.atomic.load.xor.i32.p0i32(i32* <ptr>, i32 <delta>)
7295 declare i64 @llvm.atomic.load.xor.i64.p0i32(i64* <ptr>, i64 <delta>)
7299 <p>These intrinsics bitwise the operation (and, nand, or, xor) <tt>delta</tt> to
7300 the value stored in memory at <tt>ptr</tt>. It yields the original value
7301 at <tt>ptr</tt>.</p>
7304 <p>These intrinsics take two arguments, the first a pointer to an integer value
7305 and the second an integer value. The result is also an integer value. These
7306 integer types can have any bit width, but they must all have the same bit
7307 width. The targets may only lower integer representations they support.</p>
7310 <p>These intrinsics does a series of operations atomically. They first load the
7311 value stored at <tt>ptr</tt>. They then do the bitwise
7312 operation <tt>delta</tt>, store the result to <tt>ptr</tt>. They yield the
7313 original value stored at <tt>ptr</tt>.</p>
7317 %mallocP = tail call i8* @malloc(i32 ptrtoint (i32* getelementptr (i32* null, i32 1) to i32))
7318 %ptr = bitcast i8* %mallocP to i32*
7319 store i32 0x0F0F, %ptr
7320 %result0 = call i32 @llvm.atomic.load.nand.i32.p0i32(i32* %ptr, i32 0xFF)
7321 <i>; yields {i32}:result0 = 0x0F0F</i>
7322 %result1 = call i32 @llvm.atomic.load.and.i32.p0i32(i32* %ptr, i32 0xFF)
7323 <i>; yields {i32}:result1 = 0xFFFFFFF0</i>
7324 %result2 = call i32 @llvm.atomic.load.or.i32.p0i32(i32* %ptr, i32 0F)
7325 <i>; yields {i32}:result2 = 0xF0</i>
7326 %result3 = call i32 @llvm.atomic.load.xor.i32.p0i32(i32* %ptr, i32 0F)
7327 <i>; yields {i32}:result3 = FF</i>
7328 %memval1 = load i32* %ptr <i>; yields {i32}:memval1 = F0</i>
7333 <!-- _______________________________________________________________________ -->
7334 <div class="doc_subsubsection">
7335 <a name="int_atomic_load_max">'<tt>llvm.atomic.load.max.*</tt>' Intrinsic</a><br>
7336 <a name="int_atomic_load_min">'<tt>llvm.atomic.load.min.*</tt>' Intrinsic</a><br>
7337 <a name="int_atomic_load_umax">'<tt>llvm.atomic.load.umax.*</tt>' Intrinsic</a><br>
7338 <a name="int_atomic_load_umin">'<tt>llvm.atomic.load.umin.*</tt>' Intrinsic</a><br>
7341 <div class="doc_text">
7344 <p>These are overloaded intrinsics. You can use <tt>llvm.atomic.load_max</tt>,
7345 <tt>llvm.atomic.load_min</tt>, <tt>llvm.atomic.load_umax</tt>, and
7346 <tt>llvm.atomic.load_umin</tt> on any integer bit width and for different
7347 address spaces. Not all targets support all bit widths however.</p>
7350 declare i8 @llvm.atomic.load.max.i8.p0i8(i8* <ptr>, i8 <delta>)
7351 declare i16 @llvm.atomic.load.max.i16.p0i16(i16* <ptr>, i16 <delta>)
7352 declare i32 @llvm.atomic.load.max.i32.p0i32(i32* <ptr>, i32 <delta>)
7353 declare i64 @llvm.atomic.load.max.i64.p0i64(i64* <ptr>, i64 <delta>)
7357 declare i8 @llvm.atomic.load.min.i8.p0i8(i8* <ptr>, i8 <delta>)
7358 declare i16 @llvm.atomic.load.min.i16.p0i16(i16* <ptr>, i16 <delta>)
7359 declare i32 @llvm.atomic.load.min.i32.p0i32(i32* <ptr>, i32 <delta>)
7360 declare i64 @llvm.atomic.load.min.i64.p0i64(i64* <ptr>, i64 <delta>)
7364 declare i8 @llvm.atomic.load.umax.i8.p0i8(i8* <ptr>, i8 <delta>)
7365 declare i16 @llvm.atomic.load.umax.i16.p0i16(i16* <ptr>, i16 <delta>)
7366 declare i32 @llvm.atomic.load.umax.i32.p0i32(i32* <ptr>, i32 <delta>)
7367 declare i64 @llvm.atomic.load.umax.i64.p0i64(i64* <ptr>, i64 <delta>)
7371 declare i8 @llvm.atomic.load.umin.i8.p0i8(i8* <ptr>, i8 <delta>)
7372 declare i16 @llvm.atomic.load.umin.i16.p0i16(i16* <ptr>, i16 <delta>)
7373 declare i32 @llvm.atomic.load.umin.i32.p0i32(i32* <ptr>, i32 <delta>)
7374 declare i64 @llvm.atomic.load.umin.i64.p0i64(i64* <ptr>, i64 <delta>)
7378 <p>These intrinsics takes the signed or unsigned minimum or maximum of
7379 <tt>delta</tt> and the value stored in memory at <tt>ptr</tt>. It yields the
7380 original value at <tt>ptr</tt>.</p>
7383 <p>These intrinsics take two arguments, the first a pointer to an integer value
7384 and the second an integer value. The result is also an integer value. These
7385 integer types can have any bit width, but they must all have the same bit
7386 width. The targets may only lower integer representations they support.</p>
7389 <p>These intrinsics does a series of operations atomically. They first load the
7390 value stored at <tt>ptr</tt>. They then do the signed or unsigned min or
7391 max <tt>delta</tt> and the value, store the result to <tt>ptr</tt>. They
7392 yield the original value stored at <tt>ptr</tt>.</p>
7396 %mallocP = tail call i8* @malloc(i32 ptrtoint (i32* getelementptr (i32* null, i32 1) to i32))
7397 %ptr = bitcast i8* %mallocP to i32*
7399 %result0 = call i32 @llvm.atomic.load.min.i32.p0i32(i32* %ptr, i32 -2)
7400 <i>; yields {i32}:result0 = 7</i>
7401 %result1 = call i32 @llvm.atomic.load.max.i32.p0i32(i32* %ptr, i32 8)
7402 <i>; yields {i32}:result1 = -2</i>
7403 %result2 = call i32 @llvm.atomic.load.umin.i32.p0i32(i32* %ptr, i32 10)
7404 <i>; yields {i32}:result2 = 8</i>
7405 %result3 = call i32 @llvm.atomic.load.umax.i32.p0i32(i32* %ptr, i32 30)
7406 <i>; yields {i32}:result3 = 8</i>
7407 %memval1 = load i32* %ptr <i>; yields {i32}:memval1 = 30</i>
7413 <!-- ======================================================================= -->
7414 <div class="doc_subsection">
7415 <a name="int_memorymarkers">Memory Use Markers</a>
7418 <div class="doc_text">
7420 <p>This class of intrinsics exists to information about the lifetime of memory
7421 objects and ranges where variables are immutable.</p>
7425 <!-- _______________________________________________________________________ -->
7426 <div class="doc_subsubsection">
7427 <a name="int_lifetime_start">'<tt>llvm.lifetime.start</tt>' Intrinsic</a>
7430 <div class="doc_text">
7434 declare void @llvm.lifetime.start(i64 <size>, i8* nocapture <ptr>)
7438 <p>The '<tt>llvm.lifetime.start</tt>' intrinsic specifies the start of a memory
7439 object's lifetime.</p>
7442 <p>The first argument is a constant integer representing the size of the
7443 object, or -1 if it is variable sized. The second argument is a pointer to
7447 <p>This intrinsic indicates that before this point in the code, the value of the
7448 memory pointed to by <tt>ptr</tt> is dead. This means that it is known to
7449 never be used and has an undefined value. A load from the pointer that
7450 precedes this intrinsic can be replaced with
7451 <tt>'<a href="#undefvalues">undef</a>'</tt>.</p>
7455 <!-- _______________________________________________________________________ -->
7456 <div class="doc_subsubsection">
7457 <a name="int_lifetime_end">'<tt>llvm.lifetime.end</tt>' Intrinsic</a>
7460 <div class="doc_text">
7464 declare void @llvm.lifetime.end(i64 <size>, i8* nocapture <ptr>)
7468 <p>The '<tt>llvm.lifetime.end</tt>' intrinsic specifies the end of a memory
7469 object's lifetime.</p>
7472 <p>The first argument is a constant integer representing the size of the
7473 object, or -1 if it is variable sized. The second argument is a pointer to
7477 <p>This intrinsic indicates that after this point in the code, the value of the
7478 memory pointed to by <tt>ptr</tt> is dead. This means that it is known to
7479 never be used and has an undefined value. Any stores into the memory object
7480 following this intrinsic may be removed as dead.
7484 <!-- _______________________________________________________________________ -->
7485 <div class="doc_subsubsection">
7486 <a name="int_invariant_start">'<tt>llvm.invariant.start</tt>' Intrinsic</a>
7489 <div class="doc_text">
7493 declare {}* @llvm.invariant.start(i64 <size>, i8* nocapture <ptr>) readonly
7497 <p>The '<tt>llvm.invariant.start</tt>' intrinsic specifies that the contents of
7498 a memory object will not change.</p>
7501 <p>The first argument is a constant integer representing the size of the
7502 object, or -1 if it is variable sized. The second argument is a pointer to
7506 <p>This intrinsic indicates that until an <tt>llvm.invariant.end</tt> that uses
7507 the return value, the referenced memory location is constant and
7512 <!-- _______________________________________________________________________ -->
7513 <div class="doc_subsubsection">
7514 <a name="int_invariant_end">'<tt>llvm.invariant.end</tt>' Intrinsic</a>
7517 <div class="doc_text">
7521 declare void @llvm.invariant.end({}* <start>, i64 <size>, i8* nocapture <ptr>)
7525 <p>The '<tt>llvm.invariant.end</tt>' intrinsic specifies that the contents of
7526 a memory object are mutable.</p>
7529 <p>The first argument is the matching <tt>llvm.invariant.start</tt> intrinsic.
7530 The second argument is a constant integer representing the size of the
7531 object, or -1 if it is variable sized and the third argument is a pointer
7535 <p>This intrinsic indicates that the memory is mutable again.</p>
7539 <!-- ======================================================================= -->
7540 <div class="doc_subsection">
7541 <a name="int_general">General Intrinsics</a>
7544 <div class="doc_text">
7546 <p>This class of intrinsics is designed to be generic and has no specific
7551 <!-- _______________________________________________________________________ -->
7552 <div class="doc_subsubsection">
7553 <a name="int_var_annotation">'<tt>llvm.var.annotation</tt>' Intrinsic</a>
7556 <div class="doc_text">
7560 declare void @llvm.var.annotation(i8* <val>, i8* <str>, i8* <str>, i32 <int>)
7564 <p>The '<tt>llvm.var.annotation</tt>' intrinsic.</p>
7567 <p>The first argument is a pointer to a value, the second is a pointer to a
7568 global string, the third is a pointer to a global string which is the source
7569 file name, and the last argument is the line number.</p>
7572 <p>This intrinsic allows annotation of local variables with arbitrary strings.
7573 This can be useful for special purpose optimizations that want to look for
7574 these annotations. These have no other defined use, they are ignored by code
7575 generation and optimization.</p>
7579 <!-- _______________________________________________________________________ -->
7580 <div class="doc_subsubsection">
7581 <a name="int_annotation">'<tt>llvm.annotation.*</tt>' Intrinsic</a>
7584 <div class="doc_text">
7587 <p>This is an overloaded intrinsic. You can use '<tt>llvm.annotation</tt>' on
7588 any integer bit width.</p>
7591 declare i8 @llvm.annotation.i8(i8 <val>, i8* <str>, i8* <str>, i32 <int>)
7592 declare i16 @llvm.annotation.i16(i16 <val>, i8* <str>, i8* <str>, i32 <int>)
7593 declare i32 @llvm.annotation.i32(i32 <val>, i8* <str>, i8* <str>, i32 <int>)
7594 declare i64 @llvm.annotation.i64(i64 <val>, i8* <str>, i8* <str>, i32 <int>)
7595 declare i256 @llvm.annotation.i256(i256 <val>, i8* <str>, i8* <str>, i32 <int>)
7599 <p>The '<tt>llvm.annotation</tt>' intrinsic.</p>
7602 <p>The first argument is an integer value (result of some expression), the
7603 second is a pointer to a global string, the third is a pointer to a global
7604 string which is the source file name, and the last argument is the line
7605 number. It returns the value of the first argument.</p>
7608 <p>This intrinsic allows annotations to be put on arbitrary expressions with
7609 arbitrary strings. This can be useful for special purpose optimizations that
7610 want to look for these annotations. These have no other defined use, they
7611 are ignored by code generation and optimization.</p>
7615 <!-- _______________________________________________________________________ -->
7616 <div class="doc_subsubsection">
7617 <a name="int_trap">'<tt>llvm.trap</tt>' Intrinsic</a>
7620 <div class="doc_text">
7624 declare void @llvm.trap()
7628 <p>The '<tt>llvm.trap</tt>' intrinsic.</p>
7634 <p>This intrinsics is lowered to the target dependent trap instruction. If the
7635 target does not have a trap instruction, this intrinsic will be lowered to
7636 the call of the <tt>abort()</tt> function.</p>
7640 <!-- _______________________________________________________________________ -->
7641 <div class="doc_subsubsection">
7642 <a name="int_stackprotector">'<tt>llvm.stackprotector</tt>' Intrinsic</a>
7645 <div class="doc_text">
7649 declare void @llvm.stackprotector(i8* <guard>, i8** <slot>)
7653 <p>The <tt>llvm.stackprotector</tt> intrinsic takes the <tt>guard</tt> and
7654 stores it onto the stack at <tt>slot</tt>. The stack slot is adjusted to
7655 ensure that it is placed on the stack before local variables.</p>
7658 <p>The <tt>llvm.stackprotector</tt> intrinsic requires two pointer
7659 arguments. The first argument is the value loaded from the stack
7660 guard <tt>@__stack_chk_guard</tt>. The second variable is an <tt>alloca</tt>
7661 that has enough space to hold the value of the guard.</p>
7664 <p>This intrinsic causes the prologue/epilogue inserter to force the position of
7665 the <tt>AllocaInst</tt> stack slot to be before local variables on the
7666 stack. This is to ensure that if a local variable on the stack is
7667 overwritten, it will destroy the value of the guard. When the function exits,
7668 the guard on the stack is checked against the original guard. If they're
7669 different, then the program aborts by calling the <tt>__stack_chk_fail()</tt>
7674 <!-- _______________________________________________________________________ -->
7675 <div class="doc_subsubsection">
7676 <a name="int_objectsize">'<tt>llvm.objectsize</tt>' Intrinsic</a>
7679 <div class="doc_text">
7683 declare i32 @llvm.objectsize.i32(i8* <object>, i1 <type>)
7684 declare i64 @llvm.objectsize.i64(i8* <object>, i1 <type>)
7688 <p>The <tt>llvm.objectsize</tt> intrinsic is designed to provide information
7689 to the optimizers to discover at compile time either a) when an
7690 operation like memcpy will either overflow a buffer that corresponds to
7691 an object, or b) to determine that a runtime check for overflow isn't
7692 necessary. An object in this context means an allocation of a
7693 specific class, structure, array, or other object.</p>
7696 <p>The <tt>llvm.objectsize</tt> intrinsic takes two arguments. The first
7697 argument is a pointer to or into the <tt>object</tt>. The second argument
7698 is a boolean 0 or 1. This argument determines whether you want the
7699 maximum (0) or minimum (1) bytes remaining. This needs to be a literal 0 or
7700 1, variables are not allowed.</p>
7703 <p>The <tt>llvm.objectsize</tt> intrinsic is lowered to either a constant
7704 representing the size of the object concerned or <tt>i32/i64 -1 or 0</tt>
7705 (depending on the <tt>type</tt> argument if the size cannot be determined
7706 at compile time.</p>
7710 <!-- *********************************************************************** -->
7713 <a href="http://jigsaw.w3.org/css-validator/check/referer"><img
7714 src="http://jigsaw.w3.org/css-validator/images/vcss-blue" alt="Valid CSS"></a>
7715 <a href="http://validator.w3.org/check/referer"><img
7716 src="http://www.w3.org/Icons/valid-html401-blue" alt="Valid HTML 4.01"></a>
7718 <a href="mailto:sabre@nondot.org">Chris Lattner</a><br>
7719 <a href="http://llvm.org">The LLVM Compiler Infrastructure</a><br>
7720 Last modified: $Date$