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_weak">'<tt>linker_weak</tt>' Linkage</a></li>
28 <li><a href="#linkage_internal">'<tt>internal</tt>' Linkage</a></li>
29 <li><a href="#linkage_available_externally">'<tt>available_externally</tt>' Linkage</a></li>
30 <li><a href="#linkage_linkonce">'<tt>linkonce</tt>' Linkage</a></li>
31 <li><a href="#linkage_common">'<tt>common</tt>' Linkage</a></li>
32 <li><a href="#linkage_weak">'<tt>weak</tt>' Linkage</a></li>
33 <li><a href="#linkage_appending">'<tt>appending</tt>' Linkage</a></li>
34 <li><a href="#linkage_externweak">'<tt>extern_weak</tt>' Linkage</a></li>
35 <li><a href="#linkage_linkonce_odr">'<tt>linkonce_odr</tt>' Linkage</a></li>
36 <li><a href="#linkage_weak">'<tt>weak_odr</tt>' Linkage</a></li>
37 <li><a href="#linkage_external">'<tt>externally visible</tt>' Linkage</a></li>
38 <li><a href="#linkage_dllimport">'<tt>dllimport</tt>' Linkage</a></li>
39 <li><a href="#linkage_dllexport">'<tt>dllexport</tt>' Linkage</a></li>
42 <li><a href="#callingconv">Calling Conventions</a></li>
43 <li><a href="#namedtypes">Named Types</a></li>
44 <li><a href="#globalvars">Global Variables</a></li>
45 <li><a href="#functionstructure">Functions</a></li>
46 <li><a href="#aliasstructure">Aliases</a></li>
47 <li><a href="#namedmetadatastructure">Named Metadata</a></li>
48 <li><a href="#paramattrs">Parameter Attributes</a></li>
49 <li><a href="#fnattrs">Function Attributes</a></li>
50 <li><a href="#gc">Garbage Collector Names</a></li>
51 <li><a href="#moduleasm">Module-Level Inline Assembly</a></li>
52 <li><a href="#datalayout">Data Layout</a></li>
53 <li><a href="#pointeraliasing">Pointer Aliasing Rules</a></li>
54 <li><a href="#volatile">Volatile Memory Accesses</a></li>
57 <li><a href="#typesystem">Type System</a>
59 <li><a href="#t_classifications">Type Classifications</a></li>
60 <li><a href="#t_primitive">Primitive Types</a>
62 <li><a href="#t_integer">Integer Type</a></li>
63 <li><a href="#t_floating">Floating Point Types</a></li>
64 <li><a href="#t_void">Void Type</a></li>
65 <li><a href="#t_label">Label Type</a></li>
66 <li><a href="#t_metadata">Metadata Type</a></li>
69 <li><a href="#t_derived">Derived Types</a>
71 <li><a href="#t_aggregate">Aggregate Types</a>
73 <li><a href="#t_array">Array Type</a></li>
74 <li><a href="#t_struct">Structure Type</a></li>
75 <li><a href="#t_pstruct">Packed Structure Type</a></li>
76 <li><a href="#t_union">Union Type</a></li>
77 <li><a href="#t_vector">Vector Type</a></li>
80 <li><a href="#t_function">Function Type</a></li>
81 <li><a href="#t_pointer">Pointer Type</a></li>
82 <li><a href="#t_opaque">Opaque Type</a></li>
85 <li><a href="#t_uprefs">Type Up-references</a></li>
88 <li><a href="#constants">Constants</a>
90 <li><a href="#simpleconstants">Simple Constants</a></li>
91 <li><a href="#complexconstants">Complex Constants</a></li>
92 <li><a href="#globalconstants">Global Variable and Function Addresses</a></li>
93 <li><a href="#undefvalues">Undefined Values</a></li>
94 <li><a href="#trapvalues">Trap Values</a></li>
95 <li><a href="#blockaddress">Addresses of Basic Blocks</a></li>
96 <li><a href="#constantexprs">Constant Expressions</a></li>
99 <li><a href="#othervalues">Other Values</a>
101 <li><a href="#inlineasm">Inline Assembler Expressions</a></li>
102 <li><a href="#metadata">Metadata Nodes and Metadata Strings</a></li>
105 <li><a href="#intrinsic_globals">Intrinsic Global Variables</a>
107 <li><a href="#intg_used">The '<tt>llvm.used</tt>' Global Variable</a></li>
108 <li><a href="#intg_compiler_used">The '<tt>llvm.compiler.used</tt>'
109 Global Variable</a></li>
110 <li><a href="#intg_global_ctors">The '<tt>llvm.global_ctors</tt>'
111 Global Variable</a></li>
112 <li><a href="#intg_global_dtors">The '<tt>llvm.global_dtors</tt>'
113 Global Variable</a></li>
116 <li><a href="#instref">Instruction Reference</a>
118 <li><a href="#terminators">Terminator Instructions</a>
120 <li><a href="#i_ret">'<tt>ret</tt>' Instruction</a></li>
121 <li><a href="#i_br">'<tt>br</tt>' Instruction</a></li>
122 <li><a href="#i_switch">'<tt>switch</tt>' Instruction</a></li>
123 <li><a href="#i_indirectbr">'<tt>indirectbr</tt>' Instruction</a></li>
124 <li><a href="#i_invoke">'<tt>invoke</tt>' Instruction</a></li>
125 <li><a href="#i_unwind">'<tt>unwind</tt>' Instruction</a></li>
126 <li><a href="#i_unreachable">'<tt>unreachable</tt>' Instruction</a></li>
129 <li><a href="#binaryops">Binary Operations</a>
131 <li><a href="#i_add">'<tt>add</tt>' Instruction</a></li>
132 <li><a href="#i_fadd">'<tt>fadd</tt>' Instruction</a></li>
133 <li><a href="#i_sub">'<tt>sub</tt>' Instruction</a></li>
134 <li><a href="#i_fsub">'<tt>fsub</tt>' Instruction</a></li>
135 <li><a href="#i_mul">'<tt>mul</tt>' Instruction</a></li>
136 <li><a href="#i_fmul">'<tt>fmul</tt>' Instruction</a></li>
137 <li><a href="#i_udiv">'<tt>udiv</tt>' Instruction</a></li>
138 <li><a href="#i_sdiv">'<tt>sdiv</tt>' Instruction</a></li>
139 <li><a href="#i_fdiv">'<tt>fdiv</tt>' Instruction</a></li>
140 <li><a href="#i_urem">'<tt>urem</tt>' Instruction</a></li>
141 <li><a href="#i_srem">'<tt>srem</tt>' Instruction</a></li>
142 <li><a href="#i_frem">'<tt>frem</tt>' Instruction</a></li>
145 <li><a href="#bitwiseops">Bitwise Binary Operations</a>
147 <li><a href="#i_shl">'<tt>shl</tt>' Instruction</a></li>
148 <li><a href="#i_lshr">'<tt>lshr</tt>' Instruction</a></li>
149 <li><a href="#i_ashr">'<tt>ashr</tt>' Instruction</a></li>
150 <li><a href="#i_and">'<tt>and</tt>' Instruction</a></li>
151 <li><a href="#i_or">'<tt>or</tt>' Instruction</a></li>
152 <li><a href="#i_xor">'<tt>xor</tt>' Instruction</a></li>
155 <li><a href="#vectorops">Vector Operations</a>
157 <li><a href="#i_extractelement">'<tt>extractelement</tt>' Instruction</a></li>
158 <li><a href="#i_insertelement">'<tt>insertelement</tt>' Instruction</a></li>
159 <li><a href="#i_shufflevector">'<tt>shufflevector</tt>' Instruction</a></li>
162 <li><a href="#aggregateops">Aggregate Operations</a>
164 <li><a href="#i_extractvalue">'<tt>extractvalue</tt>' Instruction</a></li>
165 <li><a href="#i_insertvalue">'<tt>insertvalue</tt>' Instruction</a></li>
168 <li><a href="#memoryops">Memory Access and Addressing Operations</a>
170 <li><a href="#i_alloca">'<tt>alloca</tt>' Instruction</a></li>
171 <li><a href="#i_load">'<tt>load</tt>' Instruction</a></li>
172 <li><a href="#i_store">'<tt>store</tt>' Instruction</a></li>
173 <li><a href="#i_getelementptr">'<tt>getelementptr</tt>' Instruction</a></li>
176 <li><a href="#convertops">Conversion Operations</a>
178 <li><a href="#i_trunc">'<tt>trunc .. to</tt>' Instruction</a></li>
179 <li><a href="#i_zext">'<tt>zext .. to</tt>' Instruction</a></li>
180 <li><a href="#i_sext">'<tt>sext .. to</tt>' Instruction</a></li>
181 <li><a href="#i_fptrunc">'<tt>fptrunc .. to</tt>' Instruction</a></li>
182 <li><a href="#i_fpext">'<tt>fpext .. to</tt>' Instruction</a></li>
183 <li><a href="#i_fptoui">'<tt>fptoui .. to</tt>' Instruction</a></li>
184 <li><a href="#i_fptosi">'<tt>fptosi .. to</tt>' Instruction</a></li>
185 <li><a href="#i_uitofp">'<tt>uitofp .. to</tt>' Instruction</a></li>
186 <li><a href="#i_sitofp">'<tt>sitofp .. to</tt>' Instruction</a></li>
187 <li><a href="#i_ptrtoint">'<tt>ptrtoint .. to</tt>' Instruction</a></li>
188 <li><a href="#i_inttoptr">'<tt>inttoptr .. to</tt>' Instruction</a></li>
189 <li><a href="#i_bitcast">'<tt>bitcast .. to</tt>' Instruction</a></li>
192 <li><a href="#otherops">Other Operations</a>
194 <li><a href="#i_icmp">'<tt>icmp</tt>' Instruction</a></li>
195 <li><a href="#i_fcmp">'<tt>fcmp</tt>' Instruction</a></li>
196 <li><a href="#i_phi">'<tt>phi</tt>' Instruction</a></li>
197 <li><a href="#i_select">'<tt>select</tt>' Instruction</a></li>
198 <li><a href="#i_call">'<tt>call</tt>' Instruction</a></li>
199 <li><a href="#i_va_arg">'<tt>va_arg</tt>' Instruction</a></li>
204 <li><a href="#intrinsics">Intrinsic Functions</a>
206 <li><a href="#int_varargs">Variable Argument Handling Intrinsics</a>
208 <li><a href="#int_va_start">'<tt>llvm.va_start</tt>' Intrinsic</a></li>
209 <li><a href="#int_va_end">'<tt>llvm.va_end</tt>' Intrinsic</a></li>
210 <li><a href="#int_va_copy">'<tt>llvm.va_copy</tt>' Intrinsic</a></li>
213 <li><a href="#int_gc">Accurate Garbage Collection Intrinsics</a>
215 <li><a href="#int_gcroot">'<tt>llvm.gcroot</tt>' Intrinsic</a></li>
216 <li><a href="#int_gcread">'<tt>llvm.gcread</tt>' Intrinsic</a></li>
217 <li><a href="#int_gcwrite">'<tt>llvm.gcwrite</tt>' Intrinsic</a></li>
220 <li><a href="#int_codegen">Code Generator Intrinsics</a>
222 <li><a href="#int_returnaddress">'<tt>llvm.returnaddress</tt>' Intrinsic</a></li>
223 <li><a href="#int_frameaddress">'<tt>llvm.frameaddress</tt>' Intrinsic</a></li>
224 <li><a href="#int_stacksave">'<tt>llvm.stacksave</tt>' Intrinsic</a></li>
225 <li><a href="#int_stackrestore">'<tt>llvm.stackrestore</tt>' Intrinsic</a></li>
226 <li><a href="#int_prefetch">'<tt>llvm.prefetch</tt>' Intrinsic</a></li>
227 <li><a href="#int_pcmarker">'<tt>llvm.pcmarker</tt>' Intrinsic</a></li>
228 <li><a href="#int_readcyclecounter">'<tt>llvm.readcyclecounter</tt>' Intrinsic</a></li>
231 <li><a href="#int_libc">Standard C Library Intrinsics</a>
233 <li><a href="#int_memcpy">'<tt>llvm.memcpy.*</tt>' Intrinsic</a></li>
234 <li><a href="#int_memmove">'<tt>llvm.memmove.*</tt>' Intrinsic</a></li>
235 <li><a href="#int_memset">'<tt>llvm.memset.*</tt>' Intrinsic</a></li>
236 <li><a href="#int_sqrt">'<tt>llvm.sqrt.*</tt>' Intrinsic</a></li>
237 <li><a href="#int_powi">'<tt>llvm.powi.*</tt>' Intrinsic</a></li>
238 <li><a href="#int_sin">'<tt>llvm.sin.*</tt>' Intrinsic</a></li>
239 <li><a href="#int_cos">'<tt>llvm.cos.*</tt>' Intrinsic</a></li>
240 <li><a href="#int_pow">'<tt>llvm.pow.*</tt>' Intrinsic</a></li>
243 <li><a href="#int_manip">Bit Manipulation Intrinsics</a>
245 <li><a href="#int_bswap">'<tt>llvm.bswap.*</tt>' Intrinsics</a></li>
246 <li><a href="#int_ctpop">'<tt>llvm.ctpop.*</tt>' Intrinsic </a></li>
247 <li><a href="#int_ctlz">'<tt>llvm.ctlz.*</tt>' Intrinsic </a></li>
248 <li><a href="#int_cttz">'<tt>llvm.cttz.*</tt>' Intrinsic </a></li>
251 <li><a href="#int_overflow">Arithmetic with Overflow Intrinsics</a>
253 <li><a href="#int_sadd_overflow">'<tt>llvm.sadd.with.overflow.*</tt> Intrinsics</a></li>
254 <li><a href="#int_uadd_overflow">'<tt>llvm.uadd.with.overflow.*</tt> Intrinsics</a></li>
255 <li><a href="#int_ssub_overflow">'<tt>llvm.ssub.with.overflow.*</tt> Intrinsics</a></li>
256 <li><a href="#int_usub_overflow">'<tt>llvm.usub.with.overflow.*</tt> Intrinsics</a></li>
257 <li><a href="#int_smul_overflow">'<tt>llvm.smul.with.overflow.*</tt> Intrinsics</a></li>
258 <li><a href="#int_umul_overflow">'<tt>llvm.umul.with.overflow.*</tt> Intrinsics</a></li>
261 <li><a href="#int_fp16">Half Precision Floating Point Intrinsics</a>
263 <li><a href="#int_convert_to_fp16">'<tt>llvm.convert.to.fp16</tt>' Intrinsic</a></li>
264 <li><a href="#int_convert_from_fp16">'<tt>llvm.convert.from.fp16</tt>' Intrinsic</a></li>
267 <li><a href="#int_debugger">Debugger intrinsics</a></li>
268 <li><a href="#int_eh">Exception Handling intrinsics</a></li>
269 <li><a href="#int_trampoline">Trampoline Intrinsic</a>
271 <li><a href="#int_it">'<tt>llvm.init.trampoline</tt>' Intrinsic</a></li>
274 <li><a href="#int_atomics">Atomic intrinsics</a>
276 <li><a href="#int_memory_barrier"><tt>llvm.memory_barrier</tt></a></li>
277 <li><a href="#int_atomic_cmp_swap"><tt>llvm.atomic.cmp.swap</tt></a></li>
278 <li><a href="#int_atomic_swap"><tt>llvm.atomic.swap</tt></a></li>
279 <li><a href="#int_atomic_load_add"><tt>llvm.atomic.load.add</tt></a></li>
280 <li><a href="#int_atomic_load_sub"><tt>llvm.atomic.load.sub</tt></a></li>
281 <li><a href="#int_atomic_load_and"><tt>llvm.atomic.load.and</tt></a></li>
282 <li><a href="#int_atomic_load_nand"><tt>llvm.atomic.load.nand</tt></a></li>
283 <li><a href="#int_atomic_load_or"><tt>llvm.atomic.load.or</tt></a></li>
284 <li><a href="#int_atomic_load_xor"><tt>llvm.atomic.load.xor</tt></a></li>
285 <li><a href="#int_atomic_load_max"><tt>llvm.atomic.load.max</tt></a></li>
286 <li><a href="#int_atomic_load_min"><tt>llvm.atomic.load.min</tt></a></li>
287 <li><a href="#int_atomic_load_umax"><tt>llvm.atomic.load.umax</tt></a></li>
288 <li><a href="#int_atomic_load_umin"><tt>llvm.atomic.load.umin</tt></a></li>
291 <li><a href="#int_memorymarkers">Memory Use Markers</a>
293 <li><a href="#int_lifetime_start"><tt>llvm.lifetime.start</tt></a></li>
294 <li><a href="#int_lifetime_end"><tt>llvm.lifetime.end</tt></a></li>
295 <li><a href="#int_invariant_start"><tt>llvm.invariant.start</tt></a></li>
296 <li><a href="#int_invariant_end"><tt>llvm.invariant.end</tt></a></li>
299 <li><a href="#int_general">General intrinsics</a>
301 <li><a href="#int_var_annotation">
302 '<tt>llvm.var.annotation</tt>' Intrinsic</a></li>
303 <li><a href="#int_annotation">
304 '<tt>llvm.annotation.*</tt>' Intrinsic</a></li>
305 <li><a href="#int_trap">
306 '<tt>llvm.trap</tt>' Intrinsic</a></li>
307 <li><a href="#int_stackprotector">
308 '<tt>llvm.stackprotector</tt>' Intrinsic</a></li>
309 <li><a href="#int_objectsize">
310 '<tt>llvm.objectsize</tt>' Intrinsic</a></li>
317 <div class="doc_author">
318 <p>Written by <a href="mailto:sabre@nondot.org">Chris Lattner</a>
319 and <a href="mailto:vadve@cs.uiuc.edu">Vikram Adve</a></p>
322 <!-- *********************************************************************** -->
323 <div class="doc_section"> <a name="abstract">Abstract </a></div>
324 <!-- *********************************************************************** -->
326 <div class="doc_text">
328 <p>This document is a reference manual for the LLVM assembly language. LLVM is
329 a Static Single Assignment (SSA) based representation that provides type
330 safety, low-level operations, flexibility, and the capability of representing
331 'all' high-level languages cleanly. It is the common code representation
332 used throughout all phases of the LLVM compilation strategy.</p>
336 <!-- *********************************************************************** -->
337 <div class="doc_section"> <a name="introduction">Introduction</a> </div>
338 <!-- *********************************************************************** -->
340 <div class="doc_text">
342 <p>The LLVM code representation is designed to be used in three different forms:
343 as an in-memory compiler IR, as an on-disk bitcode representation (suitable
344 for fast loading by a Just-In-Time compiler), and as a human readable
345 assembly language representation. This allows LLVM to provide a powerful
346 intermediate representation for efficient compiler transformations and
347 analysis, while providing a natural means to debug and visualize the
348 transformations. The three different forms of LLVM are all equivalent. This
349 document describes the human readable representation and notation.</p>
351 <p>The LLVM representation aims to be light-weight and low-level while being
352 expressive, typed, and extensible at the same time. It aims to be a
353 "universal IR" of sorts, by being at a low enough level that high-level ideas
354 may be cleanly mapped to it (similar to how microprocessors are "universal
355 IR's", allowing many source languages to be mapped to them). By providing
356 type information, LLVM can be used as the target of optimizations: for
357 example, through pointer analysis, it can be proven that a C automatic
358 variable is never accessed outside of the current function, allowing it to
359 be promoted to a simple SSA value instead of a memory location.</p>
363 <!-- _______________________________________________________________________ -->
364 <div class="doc_subsubsection"> <a name="wellformed">Well-Formedness</a> </div>
366 <div class="doc_text">
368 <p>It is important to note that this document describes 'well formed' LLVM
369 assembly language. There is a difference between what the parser accepts and
370 what is considered 'well formed'. For example, the following instruction is
371 syntactically okay, but not well formed:</p>
373 <div class="doc_code">
375 %x = <a href="#i_add">add</a> i32 1, %x
379 <p>because the definition of <tt>%x</tt> does not dominate all of its uses. The
380 LLVM infrastructure provides a verification pass that may be used to verify
381 that an LLVM module is well formed. This pass is automatically run by the
382 parser after parsing input assembly and by the optimizer before it outputs
383 bitcode. The violations pointed out by the verifier pass indicate bugs in
384 transformation passes or input to the parser.</p>
388 <!-- Describe the typesetting conventions here. -->
390 <!-- *********************************************************************** -->
391 <div class="doc_section"> <a name="identifiers">Identifiers</a> </div>
392 <!-- *********************************************************************** -->
394 <div class="doc_text">
396 <p>LLVM identifiers come in two basic types: global and local. Global
397 identifiers (functions, global variables) begin with the <tt>'@'</tt>
398 character. Local identifiers (register names, types) begin with
399 the <tt>'%'</tt> character. Additionally, there are three different formats
400 for identifiers, for different purposes:</p>
403 <li>Named values are represented as a string of characters with their prefix.
404 For example, <tt>%foo</tt>, <tt>@DivisionByZero</tt>,
405 <tt>%a.really.long.identifier</tt>. The actual regular expression used is
406 '<tt>[%@][a-zA-Z$._][a-zA-Z$._0-9]*</tt>'. Identifiers which require
407 other characters in their names can be surrounded with quotes. Special
408 characters may be escaped using <tt>"\xx"</tt> where <tt>xx</tt> is the
409 ASCII code for the character in hexadecimal. In this way, any character
410 can be used in a name value, even quotes themselves.</li>
412 <li>Unnamed values are represented as an unsigned numeric value with their
413 prefix. For example, <tt>%12</tt>, <tt>@2</tt>, <tt>%44</tt>.</li>
415 <li>Constants, which are described in a <a href="#constants">section about
416 constants</a>, below.</li>
419 <p>LLVM requires that values start with a prefix for two reasons: Compilers
420 don't need to worry about name clashes with reserved words, and the set of
421 reserved words may be expanded in the future without penalty. Additionally,
422 unnamed identifiers allow a compiler to quickly come up with a temporary
423 variable without having to avoid symbol table conflicts.</p>
425 <p>Reserved words in LLVM are very similar to reserved words in other
426 languages. There are keywords for different opcodes
427 ('<tt><a href="#i_add">add</a></tt>',
428 '<tt><a href="#i_bitcast">bitcast</a></tt>',
429 '<tt><a href="#i_ret">ret</a></tt>', etc...), for primitive type names
430 ('<tt><a href="#t_void">void</a></tt>',
431 '<tt><a href="#t_primitive">i32</a></tt>', etc...), and others. These
432 reserved words cannot conflict with variable names, because none of them
433 start with a prefix character (<tt>'%'</tt> or <tt>'@'</tt>).</p>
435 <p>Here is an example of LLVM code to multiply the integer variable
436 '<tt>%X</tt>' by 8:</p>
440 <div class="doc_code">
442 %result = <a href="#i_mul">mul</a> i32 %X, 8
446 <p>After strength reduction:</p>
448 <div class="doc_code">
450 %result = <a href="#i_shl">shl</a> i32 %X, i8 3
454 <p>And the hard way:</p>
456 <div class="doc_code">
458 %0 = <a href="#i_add">add</a> i32 %X, %X <i>; yields {i32}:%0</i>
459 %1 = <a href="#i_add">add</a> i32 %0, %0 <i>; yields {i32}:%1</i>
460 %result = <a href="#i_add">add</a> i32 %1, %1
464 <p>This last way of multiplying <tt>%X</tt> by 8 illustrates several important
465 lexical features of LLVM:</p>
468 <li>Comments are delimited with a '<tt>;</tt>' and go until the end of
471 <li>Unnamed temporaries are created when the result of a computation is not
472 assigned to a named value.</li>
474 <li>Unnamed temporaries are numbered sequentially</li>
477 <p>It also shows a convention that we follow in this document. When
478 demonstrating instructions, we will follow an instruction with a comment that
479 defines the type and name of value produced. Comments are shown in italic
484 <!-- *********************************************************************** -->
485 <div class="doc_section"> <a name="highlevel">High Level Structure</a> </div>
486 <!-- *********************************************************************** -->
488 <!-- ======================================================================= -->
489 <div class="doc_subsection"> <a name="modulestructure">Module Structure</a>
492 <div class="doc_text">
494 <p>LLVM programs are composed of "Module"s, each of which is a translation unit
495 of the input programs. Each module consists of functions, global variables,
496 and symbol table entries. Modules may be combined together with the LLVM
497 linker, which merges function (and global variable) definitions, resolves
498 forward declarations, and merges symbol table entries. Here is an example of
499 the "hello world" module:</p>
501 <div class="doc_code">
503 <i>; Declare the string constant as a global constant.</i>
504 <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>
506 <i>; External declaration of the puts function</i>
507 <a href="#functionstructure">declare</a> i32 @puts(i8*) <i>; i32 (i8*)* </i>
509 <i>; Definition of main function</i>
510 define i32 @main() { <i>; i32()* </i>
511 <i>; Convert [13 x i8]* to i8 *...</i>
512 %cast210 = <a href="#i_getelementptr">getelementptr</a> [13 x i8]* @.LC0, i64 0, i64 0 <i>; i8*</i>
514 <i>; Call puts function to write out the string to stdout.</i>
515 <a href="#i_call">call</a> i32 @puts(i8* %cast210) <i>; i32</i>
516 <a href="#i_ret">ret</a> i32 0<br>}
518 <i>; Named metadata</i>
519 !1 = metadata !{i32 41}
524 <p>This example is made up of a <a href="#globalvars">global variable</a> named
525 "<tt>.LC0</tt>", an external declaration of the "<tt>puts</tt>" function,
526 a <a href="#functionstructure">function definition</a> for
527 "<tt>main</tt>" and <a href="#namedmetadatastructure">named metadata</a>
530 <p>In general, a module is made up of a list of global values, where both
531 functions and global variables are global values. Global values are
532 represented by a pointer to a memory location (in this case, a pointer to an
533 array of char, and a pointer to a function), and have one of the
534 following <a href="#linkage">linkage types</a>.</p>
538 <!-- ======================================================================= -->
539 <div class="doc_subsection">
540 <a name="linkage">Linkage Types</a>
543 <div class="doc_text">
545 <p>All Global Variables and Functions have one of the following types of
549 <dt><tt><b><a name="linkage_private">private</a></b></tt></dt>
550 <dd>Global values with "<tt>private</tt>" linkage are only directly accessible
551 by objects in the current module. In particular, linking code into a
552 module with an private global value may cause the private to be renamed as
553 necessary to avoid collisions. Because the symbol is private to the
554 module, all references can be updated. This doesn't show up in any symbol
555 table in the object file.</dd>
557 <dt><tt><b><a name="linkage_linker_private">linker_private</a></b></tt></dt>
558 <dd>Similar to <tt>private</tt>, but the symbol is passed through the
559 assembler and removed by the linker after evaluation. Note that (unlike
560 <tt>private</tt> symbols) <tt>linker_private</tt> symbols are subject to
561 coalescing by the linker: weak symbols get merged and redefinitions are
562 rejected. However, unlike normal strong symbols, they are removed by the
563 linker from the final linked image (executable or dynamic library).
564 This is currently only used for Objective-C metadata.</dd>
566 <dt><tt><b><a name="linkage_linker_weak">linker_weak</a></b></tt></dt>
567 <dd>Global values with "<tt>linker_weak</tt>" linkage are given weak linkage,
568 but are removed by the linker after evaluation. Unlike normal weak
569 symbols, linker weak symbols are removed by the linker from the linal
570 linked image (executable or dynamic library). This is currently only used
571 for Objective-C metadata.</dd>
573 <dt><tt><b><a name="linkage_internal">internal</a></b></tt></dt>
574 <dd>Similar to <tt>private</tt>, but the value shows as a local symbol
575 (<tt>STB_LOCAL</tt> in the case of ELF) in the object file. This
576 corresponds to the notion of the '<tt>static</tt>' keyword in C.</dd>
578 <dt><tt><b><a name="linkage_available_externally">available_externally</a></b></tt></dt>
579 <dd>Globals with "<tt>available_externally</tt>" linkage are never emitted
580 into the object file corresponding to the LLVM module. They exist to
581 allow inlining and other optimizations to take place given knowledge of
582 the definition of the global, which is known to be somewhere outside the
583 module. Globals with <tt>available_externally</tt> linkage are allowed to
584 be discarded at will, and are otherwise the same as <tt>linkonce_odr</tt>.
585 This linkage type is only allowed on definitions, not declarations.</dd>
587 <dt><tt><b><a name="linkage_linkonce">linkonce</a></b></tt></dt>
588 <dd>Globals with "<tt>linkonce</tt>" linkage are merged with other globals of
589 the same name when linkage occurs. This can be used to implement
590 some forms of inline functions, templates, or other code which must be
591 generated in each translation unit that uses it, but where the body may
592 be overridden with a more definitive definition later. Unreferenced
593 <tt>linkonce</tt> globals are allowed to be discarded. Note that
594 <tt>linkonce</tt> linkage does not actually allow the optimizer to
595 inline the body of this function into callers because it doesn't know if
596 this definition of the function is the definitive definition within the
597 program or whether it will be overridden by a stronger definition.
598 To enable inlining and other optimizations, use "<tt>linkonce_odr</tt>"
601 <dt><tt><b><a name="linkage_weak">weak</a></b></tt></dt>
602 <dd>"<tt>weak</tt>" linkage has the same merging semantics as
603 <tt>linkonce</tt> linkage, except that unreferenced globals with
604 <tt>weak</tt> linkage may not be discarded. This is used for globals that
605 are declared "weak" in C source code.</dd>
607 <dt><tt><b><a name="linkage_common">common</a></b></tt></dt>
608 <dd>"<tt>common</tt>" linkage is most similar to "<tt>weak</tt>" linkage, but
609 they are used for tentative definitions in C, such as "<tt>int X;</tt>" at
611 Symbols with "<tt>common</tt>" linkage are merged in the same way as
612 <tt>weak symbols</tt>, and they may not be deleted if unreferenced.
613 <tt>common</tt> symbols may not have an explicit section,
614 must have a zero initializer, and may not be marked '<a
615 href="#globalvars"><tt>constant</tt></a>'. Functions and aliases may not
616 have common linkage.</dd>
619 <dt><tt><b><a name="linkage_appending">appending</a></b></tt></dt>
620 <dd>"<tt>appending</tt>" linkage may only be applied to global variables of
621 pointer to array type. When two global variables with appending linkage
622 are linked together, the two global arrays are appended together. This is
623 the LLVM, typesafe, equivalent of having the system linker append together
624 "sections" with identical names when .o files are linked.</dd>
626 <dt><tt><b><a name="linkage_externweak">extern_weak</a></b></tt></dt>
627 <dd>The semantics of this linkage follow the ELF object file model: the symbol
628 is weak until linked, if not linked, the symbol becomes null instead of
629 being an undefined reference.</dd>
631 <dt><tt><b><a name="linkage_linkonce_odr">linkonce_odr</a></b></tt></dt>
632 <dt><tt><b><a name="linkage_weak_odr">weak_odr</a></b></tt></dt>
633 <dd>Some languages allow differing globals to be merged, such as two functions
634 with different semantics. Other languages, such as <tt>C++</tt>, ensure
635 that only equivalent globals are ever merged (the "one definition rule" -
636 "ODR"). Such languages can use the <tt>linkonce_odr</tt>
637 and <tt>weak_odr</tt> linkage types to indicate that the global will only
638 be merged with equivalent globals. These linkage types are otherwise the
639 same as their non-<tt>odr</tt> versions.</dd>
641 <dt><tt><b><a name="linkage_external">externally visible</a></b></tt>:</dt>
642 <dd>If none of the above identifiers are used, the global is externally
643 visible, meaning that it participates in linkage and can be used to
644 resolve external symbol references.</dd>
647 <p>The next two types of linkage are targeted for Microsoft Windows platform
648 only. They are designed to support importing (exporting) symbols from (to)
649 DLLs (Dynamic Link Libraries).</p>
652 <dt><tt><b><a name="linkage_dllimport">dllimport</a></b></tt></dt>
653 <dd>"<tt>dllimport</tt>" linkage causes the compiler to reference a function
654 or variable via a global pointer to a pointer that is set up by the DLL
655 exporting the symbol. On Microsoft Windows targets, the pointer name is
656 formed by combining <code>__imp_</code> and the function or variable
659 <dt><tt><b><a name="linkage_dllexport">dllexport</a></b></tt></dt>
660 <dd>"<tt>dllexport</tt>" linkage causes the compiler to provide a global
661 pointer to a pointer in a DLL, so that it can be referenced with the
662 <tt>dllimport</tt> attribute. On Microsoft Windows targets, the pointer
663 name is formed by combining <code>__imp_</code> and the function or
667 <p>For example, since the "<tt>.LC0</tt>" variable is defined to be internal, if
668 another module defined a "<tt>.LC0</tt>" variable and was linked with this
669 one, one of the two would be renamed, preventing a collision. Since
670 "<tt>main</tt>" and "<tt>puts</tt>" are external (i.e., lacking any linkage
671 declarations), they are accessible outside of the current module.</p>
673 <p>It is illegal for a function <i>declaration</i> to have any linkage type
674 other than "externally visible", <tt>dllimport</tt>
675 or <tt>extern_weak</tt>.</p>
677 <p>Aliases can have only <tt>external</tt>, <tt>internal</tt>, <tt>weak</tt>
678 or <tt>weak_odr</tt> linkages.</p>
682 <!-- ======================================================================= -->
683 <div class="doc_subsection">
684 <a name="callingconv">Calling Conventions</a>
687 <div class="doc_text">
689 <p>LLVM <a href="#functionstructure">functions</a>, <a href="#i_call">calls</a>
690 and <a href="#i_invoke">invokes</a> can all have an optional calling
691 convention specified for the call. The calling convention of any pair of
692 dynamic caller/callee must match, or the behavior of the program is
693 undefined. The following calling conventions are supported by LLVM, and more
694 may be added in the future:</p>
697 <dt><b>"<tt>ccc</tt>" - The C calling convention</b>:</dt>
698 <dd>This calling convention (the default if no other calling convention is
699 specified) matches the target C calling conventions. This calling
700 convention supports varargs function calls and tolerates some mismatch in
701 the declared prototype and implemented declaration of the function (as
704 <dt><b>"<tt>fastcc</tt>" - The fast calling convention</b>:</dt>
705 <dd>This calling convention attempts to make calls as fast as possible
706 (e.g. by passing things in registers). This calling convention allows the
707 target to use whatever tricks it wants to produce fast code for the
708 target, without having to conform to an externally specified ABI
709 (Application Binary Interface).
710 <a href="CodeGenerator.html#tailcallopt">Tail calls can only be optimized
711 when this or the GHC convention is used.</a> This calling convention
712 does not support varargs and requires the prototype of all callees to
713 exactly match the prototype of the function definition.</dd>
715 <dt><b>"<tt>coldcc</tt>" - The cold calling convention</b>:</dt>
716 <dd>This calling convention attempts to make code in the caller as efficient
717 as possible under the assumption that the call is not commonly executed.
718 As such, these calls often preserve all registers so that the call does
719 not break any live ranges in the caller side. This calling convention
720 does not support varargs and requires the prototype of all callees to
721 exactly match the prototype of the function definition.</dd>
723 <dt><b>"<tt>cc <em>10</em></tt>" - GHC convention</b>:</dt>
724 <dd>This calling convention has been implemented specifically for use by the
725 <a href="http://www.haskell.org/ghc">Glasgow Haskell Compiler (GHC)</a>.
726 It passes everything in registers, going to extremes to achieve this by
727 disabling callee save registers. This calling convention should not be
728 used lightly but only for specific situations such as an alternative to
729 the <em>register pinning</em> performance technique often used when
730 implementing functional programming languages.At the moment only X86
731 supports this convention and it has the following limitations:
733 <li>On <em>X86-32</em> only supports up to 4 bit type parameters. No
734 floating point types are supported.</li>
735 <li>On <em>X86-64</em> only supports up to 10 bit type parameters and
736 6 floating point parameters.</li>
738 This calling convention supports
739 <a href="CodeGenerator.html#tailcallopt">tail call optimization</a> but
740 requires both the caller and callee are using it.
743 <dt><b>"<tt>cc <<em>n</em>></tt>" - Numbered convention</b>:</dt>
744 <dd>Any calling convention may be specified by number, allowing
745 target-specific calling conventions to be used. Target specific calling
746 conventions start at 64.</dd>
749 <p>More calling conventions can be added/defined on an as-needed basis, to
750 support Pascal conventions or any other well-known target-independent
755 <!-- ======================================================================= -->
756 <div class="doc_subsection">
757 <a name="visibility">Visibility Styles</a>
760 <div class="doc_text">
762 <p>All Global Variables and Functions have one of the following visibility
766 <dt><b>"<tt>default</tt>" - Default style</b>:</dt>
767 <dd>On targets that use the ELF object file format, default visibility means
768 that the declaration is visible to other modules and, in shared libraries,
769 means that the declared entity may be overridden. On Darwin, default
770 visibility means that the declaration is visible to other modules. Default
771 visibility corresponds to "external linkage" in the language.</dd>
773 <dt><b>"<tt>hidden</tt>" - Hidden style</b>:</dt>
774 <dd>Two declarations of an object with hidden visibility refer to the same
775 object if they are in the same shared object. Usually, hidden visibility
776 indicates that the symbol will not be placed into the dynamic symbol
777 table, so no other module (executable or shared library) can reference it
780 <dt><b>"<tt>protected</tt>" - Protected style</b>:</dt>
781 <dd>On ELF, protected visibility indicates that the symbol will be placed in
782 the dynamic symbol table, but that references within the defining module
783 will bind to the local symbol. That is, the symbol cannot be overridden by
789 <!-- ======================================================================= -->
790 <div class="doc_subsection">
791 <a name="namedtypes">Named Types</a>
794 <div class="doc_text">
796 <p>LLVM IR allows you to specify name aliases for certain types. This can make
797 it easier to read the IR and make the IR more condensed (particularly when
798 recursive types are involved). An example of a name specification is:</p>
800 <div class="doc_code">
802 %mytype = type { %mytype*, i32 }
806 <p>You may give a name to any <a href="#typesystem">type</a> except
807 "<a href="t_void">void</a>". Type name aliases may be used anywhere a type
808 is expected with the syntax "%mytype".</p>
810 <p>Note that type names are aliases for the structural type that they indicate,
811 and that you can therefore specify multiple names for the same type. This
812 often leads to confusing behavior when dumping out a .ll file. Since LLVM IR
813 uses structural typing, the name is not part of the type. When printing out
814 LLVM IR, the printer will pick <em>one name</em> to render all types of a
815 particular shape. This means that if you have code where two different
816 source types end up having the same LLVM type, that the dumper will sometimes
817 print the "wrong" or unexpected type. This is an important design point and
818 isn't going to change.</p>
822 <!-- ======================================================================= -->
823 <div class="doc_subsection">
824 <a name="globalvars">Global Variables</a>
827 <div class="doc_text">
829 <p>Global variables define regions of memory allocated at compilation time
830 instead of run-time. Global variables may optionally be initialized, may
831 have an explicit section to be placed in, and may have an optional explicit
832 alignment specified. A variable may be defined as "thread_local", which
833 means that it will not be shared by threads (each thread will have a
834 separated copy of the variable). A variable may be defined as a global
835 "constant," which indicates that the contents of the variable
836 will <b>never</b> be modified (enabling better optimization, allowing the
837 global data to be placed in the read-only section of an executable, etc).
838 Note that variables that need runtime initialization cannot be marked
839 "constant" as there is a store to the variable.</p>
841 <p>LLVM explicitly allows <em>declarations</em> of global variables to be marked
842 constant, even if the final definition of the global is not. This capability
843 can be used to enable slightly better optimization of the program, but
844 requires the language definition to guarantee that optimizations based on the
845 'constantness' are valid for the translation units that do not include the
848 <p>As SSA values, global variables define pointer values that are in scope
849 (i.e. they dominate) all basic blocks in the program. Global variables
850 always define a pointer to their "content" type because they describe a
851 region of memory, and all memory objects in LLVM are accessed through
854 <p>A global variable may be declared to reside in a target-specific numbered
855 address space. For targets that support them, address spaces may affect how
856 optimizations are performed and/or what target instructions are used to
857 access the variable. The default address space is zero. The address space
858 qualifier must precede any other attributes.</p>
860 <p>LLVM allows an explicit section to be specified for globals. If the target
861 supports it, it will emit globals to the section specified.</p>
863 <p>An explicit alignment may be specified for a global, which must be a power
864 of 2. If not present, or if the alignment is set to zero, the alignment of
865 the global is set by the target to whatever it feels convenient. If an
866 explicit alignment is specified, the global is forced to have exactly that
867 alignment. Targets and optimizers are not allowed to over-align the global
868 if the global has an assigned section. In this case, the extra alignment
869 could be observable: for example, code could assume that the globals are
870 densely packed in their section and try to iterate over them as an array,
871 alignment padding would break this iteration.</p>
873 <p>For example, the following defines a global in a numbered address space with
874 an initializer, section, and alignment:</p>
876 <div class="doc_code">
878 @G = addrspace(5) constant float 1.0, section "foo", align 4
885 <!-- ======================================================================= -->
886 <div class="doc_subsection">
887 <a name="functionstructure">Functions</a>
890 <div class="doc_text">
892 <p>LLVM function definitions consist of the "<tt>define</tt>" keyword, an
893 optional <a href="#linkage">linkage type</a>, an optional
894 <a href="#visibility">visibility style</a>, an optional
895 <a href="#callingconv">calling convention</a>, a return type, an optional
896 <a href="#paramattrs">parameter attribute</a> for the return type, a function
897 name, a (possibly empty) argument list (each with optional
898 <a href="#paramattrs">parameter attributes</a>), optional
899 <a href="#fnattrs">function attributes</a>, an optional section, an optional
900 alignment, an optional <a href="#gc">garbage collector name</a>, an opening
901 curly brace, a list of basic blocks, and a closing curly brace.</p>
903 <p>LLVM function declarations consist of the "<tt>declare</tt>" keyword, an
904 optional <a href="#linkage">linkage type</a>, an optional
905 <a href="#visibility">visibility style</a>, an optional
906 <a href="#callingconv">calling convention</a>, a return type, an optional
907 <a href="#paramattrs">parameter attribute</a> for the return type, a function
908 name, a possibly empty list of arguments, an optional alignment, and an
909 optional <a href="#gc">garbage collector name</a>.</p>
911 <p>A function definition contains a list of basic blocks, forming the CFG
912 (Control Flow Graph) for the function. Each basic block may optionally start
913 with a label (giving the basic block a symbol table entry), contains a list
914 of instructions, and ends with a <a href="#terminators">terminator</a>
915 instruction (such as a branch or function return).</p>
917 <p>The first basic block in a function is special in two ways: it is immediately
918 executed on entrance to the function, and it is not allowed to have
919 predecessor basic blocks (i.e. there can not be any branches to the entry
920 block of a function). Because the block can have no predecessors, it also
921 cannot have any <a href="#i_phi">PHI nodes</a>.</p>
923 <p>LLVM allows an explicit section to be specified for functions. If the target
924 supports it, it will emit functions to the section specified.</p>
926 <p>An explicit alignment may be specified for a function. If not present, or if
927 the alignment is set to zero, the alignment of the function is set by the
928 target to whatever it feels convenient. If an explicit alignment is
929 specified, the function is forced to have at least that much alignment. All
930 alignments must be a power of 2.</p>
933 <div class="doc_code">
935 define [<a href="#linkage">linkage</a>] [<a href="#visibility">visibility</a>]
936 [<a href="#callingconv">cconv</a>] [<a href="#paramattrs">ret attrs</a>]
937 <ResultType> @<FunctionName> ([argument list])
938 [<a href="#fnattrs">fn Attrs</a>] [section "name"] [align N]
939 [<a href="#gc">gc</a>] { ... }
945 <!-- ======================================================================= -->
946 <div class="doc_subsection">
947 <a name="aliasstructure">Aliases</a>
950 <div class="doc_text">
952 <p>Aliases act as "second name" for the aliasee value (which can be either
953 function, global variable, another alias or bitcast of global value). Aliases
954 may have an optional <a href="#linkage">linkage type</a>, and an
955 optional <a href="#visibility">visibility style</a>.</p>
958 <div class="doc_code">
960 @<Name> = alias [Linkage] [Visibility] <AliaseeTy> @<Aliasee>
966 <!-- ======================================================================= -->
967 <div class="doc_subsection">
968 <a name="namedmetadatastructure">Named Metadata</a>
971 <div class="doc_text">
973 <p>Named metadata is a collection of metadata. <a href="#metadata">Metadata
974 nodes</a> (but not metadata strings) and null are the only valid operands for
975 a named metadata.</p>
978 <div class="doc_code">
980 !1 = metadata !{metadata !"one"}
987 <!-- ======================================================================= -->
988 <div class="doc_subsection"><a name="paramattrs">Parameter Attributes</a></div>
990 <div class="doc_text">
992 <p>The return type and each parameter of a function type may have a set of
993 <i>parameter attributes</i> associated with them. Parameter attributes are
994 used to communicate additional information about the result or parameters of
995 a function. Parameter attributes are considered to be part of the function,
996 not of the function type, so functions with different parameter attributes
997 can have the same function type.</p>
999 <p>Parameter attributes are simple keywords that follow the type specified. If
1000 multiple parameter attributes are needed, they are space separated. For
1003 <div class="doc_code">
1005 declare i32 @printf(i8* noalias nocapture, ...)
1006 declare i32 @atoi(i8 zeroext)
1007 declare signext i8 @returns_signed_char()
1011 <p>Note that any attributes for the function result (<tt>nounwind</tt>,
1012 <tt>readonly</tt>) come immediately after the argument list.</p>
1014 <p>Currently, only the following parameter attributes are defined:</p>
1017 <dt><tt><b>zeroext</b></tt></dt>
1018 <dd>This indicates to the code generator that the parameter or return value
1019 should be zero-extended to a 32-bit value by the caller (for a parameter)
1020 or the callee (for a return value).</dd>
1022 <dt><tt><b>signext</b></tt></dt>
1023 <dd>This indicates to the code generator that the parameter or return value
1024 should be sign-extended to a 32-bit value by the caller (for a parameter)
1025 or the callee (for a return value).</dd>
1027 <dt><tt><b>inreg</b></tt></dt>
1028 <dd>This indicates that this parameter or return value should be treated in a
1029 special target-dependent fashion during while emitting code for a function
1030 call or return (usually, by putting it in a register as opposed to memory,
1031 though some targets use it to distinguish between two different kinds of
1032 registers). Use of this attribute is target-specific.</dd>
1034 <dt><tt><b><a name="byval">byval</a></b></tt></dt>
1035 <dd>This indicates that the pointer parameter should really be passed by value
1036 to the function. The attribute implies that a hidden copy of the pointee
1037 is made between the caller and the callee, so the callee is unable to
1038 modify the value in the callee. This attribute is only valid on LLVM
1039 pointer arguments. It is generally used to pass structs and arrays by
1040 value, but is also valid on pointers to scalars. The copy is considered
1041 to belong to the caller not the callee (for example,
1042 <tt><a href="#readonly">readonly</a></tt> functions should not write to
1043 <tt>byval</tt> parameters). This is not a valid attribute for return
1044 values. The byval attribute also supports specifying an alignment with
1045 the align attribute. This has a target-specific effect on the code
1046 generator that usually indicates a desired alignment for the synthesized
1049 <dt><tt><b>sret</b></tt></dt>
1050 <dd>This indicates that the pointer parameter specifies the address of a
1051 structure that is the return value of the function in the source program.
1052 This pointer must be guaranteed by the caller to be valid: loads and
1053 stores to the structure may be assumed by the callee to not to trap. This
1054 may only be applied to the first parameter. This is not a valid attribute
1055 for return values. </dd>
1057 <dt><tt><b>noalias</b></tt></dt>
1058 <dd>This indicates that the pointer does not alias any global or any other
1059 parameter. The caller is responsible for ensuring that this is the
1060 case. On a function return value, <tt>noalias</tt> additionally indicates
1061 that the pointer does not alias any other pointers visible to the
1062 caller. For further details, please see the discussion of the NoAlias
1064 <a href="http://llvm.org/docs/AliasAnalysis.html#MustMayNo">alias
1067 <dt><tt><b>nocapture</b></tt></dt>
1068 <dd>This indicates that the callee does not make any copies of the pointer
1069 that outlive the callee itself. This is not a valid attribute for return
1072 <dt><tt><b>nest</b></tt></dt>
1073 <dd>This indicates that the pointer parameter can be excised using the
1074 <a href="#int_trampoline">trampoline intrinsics</a>. This is not a valid
1075 attribute for return values.</dd>
1080 <!-- ======================================================================= -->
1081 <div class="doc_subsection">
1082 <a name="gc">Garbage Collector Names</a>
1085 <div class="doc_text">
1087 <p>Each function may specify a garbage collector name, which is simply a
1090 <div class="doc_code">
1092 define void @f() gc "name" { ... }
1096 <p>The compiler declares the supported values of <i>name</i>. Specifying a
1097 collector which will cause the compiler to alter its output in order to
1098 support the named garbage collection algorithm.</p>
1102 <!-- ======================================================================= -->
1103 <div class="doc_subsection">
1104 <a name="fnattrs">Function Attributes</a>
1107 <div class="doc_text">
1109 <p>Function attributes are set to communicate additional information about a
1110 function. Function attributes are considered to be part of the function, not
1111 of the function type, so functions with different parameter attributes can
1112 have the same function type.</p>
1114 <p>Function attributes are simple keywords that follow the type specified. If
1115 multiple attributes are needed, they are space separated. For example:</p>
1117 <div class="doc_code">
1119 define void @f() noinline { ... }
1120 define void @f() alwaysinline { ... }
1121 define void @f() alwaysinline optsize { ... }
1122 define void @f() optsize { ... }
1127 <dt><tt><b>alignstack(<<em>n</em>>)</b></tt></dt>
1128 <dd>This attribute indicates that, when emitting the prologue and epilogue,
1129 the backend should forcibly align the stack pointer. Specify the
1130 desired alignment, which must be a power of two, in parentheses.
1132 <dt><tt><b>alwaysinline</b></tt></dt>
1133 <dd>This attribute indicates that the inliner should attempt to inline this
1134 function into callers whenever possible, ignoring any active inlining size
1135 threshold for this caller.</dd>
1137 <dt><tt><b>inlinehint</b></tt></dt>
1138 <dd>This attribute indicates that the source code contained a hint that inlining
1139 this function is desirable (such as the "inline" keyword in C/C++). It
1140 is just a hint; it imposes no requirements on the inliner.</dd>
1142 <dt><tt><b>noinline</b></tt></dt>
1143 <dd>This attribute indicates that the inliner should never inline this
1144 function in any situation. This attribute may not be used together with
1145 the <tt>alwaysinline</tt> attribute.</dd>
1147 <dt><tt><b>optsize</b></tt></dt>
1148 <dd>This attribute suggests that optimization passes and code generator passes
1149 make choices that keep the code size of this function low, and otherwise
1150 do optimizations specifically to reduce code size.</dd>
1152 <dt><tt><b>noreturn</b></tt></dt>
1153 <dd>This function attribute indicates that the function never returns
1154 normally. This produces undefined behavior at runtime if the function
1155 ever does dynamically return.</dd>
1157 <dt><tt><b>nounwind</b></tt></dt>
1158 <dd>This function attribute indicates that the function never returns with an
1159 unwind or exceptional control flow. If the function does unwind, its
1160 runtime behavior is undefined.</dd>
1162 <dt><tt><b>readnone</b></tt></dt>
1163 <dd>This attribute indicates that the function computes its result (or decides
1164 to unwind an exception) based strictly on its arguments, without
1165 dereferencing any pointer arguments or otherwise accessing any mutable
1166 state (e.g. memory, control registers, etc) visible to caller functions.
1167 It does not write through any pointer arguments
1168 (including <tt><a href="#byval">byval</a></tt> arguments) and never
1169 changes any state visible to callers. This means that it cannot unwind
1170 exceptions by calling the <tt>C++</tt> exception throwing methods, but
1171 could use the <tt>unwind</tt> instruction.</dd>
1173 <dt><tt><b><a name="readonly">readonly</a></b></tt></dt>
1174 <dd>This attribute indicates that the function does not write through any
1175 pointer arguments (including <tt><a href="#byval">byval</a></tt>
1176 arguments) or otherwise modify any state (e.g. memory, control registers,
1177 etc) visible to caller functions. It may dereference pointer arguments
1178 and read state that may be set in the caller. A readonly function always
1179 returns the same value (or unwinds an exception identically) when called
1180 with the same set of arguments and global state. It cannot unwind an
1181 exception by calling the <tt>C++</tt> exception throwing methods, but may
1182 use the <tt>unwind</tt> instruction.</dd>
1184 <dt><tt><b><a name="ssp">ssp</a></b></tt></dt>
1185 <dd>This attribute indicates that the function should emit a stack smashing
1186 protector. It is in the form of a "canary"—a random value placed on
1187 the stack before the local variables that's checked upon return from the
1188 function to see if it has been overwritten. A heuristic is used to
1189 determine if a function needs stack protectors or not.<br>
1191 If a function that has an <tt>ssp</tt> attribute is inlined into a
1192 function that doesn't have an <tt>ssp</tt> attribute, then the resulting
1193 function will have an <tt>ssp</tt> attribute.</dd>
1195 <dt><tt><b>sspreq</b></tt></dt>
1196 <dd>This attribute indicates that the function should <em>always</em> emit a
1197 stack smashing protector. This overrides
1198 the <tt><a href="#ssp">ssp</a></tt> function attribute.<br>
1200 If a function that has an <tt>sspreq</tt> attribute is inlined into a
1201 function that doesn't have an <tt>sspreq</tt> attribute or which has
1202 an <tt>ssp</tt> attribute, then the resulting function will have
1203 an <tt>sspreq</tt> attribute.</dd>
1205 <dt><tt><b>noredzone</b></tt></dt>
1206 <dd>This attribute indicates that the code generator should not use a red
1207 zone, even if the target-specific ABI normally permits it.</dd>
1209 <dt><tt><b>noimplicitfloat</b></tt></dt>
1210 <dd>This attributes disables implicit floating point instructions.</dd>
1212 <dt><tt><b>naked</b></tt></dt>
1213 <dd>This attribute disables prologue / epilogue emission for the function.
1214 This can have very system-specific consequences.</dd>
1219 <!-- ======================================================================= -->
1220 <div class="doc_subsection">
1221 <a name="moduleasm">Module-Level Inline Assembly</a>
1224 <div class="doc_text">
1226 <p>Modules may contain "module-level inline asm" blocks, which corresponds to
1227 the GCC "file scope inline asm" blocks. These blocks are internally
1228 concatenated by LLVM and treated as a single unit, but may be separated in
1229 the <tt>.ll</tt> file if desired. The syntax is very simple:</p>
1231 <div class="doc_code">
1233 module asm "inline asm code goes here"
1234 module asm "more can go here"
1238 <p>The strings can contain any character by escaping non-printable characters.
1239 The escape sequence used is simply "\xx" where "xx" is the two digit hex code
1242 <p>The inline asm code is simply printed to the machine code .s file when
1243 assembly code is generated.</p>
1247 <!-- ======================================================================= -->
1248 <div class="doc_subsection">
1249 <a name="datalayout">Data Layout</a>
1252 <div class="doc_text">
1254 <p>A module may specify a target specific data layout string that specifies how
1255 data is to be laid out in memory. The syntax for the data layout is
1258 <div class="doc_code">
1260 target datalayout = "<i>layout specification</i>"
1264 <p>The <i>layout specification</i> consists of a list of specifications
1265 separated by the minus sign character ('-'). Each specification starts with
1266 a letter and may include other information after the letter to define some
1267 aspect of the data layout. The specifications accepted are as follows:</p>
1271 <dd>Specifies that the target lays out data in big-endian form. That is, the
1272 bits with the most significance have the lowest address location.</dd>
1275 <dd>Specifies that the target lays out data in little-endian form. That is,
1276 the bits with the least significance have the lowest address
1279 <dt><tt>p:<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1280 <dd>This specifies the <i>size</i> of a pointer and its <i>abi</i> and
1281 <i>preferred</i> alignments. All sizes are in bits. Specifying
1282 the <i>pref</i> alignment is optional. If omitted, the
1283 preceding <tt>:</tt> should be omitted too.</dd>
1285 <dt><tt>i<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1286 <dd>This specifies the alignment for an integer type of a given bit
1287 <i>size</i>. The value of <i>size</i> must be in the range [1,2^23).</dd>
1289 <dt><tt>v<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1290 <dd>This specifies the alignment for a vector type of a given bit
1293 <dt><tt>f<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1294 <dd>This specifies the alignment for a floating point type of a given bit
1295 <i>size</i>. Only values of <i>size</i> that are supported by the target
1296 will work. 32 (float) and 64 (double) are supported on all targets;
1297 80 or 128 (different flavors of long double) are also supported on some
1300 <dt><tt>a<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1301 <dd>This specifies the alignment for an aggregate type of a given bit
1304 <dt><tt>s<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1305 <dd>This specifies the alignment for a stack object of a given bit
1308 <dt><tt>n<i>size1</i>:<i>size2</i>:<i>size3</i>...</tt></dt>
1309 <dd>This specifies a set of native integer widths for the target CPU
1310 in bits. For example, it might contain "n32" for 32-bit PowerPC,
1311 "n32:64" for PowerPC 64, or "n8:16:32:64" for X86-64. Elements of
1312 this set are considered to support most general arithmetic
1313 operations efficiently.</dd>
1316 <p>When constructing the data layout for a given target, LLVM starts with a
1317 default set of specifications which are then (possibly) overridden by the
1318 specifications in the <tt>datalayout</tt> keyword. The default specifications
1319 are given in this list:</p>
1322 <li><tt>E</tt> - big endian</li>
1323 <li><tt>p:64:64:64</tt> - 64-bit pointers with 64-bit alignment</li>
1324 <li><tt>i1:8:8</tt> - i1 is 8-bit (byte) aligned</li>
1325 <li><tt>i8:8:8</tt> - i8 is 8-bit (byte) aligned</li>
1326 <li><tt>i16:16:16</tt> - i16 is 16-bit aligned</li>
1327 <li><tt>i32:32:32</tt> - i32 is 32-bit aligned</li>
1328 <li><tt>i64:32:64</tt> - i64 has ABI alignment of 32-bits but preferred
1329 alignment of 64-bits</li>
1330 <li><tt>f32:32:32</tt> - float is 32-bit aligned</li>
1331 <li><tt>f64:64:64</tt> - double is 64-bit aligned</li>
1332 <li><tt>v64:64:64</tt> - 64-bit vector is 64-bit aligned</li>
1333 <li><tt>v128:128:128</tt> - 128-bit vector is 128-bit aligned</li>
1334 <li><tt>a0:0:1</tt> - aggregates are 8-bit aligned</li>
1335 <li><tt>s0:64:64</tt> - stack objects are 64-bit aligned</li>
1338 <p>When LLVM is determining the alignment for a given type, it uses the
1339 following rules:</p>
1342 <li>If the type sought is an exact match for one of the specifications, that
1343 specification is used.</li>
1345 <li>If no match is found, and the type sought is an integer type, then the
1346 smallest integer type that is larger than the bitwidth of the sought type
1347 is used. If none of the specifications are larger than the bitwidth then
1348 the the largest integer type is used. For example, given the default
1349 specifications above, the i7 type will use the alignment of i8 (next
1350 largest) while both i65 and i256 will use the alignment of i64 (largest
1353 <li>If no match is found, and the type sought is a vector type, then the
1354 largest vector type that is smaller than the sought vector type will be
1355 used as a fall back. This happens because <128 x double> can be
1356 implemented in terms of 64 <2 x double>, for example.</li>
1361 <!-- ======================================================================= -->
1362 <div class="doc_subsection">
1363 <a name="pointeraliasing">Pointer Aliasing Rules</a>
1366 <div class="doc_text">
1368 <p>Any memory access must be done through a pointer value associated
1369 with an address range of the memory access, otherwise the behavior
1370 is undefined. Pointer values are associated with address ranges
1371 according to the following rules:</p>
1374 <li>A pointer value formed from a
1375 <tt><a href="#i_getelementptr">getelementptr</a></tt> instruction
1376 is associated with the addresses associated with the first operand
1377 of the <tt>getelementptr</tt>.</li>
1378 <li>An address of a global variable is associated with the address
1379 range of the variable's storage.</li>
1380 <li>The result value of an allocation instruction is associated with
1381 the address range of the allocated storage.</li>
1382 <li>A null pointer in the default address-space is associated with
1384 <li>A pointer value formed by an
1385 <tt><a href="#i_inttoptr">inttoptr</a></tt> is associated with all
1386 address ranges of all pointer values that contribute (directly or
1387 indirectly) to the computation of the pointer's value.</li>
1388 <li>The result value of a
1389 <tt><a href="#i_bitcast">bitcast</a></tt> is associated with all
1390 addresses associated with the operand of the <tt>bitcast</tt>.</li>
1391 <li>An integer constant other than zero or a pointer value returned
1392 from a function not defined within LLVM may be associated with address
1393 ranges allocated through mechanisms other than those provided by
1394 LLVM. Such ranges shall not overlap with any ranges of addresses
1395 allocated by mechanisms provided by LLVM.</li>
1398 <p>LLVM IR does not associate types with memory. The result type of a
1399 <tt><a href="#i_load">load</a></tt> merely indicates the size and
1400 alignment of the memory from which to load, as well as the
1401 interpretation of the value. The first operand type of a
1402 <tt><a href="#i_store">store</a></tt> similarly only indicates the size
1403 and alignment of the store.</p>
1405 <p>Consequently, type-based alias analysis, aka TBAA, aka
1406 <tt>-fstrict-aliasing</tt>, is not applicable to general unadorned
1407 LLVM IR. <a href="#metadata">Metadata</a> may be used to encode
1408 additional information which specialized optimization passes may use
1409 to implement type-based alias analysis.</p>
1413 <!-- ======================================================================= -->
1414 <div class="doc_subsection">
1415 <a name="volatile">Volatile Memory Accesses</a>
1418 <div class="doc_text">
1420 <p>Certain memory accesses, such as <a href="#i_load"><tt>load</tt></a>s, <a
1421 href="#i_store"><tt>store</tt></a>s, and <a
1422 href="#int_memcpy"><tt>llvm.memcpy</tt></a>s may be marked <tt>volatile</tt>.
1423 The optimizers must not change the number of volatile operations or change their
1424 order of execution relative to other volatile operations. The optimizers
1425 <i>may</i> change the order of volatile operations relative to non-volatile
1426 operations. This is not Java's "volatile" and has no cross-thread
1427 synchronization behavior.</p>
1431 <!-- *********************************************************************** -->
1432 <div class="doc_section"> <a name="typesystem">Type System</a> </div>
1433 <!-- *********************************************************************** -->
1435 <div class="doc_text">
1437 <p>The LLVM type system is one of the most important features of the
1438 intermediate representation. Being typed enables a number of optimizations
1439 to be performed on the intermediate representation directly, without having
1440 to do extra analyses on the side before the transformation. A strong type
1441 system makes it easier to read the generated code and enables novel analyses
1442 and transformations that are not feasible to perform on normal three address
1443 code representations.</p>
1447 <!-- ======================================================================= -->
1448 <div class="doc_subsection"> <a name="t_classifications">Type
1449 Classifications</a> </div>
1451 <div class="doc_text">
1453 <p>The types fall into a few useful classifications:</p>
1455 <table border="1" cellspacing="0" cellpadding="4">
1457 <tr><th>Classification</th><th>Types</th></tr>
1459 <td><a href="#t_integer">integer</a></td>
1460 <td><tt>i1, i2, i3, ... i8, ... i16, ... i32, ... i64, ... </tt></td>
1463 <td><a href="#t_floating">floating point</a></td>
1464 <td><tt>float, double, x86_fp80, fp128, ppc_fp128</tt></td>
1467 <td><a name="t_firstclass">first class</a></td>
1468 <td><a href="#t_integer">integer</a>,
1469 <a href="#t_floating">floating point</a>,
1470 <a href="#t_pointer">pointer</a>,
1471 <a href="#t_vector">vector</a>,
1472 <a href="#t_struct">structure</a>,
1473 <a href="#t_union">union</a>,
1474 <a href="#t_array">array</a>,
1475 <a href="#t_label">label</a>,
1476 <a href="#t_metadata">metadata</a>.
1480 <td><a href="#t_primitive">primitive</a></td>
1481 <td><a href="#t_label">label</a>,
1482 <a href="#t_void">void</a>,
1483 <a href="#t_floating">floating point</a>,
1484 <a href="#t_metadata">metadata</a>.</td>
1487 <td><a href="#t_derived">derived</a></td>
1488 <td><a href="#t_array">array</a>,
1489 <a href="#t_function">function</a>,
1490 <a href="#t_pointer">pointer</a>,
1491 <a href="#t_struct">structure</a>,
1492 <a href="#t_pstruct">packed structure</a>,
1493 <a href="#t_union">union</a>,
1494 <a href="#t_vector">vector</a>,
1495 <a href="#t_opaque">opaque</a>.
1501 <p>The <a href="#t_firstclass">first class</a> types are perhaps the most
1502 important. Values of these types are the only ones which can be produced by
1507 <!-- ======================================================================= -->
1508 <div class="doc_subsection"> <a name="t_primitive">Primitive Types</a> </div>
1510 <div class="doc_text">
1512 <p>The primitive types are the fundamental building blocks of the LLVM
1517 <!-- _______________________________________________________________________ -->
1518 <div class="doc_subsubsection"> <a name="t_integer">Integer Type</a> </div>
1520 <div class="doc_text">
1523 <p>The integer type is a very simple type that simply specifies an arbitrary
1524 bit width for the integer type desired. Any bit width from 1 bit to
1525 2<sup>23</sup>-1 (about 8 million) can be specified.</p>
1532 <p>The number of bits the integer will occupy is specified by the <tt>N</tt>
1536 <table class="layout">
1538 <td class="left"><tt>i1</tt></td>
1539 <td class="left">a single-bit integer.</td>
1542 <td class="left"><tt>i32</tt></td>
1543 <td class="left">a 32-bit integer.</td>
1546 <td class="left"><tt>i1942652</tt></td>
1547 <td class="left">a really big integer of over 1 million bits.</td>
1553 <!-- _______________________________________________________________________ -->
1554 <div class="doc_subsubsection"> <a name="t_floating">Floating Point Types</a> </div>
1556 <div class="doc_text">
1560 <tr><th>Type</th><th>Description</th></tr>
1561 <tr><td><tt>float</tt></td><td>32-bit floating point value</td></tr>
1562 <tr><td><tt>double</tt></td><td>64-bit floating point value</td></tr>
1563 <tr><td><tt>fp128</tt></td><td>128-bit floating point value (112-bit mantissa)</td></tr>
1564 <tr><td><tt>x86_fp80</tt></td><td>80-bit floating point value (X87)</td></tr>
1565 <tr><td><tt>ppc_fp128</tt></td><td>128-bit floating point value (two 64-bits)</td></tr>
1571 <!-- _______________________________________________________________________ -->
1572 <div class="doc_subsubsection"> <a name="t_void">Void Type</a> </div>
1574 <div class="doc_text">
1577 <p>The void type does not represent any value and has no size.</p>
1586 <!-- _______________________________________________________________________ -->
1587 <div class="doc_subsubsection"> <a name="t_label">Label Type</a> </div>
1589 <div class="doc_text">
1592 <p>The label type represents code labels.</p>
1601 <!-- _______________________________________________________________________ -->
1602 <div class="doc_subsubsection"> <a name="t_metadata">Metadata Type</a> </div>
1604 <div class="doc_text">
1607 <p>The metadata type represents embedded metadata. No derived types may be
1608 created from metadata except for <a href="#t_function">function</a>
1619 <!-- ======================================================================= -->
1620 <div class="doc_subsection"> <a name="t_derived">Derived Types</a> </div>
1622 <div class="doc_text">
1624 <p>The real power in LLVM comes from the derived types in the system. This is
1625 what allows a programmer to represent arrays, functions, pointers, and other
1626 useful types. Each of these types contain one or more element types which
1627 may be a primitive type, or another derived type. For example, it is
1628 possible to have a two dimensional array, using an array as the element type
1629 of another array.</p>
1634 <!-- _______________________________________________________________________ -->
1635 <div class="doc_subsubsection"> <a name="t_aggregate">Aggregate Types</a> </div>
1637 <div class="doc_text">
1639 <p>Aggregate Types are a subset of derived types that can contain multiple
1640 member types. <a href="#t_array">Arrays</a>,
1641 <a href="#t_struct">structs</a>, <a href="#t_vector">vectors</a> and
1642 <a href="#t_union">unions</a> are aggregate types.</p>
1648 <!-- _______________________________________________________________________ -->
1649 <div class="doc_subsubsection"> <a name="t_array">Array Type</a> </div>
1651 <div class="doc_text">
1654 <p>The array type is a very simple derived type that arranges elements
1655 sequentially in memory. The array type requires a size (number of elements)
1656 and an underlying data type.</p>
1660 [<# elements> x <elementtype>]
1663 <p>The number of elements is a constant integer value; <tt>elementtype</tt> may
1664 be any type with a size.</p>
1667 <table class="layout">
1669 <td class="left"><tt>[40 x i32]</tt></td>
1670 <td class="left">Array of 40 32-bit integer values.</td>
1673 <td class="left"><tt>[41 x i32]</tt></td>
1674 <td class="left">Array of 41 32-bit integer values.</td>
1677 <td class="left"><tt>[4 x i8]</tt></td>
1678 <td class="left">Array of 4 8-bit integer values.</td>
1681 <p>Here are some examples of multidimensional arrays:</p>
1682 <table class="layout">
1684 <td class="left"><tt>[3 x [4 x i32]]</tt></td>
1685 <td class="left">3x4 array of 32-bit integer values.</td>
1688 <td class="left"><tt>[12 x [10 x float]]</tt></td>
1689 <td class="left">12x10 array of single precision floating point values.</td>
1692 <td class="left"><tt>[2 x [3 x [4 x i16]]]</tt></td>
1693 <td class="left">2x3x4 array of 16-bit integer values.</td>
1697 <p>There is no restriction on indexing beyond the end of the array implied by
1698 a static type (though there are restrictions on indexing beyond the bounds
1699 of an allocated object in some cases). This means that single-dimension
1700 'variable sized array' addressing can be implemented in LLVM with a zero
1701 length array type. An implementation of 'pascal style arrays' in LLVM could
1702 use the type "<tt>{ i32, [0 x float]}</tt>", for example.</p>
1706 <!-- _______________________________________________________________________ -->
1707 <div class="doc_subsubsection"> <a name="t_function">Function Type</a> </div>
1709 <div class="doc_text">
1712 <p>The function type can be thought of as a function signature. It consists of
1713 a return type and a list of formal parameter types. The return type of a
1714 function type is a scalar type, a void type, a struct type, or a union
1715 type. If the return type is a struct type then all struct elements must be
1716 of first class types, and the struct must have at least one element.</p>
1720 <returntype> (<parameter list>)
1723 <p>...where '<tt><parameter list></tt>' is a comma-separated list of type
1724 specifiers. Optionally, the parameter list may include a type <tt>...</tt>,
1725 which indicates that the function takes a variable number of arguments.
1726 Variable argument functions can access their arguments with
1727 the <a href="#int_varargs">variable argument handling intrinsic</a>
1728 functions. '<tt><returntype></tt>' is any type except
1729 <a href="#t_label">label</a>.</p>
1732 <table class="layout">
1734 <td class="left"><tt>i32 (i32)</tt></td>
1735 <td class="left">function taking an <tt>i32</tt>, returning an <tt>i32</tt>
1737 </tr><tr class="layout">
1738 <td class="left"><tt>float (i16, i32 *) *
1740 <td class="left"><a href="#t_pointer">Pointer</a> to a function that takes
1741 an <tt>i16</tt> and a <a href="#t_pointer">pointer</a> to <tt>i32</tt>,
1742 returning <tt>float</tt>.
1744 </tr><tr class="layout">
1745 <td class="left"><tt>i32 (i8*, ...)</tt></td>
1746 <td class="left">A vararg function that takes at least one
1747 <a href="#t_pointer">pointer</a> to <tt>i8 </tt> (char in C),
1748 which returns an integer. This is the signature for <tt>printf</tt> in
1751 </tr><tr class="layout">
1752 <td class="left"><tt>{i32, i32} (i32)</tt></td>
1753 <td class="left">A function taking an <tt>i32</tt>, returning a
1754 <a href="#t_struct">structure</a> containing two <tt>i32</tt> values
1761 <!-- _______________________________________________________________________ -->
1762 <div class="doc_subsubsection"> <a name="t_struct">Structure Type</a> </div>
1764 <div class="doc_text">
1767 <p>The structure type is used to represent a collection of data members together
1768 in memory. The packing of the field types is defined to match the ABI of the
1769 underlying processor. The elements of a structure may be any type that has a
1772 <p>Structures in memory are accessed using '<tt><a href="#i_load">load</a></tt>'
1773 and '<tt><a href="#i_store">store</a></tt>' by getting a pointer to a field
1774 with the '<tt><a href="#i_getelementptr">getelementptr</a></tt>' instruction.
1775 Structures in registers are accessed using the
1776 '<tt><a href="#i_extractvalue">extractvalue</a></tt>' and
1777 '<tt><a href="#i_insertvalue">insertvalue</a></tt>' instructions.</p>
1780 { <type list> }
1784 <table class="layout">
1786 <td class="left"><tt>{ i32, i32, i32 }</tt></td>
1787 <td class="left">A triple of three <tt>i32</tt> values</td>
1788 </tr><tr class="layout">
1789 <td class="left"><tt>{ float, i32 (i32) * }</tt></td>
1790 <td class="left">A pair, where the first element is a <tt>float</tt> and the
1791 second element is a <a href="#t_pointer">pointer</a> to a
1792 <a href="#t_function">function</a> that takes an <tt>i32</tt>, returning
1793 an <tt>i32</tt>.</td>
1799 <!-- _______________________________________________________________________ -->
1800 <div class="doc_subsubsection"> <a name="t_pstruct">Packed Structure Type</a>
1803 <div class="doc_text">
1806 <p>The packed structure type is used to represent a collection of data members
1807 together in memory. There is no padding between fields. Further, the
1808 alignment of a packed structure is 1 byte. The elements of a packed
1809 structure may be any type that has a size.</p>
1811 <p>Structures are accessed using '<tt><a href="#i_load">load</a></tt> and
1812 '<tt><a href="#i_store">store</a></tt>' by getting a pointer to a field with
1813 the '<tt><a href="#i_getelementptr">getelementptr</a></tt>' instruction.</p>
1817 < { <type list> } >
1821 <table class="layout">
1823 <td class="left"><tt>< { i32, i32, i32 } ></tt></td>
1824 <td class="left">A triple of three <tt>i32</tt> values</td>
1825 </tr><tr class="layout">
1827 <tt>< { float, i32 (i32)* } ></tt></td>
1828 <td class="left">A pair, where the first element is a <tt>float</tt> and the
1829 second element is a <a href="#t_pointer">pointer</a> to a
1830 <a href="#t_function">function</a> that takes an <tt>i32</tt>, returning
1831 an <tt>i32</tt>.</td>
1837 <!-- _______________________________________________________________________ -->
1838 <div class="doc_subsubsection"> <a name="t_union">Union Type</a> </div>
1840 <div class="doc_text">
1843 <p>A union type describes an object with size and alignment suitable for
1844 an object of any one of a given set of types (also known as an "untagged"
1845 union). It is similar in concept and usage to a
1846 <a href="#t_struct">struct</a>, except that all members of the union
1847 have an offset of zero. The elements of a union may be any type that has a
1848 size. Unions must have at least one member - empty unions are not allowed.
1851 <p>The size of the union as a whole will be the size of its largest member,
1852 and the alignment requirements of the union as a whole will be the largest
1853 alignment requirement of any member.</p>
1855 <p>Union members are accessed using '<tt><a href="#i_load">load</a></tt> and
1856 '<tt><a href="#i_store">store</a></tt>' by getting a pointer to a field with
1857 the '<tt><a href="#i_getelementptr">getelementptr</a></tt>' instruction.
1858 Since all members are at offset zero, the getelementptr instruction does
1859 not affect the address, only the type of the resulting pointer.</p>
1863 union { <type list> }
1867 <table class="layout">
1869 <td class="left"><tt>union { i32, i32*, float }</tt></td>
1870 <td class="left">A union of three types: an <tt>i32</tt>, a pointer to
1871 an <tt>i32</tt>, and a <tt>float</tt>.</td>
1872 </tr><tr class="layout">
1874 <tt>union { float, i32 (i32) * }</tt></td>
1875 <td class="left">A union, where the first element is a <tt>float</tt> and the
1876 second element is a <a href="#t_pointer">pointer</a> to a
1877 <a href="#t_function">function</a> that takes an <tt>i32</tt>, returning
1878 an <tt>i32</tt>.</td>
1884 <!-- _______________________________________________________________________ -->
1885 <div class="doc_subsubsection"> <a name="t_pointer">Pointer Type</a> </div>
1887 <div class="doc_text">
1890 <p>The pointer type is used to specify memory locations.
1891 Pointers are commonly used to reference objects in memory.</p>
1893 <p>Pointer types may have an optional address space attribute defining the
1894 numbered address space where the pointed-to object resides. The default
1895 address space is number zero. The semantics of non-zero address
1896 spaces are target-specific.</p>
1898 <p>Note that LLVM does not permit pointers to void (<tt>void*</tt>) nor does it
1899 permit pointers to labels (<tt>label*</tt>). Use <tt>i8*</tt> instead.</p>
1907 <table class="layout">
1909 <td class="left"><tt>[4 x i32]*</tt></td>
1910 <td class="left">A <a href="#t_pointer">pointer</a> to <a
1911 href="#t_array">array</a> of four <tt>i32</tt> values.</td>
1914 <td class="left"><tt>i32 (i32*) *</tt></td>
1915 <td class="left"> A <a href="#t_pointer">pointer</a> to a <a
1916 href="#t_function">function</a> that takes an <tt>i32*</tt>, returning an
1920 <td class="left"><tt>i32 addrspace(5)*</tt></td>
1921 <td class="left">A <a href="#t_pointer">pointer</a> to an <tt>i32</tt> value
1922 that resides in address space #5.</td>
1928 <!-- _______________________________________________________________________ -->
1929 <div class="doc_subsubsection"> <a name="t_vector">Vector Type</a> </div>
1931 <div class="doc_text">
1934 <p>A vector type is a simple derived type that represents a vector of elements.
1935 Vector types are used when multiple primitive data are operated in parallel
1936 using a single instruction (SIMD). A vector type requires a size (number of
1937 elements) and an underlying primitive data type. Vector types are considered
1938 <a href="#t_firstclass">first class</a>.</p>
1942 < <# elements> x <elementtype> >
1945 <p>The number of elements is a constant integer value; elementtype may be any
1946 integer or floating point type.</p>
1949 <table class="layout">
1951 <td class="left"><tt><4 x i32></tt></td>
1952 <td class="left">Vector of 4 32-bit integer values.</td>
1955 <td class="left"><tt><8 x float></tt></td>
1956 <td class="left">Vector of 8 32-bit floating-point values.</td>
1959 <td class="left"><tt><2 x i64></tt></td>
1960 <td class="left">Vector of 2 64-bit integer values.</td>
1966 <!-- _______________________________________________________________________ -->
1967 <div class="doc_subsubsection"> <a name="t_opaque">Opaque Type</a> </div>
1968 <div class="doc_text">
1971 <p>Opaque types are used to represent unknown types in the system. This
1972 corresponds (for example) to the C notion of a forward declared structure
1973 type. In LLVM, opaque types can eventually be resolved to any type (not just
1974 a structure type).</p>
1982 <table class="layout">
1984 <td class="left"><tt>opaque</tt></td>
1985 <td class="left">An opaque type.</td>
1991 <!-- ======================================================================= -->
1992 <div class="doc_subsection">
1993 <a name="t_uprefs">Type Up-references</a>
1996 <div class="doc_text">
1999 <p>An "up reference" allows you to refer to a lexically enclosing type without
2000 requiring it to have a name. For instance, a structure declaration may
2001 contain a pointer to any of the types it is lexically a member of. Example
2002 of up references (with their equivalent as named type declarations)
2006 { \2 * } %x = type { %x* }
2007 { \2 }* %y = type { %y }*
2011 <p>An up reference is needed by the asmprinter for printing out cyclic types
2012 when there is no declared name for a type in the cycle. Because the
2013 asmprinter does not want to print out an infinite type string, it needs a
2014 syntax to handle recursive types that have no names (all names are optional
2022 <p>The level is the count of the lexical type that is being referred to.</p>
2025 <table class="layout">
2027 <td class="left"><tt>\1*</tt></td>
2028 <td class="left">Self-referential pointer.</td>
2031 <td class="left"><tt>{ { \3*, i8 }, i32 }</tt></td>
2032 <td class="left">Recursive structure where the upref refers to the out-most
2039 <!-- *********************************************************************** -->
2040 <div class="doc_section"> <a name="constants">Constants</a> </div>
2041 <!-- *********************************************************************** -->
2043 <div class="doc_text">
2045 <p>LLVM has several different basic types of constants. This section describes
2046 them all and their syntax.</p>
2050 <!-- ======================================================================= -->
2051 <div class="doc_subsection"><a name="simpleconstants">Simple Constants</a></div>
2053 <div class="doc_text">
2056 <dt><b>Boolean constants</b></dt>
2057 <dd>The two strings '<tt>true</tt>' and '<tt>false</tt>' are both valid
2058 constants of the <tt><a href="#t_integer">i1</a></tt> type.</dd>
2060 <dt><b>Integer constants</b></dt>
2061 <dd>Standard integers (such as '4') are constants of
2062 the <a href="#t_integer">integer</a> type. Negative numbers may be used
2063 with integer types.</dd>
2065 <dt><b>Floating point constants</b></dt>
2066 <dd>Floating point constants use standard decimal notation (e.g. 123.421),
2067 exponential notation (e.g. 1.23421e+2), or a more precise hexadecimal
2068 notation (see below). The assembler requires the exact decimal value of a
2069 floating-point constant. For example, the assembler accepts 1.25 but
2070 rejects 1.3 because 1.3 is a repeating decimal in binary. Floating point
2071 constants must have a <a href="#t_floating">floating point</a> type. </dd>
2073 <dt><b>Null pointer constants</b></dt>
2074 <dd>The identifier '<tt>null</tt>' is recognized as a null pointer constant
2075 and must be of <a href="#t_pointer">pointer type</a>.</dd>
2078 <p>The one non-intuitive notation for constants is the hexadecimal form of
2079 floating point constants. For example, the form '<tt>double
2080 0x432ff973cafa8000</tt>' is equivalent to (but harder to read than)
2081 '<tt>double 4.5e+15</tt>'. The only time hexadecimal floating point
2082 constants are required (and the only time that they are generated by the
2083 disassembler) is when a floating point constant must be emitted but it cannot
2084 be represented as a decimal floating point number in a reasonable number of
2085 digits. For example, NaN's, infinities, and other special values are
2086 represented in their IEEE hexadecimal format so that assembly and disassembly
2087 do not cause any bits to change in the constants.</p>
2089 <p>When using the hexadecimal form, constants of types float and double are
2090 represented using the 16-digit form shown above (which matches the IEEE754
2091 representation for double); float values must, however, be exactly
2092 representable as IEE754 single precision. Hexadecimal format is always used
2093 for long double, and there are three forms of long double. The 80-bit format
2094 used by x86 is represented as <tt>0xK</tt> followed by 20 hexadecimal digits.
2095 The 128-bit format used by PowerPC (two adjacent doubles) is represented
2096 by <tt>0xM</tt> followed by 32 hexadecimal digits. The IEEE 128-bit format
2097 is represented by <tt>0xL</tt> followed by 32 hexadecimal digits; no
2098 currently supported target uses this format. Long doubles will only work if
2099 they match the long double format on your target. All hexadecimal formats
2100 are big-endian (sign bit at the left).</p>
2104 <!-- ======================================================================= -->
2105 <div class="doc_subsection">
2106 <a name="aggregateconstants"></a> <!-- old anchor -->
2107 <a name="complexconstants">Complex Constants</a>
2110 <div class="doc_text">
2112 <p>Complex constants are a (potentially recursive) combination of simple
2113 constants and smaller complex constants.</p>
2116 <dt><b>Structure constants</b></dt>
2117 <dd>Structure constants are represented with notation similar to structure
2118 type definitions (a comma separated list of elements, surrounded by braces
2119 (<tt>{}</tt>)). For example: "<tt>{ i32 4, float 17.0, i32* @G }</tt>",
2120 where "<tt>@G</tt>" is declared as "<tt>@G = external global i32</tt>".
2121 Structure constants must have <a href="#t_struct">structure type</a>, and
2122 the number and types of elements must match those specified by the
2125 <dt><b>Union constants</b></dt>
2126 <dd>Union constants are represented with notation similar to a structure with
2127 a single element - that is, a single typed element surrounded
2128 by braces (<tt>{}</tt>)). For example: "<tt>{ i32 4 }</tt>". The
2129 <a href="#t_union">union type</a> can be initialized with a single-element
2130 struct as long as the type of the struct element matches the type of
2131 one of the union members.</dd>
2133 <dt><b>Array constants</b></dt>
2134 <dd>Array constants are represented with notation similar to array type
2135 definitions (a comma separated list of elements, surrounded by square
2136 brackets (<tt>[]</tt>)). For example: "<tt>[ i32 42, i32 11, i32 74
2137 ]</tt>". Array constants must have <a href="#t_array">array type</a>, and
2138 the number and types of elements must match those specified by the
2141 <dt><b>Vector constants</b></dt>
2142 <dd>Vector constants are represented with notation similar to vector type
2143 definitions (a comma separated list of elements, surrounded by
2144 less-than/greater-than's (<tt><></tt>)). For example: "<tt>< i32
2145 42, i32 11, i32 74, i32 100 ></tt>". Vector constants must
2146 have <a href="#t_vector">vector type</a>, and the number and types of
2147 elements must match those specified by the type.</dd>
2149 <dt><b>Zero initialization</b></dt>
2150 <dd>The string '<tt>zeroinitializer</tt>' can be used to zero initialize a
2151 value to zero of <em>any</em> type, including scalar and
2152 <a href="#t_aggregate">aggregate</a> types.
2153 This is often used to avoid having to print large zero initializers
2154 (e.g. for large arrays) and is always exactly equivalent to using explicit
2155 zero initializers.</dd>
2157 <dt><b>Metadata node</b></dt>
2158 <dd>A metadata node is a structure-like constant with
2159 <a href="#t_metadata">metadata type</a>. For example: "<tt>metadata !{
2160 i32 0, metadata !"test" }</tt>". Unlike other constants that are meant to
2161 be interpreted as part of the instruction stream, metadata is a place to
2162 attach additional information such as debug info.</dd>
2167 <!-- ======================================================================= -->
2168 <div class="doc_subsection">
2169 <a name="globalconstants">Global Variable and Function Addresses</a>
2172 <div class="doc_text">
2174 <p>The addresses of <a href="#globalvars">global variables</a>
2175 and <a href="#functionstructure">functions</a> are always implicitly valid
2176 (link-time) constants. These constants are explicitly referenced when
2177 the <a href="#identifiers">identifier for the global</a> is used and always
2178 have <a href="#t_pointer">pointer</a> type. For example, the following is a
2179 legal LLVM file:</p>
2181 <div class="doc_code">
2185 @Z = global [2 x i32*] [ i32* @X, i32* @Y ]
2191 <!-- ======================================================================= -->
2192 <div class="doc_subsection"><a name="undefvalues">Undefined Values</a></div>
2193 <div class="doc_text">
2195 <p>The string '<tt>undef</tt>' can be used anywhere a constant is expected, and
2196 indicates that the user of the value may receive an unspecified bit-pattern.
2197 Undefined values may be of any type (other than label or void) and be used
2198 anywhere a constant is permitted.</p>
2200 <p>Undefined values are useful because they indicate to the compiler that the
2201 program is well defined no matter what value is used. This gives the
2202 compiler more freedom to optimize. Here are some examples of (potentially
2203 surprising) transformations that are valid (in pseudo IR):</p>
2206 <div class="doc_code">
2218 <p>This is safe because all of the output bits are affected by the undef bits.
2219 Any output bit can have a zero or one depending on the input bits.</p>
2221 <div class="doc_code">
2234 <p>These logical operations have bits that are not always affected by the input.
2235 For example, if "%X" has a zero bit, then the output of the 'and' operation will
2236 always be a zero, no matter what the corresponding bit from the undef is. As
2237 such, it is unsafe to optimize or assume that the result of the and is undef.
2238 However, it is safe to assume that all bits of the undef could be 0, and
2239 optimize the and to 0. Likewise, it is safe to assume that all the bits of
2240 the undef operand to the or could be set, allowing the or to be folded to
2243 <div class="doc_code">
2245 %A = select undef, %X, %Y
2246 %B = select undef, 42, %Y
2247 %C = select %X, %Y, undef
2259 <p>This set of examples show that undefined select (and conditional branch)
2260 conditions can go "either way" but they have to come from one of the two
2261 operands. In the %A example, if %X and %Y were both known to have a clear low
2262 bit, then %A would have to have a cleared low bit. However, in the %C example,
2263 the optimizer is allowed to assume that the undef operand could be the same as
2264 %Y, allowing the whole select to be eliminated.</p>
2267 <div class="doc_code">
2269 %A = xor undef, undef
2288 <p>This example points out that two undef operands are not necessarily the same.
2289 This can be surprising to people (and also matches C semantics) where they
2290 assume that "X^X" is always zero, even if X is undef. This isn't true for a
2291 number of reasons, but the short answer is that an undef "variable" can
2292 arbitrarily change its value over its "live range". This is true because the
2293 "variable" doesn't actually <em>have a live range</em>. Instead, the value is
2294 logically read from arbitrary registers that happen to be around when needed,
2295 so the value is not necessarily consistent over time. In fact, %A and %C need
2296 to have the same semantics or the core LLVM "replace all uses with" concept
2299 <div class="doc_code">
2309 <p>These examples show the crucial difference between an <em>undefined
2310 value</em> and <em>undefined behavior</em>. An undefined value (like undef) is
2311 allowed to have an arbitrary bit-pattern. This means that the %A operation
2312 can be constant folded to undef because the undef could be an SNaN, and fdiv is
2313 not (currently) defined on SNaN's. However, in the second example, we can make
2314 a more aggressive assumption: because the undef is allowed to be an arbitrary
2315 value, we are allowed to assume that it could be zero. Since a divide by zero
2316 has <em>undefined behavior</em>, we are allowed to assume that the operation
2317 does not execute at all. This allows us to delete the divide and all code after
2318 it: since the undefined operation "can't happen", the optimizer can assume that
2319 it occurs in dead code.
2322 <div class="doc_code">
2324 a: store undef -> %X
2325 b: store %X -> undef
2332 <p>These examples reiterate the fdiv example: a store "of" an undefined value
2333 can be assumed to not have any effect: we can assume that the value is
2334 overwritten with bits that happen to match what was already there. However, a
2335 store "to" an undefined location could clobber arbitrary memory, therefore, it
2336 has undefined behavior.</p>
2340 <!-- ======================================================================= -->
2341 <div class="doc_subsection"><a name="trapvalues">Trap Values</a></div>
2342 <div class="doc_text">
2344 <p>Trap values are similar to <a href="#undefvalues">undef values</a>, however
2345 instead of representing an unspecified bit pattern, they represent the
2346 fact that an instruction or constant expression which cannot evoke side
2347 effects has nevertheless detected a condition which results in undefined
2350 <p>There is currently no way of representing a trap value in the IR; they
2351 only exist when produced by operations such as
2352 <a href="#i_add"><tt>add</tt></a> with the <tt>nsw</tt> flag.</p>
2354 <p>Trap value behavior is defined in terms of value <i>dependence</i>:</p>
2358 <li>Values other than <a href="#i_phi"><tt>phi</tt></a> nodes depend on
2359 their operands.</li>
2361 <li><a href="#i_phi"><tt>Phi</tt></a> nodes depend on the operand corresponding
2362 to their dynamic predecessor basic block.</li>
2364 <li>Function arguments depend on the corresponding actual argument values in
2365 the dynamic callers of their functions.</li>
2367 <li><a href="#i_call"><tt>Call</tt></a> instructions depend on the
2368 <a href="#i_ret"><tt>ret</tt></a> instructions that dynamically transfer
2369 control back to them.</li>
2371 <li><a href="#i_invoke"><tt>Invoke</tt></a> instructions depend on the
2372 <a href="#i_ret"><tt>ret</tt></a>, <a href="#i_unwind"><tt>unwind</tt></a>,
2373 or exception-throwing call instructions that dynamically transfer control
2376 <li>Non-volatile loads and stores depend on the most recent stores to all of the
2377 referenced memory addresses, following the order in the IR
2378 (including loads and stores implied by intrinsics such as
2379 <a href="#int_memcpy"><tt>@llvm.memcpy</tt></a>.)</li>
2381 <!-- TODO: In the case of multiple threads, this only applies if the store
2382 "happens-before" the load or store. -->
2384 <!-- TODO: floating-point exception state -->
2386 <li>An instruction with externally visible side effects depends on the most
2387 recent preceding instruction with externally visible side effects, following
2388 the order in the IR. (This includes volatile loads and stores.)</li>
2390 <li>An instruction <i>control-depends</i> on a
2391 <a href="#terminators">terminator instruction</a>
2392 if the terminator instruction has multiple successors and the instruction
2393 is always executed when control transfers to one of the successors, and
2394 may not be executed when control is transfered to another.</li>
2396 <li>Dependence is transitive.</li>
2401 <p>Whenever a trap value is generated, all values which depend on it evaluate
2402 to trap. If they have side effects, the evoke their side effects as if each
2403 operand with a trap value were undef. If they have externally-visible side
2404 effects, the behavior is undefined.</p>
2406 <p>Here are some examples:</p>
2408 <div class="doc_code">
2411 %trap = sub nuw i32 0, 1 ; Results in a trap value.
2412 %still_trap = and i32 %trap, 0 ; Whereas (and i32 undef, 0) would return 0.
2413 %trap_yet_again = getelementptr i32* @h, i32 %still_trap
2414 store i32 0, i32* %trap_yet_again ; undefined behavior
2416 store i32 %trap, i32* @g ; Trap value conceptually stored to memory.
2417 %trap2 = load i32* @g ; Returns a trap value, not just undef.
2419 volatile store i32 %trap, i32* @g ; External observation; undefined behavior.
2421 %narrowaddr = bitcast i32* @g to i16*
2422 %wideaddr = bitcast i32* @g to i64*
2423 %trap3 = load 16* %narrowaddr ; Returns a trap value.
2424 %trap4 = load i64* %widaddr ; Returns a trap value.
2426 %cmp = icmp i32 slt %trap, 0 ; Returns a trap value.
2427 %br i1 %cmp, %true, %end ; Branch to either destination.
2430 volatile store i32 0, i32* @g ; This is control-dependent on %cmp, so
2431 ; it has undefined behavior.
2435 %p = phi i32 [ 0, %entry ], [ 1, %true ]
2436 ; Both edges into this PHI are
2437 ; control-dependent on %cmp, so this
2438 ; always results in a trap value.
2440 volatile store i32 0, i32* @g ; %end is control-equivalent to %entry
2441 ; so this is defined (ignoring earlier
2442 ; undefined behavior in this example).
2448 <!-- ======================================================================= -->
2449 <div class="doc_subsection"><a name="blockaddress">Addresses of Basic
2451 <div class="doc_text">
2453 <p><b><tt>blockaddress(@function, %block)</tt></b></p>
2455 <p>The '<tt>blockaddress</tt>' constant computes the address of the specified
2456 basic block in the specified function, and always has an i8* type. Taking
2457 the address of the entry block is illegal.</p>
2459 <p>This value only has defined behavior when used as an operand to the
2460 '<a href="#i_indirectbr"><tt>indirectbr</tt></a>' instruction or for comparisons
2461 against null. Pointer equality tests between labels addresses is undefined
2462 behavior - though, again, comparison against null is ok, and no label is
2463 equal to the null pointer. This may also be passed around as an opaque
2464 pointer sized value as long as the bits are not inspected. This allows
2465 <tt>ptrtoint</tt> and arithmetic to be performed on these values so long as
2466 the original value is reconstituted before the <tt>indirectbr</tt>.</p>
2468 <p>Finally, some targets may provide defined semantics when
2469 using the value as the operand to an inline assembly, but that is target
2476 <!-- ======================================================================= -->
2477 <div class="doc_subsection"><a name="constantexprs">Constant Expressions</a>
2480 <div class="doc_text">
2482 <p>Constant expressions are used to allow expressions involving other constants
2483 to be used as constants. Constant expressions may be of
2484 any <a href="#t_firstclass">first class</a> type and may involve any LLVM
2485 operation that does not have side effects (e.g. load and call are not
2486 supported). The following is the syntax for constant expressions:</p>
2489 <dt><b><tt>trunc (CST to TYPE)</tt></b></dt>
2490 <dd>Truncate a constant to another type. The bit size of CST must be larger
2491 than the bit size of TYPE. Both types must be integers.</dd>
2493 <dt><b><tt>zext (CST to TYPE)</tt></b></dt>
2494 <dd>Zero extend a constant to another type. The bit size of CST must be
2495 smaller or equal to the bit size of TYPE. Both types must be
2498 <dt><b><tt>sext (CST to TYPE)</tt></b></dt>
2499 <dd>Sign extend a constant to another type. The bit size of CST must be
2500 smaller or equal to the bit size of TYPE. Both types must be
2503 <dt><b><tt>fptrunc (CST to TYPE)</tt></b></dt>
2504 <dd>Truncate a floating point constant to another floating point type. The
2505 size of CST must be larger than the size of TYPE. Both types must be
2506 floating point.</dd>
2508 <dt><b><tt>fpext (CST to TYPE)</tt></b></dt>
2509 <dd>Floating point extend a constant to another type. The size of CST must be
2510 smaller or equal to the size of TYPE. Both types must be floating
2513 <dt><b><tt>fptoui (CST to TYPE)</tt></b></dt>
2514 <dd>Convert a floating point constant to the corresponding unsigned integer
2515 constant. TYPE must be a scalar or vector integer type. CST must be of
2516 scalar or vector floating point type. Both CST and TYPE must be scalars,
2517 or vectors of the same number of elements. If the value won't fit in the
2518 integer type, the results are undefined.</dd>
2520 <dt><b><tt>fptosi (CST to TYPE)</tt></b></dt>
2521 <dd>Convert a floating point constant to the corresponding signed integer
2522 constant. TYPE must be a scalar or vector integer type. CST must be of
2523 scalar or vector floating point type. Both CST and TYPE must be scalars,
2524 or vectors of the same number of elements. If the value won't fit in the
2525 integer type, the results are undefined.</dd>
2527 <dt><b><tt>uitofp (CST to TYPE)</tt></b></dt>
2528 <dd>Convert an unsigned integer constant to the corresponding floating point
2529 constant. TYPE must be a scalar or vector floating point type. CST must be
2530 of scalar or vector integer type. Both CST and TYPE must be scalars, or
2531 vectors of the same number of elements. If the value won't fit in the
2532 floating point type, the results are undefined.</dd>
2534 <dt><b><tt>sitofp (CST to TYPE)</tt></b></dt>
2535 <dd>Convert a signed integer constant to the corresponding floating point
2536 constant. TYPE must be a scalar or vector floating point type. CST must be
2537 of scalar or vector integer type. Both CST and TYPE must be scalars, or
2538 vectors of the same number of elements. If the value won't fit in the
2539 floating point type, the results are undefined.</dd>
2541 <dt><b><tt>ptrtoint (CST to TYPE)</tt></b></dt>
2542 <dd>Convert a pointer typed constant to the corresponding integer constant
2543 <tt>TYPE</tt> must be an integer type. <tt>CST</tt> must be of pointer
2544 type. The <tt>CST</tt> value is zero extended, truncated, or unchanged to
2545 make it fit in <tt>TYPE</tt>.</dd>
2547 <dt><b><tt>inttoptr (CST to TYPE)</tt></b></dt>
2548 <dd>Convert a integer constant to a pointer constant. TYPE must be a pointer
2549 type. CST must be of integer type. The CST value is zero extended,
2550 truncated, or unchanged to make it fit in a pointer size. This one is
2551 <i>really</i> dangerous!</dd>
2553 <dt><b><tt>bitcast (CST to TYPE)</tt></b></dt>
2554 <dd>Convert a constant, CST, to another TYPE. The constraints of the operands
2555 are the same as those for the <a href="#i_bitcast">bitcast
2556 instruction</a>.</dd>
2558 <dt><b><tt>getelementptr (CSTPTR, IDX0, IDX1, ...)</tt></b></dt>
2559 <dt><b><tt>getelementptr inbounds (CSTPTR, IDX0, IDX1, ...)</tt></b></dt>
2560 <dd>Perform the <a href="#i_getelementptr">getelementptr operation</a> on
2561 constants. As with the <a href="#i_getelementptr">getelementptr</a>
2562 instruction, the index list may have zero or more indexes, which are
2563 required to make sense for the type of "CSTPTR".</dd>
2565 <dt><b><tt>select (COND, VAL1, VAL2)</tt></b></dt>
2566 <dd>Perform the <a href="#i_select">select operation</a> on constants.</dd>
2568 <dt><b><tt>icmp COND (VAL1, VAL2)</tt></b></dt>
2569 <dd>Performs the <a href="#i_icmp">icmp operation</a> on constants.</dd>
2571 <dt><b><tt>fcmp COND (VAL1, VAL2)</tt></b></dt>
2572 <dd>Performs the <a href="#i_fcmp">fcmp operation</a> on constants.</dd>
2574 <dt><b><tt>extractelement (VAL, IDX)</tt></b></dt>
2575 <dd>Perform the <a href="#i_extractelement">extractelement operation</a> on
2578 <dt><b><tt>insertelement (VAL, ELT, IDX)</tt></b></dt>
2579 <dd>Perform the <a href="#i_insertelement">insertelement operation</a> on
2582 <dt><b><tt>shufflevector (VEC1, VEC2, IDXMASK)</tt></b></dt>
2583 <dd>Perform the <a href="#i_shufflevector">shufflevector operation</a> on
2586 <dt><b><tt>extractvalue (VAL, IDX0, IDX1, ...)</tt></b></dt>
2587 <dd>Perform the <a href="#i_extractvalue">extractvalue operation</a> on
2588 constants. The index list is interpreted in a similar manner as indices in
2589 a '<a href="#i_getelementptr">getelementptr</a>' operation. At least one
2590 index value must be specified.</dd>
2592 <dt><b><tt>insertvalue (VAL, ELT, IDX0, IDX1, ...)</tt></b></dt>
2593 <dd>Perform the <a href="#i_insertvalue">insertvalue operation</a> on
2594 constants. The index list is interpreted in a similar manner as indices in
2595 a '<a href="#i_getelementptr">getelementptr</a>' operation. At least one
2596 index value must be specified.</dd>
2598 <dt><b><tt>OPCODE (LHS, RHS)</tt></b></dt>
2599 <dd>Perform the specified operation of the LHS and RHS constants. OPCODE may
2600 be any of the <a href="#binaryops">binary</a>
2601 or <a href="#bitwiseops">bitwise binary</a> operations. The constraints
2602 on operands are the same as those for the corresponding instruction
2603 (e.g. no bitwise operations on floating point values are allowed).</dd>
2608 <!-- *********************************************************************** -->
2609 <div class="doc_section"> <a name="othervalues">Other Values</a> </div>
2610 <!-- *********************************************************************** -->
2612 <!-- ======================================================================= -->
2613 <div class="doc_subsection">
2614 <a name="inlineasm">Inline Assembler Expressions</a>
2617 <div class="doc_text">
2619 <p>LLVM supports inline assembler expressions (as opposed
2620 to <a href="#moduleasm"> Module-Level Inline Assembly</a>) through the use of
2621 a special value. This value represents the inline assembler as a string
2622 (containing the instructions to emit), a list of operand constraints (stored
2623 as a string), a flag that indicates whether or not the inline asm
2624 expression has side effects, and a flag indicating whether the function
2625 containing the asm needs to align its stack conservatively. An example
2626 inline assembler expression is:</p>
2628 <div class="doc_code">
2630 i32 (i32) asm "bswap $0", "=r,r"
2634 <p>Inline assembler expressions may <b>only</b> be used as the callee operand of
2635 a <a href="#i_call"><tt>call</tt> instruction</a>. Thus, typically we
2638 <div class="doc_code">
2640 %X = call i32 asm "<a href="#int_bswap">bswap</a> $0", "=r,r"(i32 %Y)
2644 <p>Inline asms with side effects not visible in the constraint list must be
2645 marked as having side effects. This is done through the use of the
2646 '<tt>sideeffect</tt>' keyword, like so:</p>
2648 <div class="doc_code">
2650 call void asm sideeffect "eieio", ""()
2654 <p>In some cases inline asms will contain code that will not work unless the
2655 stack is aligned in some way, such as calls or SSE instructions on x86,
2656 yet will not contain code that does that alignment within the asm.
2657 The compiler should make conservative assumptions about what the asm might
2658 contain and should generate its usual stack alignment code in the prologue
2659 if the '<tt>alignstack</tt>' keyword is present:</p>
2661 <div class="doc_code">
2663 call void asm alignstack "eieio", ""()
2667 <p>If both keywords appear the '<tt>sideeffect</tt>' keyword must come
2670 <p>TODO: The format of the asm and constraints string still need to be
2671 documented here. Constraints on what can be done (e.g. duplication, moving,
2672 etc need to be documented). This is probably best done by reference to
2673 another document that covers inline asm from a holistic perspective.</p>
2676 <div class="doc_subsubsection">
2677 <a name="inlineasm_md">Inline Asm Metadata</a>
2680 <div class="doc_text">
2682 <p>The call instructions that wrap inline asm nodes may have a "!srcloc" MDNode
2683 attached to it that contains a constant integer. If present, the code
2684 generator will use the integer as the location cookie value when report
2685 errors through the LLVMContext error reporting mechanisms. This allows a
2686 front-end to correlate backend errors that occur with inline asm back to the
2687 source code that produced it. For example:</p>
2689 <div class="doc_code">
2691 call void asm sideeffect "something bad", ""()<b>, !srcloc !42</b>
2693 !42 = !{ i32 1234567 }
2697 <p>It is up to the front-end to make sense of the magic numbers it places in the
2702 <!-- ======================================================================= -->
2703 <div class="doc_subsection"><a name="metadata">Metadata Nodes and Metadata
2707 <div class="doc_text">
2709 <p>LLVM IR allows metadata to be attached to instructions in the program that
2710 can convey extra information about the code to the optimizers and code
2711 generator. One example application of metadata is source-level debug
2712 information. There are two metadata primitives: strings and nodes. All
2713 metadata has the <tt>metadata</tt> type and is identified in syntax by a
2714 preceding exclamation point ('<tt>!</tt>').</p>
2716 <p>A metadata string is a string surrounded by double quotes. It can contain
2717 any character by escaping non-printable characters with "\xx" where "xx" is
2718 the two digit hex code. For example: "<tt>!"test\00"</tt>".</p>
2720 <p>Metadata nodes are represented with notation similar to structure constants
2721 (a comma separated list of elements, surrounded by braces and preceded by an
2722 exclamation point). For example: "<tt>!{ metadata !"test\00", i32
2723 10}</tt>". Metadata nodes can have any values as their operand.</p>
2725 <p>A <a href="#namedmetadatastructure">named metadata</a> is a collection of
2726 metadata nodes, which can be looked up in the module symbol table. For
2727 example: "<tt>!foo = metadata !{!4, !3}</tt>".
2729 <p>Metadata can be used as function arguments. Here <tt>llvm.dbg.value</tt>
2730 function is using two metadata arguments.
2732 <div class="doc_code">
2734 call void @llvm.dbg.value(metadata !24, i64 0, metadata !25)
2738 <p>Metadata can be attached with an instruction. Here metadata <tt>!21</tt> is
2739 attached with <tt>add</tt> instruction using <tt>!dbg</tt> identifier.
2741 <div class="doc_code">
2743 %indvar.next = add i64 %indvar, 1, !dbg !21
2749 <!-- *********************************************************************** -->
2750 <div class="doc_section">
2751 <a name="intrinsic_globals">Intrinsic Global Variables</a>
2753 <!-- *********************************************************************** -->
2755 <p>LLVM has a number of "magic" global variables that contain data that affect
2756 code generation or other IR semantics. These are documented here. All globals
2757 of this sort should have a section specified as "<tt>llvm.metadata</tt>". This
2758 section and all globals that start with "<tt>llvm.</tt>" are reserved for use
2761 <!-- ======================================================================= -->
2762 <div class="doc_subsection">
2763 <a name="intg_used">The '<tt>llvm.used</tt>' Global Variable</a>
2766 <div class="doc_text">
2768 <p>The <tt>@llvm.used</tt> global is an array with i8* element type which has <a
2769 href="#linkage_appending">appending linkage</a>. This array contains a list of
2770 pointers to global variables and functions which may optionally have a pointer
2771 cast formed of bitcast or getelementptr. For example, a legal use of it is:</p>
2777 @llvm.used = appending global [2 x i8*] [
2779 i8* bitcast (i32* @Y to i8*)
2780 ], section "llvm.metadata"
2783 <p>If a global variable appears in the <tt>@llvm.used</tt> list, then the
2784 compiler, assembler, and linker are required to treat the symbol as if there is
2785 a reference to the global that it cannot see. For example, if a variable has
2786 internal linkage and no references other than that from the <tt>@llvm.used</tt>
2787 list, it cannot be deleted. This is commonly used to represent references from
2788 inline asms and other things the compiler cannot "see", and corresponds to
2789 "attribute((used))" in GNU C.</p>
2791 <p>On some targets, the code generator must emit a directive to the assembler or
2792 object file to prevent the assembler and linker from molesting the symbol.</p>
2796 <!-- ======================================================================= -->
2797 <div class="doc_subsection">
2798 <a name="intg_compiler_used">The '<tt>llvm.compiler.used</tt>' Global Variable</a>
2801 <div class="doc_text">
2803 <p>The <tt>@llvm.compiler.used</tt> directive is the same as the
2804 <tt>@llvm.used</tt> directive, except that it only prevents the compiler from
2805 touching the symbol. On targets that support it, this allows an intelligent
2806 linker to optimize references to the symbol without being impeded as it would be
2807 by <tt>@llvm.used</tt>.</p>
2809 <p>This is a rare construct that should only be used in rare circumstances, and
2810 should not be exposed to source languages.</p>
2814 <!-- ======================================================================= -->
2815 <div class="doc_subsection">
2816 <a name="intg_global_ctors">The '<tt>llvm.global_ctors</tt>' Global Variable</a>
2819 <div class="doc_text">
2821 %0 = type { i32, void ()* }
2822 @llvm.global_ctors = appending global [1 x %0] [%0 { i32 65535, void ()* @ctor }]
2824 <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.
2829 <!-- ======================================================================= -->
2830 <div class="doc_subsection">
2831 <a name="intg_global_dtors">The '<tt>llvm.global_dtors</tt>' Global Variable</a>
2834 <div class="doc_text">
2836 %0 = type { i32, void ()* }
2837 @llvm.global_dtors = appending global [1 x %0] [%0 { i32 65535, void ()* @dtor }]
2840 <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.
2846 <!-- *********************************************************************** -->
2847 <div class="doc_section"> <a name="instref">Instruction Reference</a> </div>
2848 <!-- *********************************************************************** -->
2850 <div class="doc_text">
2852 <p>The LLVM instruction set consists of several different classifications of
2853 instructions: <a href="#terminators">terminator
2854 instructions</a>, <a href="#binaryops">binary instructions</a>,
2855 <a href="#bitwiseops">bitwise binary instructions</a>,
2856 <a href="#memoryops">memory instructions</a>, and
2857 <a href="#otherops">other instructions</a>.</p>
2861 <!-- ======================================================================= -->
2862 <div class="doc_subsection"> <a name="terminators">Terminator
2863 Instructions</a> </div>
2865 <div class="doc_text">
2867 <p>As mentioned <a href="#functionstructure">previously</a>, every basic block
2868 in a program ends with a "Terminator" instruction, which indicates which
2869 block should be executed after the current block is finished. These
2870 terminator instructions typically yield a '<tt>void</tt>' value: they produce
2871 control flow, not values (the one exception being the
2872 '<a href="#i_invoke"><tt>invoke</tt></a>' instruction).</p>
2874 <p>There are seven different terminator instructions: the
2875 '<a href="#i_ret"><tt>ret</tt></a>' instruction, the
2876 '<a href="#i_br"><tt>br</tt></a>' instruction, the
2877 '<a href="#i_switch"><tt>switch</tt></a>' instruction, the
2878 '<a href="#i_indirectbr">'<tt>indirectbr</tt></a>' Instruction, the
2879 '<a href="#i_invoke"><tt>invoke</tt></a>' instruction, the
2880 '<a href="#i_unwind"><tt>unwind</tt></a>' instruction, and the
2881 '<a href="#i_unreachable"><tt>unreachable</tt></a>' instruction.</p>
2885 <!-- _______________________________________________________________________ -->
2886 <div class="doc_subsubsection"> <a name="i_ret">'<tt>ret</tt>'
2887 Instruction</a> </div>
2889 <div class="doc_text">
2893 ret <type> <value> <i>; Return a value from a non-void function</i>
2894 ret void <i>; Return from void function</i>
2898 <p>The '<tt>ret</tt>' instruction is used to return control flow (and optionally
2899 a value) from a function back to the caller.</p>
2901 <p>There are two forms of the '<tt>ret</tt>' instruction: one that returns a
2902 value and then causes control flow, and one that just causes control flow to
2906 <p>The '<tt>ret</tt>' instruction optionally accepts a single argument, the
2907 return value. The type of the return value must be a
2908 '<a href="#t_firstclass">first class</a>' type.</p>
2910 <p>A function is not <a href="#wellformed">well formed</a> if it it has a
2911 non-void return type and contains a '<tt>ret</tt>' instruction with no return
2912 value or a return value with a type that does not match its type, or if it
2913 has a void return type and contains a '<tt>ret</tt>' instruction with a
2917 <p>When the '<tt>ret</tt>' instruction is executed, control flow returns back to
2918 the calling function's context. If the caller is a
2919 "<a href="#i_call"><tt>call</tt></a>" instruction, execution continues at the
2920 instruction after the call. If the caller was an
2921 "<a href="#i_invoke"><tt>invoke</tt></a>" instruction, execution continues at
2922 the beginning of the "normal" destination block. If the instruction returns
2923 a value, that value shall set the call or invoke instruction's return
2928 ret i32 5 <i>; Return an integer value of 5</i>
2929 ret void <i>; Return from a void function</i>
2930 ret { i32, i8 } { i32 4, i8 2 } <i>; Return a struct of values 4 and 2</i>
2934 <!-- _______________________________________________________________________ -->
2935 <div class="doc_subsubsection"> <a name="i_br">'<tt>br</tt>' Instruction</a> </div>
2937 <div class="doc_text">
2941 br i1 <cond>, label <iftrue>, label <iffalse><br> br label <dest> <i>; Unconditional branch</i>
2945 <p>The '<tt>br</tt>' instruction is used to cause control flow to transfer to a
2946 different basic block in the current function. There are two forms of this
2947 instruction, corresponding to a conditional branch and an unconditional
2951 <p>The conditional branch form of the '<tt>br</tt>' instruction takes a single
2952 '<tt>i1</tt>' value and two '<tt>label</tt>' values. The unconditional form
2953 of the '<tt>br</tt>' instruction takes a single '<tt>label</tt>' value as a
2957 <p>Upon execution of a conditional '<tt>br</tt>' instruction, the '<tt>i1</tt>'
2958 argument is evaluated. If the value is <tt>true</tt>, control flows to the
2959 '<tt>iftrue</tt>' <tt>label</tt> argument. If "cond" is <tt>false</tt>,
2960 control flows to the '<tt>iffalse</tt>' <tt>label</tt> argument.</p>
2965 %cond = <a href="#i_icmp">icmp</a> eq i32 %a, %b
2966 br i1 %cond, label %IfEqual, label %IfUnequal
2968 <a href="#i_ret">ret</a> i32 1
2970 <a href="#i_ret">ret</a> i32 0
2975 <!-- _______________________________________________________________________ -->
2976 <div class="doc_subsubsection">
2977 <a name="i_switch">'<tt>switch</tt>' Instruction</a>
2980 <div class="doc_text">
2984 switch <intty> <value>, label <defaultdest> [ <intty> <val>, label <dest> ... ]
2988 <p>The '<tt>switch</tt>' instruction is used to transfer control flow to one of
2989 several different places. It is a generalization of the '<tt>br</tt>'
2990 instruction, allowing a branch to occur to one of many possible
2994 <p>The '<tt>switch</tt>' instruction uses three parameters: an integer
2995 comparison value '<tt>value</tt>', a default '<tt>label</tt>' destination,
2996 and an array of pairs of comparison value constants and '<tt>label</tt>'s.
2997 The table is not allowed to contain duplicate constant entries.</p>
3000 <p>The <tt>switch</tt> instruction specifies a table of values and
3001 destinations. When the '<tt>switch</tt>' instruction is executed, this table
3002 is searched for the given value. If the value is found, control flow is
3003 transferred to the corresponding destination; otherwise, control flow is
3004 transferred to the default destination.</p>
3006 <h5>Implementation:</h5>
3007 <p>Depending on properties of the target machine and the particular
3008 <tt>switch</tt> instruction, this instruction may be code generated in
3009 different ways. For example, it could be generated as a series of chained
3010 conditional branches or with a lookup table.</p>
3014 <i>; Emulate a conditional br instruction</i>
3015 %Val = <a href="#i_zext">zext</a> i1 %value to i32
3016 switch i32 %Val, label %truedest [ i32 0, label %falsedest ]
3018 <i>; Emulate an unconditional br instruction</i>
3019 switch i32 0, label %dest [ ]
3021 <i>; Implement a jump table:</i>
3022 switch i32 %val, label %otherwise [ i32 0, label %onzero
3024 i32 2, label %ontwo ]
3030 <!-- _______________________________________________________________________ -->
3031 <div class="doc_subsubsection">
3032 <a name="i_indirectbr">'<tt>indirectbr</tt>' Instruction</a>
3035 <div class="doc_text">
3039 indirectbr <somety>* <address>, [ label <dest1>, label <dest2>, ... ]
3044 <p>The '<tt>indirectbr</tt>' instruction implements an indirect branch to a label
3045 within the current function, whose address is specified by
3046 "<tt>address</tt>". Address must be derived from a <a
3047 href="#blockaddress">blockaddress</a> constant.</p>
3051 <p>The '<tt>address</tt>' argument is the address of the label to jump to. The
3052 rest of the arguments indicate the full set of possible destinations that the
3053 address may point to. Blocks are allowed to occur multiple times in the
3054 destination list, though this isn't particularly useful.</p>
3056 <p>This destination list is required so that dataflow analysis has an accurate
3057 understanding of the CFG.</p>
3061 <p>Control transfers to the block specified in the address argument. All
3062 possible destination blocks must be listed in the label list, otherwise this
3063 instruction has undefined behavior. This implies that jumps to labels
3064 defined in other functions have undefined behavior as well.</p>
3066 <h5>Implementation:</h5>
3068 <p>This is typically implemented with a jump through a register.</p>
3072 indirectbr i8* %Addr, [ label %bb1, label %bb2, label %bb3 ]
3078 <!-- _______________________________________________________________________ -->
3079 <div class="doc_subsubsection">
3080 <a name="i_invoke">'<tt>invoke</tt>' Instruction</a>
3083 <div class="doc_text">
3087 <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>]
3088 to label <normal label> unwind label <exception label>
3092 <p>The '<tt>invoke</tt>' instruction causes control to transfer to a specified
3093 function, with the possibility of control flow transfer to either the
3094 '<tt>normal</tt>' label or the '<tt>exception</tt>' label. If the callee
3095 function returns with the "<tt><a href="#i_ret">ret</a></tt>" instruction,
3096 control flow will return to the "normal" label. If the callee (or any
3097 indirect callees) returns with the "<a href="#i_unwind"><tt>unwind</tt></a>"
3098 instruction, control is interrupted and continued at the dynamically nearest
3099 "exception" label.</p>
3102 <p>This instruction requires several arguments:</p>
3105 <li>The optional "cconv" marker indicates which <a href="#callingconv">calling
3106 convention</a> the call should use. If none is specified, the call
3107 defaults to using C calling conventions.</li>
3109 <li>The optional <a href="#paramattrs">Parameter Attributes</a> list for
3110 return values. Only '<tt>zeroext</tt>', '<tt>signext</tt>', and
3111 '<tt>inreg</tt>' attributes are valid here.</li>
3113 <li>'<tt>ptr to function ty</tt>': shall be the signature of the pointer to
3114 function value being invoked. In most cases, this is a direct function
3115 invocation, but indirect <tt>invoke</tt>s are just as possible, branching
3116 off an arbitrary pointer to function value.</li>
3118 <li>'<tt>function ptr val</tt>': An LLVM value containing a pointer to a
3119 function to be invoked. </li>
3121 <li>'<tt>function args</tt>': argument list whose types match the function
3122 signature argument types and parameter attributes. All arguments must be
3123 of <a href="#t_firstclass">first class</a> type. If the function
3124 signature indicates the function accepts a variable number of arguments,
3125 the extra arguments can be specified.</li>
3127 <li>'<tt>normal label</tt>': the label reached when the called function
3128 executes a '<tt><a href="#i_ret">ret</a></tt>' instruction. </li>
3130 <li>'<tt>exception label</tt>': the label reached when a callee returns with
3131 the <a href="#i_unwind"><tt>unwind</tt></a> instruction. </li>
3133 <li>The optional <a href="#fnattrs">function attributes</a> list. Only
3134 '<tt>noreturn</tt>', '<tt>nounwind</tt>', '<tt>readonly</tt>' and
3135 '<tt>readnone</tt>' attributes are valid here.</li>
3139 <p>This instruction is designed to operate as a standard
3140 '<tt><a href="#i_call">call</a></tt>' instruction in most regards. The
3141 primary difference is that it establishes an association with a label, which
3142 is used by the runtime library to unwind the stack.</p>
3144 <p>This instruction is used in languages with destructors to ensure that proper
3145 cleanup is performed in the case of either a <tt>longjmp</tt> or a thrown
3146 exception. Additionally, this is important for implementation of
3147 '<tt>catch</tt>' clauses in high-level languages that support them.</p>
3149 <p>For the purposes of the SSA form, the definition of the value returned by the
3150 '<tt>invoke</tt>' instruction is deemed to occur on the edge from the current
3151 block to the "normal" label. If the callee unwinds then no return value is
3154 <p>Note that the code generator does not yet completely support unwind, and
3155 that the invoke/unwind semantics are likely to change in future versions.</p>
3159 %retval = invoke i32 @Test(i32 15) to label %Continue
3160 unwind label %TestCleanup <i>; {i32}:retval set</i>
3161 %retval = invoke <a href="#callingconv">coldcc</a> i32 %Testfnptr(i32 15) to label %Continue
3162 unwind label %TestCleanup <i>; {i32}:retval set</i>
3167 <!-- _______________________________________________________________________ -->
3169 <div class="doc_subsubsection"> <a name="i_unwind">'<tt>unwind</tt>'
3170 Instruction</a> </div>
3172 <div class="doc_text">
3180 <p>The '<tt>unwind</tt>' instruction unwinds the stack, continuing control flow
3181 at the first callee in the dynamic call stack which used
3182 an <a href="#i_invoke"><tt>invoke</tt></a> instruction to perform the call.
3183 This is primarily used to implement exception handling.</p>
3186 <p>The '<tt>unwind</tt>' instruction causes execution of the current function to
3187 immediately halt. The dynamic call stack is then searched for the
3188 first <a href="#i_invoke"><tt>invoke</tt></a> instruction on the call stack.
3189 Once found, execution continues at the "exceptional" destination block
3190 specified by the <tt>invoke</tt> instruction. If there is no <tt>invoke</tt>
3191 instruction in the dynamic call chain, undefined behavior results.</p>
3193 <p>Note that the code generator does not yet completely support unwind, and
3194 that the invoke/unwind semantics are likely to change in future versions.</p>
3198 <!-- _______________________________________________________________________ -->
3200 <div class="doc_subsubsection"> <a name="i_unreachable">'<tt>unreachable</tt>'
3201 Instruction</a> </div>
3203 <div class="doc_text">
3211 <p>The '<tt>unreachable</tt>' instruction has no defined semantics. This
3212 instruction is used to inform the optimizer that a particular portion of the
3213 code is not reachable. This can be used to indicate that the code after a
3214 no-return function cannot be reached, and other facts.</p>
3217 <p>The '<tt>unreachable</tt>' instruction has no defined semantics.</p>
3221 <!-- ======================================================================= -->
3222 <div class="doc_subsection"> <a name="binaryops">Binary Operations</a> </div>
3224 <div class="doc_text">
3226 <p>Binary operators are used to do most of the computation in a program. They
3227 require two operands of the same type, execute an operation on them, and
3228 produce a single value. The operands might represent multiple data, as is
3229 the case with the <a href="#t_vector">vector</a> data type. The result value
3230 has the same type as its operands.</p>
3232 <p>There are several different binary operators:</p>
3236 <!-- _______________________________________________________________________ -->
3237 <div class="doc_subsubsection">
3238 <a name="i_add">'<tt>add</tt>' Instruction</a>
3241 <div class="doc_text">
3245 <result> = add <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3246 <result> = add nuw <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3247 <result> = add nsw <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3248 <result> = add nuw nsw <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3252 <p>The '<tt>add</tt>' instruction returns the sum of its two operands.</p>
3255 <p>The two arguments to the '<tt>add</tt>' instruction must
3256 be <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of
3257 integer values. Both arguments must have identical types.</p>
3260 <p>The value produced is the integer sum of the two operands.</p>
3262 <p>If the sum has unsigned overflow, the result returned is the mathematical
3263 result modulo 2<sup>n</sup>, where n is the bit width of the result.</p>
3265 <p>Because LLVM integers use a two's complement representation, this instruction
3266 is appropriate for both signed and unsigned integers.</p>
3268 <p><tt>nuw</tt> and <tt>nsw</tt> stand for "No Unsigned Wrap"
3269 and "No Signed Wrap", respectively. If the <tt>nuw</tt> and/or
3270 <tt>nsw</tt> keywords are present, the result value of the <tt>add</tt>
3271 is a <a href="#trapvalues">trap value</a> if unsigned and/or signed overflow,
3272 respectively, occurs.</p>
3276 <result> = add i32 4, %var <i>; yields {i32}:result = 4 + %var</i>
3281 <!-- _______________________________________________________________________ -->
3282 <div class="doc_subsubsection">
3283 <a name="i_fadd">'<tt>fadd</tt>' Instruction</a>
3286 <div class="doc_text">
3290 <result> = fadd <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3294 <p>The '<tt>fadd</tt>' instruction returns the sum of its two operands.</p>
3297 <p>The two arguments to the '<tt>fadd</tt>' instruction must be
3298 <a href="#t_floating">floating point</a> or <a href="#t_vector">vector</a> of
3299 floating point values. Both arguments must have identical types.</p>
3302 <p>The value produced is the floating point sum of the two operands.</p>
3306 <result> = fadd float 4.0, %var <i>; yields {float}:result = 4.0 + %var</i>
3311 <!-- _______________________________________________________________________ -->
3312 <div class="doc_subsubsection">
3313 <a name="i_sub">'<tt>sub</tt>' Instruction</a>
3316 <div class="doc_text">
3320 <result> = sub <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3321 <result> = sub nuw <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3322 <result> = sub nsw <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3323 <result> = sub nuw nsw <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3327 <p>The '<tt>sub</tt>' instruction returns the difference of its two
3330 <p>Note that the '<tt>sub</tt>' instruction is used to represent the
3331 '<tt>neg</tt>' instruction present in most other intermediate
3332 representations.</p>
3335 <p>The two arguments to the '<tt>sub</tt>' instruction must
3336 be <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of
3337 integer values. Both arguments must have identical types.</p>
3340 <p>The value produced is the integer difference of the two operands.</p>
3342 <p>If the difference has unsigned overflow, the result returned is the
3343 mathematical result modulo 2<sup>n</sup>, where n is the bit width of the
3346 <p>Because LLVM integers use a two's complement representation, this instruction
3347 is appropriate for both signed and unsigned integers.</p>
3349 <p><tt>nuw</tt> and <tt>nsw</tt> stand for "No Unsigned Wrap"
3350 and "No Signed Wrap", respectively. If the <tt>nuw</tt> and/or
3351 <tt>nsw</tt> keywords are present, the result value of the <tt>sub</tt>
3352 is a <a href="#trapvalues">trap value</a> if unsigned and/or signed overflow,
3353 respectively, occurs.</p>
3357 <result> = sub i32 4, %var <i>; yields {i32}:result = 4 - %var</i>
3358 <result> = sub i32 0, %val <i>; yields {i32}:result = -%var</i>
3363 <!-- _______________________________________________________________________ -->
3364 <div class="doc_subsubsection">
3365 <a name="i_fsub">'<tt>fsub</tt>' Instruction</a>
3368 <div class="doc_text">
3372 <result> = fsub <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3376 <p>The '<tt>fsub</tt>' instruction returns the difference of its two
3379 <p>Note that the '<tt>fsub</tt>' instruction is used to represent the
3380 '<tt>fneg</tt>' instruction present in most other intermediate
3381 representations.</p>
3384 <p>The two arguments to the '<tt>fsub</tt>' instruction must be
3385 <a href="#t_floating">floating point</a> or <a href="#t_vector">vector</a> of
3386 floating point values. Both arguments must have identical types.</p>
3389 <p>The value produced is the floating point difference of the two operands.</p>
3393 <result> = fsub float 4.0, %var <i>; yields {float}:result = 4.0 - %var</i>
3394 <result> = fsub float -0.0, %val <i>; yields {float}:result = -%var</i>
3399 <!-- _______________________________________________________________________ -->
3400 <div class="doc_subsubsection">
3401 <a name="i_mul">'<tt>mul</tt>' Instruction</a>
3404 <div class="doc_text">
3408 <result> = mul <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3409 <result> = mul nuw <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3410 <result> = mul nsw <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3411 <result> = mul nuw nsw <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3415 <p>The '<tt>mul</tt>' instruction returns the product of its two operands.</p>
3418 <p>The two arguments to the '<tt>mul</tt>' instruction must
3419 be <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of
3420 integer values. Both arguments must have identical types.</p>
3423 <p>The value produced is the integer product of the two operands.</p>
3425 <p>If the result of the multiplication has unsigned overflow, the result
3426 returned is the mathematical result modulo 2<sup>n</sup>, where n is the bit
3427 width of the result.</p>
3429 <p>Because LLVM integers use a two's complement representation, and the result
3430 is the same width as the operands, this instruction returns the correct
3431 result for both signed and unsigned integers. If a full product
3432 (e.g. <tt>i32</tt>x<tt>i32</tt>-><tt>i64</tt>) is needed, the operands should
3433 be sign-extended or zero-extended as appropriate to the width of the full
3436 <p><tt>nuw</tt> and <tt>nsw</tt> stand for "No Unsigned Wrap"
3437 and "No Signed Wrap", respectively. If the <tt>nuw</tt> and/or
3438 <tt>nsw</tt> keywords are present, the result value of the <tt>mul</tt>
3439 is a <a href="#trapvalues">trap value</a> if unsigned and/or signed overflow,
3440 respectively, occurs.</p>
3444 <result> = mul i32 4, %var <i>; yields {i32}:result = 4 * %var</i>
3449 <!-- _______________________________________________________________________ -->
3450 <div class="doc_subsubsection">
3451 <a name="i_fmul">'<tt>fmul</tt>' Instruction</a>
3454 <div class="doc_text">
3458 <result> = fmul <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3462 <p>The '<tt>fmul</tt>' instruction returns the product of its two operands.</p>
3465 <p>The two arguments to the '<tt>fmul</tt>' instruction must be
3466 <a href="#t_floating">floating point</a> or <a href="#t_vector">vector</a> of
3467 floating point values. Both arguments must have identical types.</p>
3470 <p>The value produced is the floating point product of the two operands.</p>
3474 <result> = fmul float 4.0, %var <i>; yields {float}:result = 4.0 * %var</i>
3479 <!-- _______________________________________________________________________ -->
3480 <div class="doc_subsubsection"> <a name="i_udiv">'<tt>udiv</tt>' Instruction
3483 <div class="doc_text">
3487 <result> = udiv <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3491 <p>The '<tt>udiv</tt>' instruction returns the quotient of its two operands.</p>
3494 <p>The two arguments to the '<tt>udiv</tt>' instruction must be
3495 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
3496 values. Both arguments must have identical types.</p>
3499 <p>The value produced is the unsigned integer quotient of the two operands.</p>
3501 <p>Note that unsigned integer division and signed integer division are distinct
3502 operations; for signed integer division, use '<tt>sdiv</tt>'.</p>
3504 <p>Division by zero leads to undefined behavior.</p>
3508 <result> = udiv i32 4, %var <i>; yields {i32}:result = 4 / %var</i>
3513 <!-- _______________________________________________________________________ -->
3514 <div class="doc_subsubsection"> <a name="i_sdiv">'<tt>sdiv</tt>' Instruction
3517 <div class="doc_text">
3521 <result> = sdiv <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3522 <result> = sdiv exact <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3526 <p>The '<tt>sdiv</tt>' instruction returns the quotient of its two operands.</p>
3529 <p>The two arguments to the '<tt>sdiv</tt>' instruction must be
3530 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
3531 values. Both arguments must have identical types.</p>
3534 <p>The value produced is the signed integer quotient of the two operands rounded
3537 <p>Note that signed integer division and unsigned integer division are distinct
3538 operations; for unsigned integer division, use '<tt>udiv</tt>'.</p>
3540 <p>Division by zero leads to undefined behavior. Overflow also leads to
3541 undefined behavior; this is a rare case, but can occur, for example, by doing
3542 a 32-bit division of -2147483648 by -1.</p>
3544 <p>If the <tt>exact</tt> keyword is present, the result value of the
3545 <tt>sdiv</tt> is a <a href="#trapvalues">trap value</a> if the result would
3546 be rounded or if overflow would occur.</p>
3550 <result> = sdiv i32 4, %var <i>; yields {i32}:result = 4 / %var</i>
3555 <!-- _______________________________________________________________________ -->
3556 <div class="doc_subsubsection"> <a name="i_fdiv">'<tt>fdiv</tt>'
3557 Instruction</a> </div>
3559 <div class="doc_text">
3563 <result> = fdiv <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3567 <p>The '<tt>fdiv</tt>' instruction returns the quotient of its two operands.</p>
3570 <p>The two arguments to the '<tt>fdiv</tt>' instruction must be
3571 <a href="#t_floating">floating point</a> or <a href="#t_vector">vector</a> of
3572 floating point values. Both arguments must have identical types.</p>
3575 <p>The value produced is the floating point quotient of the two operands.</p>
3579 <result> = fdiv float 4.0, %var <i>; yields {float}:result = 4.0 / %var</i>
3584 <!-- _______________________________________________________________________ -->
3585 <div class="doc_subsubsection"> <a name="i_urem">'<tt>urem</tt>' Instruction</a>
3588 <div class="doc_text">
3592 <result> = urem <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3596 <p>The '<tt>urem</tt>' instruction returns the remainder from the unsigned
3597 division of its two arguments.</p>
3600 <p>The two arguments to the '<tt>urem</tt>' instruction must be
3601 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
3602 values. Both arguments must have identical types.</p>
3605 <p>This instruction returns the unsigned integer <i>remainder</i> of a division.
3606 This instruction always performs an unsigned division to get the
3609 <p>Note that unsigned integer remainder and signed integer remainder are
3610 distinct operations; for signed integer remainder, use '<tt>srem</tt>'.</p>
3612 <p>Taking the remainder of a division by zero leads to undefined behavior.</p>
3616 <result> = urem i32 4, %var <i>; yields {i32}:result = 4 % %var</i>
3621 <!-- _______________________________________________________________________ -->
3622 <div class="doc_subsubsection">
3623 <a name="i_srem">'<tt>srem</tt>' Instruction</a>
3626 <div class="doc_text">
3630 <result> = srem <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3634 <p>The '<tt>srem</tt>' instruction returns the remainder from the signed
3635 division of its two operands. This instruction can also take
3636 <a href="#t_vector">vector</a> versions of the values in which case the
3637 elements must be integers.</p>
3640 <p>The two arguments to the '<tt>srem</tt>' instruction must be
3641 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
3642 values. Both arguments must have identical types.</p>
3645 <p>This instruction returns the <i>remainder</i> of a division (where the result
3646 has the same sign as the dividend, <tt>op1</tt>), not the <i>modulo</i>
3647 operator (where the result has the same sign as the divisor, <tt>op2</tt>) of
3648 a value. For more information about the difference,
3649 see <a href="http://mathforum.org/dr.math/problems/anne.4.28.99.html">The
3650 Math Forum</a>. For a table of how this is implemented in various languages,
3651 please see <a href="http://en.wikipedia.org/wiki/Modulo_operation">
3652 Wikipedia: modulo operation</a>.</p>
3654 <p>Note that signed integer remainder and unsigned integer remainder are
3655 distinct operations; for unsigned integer remainder, use '<tt>urem</tt>'.</p>
3657 <p>Taking the remainder of a division by zero leads to undefined behavior.
3658 Overflow also leads to undefined behavior; this is a rare case, but can
3659 occur, for example, by taking the remainder of a 32-bit division of
3660 -2147483648 by -1. (The remainder doesn't actually overflow, but this rule
3661 lets srem be implemented using instructions that return both the result of
3662 the division and the remainder.)</p>
3666 <result> = srem i32 4, %var <i>; yields {i32}:result = 4 % %var</i>
3671 <!-- _______________________________________________________________________ -->
3672 <div class="doc_subsubsection">
3673 <a name="i_frem">'<tt>frem</tt>' Instruction</a> </div>
3675 <div class="doc_text">
3679 <result> = frem <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3683 <p>The '<tt>frem</tt>' instruction returns the remainder from the division of
3684 its two operands.</p>
3687 <p>The two arguments to the '<tt>frem</tt>' instruction must be
3688 <a href="#t_floating">floating point</a> or <a href="#t_vector">vector</a> of
3689 floating point values. Both arguments must have identical types.</p>
3692 <p>This instruction returns the <i>remainder</i> of a division. The remainder
3693 has the same sign as the dividend.</p>
3697 <result> = frem float 4.0, %var <i>; yields {float}:result = 4.0 % %var</i>
3702 <!-- ======================================================================= -->
3703 <div class="doc_subsection"> <a name="bitwiseops">Bitwise Binary
3704 Operations</a> </div>
3706 <div class="doc_text">
3708 <p>Bitwise binary operators are used to do various forms of bit-twiddling in a
3709 program. They are generally very efficient instructions and can commonly be
3710 strength reduced from other instructions. They require two operands of the
3711 same type, execute an operation on them, and produce a single value. The
3712 resulting value is the same type as its operands.</p>
3716 <!-- _______________________________________________________________________ -->
3717 <div class="doc_subsubsection"> <a name="i_shl">'<tt>shl</tt>'
3718 Instruction</a> </div>
3720 <div class="doc_text">
3724 <result> = shl <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3728 <p>The '<tt>shl</tt>' instruction returns the first operand shifted to the left
3729 a specified number of bits.</p>
3732 <p>Both arguments to the '<tt>shl</tt>' instruction must be the
3733 same <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of
3734 integer type. '<tt>op2</tt>' is treated as an unsigned value.</p>
3737 <p>The value produced is <tt>op1</tt> * 2<sup><tt>op2</tt></sup> mod
3738 2<sup>n</sup>, where <tt>n</tt> is the width of the result. If <tt>op2</tt>
3739 is (statically or dynamically) negative or equal to or larger than the number
3740 of bits in <tt>op1</tt>, the result is undefined. If the arguments are
3741 vectors, each vector element of <tt>op1</tt> is shifted by the corresponding
3742 shift amount in <tt>op2</tt>.</p>
3746 <result> = shl i32 4, %var <i>; yields {i32}: 4 << %var</i>
3747 <result> = shl i32 4, 2 <i>; yields {i32}: 16</i>
3748 <result> = shl i32 1, 10 <i>; yields {i32}: 1024</i>
3749 <result> = shl i32 1, 32 <i>; undefined</i>
3750 <result> = shl <2 x i32> < i32 1, i32 1>, < i32 1, i32 2> <i>; yields: result=<2 x i32> < i32 2, i32 4></i>
3755 <!-- _______________________________________________________________________ -->
3756 <div class="doc_subsubsection"> <a name="i_lshr">'<tt>lshr</tt>'
3757 Instruction</a> </div>
3759 <div class="doc_text">
3763 <result> = lshr <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3767 <p>The '<tt>lshr</tt>' instruction (logical shift right) returns the first
3768 operand shifted to the right a specified number of bits with zero fill.</p>
3771 <p>Both arguments to the '<tt>lshr</tt>' instruction must be the same
3772 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
3773 type. '<tt>op2</tt>' is treated as an unsigned value.</p>
3776 <p>This instruction always performs a logical shift right operation. The most
3777 significant bits of the result will be filled with zero bits after the shift.
3778 If <tt>op2</tt> is (statically or dynamically) equal to or larger than the
3779 number of bits in <tt>op1</tt>, the result is undefined. If the arguments are
3780 vectors, each vector element of <tt>op1</tt> is shifted by the corresponding
3781 shift amount in <tt>op2</tt>.</p>
3785 <result> = lshr i32 4, 1 <i>; yields {i32}:result = 2</i>
3786 <result> = lshr i32 4, 2 <i>; yields {i32}:result = 1</i>
3787 <result> = lshr i8 4, 3 <i>; yields {i8}:result = 0</i>
3788 <result> = lshr i8 -2, 1 <i>; yields {i8}:result = 0x7FFFFFFF </i>
3789 <result> = lshr i32 1, 32 <i>; undefined</i>
3790 <result> = lshr <2 x i32> < i32 -2, i32 4>, < i32 1, i32 2> <i>; yields: result=<2 x i32> < i32 0x7FFFFFFF, i32 1></i>
3795 <!-- _______________________________________________________________________ -->
3796 <div class="doc_subsubsection"> <a name="i_ashr">'<tt>ashr</tt>'
3797 Instruction</a> </div>
3798 <div class="doc_text">
3802 <result> = ashr <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3806 <p>The '<tt>ashr</tt>' instruction (arithmetic shift right) returns the first
3807 operand shifted to the right a specified number of bits with sign
3811 <p>Both arguments to the '<tt>ashr</tt>' instruction must be the same
3812 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
3813 type. '<tt>op2</tt>' is treated as an unsigned value.</p>
3816 <p>This instruction always performs an arithmetic shift right operation, The
3817 most significant bits of the result will be filled with the sign bit
3818 of <tt>op1</tt>. If <tt>op2</tt> is (statically or dynamically) equal to or
3819 larger than the number of bits in <tt>op1</tt>, the result is undefined. If
3820 the arguments are vectors, each vector element of <tt>op1</tt> is shifted by
3821 the corresponding shift amount in <tt>op2</tt>.</p>
3825 <result> = ashr i32 4, 1 <i>; yields {i32}:result = 2</i>
3826 <result> = ashr i32 4, 2 <i>; yields {i32}:result = 1</i>
3827 <result> = ashr i8 4, 3 <i>; yields {i8}:result = 0</i>
3828 <result> = ashr i8 -2, 1 <i>; yields {i8}:result = -1</i>
3829 <result> = ashr i32 1, 32 <i>; undefined</i>
3830 <result> = ashr <2 x i32> < i32 -2, i32 4>, < i32 1, i32 3> <i>; yields: result=<2 x i32> < i32 -1, i32 0></i>
3835 <!-- _______________________________________________________________________ -->
3836 <div class="doc_subsubsection"> <a name="i_and">'<tt>and</tt>'
3837 Instruction</a> </div>
3839 <div class="doc_text">
3843 <result> = and <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3847 <p>The '<tt>and</tt>' instruction returns the bitwise logical and of its two
3851 <p>The two arguments to the '<tt>and</tt>' instruction must be
3852 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
3853 values. Both arguments must have identical types.</p>
3856 <p>The truth table used for the '<tt>and</tt>' instruction is:</p>
3858 <table border="1" cellspacing="0" cellpadding="4">
3890 <result> = and i32 4, %var <i>; yields {i32}:result = 4 & %var</i>
3891 <result> = and i32 15, 40 <i>; yields {i32}:result = 8</i>
3892 <result> = and i32 4, 8 <i>; yields {i32}:result = 0</i>
3895 <!-- _______________________________________________________________________ -->
3896 <div class="doc_subsubsection"> <a name="i_or">'<tt>or</tt>' Instruction</a> </div>
3898 <div class="doc_text">
3902 <result> = or <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3906 <p>The '<tt>or</tt>' instruction returns the bitwise logical inclusive or of its
3910 <p>The two arguments to the '<tt>or</tt>' instruction must be
3911 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
3912 values. Both arguments must have identical types.</p>
3915 <p>The truth table used for the '<tt>or</tt>' instruction is:</p>
3917 <table border="1" cellspacing="0" cellpadding="4">
3949 <result> = or i32 4, %var <i>; yields {i32}:result = 4 | %var</i>
3950 <result> = or i32 15, 40 <i>; yields {i32}:result = 47</i>
3951 <result> = or i32 4, 8 <i>; yields {i32}:result = 12</i>
3956 <!-- _______________________________________________________________________ -->
3957 <div class="doc_subsubsection"> <a name="i_xor">'<tt>xor</tt>'
3958 Instruction</a> </div>
3960 <div class="doc_text">
3964 <result> = xor <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3968 <p>The '<tt>xor</tt>' instruction returns the bitwise logical exclusive or of
3969 its two operands. The <tt>xor</tt> is used to implement the "one's
3970 complement" operation, which is the "~" operator in C.</p>
3973 <p>The two arguments to the '<tt>xor</tt>' instruction must be
3974 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
3975 values. Both arguments must have identical types.</p>
3978 <p>The truth table used for the '<tt>xor</tt>' instruction is:</p>
3980 <table border="1" cellspacing="0" cellpadding="4">
4012 <result> = xor i32 4, %var <i>; yields {i32}:result = 4 ^ %var</i>
4013 <result> = xor i32 15, 40 <i>; yields {i32}:result = 39</i>
4014 <result> = xor i32 4, 8 <i>; yields {i32}:result = 12</i>
4015 <result> = xor i32 %V, -1 <i>; yields {i32}:result = ~%V</i>
4020 <!-- ======================================================================= -->
4021 <div class="doc_subsection">
4022 <a name="vectorops">Vector Operations</a>
4025 <div class="doc_text">
4027 <p>LLVM supports several instructions to represent vector operations in a
4028 target-independent manner. These instructions cover the element-access and
4029 vector-specific operations needed to process vectors effectively. While LLVM
4030 does directly support these vector operations, many sophisticated algorithms
4031 will want to use target-specific intrinsics to take full advantage of a
4032 specific target.</p>
4036 <!-- _______________________________________________________________________ -->
4037 <div class="doc_subsubsection">
4038 <a name="i_extractelement">'<tt>extractelement</tt>' Instruction</a>
4041 <div class="doc_text">
4045 <result> = extractelement <n x <ty>> <val>, i32 <idx> <i>; yields <ty></i>
4049 <p>The '<tt>extractelement</tt>' instruction extracts a single scalar element
4050 from a vector at a specified index.</p>
4054 <p>The first operand of an '<tt>extractelement</tt>' instruction is a value
4055 of <a href="#t_vector">vector</a> type. The second operand is an index
4056 indicating the position from which to extract the element. The index may be
4060 <p>The result is a scalar of the same type as the element type of
4061 <tt>val</tt>. Its value is the value at position <tt>idx</tt> of
4062 <tt>val</tt>. If <tt>idx</tt> exceeds the length of <tt>val</tt>, the
4063 results are undefined.</p>
4067 <result> = extractelement <4 x i32> %vec, i32 0 <i>; yields i32</i>
4072 <!-- _______________________________________________________________________ -->
4073 <div class="doc_subsubsection">
4074 <a name="i_insertelement">'<tt>insertelement</tt>' Instruction</a>
4077 <div class="doc_text">
4081 <result> = insertelement <n x <ty>> <val>, <ty> <elt>, i32 <idx> <i>; yields <n x <ty>></i>
4085 <p>The '<tt>insertelement</tt>' instruction inserts a scalar element into a
4086 vector at a specified index.</p>
4089 <p>The first operand of an '<tt>insertelement</tt>' instruction is a value
4090 of <a href="#t_vector">vector</a> type. The second operand is a scalar value
4091 whose type must equal the element type of the first operand. The third
4092 operand is an index indicating the position at which to insert the value.
4093 The index may be a variable.</p>
4096 <p>The result is a vector of the same type as <tt>val</tt>. Its element values
4097 are those of <tt>val</tt> except at position <tt>idx</tt>, where it gets the
4098 value <tt>elt</tt>. If <tt>idx</tt> exceeds the length of <tt>val</tt>, the
4099 results are undefined.</p>
4103 <result> = insertelement <4 x i32> %vec, i32 1, i32 0 <i>; yields <4 x i32></i>
4108 <!-- _______________________________________________________________________ -->
4109 <div class="doc_subsubsection">
4110 <a name="i_shufflevector">'<tt>shufflevector</tt>' Instruction</a>
4113 <div class="doc_text">
4117 <result> = shufflevector <n x <ty>> <v1>, <n x <ty>> <v2>, <m x i32> <mask> <i>; yields <m x <ty>></i>
4121 <p>The '<tt>shufflevector</tt>' instruction constructs a permutation of elements
4122 from two input vectors, returning a vector with the same element type as the
4123 input and length that is the same as the shuffle mask.</p>
4126 <p>The first two operands of a '<tt>shufflevector</tt>' instruction are vectors
4127 with types that match each other. The third argument is a shuffle mask whose
4128 element type is always 'i32'. The result of the instruction is a vector
4129 whose length is the same as the shuffle mask and whose element type is the
4130 same as the element type of the first two operands.</p>
4132 <p>The shuffle mask operand is required to be a constant vector with either
4133 constant integer or undef values.</p>
4136 <p>The elements of the two input vectors are numbered from left to right across
4137 both of the vectors. The shuffle mask operand specifies, for each element of
4138 the result vector, which element of the two input vectors the result element
4139 gets. The element selector may be undef (meaning "don't care") and the
4140 second operand may be undef if performing a shuffle from only one vector.</p>
4144 <result> = shufflevector <4 x i32> %v1, <4 x i32> %v2,
4145 <4 x i32> <i32 0, i32 4, i32 1, i32 5> <i>; yields <4 x i32></i>
4146 <result> = shufflevector <4 x i32> %v1, <4 x i32> undef,
4147 <4 x i32> <i32 0, i32 1, i32 2, i32 3> <i>; yields <4 x i32></i> - Identity shuffle.
4148 <result> = shufflevector <8 x i32> %v1, <8 x i32> undef,
4149 <4 x i32> <i32 0, i32 1, i32 2, i32 3> <i>; yields <4 x i32></i>
4150 <result> = shufflevector <4 x i32> %v1, <4 x i32> %v2,
4151 <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>
4156 <!-- ======================================================================= -->
4157 <div class="doc_subsection">
4158 <a name="aggregateops">Aggregate Operations</a>
4161 <div class="doc_text">
4163 <p>LLVM supports several instructions for working with
4164 <a href="#t_aggregate">aggregate</a> values.</p>
4168 <!-- _______________________________________________________________________ -->
4169 <div class="doc_subsubsection">
4170 <a name="i_extractvalue">'<tt>extractvalue</tt>' Instruction</a>
4173 <div class="doc_text">
4177 <result> = extractvalue <aggregate type> <val>, <idx>{, <idx>}*
4181 <p>The '<tt>extractvalue</tt>' instruction extracts the value of a member field
4182 from an <a href="#t_aggregate">aggregate</a> value.</p>
4185 <p>The first operand of an '<tt>extractvalue</tt>' instruction is a value
4186 of <a href="#t_struct">struct</a>, <a href="#t_union">union</a> or
4187 <a href="#t_array">array</a> type. The operands are constant indices to
4188 specify which value to extract in a similar manner as indices in a
4189 '<tt><a href="#i_getelementptr">getelementptr</a></tt>' instruction.</p>
4192 <p>The result is the value at the position in the aggregate specified by the
4197 <result> = extractvalue {i32, float} %agg, 0 <i>; yields i32</i>
4202 <!-- _______________________________________________________________________ -->
4203 <div class="doc_subsubsection">
4204 <a name="i_insertvalue">'<tt>insertvalue</tt>' Instruction</a>
4207 <div class="doc_text">
4211 <result> = insertvalue <aggregate type> <val>, <ty> <elt>, <idx> <i>; yields <aggregate type></i>
4215 <p>The '<tt>insertvalue</tt>' instruction inserts a value into a member field
4216 in an <a href="#t_aggregate">aggregate</a> value.</p>
4219 <p>The first operand of an '<tt>insertvalue</tt>' instruction is a value
4220 of <a href="#t_struct">struct</a>, <a href="#t_union">union</a> or
4221 <a href="#t_array">array</a> type. The second operand is a first-class
4222 value to insert. The following operands are constant indices indicating
4223 the position at which to insert the value in a similar manner as indices in a
4224 '<tt><a href="#i_getelementptr">getelementptr</a></tt>' instruction. The
4225 value to insert must have the same type as the value identified by the
4229 <p>The result is an aggregate of the same type as <tt>val</tt>. Its value is
4230 that of <tt>val</tt> except that the value at the position specified by the
4231 indices is that of <tt>elt</tt>.</p>
4235 %agg1 = insertvalue {i32, float} undef, i32 1, 0 <i>; yields {i32 1, float undef}</i>
4236 %agg2 = insertvalue {i32, float} %agg1, float %val, 1 <i>; yields {i32 1, float %val}</i>
4242 <!-- ======================================================================= -->
4243 <div class="doc_subsection">
4244 <a name="memoryops">Memory Access and Addressing Operations</a>
4247 <div class="doc_text">
4249 <p>A key design point of an SSA-based representation is how it represents
4250 memory. In LLVM, no memory locations are in SSA form, which makes things
4251 very simple. This section describes how to read, write, and allocate
4256 <!-- _______________________________________________________________________ -->
4257 <div class="doc_subsubsection">
4258 <a name="i_alloca">'<tt>alloca</tt>' Instruction</a>
4261 <div class="doc_text">
4265 <result> = alloca <type>[, <ty> <NumElements>][, align <alignment>] <i>; yields {type*}:result</i>
4269 <p>The '<tt>alloca</tt>' instruction allocates memory on the stack frame of the
4270 currently executing function, to be automatically released when this function
4271 returns to its caller. The object is always allocated in the generic address
4272 space (address space zero).</p>
4275 <p>The '<tt>alloca</tt>' instruction
4276 allocates <tt>sizeof(<type>)*NumElements</tt> bytes of memory on the
4277 runtime stack, returning a pointer of the appropriate type to the program.
4278 If "NumElements" is specified, it is the number of elements allocated,
4279 otherwise "NumElements" is defaulted to be one. If a constant alignment is
4280 specified, the value result of the allocation is guaranteed to be aligned to
4281 at least that boundary. If not specified, or if zero, the target can choose
4282 to align the allocation on any convenient boundary compatible with the
4285 <p>'<tt>type</tt>' may be any sized type.</p>
4288 <p>Memory is allocated; a pointer is returned. The operation is undefined if
4289 there is insufficient stack space for the allocation. '<tt>alloca</tt>'d
4290 memory is automatically released when the function returns. The
4291 '<tt>alloca</tt>' instruction is commonly used to represent automatic
4292 variables that must have an address available. When the function returns
4293 (either with the <tt><a href="#i_ret">ret</a></tt>
4294 or <tt><a href="#i_unwind">unwind</a></tt> instructions), the memory is
4295 reclaimed. Allocating zero bytes is legal, but the result is undefined.</p>
4299 %ptr = alloca i32 <i>; yields {i32*}:ptr</i>
4300 %ptr = alloca i32, i32 4 <i>; yields {i32*}:ptr</i>
4301 %ptr = alloca i32, i32 4, align 1024 <i>; yields {i32*}:ptr</i>
4302 %ptr = alloca i32, align 1024 <i>; yields {i32*}:ptr</i>
4307 <!-- _______________________________________________________________________ -->
4308 <div class="doc_subsubsection"> <a name="i_load">'<tt>load</tt>'
4309 Instruction</a> </div>
4311 <div class="doc_text">
4315 <result> = load <ty>* <pointer>[, align <alignment>][, !nontemporal !<index>]
4316 <result> = volatile load <ty>* <pointer>[, align <alignment>][, !nontemporal !<index>]
4317 !<index> = !{ i32 1 }
4321 <p>The '<tt>load</tt>' instruction is used to read from memory.</p>
4324 <p>The argument to the '<tt>load</tt>' instruction specifies the memory address
4325 from which to load. The pointer must point to
4326 a <a href="#t_firstclass">first class</a> type. If the <tt>load</tt> is
4327 marked as <tt>volatile</tt>, then the optimizer is not allowed to modify the
4328 number or order of execution of this <tt>load</tt> with other <a
4329 href="#volatile">volatile operations</a>.</p>
4331 <p>The optional constant <tt>align</tt> argument specifies the alignment of the
4332 operation (that is, the alignment of the memory address). A value of 0 or an
4333 omitted <tt>align</tt> argument means that the operation has the preferential
4334 alignment for the target. It is the responsibility of the code emitter to
4335 ensure that the alignment information is correct. Overestimating the
4336 alignment results in undefined behavior. Underestimating the alignment may
4337 produce less efficient code. An alignment of 1 is always safe.</p>
4339 <p>The optional <tt>!nontemporal</tt> metadata must reference a single
4340 metatadata name <index> corresponding to a metadata node with
4341 one <tt>i32</tt> entry of value 1. The existence of
4342 the <tt>!nontemporal</tt> metatadata on the instruction tells the optimizer
4343 and code generator that this load is not expected to be reused in the cache.
4344 The code generator may select special instructions to save cache bandwidth,
4345 such as the <tt>MOVNT</tt> instruction on x86.</p>
4348 <p>The location of memory pointed to is loaded. If the value being loaded is of
4349 scalar type then the number of bytes read does not exceed the minimum number
4350 of bytes needed to hold all bits of the type. For example, loading an
4351 <tt>i24</tt> reads at most three bytes. When loading a value of a type like
4352 <tt>i20</tt> with a size that is not an integral number of bytes, the result
4353 is undefined if the value was not originally written using a store of the
4358 %ptr = <a href="#i_alloca">alloca</a> i32 <i>; yields {i32*}:ptr</i>
4359 <a href="#i_store">store</a> i32 3, i32* %ptr <i>; yields {void}</i>
4360 %val = load i32* %ptr <i>; yields {i32}:val = i32 3</i>
4365 <!-- _______________________________________________________________________ -->
4366 <div class="doc_subsubsection"> <a name="i_store">'<tt>store</tt>'
4367 Instruction</a> </div>
4369 <div class="doc_text">
4373 store <ty> <value>, <ty>* <pointer>[, align <alignment>][, !nontemporal !<index>] <i>; yields {void}</i>
4374 volatile store <ty> <value>, <ty>* <pointer>[, align <alignment>][, !nontemporal !<index>] <i>; yields {void}</i>
4378 <p>The '<tt>store</tt>' instruction is used to write to memory.</p>
4381 <p>There are two arguments to the '<tt>store</tt>' instruction: a value to store
4382 and an address at which to store it. The type of the
4383 '<tt><pointer></tt>' operand must be a pointer to
4384 the <a href="#t_firstclass">first class</a> type of the
4385 '<tt><value></tt>' operand. If the <tt>store</tt> is marked as
4386 <tt>volatile</tt>, then the optimizer is not allowed to modify the number or
4387 order of execution of this <tt>store</tt> with other <a
4388 href="#volatile">volatile operations</a>.</p>
4390 <p>The optional constant "align" argument specifies the alignment of the
4391 operation (that is, the alignment of the memory address). A value of 0 or an
4392 omitted "align" argument means that the operation has the preferential
4393 alignment for the target. It is the responsibility of the code emitter to
4394 ensure that the alignment information is correct. Overestimating the
4395 alignment results in an undefined behavior. Underestimating the alignment may
4396 produce less efficient code. An alignment of 1 is always safe.</p>
4398 <p>The optional !nontemporal metadata must reference a single metatadata
4399 name <index> corresponding to a metadata node with one i32 entry of
4400 value 1. The existence of the !nontemporal metatadata on the
4401 instruction tells the optimizer and code generator that this load is
4402 not expected to be reused in the cache. The code generator may
4403 select special instructions to save cache bandwidth, such as the
4404 MOVNT instruction on x86.</p>
4408 <p>The contents of memory are updated to contain '<tt><value></tt>' at the
4409 location specified by the '<tt><pointer></tt>' operand. If
4410 '<tt><value></tt>' is of scalar type then the number of bytes written
4411 does not exceed the minimum number of bytes needed to hold all bits of the
4412 type. For example, storing an <tt>i24</tt> writes at most three bytes. When
4413 writing a value of a type like <tt>i20</tt> with a size that is not an
4414 integral number of bytes, it is unspecified what happens to the extra bits
4415 that do not belong to the type, but they will typically be overwritten.</p>
4419 %ptr = <a href="#i_alloca">alloca</a> i32 <i>; yields {i32*}:ptr</i>
4420 store i32 3, i32* %ptr <i>; yields {void}</i>
4421 %val = <a href="#i_load">load</a> i32* %ptr <i>; yields {i32}:val = i32 3</i>
4426 <!-- _______________________________________________________________________ -->
4427 <div class="doc_subsubsection">
4428 <a name="i_getelementptr">'<tt>getelementptr</tt>' Instruction</a>
4431 <div class="doc_text">
4435 <result> = getelementptr <pty>* <ptrval>{, <ty> <idx>}*
4436 <result> = getelementptr inbounds <pty>* <ptrval>{, <ty> <idx>}*
4440 <p>The '<tt>getelementptr</tt>' instruction is used to get the address of a
4441 subelement of an <a href="#t_aggregate">aggregate</a> data structure.
4442 It performs address calculation only and does not access memory.</p>
4445 <p>The first argument is always a pointer, and forms the basis of the
4446 calculation. The remaining arguments are indices that indicate which of the
4447 elements of the aggregate object are indexed. The interpretation of each
4448 index is dependent on the type being indexed into. The first index always
4449 indexes the pointer value given as the first argument, the second index
4450 indexes a value of the type pointed to (not necessarily the value directly
4451 pointed to, since the first index can be non-zero), etc. The first type
4452 indexed into must be a pointer value, subsequent types can be arrays,
4453 vectors, structs and unions. Note that subsequent types being indexed into
4454 can never be pointers, since that would require loading the pointer before
4455 continuing calculation.</p>
4457 <p>The type of each index argument depends on the type it is indexing into.
4458 When indexing into a (optionally packed) structure or union, only <tt>i32</tt>
4459 integer <b>constants</b> are allowed. When indexing into an array, pointer
4460 or vector, integers of any width are allowed, and they are not required to be
4463 <p>For example, let's consider a C code fragment and how it gets compiled to
4466 <div class="doc_code">
4479 int *foo(struct ST *s) {
4480 return &s[1].Z.B[5][13];
4485 <p>The LLVM code generated by the GCC frontend is:</p>
4487 <div class="doc_code">
4489 %RT = <a href="#namedtypes">type</a> { i8 , [10 x [20 x i32]], i8 }
4490 %ST = <a href="#namedtypes">type</a> { i32, double, %RT }
4492 define i32* @foo(%ST* %s) {
4494 %reg = getelementptr %ST* %s, i32 1, i32 2, i32 1, i32 5, i32 13
4501 <p>In the example above, the first index is indexing into the '<tt>%ST*</tt>'
4502 type, which is a pointer, yielding a '<tt>%ST</tt>' = '<tt>{ i32, double, %RT
4503 }</tt>' type, a structure. The second index indexes into the third element
4504 of the structure, yielding a '<tt>%RT</tt>' = '<tt>{ i8 , [10 x [20 x i32]],
4505 i8 }</tt>' type, another structure. The third index indexes into the second
4506 element of the structure, yielding a '<tt>[10 x [20 x i32]]</tt>' type, an
4507 array. The two dimensions of the array are subscripted into, yielding an
4508 '<tt>i32</tt>' type. The '<tt>getelementptr</tt>' instruction returns a
4509 pointer to this element, thus computing a value of '<tt>i32*</tt>' type.</p>
4511 <p>Note that it is perfectly legal to index partially through a structure,
4512 returning a pointer to an inner element. Because of this, the LLVM code for
4513 the given testcase is equivalent to:</p>
4516 define i32* @foo(%ST* %s) {
4517 %t1 = getelementptr %ST* %s, i32 1 <i>; yields %ST*:%t1</i>
4518 %t2 = getelementptr %ST* %t1, i32 0, i32 2 <i>; yields %RT*:%t2</i>
4519 %t3 = getelementptr %RT* %t2, i32 0, i32 1 <i>; yields [10 x [20 x i32]]*:%t3</i>
4520 %t4 = getelementptr [10 x [20 x i32]]* %t3, i32 0, i32 5 <i>; yields [20 x i32]*:%t4</i>
4521 %t5 = getelementptr [20 x i32]* %t4, i32 0, i32 13 <i>; yields i32*:%t5</i>
4526 <p>If the <tt>inbounds</tt> keyword is present, the result value of the
4527 <tt>getelementptr</tt> is a <a href="#trapvalues">trap value</a> if the
4528 base pointer is not an <i>in bounds</i> address of an allocated object,
4529 or if any of the addresses that would be formed by successive addition of
4530 the offsets implied by the indices to the base address with infinitely
4531 precise arithmetic are not an <i>in bounds</i> address of that allocated
4532 object. The <i>in bounds</i> addresses for an allocated object are all
4533 the addresses that point into the object, plus the address one byte past
4536 <p>If the <tt>inbounds</tt> keyword is not present, the offsets are added to
4537 the base address with silently-wrapping two's complement arithmetic, and
4538 the result value of the <tt>getelementptr</tt> may be outside the object
4539 pointed to by the base pointer. The result value may not necessarily be
4540 used to access memory though, even if it happens to point into allocated
4541 storage. See the <a href="#pointeraliasing">Pointer Aliasing Rules</a>
4542 section for more information.</p>
4544 <p>The getelementptr instruction is often confusing. For some more insight into
4545 how it works, see <a href="GetElementPtr.html">the getelementptr FAQ</a>.</p>
4549 <i>; yields [12 x i8]*:aptr</i>
4550 %aptr = getelementptr {i32, [12 x i8]}* %saptr, i64 0, i32 1
4551 <i>; yields i8*:vptr</i>
4552 %vptr = getelementptr {i32, <2 x i8>}* %svptr, i64 0, i32 1, i32 1
4553 <i>; yields i8*:eptr</i>
4554 %eptr = getelementptr [12 x i8]* %aptr, i64 0, i32 1
4555 <i>; yields i32*:iptr</i>
4556 %iptr = getelementptr [10 x i32]* @arr, i16 0, i16 0
4561 <!-- ======================================================================= -->
4562 <div class="doc_subsection"> <a name="convertops">Conversion Operations</a>
4565 <div class="doc_text">
4567 <p>The instructions in this category are the conversion instructions (casting)
4568 which all take a single operand and a type. They perform various bit
4569 conversions on the operand.</p>
4573 <!-- _______________________________________________________________________ -->
4574 <div class="doc_subsubsection">
4575 <a name="i_trunc">'<tt>trunc .. to</tt>' Instruction</a>
4577 <div class="doc_text">
4581 <result> = trunc <ty> <value> to <ty2> <i>; yields ty2</i>
4585 <p>The '<tt>trunc</tt>' instruction truncates its operand to the
4586 type <tt>ty2</tt>.</p>
4589 <p>The '<tt>trunc</tt>' instruction takes a <tt>value</tt> to trunc, which must
4590 be an <a href="#t_integer">integer</a> type, and a type that specifies the
4591 size and type of the result, which must be
4592 an <a href="#t_integer">integer</a> type. The bit size of <tt>value</tt> must
4593 be larger than the bit size of <tt>ty2</tt>. Equal sized types are not
4597 <p>The '<tt>trunc</tt>' instruction truncates the high order bits
4598 in <tt>value</tt> and converts the remaining bits to <tt>ty2</tt>. Since the
4599 source size must be larger than the destination size, <tt>trunc</tt> cannot
4600 be a <i>no-op cast</i>. It will always truncate bits.</p>
4604 %X = trunc i32 257 to i8 <i>; yields i8:1</i>
4605 %Y = trunc i32 123 to i1 <i>; yields i1:true</i>
4606 %Z = trunc i32 122 to i1 <i>; yields i1:false</i>
4611 <!-- _______________________________________________________________________ -->
4612 <div class="doc_subsubsection">
4613 <a name="i_zext">'<tt>zext .. to</tt>' Instruction</a>
4615 <div class="doc_text">
4619 <result> = zext <ty> <value> to <ty2> <i>; yields ty2</i>
4623 <p>The '<tt>zext</tt>' instruction zero extends its operand to type
4628 <p>The '<tt>zext</tt>' instruction takes a value to cast, which must be of
4629 <a href="#t_integer">integer</a> type, and a type to cast it to, which must
4630 also be of <a href="#t_integer">integer</a> type. The bit size of the
4631 <tt>value</tt> must be smaller than the bit size of the destination type,
4635 <p>The <tt>zext</tt> fills the high order bits of the <tt>value</tt> with zero
4636 bits until it reaches the size of the destination type, <tt>ty2</tt>.</p>
4638 <p>When zero extending from i1, the result will always be either 0 or 1.</p>
4642 %X = zext i32 257 to i64 <i>; yields i64:257</i>
4643 %Y = zext i1 true to i32 <i>; yields i32:1</i>
4648 <!-- _______________________________________________________________________ -->
4649 <div class="doc_subsubsection">
4650 <a name="i_sext">'<tt>sext .. to</tt>' Instruction</a>
4652 <div class="doc_text">
4656 <result> = sext <ty> <value> to <ty2> <i>; yields ty2</i>
4660 <p>The '<tt>sext</tt>' sign extends <tt>value</tt> to the type <tt>ty2</tt>.</p>
4663 <p>The '<tt>sext</tt>' instruction takes a value to cast, which must be of
4664 <a href="#t_integer">integer</a> type, and a type to cast it to, which must
4665 also be of <a href="#t_integer">integer</a> type. The bit size of the
4666 <tt>value</tt> must be smaller than the bit size of the destination type,
4670 <p>The '<tt>sext</tt>' instruction performs a sign extension by copying the sign
4671 bit (highest order bit) of the <tt>value</tt> until it reaches the bit size
4672 of the type <tt>ty2</tt>.</p>
4674 <p>When sign extending from i1, the extension always results in -1 or 0.</p>
4678 %X = sext i8 -1 to i16 <i>; yields i16 :65535</i>
4679 %Y = sext i1 true to i32 <i>; yields i32:-1</i>
4684 <!-- _______________________________________________________________________ -->
4685 <div class="doc_subsubsection">
4686 <a name="i_fptrunc">'<tt>fptrunc .. to</tt>' Instruction</a>
4689 <div class="doc_text">
4693 <result> = fptrunc <ty> <value> to <ty2> <i>; yields ty2</i>
4697 <p>The '<tt>fptrunc</tt>' instruction truncates <tt>value</tt> to type
4701 <p>The '<tt>fptrunc</tt>' instruction takes a <a href="#t_floating">floating
4702 point</a> value to cast and a <a href="#t_floating">floating point</a> type
4703 to cast it to. The size of <tt>value</tt> must be larger than the size of
4704 <tt>ty2</tt>. This implies that <tt>fptrunc</tt> cannot be used to make a
4705 <i>no-op cast</i>.</p>
4708 <p>The '<tt>fptrunc</tt>' instruction truncates a <tt>value</tt> from a larger
4709 <a href="#t_floating">floating point</a> type to a smaller
4710 <a href="#t_floating">floating point</a> type. If the value cannot fit
4711 within the destination type, <tt>ty2</tt>, then the results are
4716 %X = fptrunc double 123.0 to float <i>; yields float:123.0</i>
4717 %Y = fptrunc double 1.0E+300 to float <i>; yields undefined</i>
4722 <!-- _______________________________________________________________________ -->
4723 <div class="doc_subsubsection">
4724 <a name="i_fpext">'<tt>fpext .. to</tt>' Instruction</a>
4726 <div class="doc_text">
4730 <result> = fpext <ty> <value> to <ty2> <i>; yields ty2</i>
4734 <p>The '<tt>fpext</tt>' extends a floating point <tt>value</tt> to a larger
4735 floating point value.</p>
4738 <p>The '<tt>fpext</tt>' instruction takes a
4739 <a href="#t_floating">floating point</a> <tt>value</tt> to cast, and
4740 a <a href="#t_floating">floating point</a> type to cast it to. The source
4741 type must be smaller than the destination type.</p>
4744 <p>The '<tt>fpext</tt>' instruction extends the <tt>value</tt> from a smaller
4745 <a href="#t_floating">floating point</a> type to a larger
4746 <a href="#t_floating">floating point</a> type. The <tt>fpext</tt> cannot be
4747 used to make a <i>no-op cast</i> because it always changes bits. Use
4748 <tt>bitcast</tt> to make a <i>no-op cast</i> for a floating point cast.</p>
4752 %X = fpext float 3.1415 to double <i>; yields double:3.1415</i>
4753 %Y = fpext float 1.0 to float <i>; yields float:1.0 (no-op)</i>
4758 <!-- _______________________________________________________________________ -->
4759 <div class="doc_subsubsection">
4760 <a name="i_fptoui">'<tt>fptoui .. to</tt>' Instruction</a>
4762 <div class="doc_text">
4766 <result> = fptoui <ty> <value> to <ty2> <i>; yields ty2</i>
4770 <p>The '<tt>fptoui</tt>' converts a floating point <tt>value</tt> to its
4771 unsigned integer equivalent of type <tt>ty2</tt>.</p>
4774 <p>The '<tt>fptoui</tt>' instruction takes a value to cast, which must be a
4775 scalar or vector <a href="#t_floating">floating point</a> value, and a type
4776 to cast it to <tt>ty2</tt>, which must be an <a href="#t_integer">integer</a>
4777 type. If <tt>ty</tt> is a vector floating point type, <tt>ty2</tt> must be a
4778 vector integer type with the same number of elements as <tt>ty</tt></p>
4781 <p>The '<tt>fptoui</tt>' instruction converts its
4782 <a href="#t_floating">floating point</a> operand into the nearest (rounding
4783 towards zero) unsigned integer value. If the value cannot fit
4784 in <tt>ty2</tt>, the results are undefined.</p>
4788 %X = fptoui double 123.0 to i32 <i>; yields i32:123</i>
4789 %Y = fptoui float 1.0E+300 to i1 <i>; yields undefined:1</i>
4790 %Z = fptoui float 1.04E+17 to i8 <i>; yields undefined:1</i>
4795 <!-- _______________________________________________________________________ -->
4796 <div class="doc_subsubsection">
4797 <a name="i_fptosi">'<tt>fptosi .. to</tt>' Instruction</a>
4799 <div class="doc_text">
4803 <result> = fptosi <ty> <value> to <ty2> <i>; yields ty2</i>
4807 <p>The '<tt>fptosi</tt>' instruction converts
4808 <a href="#t_floating">floating point</a> <tt>value</tt> to
4809 type <tt>ty2</tt>.</p>
4812 <p>The '<tt>fptosi</tt>' instruction takes a value to cast, which must be a
4813 scalar or vector <a href="#t_floating">floating point</a> value, and a type
4814 to cast it to <tt>ty2</tt>, which must be an <a href="#t_integer">integer</a>
4815 type. If <tt>ty</tt> is a vector floating point type, <tt>ty2</tt> must be a
4816 vector integer type with the same number of elements as <tt>ty</tt></p>
4819 <p>The '<tt>fptosi</tt>' instruction converts its
4820 <a href="#t_floating">floating point</a> operand into the nearest (rounding
4821 towards zero) signed integer value. If the value cannot fit in <tt>ty2</tt>,
4822 the results are undefined.</p>
4826 %X = fptosi double -123.0 to i32 <i>; yields i32:-123</i>
4827 %Y = fptosi float 1.0E-247 to i1 <i>; yields undefined:1</i>
4828 %Z = fptosi float 1.04E+17 to i8 <i>; yields undefined:1</i>
4833 <!-- _______________________________________________________________________ -->
4834 <div class="doc_subsubsection">
4835 <a name="i_uitofp">'<tt>uitofp .. to</tt>' Instruction</a>
4837 <div class="doc_text">
4841 <result> = uitofp <ty> <value> to <ty2> <i>; yields ty2</i>
4845 <p>The '<tt>uitofp</tt>' instruction regards <tt>value</tt> as an unsigned
4846 integer and converts that value to the <tt>ty2</tt> type.</p>
4849 <p>The '<tt>uitofp</tt>' instruction takes a value to cast, which must be a
4850 scalar or vector <a href="#t_integer">integer</a> value, and a type to cast
4851 it to <tt>ty2</tt>, which must be an <a href="#t_floating">floating point</a>
4852 type. If <tt>ty</tt> is a vector integer type, <tt>ty2</tt> must be a vector
4853 floating point type with the same number of elements as <tt>ty</tt></p>
4856 <p>The '<tt>uitofp</tt>' instruction interprets its operand as an unsigned
4857 integer quantity and converts it to the corresponding floating point
4858 value. If the value cannot fit in the floating point value, the results are
4863 %X = uitofp i32 257 to float <i>; yields float:257.0</i>
4864 %Y = uitofp i8 -1 to double <i>; yields double:255.0</i>
4869 <!-- _______________________________________________________________________ -->
4870 <div class="doc_subsubsection">
4871 <a name="i_sitofp">'<tt>sitofp .. to</tt>' Instruction</a>
4873 <div class="doc_text">
4877 <result> = sitofp <ty> <value> to <ty2> <i>; yields ty2</i>
4881 <p>The '<tt>sitofp</tt>' instruction regards <tt>value</tt> as a signed integer
4882 and converts that value to the <tt>ty2</tt> type.</p>
4885 <p>The '<tt>sitofp</tt>' instruction takes a value to cast, which must be a
4886 scalar or vector <a href="#t_integer">integer</a> value, and a type to cast
4887 it to <tt>ty2</tt>, which must be an <a href="#t_floating">floating point</a>
4888 type. If <tt>ty</tt> is a vector integer type, <tt>ty2</tt> must be a vector
4889 floating point type with the same number of elements as <tt>ty</tt></p>
4892 <p>The '<tt>sitofp</tt>' instruction interprets its operand as a signed integer
4893 quantity and converts it to the corresponding floating point value. If the
4894 value cannot fit in the floating point value, the results are undefined.</p>
4898 %X = sitofp i32 257 to float <i>; yields float:257.0</i>
4899 %Y = sitofp i8 -1 to double <i>; yields double:-1.0</i>
4904 <!-- _______________________________________________________________________ -->
4905 <div class="doc_subsubsection">
4906 <a name="i_ptrtoint">'<tt>ptrtoint .. to</tt>' Instruction</a>
4908 <div class="doc_text">
4912 <result> = ptrtoint <ty> <value> to <ty2> <i>; yields ty2</i>
4916 <p>The '<tt>ptrtoint</tt>' instruction converts the pointer <tt>value</tt> to
4917 the integer type <tt>ty2</tt>.</p>
4920 <p>The '<tt>ptrtoint</tt>' instruction takes a <tt>value</tt> to cast, which
4921 must be a <a href="#t_pointer">pointer</a> value, and a type to cast it to
4922 <tt>ty2</tt>, which must be an <a href="#t_integer">integer</a> type.</p>
4925 <p>The '<tt>ptrtoint</tt>' instruction converts <tt>value</tt> to integer type
4926 <tt>ty2</tt> by interpreting the pointer value as an integer and either
4927 truncating or zero extending that value to the size of the integer type. If
4928 <tt>value</tt> is smaller than <tt>ty2</tt> then a zero extension is done. If
4929 <tt>value</tt> is larger than <tt>ty2</tt> then a truncation is done. If they
4930 are the same size, then nothing is done (<i>no-op cast</i>) other than a type
4935 %X = ptrtoint i32* %X to i8 <i>; yields truncation on 32-bit architecture</i>
4936 %Y = ptrtoint i32* %x to i64 <i>; yields zero extension on 32-bit architecture</i>
4941 <!-- _______________________________________________________________________ -->
4942 <div class="doc_subsubsection">
4943 <a name="i_inttoptr">'<tt>inttoptr .. to</tt>' Instruction</a>
4945 <div class="doc_text">
4949 <result> = inttoptr <ty> <value> to <ty2> <i>; yields ty2</i>
4953 <p>The '<tt>inttoptr</tt>' instruction converts an integer <tt>value</tt> to a
4954 pointer type, <tt>ty2</tt>.</p>
4957 <p>The '<tt>inttoptr</tt>' instruction takes an <a href="#t_integer">integer</a>
4958 value to cast, and a type to cast it to, which must be a
4959 <a href="#t_pointer">pointer</a> type.</p>
4962 <p>The '<tt>inttoptr</tt>' instruction converts <tt>value</tt> to type
4963 <tt>ty2</tt> by applying either a zero extension or a truncation depending on
4964 the size of the integer <tt>value</tt>. If <tt>value</tt> is larger than the
4965 size of a pointer then a truncation is done. If <tt>value</tt> is smaller
4966 than the size of a pointer then a zero extension is done. If they are the
4967 same size, nothing is done (<i>no-op cast</i>).</p>
4971 %X = inttoptr i32 255 to i32* <i>; yields zero extension on 64-bit architecture</i>
4972 %Y = inttoptr i32 255 to i32* <i>; yields no-op on 32-bit architecture</i>
4973 %Z = inttoptr i64 0 to i32* <i>; yields truncation on 32-bit architecture</i>
4978 <!-- _______________________________________________________________________ -->
4979 <div class="doc_subsubsection">
4980 <a name="i_bitcast">'<tt>bitcast .. to</tt>' Instruction</a>
4982 <div class="doc_text">
4986 <result> = bitcast <ty> <value> to <ty2> <i>; yields ty2</i>
4990 <p>The '<tt>bitcast</tt>' instruction converts <tt>value</tt> to type
4991 <tt>ty2</tt> without changing any bits.</p>
4994 <p>The '<tt>bitcast</tt>' instruction takes a value to cast, which must be a
4995 non-aggregate first class value, and a type to cast it to, which must also be
4996 a non-aggregate <a href="#t_firstclass">first class</a> type. The bit sizes
4997 of <tt>value</tt> and the destination type, <tt>ty2</tt>, must be
4998 identical. If the source type is a pointer, the destination type must also be
4999 a pointer. This instruction supports bitwise conversion of vectors to
5000 integers and to vectors of other types (as long as they have the same
5004 <p>The '<tt>bitcast</tt>' instruction converts <tt>value</tt> to type
5005 <tt>ty2</tt>. It is always a <i>no-op cast</i> because no bits change with
5006 this conversion. The conversion is done as if the <tt>value</tt> had been
5007 stored to memory and read back as type <tt>ty2</tt>. Pointer types may only
5008 be converted to other pointer types with this instruction. To convert
5009 pointers to other types, use the <a href="#i_inttoptr">inttoptr</a> or
5010 <a href="#i_ptrtoint">ptrtoint</a> instructions first.</p>
5014 %X = bitcast i8 255 to i8 <i>; yields i8 :-1</i>
5015 %Y = bitcast i32* %x to sint* <i>; yields sint*:%x</i>
5016 %Z = bitcast <2 x int> %V to i64; <i>; yields i64: %V</i>
5021 <!-- ======================================================================= -->
5022 <div class="doc_subsection"> <a name="otherops">Other Operations</a> </div>
5024 <div class="doc_text">
5026 <p>The instructions in this category are the "miscellaneous" instructions, which
5027 defy better classification.</p>
5031 <!-- _______________________________________________________________________ -->
5032 <div class="doc_subsubsection"><a name="i_icmp">'<tt>icmp</tt>' Instruction</a>
5035 <div class="doc_text">
5039 <result> = icmp <cond> <ty> <op1>, <op2> <i>; yields {i1} or {<N x i1>}:result</i>
5043 <p>The '<tt>icmp</tt>' instruction returns a boolean value or a vector of
5044 boolean values based on comparison of its two integer, integer vector, or
5045 pointer operands.</p>
5048 <p>The '<tt>icmp</tt>' instruction takes three operands. The first operand is
5049 the condition code indicating the kind of comparison to perform. It is not a
5050 value, just a keyword. The possible condition code are:</p>
5053 <li><tt>eq</tt>: equal</li>
5054 <li><tt>ne</tt>: not equal </li>
5055 <li><tt>ugt</tt>: unsigned greater than</li>
5056 <li><tt>uge</tt>: unsigned greater or equal</li>
5057 <li><tt>ult</tt>: unsigned less than</li>
5058 <li><tt>ule</tt>: unsigned less or equal</li>
5059 <li><tt>sgt</tt>: signed greater than</li>
5060 <li><tt>sge</tt>: signed greater or equal</li>
5061 <li><tt>slt</tt>: signed less than</li>
5062 <li><tt>sle</tt>: signed less or equal</li>
5065 <p>The remaining two arguments must be <a href="#t_integer">integer</a> or
5066 <a href="#t_pointer">pointer</a> or integer <a href="#t_vector">vector</a>
5067 typed. They must also be identical types.</p>
5070 <p>The '<tt>icmp</tt>' compares <tt>op1</tt> and <tt>op2</tt> according to the
5071 condition code given as <tt>cond</tt>. The comparison performed always yields
5072 either an <a href="#t_integer"><tt>i1</tt></a> or vector of <tt>i1</tt>
5073 result, as follows:</p>
5076 <li><tt>eq</tt>: yields <tt>true</tt> if the operands are equal,
5077 <tt>false</tt> otherwise. No sign interpretation is necessary or
5080 <li><tt>ne</tt>: yields <tt>true</tt> if the operands are unequal,
5081 <tt>false</tt> otherwise. No sign interpretation is necessary or
5084 <li><tt>ugt</tt>: interprets the operands as unsigned values and yields
5085 <tt>true</tt> if <tt>op1</tt> is greater than <tt>op2</tt>.</li>
5087 <li><tt>uge</tt>: interprets the operands as unsigned values and yields
5088 <tt>true</tt> if <tt>op1</tt> is greater than or equal
5089 to <tt>op2</tt>.</li>
5091 <li><tt>ult</tt>: interprets the operands as unsigned values and yields
5092 <tt>true</tt> if <tt>op1</tt> is less than <tt>op2</tt>.</li>
5094 <li><tt>ule</tt>: interprets the operands as unsigned values and yields
5095 <tt>true</tt> if <tt>op1</tt> is less than or equal to <tt>op2</tt>.</li>
5097 <li><tt>sgt</tt>: interprets the operands as signed values and yields
5098 <tt>true</tt> if <tt>op1</tt> is greater than <tt>op2</tt>.</li>
5100 <li><tt>sge</tt>: interprets the operands as signed values and yields
5101 <tt>true</tt> if <tt>op1</tt> is greater than or equal
5102 to <tt>op2</tt>.</li>
5104 <li><tt>slt</tt>: interprets the operands as signed values and yields
5105 <tt>true</tt> if <tt>op1</tt> is less than <tt>op2</tt>.</li>
5107 <li><tt>sle</tt>: interprets the operands as signed values and yields
5108 <tt>true</tt> if <tt>op1</tt> is less than or equal to <tt>op2</tt>.</li>
5111 <p>If the operands are <a href="#t_pointer">pointer</a> typed, the pointer
5112 values are compared as if they were integers.</p>
5114 <p>If the operands are integer vectors, then they are compared element by
5115 element. The result is an <tt>i1</tt> vector with the same number of elements
5116 as the values being compared. Otherwise, the result is an <tt>i1</tt>.</p>
5120 <result> = icmp eq i32 4, 5 <i>; yields: result=false</i>
5121 <result> = icmp ne float* %X, %X <i>; yields: result=false</i>
5122 <result> = icmp ult i16 4, 5 <i>; yields: result=true</i>
5123 <result> = icmp sgt i16 4, 5 <i>; yields: result=false</i>
5124 <result> = icmp ule i16 -4, 5 <i>; yields: result=false</i>
5125 <result> = icmp sge i16 4, 5 <i>; yields: result=false</i>
5128 <p>Note that the code generator does not yet support vector types with
5129 the <tt>icmp</tt> instruction.</p>
5133 <!-- _______________________________________________________________________ -->
5134 <div class="doc_subsubsection"><a name="i_fcmp">'<tt>fcmp</tt>' Instruction</a>
5137 <div class="doc_text">
5141 <result> = fcmp <cond> <ty> <op1>, <op2> <i>; yields {i1} or {<N x i1>}:result</i>
5145 <p>The '<tt>fcmp</tt>' instruction returns a boolean value or vector of boolean
5146 values based on comparison of its operands.</p>
5148 <p>If the operands are floating point scalars, then the result type is a boolean
5149 (<a href="#t_integer"><tt>i1</tt></a>).</p>
5151 <p>If the operands are floating point vectors, then the result type is a vector
5152 of boolean with the same number of elements as the operands being
5156 <p>The '<tt>fcmp</tt>' instruction takes three operands. The first operand is
5157 the condition code indicating the kind of comparison to perform. It is not a
5158 value, just a keyword. The possible condition code are:</p>
5161 <li><tt>false</tt>: no comparison, always returns false</li>
5162 <li><tt>oeq</tt>: ordered and equal</li>
5163 <li><tt>ogt</tt>: ordered and greater than </li>
5164 <li><tt>oge</tt>: ordered and greater than or equal</li>
5165 <li><tt>olt</tt>: ordered and less than </li>
5166 <li><tt>ole</tt>: ordered and less than or equal</li>
5167 <li><tt>one</tt>: ordered and not equal</li>
5168 <li><tt>ord</tt>: ordered (no nans)</li>
5169 <li><tt>ueq</tt>: unordered or equal</li>
5170 <li><tt>ugt</tt>: unordered or greater than </li>
5171 <li><tt>uge</tt>: unordered or greater than or equal</li>
5172 <li><tt>ult</tt>: unordered or less than </li>
5173 <li><tt>ule</tt>: unordered or less than or equal</li>
5174 <li><tt>une</tt>: unordered or not equal</li>
5175 <li><tt>uno</tt>: unordered (either nans)</li>
5176 <li><tt>true</tt>: no comparison, always returns true</li>
5179 <p><i>Ordered</i> means that neither operand is a QNAN while
5180 <i>unordered</i> means that either operand may be a QNAN.</p>
5182 <p>Each of <tt>val1</tt> and <tt>val2</tt> arguments must be either
5183 a <a href="#t_floating">floating point</a> type or
5184 a <a href="#t_vector">vector</a> of floating point type. They must have
5185 identical types.</p>
5188 <p>The '<tt>fcmp</tt>' instruction compares <tt>op1</tt> and <tt>op2</tt>
5189 according to the condition code given as <tt>cond</tt>. If the operands are
5190 vectors, then the vectors are compared element by element. Each comparison
5191 performed always yields an <a href="#t_integer">i1</a> result, as
5195 <li><tt>false</tt>: always yields <tt>false</tt>, regardless of operands.</li>
5197 <li><tt>oeq</tt>: yields <tt>true</tt> if both operands are not a QNAN and
5198 <tt>op1</tt> is equal to <tt>op2</tt>.</li>
5200 <li><tt>ogt</tt>: yields <tt>true</tt> if both operands are not a QNAN and
5201 <tt>op1</tt> is greater than <tt>op2</tt>.</li>
5203 <li><tt>oge</tt>: yields <tt>true</tt> if both operands are not a QNAN and
5204 <tt>op1</tt> is greater than or equal to <tt>op2</tt>.</li>
5206 <li><tt>olt</tt>: yields <tt>true</tt> if both operands are not a QNAN and
5207 <tt>op1</tt> is less than <tt>op2</tt>.</li>
5209 <li><tt>ole</tt>: yields <tt>true</tt> if both operands are not a QNAN and
5210 <tt>op1</tt> is less than or equal to <tt>op2</tt>.</li>
5212 <li><tt>one</tt>: yields <tt>true</tt> if both operands are not a QNAN and
5213 <tt>op1</tt> is not equal to <tt>op2</tt>.</li>
5215 <li><tt>ord</tt>: yields <tt>true</tt> if both operands are not a QNAN.</li>
5217 <li><tt>ueq</tt>: yields <tt>true</tt> if either operand is a QNAN or
5218 <tt>op1</tt> is equal to <tt>op2</tt>.</li>
5220 <li><tt>ugt</tt>: yields <tt>true</tt> if either operand is a QNAN or
5221 <tt>op1</tt> is greater than <tt>op2</tt>.</li>
5223 <li><tt>uge</tt>: yields <tt>true</tt> if either operand is a QNAN or
5224 <tt>op1</tt> is greater than or equal to <tt>op2</tt>.</li>
5226 <li><tt>ult</tt>: yields <tt>true</tt> if either operand is a QNAN or
5227 <tt>op1</tt> is less than <tt>op2</tt>.</li>
5229 <li><tt>ule</tt>: yields <tt>true</tt> if either operand is a QNAN or
5230 <tt>op1</tt> is less than or equal to <tt>op2</tt>.</li>
5232 <li><tt>une</tt>: yields <tt>true</tt> if either operand is a QNAN or
5233 <tt>op1</tt> is not equal to <tt>op2</tt>.</li>
5235 <li><tt>uno</tt>: yields <tt>true</tt> if either operand is a QNAN.</li>
5237 <li><tt>true</tt>: always yields <tt>true</tt>, regardless of operands.</li>
5242 <result> = fcmp oeq float 4.0, 5.0 <i>; yields: result=false</i>
5243 <result> = fcmp one float 4.0, 5.0 <i>; yields: result=true</i>
5244 <result> = fcmp olt float 4.0, 5.0 <i>; yields: result=true</i>
5245 <result> = fcmp ueq double 1.0, 2.0 <i>; yields: result=false</i>
5248 <p>Note that the code generator does not yet support vector types with
5249 the <tt>fcmp</tt> instruction.</p>
5253 <!-- _______________________________________________________________________ -->
5254 <div class="doc_subsubsection">
5255 <a name="i_phi">'<tt>phi</tt>' Instruction</a>
5258 <div class="doc_text">
5262 <result> = phi <ty> [ <val0>, <label0>], ...
5266 <p>The '<tt>phi</tt>' instruction is used to implement the φ node in the
5267 SSA graph representing the function.</p>
5270 <p>The type of the incoming values is specified with the first type field. After
5271 this, the '<tt>phi</tt>' instruction takes a list of pairs as arguments, with
5272 one pair for each predecessor basic block of the current block. Only values
5273 of <a href="#t_firstclass">first class</a> type may be used as the value
5274 arguments to the PHI node. Only labels may be used as the label
5277 <p>There must be no non-phi instructions between the start of a basic block and
5278 the PHI instructions: i.e. PHI instructions must be first in a basic
5281 <p>For the purposes of the SSA form, the use of each incoming value is deemed to
5282 occur on the edge from the corresponding predecessor block to the current
5283 block (but after any definition of an '<tt>invoke</tt>' instruction's return
5284 value on the same edge).</p>
5287 <p>At runtime, the '<tt>phi</tt>' instruction logically takes on the value
5288 specified by the pair corresponding to the predecessor basic block that
5289 executed just prior to the current block.</p>
5293 Loop: ; Infinite loop that counts from 0 on up...
5294 %indvar = phi i32 [ 0, %LoopHeader ], [ %nextindvar, %Loop ]
5295 %nextindvar = add i32 %indvar, 1
5301 <!-- _______________________________________________________________________ -->
5302 <div class="doc_subsubsection">
5303 <a name="i_select">'<tt>select</tt>' Instruction</a>
5306 <div class="doc_text">
5310 <result> = select <i>selty</i> <cond>, <ty> <val1>, <ty> <val2> <i>; yields ty</i>
5312 <i>selty</i> is either i1 or {<N x i1>}
5316 <p>The '<tt>select</tt>' instruction is used to choose one value based on a
5317 condition, without branching.</p>
5321 <p>The '<tt>select</tt>' instruction requires an 'i1' value or a vector of 'i1'
5322 values indicating the condition, and two values of the
5323 same <a href="#t_firstclass">first class</a> type. If the val1/val2 are
5324 vectors and the condition is a scalar, then entire vectors are selected, not
5325 individual elements.</p>
5328 <p>If the condition is an i1 and it evaluates to 1, the instruction returns the
5329 first value argument; otherwise, it returns the second value argument.</p>
5331 <p>If the condition is a vector of i1, then the value arguments must be vectors
5332 of the same size, and the selection is done element by element.</p>
5336 %X = select i1 true, i8 17, i8 42 <i>; yields i8:17</i>
5339 <p>Note that the code generator does not yet support conditions
5340 with vector type.</p>
5344 <!-- _______________________________________________________________________ -->
5345 <div class="doc_subsubsection">
5346 <a name="i_call">'<tt>call</tt>' Instruction</a>
5349 <div class="doc_text">
5353 <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>]
5357 <p>The '<tt>call</tt>' instruction represents a simple function call.</p>
5360 <p>This instruction requires several arguments:</p>
5363 <li>The optional "tail" marker indicates that the callee function does not
5364 access any allocas or varargs in the caller. Note that calls may be
5365 marked "tail" even if they do not occur before
5366 a <a href="#i_ret"><tt>ret</tt></a> instruction. If the "tail" marker is
5367 present, the function call is eligible for tail call optimization,
5368 but <a href="CodeGenerator.html#tailcallopt">might not in fact be
5369 optimized into a jump</a>. The code generator may optimize calls marked
5370 "tail" with either 1) automatic <a href="CodeGenerator.html#sibcallopt">
5371 sibling call optimization</a> when the caller and callee have
5372 matching signatures, or 2) forced tail call optimization when the
5373 following extra requirements are met:
5375 <li>Caller and callee both have the calling
5376 convention <tt>fastcc</tt>.</li>
5377 <li>The call is in tail position (ret immediately follows call and ret
5378 uses value of call or is void).</li>
5379 <li>Option <tt>-tailcallopt</tt> is enabled,
5380 or <code>llvm::GuaranteedTailCallOpt</code> is <code>true</code>.</li>
5381 <li><a href="CodeGenerator.html#tailcallopt">Platform specific
5382 constraints are met.</a></li>
5386 <li>The optional "cconv" marker indicates which <a href="#callingconv">calling
5387 convention</a> the call should use. If none is specified, the call
5388 defaults to using C calling conventions. The calling convention of the
5389 call must match the calling convention of the target function, or else the
5390 behavior is undefined.</li>
5392 <li>The optional <a href="#paramattrs">Parameter Attributes</a> list for
5393 return values. Only '<tt>zeroext</tt>', '<tt>signext</tt>', and
5394 '<tt>inreg</tt>' attributes are valid here.</li>
5396 <li>'<tt>ty</tt>': the type of the call instruction itself which is also the
5397 type of the return value. Functions that return no value are marked
5398 <tt><a href="#t_void">void</a></tt>.</li>
5400 <li>'<tt>fnty</tt>': shall be the signature of the pointer to function value
5401 being invoked. The argument types must match the types implied by this
5402 signature. This type can be omitted if the function is not varargs and if
5403 the function type does not return a pointer to a function.</li>
5405 <li>'<tt>fnptrval</tt>': An LLVM value containing a pointer to a function to
5406 be invoked. In most cases, this is a direct function invocation, but
5407 indirect <tt>call</tt>s are just as possible, calling an arbitrary pointer
5408 to function value.</li>
5410 <li>'<tt>function args</tt>': argument list whose types match the function
5411 signature argument types and parameter attributes. All arguments must be
5412 of <a href="#t_firstclass">first class</a> type. If the function
5413 signature indicates the function accepts a variable number of arguments,
5414 the extra arguments can be specified.</li>
5416 <li>The optional <a href="#fnattrs">function attributes</a> list. Only
5417 '<tt>noreturn</tt>', '<tt>nounwind</tt>', '<tt>readonly</tt>' and
5418 '<tt>readnone</tt>' attributes are valid here.</li>
5422 <p>The '<tt>call</tt>' instruction is used to cause control flow to transfer to
5423 a specified function, with its incoming arguments bound to the specified
5424 values. Upon a '<tt><a href="#i_ret">ret</a></tt>' instruction in the called
5425 function, control flow continues with the instruction after the function
5426 call, and the return value of the function is bound to the result
5431 %retval = call i32 @test(i32 %argc)
5432 call i32 (i8*, ...)* @printf(i8* %msg, i32 12, i8 42) <i>; yields i32</i>
5433 %X = tail call i32 @foo() <i>; yields i32</i>
5434 %Y = tail call <a href="#callingconv">fastcc</a> i32 @foo() <i>; yields i32</i>
5435 call void %foo(i8 97 signext)
5437 %struct.A = type { i32, i8 }
5438 %r = call %struct.A @foo() <i>; yields { 32, i8 }</i>
5439 %gr = extractvalue %struct.A %r, 0 <i>; yields i32</i>
5440 %gr1 = extractvalue %struct.A %r, 1 <i>; yields i8</i>
5441 %Z = call void @foo() noreturn <i>; indicates that %foo never returns normally</i>
5442 %ZZ = call zeroext i32 @bar() <i>; Return value is %zero extended</i>
5445 <p>llvm treats calls to some functions with names and arguments that match the
5446 standard C99 library as being the C99 library functions, and may perform
5447 optimizations or generate code for them under that assumption. This is
5448 something we'd like to change in the future to provide better support for
5449 freestanding environments and non-C-based languages.</p>
5453 <!-- _______________________________________________________________________ -->
5454 <div class="doc_subsubsection">
5455 <a name="i_va_arg">'<tt>va_arg</tt>' Instruction</a>
5458 <div class="doc_text">
5462 <resultval> = va_arg <va_list*> <arglist>, <argty>
5466 <p>The '<tt>va_arg</tt>' instruction is used to access arguments passed through
5467 the "variable argument" area of a function call. It is used to implement the
5468 <tt>va_arg</tt> macro in C.</p>
5471 <p>This instruction takes a <tt>va_list*</tt> value and the type of the
5472 argument. It returns a value of the specified argument type and increments
5473 the <tt>va_list</tt> to point to the next argument. The actual type
5474 of <tt>va_list</tt> is target specific.</p>
5477 <p>The '<tt>va_arg</tt>' instruction loads an argument of the specified type
5478 from the specified <tt>va_list</tt> and causes the <tt>va_list</tt> to point
5479 to the next argument. For more information, see the variable argument
5480 handling <a href="#int_varargs">Intrinsic Functions</a>.</p>
5482 <p>It is legal for this instruction to be called in a function which does not
5483 take a variable number of arguments, for example, the <tt>vfprintf</tt>
5486 <p><tt>va_arg</tt> is an LLVM instruction instead of
5487 an <a href="#intrinsics">intrinsic function</a> because it takes a type as an
5491 <p>See the <a href="#int_varargs">variable argument processing</a> section.</p>
5493 <p>Note that the code generator does not yet fully support va_arg on many
5494 targets. Also, it does not currently support va_arg with aggregate types on
5499 <!-- *********************************************************************** -->
5500 <div class="doc_section"> <a name="intrinsics">Intrinsic Functions</a> </div>
5501 <!-- *********************************************************************** -->
5503 <div class="doc_text">
5505 <p>LLVM supports the notion of an "intrinsic function". These functions have
5506 well known names and semantics and are required to follow certain
5507 restrictions. Overall, these intrinsics represent an extension mechanism for
5508 the LLVM language that does not require changing all of the transformations
5509 in LLVM when adding to the language (or the bitcode reader/writer, the
5510 parser, etc...).</p>
5512 <p>Intrinsic function names must all start with an "<tt>llvm.</tt>" prefix. This
5513 prefix is reserved in LLVM for intrinsic names; thus, function names may not
5514 begin with this prefix. Intrinsic functions must always be external
5515 functions: you cannot define the body of intrinsic functions. Intrinsic
5516 functions may only be used in call or invoke instructions: it is illegal to
5517 take the address of an intrinsic function. Additionally, because intrinsic
5518 functions are part of the LLVM language, it is required if any are added that
5519 they be documented here.</p>
5521 <p>Some intrinsic functions can be overloaded, i.e., the intrinsic represents a
5522 family of functions that perform the same operation but on different data
5523 types. Because LLVM can represent over 8 million different integer types,
5524 overloading is used commonly to allow an intrinsic function to operate on any
5525 integer type. One or more of the argument types or the result type can be
5526 overloaded to accept any integer type. Argument types may also be defined as
5527 exactly matching a previous argument's type or the result type. This allows
5528 an intrinsic function which accepts multiple arguments, but needs all of them
5529 to be of the same type, to only be overloaded with respect to a single
5530 argument or the result.</p>
5532 <p>Overloaded intrinsics will have the names of its overloaded argument types
5533 encoded into its function name, each preceded by a period. Only those types
5534 which are overloaded result in a name suffix. Arguments whose type is matched
5535 against another type do not. For example, the <tt>llvm.ctpop</tt> function
5536 can take an integer of any width and returns an integer of exactly the same
5537 integer width. This leads to a family of functions such as
5538 <tt>i8 @llvm.ctpop.i8(i8 %val)</tt> and <tt>i29 @llvm.ctpop.i29(i29
5539 %val)</tt>. Only one type, the return type, is overloaded, and only one type
5540 suffix is required. Because the argument's type is matched against the return
5541 type, it does not require its own name suffix.</p>
5543 <p>To learn how to add an intrinsic function, please see the
5544 <a href="ExtendingLLVM.html">Extending LLVM Guide</a>.</p>
5548 <!-- ======================================================================= -->
5549 <div class="doc_subsection">
5550 <a name="int_varargs">Variable Argument Handling Intrinsics</a>
5553 <div class="doc_text">
5555 <p>Variable argument support is defined in LLVM with
5556 the <a href="#i_va_arg"><tt>va_arg</tt></a> instruction and these three
5557 intrinsic functions. These functions are related to the similarly named
5558 macros defined in the <tt><stdarg.h></tt> header file.</p>
5560 <p>All of these functions operate on arguments that use a target-specific value
5561 type "<tt>va_list</tt>". The LLVM assembly language reference manual does
5562 not define what this type is, so all transformations should be prepared to
5563 handle these functions regardless of the type used.</p>
5565 <p>This example shows how the <a href="#i_va_arg"><tt>va_arg</tt></a>
5566 instruction and the variable argument handling intrinsic functions are
5569 <div class="doc_code">
5571 define i32 @test(i32 %X, ...) {
5572 ; Initialize variable argument processing
5574 %ap2 = bitcast i8** %ap to i8*
5575 call void @llvm.va_start(i8* %ap2)
5577 ; Read a single integer argument
5578 %tmp = va_arg i8** %ap, i32
5580 ; Demonstrate usage of llvm.va_copy and llvm.va_end
5582 %aq2 = bitcast i8** %aq to i8*
5583 call void @llvm.va_copy(i8* %aq2, i8* %ap2)
5584 call void @llvm.va_end(i8* %aq2)
5586 ; Stop processing of arguments.
5587 call void @llvm.va_end(i8* %ap2)
5591 declare void @llvm.va_start(i8*)
5592 declare void @llvm.va_copy(i8*, i8*)
5593 declare void @llvm.va_end(i8*)
5599 <!-- _______________________________________________________________________ -->
5600 <div class="doc_subsubsection">
5601 <a name="int_va_start">'<tt>llvm.va_start</tt>' Intrinsic</a>
5605 <div class="doc_text">
5609 declare void %llvm.va_start(i8* <arglist>)
5613 <p>The '<tt>llvm.va_start</tt>' intrinsic initializes <tt>*<arglist></tt>
5614 for subsequent use by <tt><a href="#i_va_arg">va_arg</a></tt>.</p>
5617 <p>The argument is a pointer to a <tt>va_list</tt> element to initialize.</p>
5620 <p>The '<tt>llvm.va_start</tt>' intrinsic works just like the <tt>va_start</tt>
5621 macro available in C. In a target-dependent way, it initializes
5622 the <tt>va_list</tt> element to which the argument points, so that the next
5623 call to <tt>va_arg</tt> will produce the first variable argument passed to
5624 the function. Unlike the C <tt>va_start</tt> macro, this intrinsic does not
5625 need to know the last argument of the function as the compiler can figure
5630 <!-- _______________________________________________________________________ -->
5631 <div class="doc_subsubsection">
5632 <a name="int_va_end">'<tt>llvm.va_end</tt>' Intrinsic</a>
5635 <div class="doc_text">
5639 declare void @llvm.va_end(i8* <arglist>)
5643 <p>The '<tt>llvm.va_end</tt>' intrinsic destroys <tt>*<arglist></tt>,
5644 which has been initialized previously
5645 with <tt><a href="#int_va_start">llvm.va_start</a></tt>
5646 or <tt><a href="#i_va_copy">llvm.va_copy</a></tt>.</p>
5649 <p>The argument is a pointer to a <tt>va_list</tt> to destroy.</p>
5652 <p>The '<tt>llvm.va_end</tt>' intrinsic works just like the <tt>va_end</tt>
5653 macro available in C. In a target-dependent way, it destroys
5654 the <tt>va_list</tt> element to which the argument points. Calls
5655 to <a href="#int_va_start"><tt>llvm.va_start</tt></a>
5656 and <a href="#int_va_copy"> <tt>llvm.va_copy</tt></a> must be matched exactly
5657 with calls to <tt>llvm.va_end</tt>.</p>
5661 <!-- _______________________________________________________________________ -->
5662 <div class="doc_subsubsection">
5663 <a name="int_va_copy">'<tt>llvm.va_copy</tt>' Intrinsic</a>
5666 <div class="doc_text">
5670 declare void @llvm.va_copy(i8* <destarglist>, i8* <srcarglist>)
5674 <p>The '<tt>llvm.va_copy</tt>' intrinsic copies the current argument position
5675 from the source argument list to the destination argument list.</p>
5678 <p>The first argument is a pointer to a <tt>va_list</tt> element to initialize.
5679 The second argument is a pointer to a <tt>va_list</tt> element to copy
5683 <p>The '<tt>llvm.va_copy</tt>' intrinsic works just like the <tt>va_copy</tt>
5684 macro available in C. In a target-dependent way, it copies the
5685 source <tt>va_list</tt> element into the destination <tt>va_list</tt>
5686 element. This intrinsic is necessary because
5687 the <tt><a href="#int_va_start"> llvm.va_start</a></tt> intrinsic may be
5688 arbitrarily complex and require, for example, memory allocation.</p>
5692 <!-- ======================================================================= -->
5693 <div class="doc_subsection">
5694 <a name="int_gc">Accurate Garbage Collection Intrinsics</a>
5697 <div class="doc_text">
5699 <p>LLVM support for <a href="GarbageCollection.html">Accurate Garbage
5700 Collection</a> (GC) requires the implementation and generation of these
5701 intrinsics. These intrinsics allow identification of <a href="#int_gcroot">GC
5702 roots on the stack</a>, as well as garbage collector implementations that
5703 require <a href="#int_gcread">read</a> and <a href="#int_gcwrite">write</a>
5704 barriers. Front-ends for type-safe garbage collected languages should generate
5705 these intrinsics to make use of the LLVM garbage collectors. For more details,
5706 see <a href="GarbageCollection.html">Accurate Garbage Collection with
5709 <p>The garbage collection intrinsics only operate on objects in the generic
5710 address space (address space zero).</p>
5714 <!-- _______________________________________________________________________ -->
5715 <div class="doc_subsubsection">
5716 <a name="int_gcroot">'<tt>llvm.gcroot</tt>' Intrinsic</a>
5719 <div class="doc_text">
5723 declare void @llvm.gcroot(i8** %ptrloc, i8* %metadata)
5727 <p>The '<tt>llvm.gcroot</tt>' intrinsic declares the existence of a GC root to
5728 the code generator, and allows some metadata to be associated with it.</p>
5731 <p>The first argument specifies the address of a stack object that contains the
5732 root pointer. The second pointer (which must be either a constant or a
5733 global value address) contains the meta-data to be associated with the
5737 <p>At runtime, a call to this intrinsic stores a null pointer into the "ptrloc"
5738 location. At compile-time, the code generator generates information to allow
5739 the runtime to find the pointer at GC safe points. The '<tt>llvm.gcroot</tt>'
5740 intrinsic may only be used in a function which <a href="#gc">specifies a GC
5745 <!-- _______________________________________________________________________ -->
5746 <div class="doc_subsubsection">
5747 <a name="int_gcread">'<tt>llvm.gcread</tt>' Intrinsic</a>
5750 <div class="doc_text">
5754 declare i8* @llvm.gcread(i8* %ObjPtr, i8** %Ptr)
5758 <p>The '<tt>llvm.gcread</tt>' intrinsic identifies reads of references from heap
5759 locations, allowing garbage collector implementations that require read
5763 <p>The second argument is the address to read from, which should be an address
5764 allocated from the garbage collector. The first object is a pointer to the
5765 start of the referenced object, if needed by the language runtime (otherwise
5769 <p>The '<tt>llvm.gcread</tt>' intrinsic has the same semantics as a load
5770 instruction, but may be replaced with substantially more complex code by the
5771 garbage collector runtime, as needed. The '<tt>llvm.gcread</tt>' intrinsic
5772 may only be used in a function which <a href="#gc">specifies a GC
5777 <!-- _______________________________________________________________________ -->
5778 <div class="doc_subsubsection">
5779 <a name="int_gcwrite">'<tt>llvm.gcwrite</tt>' Intrinsic</a>
5782 <div class="doc_text">
5786 declare void @llvm.gcwrite(i8* %P1, i8* %Obj, i8** %P2)
5790 <p>The '<tt>llvm.gcwrite</tt>' intrinsic identifies writes of references to heap
5791 locations, allowing garbage collector implementations that require write
5792 barriers (such as generational or reference counting collectors).</p>
5795 <p>The first argument is the reference to store, the second is the start of the
5796 object to store it to, and the third is the address of the field of Obj to
5797 store to. If the runtime does not require a pointer to the object, Obj may
5801 <p>The '<tt>llvm.gcwrite</tt>' intrinsic has the same semantics as a store
5802 instruction, but may be replaced with substantially more complex code by the
5803 garbage collector runtime, as needed. The '<tt>llvm.gcwrite</tt>' intrinsic
5804 may only be used in a function which <a href="#gc">specifies a GC
5809 <!-- ======================================================================= -->
5810 <div class="doc_subsection">
5811 <a name="int_codegen">Code Generator Intrinsics</a>
5814 <div class="doc_text">
5816 <p>These intrinsics are provided by LLVM to expose special features that may
5817 only be implemented with code generator support.</p>
5821 <!-- _______________________________________________________________________ -->
5822 <div class="doc_subsubsection">
5823 <a name="int_returnaddress">'<tt>llvm.returnaddress</tt>' Intrinsic</a>
5826 <div class="doc_text">
5830 declare i8 *@llvm.returnaddress(i32 <level>)
5834 <p>The '<tt>llvm.returnaddress</tt>' intrinsic attempts to compute a
5835 target-specific value indicating the return address of the current function
5836 or one of its callers.</p>
5839 <p>The argument to this intrinsic indicates which function to return the address
5840 for. Zero indicates the calling function, one indicates its caller, etc.
5841 The argument is <b>required</b> to be a constant integer value.</p>
5844 <p>The '<tt>llvm.returnaddress</tt>' intrinsic either returns a pointer
5845 indicating the return address of the specified call frame, or zero if it
5846 cannot be identified. The value returned by this intrinsic is likely to be
5847 incorrect or 0 for arguments other than zero, so it should only be used for
5848 debugging purposes.</p>
5850 <p>Note that calling this intrinsic does not prevent function inlining or other
5851 aggressive transformations, so the value returned may not be that of the
5852 obvious source-language caller.</p>
5856 <!-- _______________________________________________________________________ -->
5857 <div class="doc_subsubsection">
5858 <a name="int_frameaddress">'<tt>llvm.frameaddress</tt>' Intrinsic</a>
5861 <div class="doc_text">
5865 declare i8* @llvm.frameaddress(i32 <level>)
5869 <p>The '<tt>llvm.frameaddress</tt>' intrinsic attempts to return the
5870 target-specific frame pointer value for the specified stack frame.</p>
5873 <p>The argument to this intrinsic indicates which function to return the frame
5874 pointer for. Zero indicates the calling function, one indicates its caller,
5875 etc. The argument is <b>required</b> to be a constant integer value.</p>
5878 <p>The '<tt>llvm.frameaddress</tt>' intrinsic either returns a pointer
5879 indicating the frame address of the specified call frame, or zero if it
5880 cannot be identified. The value returned by this intrinsic is likely to be
5881 incorrect or 0 for arguments other than zero, so it should only be used for
5882 debugging purposes.</p>
5884 <p>Note that calling this intrinsic does not prevent function inlining or other
5885 aggressive transformations, so the value returned may not be that of the
5886 obvious source-language caller.</p>
5890 <!-- _______________________________________________________________________ -->
5891 <div class="doc_subsubsection">
5892 <a name="int_stacksave">'<tt>llvm.stacksave</tt>' Intrinsic</a>
5895 <div class="doc_text">
5899 declare i8* @llvm.stacksave()
5903 <p>The '<tt>llvm.stacksave</tt>' intrinsic is used to remember the current state
5904 of the function stack, for use
5905 with <a href="#int_stackrestore"> <tt>llvm.stackrestore</tt></a>. This is
5906 useful for implementing language features like scoped automatic variable
5907 sized arrays in C99.</p>
5910 <p>This intrinsic returns a opaque pointer value that can be passed
5911 to <a href="#int_stackrestore"><tt>llvm.stackrestore</tt></a>. When
5912 an <tt>llvm.stackrestore</tt> intrinsic is executed with a value saved
5913 from <tt>llvm.stacksave</tt>, it effectively restores the state of the stack
5914 to the state it was in when the <tt>llvm.stacksave</tt> intrinsic executed.
5915 In practice, this pops any <a href="#i_alloca">alloca</a> blocks from the
5916 stack that were allocated after the <tt>llvm.stacksave</tt> was executed.</p>
5920 <!-- _______________________________________________________________________ -->
5921 <div class="doc_subsubsection">
5922 <a name="int_stackrestore">'<tt>llvm.stackrestore</tt>' Intrinsic</a>
5925 <div class="doc_text">
5929 declare void @llvm.stackrestore(i8* %ptr)
5933 <p>The '<tt>llvm.stackrestore</tt>' intrinsic is used to restore the state of
5934 the function stack to the state it was in when the
5935 corresponding <a href="#int_stacksave"><tt>llvm.stacksave</tt></a> intrinsic
5936 executed. This is useful for implementing language features like scoped
5937 automatic variable sized arrays in C99.</p>
5940 <p>See the description
5941 for <a href="#int_stacksave"><tt>llvm.stacksave</tt></a>.</p>
5945 <!-- _______________________________________________________________________ -->
5946 <div class="doc_subsubsection">
5947 <a name="int_prefetch">'<tt>llvm.prefetch</tt>' Intrinsic</a>
5950 <div class="doc_text">
5954 declare void @llvm.prefetch(i8* <address>, i32 <rw>, i32 <locality>)
5958 <p>The '<tt>llvm.prefetch</tt>' intrinsic is a hint to the code generator to
5959 insert a prefetch instruction if supported; otherwise, it is a noop.
5960 Prefetches have no effect on the behavior of the program but can change its
5961 performance characteristics.</p>
5964 <p><tt>address</tt> is the address to be prefetched, <tt>rw</tt> is the
5965 specifier determining if the fetch should be for a read (0) or write (1),
5966 and <tt>locality</tt> is a temporal locality specifier ranging from (0) - no
5967 locality, to (3) - extremely local keep in cache. The <tt>rw</tt>
5968 and <tt>locality</tt> arguments must be constant integers.</p>
5971 <p>This intrinsic does not modify the behavior of the program. In particular,
5972 prefetches cannot trap and do not produce a value. On targets that support
5973 this intrinsic, the prefetch can provide hints to the processor cache for
5974 better performance.</p>
5978 <!-- _______________________________________________________________________ -->
5979 <div class="doc_subsubsection">
5980 <a name="int_pcmarker">'<tt>llvm.pcmarker</tt>' Intrinsic</a>
5983 <div class="doc_text">
5987 declare void @llvm.pcmarker(i32 <id>)
5991 <p>The '<tt>llvm.pcmarker</tt>' intrinsic is a method to export a Program
5992 Counter (PC) in a region of code to simulators and other tools. The method
5993 is target specific, but it is expected that the marker will use exported
5994 symbols to transmit the PC of the marker. The marker makes no guarantees
5995 that it will remain with any specific instruction after optimizations. It is
5996 possible that the presence of a marker will inhibit optimizations. The
5997 intended use is to be inserted after optimizations to allow correlations of
5998 simulation runs.</p>
6001 <p><tt>id</tt> is a numerical id identifying the marker.</p>
6004 <p>This intrinsic does not modify the behavior of the program. Backends that do
6005 not support this intrinsic may ignore it.</p>
6009 <!-- _______________________________________________________________________ -->
6010 <div class="doc_subsubsection">
6011 <a name="int_readcyclecounter">'<tt>llvm.readcyclecounter</tt>' Intrinsic</a>
6014 <div class="doc_text">
6018 declare i64 @llvm.readcyclecounter()
6022 <p>The '<tt>llvm.readcyclecounter</tt>' intrinsic provides access to the cycle
6023 counter register (or similar low latency, high accuracy clocks) on those
6024 targets that support it. On X86, it should map to RDTSC. On Alpha, it
6025 should map to RPCC. As the backing counters overflow quickly (on the order
6026 of 9 seconds on alpha), this should only be used for small timings.</p>
6029 <p>When directly supported, reading the cycle counter should not modify any
6030 memory. Implementations are allowed to either return a application specific
6031 value or a system wide value. On backends without support, this is lowered
6032 to a constant 0.</p>
6036 <!-- ======================================================================= -->
6037 <div class="doc_subsection">
6038 <a name="int_libc">Standard C Library Intrinsics</a>
6041 <div class="doc_text">
6043 <p>LLVM provides intrinsics for a few important standard C library functions.
6044 These intrinsics allow source-language front-ends to pass information about
6045 the alignment of the pointer arguments to the code generator, providing
6046 opportunity for more efficient code generation.</p>
6050 <!-- _______________________________________________________________________ -->
6051 <div class="doc_subsubsection">
6052 <a name="int_memcpy">'<tt>llvm.memcpy</tt>' Intrinsic</a>
6055 <div class="doc_text">
6058 <p>This is an overloaded intrinsic. You can use <tt>llvm.memcpy</tt> on any
6059 integer bit width and for different address spaces. Not all targets support
6060 all bit widths however.</p>
6063 declare void @llvm.memcpy.p0i8.p0i8.i32(i8* <dest>, i8* <src>,
6064 i32 <len>, i32 <align>, i1 <isvolatile>)
6065 declare void @llvm.memcpy.p0i8.p0i8.i64(i8* <dest>, i8* <src>,
6066 i64 <len>, i32 <align>, i1 <isvolatile>)
6070 <p>The '<tt>llvm.memcpy.*</tt>' intrinsics copy a block of memory from the
6071 source location to the destination location.</p>
6073 <p>Note that, unlike the standard libc function, the <tt>llvm.memcpy.*</tt>
6074 intrinsics do not return a value, takes extra alignment/isvolatile arguments
6075 and the pointers can be in specified address spaces.</p>
6079 <p>The first argument is a pointer to the destination, the second is a pointer
6080 to the source. The third argument is an integer argument specifying the
6081 number of bytes to copy, the fourth argument is the alignment of the
6082 source and destination locations, and the fifth is a boolean indicating a
6083 volatile access.</p>
6085 <p>If the call to this intrinsic has an alignment value that is not 0 or 1,
6086 then the caller guarantees that both the source and destination pointers are
6087 aligned to that boundary.</p>
6089 <p>If the <tt>isvolatile</tt> parameter is <tt>true</tt>, the
6090 <tt>llvm.memcpy</tt> call is a <a href="#volatile">volatile operation</a>.
6091 The detailed access behavior is not very cleanly specified and it is unwise
6092 to depend on it.</p>
6096 <p>The '<tt>llvm.memcpy.*</tt>' intrinsics copy a block of memory from the
6097 source location to the destination location, which are not allowed to
6098 overlap. It copies "len" bytes of memory over. If the argument is known to
6099 be aligned to some boundary, this can be specified as the fourth argument,
6100 otherwise it should be set to 0 or 1.</p>
6104 <!-- _______________________________________________________________________ -->
6105 <div class="doc_subsubsection">
6106 <a name="int_memmove">'<tt>llvm.memmove</tt>' Intrinsic</a>
6109 <div class="doc_text">
6112 <p>This is an overloaded intrinsic. You can use llvm.memmove on any integer bit
6113 width and for different address space. Not all targets support all bit
6117 declare void @llvm.memmove.p0i8.p0i8.i32(i8* <dest>, i8* <src>,
6118 i32 <len>, i32 <align>, i1 <isvolatile>)
6119 declare void @llvm.memmove.p0i8.p0i8.i64(i8* <dest>, i8* <src>,
6120 i64 <len>, i32 <align>, i1 <isvolatile>)
6124 <p>The '<tt>llvm.memmove.*</tt>' intrinsics move a block of memory from the
6125 source location to the destination location. It is similar to the
6126 '<tt>llvm.memcpy</tt>' intrinsic but allows the two memory locations to
6129 <p>Note that, unlike the standard libc function, the <tt>llvm.memmove.*</tt>
6130 intrinsics do not return a value, takes extra alignment/isvolatile arguments
6131 and the pointers can be in specified address spaces.</p>
6135 <p>The first argument is a pointer to the destination, the second is a pointer
6136 to the source. The third argument is an integer argument specifying the
6137 number of bytes to copy, the fourth argument is the alignment of the
6138 source and destination locations, and the fifth is a boolean indicating a
6139 volatile access.</p>
6141 <p>If the call to this intrinsic has an alignment value that is not 0 or 1,
6142 then the caller guarantees that the source and destination pointers are
6143 aligned to that boundary.</p>
6145 <p>If the <tt>isvolatile</tt> parameter is <tt>true</tt>, the
6146 <tt>llvm.memmove</tt> call is a <a href="#volatile">volatile operation</a>.
6147 The detailed access behavior is not very cleanly specified and it is unwise
6148 to depend on it.</p>
6152 <p>The '<tt>llvm.memmove.*</tt>' intrinsics copy a block of memory from the
6153 source location to the destination location, which may overlap. It copies
6154 "len" bytes of memory over. If the argument is known to be aligned to some
6155 boundary, this can be specified as the fourth argument, otherwise it should
6156 be set to 0 or 1.</p>
6160 <!-- _______________________________________________________________________ -->
6161 <div class="doc_subsubsection">
6162 <a name="int_memset">'<tt>llvm.memset.*</tt>' Intrinsics</a>
6165 <div class="doc_text">
6168 <p>This is an overloaded intrinsic. You can use llvm.memset on any integer bit
6169 width and for different address spaces. Not all targets support all bit
6173 declare void @llvm.memset.p0i8.i32(i8* <dest>, i8 <val>,
6174 i32 <len>, i32 <align>, i1 <isvolatile>)
6175 declare void @llvm.memset.p0i8.i64(i8* <dest>, i8 <val>,
6176 i64 <len>, i32 <align>, i1 <isvolatile>)
6180 <p>The '<tt>llvm.memset.*</tt>' intrinsics fill a block of memory with a
6181 particular byte value.</p>
6183 <p>Note that, unlike the standard libc function, the <tt>llvm.memset</tt>
6184 intrinsic does not return a value, takes extra alignment/volatile arguments,
6185 and the destination can be in an arbitrary address space.</p>
6188 <p>The first argument is a pointer to the destination to fill, the second is the
6189 byte value to fill it with, the third argument is an integer argument
6190 specifying the number of bytes to fill, and the fourth argument is the known
6191 alignment of destination location.</p>
6193 <p>If the call to this intrinsic has an alignment value that is not 0 or 1,
6194 then the caller guarantees that the destination pointer is aligned to that
6197 <p>If the <tt>isvolatile</tt> parameter is <tt>true</tt>, the
6198 <tt>llvm.memset</tt> call is a <a href="#volatile">volatile operation</a>.
6199 The detailed access behavior is not very cleanly specified and it is unwise
6200 to depend on it.</p>
6203 <p>The '<tt>llvm.memset.*</tt>' intrinsics fill "len" bytes of memory starting
6204 at the destination location. If the argument is known to be aligned to some
6205 boundary, this can be specified as the fourth argument, otherwise it should
6206 be set to 0 or 1.</p>
6210 <!-- _______________________________________________________________________ -->
6211 <div class="doc_subsubsection">
6212 <a name="int_sqrt">'<tt>llvm.sqrt.*</tt>' Intrinsic</a>
6215 <div class="doc_text">
6218 <p>This is an overloaded intrinsic. You can use <tt>llvm.sqrt</tt> on any
6219 floating point or vector of floating point type. Not all targets support all
6223 declare float @llvm.sqrt.f32(float %Val)
6224 declare double @llvm.sqrt.f64(double %Val)
6225 declare x86_fp80 @llvm.sqrt.f80(x86_fp80 %Val)
6226 declare fp128 @llvm.sqrt.f128(fp128 %Val)
6227 declare ppc_fp128 @llvm.sqrt.ppcf128(ppc_fp128 %Val)
6231 <p>The '<tt>llvm.sqrt</tt>' intrinsics return the sqrt of the specified operand,
6232 returning the same value as the libm '<tt>sqrt</tt>' functions would.
6233 Unlike <tt>sqrt</tt> in libm, however, <tt>llvm.sqrt</tt> has undefined
6234 behavior for negative numbers other than -0.0 (which allows for better
6235 optimization, because there is no need to worry about errno being
6236 set). <tt>llvm.sqrt(-0.0)</tt> is defined to return -0.0 like IEEE sqrt.</p>
6239 <p>The argument and return value are floating point numbers of the same
6243 <p>This function returns the sqrt of the specified operand if it is a
6244 nonnegative floating point number.</p>
6248 <!-- _______________________________________________________________________ -->
6249 <div class="doc_subsubsection">
6250 <a name="int_powi">'<tt>llvm.powi.*</tt>' Intrinsic</a>
6253 <div class="doc_text">
6256 <p>This is an overloaded intrinsic. You can use <tt>llvm.powi</tt> on any
6257 floating point or vector of floating point type. Not all targets support all
6261 declare float @llvm.powi.f32(float %Val, i32 %power)
6262 declare double @llvm.powi.f64(double %Val, i32 %power)
6263 declare x86_fp80 @llvm.powi.f80(x86_fp80 %Val, i32 %power)
6264 declare fp128 @llvm.powi.f128(fp128 %Val, i32 %power)
6265 declare ppc_fp128 @llvm.powi.ppcf128(ppc_fp128 %Val, i32 %power)
6269 <p>The '<tt>llvm.powi.*</tt>' intrinsics return the first operand raised to the
6270 specified (positive or negative) power. The order of evaluation of
6271 multiplications is not defined. When a vector of floating point type is
6272 used, the second argument remains a scalar integer value.</p>
6275 <p>The second argument is an integer power, and the first is a value to raise to
6279 <p>This function returns the first value raised to the second power with an
6280 unspecified sequence of rounding operations.</p>
6284 <!-- _______________________________________________________________________ -->
6285 <div class="doc_subsubsection">
6286 <a name="int_sin">'<tt>llvm.sin.*</tt>' Intrinsic</a>
6289 <div class="doc_text">
6292 <p>This is an overloaded intrinsic. You can use <tt>llvm.sin</tt> on any
6293 floating point or vector of floating point type. Not all targets support all
6297 declare float @llvm.sin.f32(float %Val)
6298 declare double @llvm.sin.f64(double %Val)
6299 declare x86_fp80 @llvm.sin.f80(x86_fp80 %Val)
6300 declare fp128 @llvm.sin.f128(fp128 %Val)
6301 declare ppc_fp128 @llvm.sin.ppcf128(ppc_fp128 %Val)
6305 <p>The '<tt>llvm.sin.*</tt>' intrinsics return the sine of the operand.</p>
6308 <p>The argument and return value are floating point numbers of the same
6312 <p>This function returns the sine of the specified operand, returning the same
6313 values as the libm <tt>sin</tt> functions would, and handles error conditions
6314 in the same way.</p>
6318 <!-- _______________________________________________________________________ -->
6319 <div class="doc_subsubsection">
6320 <a name="int_cos">'<tt>llvm.cos.*</tt>' Intrinsic</a>
6323 <div class="doc_text">
6326 <p>This is an overloaded intrinsic. You can use <tt>llvm.cos</tt> on any
6327 floating point or vector of floating point type. Not all targets support all
6331 declare float @llvm.cos.f32(float %Val)
6332 declare double @llvm.cos.f64(double %Val)
6333 declare x86_fp80 @llvm.cos.f80(x86_fp80 %Val)
6334 declare fp128 @llvm.cos.f128(fp128 %Val)
6335 declare ppc_fp128 @llvm.cos.ppcf128(ppc_fp128 %Val)
6339 <p>The '<tt>llvm.cos.*</tt>' intrinsics return the cosine of the operand.</p>
6342 <p>The argument and return value are floating point numbers of the same
6346 <p>This function returns the cosine of the specified operand, returning the same
6347 values as the libm <tt>cos</tt> functions would, and handles error conditions
6348 in the same way.</p>
6352 <!-- _______________________________________________________________________ -->
6353 <div class="doc_subsubsection">
6354 <a name="int_pow">'<tt>llvm.pow.*</tt>' Intrinsic</a>
6357 <div class="doc_text">
6360 <p>This is an overloaded intrinsic. You can use <tt>llvm.pow</tt> on any
6361 floating point or vector of floating point type. Not all targets support all
6365 declare float @llvm.pow.f32(float %Val, float %Power)
6366 declare double @llvm.pow.f64(double %Val, double %Power)
6367 declare x86_fp80 @llvm.pow.f80(x86_fp80 %Val, x86_fp80 %Power)
6368 declare fp128 @llvm.pow.f128(fp128 %Val, fp128 %Power)
6369 declare ppc_fp128 @llvm.pow.ppcf128(ppc_fp128 %Val, ppc_fp128 Power)
6373 <p>The '<tt>llvm.pow.*</tt>' intrinsics return the first operand raised to the
6374 specified (positive or negative) power.</p>
6377 <p>The second argument is a floating point power, and the first is a value to
6378 raise to that power.</p>
6381 <p>This function returns the first value raised to the second power, returning
6382 the same values as the libm <tt>pow</tt> functions would, and handles error
6383 conditions in the same way.</p>
6387 <!-- ======================================================================= -->
6388 <div class="doc_subsection">
6389 <a name="int_manip">Bit Manipulation Intrinsics</a>
6392 <div class="doc_text">
6394 <p>LLVM provides intrinsics for a few important bit manipulation operations.
6395 These allow efficient code generation for some algorithms.</p>
6399 <!-- _______________________________________________________________________ -->
6400 <div class="doc_subsubsection">
6401 <a name="int_bswap">'<tt>llvm.bswap.*</tt>' Intrinsics</a>
6404 <div class="doc_text">
6407 <p>This is an overloaded intrinsic function. You can use bswap on any integer
6408 type that is an even number of bytes (i.e. BitWidth % 16 == 0).</p>
6411 declare i16 @llvm.bswap.i16(i16 <id>)
6412 declare i32 @llvm.bswap.i32(i32 <id>)
6413 declare i64 @llvm.bswap.i64(i64 <id>)
6417 <p>The '<tt>llvm.bswap</tt>' family of intrinsics is used to byte swap integer
6418 values with an even number of bytes (positive multiple of 16 bits). These
6419 are useful for performing operations on data that is not in the target's
6420 native byte order.</p>
6423 <p>The <tt>llvm.bswap.i16</tt> intrinsic returns an i16 value that has the high
6424 and low byte of the input i16 swapped. Similarly,
6425 the <tt>llvm.bswap.i32</tt> intrinsic returns an i32 value that has the four
6426 bytes of the input i32 swapped, so that if the input bytes are numbered 0, 1,
6427 2, 3 then the returned i32 will have its bytes in 3, 2, 1, 0 order.
6428 The <tt>llvm.bswap.i48</tt>, <tt>llvm.bswap.i64</tt> and other intrinsics
6429 extend this concept to additional even-byte lengths (6 bytes, 8 bytes and
6430 more, respectively).</p>
6434 <!-- _______________________________________________________________________ -->
6435 <div class="doc_subsubsection">
6436 <a name="int_ctpop">'<tt>llvm.ctpop.*</tt>' Intrinsic</a>
6439 <div class="doc_text">
6442 <p>This is an overloaded intrinsic. You can use llvm.ctpop on any integer bit
6443 width. Not all targets support all bit widths however.</p>
6446 declare i8 @llvm.ctpop.i8(i8 <src>)
6447 declare i16 @llvm.ctpop.i16(i16 <src>)
6448 declare i32 @llvm.ctpop.i32(i32 <src>)
6449 declare i64 @llvm.ctpop.i64(i64 <src>)
6450 declare i256 @llvm.ctpop.i256(i256 <src>)
6454 <p>The '<tt>llvm.ctpop</tt>' family of intrinsics counts the number of bits set
6458 <p>The only argument is the value to be counted. The argument may be of any
6459 integer type. The return type must match the argument type.</p>
6462 <p>The '<tt>llvm.ctpop</tt>' intrinsic counts the 1's in a variable.</p>
6466 <!-- _______________________________________________________________________ -->
6467 <div class="doc_subsubsection">
6468 <a name="int_ctlz">'<tt>llvm.ctlz.*</tt>' Intrinsic</a>
6471 <div class="doc_text">
6474 <p>This is an overloaded intrinsic. You can use <tt>llvm.ctlz</tt> on any
6475 integer bit width. Not all targets support all bit widths however.</p>
6478 declare i8 @llvm.ctlz.i8 (i8 <src>)
6479 declare i16 @llvm.ctlz.i16(i16 <src>)
6480 declare i32 @llvm.ctlz.i32(i32 <src>)
6481 declare i64 @llvm.ctlz.i64(i64 <src>)
6482 declare i256 @llvm.ctlz.i256(i256 <src>)
6486 <p>The '<tt>llvm.ctlz</tt>' family of intrinsic functions counts the number of
6487 leading zeros in a variable.</p>
6490 <p>The only argument is the value to be counted. The argument may be of any
6491 integer type. The return type must match the argument type.</p>
6494 <p>The '<tt>llvm.ctlz</tt>' intrinsic counts the leading (most significant)
6495 zeros in a variable. If the src == 0 then the result is the size in bits of
6496 the type of src. For example, <tt>llvm.ctlz(i32 2) = 30</tt>.</p>
6500 <!-- _______________________________________________________________________ -->
6501 <div class="doc_subsubsection">
6502 <a name="int_cttz">'<tt>llvm.cttz.*</tt>' Intrinsic</a>
6505 <div class="doc_text">
6508 <p>This is an overloaded intrinsic. You can use <tt>llvm.cttz</tt> on any
6509 integer bit width. Not all targets support all bit widths however.</p>
6512 declare i8 @llvm.cttz.i8 (i8 <src>)
6513 declare i16 @llvm.cttz.i16(i16 <src>)
6514 declare i32 @llvm.cttz.i32(i32 <src>)
6515 declare i64 @llvm.cttz.i64(i64 <src>)
6516 declare i256 @llvm.cttz.i256(i256 <src>)
6520 <p>The '<tt>llvm.cttz</tt>' family of intrinsic functions counts the number of
6524 <p>The only argument is the value to be counted. The argument may be of any
6525 integer type. The return type must match the argument type.</p>
6528 <p>The '<tt>llvm.cttz</tt>' intrinsic counts the trailing (least significant)
6529 zeros in a variable. If the src == 0 then the result is the size in bits of
6530 the type of src. For example, <tt>llvm.cttz(2) = 1</tt>.</p>
6534 <!-- ======================================================================= -->
6535 <div class="doc_subsection">
6536 <a name="int_overflow">Arithmetic with Overflow Intrinsics</a>
6539 <div class="doc_text">
6541 <p>LLVM provides intrinsics for some arithmetic with overflow operations.</p>
6545 <!-- _______________________________________________________________________ -->
6546 <div class="doc_subsubsection">
6547 <a name="int_sadd_overflow">'<tt>llvm.sadd.with.overflow.*</tt>' Intrinsics</a>
6550 <div class="doc_text">
6553 <p>This is an overloaded intrinsic. You can use <tt>llvm.sadd.with.overflow</tt>
6554 on any integer bit width.</p>
6557 declare {i16, i1} @llvm.sadd.with.overflow.i16(i16 %a, i16 %b)
6558 declare {i32, i1} @llvm.sadd.with.overflow.i32(i32 %a, i32 %b)
6559 declare {i64, i1} @llvm.sadd.with.overflow.i64(i64 %a, i64 %b)
6563 <p>The '<tt>llvm.sadd.with.overflow</tt>' family of intrinsic functions perform
6564 a signed addition of the two arguments, and indicate whether an overflow
6565 occurred during the signed summation.</p>
6568 <p>The arguments (%a and %b) and the first element of the result structure may
6569 be of integer types of any bit width, but they must have the same bit
6570 width. The second element of the result structure must be of
6571 type <tt>i1</tt>. <tt>%a</tt> and <tt>%b</tt> are the two values that will
6572 undergo signed addition.</p>
6575 <p>The '<tt>llvm.sadd.with.overflow</tt>' family of intrinsic functions perform
6576 a signed addition of the two variables. They return a structure — the
6577 first element of which is the signed summation, and the second element of
6578 which is a bit specifying if the signed summation resulted in an
6583 %res = call {i32, i1} @llvm.sadd.with.overflow.i32(i32 %a, i32 %b)
6584 %sum = extractvalue {i32, i1} %res, 0
6585 %obit = extractvalue {i32, i1} %res, 1
6586 br i1 %obit, label %overflow, label %normal
6591 <!-- _______________________________________________________________________ -->
6592 <div class="doc_subsubsection">
6593 <a name="int_uadd_overflow">'<tt>llvm.uadd.with.overflow.*</tt>' Intrinsics</a>
6596 <div class="doc_text">
6599 <p>This is an overloaded intrinsic. You can use <tt>llvm.uadd.with.overflow</tt>
6600 on any integer bit width.</p>
6603 declare {i16, i1} @llvm.uadd.with.overflow.i16(i16 %a, i16 %b)
6604 declare {i32, i1} @llvm.uadd.with.overflow.i32(i32 %a, i32 %b)
6605 declare {i64, i1} @llvm.uadd.with.overflow.i64(i64 %a, i64 %b)
6609 <p>The '<tt>llvm.uadd.with.overflow</tt>' family of intrinsic functions perform
6610 an unsigned addition of the two arguments, and indicate whether a carry
6611 occurred during the unsigned summation.</p>
6614 <p>The arguments (%a and %b) and the first element of the result structure may
6615 be of integer types of any bit width, but they must have the same bit
6616 width. The second element of the result structure must be of
6617 type <tt>i1</tt>. <tt>%a</tt> and <tt>%b</tt> are the two values that will
6618 undergo unsigned addition.</p>
6621 <p>The '<tt>llvm.uadd.with.overflow</tt>' family of intrinsic functions perform
6622 an unsigned addition of the two arguments. They return a structure —
6623 the first element of which is the sum, and the second element of which is a
6624 bit specifying if the unsigned summation resulted in a carry.</p>
6628 %res = call {i32, i1} @llvm.uadd.with.overflow.i32(i32 %a, i32 %b)
6629 %sum = extractvalue {i32, i1} %res, 0
6630 %obit = extractvalue {i32, i1} %res, 1
6631 br i1 %obit, label %carry, label %normal
6636 <!-- _______________________________________________________________________ -->
6637 <div class="doc_subsubsection">
6638 <a name="int_ssub_overflow">'<tt>llvm.ssub.with.overflow.*</tt>' Intrinsics</a>
6641 <div class="doc_text">
6644 <p>This is an overloaded intrinsic. You can use <tt>llvm.ssub.with.overflow</tt>
6645 on any integer bit width.</p>
6648 declare {i16, i1} @llvm.ssub.with.overflow.i16(i16 %a, i16 %b)
6649 declare {i32, i1} @llvm.ssub.with.overflow.i32(i32 %a, i32 %b)
6650 declare {i64, i1} @llvm.ssub.with.overflow.i64(i64 %a, i64 %b)
6654 <p>The '<tt>llvm.ssub.with.overflow</tt>' family of intrinsic functions perform
6655 a signed subtraction of the two arguments, and indicate whether an overflow
6656 occurred during the signed subtraction.</p>
6659 <p>The arguments (%a and %b) and the first element of the result structure may
6660 be of integer types of any bit width, but they must have the same bit
6661 width. The second element of the result structure must be of
6662 type <tt>i1</tt>. <tt>%a</tt> and <tt>%b</tt> are the two values that will
6663 undergo signed subtraction.</p>
6666 <p>The '<tt>llvm.ssub.with.overflow</tt>' family of intrinsic functions perform
6667 a signed subtraction of the two arguments. They return a structure —
6668 the first element of which is the subtraction, and the second element of
6669 which is a bit specifying if the signed subtraction resulted in an
6674 %res = call {i32, i1} @llvm.ssub.with.overflow.i32(i32 %a, i32 %b)
6675 %sum = extractvalue {i32, i1} %res, 0
6676 %obit = extractvalue {i32, i1} %res, 1
6677 br i1 %obit, label %overflow, label %normal
6682 <!-- _______________________________________________________________________ -->
6683 <div class="doc_subsubsection">
6684 <a name="int_usub_overflow">'<tt>llvm.usub.with.overflow.*</tt>' Intrinsics</a>
6687 <div class="doc_text">
6690 <p>This is an overloaded intrinsic. You can use <tt>llvm.usub.with.overflow</tt>
6691 on any integer bit width.</p>
6694 declare {i16, i1} @llvm.usub.with.overflow.i16(i16 %a, i16 %b)
6695 declare {i32, i1} @llvm.usub.with.overflow.i32(i32 %a, i32 %b)
6696 declare {i64, i1} @llvm.usub.with.overflow.i64(i64 %a, i64 %b)
6700 <p>The '<tt>llvm.usub.with.overflow</tt>' family of intrinsic functions perform
6701 an unsigned subtraction of the two arguments, and indicate whether an
6702 overflow occurred during the unsigned subtraction.</p>
6705 <p>The arguments (%a and %b) and the first element of the result structure may
6706 be of integer types of any bit width, but they must have the same bit
6707 width. The second element of the result structure must be of
6708 type <tt>i1</tt>. <tt>%a</tt> and <tt>%b</tt> are the two values that will
6709 undergo unsigned subtraction.</p>
6712 <p>The '<tt>llvm.usub.with.overflow</tt>' family of intrinsic functions perform
6713 an unsigned subtraction of the two arguments. They return a structure —
6714 the first element of which is the subtraction, and the second element of
6715 which is a bit specifying if the unsigned subtraction resulted in an
6720 %res = call {i32, i1} @llvm.usub.with.overflow.i32(i32 %a, i32 %b)
6721 %sum = extractvalue {i32, i1} %res, 0
6722 %obit = extractvalue {i32, i1} %res, 1
6723 br i1 %obit, label %overflow, label %normal
6728 <!-- _______________________________________________________________________ -->
6729 <div class="doc_subsubsection">
6730 <a name="int_smul_overflow">'<tt>llvm.smul.with.overflow.*</tt>' Intrinsics</a>
6733 <div class="doc_text">
6736 <p>This is an overloaded intrinsic. You can use <tt>llvm.smul.with.overflow</tt>
6737 on any integer bit width.</p>
6740 declare {i16, i1} @llvm.smul.with.overflow.i16(i16 %a, i16 %b)
6741 declare {i32, i1} @llvm.smul.with.overflow.i32(i32 %a, i32 %b)
6742 declare {i64, i1} @llvm.smul.with.overflow.i64(i64 %a, i64 %b)
6747 <p>The '<tt>llvm.smul.with.overflow</tt>' family of intrinsic functions perform
6748 a signed multiplication of the two arguments, and indicate whether an
6749 overflow occurred during the signed multiplication.</p>
6752 <p>The arguments (%a and %b) and the first element of the result structure may
6753 be of integer types of any bit width, but they must have the same bit
6754 width. The second element of the result structure must be of
6755 type <tt>i1</tt>. <tt>%a</tt> and <tt>%b</tt> are the two values that will
6756 undergo signed multiplication.</p>
6759 <p>The '<tt>llvm.smul.with.overflow</tt>' family of intrinsic functions perform
6760 a signed multiplication of the two arguments. They return a structure —
6761 the first element of which is the multiplication, and the second element of
6762 which is a bit specifying if the signed multiplication resulted in an
6767 %res = call {i32, i1} @llvm.smul.with.overflow.i32(i32 %a, i32 %b)
6768 %sum = extractvalue {i32, i1} %res, 0
6769 %obit = extractvalue {i32, i1} %res, 1
6770 br i1 %obit, label %overflow, label %normal
6775 <!-- _______________________________________________________________________ -->
6776 <div class="doc_subsubsection">
6777 <a name="int_umul_overflow">'<tt>llvm.umul.with.overflow.*</tt>' Intrinsics</a>
6780 <div class="doc_text">
6783 <p>This is an overloaded intrinsic. You can use <tt>llvm.umul.with.overflow</tt>
6784 on any integer bit width.</p>
6787 declare {i16, i1} @llvm.umul.with.overflow.i16(i16 %a, i16 %b)
6788 declare {i32, i1} @llvm.umul.with.overflow.i32(i32 %a, i32 %b)
6789 declare {i64, i1} @llvm.umul.with.overflow.i64(i64 %a, i64 %b)
6793 <p>The '<tt>llvm.umul.with.overflow</tt>' family of intrinsic functions perform
6794 a unsigned multiplication of the two arguments, and indicate whether an
6795 overflow occurred during the unsigned multiplication.</p>
6798 <p>The arguments (%a and %b) and the first element of the result structure may
6799 be of integer types of any bit width, but they must have the same bit
6800 width. The second element of the result structure must be of
6801 type <tt>i1</tt>. <tt>%a</tt> and <tt>%b</tt> are the two values that will
6802 undergo unsigned multiplication.</p>
6805 <p>The '<tt>llvm.umul.with.overflow</tt>' family of intrinsic functions perform
6806 an unsigned multiplication of the two arguments. They return a structure
6807 — the first element of which is the multiplication, and the second
6808 element of which is a bit specifying if the unsigned multiplication resulted
6813 %res = call {i32, i1} @llvm.umul.with.overflow.i32(i32 %a, i32 %b)
6814 %sum = extractvalue {i32, i1} %res, 0
6815 %obit = extractvalue {i32, i1} %res, 1
6816 br i1 %obit, label %overflow, label %normal
6821 <!-- ======================================================================= -->
6822 <div class="doc_subsection">
6823 <a name="int_fp16">Half Precision Floating Point Intrinsics</a>
6826 <div class="doc_text">
6828 <p>Half precision floating point is a storage-only format. This means that it is
6829 a dense encoding (in memory) but does not support computation in the
6832 <p>This means that code must first load the half-precision floating point
6833 value as an i16, then convert it to float with <a
6834 href="#int_convert_from_fp16"><tt>llvm.convert.from.fp16</tt></a>.
6835 Computation can then be performed on the float value (including extending to
6836 double etc). To store the value back to memory, it is first converted to
6837 float if needed, then converted to i16 with
6838 <a href="#int_convert_to_fp16"><tt>llvm.convert.to.fp16</tt></a>, then
6839 storing as an i16 value.</p>
6842 <!-- _______________________________________________________________________ -->
6843 <div class="doc_subsubsection">
6844 <a name="int_convert_to_fp16">'<tt>llvm.convert.to.fp16</tt>' Intrinsic</a>
6847 <div class="doc_text">
6851 declare i16 @llvm.convert.to.fp16(f32 %a)
6855 <p>The '<tt>llvm.convert.to.fp16</tt>' intrinsic function performs
6856 a conversion from single precision floating point format to half precision
6857 floating point format.</p>
6860 <p>The intrinsic function contains single argument - the value to be
6864 <p>The '<tt>llvm.convert.to.fp16</tt>' intrinsic function performs
6865 a conversion from single precision floating point format to half precision
6866 floating point format. The return value is an <tt>i16</tt> which
6867 contains the converted number.</p>
6871 %res = call i16 @llvm.convert.to.fp16(f32 %a)
6872 store i16 %res, i16* @x, align 2
6877 <!-- _______________________________________________________________________ -->
6878 <div class="doc_subsubsection">
6879 <a name="int_convert_from_fp16">'<tt>llvm.convert.from.fp16</tt>' Intrinsic</a>
6882 <div class="doc_text">
6886 declare f32 @llvm.convert.from.fp16(i16 %a)
6890 <p>The '<tt>llvm.convert.from.fp16</tt>' intrinsic function performs
6891 a conversion from half precision floating point format to single precision
6892 floating point format.</p>
6895 <p>The intrinsic function contains single argument - the value to be
6899 <p>The '<tt>llvm.convert.from.fp16</tt>' intrinsic function performs a
6900 conversion from half single precision floating point format to single
6901 precision floating point format. The input half-float value is represented by
6902 an <tt>i16</tt> value.</p>
6906 %a = load i16* @x, align 2
6907 %res = call f32 @llvm.convert.from.fp16(i16 %a)
6912 <!-- ======================================================================= -->
6913 <div class="doc_subsection">
6914 <a name="int_debugger">Debugger Intrinsics</a>
6917 <div class="doc_text">
6919 <p>The LLVM debugger intrinsics (which all start with <tt>llvm.dbg.</tt>
6920 prefix), are described in
6921 the <a href="SourceLevelDebugging.html#format_common_intrinsics">LLVM Source
6922 Level Debugging</a> document.</p>
6926 <!-- ======================================================================= -->
6927 <div class="doc_subsection">
6928 <a name="int_eh">Exception Handling Intrinsics</a>
6931 <div class="doc_text">
6933 <p>The LLVM exception handling intrinsics (which all start with
6934 <tt>llvm.eh.</tt> prefix), are described in
6935 the <a href="ExceptionHandling.html#format_common_intrinsics">LLVM Exception
6936 Handling</a> document.</p>
6940 <!-- ======================================================================= -->
6941 <div class="doc_subsection">
6942 <a name="int_trampoline">Trampoline Intrinsic</a>
6945 <div class="doc_text">
6947 <p>This intrinsic makes it possible to excise one parameter, marked with
6948 the <tt>nest</tt> attribute, from a function. The result is a callable
6949 function pointer lacking the nest parameter - the caller does not need to
6950 provide a value for it. Instead, the value to use is stored in advance in a
6951 "trampoline", a block of memory usually allocated on the stack, which also
6952 contains code to splice the nest value into the argument list. This is used
6953 to implement the GCC nested function address extension.</p>
6955 <p>For example, if the function is
6956 <tt>i32 f(i8* nest %c, i32 %x, i32 %y)</tt> then the resulting function
6957 pointer has signature <tt>i32 (i32, i32)*</tt>. It can be created as
6960 <div class="doc_code">
6962 %tramp = alloca [10 x i8], align 4 ; size and alignment only correct for X86
6963 %tramp1 = getelementptr [10 x i8]* %tramp, i32 0, i32 0
6964 %p = call i8* @llvm.init.trampoline(i8* %tramp1, i8* bitcast (i32 (i8* nest , i32, i32)* @f to i8*), i8* %nval)
6965 %fp = bitcast i8* %p to i32 (i32, i32)*
6969 <p>The call <tt>%val = call i32 %fp(i32 %x, i32 %y)</tt> is then equivalent
6970 to <tt>%val = call i32 %f(i8* %nval, i32 %x, i32 %y)</tt>.</p>
6974 <!-- _______________________________________________________________________ -->
6975 <div class="doc_subsubsection">
6976 <a name="int_it">'<tt>llvm.init.trampoline</tt>' Intrinsic</a>
6979 <div class="doc_text">
6983 declare i8* @llvm.init.trampoline(i8* <tramp>, i8* <func>, i8* <nval>)
6987 <p>This fills the memory pointed to by <tt>tramp</tt> with code and returns a
6988 function pointer suitable for executing it.</p>
6991 <p>The <tt>llvm.init.trampoline</tt> intrinsic takes three arguments, all
6992 pointers. The <tt>tramp</tt> argument must point to a sufficiently large and
6993 sufficiently aligned block of memory; this memory is written to by the
6994 intrinsic. Note that the size and the alignment are target-specific - LLVM
6995 currently provides no portable way of determining them, so a front-end that
6996 generates this intrinsic needs to have some target-specific knowledge.
6997 The <tt>func</tt> argument must hold a function bitcast to
6998 an <tt>i8*</tt>.</p>
7001 <p>The block of memory pointed to by <tt>tramp</tt> is filled with target
7002 dependent code, turning it into a function. A pointer to this function is
7003 returned, but needs to be bitcast to an <a href="#int_trampoline">appropriate
7004 function pointer type</a> before being called. The new function's signature
7005 is the same as that of <tt>func</tt> with any arguments marked with
7006 the <tt>nest</tt> attribute removed. At most one such <tt>nest</tt> argument
7007 is allowed, and it must be of pointer type. Calling the new function is
7008 equivalent to calling <tt>func</tt> with the same argument list, but
7009 with <tt>nval</tt> used for the missing <tt>nest</tt> argument. If, after
7010 calling <tt>llvm.init.trampoline</tt>, the memory pointed to
7011 by <tt>tramp</tt> is modified, then the effect of any later call to the
7012 returned function pointer is undefined.</p>
7016 <!-- ======================================================================= -->
7017 <div class="doc_subsection">
7018 <a name="int_atomics">Atomic Operations and Synchronization Intrinsics</a>
7021 <div class="doc_text">
7023 <p>These intrinsic functions expand the "universal IR" of LLVM to represent
7024 hardware constructs for atomic operations and memory synchronization. This
7025 provides an interface to the hardware, not an interface to the programmer. It
7026 is aimed at a low enough level to allow any programming models or APIs
7027 (Application Programming Interfaces) which need atomic behaviors to map
7028 cleanly onto it. It is also modeled primarily on hardware behavior. Just as
7029 hardware provides a "universal IR" for source languages, it also provides a
7030 starting point for developing a "universal" atomic operation and
7031 synchronization IR.</p>
7033 <p>These do <em>not</em> form an API such as high-level threading libraries,
7034 software transaction memory systems, atomic primitives, and intrinsic
7035 functions as found in BSD, GNU libc, atomic_ops, APR, and other system and
7036 application libraries. The hardware interface provided by LLVM should allow
7037 a clean implementation of all of these APIs and parallel programming models.
7038 No one model or paradigm should be selected above others unless the hardware
7039 itself ubiquitously does so.</p>
7043 <!-- _______________________________________________________________________ -->
7044 <div class="doc_subsubsection">
7045 <a name="int_memory_barrier">'<tt>llvm.memory.barrier</tt>' Intrinsic</a>
7047 <div class="doc_text">
7050 declare void @llvm.memory.barrier(i1 <ll>, i1 <ls>, i1 <sl>, i1 <ss>, i1 <device>)
7054 <p>The <tt>llvm.memory.barrier</tt> intrinsic guarantees ordering between
7055 specific pairs of memory access types.</p>
7058 <p>The <tt>llvm.memory.barrier</tt> intrinsic requires five boolean arguments.
7059 The first four arguments enables a specific barrier as listed below. The
7060 fifth argument specifies that the barrier applies to io or device or uncached
7064 <li><tt>ll</tt>: load-load barrier</li>
7065 <li><tt>ls</tt>: load-store barrier</li>
7066 <li><tt>sl</tt>: store-load barrier</li>
7067 <li><tt>ss</tt>: store-store barrier</li>
7068 <li><tt>device</tt>: barrier applies to device and uncached memory also.</li>
7072 <p>This intrinsic causes the system to enforce some ordering constraints upon
7073 the loads and stores of the program. This barrier does not
7074 indicate <em>when</em> any events will occur, it only enforces
7075 an <em>order</em> in which they occur. For any of the specified pairs of load
7076 and store operations (f.ex. load-load, or store-load), all of the first
7077 operations preceding the barrier will complete before any of the second
7078 operations succeeding the barrier begin. Specifically the semantics for each
7079 pairing is as follows:</p>
7082 <li><tt>ll</tt>: All loads before the barrier must complete before any load
7083 after the barrier begins.</li>
7084 <li><tt>ls</tt>: All loads before the barrier must complete before any
7085 store after the barrier begins.</li>
7086 <li><tt>ss</tt>: All stores before the barrier must complete before any
7087 store after the barrier begins.</li>
7088 <li><tt>sl</tt>: All stores before the barrier must complete before any
7089 load after the barrier begins.</li>
7092 <p>These semantics are applied with a logical "and" behavior when more than one
7093 is enabled in a single memory barrier intrinsic.</p>
7095 <p>Backends may implement stronger barriers than those requested when they do
7096 not support as fine grained a barrier as requested. Some architectures do
7097 not need all types of barriers and on such architectures, these become
7102 %mallocP = tail call i8* @malloc(i32 ptrtoint (i32* getelementptr (i32* null, i32 1) to i32))
7103 %ptr = bitcast i8* %mallocP to i32*
7106 %result1 = load i32* %ptr <i>; yields {i32}:result1 = 4</i>
7107 call void @llvm.memory.barrier(i1 false, i1 true, i1 false, i1 false)
7108 <i>; guarantee the above finishes</i>
7109 store i32 8, %ptr <i>; before this begins</i>
7114 <!-- _______________________________________________________________________ -->
7115 <div class="doc_subsubsection">
7116 <a name="int_atomic_cmp_swap">'<tt>llvm.atomic.cmp.swap.*</tt>' Intrinsic</a>
7119 <div class="doc_text">
7122 <p>This is an overloaded intrinsic. You can use <tt>llvm.atomic.cmp.swap</tt> on
7123 any integer bit width and for different address spaces. Not all targets
7124 support all bit widths however.</p>
7127 declare i8 @llvm.atomic.cmp.swap.i8.p0i8(i8* <ptr>, i8 <cmp>, i8 <val>)
7128 declare i16 @llvm.atomic.cmp.swap.i16.p0i16(i16* <ptr>, i16 <cmp>, i16 <val>)
7129 declare i32 @llvm.atomic.cmp.swap.i32.p0i32(i32* <ptr>, i32 <cmp>, i32 <val>)
7130 declare i64 @llvm.atomic.cmp.swap.i64.p0i64(i64* <ptr>, i64 <cmp>, i64 <val>)
7134 <p>This loads a value in memory and compares it to a given value. If they are
7135 equal, it stores a new value into the memory.</p>
7138 <p>The <tt>llvm.atomic.cmp.swap</tt> intrinsic takes three arguments. The result
7139 as well as both <tt>cmp</tt> and <tt>val</tt> must be integer values with the
7140 same bit width. The <tt>ptr</tt> argument must be a pointer to a value of
7141 this integer type. While any bit width integer may be used, targets may only
7142 lower representations they support in hardware.</p>
7145 <p>This entire intrinsic must be executed atomically. It first loads the value
7146 in memory pointed to by <tt>ptr</tt> and compares it with the
7147 value <tt>cmp</tt>. If they are equal, <tt>val</tt> is stored into the
7148 memory. The loaded value is yielded in all cases. This provides the
7149 equivalent of an atomic compare-and-swap operation within the SSA
7154 %mallocP = tail call i8* @malloc(i32 ptrtoint (i32* getelementptr (i32* null, i32 1) to i32))
7155 %ptr = bitcast i8* %mallocP to i32*
7158 %val1 = add i32 4, 4
7159 %result1 = call i32 @llvm.atomic.cmp.swap.i32.p0i32(i32* %ptr, i32 4, %val1)
7160 <i>; yields {i32}:result1 = 4</i>
7161 %stored1 = icmp eq i32 %result1, 4 <i>; yields {i1}:stored1 = true</i>
7162 %memval1 = load i32* %ptr <i>; yields {i32}:memval1 = 8</i>
7164 %val2 = add i32 1, 1
7165 %result2 = call i32 @llvm.atomic.cmp.swap.i32.p0i32(i32* %ptr, i32 5, %val2)
7166 <i>; yields {i32}:result2 = 8</i>
7167 %stored2 = icmp eq i32 %result2, 5 <i>; yields {i1}:stored2 = false</i>
7169 %memval2 = load i32* %ptr <i>; yields {i32}:memval2 = 8</i>
7174 <!-- _______________________________________________________________________ -->
7175 <div class="doc_subsubsection">
7176 <a name="int_atomic_swap">'<tt>llvm.atomic.swap.*</tt>' Intrinsic</a>
7178 <div class="doc_text">
7181 <p>This is an overloaded intrinsic. You can use <tt>llvm.atomic.swap</tt> on any
7182 integer bit width. Not all targets support all bit widths however.</p>
7185 declare i8 @llvm.atomic.swap.i8.p0i8(i8* <ptr>, i8 <val>)
7186 declare i16 @llvm.atomic.swap.i16.p0i16(i16* <ptr>, i16 <val>)
7187 declare i32 @llvm.atomic.swap.i32.p0i32(i32* <ptr>, i32 <val>)
7188 declare i64 @llvm.atomic.swap.i64.p0i64(i64* <ptr>, i64 <val>)
7192 <p>This intrinsic loads the value stored in memory at <tt>ptr</tt> and yields
7193 the value from memory. It then stores the value in <tt>val</tt> in the memory
7194 at <tt>ptr</tt>.</p>
7197 <p>The <tt>llvm.atomic.swap</tt> intrinsic takes two arguments. Both
7198 the <tt>val</tt> argument and the result must be integers of the same bit
7199 width. The first argument, <tt>ptr</tt>, must be a pointer to a value of this
7200 integer type. The targets may only lower integer representations they
7204 <p>This intrinsic loads the value pointed to by <tt>ptr</tt>, yields it, and
7205 stores <tt>val</tt> back into <tt>ptr</tt> atomically. This provides the
7206 equivalent of an atomic swap operation within the SSA framework.</p>
7210 %mallocP = tail call i8* @malloc(i32 ptrtoint (i32* getelementptr (i32* null, i32 1) to i32))
7211 %ptr = bitcast i8* %mallocP to i32*
7214 %val1 = add i32 4, 4
7215 %result1 = call i32 @llvm.atomic.swap.i32.p0i32(i32* %ptr, i32 %val1)
7216 <i>; yields {i32}:result1 = 4</i>
7217 %stored1 = icmp eq i32 %result1, 4 <i>; yields {i1}:stored1 = true</i>
7218 %memval1 = load i32* %ptr <i>; yields {i32}:memval1 = 8</i>
7220 %val2 = add i32 1, 1
7221 %result2 = call i32 @llvm.atomic.swap.i32.p0i32(i32* %ptr, i32 %val2)
7222 <i>; yields {i32}:result2 = 8</i>
7224 %stored2 = icmp eq i32 %result2, 8 <i>; yields {i1}:stored2 = true</i>
7225 %memval2 = load i32* %ptr <i>; yields {i32}:memval2 = 2</i>
7230 <!-- _______________________________________________________________________ -->
7231 <div class="doc_subsubsection">
7232 <a name="int_atomic_load_add">'<tt>llvm.atomic.load.add.*</tt>' Intrinsic</a>
7236 <div class="doc_text">
7239 <p>This is an overloaded intrinsic. You can use <tt>llvm.atomic.load.add</tt> on
7240 any integer bit width. Not all targets support all bit widths however.</p>
7243 declare i8 @llvm.atomic.load.add.i8.p0i8(i8* <ptr>, i8 <delta>)
7244 declare i16 @llvm.atomic.load.add.i16.p0i16(i16* <ptr>, i16 <delta>)
7245 declare i32 @llvm.atomic.load.add.i32.p0i32(i32* <ptr>, i32 <delta>)
7246 declare i64 @llvm.atomic.load.add.i64.p0i64(i64* <ptr>, i64 <delta>)
7250 <p>This intrinsic adds <tt>delta</tt> to the value stored in memory
7251 at <tt>ptr</tt>. It yields the original value at <tt>ptr</tt>.</p>
7254 <p>The intrinsic takes two arguments, the first a pointer to an integer value
7255 and the second an integer value. The result is also an integer value. These
7256 integer types can have any bit width, but they must all have the same bit
7257 width. The targets may only lower integer representations they support.</p>
7260 <p>This intrinsic does a series of operations atomically. It first loads the
7261 value stored at <tt>ptr</tt>. It then adds <tt>delta</tt>, stores the result
7262 to <tt>ptr</tt>. It yields the original value stored at <tt>ptr</tt>.</p>
7266 %mallocP = tail call i8* @malloc(i32 ptrtoint (i32* getelementptr (i32* null, i32 1) to i32))
7267 %ptr = bitcast i8* %mallocP to i32*
7269 %result1 = call i32 @llvm.atomic.load.add.i32.p0i32(i32* %ptr, i32 4)
7270 <i>; yields {i32}:result1 = 4</i>
7271 %result2 = call i32 @llvm.atomic.load.add.i32.p0i32(i32* %ptr, i32 2)
7272 <i>; yields {i32}:result2 = 8</i>
7273 %result3 = call i32 @llvm.atomic.load.add.i32.p0i32(i32* %ptr, i32 5)
7274 <i>; yields {i32}:result3 = 10</i>
7275 %memval1 = load i32* %ptr <i>; yields {i32}:memval1 = 15</i>
7280 <!-- _______________________________________________________________________ -->
7281 <div class="doc_subsubsection">
7282 <a name="int_atomic_load_sub">'<tt>llvm.atomic.load.sub.*</tt>' Intrinsic</a>
7286 <div class="doc_text">
7289 <p>This is an overloaded intrinsic. You can use <tt>llvm.atomic.load.sub</tt> on
7290 any integer bit width and for different address spaces. Not all targets
7291 support all bit widths however.</p>
7294 declare i8 @llvm.atomic.load.sub.i8.p0i32(i8* <ptr>, i8 <delta>)
7295 declare i16 @llvm.atomic.load.sub.i16.p0i32(i16* <ptr>, i16 <delta>)
7296 declare i32 @llvm.atomic.load.sub.i32.p0i32(i32* <ptr>, i32 <delta>)
7297 declare i64 @llvm.atomic.load.sub.i64.p0i32(i64* <ptr>, i64 <delta>)
7301 <p>This intrinsic subtracts <tt>delta</tt> to the value stored in memory at
7302 <tt>ptr</tt>. It yields the original value at <tt>ptr</tt>.</p>
7305 <p>The intrinsic takes two arguments, the first a pointer to an integer value
7306 and the second an integer value. The result is also an integer value. These
7307 integer types can have any bit width, but they must all have the same bit
7308 width. The targets may only lower integer representations they support.</p>
7311 <p>This intrinsic does a series of operations atomically. It first loads the
7312 value stored at <tt>ptr</tt>. It then subtracts <tt>delta</tt>, stores the
7313 result to <tt>ptr</tt>. It yields the original value stored
7314 at <tt>ptr</tt>.</p>
7318 %mallocP = tail call i8* @malloc(i32 ptrtoint (i32* getelementptr (i32* null, i32 1) to i32))
7319 %ptr = bitcast i8* %mallocP to i32*
7321 %result1 = call i32 @llvm.atomic.load.sub.i32.p0i32(i32* %ptr, i32 4)
7322 <i>; yields {i32}:result1 = 8</i>
7323 %result2 = call i32 @llvm.atomic.load.sub.i32.p0i32(i32* %ptr, i32 2)
7324 <i>; yields {i32}:result2 = 4</i>
7325 %result3 = call i32 @llvm.atomic.load.sub.i32.p0i32(i32* %ptr, i32 5)
7326 <i>; yields {i32}:result3 = 2</i>
7327 %memval1 = load i32* %ptr <i>; yields {i32}:memval1 = -3</i>
7332 <!-- _______________________________________________________________________ -->
7333 <div class="doc_subsubsection">
7334 <a name="int_atomic_load_and">'<tt>llvm.atomic.load.and.*</tt>' Intrinsic</a><br>
7335 <a name="int_atomic_load_nand">'<tt>llvm.atomic.load.nand.*</tt>' Intrinsic</a><br>
7336 <a name="int_atomic_load_or">'<tt>llvm.atomic.load.or.*</tt>' Intrinsic</a><br>
7337 <a name="int_atomic_load_xor">'<tt>llvm.atomic.load.xor.*</tt>' Intrinsic</a><br>
7340 <div class="doc_text">
7343 <p>These are overloaded intrinsics. You can
7344 use <tt>llvm.atomic.load_and</tt>, <tt>llvm.atomic.load_nand</tt>,
7345 <tt>llvm.atomic.load_or</tt>, and <tt>llvm.atomic.load_xor</tt> on any integer
7346 bit width and for different address spaces. Not all targets support all bit
7350 declare i8 @llvm.atomic.load.and.i8.p0i8(i8* <ptr>, i8 <delta>)
7351 declare i16 @llvm.atomic.load.and.i16.p0i16(i16* <ptr>, i16 <delta>)
7352 declare i32 @llvm.atomic.load.and.i32.p0i32(i32* <ptr>, i32 <delta>)
7353 declare i64 @llvm.atomic.load.and.i64.p0i64(i64* <ptr>, i64 <delta>)
7357 declare i8 @llvm.atomic.load.or.i8.p0i8(i8* <ptr>, i8 <delta>)
7358 declare i16 @llvm.atomic.load.or.i16.p0i16(i16* <ptr>, i16 <delta>)
7359 declare i32 @llvm.atomic.load.or.i32.p0i32(i32* <ptr>, i32 <delta>)
7360 declare i64 @llvm.atomic.load.or.i64.p0i64(i64* <ptr>, i64 <delta>)
7364 declare i8 @llvm.atomic.load.nand.i8.p0i32(i8* <ptr>, i8 <delta>)
7365 declare i16 @llvm.atomic.load.nand.i16.p0i32(i16* <ptr>, i16 <delta>)
7366 declare i32 @llvm.atomic.load.nand.i32.p0i32(i32* <ptr>, i32 <delta>)
7367 declare i64 @llvm.atomic.load.nand.i64.p0i32(i64* <ptr>, i64 <delta>)
7371 declare i8 @llvm.atomic.load.xor.i8.p0i32(i8* <ptr>, i8 <delta>)
7372 declare i16 @llvm.atomic.load.xor.i16.p0i32(i16* <ptr>, i16 <delta>)
7373 declare i32 @llvm.atomic.load.xor.i32.p0i32(i32* <ptr>, i32 <delta>)
7374 declare i64 @llvm.atomic.load.xor.i64.p0i32(i64* <ptr>, i64 <delta>)
7378 <p>These intrinsics bitwise the operation (and, nand, or, xor) <tt>delta</tt> to
7379 the value stored in memory at <tt>ptr</tt>. It yields the original value
7380 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 bitwise
7391 operation <tt>delta</tt>, store the result to <tt>ptr</tt>. They yield the
7392 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*
7398 store i32 0x0F0F, %ptr
7399 %result0 = call i32 @llvm.atomic.load.nand.i32.p0i32(i32* %ptr, i32 0xFF)
7400 <i>; yields {i32}:result0 = 0x0F0F</i>
7401 %result1 = call i32 @llvm.atomic.load.and.i32.p0i32(i32* %ptr, i32 0xFF)
7402 <i>; yields {i32}:result1 = 0xFFFFFFF0</i>
7403 %result2 = call i32 @llvm.atomic.load.or.i32.p0i32(i32* %ptr, i32 0F)
7404 <i>; yields {i32}:result2 = 0xF0</i>
7405 %result3 = call i32 @llvm.atomic.load.xor.i32.p0i32(i32* %ptr, i32 0F)
7406 <i>; yields {i32}:result3 = FF</i>
7407 %memval1 = load i32* %ptr <i>; yields {i32}:memval1 = F0</i>
7412 <!-- _______________________________________________________________________ -->
7413 <div class="doc_subsubsection">
7414 <a name="int_atomic_load_max">'<tt>llvm.atomic.load.max.*</tt>' Intrinsic</a><br>
7415 <a name="int_atomic_load_min">'<tt>llvm.atomic.load.min.*</tt>' Intrinsic</a><br>
7416 <a name="int_atomic_load_umax">'<tt>llvm.atomic.load.umax.*</tt>' Intrinsic</a><br>
7417 <a name="int_atomic_load_umin">'<tt>llvm.atomic.load.umin.*</tt>' Intrinsic</a><br>
7420 <div class="doc_text">
7423 <p>These are overloaded intrinsics. You can use <tt>llvm.atomic.load_max</tt>,
7424 <tt>llvm.atomic.load_min</tt>, <tt>llvm.atomic.load_umax</tt>, and
7425 <tt>llvm.atomic.load_umin</tt> on any integer bit width and for different
7426 address spaces. Not all targets support all bit widths however.</p>
7429 declare i8 @llvm.atomic.load.max.i8.p0i8(i8* <ptr>, i8 <delta>)
7430 declare i16 @llvm.atomic.load.max.i16.p0i16(i16* <ptr>, i16 <delta>)
7431 declare i32 @llvm.atomic.load.max.i32.p0i32(i32* <ptr>, i32 <delta>)
7432 declare i64 @llvm.atomic.load.max.i64.p0i64(i64* <ptr>, i64 <delta>)
7436 declare i8 @llvm.atomic.load.min.i8.p0i8(i8* <ptr>, i8 <delta>)
7437 declare i16 @llvm.atomic.load.min.i16.p0i16(i16* <ptr>, i16 <delta>)
7438 declare i32 @llvm.atomic.load.min.i32.p0i32(i32* <ptr>, i32 <delta>)
7439 declare i64 @llvm.atomic.load.min.i64.p0i64(i64* <ptr>, i64 <delta>)
7443 declare i8 @llvm.atomic.load.umax.i8.p0i8(i8* <ptr>, i8 <delta>)
7444 declare i16 @llvm.atomic.load.umax.i16.p0i16(i16* <ptr>, i16 <delta>)
7445 declare i32 @llvm.atomic.load.umax.i32.p0i32(i32* <ptr>, i32 <delta>)
7446 declare i64 @llvm.atomic.load.umax.i64.p0i64(i64* <ptr>, i64 <delta>)
7450 declare i8 @llvm.atomic.load.umin.i8.p0i8(i8* <ptr>, i8 <delta>)
7451 declare i16 @llvm.atomic.load.umin.i16.p0i16(i16* <ptr>, i16 <delta>)
7452 declare i32 @llvm.atomic.load.umin.i32.p0i32(i32* <ptr>, i32 <delta>)
7453 declare i64 @llvm.atomic.load.umin.i64.p0i64(i64* <ptr>, i64 <delta>)
7457 <p>These intrinsics takes the signed or unsigned minimum or maximum of
7458 <tt>delta</tt> and the value stored in memory at <tt>ptr</tt>. It yields the
7459 original value at <tt>ptr</tt>.</p>
7462 <p>These intrinsics take two arguments, the first a pointer to an integer value
7463 and the second an integer value. The result is also an integer value. These
7464 integer types can have any bit width, but they must all have the same bit
7465 width. The targets may only lower integer representations they support.</p>
7468 <p>These intrinsics does a series of operations atomically. They first load the
7469 value stored at <tt>ptr</tt>. They then do the signed or unsigned min or
7470 max <tt>delta</tt> and the value, store the result to <tt>ptr</tt>. They
7471 yield the original value stored at <tt>ptr</tt>.</p>
7475 %mallocP = tail call i8* @malloc(i32 ptrtoint (i32* getelementptr (i32* null, i32 1) to i32))
7476 %ptr = bitcast i8* %mallocP to i32*
7478 %result0 = call i32 @llvm.atomic.load.min.i32.p0i32(i32* %ptr, i32 -2)
7479 <i>; yields {i32}:result0 = 7</i>
7480 %result1 = call i32 @llvm.atomic.load.max.i32.p0i32(i32* %ptr, i32 8)
7481 <i>; yields {i32}:result1 = -2</i>
7482 %result2 = call i32 @llvm.atomic.load.umin.i32.p0i32(i32* %ptr, i32 10)
7483 <i>; yields {i32}:result2 = 8</i>
7484 %result3 = call i32 @llvm.atomic.load.umax.i32.p0i32(i32* %ptr, i32 30)
7485 <i>; yields {i32}:result3 = 8</i>
7486 %memval1 = load i32* %ptr <i>; yields {i32}:memval1 = 30</i>
7492 <!-- ======================================================================= -->
7493 <div class="doc_subsection">
7494 <a name="int_memorymarkers">Memory Use Markers</a>
7497 <div class="doc_text">
7499 <p>This class of intrinsics exists to information about the lifetime of memory
7500 objects and ranges where variables are immutable.</p>
7504 <!-- _______________________________________________________________________ -->
7505 <div class="doc_subsubsection">
7506 <a name="int_lifetime_start">'<tt>llvm.lifetime.start</tt>' Intrinsic</a>
7509 <div class="doc_text">
7513 declare void @llvm.lifetime.start(i64 <size>, i8* nocapture <ptr>)
7517 <p>The '<tt>llvm.lifetime.start</tt>' intrinsic specifies the start of a memory
7518 object's lifetime.</p>
7521 <p>The first argument is a constant integer representing the size of the
7522 object, or -1 if it is variable sized. The second argument is a pointer to
7526 <p>This intrinsic indicates that before this point in the code, the value of the
7527 memory pointed to by <tt>ptr</tt> is dead. This means that it is known to
7528 never be used and has an undefined value. A load from the pointer that
7529 precedes this intrinsic can be replaced with
7530 <tt>'<a href="#undefvalues">undef</a>'</tt>.</p>
7534 <!-- _______________________________________________________________________ -->
7535 <div class="doc_subsubsection">
7536 <a name="int_lifetime_end">'<tt>llvm.lifetime.end</tt>' Intrinsic</a>
7539 <div class="doc_text">
7543 declare void @llvm.lifetime.end(i64 <size>, i8* nocapture <ptr>)
7547 <p>The '<tt>llvm.lifetime.end</tt>' intrinsic specifies the end of a memory
7548 object's lifetime.</p>
7551 <p>The first argument is a constant integer representing the size of the
7552 object, or -1 if it is variable sized. The second argument is a pointer to
7556 <p>This intrinsic indicates that after this point in the code, the value of the
7557 memory pointed to by <tt>ptr</tt> is dead. This means that it is known to
7558 never be used and has an undefined value. Any stores into the memory object
7559 following this intrinsic may be removed as dead.
7563 <!-- _______________________________________________________________________ -->
7564 <div class="doc_subsubsection">
7565 <a name="int_invariant_start">'<tt>llvm.invariant.start</tt>' Intrinsic</a>
7568 <div class="doc_text">
7572 declare {}* @llvm.invariant.start(i64 <size>, i8* nocapture <ptr>) readonly
7576 <p>The '<tt>llvm.invariant.start</tt>' intrinsic specifies that the contents of
7577 a memory object will not change.</p>
7580 <p>The first argument is a constant integer representing the size of the
7581 object, or -1 if it is variable sized. The second argument is a pointer to
7585 <p>This intrinsic indicates that until an <tt>llvm.invariant.end</tt> that uses
7586 the return value, the referenced memory location is constant and
7591 <!-- _______________________________________________________________________ -->
7592 <div class="doc_subsubsection">
7593 <a name="int_invariant_end">'<tt>llvm.invariant.end</tt>' Intrinsic</a>
7596 <div class="doc_text">
7600 declare void @llvm.invariant.end({}* <start>, i64 <size>, i8* nocapture <ptr>)
7604 <p>The '<tt>llvm.invariant.end</tt>' intrinsic specifies that the contents of
7605 a memory object are mutable.</p>
7608 <p>The first argument is the matching <tt>llvm.invariant.start</tt> intrinsic.
7609 The second argument is a constant integer representing the size of the
7610 object, or -1 if it is variable sized and the third argument is a pointer
7614 <p>This intrinsic indicates that the memory is mutable again.</p>
7618 <!-- ======================================================================= -->
7619 <div class="doc_subsection">
7620 <a name="int_general">General Intrinsics</a>
7623 <div class="doc_text">
7625 <p>This class of intrinsics is designed to be generic and has no specific
7630 <!-- _______________________________________________________________________ -->
7631 <div class="doc_subsubsection">
7632 <a name="int_var_annotation">'<tt>llvm.var.annotation</tt>' Intrinsic</a>
7635 <div class="doc_text">
7639 declare void @llvm.var.annotation(i8* <val>, i8* <str>, i8* <str>, i32 <int>)
7643 <p>The '<tt>llvm.var.annotation</tt>' intrinsic.</p>
7646 <p>The first argument is a pointer to a value, the second is a pointer to a
7647 global string, the third is a pointer to a global string which is the source
7648 file name, and the last argument is the line number.</p>
7651 <p>This intrinsic allows annotation of local variables with arbitrary strings.
7652 This can be useful for special purpose optimizations that want to look for
7653 these annotations. These have no other defined use, they are ignored by code
7654 generation and optimization.</p>
7658 <!-- _______________________________________________________________________ -->
7659 <div class="doc_subsubsection">
7660 <a name="int_annotation">'<tt>llvm.annotation.*</tt>' Intrinsic</a>
7663 <div class="doc_text">
7666 <p>This is an overloaded intrinsic. You can use '<tt>llvm.annotation</tt>' on
7667 any integer bit width.</p>
7670 declare i8 @llvm.annotation.i8(i8 <val>, i8* <str>, i8* <str>, i32 <int>)
7671 declare i16 @llvm.annotation.i16(i16 <val>, i8* <str>, i8* <str>, i32 <int>)
7672 declare i32 @llvm.annotation.i32(i32 <val>, i8* <str>, i8* <str>, i32 <int>)
7673 declare i64 @llvm.annotation.i64(i64 <val>, i8* <str>, i8* <str>, i32 <int>)
7674 declare i256 @llvm.annotation.i256(i256 <val>, i8* <str>, i8* <str>, i32 <int>)
7678 <p>The '<tt>llvm.annotation</tt>' intrinsic.</p>
7681 <p>The first argument is an integer value (result of some expression), the
7682 second is a pointer to a global string, the third is a pointer to a global
7683 string which is the source file name, and the last argument is the line
7684 number. It returns the value of the first argument.</p>
7687 <p>This intrinsic allows annotations to be put on arbitrary expressions with
7688 arbitrary strings. This can be useful for special purpose optimizations that
7689 want to look for these annotations. These have no other defined use, they
7690 are ignored by code generation and optimization.</p>
7694 <!-- _______________________________________________________________________ -->
7695 <div class="doc_subsubsection">
7696 <a name="int_trap">'<tt>llvm.trap</tt>' Intrinsic</a>
7699 <div class="doc_text">
7703 declare void @llvm.trap()
7707 <p>The '<tt>llvm.trap</tt>' intrinsic.</p>
7713 <p>This intrinsics is lowered to the target dependent trap instruction. If the
7714 target does not have a trap instruction, this intrinsic will be lowered to
7715 the call of the <tt>abort()</tt> function.</p>
7719 <!-- _______________________________________________________________________ -->
7720 <div class="doc_subsubsection">
7721 <a name="int_stackprotector">'<tt>llvm.stackprotector</tt>' Intrinsic</a>
7724 <div class="doc_text">
7728 declare void @llvm.stackprotector(i8* <guard>, i8** <slot>)
7732 <p>The <tt>llvm.stackprotector</tt> intrinsic takes the <tt>guard</tt> and
7733 stores it onto the stack at <tt>slot</tt>. The stack slot is adjusted to
7734 ensure that it is placed on the stack before local variables.</p>
7737 <p>The <tt>llvm.stackprotector</tt> intrinsic requires two pointer
7738 arguments. The first argument is the value loaded from the stack
7739 guard <tt>@__stack_chk_guard</tt>. The second variable is an <tt>alloca</tt>
7740 that has enough space to hold the value of the guard.</p>
7743 <p>This intrinsic causes the prologue/epilogue inserter to force the position of
7744 the <tt>AllocaInst</tt> stack slot to be before local variables on the
7745 stack. This is to ensure that if a local variable on the stack is
7746 overwritten, it will destroy the value of the guard. When the function exits,
7747 the guard on the stack is checked against the original guard. If they're
7748 different, then the program aborts by calling the <tt>__stack_chk_fail()</tt>
7753 <!-- _______________________________________________________________________ -->
7754 <div class="doc_subsubsection">
7755 <a name="int_objectsize">'<tt>llvm.objectsize</tt>' Intrinsic</a>
7758 <div class="doc_text">
7762 declare i32 @llvm.objectsize.i32(i8* <object>, i1 <type>)
7763 declare i64 @llvm.objectsize.i64(i8* <object>, i1 <type>)
7767 <p>The <tt>llvm.objectsize</tt> intrinsic is designed to provide information
7768 to the optimizers to discover at compile time either a) when an
7769 operation like memcpy will either overflow a buffer that corresponds to
7770 an object, or b) to determine that a runtime check for overflow isn't
7771 necessary. An object in this context means an allocation of a
7772 specific class, structure, array, or other object.</p>
7775 <p>The <tt>llvm.objectsize</tt> intrinsic takes two arguments. The first
7776 argument is a pointer to or into the <tt>object</tt>. The second argument
7777 is a boolean 0 or 1. This argument determines whether you want the
7778 maximum (0) or minimum (1) bytes remaining. This needs to be a literal 0 or
7779 1, variables are not allowed.</p>
7782 <p>The <tt>llvm.objectsize</tt> intrinsic is lowered to either a constant
7783 representing the size of the object concerned or <tt>i32/i64 -1 or 0</tt>
7784 (depending on the <tt>type</tt> argument if the size cannot be determined
7785 at compile time.</p>
7789 <!-- *********************************************************************** -->
7792 <a href="http://jigsaw.w3.org/css-validator/check/referer"><img
7793 src="http://jigsaw.w3.org/css-validator/images/vcss-blue" alt="Valid CSS"></a>
7794 <a href="http://validator.w3.org/check/referer"><img
7795 src="http://www.w3.org/Icons/valid-html401-blue" alt="Valid HTML 4.01"></a>
7797 <a href="mailto:sabre@nondot.org">Chris Lattner</a><br>
7798 <a href="http://llvm.org">The LLVM Compiler Infrastructure</a><br>
7799 Last modified: $Date$