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 <h1>LLVM Language Reference Manual</h1>
17 <li><a href="#abstract">Abstract</a></li>
18 <li><a href="#introduction">Introduction</a></li>
19 <li><a href="#identifiers">Identifiers</a></li>
20 <li><a href="#highlevel">High Level Structure</a>
22 <li><a href="#modulestructure">Module Structure</a></li>
23 <li><a href="#linkage">Linkage Types</a>
25 <li><a href="#linkage_private">'<tt>private</tt>' Linkage</a></li>
26 <li><a href="#linkage_linker_private">'<tt>linker_private</tt>' Linkage</a></li>
27 <li><a href="#linkage_linker_private_weak">'<tt>linker_private_weak</tt>' Linkage</a></li>
28 <li><a href="#linkage_linker_private_weak_def_auto">'<tt>linker_private_weak_def_auto</tt>' Linkage</a></li>
29 <li><a href="#linkage_internal">'<tt>internal</tt>' Linkage</a></li>
30 <li><a href="#linkage_available_externally">'<tt>available_externally</tt>' Linkage</a></li>
31 <li><a href="#linkage_linkonce">'<tt>linkonce</tt>' Linkage</a></li>
32 <li><a href="#linkage_common">'<tt>common</tt>' Linkage</a></li>
33 <li><a href="#linkage_weak">'<tt>weak</tt>' Linkage</a></li>
34 <li><a href="#linkage_appending">'<tt>appending</tt>' Linkage</a></li>
35 <li><a href="#linkage_externweak">'<tt>extern_weak</tt>' Linkage</a></li>
36 <li><a href="#linkage_linkonce_odr">'<tt>linkonce_odr</tt>' Linkage</a></li>
37 <li><a href="#linkage_weak">'<tt>weak_odr</tt>' Linkage</a></li>
38 <li><a href="#linkage_external">'<tt>external</tt>' Linkage</a></li>
39 <li><a href="#linkage_dllimport">'<tt>dllimport</tt>' Linkage</a></li>
40 <li><a href="#linkage_dllexport">'<tt>dllexport</tt>' Linkage</a></li>
43 <li><a href="#callingconv">Calling Conventions</a></li>
44 <li><a href="#namedtypes">Named Types</a></li>
45 <li><a href="#globalvars">Global Variables</a></li>
46 <li><a href="#functionstructure">Functions</a></li>
47 <li><a href="#aliasstructure">Aliases</a></li>
48 <li><a href="#namedmetadatastructure">Named Metadata</a></li>
49 <li><a href="#paramattrs">Parameter Attributes</a></li>
50 <li><a href="#fnattrs">Function Attributes</a></li>
51 <li><a href="#gc">Garbage Collector Names</a></li>
52 <li><a href="#moduleasm">Module-Level Inline Assembly</a></li>
53 <li><a href="#datalayout">Data Layout</a></li>
54 <li><a href="#pointeraliasing">Pointer Aliasing Rules</a></li>
55 <li><a href="#volatile">Volatile Memory Accesses</a></li>
56 <li><a href="#memmodel">Memory Model for Concurrent Operations</a></li>
57 <li><a href="#ordering">Atomic Memory Ordering Constraints</a></li>
60 <li><a href="#typesystem">Type System</a>
62 <li><a href="#t_classifications">Type Classifications</a></li>
63 <li><a href="#t_primitive">Primitive Types</a>
65 <li><a href="#t_integer">Integer Type</a></li>
66 <li><a href="#t_floating">Floating Point Types</a></li>
67 <li><a href="#t_x86mmx">X86mmx Type</a></li>
68 <li><a href="#t_void">Void Type</a></li>
69 <li><a href="#t_label">Label Type</a></li>
70 <li><a href="#t_metadata">Metadata Type</a></li>
73 <li><a href="#t_derived">Derived Types</a>
75 <li><a href="#t_aggregate">Aggregate Types</a>
77 <li><a href="#t_array">Array Type</a></li>
78 <li><a href="#t_struct">Structure Type</a></li>
79 <li><a href="#t_opaque">Opaque Structure Types</a></li>
80 <li><a href="#t_vector">Vector Type</a></li>
83 <li><a href="#t_function">Function Type</a></li>
84 <li><a href="#t_pointer">Pointer Type</a></li>
89 <li><a href="#constants">Constants</a>
91 <li><a href="#simpleconstants">Simple Constants</a></li>
92 <li><a href="#complexconstants">Complex Constants</a></li>
93 <li><a href="#globalconstants">Global Variable and Function Addresses</a></li>
94 <li><a href="#undefvalues">Undefined Values</a></li>
95 <li><a href="#poisonvalues">Poison Values</a></li>
96 <li><a href="#blockaddress">Addresses of Basic Blocks</a></li>
97 <li><a href="#constantexprs">Constant Expressions</a></li>
100 <li><a href="#othervalues">Other Values</a>
102 <li><a href="#inlineasm">Inline Assembler Expressions</a></li>
103 <li><a href="#metadata">Metadata Nodes and Metadata Strings</a>
105 <li><a href="#tbaa">'<tt>tbaa</tt>' Metadata</a></li>
106 <li><a href="#fpaccuracy">'<tt>fpaccuracy</tt>' Metadata</a></li>
109 <li><a href="#module_flags">Module Flags Metadata</a>
111 <li><a href="#objc_metadata">Objective-C Metadata</a></li>
116 <li><a href="#intrinsic_globals">Intrinsic Global Variables</a>
118 <li><a href="#intg_used">The '<tt>llvm.used</tt>' Global Variable</a></li>
119 <li><a href="#intg_compiler_used">The '<tt>llvm.compiler.used</tt>'
120 Global Variable</a></li>
121 <li><a href="#intg_global_ctors">The '<tt>llvm.global_ctors</tt>'
122 Global Variable</a></li>
123 <li><a href="#intg_global_dtors">The '<tt>llvm.global_dtors</tt>'
124 Global Variable</a></li>
127 <li><a href="#instref">Instruction Reference</a>
129 <li><a href="#terminators">Terminator Instructions</a>
131 <li><a href="#i_ret">'<tt>ret</tt>' Instruction</a></li>
132 <li><a href="#i_br">'<tt>br</tt>' Instruction</a></li>
133 <li><a href="#i_switch">'<tt>switch</tt>' Instruction</a></li>
134 <li><a href="#i_indirectbr">'<tt>indirectbr</tt>' Instruction</a></li>
135 <li><a href="#i_invoke">'<tt>invoke</tt>' Instruction</a></li>
136 <li><a href="#i_unwind">'<tt>unwind</tt>' Instruction</a></li>
137 <li><a href="#i_resume">'<tt>resume</tt>' Instruction</a></li>
138 <li><a href="#i_unreachable">'<tt>unreachable</tt>' Instruction</a></li>
141 <li><a href="#binaryops">Binary Operations</a>
143 <li><a href="#i_add">'<tt>add</tt>' Instruction</a></li>
144 <li><a href="#i_fadd">'<tt>fadd</tt>' Instruction</a></li>
145 <li><a href="#i_sub">'<tt>sub</tt>' Instruction</a></li>
146 <li><a href="#i_fsub">'<tt>fsub</tt>' Instruction</a></li>
147 <li><a href="#i_mul">'<tt>mul</tt>' Instruction</a></li>
148 <li><a href="#i_fmul">'<tt>fmul</tt>' Instruction</a></li>
149 <li><a href="#i_udiv">'<tt>udiv</tt>' Instruction</a></li>
150 <li><a href="#i_sdiv">'<tt>sdiv</tt>' Instruction</a></li>
151 <li><a href="#i_fdiv">'<tt>fdiv</tt>' Instruction</a></li>
152 <li><a href="#i_urem">'<tt>urem</tt>' Instruction</a></li>
153 <li><a href="#i_srem">'<tt>srem</tt>' Instruction</a></li>
154 <li><a href="#i_frem">'<tt>frem</tt>' Instruction</a></li>
157 <li><a href="#bitwiseops">Bitwise Binary Operations</a>
159 <li><a href="#i_shl">'<tt>shl</tt>' Instruction</a></li>
160 <li><a href="#i_lshr">'<tt>lshr</tt>' Instruction</a></li>
161 <li><a href="#i_ashr">'<tt>ashr</tt>' Instruction</a></li>
162 <li><a href="#i_and">'<tt>and</tt>' Instruction</a></li>
163 <li><a href="#i_or">'<tt>or</tt>' Instruction</a></li>
164 <li><a href="#i_xor">'<tt>xor</tt>' Instruction</a></li>
167 <li><a href="#vectorops">Vector Operations</a>
169 <li><a href="#i_extractelement">'<tt>extractelement</tt>' Instruction</a></li>
170 <li><a href="#i_insertelement">'<tt>insertelement</tt>' Instruction</a></li>
171 <li><a href="#i_shufflevector">'<tt>shufflevector</tt>' Instruction</a></li>
174 <li><a href="#aggregateops">Aggregate Operations</a>
176 <li><a href="#i_extractvalue">'<tt>extractvalue</tt>' Instruction</a></li>
177 <li><a href="#i_insertvalue">'<tt>insertvalue</tt>' Instruction</a></li>
180 <li><a href="#memoryops">Memory Access and Addressing Operations</a>
182 <li><a href="#i_alloca">'<tt>alloca</tt>' Instruction</a></li>
183 <li><a href="#i_load">'<tt>load</tt>' Instruction</a></li>
184 <li><a href="#i_store">'<tt>store</tt>' Instruction</a></li>
185 <li><a href="#i_fence">'<tt>fence</tt>' Instruction</a></li>
186 <li><a href="#i_cmpxchg">'<tt>cmpxchg</tt>' Instruction</a></li>
187 <li><a href="#i_atomicrmw">'<tt>atomicrmw</tt>' Instruction</a></li>
188 <li><a href="#i_getelementptr">'<tt>getelementptr</tt>' Instruction</a></li>
191 <li><a href="#convertops">Conversion Operations</a>
193 <li><a href="#i_trunc">'<tt>trunc .. to</tt>' Instruction</a></li>
194 <li><a href="#i_zext">'<tt>zext .. to</tt>' Instruction</a></li>
195 <li><a href="#i_sext">'<tt>sext .. to</tt>' Instruction</a></li>
196 <li><a href="#i_fptrunc">'<tt>fptrunc .. to</tt>' Instruction</a></li>
197 <li><a href="#i_fpext">'<tt>fpext .. to</tt>' Instruction</a></li>
198 <li><a href="#i_fptoui">'<tt>fptoui .. to</tt>' Instruction</a></li>
199 <li><a href="#i_fptosi">'<tt>fptosi .. to</tt>' Instruction</a></li>
200 <li><a href="#i_uitofp">'<tt>uitofp .. to</tt>' Instruction</a></li>
201 <li><a href="#i_sitofp">'<tt>sitofp .. to</tt>' Instruction</a></li>
202 <li><a href="#i_ptrtoint">'<tt>ptrtoint .. to</tt>' Instruction</a></li>
203 <li><a href="#i_inttoptr">'<tt>inttoptr .. to</tt>' Instruction</a></li>
204 <li><a href="#i_bitcast">'<tt>bitcast .. to</tt>' Instruction</a></li>
207 <li><a href="#otherops">Other Operations</a>
209 <li><a href="#i_icmp">'<tt>icmp</tt>' Instruction</a></li>
210 <li><a href="#i_fcmp">'<tt>fcmp</tt>' Instruction</a></li>
211 <li><a href="#i_phi">'<tt>phi</tt>' Instruction</a></li>
212 <li><a href="#i_select">'<tt>select</tt>' Instruction</a></li>
213 <li><a href="#i_call">'<tt>call</tt>' Instruction</a></li>
214 <li><a href="#i_va_arg">'<tt>va_arg</tt>' Instruction</a></li>
215 <li><a href="#i_landingpad">'<tt>landingpad</tt>' Instruction</a></li>
220 <li><a href="#intrinsics">Intrinsic Functions</a>
222 <li><a href="#int_varargs">Variable Argument Handling Intrinsics</a>
224 <li><a href="#int_va_start">'<tt>llvm.va_start</tt>' Intrinsic</a></li>
225 <li><a href="#int_va_end">'<tt>llvm.va_end</tt>' Intrinsic</a></li>
226 <li><a href="#int_va_copy">'<tt>llvm.va_copy</tt>' Intrinsic</a></li>
229 <li><a href="#int_gc">Accurate Garbage Collection Intrinsics</a>
231 <li><a href="#int_gcroot">'<tt>llvm.gcroot</tt>' Intrinsic</a></li>
232 <li><a href="#int_gcread">'<tt>llvm.gcread</tt>' Intrinsic</a></li>
233 <li><a href="#int_gcwrite">'<tt>llvm.gcwrite</tt>' Intrinsic</a></li>
236 <li><a href="#int_codegen">Code Generator Intrinsics</a>
238 <li><a href="#int_returnaddress">'<tt>llvm.returnaddress</tt>' Intrinsic</a></li>
239 <li><a href="#int_frameaddress">'<tt>llvm.frameaddress</tt>' Intrinsic</a></li>
240 <li><a href="#int_stacksave">'<tt>llvm.stacksave</tt>' Intrinsic</a></li>
241 <li><a href="#int_stackrestore">'<tt>llvm.stackrestore</tt>' Intrinsic</a></li>
242 <li><a href="#int_prefetch">'<tt>llvm.prefetch</tt>' Intrinsic</a></li>
243 <li><a href="#int_pcmarker">'<tt>llvm.pcmarker</tt>' Intrinsic</a></li>
244 <li><a href="#int_readcyclecounter">'<tt>llvm.readcyclecounter</tt>' Intrinsic</a></li>
247 <li><a href="#int_libc">Standard C Library Intrinsics</a>
249 <li><a href="#int_memcpy">'<tt>llvm.memcpy.*</tt>' Intrinsic</a></li>
250 <li><a href="#int_memmove">'<tt>llvm.memmove.*</tt>' Intrinsic</a></li>
251 <li><a href="#int_memset">'<tt>llvm.memset.*</tt>' Intrinsic</a></li>
252 <li><a href="#int_sqrt">'<tt>llvm.sqrt.*</tt>' Intrinsic</a></li>
253 <li><a href="#int_powi">'<tt>llvm.powi.*</tt>' Intrinsic</a></li>
254 <li><a href="#int_sin">'<tt>llvm.sin.*</tt>' Intrinsic</a></li>
255 <li><a href="#int_cos">'<tt>llvm.cos.*</tt>' Intrinsic</a></li>
256 <li><a href="#int_pow">'<tt>llvm.pow.*</tt>' Intrinsic</a></li>
257 <li><a href="#int_exp">'<tt>llvm.exp.*</tt>' Intrinsic</a></li>
258 <li><a href="#int_log">'<tt>llvm.log.*</tt>' Intrinsic</a></li>
259 <li><a href="#int_fma">'<tt>llvm.fma.*</tt>' Intrinsic</a></li>
262 <li><a href="#int_manip">Bit Manipulation Intrinsics</a>
264 <li><a href="#int_bswap">'<tt>llvm.bswap.*</tt>' Intrinsics</a></li>
265 <li><a href="#int_ctpop">'<tt>llvm.ctpop.*</tt>' Intrinsic </a></li>
266 <li><a href="#int_ctlz">'<tt>llvm.ctlz.*</tt>' Intrinsic </a></li>
267 <li><a href="#int_cttz">'<tt>llvm.cttz.*</tt>' Intrinsic </a></li>
270 <li><a href="#int_overflow">Arithmetic with Overflow Intrinsics</a>
272 <li><a href="#int_sadd_overflow">'<tt>llvm.sadd.with.overflow.*</tt> Intrinsics</a></li>
273 <li><a href="#int_uadd_overflow">'<tt>llvm.uadd.with.overflow.*</tt> Intrinsics</a></li>
274 <li><a href="#int_ssub_overflow">'<tt>llvm.ssub.with.overflow.*</tt> Intrinsics</a></li>
275 <li><a href="#int_usub_overflow">'<tt>llvm.usub.with.overflow.*</tt> Intrinsics</a></li>
276 <li><a href="#int_smul_overflow">'<tt>llvm.smul.with.overflow.*</tt> Intrinsics</a></li>
277 <li><a href="#int_umul_overflow">'<tt>llvm.umul.with.overflow.*</tt> Intrinsics</a></li>
280 <li><a href="#int_fp16">Half Precision Floating Point Intrinsics</a>
282 <li><a href="#int_convert_to_fp16">'<tt>llvm.convert.to.fp16</tt>' Intrinsic</a></li>
283 <li><a href="#int_convert_from_fp16">'<tt>llvm.convert.from.fp16</tt>' Intrinsic</a></li>
286 <li><a href="#int_debugger">Debugger intrinsics</a></li>
287 <li><a href="#int_eh">Exception Handling intrinsics</a></li>
288 <li><a href="#int_trampoline">Trampoline Intrinsics</a>
290 <li><a href="#int_it">'<tt>llvm.init.trampoline</tt>' Intrinsic</a></li>
291 <li><a href="#int_at">'<tt>llvm.adjust.trampoline</tt>' Intrinsic</a></li>
294 <li><a href="#int_memorymarkers">Memory Use Markers</a>
296 <li><a href="#int_lifetime_start">'<tt>llvm.lifetime.start</tt>' Intrinsic</a></li>
297 <li><a href="#int_lifetime_end">'<tt>llvm.lifetime.end</tt>' Intrinsic</a></li>
298 <li><a href="#int_invariant_start">'<tt>llvm.invariant.start</tt>' Intrinsic</a></li>
299 <li><a href="#int_invariant_end">'<tt>llvm.invariant.end</tt>' Intrinsic</a></li>
302 <li><a href="#int_general">General intrinsics</a>
304 <li><a href="#int_var_annotation">
305 '<tt>llvm.var.annotation</tt>' Intrinsic</a></li>
306 <li><a href="#int_annotation">
307 '<tt>llvm.annotation.*</tt>' Intrinsic</a></li>
308 <li><a href="#int_trap">
309 '<tt>llvm.trap</tt>' Intrinsic</a></li>
310 <li><a href="#int_stackprotector">
311 '<tt>llvm.stackprotector</tt>' Intrinsic</a></li>
312 <li><a href="#int_objectsize">
313 '<tt>llvm.objectsize</tt>' Intrinsic</a></li>
314 <li><a href="#int_expect">
315 '<tt>llvm.expect</tt>' Intrinsic</a></li>
322 <div class="doc_author">
323 <p>Written by <a href="mailto:sabre@nondot.org">Chris Lattner</a>
324 and <a href="mailto:vadve@cs.uiuc.edu">Vikram Adve</a></p>
327 <!-- *********************************************************************** -->
328 <h2><a name="abstract">Abstract</a></h2>
329 <!-- *********************************************************************** -->
333 <p>This document is a reference manual for the LLVM assembly language. LLVM is
334 a Static Single Assignment (SSA) based representation that provides type
335 safety, low-level operations, flexibility, and the capability of representing
336 'all' high-level languages cleanly. It is the common code representation
337 used throughout all phases of the LLVM compilation strategy.</p>
341 <!-- *********************************************************************** -->
342 <h2><a name="introduction">Introduction</a></h2>
343 <!-- *********************************************************************** -->
347 <p>The LLVM code representation is designed to be used in three different forms:
348 as an in-memory compiler IR, as an on-disk bitcode representation (suitable
349 for fast loading by a Just-In-Time compiler), and as a human readable
350 assembly language representation. This allows LLVM to provide a powerful
351 intermediate representation for efficient compiler transformations and
352 analysis, while providing a natural means to debug and visualize the
353 transformations. The three different forms of LLVM are all equivalent. This
354 document describes the human readable representation and notation.</p>
356 <p>The LLVM representation aims to be light-weight and low-level while being
357 expressive, typed, and extensible at the same time. It aims to be a
358 "universal IR" of sorts, by being at a low enough level that high-level ideas
359 may be cleanly mapped to it (similar to how microprocessors are "universal
360 IR's", allowing many source languages to be mapped to them). By providing
361 type information, LLVM can be used as the target of optimizations: for
362 example, through pointer analysis, it can be proven that a C automatic
363 variable is never accessed outside of the current function, allowing it to
364 be promoted to a simple SSA value instead of a memory location.</p>
366 <!-- _______________________________________________________________________ -->
368 <a name="wellformed">Well-Formedness</a>
373 <p>It is important to note that this document describes 'well formed' LLVM
374 assembly language. There is a difference between what the parser accepts and
375 what is considered 'well formed'. For example, the following instruction is
376 syntactically okay, but not well formed:</p>
378 <pre class="doc_code">
379 %x = <a href="#i_add">add</a> i32 1, %x
382 <p>because the definition of <tt>%x</tt> does not dominate all of its uses. The
383 LLVM infrastructure provides a verification pass that may be used to verify
384 that an LLVM module is well formed. This pass is automatically run by the
385 parser after parsing input assembly and by the optimizer before it outputs
386 bitcode. The violations pointed out by the verifier pass indicate bugs in
387 transformation passes or input to the parser.</p>
393 <!-- Describe the typesetting conventions here. -->
395 <!-- *********************************************************************** -->
396 <h2><a name="identifiers">Identifiers</a></h2>
397 <!-- *********************************************************************** -->
401 <p>LLVM identifiers come in two basic types: global and local. Global
402 identifiers (functions, global variables) begin with the <tt>'@'</tt>
403 character. Local identifiers (register names, types) begin with
404 the <tt>'%'</tt> character. Additionally, there are three different formats
405 for identifiers, for different purposes:</p>
408 <li>Named values are represented as a string of characters with their prefix.
409 For example, <tt>%foo</tt>, <tt>@DivisionByZero</tt>,
410 <tt>%a.really.long.identifier</tt>. The actual regular expression used is
411 '<tt>[%@][a-zA-Z$._][a-zA-Z$._0-9]*</tt>'. Identifiers which require
412 other characters in their names can be surrounded with quotes. Special
413 characters may be escaped using <tt>"\xx"</tt> where <tt>xx</tt> is the
414 ASCII code for the character in hexadecimal. In this way, any character
415 can be used in a name value, even quotes themselves.</li>
417 <li>Unnamed values are represented as an unsigned numeric value with their
418 prefix. For example, <tt>%12</tt>, <tt>@2</tt>, <tt>%44</tt>.</li>
420 <li>Constants, which are described in a <a href="#constants">section about
421 constants</a>, below.</li>
424 <p>LLVM requires that values start with a prefix for two reasons: Compilers
425 don't need to worry about name clashes with reserved words, and the set of
426 reserved words may be expanded in the future without penalty. Additionally,
427 unnamed identifiers allow a compiler to quickly come up with a temporary
428 variable without having to avoid symbol table conflicts.</p>
430 <p>Reserved words in LLVM are very similar to reserved words in other
431 languages. There are keywords for different opcodes
432 ('<tt><a href="#i_add">add</a></tt>',
433 '<tt><a href="#i_bitcast">bitcast</a></tt>',
434 '<tt><a href="#i_ret">ret</a></tt>', etc...), for primitive type names
435 ('<tt><a href="#t_void">void</a></tt>',
436 '<tt><a href="#t_primitive">i32</a></tt>', etc...), and others. These
437 reserved words cannot conflict with variable names, because none of them
438 start with a prefix character (<tt>'%'</tt> or <tt>'@'</tt>).</p>
440 <p>Here is an example of LLVM code to multiply the integer variable
441 '<tt>%X</tt>' by 8:</p>
445 <pre class="doc_code">
446 %result = <a href="#i_mul">mul</a> i32 %X, 8
449 <p>After strength reduction:</p>
451 <pre class="doc_code">
452 %result = <a href="#i_shl">shl</a> i32 %X, i8 3
455 <p>And the hard way:</p>
457 <pre 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
463 <p>This last way of multiplying <tt>%X</tt> by 8 illustrates several important
464 lexical features of LLVM:</p>
467 <li>Comments are delimited with a '<tt>;</tt>' and go until the end of
470 <li>Unnamed temporaries are created when the result of a computation is not
471 assigned to a named value.</li>
473 <li>Unnamed temporaries are numbered sequentially</li>
476 <p>It also shows a convention that we follow in this document. When
477 demonstrating instructions, we will follow an instruction with a comment that
478 defines the type and name of value produced. Comments are shown in italic
483 <!-- *********************************************************************** -->
484 <h2><a name="highlevel">High Level Structure</a></h2>
485 <!-- *********************************************************************** -->
487 <!-- ======================================================================= -->
489 <a name="modulestructure">Module Structure</a>
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 <pre class="doc_code">
502 <i>; Declare the string constant as a global constant.</i>
503 <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>
505 <i>; External declaration of the puts function</i>
506 <a href="#functionstructure">declare</a> i32 @puts(i8*) <i>; i32 (i8*)* </i>
508 <i>; Definition of main function</i>
509 define i32 @main() { <i>; i32()* </i>
510 <i>; Convert [13 x i8]* to i8 *...</i>
511 %cast210 = <a href="#i_getelementptr">getelementptr</a> [13 x i8]* @.LC0, i64 0, i64 0 <i>; i8*</i>
513 <i>; Call puts function to write out the string to stdout.</i>
514 <a href="#i_call">call</a> i32 @puts(i8* %cast210) <i>; i32</i>
515 <a href="#i_ret">ret</a> i32 0
518 <i>; Named metadata</i>
519 !1 = metadata !{i32 41}
523 <p>This example is made up of a <a href="#globalvars">global variable</a> named
524 "<tt>.LC0</tt>", an external declaration of the "<tt>puts</tt>" function,
525 a <a href="#functionstructure">function definition</a> for
526 "<tt>main</tt>" and <a href="#namedmetadatastructure">named metadata</a>
529 <p>In general, a module is made up of a list of global values, where both
530 functions and global variables are global values. Global values are
531 represented by a pointer to a memory location (in this case, a pointer to an
532 array of char, and a pointer to a function), and have one of the
533 following <a href="#linkage">linkage types</a>.</p>
537 <!-- ======================================================================= -->
539 <a name="linkage">Linkage Types</a>
544 <p>All Global Variables and Functions have one of the following types of
548 <dt><tt><b><a name="linkage_private">private</a></b></tt></dt>
549 <dd>Global values with "<tt>private</tt>" linkage are only directly accessible
550 by objects in the current module. In particular, linking code into a
551 module with an private global value may cause the private to be renamed as
552 necessary to avoid collisions. Because the symbol is private to the
553 module, all references can be updated. This doesn't show up in any symbol
554 table in the object file.</dd>
556 <dt><tt><b><a name="linkage_linker_private">linker_private</a></b></tt></dt>
557 <dd>Similar to <tt>private</tt>, but the symbol is passed through the
558 assembler and evaluated by the linker. Unlike normal strong symbols, they
559 are removed by the linker from the final linked image (executable or
560 dynamic library).</dd>
562 <dt><tt><b><a name="linkage_linker_private_weak">linker_private_weak</a></b></tt></dt>
563 <dd>Similar to "<tt>linker_private</tt>", but the symbol is weak. Note that
564 <tt>linker_private_weak</tt> symbols are subject to coalescing by the
565 linker. The symbols are removed by the linker from the final linked image
566 (executable or dynamic library).</dd>
568 <dt><tt><b><a name="linkage_linker_private_weak_def_auto">linker_private_weak_def_auto</a></b></tt></dt>
569 <dd>Similar to "<tt>linker_private_weak</tt>", but it's known that the address
570 of the object is not taken. For instance, functions that had an inline
571 definition, but the compiler decided not to inline it. Note,
572 unlike <tt>linker_private</tt> and <tt>linker_private_weak</tt>,
573 <tt>linker_private_weak_def_auto</tt> may have only <tt>default</tt>
574 visibility. The symbols are removed by the linker from the final linked
575 image (executable or dynamic library).</dd>
577 <dt><tt><b><a name="linkage_internal">internal</a></b></tt></dt>
578 <dd>Similar to private, but the value shows as a local symbol
579 (<tt>STB_LOCAL</tt> in the case of ELF) in the object file. This
580 corresponds to the notion of the '<tt>static</tt>' keyword in C.</dd>
582 <dt><tt><b><a name="linkage_available_externally">available_externally</a></b></tt></dt>
583 <dd>Globals with "<tt>available_externally</tt>" linkage are never emitted
584 into the object file corresponding to the LLVM module. They exist to
585 allow inlining and other optimizations to take place given knowledge of
586 the definition of the global, which is known to be somewhere outside the
587 module. Globals with <tt>available_externally</tt> linkage are allowed to
588 be discarded at will, and are otherwise the same as <tt>linkonce_odr</tt>.
589 This linkage type is only allowed on definitions, not declarations.</dd>
591 <dt><tt><b><a name="linkage_linkonce">linkonce</a></b></tt></dt>
592 <dd>Globals with "<tt>linkonce</tt>" linkage are merged with other globals of
593 the same name when linkage occurs. This can be used to implement
594 some forms of inline functions, templates, or other code which must be
595 generated in each translation unit that uses it, but where the body may
596 be overridden with a more definitive definition later. Unreferenced
597 <tt>linkonce</tt> globals are allowed to be discarded. Note that
598 <tt>linkonce</tt> linkage does not actually allow the optimizer to
599 inline the body of this function into callers because it doesn't know if
600 this definition of the function is the definitive definition within the
601 program or whether it will be overridden by a stronger definition.
602 To enable inlining and other optimizations, use "<tt>linkonce_odr</tt>"
605 <dt><tt><b><a name="linkage_weak">weak</a></b></tt></dt>
606 <dd>"<tt>weak</tt>" linkage has the same merging semantics as
607 <tt>linkonce</tt> linkage, except that unreferenced globals with
608 <tt>weak</tt> linkage may not be discarded. This is used for globals that
609 are declared "weak" in C source code.</dd>
611 <dt><tt><b><a name="linkage_common">common</a></b></tt></dt>
612 <dd>"<tt>common</tt>" linkage is most similar to "<tt>weak</tt>" linkage, but
613 they are used for tentative definitions in C, such as "<tt>int X;</tt>" at
615 Symbols with "<tt>common</tt>" linkage are merged in the same way as
616 <tt>weak symbols</tt>, and they may not be deleted if unreferenced.
617 <tt>common</tt> symbols may not have an explicit section,
618 must have a zero initializer, and may not be marked '<a
619 href="#globalvars"><tt>constant</tt></a>'. Functions and aliases may not
620 have common linkage.</dd>
623 <dt><tt><b><a name="linkage_appending">appending</a></b></tt></dt>
624 <dd>"<tt>appending</tt>" linkage may only be applied to global variables of
625 pointer to array type. When two global variables with appending linkage
626 are linked together, the two global arrays are appended together. This is
627 the LLVM, typesafe, equivalent of having the system linker append together
628 "sections" with identical names when .o files are linked.</dd>
630 <dt><tt><b><a name="linkage_externweak">extern_weak</a></b></tt></dt>
631 <dd>The semantics of this linkage follow the ELF object file model: the symbol
632 is weak until linked, if not linked, the symbol becomes null instead of
633 being an undefined reference.</dd>
635 <dt><tt><b><a name="linkage_linkonce_odr">linkonce_odr</a></b></tt></dt>
636 <dt><tt><b><a name="linkage_weak_odr">weak_odr</a></b></tt></dt>
637 <dd>Some languages allow differing globals to be merged, such as two functions
638 with different semantics. Other languages, such as <tt>C++</tt>, ensure
639 that only equivalent globals are ever merged (the "one definition rule"
640 — "ODR"). Such languages can use the <tt>linkonce_odr</tt>
641 and <tt>weak_odr</tt> linkage types to indicate that the global will only
642 be merged with equivalent globals. These linkage types are otherwise the
643 same as their non-<tt>odr</tt> versions.</dd>
645 <dt><tt><b><a name="linkage_external">external</a></b></tt></dt>
646 <dd>If none of the above identifiers are used, the global is externally
647 visible, meaning that it participates in linkage and can be used to
648 resolve external symbol references.</dd>
651 <p>The next two types of linkage are targeted for Microsoft Windows platform
652 only. They are designed to support importing (exporting) symbols from (to)
653 DLLs (Dynamic Link Libraries).</p>
656 <dt><tt><b><a name="linkage_dllimport">dllimport</a></b></tt></dt>
657 <dd>"<tt>dllimport</tt>" linkage causes the compiler to reference a function
658 or variable via a global pointer to a pointer that is set up by the DLL
659 exporting the symbol. On Microsoft Windows targets, the pointer name is
660 formed by combining <code>__imp_</code> and the function or variable
663 <dt><tt><b><a name="linkage_dllexport">dllexport</a></b></tt></dt>
664 <dd>"<tt>dllexport</tt>" linkage causes the compiler to provide a global
665 pointer to a pointer in a DLL, so that it can be referenced with the
666 <tt>dllimport</tt> attribute. On Microsoft Windows targets, the pointer
667 name is formed by combining <code>__imp_</code> and the function or
671 <p>For example, since the "<tt>.LC0</tt>" variable is defined to be internal, if
672 another module defined a "<tt>.LC0</tt>" variable and was linked with this
673 one, one of the two would be renamed, preventing a collision. Since
674 "<tt>main</tt>" and "<tt>puts</tt>" are external (i.e., lacking any linkage
675 declarations), they are accessible outside of the current module.</p>
677 <p>It is illegal for a function <i>declaration</i> to have any linkage type
678 other than <tt>external</tt>, <tt>dllimport</tt>
679 or <tt>extern_weak</tt>.</p>
681 <p>Aliases can have only <tt>external</tt>, <tt>internal</tt>, <tt>weak</tt>
682 or <tt>weak_odr</tt> linkages.</p>
686 <!-- ======================================================================= -->
688 <a name="callingconv">Calling Conventions</a>
693 <p>LLVM <a href="#functionstructure">functions</a>, <a href="#i_call">calls</a>
694 and <a href="#i_invoke">invokes</a> can all have an optional calling
695 convention specified for the call. The calling convention of any pair of
696 dynamic caller/callee must match, or the behavior of the program is
697 undefined. The following calling conventions are supported by LLVM, and more
698 may be added in the future:</p>
701 <dt><b>"<tt>ccc</tt>" - The C calling convention</b>:</dt>
702 <dd>This calling convention (the default if no other calling convention is
703 specified) matches the target C calling conventions. This calling
704 convention supports varargs function calls and tolerates some mismatch in
705 the declared prototype and implemented declaration of the function (as
708 <dt><b>"<tt>fastcc</tt>" - The fast calling convention</b>:</dt>
709 <dd>This calling convention attempts to make calls as fast as possible
710 (e.g. by passing things in registers). This calling convention allows the
711 target to use whatever tricks it wants to produce fast code for the
712 target, without having to conform to an externally specified ABI
713 (Application Binary Interface).
714 <a href="CodeGenerator.html#tailcallopt">Tail calls can only be optimized
715 when this or the GHC convention is used.</a> This calling convention
716 does not support varargs and requires the prototype of all callees to
717 exactly match the prototype of the function definition.</dd>
719 <dt><b>"<tt>coldcc</tt>" - The cold calling convention</b>:</dt>
720 <dd>This calling convention attempts to make code in the caller as efficient
721 as possible under the assumption that the call is not commonly executed.
722 As such, these calls often preserve all registers so that the call does
723 not break any live ranges in the caller side. This calling convention
724 does not support varargs and requires the prototype of all callees to
725 exactly match the prototype of the function definition.</dd>
727 <dt><b>"<tt>cc <em>10</em></tt>" - GHC convention</b>:</dt>
728 <dd>This calling convention has been implemented specifically for use by the
729 <a href="http://www.haskell.org/ghc">Glasgow Haskell Compiler (GHC)</a>.
730 It passes everything in registers, going to extremes to achieve this by
731 disabling callee save registers. This calling convention should not be
732 used lightly but only for specific situations such as an alternative to
733 the <em>register pinning</em> performance technique often used when
734 implementing functional programming languages.At the moment only X86
735 supports this convention and it has the following limitations:
737 <li>On <em>X86-32</em> only supports up to 4 bit type parameters. No
738 floating point types are supported.</li>
739 <li>On <em>X86-64</em> only supports up to 10 bit type parameters and
740 6 floating point parameters.</li>
742 This calling convention supports
743 <a href="CodeGenerator.html#tailcallopt">tail call optimization</a> but
744 requires both the caller and callee are using it.
747 <dt><b>"<tt>cc <<em>n</em>></tt>" - Numbered convention</b>:</dt>
748 <dd>Any calling convention may be specified by number, allowing
749 target-specific calling conventions to be used. Target specific calling
750 conventions start at 64.</dd>
753 <p>More calling conventions can be added/defined on an as-needed basis, to
754 support Pascal conventions or any other well-known target-independent
759 <!-- ======================================================================= -->
761 <a name="visibility">Visibility Styles</a>
766 <p>All Global Variables and Functions have one of the following visibility
770 <dt><b>"<tt>default</tt>" - Default style</b>:</dt>
771 <dd>On targets that use the ELF object file format, default visibility means
772 that the declaration is visible to other modules and, in shared libraries,
773 means that the declared entity may be overridden. On Darwin, default
774 visibility means that the declaration is visible to other modules. Default
775 visibility corresponds to "external linkage" in the language.</dd>
777 <dt><b>"<tt>hidden</tt>" - Hidden style</b>:</dt>
778 <dd>Two declarations of an object with hidden visibility refer to the same
779 object if they are in the same shared object. Usually, hidden visibility
780 indicates that the symbol will not be placed into the dynamic symbol
781 table, so no other module (executable or shared library) can reference it
784 <dt><b>"<tt>protected</tt>" - Protected style</b>:</dt>
785 <dd>On ELF, protected visibility indicates that the symbol will be placed in
786 the dynamic symbol table, but that references within the defining module
787 will bind to the local symbol. That is, the symbol cannot be overridden by
793 <!-- ======================================================================= -->
795 <a name="namedtypes">Named Types</a>
800 <p>LLVM IR allows you to specify name aliases for certain types. This can make
801 it easier to read the IR and make the IR more condensed (particularly when
802 recursive types are involved). An example of a name specification is:</p>
804 <pre class="doc_code">
805 %mytype = type { %mytype*, i32 }
808 <p>You may give a name to any <a href="#typesystem">type</a> except
809 "<a href="#t_void">void</a>". Type name aliases may be used anywhere a type
810 is expected with the syntax "%mytype".</p>
812 <p>Note that type names are aliases for the structural type that they indicate,
813 and that you can therefore specify multiple names for the same type. This
814 often leads to confusing behavior when dumping out a .ll file. Since LLVM IR
815 uses structural typing, the name is not part of the type. When printing out
816 LLVM IR, the printer will pick <em>one name</em> to render all types of a
817 particular shape. This means that if you have code where two different
818 source types end up having the same LLVM type, that the dumper will sometimes
819 print the "wrong" or unexpected type. This is an important design point and
820 isn't going to change.</p>
824 <!-- ======================================================================= -->
826 <a name="globalvars">Global Variables</a>
831 <p>Global variables define regions of memory allocated at compilation time
832 instead of run-time. Global variables may optionally be initialized, may
833 have an explicit section to be placed in, and may have an optional explicit
834 alignment specified. A variable may be defined as "thread_local", which
835 means that it will not be shared by threads (each thread will have a
836 separated copy of the variable). A variable may be defined as a global
837 "constant," which indicates that the contents of the variable
838 will <b>never</b> be modified (enabling better optimization, allowing the
839 global data to be placed in the read-only section of an executable, etc).
840 Note that variables that need runtime initialization cannot be marked
841 "constant" as there is a store to the variable.</p>
843 <p>LLVM explicitly allows <em>declarations</em> of global variables to be marked
844 constant, even if the final definition of the global is not. This capability
845 can be used to enable slightly better optimization of the program, but
846 requires the language definition to guarantee that optimizations based on the
847 'constantness' are valid for the translation units that do not include the
850 <p>As SSA values, global variables define pointer values that are in scope
851 (i.e. they dominate) all basic blocks in the program. Global variables
852 always define a pointer to their "content" type because they describe a
853 region of memory, and all memory objects in LLVM are accessed through
856 <p>Global variables can be marked with <tt>unnamed_addr</tt> which indicates
857 that the address is not significant, only the content. Constants marked
858 like this can be merged with other constants if they have the same
859 initializer. Note that a constant with significant address <em>can</em>
860 be merged with a <tt>unnamed_addr</tt> constant, the result being a
861 constant whose address is significant.</p>
863 <p>A global variable may be declared to reside in a target-specific numbered
864 address space. For targets that support them, address spaces may affect how
865 optimizations are performed and/or what target instructions are used to
866 access the variable. The default address space is zero. The address space
867 qualifier must precede any other attributes.</p>
869 <p>LLVM allows an explicit section to be specified for globals. If the target
870 supports it, it will emit globals to the section specified.</p>
872 <p>An explicit alignment may be specified for a global, which must be a power
873 of 2. If not present, or if the alignment is set to zero, the alignment of
874 the global is set by the target to whatever it feels convenient. If an
875 explicit alignment is specified, the global is forced to have exactly that
876 alignment. Targets and optimizers are not allowed to over-align the global
877 if the global has an assigned section. In this case, the extra alignment
878 could be observable: for example, code could assume that the globals are
879 densely packed in their section and try to iterate over them as an array,
880 alignment padding would break this iteration.</p>
882 <p>For example, the following defines a global in a numbered address space with
883 an initializer, section, and alignment:</p>
885 <pre class="doc_code">
886 @G = addrspace(5) constant float 1.0, section "foo", align 4
892 <!-- ======================================================================= -->
894 <a name="functionstructure">Functions</a>
899 <p>LLVM function definitions consist of the "<tt>define</tt>" keyword, an
900 optional <a href="#linkage">linkage type</a>, an optional
901 <a href="#visibility">visibility style</a>, an optional
902 <a href="#callingconv">calling convention</a>,
903 an optional <tt>unnamed_addr</tt> attribute, a return type, an optional
904 <a href="#paramattrs">parameter attribute</a> for the return type, a function
905 name, a (possibly empty) argument list (each with optional
906 <a href="#paramattrs">parameter attributes</a>), optional
907 <a href="#fnattrs">function attributes</a>, an optional section, an optional
908 alignment, an optional <a href="#gc">garbage collector name</a>, an opening
909 curly brace, a list of basic blocks, and a closing curly brace.</p>
911 <p>LLVM function declarations consist of the "<tt>declare</tt>" keyword, an
912 optional <a href="#linkage">linkage type</a>, an optional
913 <a href="#visibility">visibility style</a>, an optional
914 <a href="#callingconv">calling convention</a>,
915 an optional <tt>unnamed_addr</tt> attribute, a return type, an optional
916 <a href="#paramattrs">parameter attribute</a> for the return type, a function
917 name, a possibly empty list of arguments, an optional alignment, and an
918 optional <a href="#gc">garbage collector name</a>.</p>
920 <p>A function definition contains a list of basic blocks, forming the CFG
921 (Control Flow Graph) for the function. Each basic block may optionally start
922 with a label (giving the basic block a symbol table entry), contains a list
923 of instructions, and ends with a <a href="#terminators">terminator</a>
924 instruction (such as a branch or function return).</p>
926 <p>The first basic block in a function is special in two ways: it is immediately
927 executed on entrance to the function, and it is not allowed to have
928 predecessor basic blocks (i.e. there can not be any branches to the entry
929 block of a function). Because the block can have no predecessors, it also
930 cannot have any <a href="#i_phi">PHI nodes</a>.</p>
932 <p>LLVM allows an explicit section to be specified for functions. If the target
933 supports it, it will emit functions to the section specified.</p>
935 <p>An explicit alignment may be specified for a function. If not present, or if
936 the alignment is set to zero, the alignment of the function is set by the
937 target to whatever it feels convenient. If an explicit alignment is
938 specified, the function is forced to have at least that much alignment. All
939 alignments must be a power of 2.</p>
941 <p>If the <tt>unnamed_addr</tt> attribute is given, the address is know to not
942 be significant and two identical functions can be merged.</p>
945 <pre class="doc_code">
946 define [<a href="#linkage">linkage</a>] [<a href="#visibility">visibility</a>]
947 [<a href="#callingconv">cconv</a>] [<a href="#paramattrs">ret attrs</a>]
948 <ResultType> @<FunctionName> ([argument list])
949 [<a href="#fnattrs">fn Attrs</a>] [section "name"] [align N]
950 [<a href="#gc">gc</a>] { ... }
955 <!-- ======================================================================= -->
957 <a name="aliasstructure">Aliases</a>
962 <p>Aliases act as "second name" for the aliasee value (which can be either
963 function, global variable, another alias or bitcast of global value). Aliases
964 may have an optional <a href="#linkage">linkage type</a>, and an
965 optional <a href="#visibility">visibility style</a>.</p>
968 <pre class="doc_code">
969 @<Name> = alias [Linkage] [Visibility] <AliaseeTy> @<Aliasee>
974 <!-- ======================================================================= -->
976 <a name="namedmetadatastructure">Named Metadata</a>
981 <p>Named metadata is a collection of metadata. <a href="#metadata">Metadata
982 nodes</a> (but not metadata strings) are the only valid operands for
983 a named metadata.</p>
986 <pre class="doc_code">
987 ; Some unnamed metadata nodes, which are referenced by the named metadata.
988 !0 = metadata !{metadata !"zero"}
989 !1 = metadata !{metadata !"one"}
990 !2 = metadata !{metadata !"two"}
992 !name = !{!0, !1, !2}
997 <!-- ======================================================================= -->
999 <a name="paramattrs">Parameter Attributes</a>
1004 <p>The return type and each parameter of a function type may have a set of
1005 <i>parameter attributes</i> associated with them. Parameter attributes are
1006 used to communicate additional information about the result or parameters of
1007 a function. Parameter attributes are considered to be part of the function,
1008 not of the function type, so functions with different parameter attributes
1009 can have the same function type.</p>
1011 <p>Parameter attributes are simple keywords that follow the type specified. If
1012 multiple parameter attributes are needed, they are space separated. For
1015 <pre class="doc_code">
1016 declare i32 @printf(i8* noalias nocapture, ...)
1017 declare i32 @atoi(i8 zeroext)
1018 declare signext i8 @returns_signed_char()
1021 <p>Note that any attributes for the function result (<tt>nounwind</tt>,
1022 <tt>readonly</tt>) come immediately after the argument list.</p>
1024 <p>Currently, only the following parameter attributes are defined:</p>
1027 <dt><tt><b>zeroext</b></tt></dt>
1028 <dd>This indicates to the code generator that the parameter or return value
1029 should be zero-extended to the extent required by the target's ABI (which
1030 is usually 32-bits, but is 8-bits for a i1 on x86-64) by the caller (for a
1031 parameter) or the callee (for a return value).</dd>
1033 <dt><tt><b>signext</b></tt></dt>
1034 <dd>This indicates to the code generator that the parameter or return value
1035 should be sign-extended to the extent required by the target's ABI (which
1036 is usually 32-bits) by the caller (for a parameter) or the callee (for a
1039 <dt><tt><b>inreg</b></tt></dt>
1040 <dd>This indicates that this parameter or return value should be treated in a
1041 special target-dependent fashion during while emitting code for a function
1042 call or return (usually, by putting it in a register as opposed to memory,
1043 though some targets use it to distinguish between two different kinds of
1044 registers). Use of this attribute is target-specific.</dd>
1046 <dt><tt><b><a name="byval">byval</a></b></tt></dt>
1047 <dd><p>This indicates that the pointer parameter should really be passed by
1048 value to the function. The attribute implies that a hidden copy of the
1050 is made between the caller and the callee, so the callee is unable to
1051 modify the value in the callee. This attribute is only valid on LLVM
1052 pointer arguments. It is generally used to pass structs and arrays by
1053 value, but is also valid on pointers to scalars. The copy is considered
1054 to belong to the caller not the callee (for example,
1055 <tt><a href="#readonly">readonly</a></tt> functions should not write to
1056 <tt>byval</tt> parameters). This is not a valid attribute for return
1059 <p>The byval attribute also supports specifying an alignment with
1060 the align attribute. It indicates the alignment of the stack slot to
1061 form and the known alignment of the pointer specified to the call site. If
1062 the alignment is not specified, then the code generator makes a
1063 target-specific assumption.</p></dd>
1065 <dt><tt><b><a name="sret">sret</a></b></tt></dt>
1066 <dd>This indicates that the pointer parameter specifies the address of a
1067 structure that is the return value of the function in the source program.
1068 This pointer must be guaranteed by the caller to be valid: loads and
1069 stores to the structure may be assumed by the callee to not to trap. This
1070 may only be applied to the first parameter. This is not a valid attribute
1071 for return values. </dd>
1073 <dt><tt><b><a name="noalias">noalias</a></b></tt></dt>
1074 <dd>This indicates that pointer values
1075 <a href="#pointeraliasing"><i>based</i></a> on the argument or return
1076 value do not alias pointer values which are not <i>based</i> on it,
1077 ignoring certain "irrelevant" dependencies.
1078 For a call to the parent function, dependencies between memory
1079 references from before or after the call and from those during the call
1080 are "irrelevant" to the <tt>noalias</tt> keyword for the arguments and
1081 return value used in that call.
1082 The caller shares the responsibility with the callee for ensuring that
1083 these requirements are met.
1084 For further details, please see the discussion of the NoAlias response in
1085 <a href="AliasAnalysis.html#MustMayNo">alias analysis</a>.<br>
1087 Note that this definition of <tt>noalias</tt> is intentionally
1088 similar to the definition of <tt>restrict</tt> in C99 for function
1089 arguments, though it is slightly weaker.
1091 For function return values, C99's <tt>restrict</tt> is not meaningful,
1092 while LLVM's <tt>noalias</tt> is.
1095 <dt><tt><b><a name="nocapture">nocapture</a></b></tt></dt>
1096 <dd>This indicates that the callee does not make any copies of the pointer
1097 that outlive the callee itself. This is not a valid attribute for return
1100 <dt><tt><b><a name="nest">nest</a></b></tt></dt>
1101 <dd>This indicates that the pointer parameter can be excised using the
1102 <a href="#int_trampoline">trampoline intrinsics</a>. This is not a valid
1103 attribute for return values.</dd>
1108 <!-- ======================================================================= -->
1110 <a name="gc">Garbage Collector Names</a>
1115 <p>Each function may specify a garbage collector name, which is simply a
1118 <pre class="doc_code">
1119 define void @f() gc "name" { ... }
1122 <p>The compiler declares the supported values of <i>name</i>. Specifying a
1123 collector which will cause the compiler to alter its output in order to
1124 support the named garbage collection algorithm.</p>
1128 <!-- ======================================================================= -->
1130 <a name="fnattrs">Function Attributes</a>
1135 <p>Function attributes are set to communicate additional information about a
1136 function. Function attributes are considered to be part of the function, not
1137 of the function type, so functions with different parameter attributes can
1138 have the same function type.</p>
1140 <p>Function attributes are simple keywords that follow the type specified. If
1141 multiple attributes are needed, they are space separated. For example:</p>
1143 <pre class="doc_code">
1144 define void @f() noinline { ... }
1145 define void @f() alwaysinline { ... }
1146 define void @f() alwaysinline optsize { ... }
1147 define void @f() optsize { ... }
1151 <dt><tt><b>alignstack(<<em>n</em>>)</b></tt></dt>
1152 <dd>This attribute indicates that, when emitting the prologue and epilogue,
1153 the backend should forcibly align the stack pointer. Specify the
1154 desired alignment, which must be a power of two, in parentheses.
1156 <dt><tt><b>alwaysinline</b></tt></dt>
1157 <dd>This attribute indicates that the inliner should attempt to inline this
1158 function into callers whenever possible, ignoring any active inlining size
1159 threshold for this caller.</dd>
1161 <dt><tt><b>nonlazybind</b></tt></dt>
1162 <dd>This attribute suppresses lazy symbol binding for the function. This
1163 may make calls to the function faster, at the cost of extra program
1164 startup time if the function is not called during program startup.</dd>
1166 <dt><tt><b>inlinehint</b></tt></dt>
1167 <dd>This attribute indicates that the source code contained a hint that inlining
1168 this function is desirable (such as the "inline" keyword in C/C++). It
1169 is just a hint; it imposes no requirements on the inliner.</dd>
1171 <dt><tt><b>naked</b></tt></dt>
1172 <dd>This attribute disables prologue / epilogue emission for the function.
1173 This can have very system-specific consequences.</dd>
1175 <dt><tt><b>noimplicitfloat</b></tt></dt>
1176 <dd>This attributes disables implicit floating point instructions.</dd>
1178 <dt><tt><b>noinline</b></tt></dt>
1179 <dd>This attribute indicates that the inliner should never inline this
1180 function in any situation. This attribute may not be used together with
1181 the <tt>alwaysinline</tt> attribute.</dd>
1183 <dt><tt><b>noredzone</b></tt></dt>
1184 <dd>This attribute indicates that the code generator should not use a red
1185 zone, even if the target-specific ABI normally permits it.</dd>
1187 <dt><tt><b>noreturn</b></tt></dt>
1188 <dd>This function attribute indicates that the function never returns
1189 normally. This produces undefined behavior at runtime if the function
1190 ever does dynamically return.</dd>
1192 <dt><tt><b>nounwind</b></tt></dt>
1193 <dd>This function attribute indicates that the function never returns with an
1194 unwind or exceptional control flow. If the function does unwind, its
1195 runtime behavior is undefined.</dd>
1197 <dt><tt><b>optsize</b></tt></dt>
1198 <dd>This attribute suggests that optimization passes and code generator passes
1199 make choices that keep the code size of this function low, and otherwise
1200 do optimizations specifically to reduce code size.</dd>
1202 <dt><tt><b>readnone</b></tt></dt>
1203 <dd>This attribute indicates that the function computes its result (or decides
1204 to unwind an exception) based strictly on its arguments, without
1205 dereferencing any pointer arguments or otherwise accessing any mutable
1206 state (e.g. memory, control registers, etc) visible to caller functions.
1207 It does not write through any pointer arguments
1208 (including <tt><a href="#byval">byval</a></tt> arguments) and never
1209 changes any state visible to callers. This means that it cannot unwind
1210 exceptions by calling the <tt>C++</tt> exception throwing methods, but
1211 could use the <tt>unwind</tt> instruction.</dd>
1213 <dt><tt><b><a name="readonly">readonly</a></b></tt></dt>
1214 <dd>This attribute indicates that the function does not write through any
1215 pointer arguments (including <tt><a href="#byval">byval</a></tt>
1216 arguments) or otherwise modify any state (e.g. memory, control registers,
1217 etc) visible to caller functions. It may dereference pointer arguments
1218 and read state that may be set in the caller. A readonly function always
1219 returns the same value (or unwinds an exception identically) when called
1220 with the same set of arguments and global state. It cannot unwind an
1221 exception by calling the <tt>C++</tt> exception throwing methods, but may
1222 use the <tt>unwind</tt> instruction.</dd>
1224 <dt><tt><b><a name="returns_twice">returns_twice</a></b></tt></dt>
1225 <dd>This attribute indicates that this function can return twice. The
1226 C <code>setjmp</code> is an example of such a function. The compiler
1227 disables some optimizations (like tail calls) in the caller of these
1230 <dt><tt><b><a name="ssp">ssp</a></b></tt></dt>
1231 <dd>This attribute indicates that the function should emit a stack smashing
1232 protector. It is in the form of a "canary"—a random value placed on
1233 the stack before the local variables that's checked upon return from the
1234 function to see if it has been overwritten. A heuristic is used to
1235 determine if a function needs stack protectors or not.<br>
1237 If a function that has an <tt>ssp</tt> attribute is inlined into a
1238 function that doesn't have an <tt>ssp</tt> attribute, then the resulting
1239 function will have an <tt>ssp</tt> attribute.</dd>
1241 <dt><tt><b>sspreq</b></tt></dt>
1242 <dd>This attribute indicates that the function should <em>always</em> emit a
1243 stack smashing protector. This overrides
1244 the <tt><a href="#ssp">ssp</a></tt> function attribute.<br>
1246 If a function that has an <tt>sspreq</tt> attribute is inlined into a
1247 function that doesn't have an <tt>sspreq</tt> attribute or which has
1248 an <tt>ssp</tt> attribute, then the resulting function will have
1249 an <tt>sspreq</tt> attribute.</dd>
1251 <dt><tt><b><a name="uwtable">uwtable</a></b></tt></dt>
1252 <dd>This attribute indicates that the ABI being targeted requires that
1253 an unwind table entry be produce for this function even if we can
1254 show that no exceptions passes by it. This is normally the case for
1255 the ELF x86-64 abi, but it can be disabled for some compilation
1261 <!-- ======================================================================= -->
1263 <a name="moduleasm">Module-Level Inline Assembly</a>
1268 <p>Modules may contain "module-level inline asm" blocks, which corresponds to
1269 the GCC "file scope inline asm" blocks. These blocks are internally
1270 concatenated by LLVM and treated as a single unit, but may be separated in
1271 the <tt>.ll</tt> file if desired. The syntax is very simple:</p>
1273 <pre class="doc_code">
1274 module asm "inline asm code goes here"
1275 module asm "more can go here"
1278 <p>The strings can contain any character by escaping non-printable characters.
1279 The escape sequence used is simply "\xx" where "xx" is the two digit hex code
1282 <p>The inline asm code is simply printed to the machine code .s file when
1283 assembly code is generated.</p>
1287 <!-- ======================================================================= -->
1289 <a name="datalayout">Data Layout</a>
1294 <p>A module may specify a target specific data layout string that specifies how
1295 data is to be laid out in memory. The syntax for the data layout is
1298 <pre class="doc_code">
1299 target datalayout = "<i>layout specification</i>"
1302 <p>The <i>layout specification</i> consists of a list of specifications
1303 separated by the minus sign character ('-'). Each specification starts with
1304 a letter and may include other information after the letter to define some
1305 aspect of the data layout. The specifications accepted are as follows:</p>
1309 <dd>Specifies that the target lays out data in big-endian form. That is, the
1310 bits with the most significance have the lowest address location.</dd>
1313 <dd>Specifies that the target lays out data in little-endian form. That is,
1314 the bits with the least significance have the lowest address
1317 <dt><tt>S<i>size</i></tt></dt>
1318 <dd>Specifies the natural alignment of the stack in bits. Alignment promotion
1319 of stack variables is limited to the natural stack alignment to avoid
1320 dynamic stack realignment. The stack alignment must be a multiple of
1321 8-bits. If omitted, the natural stack alignment defaults to "unspecified",
1322 which does not prevent any alignment promotions.</dd>
1324 <dt><tt>p:<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1325 <dd>This specifies the <i>size</i> of a pointer and its <i>abi</i> and
1326 <i>preferred</i> alignments. All sizes are in bits. Specifying
1327 the <i>pref</i> alignment is optional. If omitted, the
1328 preceding <tt>:</tt> should be omitted too.</dd>
1330 <dt><tt>i<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1331 <dd>This specifies the alignment for an integer type of a given bit
1332 <i>size</i>. The value of <i>size</i> must be in the range [1,2^23).</dd>
1334 <dt><tt>v<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1335 <dd>This specifies the alignment for a vector type of a given bit
1338 <dt><tt>f<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1339 <dd>This specifies the alignment for a floating point type of a given bit
1340 <i>size</i>. Only values of <i>size</i> that are supported by the target
1341 will work. 32 (float) and 64 (double) are supported on all targets;
1342 80 or 128 (different flavors of long double) are also supported on some
1345 <dt><tt>a<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1346 <dd>This specifies the alignment for an aggregate type of a given bit
1349 <dt><tt>s<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1350 <dd>This specifies the alignment for a stack object of a given bit
1353 <dt><tt>n<i>size1</i>:<i>size2</i>:<i>size3</i>...</tt></dt>
1354 <dd>This specifies a set of native integer widths for the target CPU
1355 in bits. For example, it might contain "n32" for 32-bit PowerPC,
1356 "n32:64" for PowerPC 64, or "n8:16:32:64" for X86-64. Elements of
1357 this set are considered to support most general arithmetic
1358 operations efficiently.</dd>
1361 <p>When constructing the data layout for a given target, LLVM starts with a
1362 default set of specifications which are then (possibly) overridden by the
1363 specifications in the <tt>datalayout</tt> keyword. The default specifications
1364 are given in this list:</p>
1367 <li><tt>E</tt> - big endian</li>
1368 <li><tt>p:64:64:64</tt> - 64-bit pointers with 64-bit alignment</li>
1369 <li><tt>i1:8:8</tt> - i1 is 8-bit (byte) aligned</li>
1370 <li><tt>i8:8:8</tt> - i8 is 8-bit (byte) aligned</li>
1371 <li><tt>i16:16:16</tt> - i16 is 16-bit aligned</li>
1372 <li><tt>i32:32:32</tt> - i32 is 32-bit aligned</li>
1373 <li><tt>i64:32:64</tt> - i64 has ABI alignment of 32-bits but preferred
1374 alignment of 64-bits</li>
1375 <li><tt>f32:32:32</tt> - float is 32-bit aligned</li>
1376 <li><tt>f64:64:64</tt> - double is 64-bit aligned</li>
1377 <li><tt>v64:64:64</tt> - 64-bit vector is 64-bit aligned</li>
1378 <li><tt>v128:128:128</tt> - 128-bit vector is 128-bit aligned</li>
1379 <li><tt>a0:0:1</tt> - aggregates are 8-bit aligned</li>
1380 <li><tt>s0:64:64</tt> - stack objects are 64-bit aligned</li>
1383 <p>When LLVM is determining the alignment for a given type, it uses the
1384 following rules:</p>
1387 <li>If the type sought is an exact match for one of the specifications, that
1388 specification is used.</li>
1390 <li>If no match is found, and the type sought is an integer type, then the
1391 smallest integer type that is larger than the bitwidth of the sought type
1392 is used. If none of the specifications are larger than the bitwidth then
1393 the the largest integer type is used. For example, given the default
1394 specifications above, the i7 type will use the alignment of i8 (next
1395 largest) while both i65 and i256 will use the alignment of i64 (largest
1398 <li>If no match is found, and the type sought is a vector type, then the
1399 largest vector type that is smaller than the sought vector type will be
1400 used as a fall back. This happens because <128 x double> can be
1401 implemented in terms of 64 <2 x double>, for example.</li>
1404 <p>The function of the data layout string may not be what you expect. Notably,
1405 this is not a specification from the frontend of what alignment the code
1406 generator should use.</p>
1408 <p>Instead, if specified, the target data layout is required to match what the
1409 ultimate <em>code generator</em> expects. This string is used by the
1410 mid-level optimizers to
1411 improve code, and this only works if it matches what the ultimate code
1412 generator uses. If you would like to generate IR that does not embed this
1413 target-specific detail into the IR, then you don't have to specify the
1414 string. This will disable some optimizations that require precise layout
1415 information, but this also prevents those optimizations from introducing
1416 target specificity into the IR.</p>
1422 <!-- ======================================================================= -->
1424 <a name="pointeraliasing">Pointer Aliasing Rules</a>
1429 <p>Any memory access must be done through a pointer value associated
1430 with an address range of the memory access, otherwise the behavior
1431 is undefined. Pointer values are associated with address ranges
1432 according to the following rules:</p>
1435 <li>A pointer value is associated with the addresses associated with
1436 any value it is <i>based</i> on.
1437 <li>An address of a global variable is associated with the address
1438 range of the variable's storage.</li>
1439 <li>The result value of an allocation instruction is associated with
1440 the address range of the allocated storage.</li>
1441 <li>A null pointer in the default address-space is associated with
1443 <li>An integer constant other than zero or a pointer value returned
1444 from a function not defined within LLVM may be associated with address
1445 ranges allocated through mechanisms other than those provided by
1446 LLVM. Such ranges shall not overlap with any ranges of addresses
1447 allocated by mechanisms provided by LLVM.</li>
1450 <p>A pointer value is <i>based</i> on another pointer value according
1451 to the following rules:</p>
1454 <li>A pointer value formed from a
1455 <tt><a href="#i_getelementptr">getelementptr</a></tt> operation
1456 is <i>based</i> on the first operand of the <tt>getelementptr</tt>.</li>
1457 <li>The result value of a
1458 <tt><a href="#i_bitcast">bitcast</a></tt> is <i>based</i> on the operand
1459 of the <tt>bitcast</tt>.</li>
1460 <li>A pointer value formed by an
1461 <tt><a href="#i_inttoptr">inttoptr</a></tt> is <i>based</i> on all
1462 pointer values that contribute (directly or indirectly) to the
1463 computation of the pointer's value.</li>
1464 <li>The "<i>based</i> on" relationship is transitive.</li>
1467 <p>Note that this definition of <i>"based"</i> is intentionally
1468 similar to the definition of <i>"based"</i> in C99, though it is
1469 slightly weaker.</p>
1471 <p>LLVM IR does not associate types with memory. The result type of a
1472 <tt><a href="#i_load">load</a></tt> merely indicates the size and
1473 alignment of the memory from which to load, as well as the
1474 interpretation of the value. The first operand type of a
1475 <tt><a href="#i_store">store</a></tt> similarly only indicates the size
1476 and alignment of the store.</p>
1478 <p>Consequently, type-based alias analysis, aka TBAA, aka
1479 <tt>-fstrict-aliasing</tt>, is not applicable to general unadorned
1480 LLVM IR. <a href="#metadata">Metadata</a> may be used to encode
1481 additional information which specialized optimization passes may use
1482 to implement type-based alias analysis.</p>
1486 <!-- ======================================================================= -->
1488 <a name="volatile">Volatile Memory Accesses</a>
1493 <p>Certain memory accesses, such as <a href="#i_load"><tt>load</tt></a>s, <a
1494 href="#i_store"><tt>store</tt></a>s, and <a
1495 href="#int_memcpy"><tt>llvm.memcpy</tt></a>s may be marked <tt>volatile</tt>.
1496 The optimizers must not change the number of volatile operations or change their
1497 order of execution relative to other volatile operations. The optimizers
1498 <i>may</i> change the order of volatile operations relative to non-volatile
1499 operations. This is not Java's "volatile" and has no cross-thread
1500 synchronization behavior.</p>
1504 <!-- ======================================================================= -->
1506 <a name="memmodel">Memory Model for Concurrent Operations</a>
1511 <p>The LLVM IR does not define any way to start parallel threads of execution
1512 or to register signal handlers. Nonetheless, there are platform-specific
1513 ways to create them, and we define LLVM IR's behavior in their presence. This
1514 model is inspired by the C++0x memory model.</p>
1516 <p>For a more informal introduction to this model, see the
1517 <a href="Atomics.html">LLVM Atomic Instructions and Concurrency Guide</a>.
1519 <p>We define a <i>happens-before</i> partial order as the least partial order
1522 <li>Is a superset of single-thread program order, and</li>
1523 <li>When a <i>synchronizes-with</i> <tt>b</tt>, includes an edge from
1524 <tt>a</tt> to <tt>b</tt>. <i>Synchronizes-with</i> pairs are introduced
1525 by platform-specific techniques, like pthread locks, thread
1526 creation, thread joining, etc., and by atomic instructions.
1527 (See also <a href="#ordering">Atomic Memory Ordering Constraints</a>).
1531 <p>Note that program order does not introduce <i>happens-before</i> edges
1532 between a thread and signals executing inside that thread.</p>
1534 <p>Every (defined) read operation (load instructions, memcpy, atomic
1535 loads/read-modify-writes, etc.) <var>R</var> reads a series of bytes written by
1536 (defined) write operations (store instructions, atomic
1537 stores/read-modify-writes, memcpy, etc.). For the purposes of this section,
1538 initialized globals are considered to have a write of the initializer which is
1539 atomic and happens before any other read or write of the memory in question.
1540 For each byte of a read <var>R</var>, <var>R<sub>byte</sub></var> may see
1541 any write to the same byte, except:</p>
1544 <li>If <var>write<sub>1</sub></var> happens before
1545 <var>write<sub>2</sub></var>, and <var>write<sub>2</sub></var> happens
1546 before <var>R<sub>byte</sub></var>, then <var>R<sub>byte</sub></var>
1547 does not see <var>write<sub>1</sub></var>.
1548 <li>If <var>R<sub>byte</sub></var> happens before
1549 <var>write<sub>3</sub></var>, then <var>R<sub>byte</sub></var> does not
1550 see <var>write<sub>3</sub></var>.
1553 <p>Given that definition, <var>R<sub>byte</sub></var> is defined as follows:
1555 <li>If <var>R</var> is volatile, the result is target-dependent. (Volatile
1556 is supposed to give guarantees which can support
1557 <code>sig_atomic_t</code> in C/C++, and may be used for accesses to
1558 addresses which do not behave like normal memory. It does not generally
1559 provide cross-thread synchronization.)
1560 <li>Otherwise, if there is no write to the same byte that happens before
1561 <var>R<sub>byte</sub></var>, <var>R<sub>byte</sub></var> returns
1562 <tt>undef</tt> for that byte.
1563 <li>Otherwise, if <var>R<sub>byte</sub></var> may see exactly one write,
1564 <var>R<sub>byte</sub></var> returns the value written by that
1566 <li>Otherwise, if <var>R</var> is atomic, and all the writes
1567 <var>R<sub>byte</sub></var> may see are atomic, it chooses one of the
1568 values written. See the <a href="#ordering">Atomic Memory Ordering
1569 Constraints</a> section for additional constraints on how the choice
1571 <li>Otherwise <var>R<sub>byte</sub></var> returns <tt>undef</tt>.</li>
1574 <p><var>R</var> returns the value composed of the series of bytes it read.
1575 This implies that some bytes within the value may be <tt>undef</tt>
1576 <b>without</b> the entire value being <tt>undef</tt>. Note that this only
1577 defines the semantics of the operation; it doesn't mean that targets will
1578 emit more than one instruction to read the series of bytes.</p>
1580 <p>Note that in cases where none of the atomic intrinsics are used, this model
1581 places only one restriction on IR transformations on top of what is required
1582 for single-threaded execution: introducing a store to a byte which might not
1583 otherwise be stored is not allowed in general. (Specifically, in the case
1584 where another thread might write to and read from an address, introducing a
1585 store can change a load that may see exactly one write into a load that may
1586 see multiple writes.)</p>
1588 <!-- FIXME: This model assumes all targets where concurrency is relevant have
1589 a byte-size store which doesn't affect adjacent bytes. As far as I can tell,
1590 none of the backends currently in the tree fall into this category; however,
1591 there might be targets which care. If there are, we want a paragraph
1594 Targets may specify that stores narrower than a certain width are not
1595 available; on such a target, for the purposes of this model, treat any
1596 non-atomic write with an alignment or width less than the minimum width
1597 as if it writes to the relevant surrounding bytes.
1602 <!-- ======================================================================= -->
1604 <a name="ordering">Atomic Memory Ordering Constraints</a>
1609 <p>Atomic instructions (<a href="#i_cmpxchg"><code>cmpxchg</code></a>,
1610 <a href="#i_atomicrmw"><code>atomicrmw</code></a>,
1611 <a href="#i_fence"><code>fence</code></a>,
1612 <a href="#i_load"><code>atomic load</code></a>, and
1613 <a href="#i_store"><code>atomic store</code></a>) take an ordering parameter
1614 that determines which other atomic instructions on the same address they
1615 <i>synchronize with</i>. These semantics are borrowed from Java and C++0x,
1616 but are somewhat more colloquial. If these descriptions aren't precise enough,
1617 check those specs (see spec references in the
1618 <a href="Atomic.html#introduction">atomics guide</a>).
1619 <a href="#i_fence"><code>fence</code></a> instructions
1620 treat these orderings somewhat differently since they don't take an address.
1621 See that instruction's documentation for details.</p>
1623 <p>For a simpler introduction to the ordering constraints, see the
1624 <a href="Atomics.html">LLVM Atomic Instructions and Concurrency Guide</a>.</p>
1627 <dt><code>unordered</code></dt>
1628 <dd>The set of values that can be read is governed by the happens-before
1629 partial order. A value cannot be read unless some operation wrote it.
1630 This is intended to provide a guarantee strong enough to model Java's
1631 non-volatile shared variables. This ordering cannot be specified for
1632 read-modify-write operations; it is not strong enough to make them atomic
1633 in any interesting way.</dd>
1634 <dt><code>monotonic</code></dt>
1635 <dd>In addition to the guarantees of <code>unordered</code>, there is a single
1636 total order for modifications by <code>monotonic</code> operations on each
1637 address. All modification orders must be compatible with the happens-before
1638 order. There is no guarantee that the modification orders can be combined to
1639 a global total order for the whole program (and this often will not be
1640 possible). The read in an atomic read-modify-write operation
1641 (<a href="#i_cmpxchg"><code>cmpxchg</code></a> and
1642 <a href="#i_atomicrmw"><code>atomicrmw</code></a>)
1643 reads the value in the modification order immediately before the value it
1644 writes. If one atomic read happens before another atomic read of the same
1645 address, the later read must see the same value or a later value in the
1646 address's modification order. This disallows reordering of
1647 <code>monotonic</code> (or stronger) operations on the same address. If an
1648 address is written <code>monotonic</code>ally by one thread, and other threads
1649 <code>monotonic</code>ally read that address repeatedly, the other threads must
1650 eventually see the write. This corresponds to the C++0x/C1x
1651 <code>memory_order_relaxed</code>.</dd>
1652 <dt><code>acquire</code></dt>
1653 <dd>In addition to the guarantees of <code>monotonic</code>,
1654 a <i>synchronizes-with</i> edge may be formed with a <code>release</code>
1655 operation. This is intended to model C++'s <code>memory_order_acquire</code>.</dd>
1656 <dt><code>release</code></dt>
1657 <dd>In addition to the guarantees of <code>monotonic</code>, if this operation
1658 writes a value which is subsequently read by an <code>acquire</code> operation,
1659 it <i>synchronizes-with</i> that operation. (This isn't a complete
1660 description; see the C++0x definition of a release sequence.) This corresponds
1661 to the C++0x/C1x <code>memory_order_release</code>.</dd>
1662 <dt><code>acq_rel</code> (acquire+release)</dt><dd>Acts as both an
1663 <code>acquire</code> and <code>release</code> operation on its address.
1664 This corresponds to the C++0x/C1x <code>memory_order_acq_rel</code>.</dd>
1665 <dt><code>seq_cst</code> (sequentially consistent)</dt><dd>
1666 <dd>In addition to the guarantees of <code>acq_rel</code>
1667 (<code>acquire</code> for an operation which only reads, <code>release</code>
1668 for an operation which only writes), there is a global total order on all
1669 sequentially-consistent operations on all addresses, which is consistent with
1670 the <i>happens-before</i> partial order and with the modification orders of
1671 all the affected addresses. Each sequentially-consistent read sees the last
1672 preceding write to the same address in this global order. This corresponds
1673 to the C++0x/C1x <code>memory_order_seq_cst</code> and Java volatile.</dd>
1676 <p id="singlethread">If an atomic operation is marked <code>singlethread</code>,
1677 it only <i>synchronizes with</i> or participates in modification and seq_cst
1678 total orderings with other operations running in the same thread (for example,
1679 in signal handlers).</p>
1685 <!-- *********************************************************************** -->
1686 <h2><a name="typesystem">Type System</a></h2>
1687 <!-- *********************************************************************** -->
1691 <p>The LLVM type system is one of the most important features of the
1692 intermediate representation. Being typed enables a number of optimizations
1693 to be performed on the intermediate representation directly, without having
1694 to do extra analyses on the side before the transformation. A strong type
1695 system makes it easier to read the generated code and enables novel analyses
1696 and transformations that are not feasible to perform on normal three address
1697 code representations.</p>
1699 <!-- ======================================================================= -->
1701 <a name="t_classifications">Type Classifications</a>
1706 <p>The types fall into a few useful classifications:</p>
1708 <table border="1" cellspacing="0" cellpadding="4">
1710 <tr><th>Classification</th><th>Types</th></tr>
1712 <td><a href="#t_integer">integer</a></td>
1713 <td><tt>i1, i2, i3, ... i8, ... i16, ... i32, ... i64, ... </tt></td>
1716 <td><a href="#t_floating">floating point</a></td>
1717 <td><tt>half, float, double, x86_fp80, fp128, ppc_fp128</tt></td>
1720 <td><a name="t_firstclass">first class</a></td>
1721 <td><a href="#t_integer">integer</a>,
1722 <a href="#t_floating">floating point</a>,
1723 <a href="#t_pointer">pointer</a>,
1724 <a href="#t_vector">vector</a>,
1725 <a href="#t_struct">structure</a>,
1726 <a href="#t_array">array</a>,
1727 <a href="#t_label">label</a>,
1728 <a href="#t_metadata">metadata</a>.
1732 <td><a href="#t_primitive">primitive</a></td>
1733 <td><a href="#t_label">label</a>,
1734 <a href="#t_void">void</a>,
1735 <a href="#t_integer">integer</a>,
1736 <a href="#t_floating">floating point</a>,
1737 <a href="#t_x86mmx">x86mmx</a>,
1738 <a href="#t_metadata">metadata</a>.</td>
1741 <td><a href="#t_derived">derived</a></td>
1742 <td><a href="#t_array">array</a>,
1743 <a href="#t_function">function</a>,
1744 <a href="#t_pointer">pointer</a>,
1745 <a href="#t_struct">structure</a>,
1746 <a href="#t_vector">vector</a>,
1747 <a href="#t_opaque">opaque</a>.
1753 <p>The <a href="#t_firstclass">first class</a> types are perhaps the most
1754 important. Values of these types are the only ones which can be produced by
1759 <!-- ======================================================================= -->
1761 <a name="t_primitive">Primitive Types</a>
1766 <p>The primitive types are the fundamental building blocks of the LLVM
1769 <!-- _______________________________________________________________________ -->
1771 <a name="t_integer">Integer Type</a>
1777 <p>The integer type is a very simple type that simply specifies an arbitrary
1778 bit width for the integer type desired. Any bit width from 1 bit to
1779 2<sup>23</sup>-1 (about 8 million) can be specified.</p>
1786 <p>The number of bits the integer will occupy is specified by the <tt>N</tt>
1790 <table class="layout">
1792 <td class="left"><tt>i1</tt></td>
1793 <td class="left">a single-bit integer.</td>
1796 <td class="left"><tt>i32</tt></td>
1797 <td class="left">a 32-bit integer.</td>
1800 <td class="left"><tt>i1942652</tt></td>
1801 <td class="left">a really big integer of over 1 million bits.</td>
1807 <!-- _______________________________________________________________________ -->
1809 <a name="t_floating">Floating Point Types</a>
1816 <tr><th>Type</th><th>Description</th></tr>
1817 <tr><td><tt>half</tt></td><td>16-bit floating point value</td></tr>
1818 <tr><td><tt>float</tt></td><td>32-bit floating point value</td></tr>
1819 <tr><td><tt>double</tt></td><td>64-bit floating point value</td></tr>
1820 <tr><td><tt>fp128</tt></td><td>128-bit floating point value (112-bit mantissa)</td></tr>
1821 <tr><td><tt>x86_fp80</tt></td><td>80-bit floating point value (X87)</td></tr>
1822 <tr><td><tt>ppc_fp128</tt></td><td>128-bit floating point value (two 64-bits)</td></tr>
1828 <!-- _______________________________________________________________________ -->
1830 <a name="t_x86mmx">X86mmx Type</a>
1836 <p>The x86mmx type represents a value held in an MMX register on an x86 machine. The operations allowed on it are quite limited: parameters and return values, load and store, and bitcast. User-specified MMX instructions are represented as intrinsic or asm calls with arguments and/or results of this type. There are no arrays, vectors or constants of this type.</p>
1845 <!-- _______________________________________________________________________ -->
1847 <a name="t_void">Void Type</a>
1853 <p>The void type does not represent any value and has no size.</p>
1862 <!-- _______________________________________________________________________ -->
1864 <a name="t_label">Label Type</a>
1870 <p>The label type represents code labels.</p>
1879 <!-- _______________________________________________________________________ -->
1881 <a name="t_metadata">Metadata Type</a>
1887 <p>The metadata type represents embedded metadata. No derived types may be
1888 created from metadata except for <a href="#t_function">function</a>
1900 <!-- ======================================================================= -->
1902 <a name="t_derived">Derived Types</a>
1907 <p>The real power in LLVM comes from the derived types in the system. This is
1908 what allows a programmer to represent arrays, functions, pointers, and other
1909 useful types. Each of these types contain one or more element types which
1910 may be a primitive type, or another derived type. For example, it is
1911 possible to have a two dimensional array, using an array as the element type
1912 of another array.</p>
1914 <!-- _______________________________________________________________________ -->
1916 <a name="t_aggregate">Aggregate Types</a>
1921 <p>Aggregate Types are a subset of derived types that can contain multiple
1922 member types. <a href="#t_array">Arrays</a> and
1923 <a href="#t_struct">structs</a> are aggregate types.
1924 <a href="#t_vector">Vectors</a> are not considered to be aggregate types.</p>
1928 <!-- _______________________________________________________________________ -->
1930 <a name="t_array">Array Type</a>
1936 <p>The array type is a very simple derived type that arranges elements
1937 sequentially in memory. The array type requires a size (number of elements)
1938 and an underlying data type.</p>
1942 [<# elements> x <elementtype>]
1945 <p>The number of elements is a constant integer value; <tt>elementtype</tt> may
1946 be any type with a size.</p>
1949 <table class="layout">
1951 <td class="left"><tt>[40 x i32]</tt></td>
1952 <td class="left">Array of 40 32-bit integer values.</td>
1955 <td class="left"><tt>[41 x i32]</tt></td>
1956 <td class="left">Array of 41 32-bit integer values.</td>
1959 <td class="left"><tt>[4 x i8]</tt></td>
1960 <td class="left">Array of 4 8-bit integer values.</td>
1963 <p>Here are some examples of multidimensional arrays:</p>
1964 <table class="layout">
1966 <td class="left"><tt>[3 x [4 x i32]]</tt></td>
1967 <td class="left">3x4 array of 32-bit integer values.</td>
1970 <td class="left"><tt>[12 x [10 x float]]</tt></td>
1971 <td class="left">12x10 array of single precision floating point values.</td>
1974 <td class="left"><tt>[2 x [3 x [4 x i16]]]</tt></td>
1975 <td class="left">2x3x4 array of 16-bit integer values.</td>
1979 <p>There is no restriction on indexing beyond the end of the array implied by
1980 a static type (though there are restrictions on indexing beyond the bounds
1981 of an allocated object in some cases). This means that single-dimension
1982 'variable sized array' addressing can be implemented in LLVM with a zero
1983 length array type. An implementation of 'pascal style arrays' in LLVM could
1984 use the type "<tt>{ i32, [0 x float]}</tt>", for example.</p>
1988 <!-- _______________________________________________________________________ -->
1990 <a name="t_function">Function Type</a>
1996 <p>The function type can be thought of as a function signature. It consists of
1997 a return type and a list of formal parameter types. The return type of a
1998 function type is a first class type or a void type.</p>
2002 <returntype> (<parameter list>)
2005 <p>...where '<tt><parameter list></tt>' is a comma-separated list of type
2006 specifiers. Optionally, the parameter list may include a type <tt>...</tt>,
2007 which indicates that the function takes a variable number of arguments.
2008 Variable argument functions can access their arguments with
2009 the <a href="#int_varargs">variable argument handling intrinsic</a>
2010 functions. '<tt><returntype></tt>' is any type except
2011 <a href="#t_label">label</a>.</p>
2014 <table class="layout">
2016 <td class="left"><tt>i32 (i32)</tt></td>
2017 <td class="left">function taking an <tt>i32</tt>, returning an <tt>i32</tt>
2019 </tr><tr class="layout">
2020 <td class="left"><tt>float (i16, i32 *) *
2022 <td class="left"><a href="#t_pointer">Pointer</a> to a function that takes
2023 an <tt>i16</tt> and a <a href="#t_pointer">pointer</a> to <tt>i32</tt>,
2024 returning <tt>float</tt>.
2026 </tr><tr class="layout">
2027 <td class="left"><tt>i32 (i8*, ...)</tt></td>
2028 <td class="left">A vararg function that takes at least one
2029 <a href="#t_pointer">pointer</a> to <tt>i8 </tt> (char in C),
2030 which returns an integer. This is the signature for <tt>printf</tt> in
2033 </tr><tr class="layout">
2034 <td class="left"><tt>{i32, i32} (i32)</tt></td>
2035 <td class="left">A function taking an <tt>i32</tt>, returning a
2036 <a href="#t_struct">structure</a> containing two <tt>i32</tt> values
2043 <!-- _______________________________________________________________________ -->
2045 <a name="t_struct">Structure Type</a>
2051 <p>The structure type is used to represent a collection of data members together
2052 in memory. The elements of a structure may be any type that has a size.</p>
2054 <p>Structures in memory are accessed using '<tt><a href="#i_load">load</a></tt>'
2055 and '<tt><a href="#i_store">store</a></tt>' by getting a pointer to a field
2056 with the '<tt><a href="#i_getelementptr">getelementptr</a></tt>' instruction.
2057 Structures in registers are accessed using the
2058 '<tt><a href="#i_extractvalue">extractvalue</a></tt>' and
2059 '<tt><a href="#i_insertvalue">insertvalue</a></tt>' instructions.</p>
2061 <p>Structures may optionally be "packed" structures, which indicate that the
2062 alignment of the struct is one byte, and that there is no padding between
2063 the elements. In non-packed structs, padding between field types is inserted
2064 as defined by the TargetData string in the module, which is required to match
2065 what the underlying code generator expects.</p>
2067 <p>Structures can either be "literal" or "identified". A literal structure is
2068 defined inline with other types (e.g. <tt>{i32, i32}*</tt>) whereas identified
2069 types are always defined at the top level with a name. Literal types are
2070 uniqued by their contents and can never be recursive or opaque since there is
2071 no way to write one. Identified types can be recursive, can be opaqued, and are
2077 %T1 = type { <type list> } <i>; Identified normal struct type</i>
2078 %T2 = type <{ <type list> }> <i>; Identified packed struct type</i>
2082 <table class="layout">
2084 <td class="left"><tt>{ i32, i32, i32 }</tt></td>
2085 <td class="left">A triple of three <tt>i32</tt> values</td>
2088 <td class="left"><tt>{ float, i32 (i32) * }</tt></td>
2089 <td class="left">A pair, where the first element is a <tt>float</tt> and the
2090 second element is a <a href="#t_pointer">pointer</a> to a
2091 <a href="#t_function">function</a> that takes an <tt>i32</tt>, returning
2092 an <tt>i32</tt>.</td>
2095 <td class="left"><tt><{ i8, i32 }></tt></td>
2096 <td class="left">A packed struct known to be 5 bytes in size.</td>
2102 <!-- _______________________________________________________________________ -->
2104 <a name="t_opaque">Opaque Structure Types</a>
2110 <p>Opaque structure types are used to represent named structure types that do
2111 not have a body specified. This corresponds (for example) to the C notion of
2112 a forward declared structure.</p>
2121 <table class="layout">
2123 <td class="left"><tt>opaque</tt></td>
2124 <td class="left">An opaque type.</td>
2132 <!-- _______________________________________________________________________ -->
2134 <a name="t_pointer">Pointer Type</a>
2140 <p>The pointer type is used to specify memory locations.
2141 Pointers are commonly used to reference objects in memory.</p>
2143 <p>Pointer types may have an optional address space attribute defining the
2144 numbered address space where the pointed-to object resides. The default
2145 address space is number zero. The semantics of non-zero address
2146 spaces are target-specific.</p>
2148 <p>Note that LLVM does not permit pointers to void (<tt>void*</tt>) nor does it
2149 permit pointers to labels (<tt>label*</tt>). Use <tt>i8*</tt> instead.</p>
2157 <table class="layout">
2159 <td class="left"><tt>[4 x i32]*</tt></td>
2160 <td class="left">A <a href="#t_pointer">pointer</a> to <a
2161 href="#t_array">array</a> of four <tt>i32</tt> values.</td>
2164 <td class="left"><tt>i32 (i32*) *</tt></td>
2165 <td class="left"> A <a href="#t_pointer">pointer</a> to a <a
2166 href="#t_function">function</a> that takes an <tt>i32*</tt>, returning an
2170 <td class="left"><tt>i32 addrspace(5)*</tt></td>
2171 <td class="left">A <a href="#t_pointer">pointer</a> to an <tt>i32</tt> value
2172 that resides in address space #5.</td>
2178 <!-- _______________________________________________________________________ -->
2180 <a name="t_vector">Vector Type</a>
2186 <p>A vector type is a simple derived type that represents a vector of elements.
2187 Vector types are used when multiple primitive data are operated in parallel
2188 using a single instruction (SIMD). A vector type requires a size (number of
2189 elements) and an underlying primitive data type. Vector types are considered
2190 <a href="#t_firstclass">first class</a>.</p>
2194 < <# elements> x <elementtype> >
2197 <p>The number of elements is a constant integer value larger than 0; elementtype
2198 may be any integer or floating point type, or a pointer to these types.
2199 Vectors of size zero are not allowed. </p>
2202 <table class="layout">
2204 <td class="left"><tt><4 x i32></tt></td>
2205 <td class="left">Vector of 4 32-bit integer values.</td>
2208 <td class="left"><tt><8 x float></tt></td>
2209 <td class="left">Vector of 8 32-bit floating-point values.</td>
2212 <td class="left"><tt><2 x i64></tt></td>
2213 <td class="left">Vector of 2 64-bit integer values.</td>
2216 <td class="left"><tt><4 x i64*></tt></td>
2217 <td class="left">Vector of 4 pointers to 64-bit integer values.</td>
2227 <!-- *********************************************************************** -->
2228 <h2><a name="constants">Constants</a></h2>
2229 <!-- *********************************************************************** -->
2233 <p>LLVM has several different basic types of constants. This section describes
2234 them all and their syntax.</p>
2236 <!-- ======================================================================= -->
2238 <a name="simpleconstants">Simple Constants</a>
2244 <dt><b>Boolean constants</b></dt>
2245 <dd>The two strings '<tt>true</tt>' and '<tt>false</tt>' are both valid
2246 constants of the <tt><a href="#t_integer">i1</a></tt> type.</dd>
2248 <dt><b>Integer constants</b></dt>
2249 <dd>Standard integers (such as '4') are constants of
2250 the <a href="#t_integer">integer</a> type. Negative numbers may be used
2251 with integer types.</dd>
2253 <dt><b>Floating point constants</b></dt>
2254 <dd>Floating point constants use standard decimal notation (e.g. 123.421),
2255 exponential notation (e.g. 1.23421e+2), or a more precise hexadecimal
2256 notation (see below). The assembler requires the exact decimal value of a
2257 floating-point constant. For example, the assembler accepts 1.25 but
2258 rejects 1.3 because 1.3 is a repeating decimal in binary. Floating point
2259 constants must have a <a href="#t_floating">floating point</a> type. </dd>
2261 <dt><b>Null pointer constants</b></dt>
2262 <dd>The identifier '<tt>null</tt>' is recognized as a null pointer constant
2263 and must be of <a href="#t_pointer">pointer type</a>.</dd>
2266 <p>The one non-intuitive notation for constants is the hexadecimal form of
2267 floating point constants. For example, the form '<tt>double
2268 0x432ff973cafa8000</tt>' is equivalent to (but harder to read than)
2269 '<tt>double 4.5e+15</tt>'. The only time hexadecimal floating point
2270 constants are required (and the only time that they are generated by the
2271 disassembler) is when a floating point constant must be emitted but it cannot
2272 be represented as a decimal floating point number in a reasonable number of
2273 digits. For example, NaN's, infinities, and other special values are
2274 represented in their IEEE hexadecimal format so that assembly and disassembly
2275 do not cause any bits to change in the constants.</p>
2277 <p>When using the hexadecimal form, constants of types half, float, and double are
2278 represented using the 16-digit form shown above (which matches the IEEE754
2279 representation for double); half and float values must, however, be exactly
2280 representable as IEE754 half and single precision, respectively.
2281 Hexadecimal format is always used
2282 for long double, and there are three forms of long double. The 80-bit format
2283 used by x86 is represented as <tt>0xK</tt> followed by 20 hexadecimal digits.
2284 The 128-bit format used by PowerPC (two adjacent doubles) is represented
2285 by <tt>0xM</tt> followed by 32 hexadecimal digits. The IEEE 128-bit format
2286 is represented by <tt>0xL</tt> followed by 32 hexadecimal digits; no
2287 currently supported target uses this format. Long doubles will only work if
2288 they match the long double format on your target. All hexadecimal formats
2289 are big-endian (sign bit at the left).</p>
2291 <p>There are no constants of type x86mmx.</p>
2294 <!-- ======================================================================= -->
2296 <a name="aggregateconstants"></a> <!-- old anchor -->
2297 <a name="complexconstants">Complex Constants</a>
2302 <p>Complex constants are a (potentially recursive) combination of simple
2303 constants and smaller complex constants.</p>
2306 <dt><b>Structure constants</b></dt>
2307 <dd>Structure constants are represented with notation similar to structure
2308 type definitions (a comma separated list of elements, surrounded by braces
2309 (<tt>{}</tt>)). For example: "<tt>{ i32 4, float 17.0, i32* @G }</tt>",
2310 where "<tt>@G</tt>" is declared as "<tt>@G = external global i32</tt>".
2311 Structure constants must have <a href="#t_struct">structure type</a>, and
2312 the number and types of elements must match those specified by the
2315 <dt><b>Array constants</b></dt>
2316 <dd>Array constants are represented with notation similar to array type
2317 definitions (a comma separated list of elements, surrounded by square
2318 brackets (<tt>[]</tt>)). For example: "<tt>[ i32 42, i32 11, i32 74
2319 ]</tt>". Array constants must have <a href="#t_array">array type</a>, and
2320 the number and types of elements must match those specified by the
2323 <dt><b>Vector constants</b></dt>
2324 <dd>Vector constants are represented with notation similar to vector type
2325 definitions (a comma separated list of elements, surrounded by
2326 less-than/greater-than's (<tt><></tt>)). For example: "<tt>< i32
2327 42, i32 11, i32 74, i32 100 ></tt>". Vector constants must
2328 have <a href="#t_vector">vector type</a>, and the number and types of
2329 elements must match those specified by the type.</dd>
2331 <dt><b>Zero initialization</b></dt>
2332 <dd>The string '<tt>zeroinitializer</tt>' can be used to zero initialize a
2333 value to zero of <em>any</em> type, including scalar and
2334 <a href="#t_aggregate">aggregate</a> types.
2335 This is often used to avoid having to print large zero initializers
2336 (e.g. for large arrays) and is always exactly equivalent to using explicit
2337 zero initializers.</dd>
2339 <dt><b>Metadata node</b></dt>
2340 <dd>A metadata node is a structure-like constant with
2341 <a href="#t_metadata">metadata type</a>. For example: "<tt>metadata !{
2342 i32 0, metadata !"test" }</tt>". Unlike other constants that are meant to
2343 be interpreted as part of the instruction stream, metadata is a place to
2344 attach additional information such as debug info.</dd>
2349 <!-- ======================================================================= -->
2351 <a name="globalconstants">Global Variable and Function Addresses</a>
2356 <p>The addresses of <a href="#globalvars">global variables</a>
2357 and <a href="#functionstructure">functions</a> are always implicitly valid
2358 (link-time) constants. These constants are explicitly referenced when
2359 the <a href="#identifiers">identifier for the global</a> is used and always
2360 have <a href="#t_pointer">pointer</a> type. For example, the following is a
2361 legal LLVM file:</p>
2363 <pre class="doc_code">
2366 @Z = global [2 x i32*] [ i32* @X, i32* @Y ]
2371 <!-- ======================================================================= -->
2373 <a name="undefvalues">Undefined Values</a>
2378 <p>The string '<tt>undef</tt>' can be used anywhere a constant is expected, and
2379 indicates that the user of the value may receive an unspecified bit-pattern.
2380 Undefined values may be of any type (other than '<tt>label</tt>'
2381 or '<tt>void</tt>') and be used anywhere a constant is permitted.</p>
2383 <p>Undefined values are useful because they indicate to the compiler that the
2384 program is well defined no matter what value is used. This gives the
2385 compiler more freedom to optimize. Here are some examples of (potentially
2386 surprising) transformations that are valid (in pseudo IR):</p>
2389 <pre class="doc_code">
2399 <p>This is safe because all of the output bits are affected by the undef bits.
2400 Any output bit can have a zero or one depending on the input bits.</p>
2402 <pre class="doc_code">
2413 <p>These logical operations have bits that are not always affected by the input.
2414 For example, if <tt>%X</tt> has a zero bit, then the output of the
2415 '<tt>and</tt>' operation will always be a zero for that bit, no matter what
2416 the corresponding bit from the '<tt>undef</tt>' is. As such, it is unsafe to
2417 optimize or assume that the result of the '<tt>and</tt>' is '<tt>undef</tt>'.
2418 However, it is safe to assume that all bits of the '<tt>undef</tt>' could be
2419 0, and optimize the '<tt>and</tt>' to 0. Likewise, it is safe to assume that
2420 all the bits of the '<tt>undef</tt>' operand to the '<tt>or</tt>' could be
2421 set, allowing the '<tt>or</tt>' to be folded to -1.</p>
2423 <pre class="doc_code">
2424 %A = select undef, %X, %Y
2425 %B = select undef, 42, %Y
2426 %C = select %X, %Y, undef
2437 <p>This set of examples shows that undefined '<tt>select</tt>' (and conditional
2438 branch) conditions can go <em>either way</em>, but they have to come from one
2439 of the two operands. In the <tt>%A</tt> example, if <tt>%X</tt> and
2440 <tt>%Y</tt> were both known to have a clear low bit, then <tt>%A</tt> would
2441 have to have a cleared low bit. However, in the <tt>%C</tt> example, the
2442 optimizer is allowed to assume that the '<tt>undef</tt>' operand could be the
2443 same as <tt>%Y</tt>, allowing the whole '<tt>select</tt>' to be
2446 <pre class="doc_code">
2447 %A = xor undef, undef
2465 <p>This example points out that two '<tt>undef</tt>' operands are not
2466 necessarily the same. This can be surprising to people (and also matches C
2467 semantics) where they assume that "<tt>X^X</tt>" is always zero, even
2468 if <tt>X</tt> is undefined. This isn't true for a number of reasons, but the
2469 short answer is that an '<tt>undef</tt>' "variable" can arbitrarily change
2470 its value over its "live range". This is true because the variable doesn't
2471 actually <em>have a live range</em>. Instead, the value is logically read
2472 from arbitrary registers that happen to be around when needed, so the value
2473 is not necessarily consistent over time. In fact, <tt>%A</tt> and <tt>%C</tt>
2474 need to have the same semantics or the core LLVM "replace all uses with"
2475 concept would not hold.</p>
2477 <pre class="doc_code">
2485 <p>These examples show the crucial difference between an <em>undefined
2486 value</em> and <em>undefined behavior</em>. An undefined value (like
2487 '<tt>undef</tt>') is allowed to have an arbitrary bit-pattern. This means that
2488 the <tt>%A</tt> operation can be constant folded to '<tt>undef</tt>', because
2489 the '<tt>undef</tt>' could be an SNaN, and <tt>fdiv</tt> is not (currently)
2490 defined on SNaN's. However, in the second example, we can make a more
2491 aggressive assumption: because the <tt>undef</tt> is allowed to be an
2492 arbitrary value, we are allowed to assume that it could be zero. Since a
2493 divide by zero has <em>undefined behavior</em>, we are allowed to assume that
2494 the operation does not execute at all. This allows us to delete the divide and
2495 all code after it. Because the undefined operation "can't happen", the
2496 optimizer can assume that it occurs in dead code.</p>
2498 <pre class="doc_code">
2499 a: store undef -> %X
2500 b: store %X -> undef
2506 <p>These examples reiterate the <tt>fdiv</tt> example: a store <em>of</em> an
2507 undefined value can be assumed to not have any effect; we can assume that the
2508 value is overwritten with bits that happen to match what was already there.
2509 However, a store <em>to</em> an undefined location could clobber arbitrary
2510 memory, therefore, it has undefined behavior.</p>
2514 <!-- ======================================================================= -->
2516 <a name="poisonvalues">Poison Values</a>
2521 <p>Poison values are similar to <a href="#undefvalues">undef values</a>, however
2522 they also represent the fact that an instruction or constant expression which
2523 cannot evoke side effects has nevertheless detected a condition which results
2524 in undefined behavior.</p>
2526 <p>There is currently no way of representing a poison value in the IR; they
2527 only exist when produced by operations such as
2528 <a href="#i_add"><tt>add</tt></a> with the <tt>nsw</tt> flag.</p>
2530 <p>Poison value behavior is defined in terms of value <i>dependence</i>:</p>
2533 <li>Values other than <a href="#i_phi"><tt>phi</tt></a> nodes depend on
2534 their operands.</li>
2536 <li><a href="#i_phi"><tt>Phi</tt></a> nodes depend on the operand corresponding
2537 to their dynamic predecessor basic block.</li>
2539 <li>Function arguments depend on the corresponding actual argument values in
2540 the dynamic callers of their functions.</li>
2542 <li><a href="#i_call"><tt>Call</tt></a> instructions depend on the
2543 <a href="#i_ret"><tt>ret</tt></a> instructions that dynamically transfer
2544 control back to them.</li>
2546 <li><a href="#i_invoke"><tt>Invoke</tt></a> instructions depend on the
2547 <a href="#i_ret"><tt>ret</tt></a>, <a href="#i_unwind"><tt>unwind</tt></a>,
2548 or exception-throwing call instructions that dynamically transfer control
2551 <li>Non-volatile loads and stores depend on the most recent stores to all of the
2552 referenced memory addresses, following the order in the IR
2553 (including loads and stores implied by intrinsics such as
2554 <a href="#int_memcpy"><tt>@llvm.memcpy</tt></a>.)</li>
2556 <!-- TODO: In the case of multiple threads, this only applies if the store
2557 "happens-before" the load or store. -->
2559 <!-- TODO: floating-point exception state -->
2561 <li>An instruction with externally visible side effects depends on the most
2562 recent preceding instruction with externally visible side effects, following
2563 the order in the IR. (This includes
2564 <a href="#volatile">volatile operations</a>.)</li>
2566 <li>An instruction <i>control-depends</i> on a
2567 <a href="#terminators">terminator instruction</a>
2568 if the terminator instruction has multiple successors and the instruction
2569 is always executed when control transfers to one of the successors, and
2570 may not be executed when control is transferred to another.</li>
2572 <li>Additionally, an instruction also <i>control-depends</i> on a terminator
2573 instruction if the set of instructions it otherwise depends on would be
2574 different if the terminator had transferred control to a different
2577 <li>Dependence is transitive.</li>
2581 <p>Poison Values have the same behavior as <a href="#undefvalues">undef values</a>,
2582 with the additional affect that any instruction which has a <i>dependence</i>
2583 on a poison value has undefined behavior.</p>
2585 <p>Here are some examples:</p>
2587 <pre class="doc_code">
2589 %poison = sub nuw i32 0, 1 ; Results in a poison value.
2590 %still_poison = and i32 %poison, 0 ; 0, but also poison.
2591 %poison_yet_again = getelementptr i32* @h, i32 %still_poison
2592 store i32 0, i32* %poison_yet_again ; memory at @h[0] is poisoned
2594 store i32 %poison, i32* @g ; Poison value stored to memory.
2595 %poison2 = load i32* @g ; Poison value loaded back from memory.
2597 store volatile i32 %poison, i32* @g ; External observation; undefined behavior.
2599 %narrowaddr = bitcast i32* @g to i16*
2600 %wideaddr = bitcast i32* @g to i64*
2601 %poison3 = load i16* %narrowaddr ; Returns a poison value.
2602 %poison4 = load i64* %wideaddr ; Returns a poison value.
2604 %cmp = icmp slt i32 %poison, 0 ; Returns a poison value.
2605 br i1 %cmp, label %true, label %end ; Branch to either destination.
2608 store volatile i32 0, i32* @g ; This is control-dependent on %cmp, so
2609 ; it has undefined behavior.
2613 %p = phi i32 [ 0, %entry ], [ 1, %true ]
2614 ; Both edges into this PHI are
2615 ; control-dependent on %cmp, so this
2616 ; always results in a poison value.
2618 store volatile i32 0, i32* @g ; This would depend on the store in %true
2619 ; if %cmp is true, or the store in %entry
2620 ; otherwise, so this is undefined behavior.
2622 br i1 %cmp, label %second_true, label %second_end
2623 ; The same branch again, but this time the
2624 ; true block doesn't have side effects.
2631 store volatile i32 0, i32* @g ; This time, the instruction always depends
2632 ; on the store in %end. Also, it is
2633 ; control-equivalent to %end, so this is
2634 ; well-defined (ignoring earlier undefined
2635 ; behavior in this example).
2640 <!-- ======================================================================= -->
2642 <a name="blockaddress">Addresses of Basic Blocks</a>
2647 <p><b><tt>blockaddress(@function, %block)</tt></b></p>
2649 <p>The '<tt>blockaddress</tt>' constant computes the address of the specified
2650 basic block in the specified function, and always has an i8* type. Taking
2651 the address of the entry block is illegal.</p>
2653 <p>This value only has defined behavior when used as an operand to the
2654 '<a href="#i_indirectbr"><tt>indirectbr</tt></a>' instruction, or for
2655 comparisons against null. Pointer equality tests between labels addresses
2656 results in undefined behavior — though, again, comparison against null
2657 is ok, and no label is equal to the null pointer. This may be passed around
2658 as an opaque pointer sized value as long as the bits are not inspected. This
2659 allows <tt>ptrtoint</tt> and arithmetic to be performed on these values so
2660 long as the original value is reconstituted before the <tt>indirectbr</tt>
2663 <p>Finally, some targets may provide defined semantics when using the value as
2664 the operand to an inline assembly, but that is target specific.</p>
2669 <!-- ======================================================================= -->
2671 <a name="constantexprs">Constant Expressions</a>
2676 <p>Constant expressions are used to allow expressions involving other constants
2677 to be used as constants. Constant expressions may be of
2678 any <a href="#t_firstclass">first class</a> type and may involve any LLVM
2679 operation that does not have side effects (e.g. load and call are not
2680 supported). The following is the syntax for constant expressions:</p>
2683 <dt><b><tt>trunc (CST to TYPE)</tt></b></dt>
2684 <dd>Truncate a constant to another type. The bit size of CST must be larger
2685 than the bit size of TYPE. Both types must be integers.</dd>
2687 <dt><b><tt>zext (CST to TYPE)</tt></b></dt>
2688 <dd>Zero extend a constant to another type. The bit size of CST must be
2689 smaller than the bit size of TYPE. Both types must be integers.</dd>
2691 <dt><b><tt>sext (CST to TYPE)</tt></b></dt>
2692 <dd>Sign extend a constant to another type. The bit size of CST must be
2693 smaller than the bit size of TYPE. Both types must be integers.</dd>
2695 <dt><b><tt>fptrunc (CST to TYPE)</tt></b></dt>
2696 <dd>Truncate a floating point constant to another floating point type. The
2697 size of CST must be larger than the size of TYPE. Both types must be
2698 floating point.</dd>
2700 <dt><b><tt>fpext (CST to TYPE)</tt></b></dt>
2701 <dd>Floating point extend a constant to another type. The size of CST must be
2702 smaller or equal to the size of TYPE. Both types must be floating
2705 <dt><b><tt>fptoui (CST to TYPE)</tt></b></dt>
2706 <dd>Convert a floating point constant to the corresponding unsigned integer
2707 constant. TYPE must be a scalar or vector integer type. CST must be of
2708 scalar or vector floating point type. Both CST and TYPE must be scalars,
2709 or vectors of the same number of elements. If the value won't fit in the
2710 integer type, the results are undefined.</dd>
2712 <dt><b><tt>fptosi (CST to TYPE)</tt></b></dt>
2713 <dd>Convert a floating point constant to the corresponding signed integer
2714 constant. TYPE must be a scalar or vector integer type. CST must be of
2715 scalar or vector floating point type. Both CST and TYPE must be scalars,
2716 or vectors of the same number of elements. If the value won't fit in the
2717 integer type, the results are undefined.</dd>
2719 <dt><b><tt>uitofp (CST to TYPE)</tt></b></dt>
2720 <dd>Convert an unsigned integer constant to the corresponding floating point
2721 constant. TYPE must be a scalar or vector floating point type. CST must be
2722 of scalar or vector integer type. Both CST and TYPE must be scalars, or
2723 vectors of the same number of elements. If the value won't fit in the
2724 floating point type, the results are undefined.</dd>
2726 <dt><b><tt>sitofp (CST to TYPE)</tt></b></dt>
2727 <dd>Convert a signed integer constant to the corresponding floating point
2728 constant. TYPE must be a scalar or vector floating point type. CST must be
2729 of scalar or vector integer type. Both CST and TYPE must be scalars, or
2730 vectors of the same number of elements. If the value won't fit in the
2731 floating point type, the results are undefined.</dd>
2733 <dt><b><tt>ptrtoint (CST to TYPE)</tt></b></dt>
2734 <dd>Convert a pointer typed constant to the corresponding integer constant
2735 <tt>TYPE</tt> must be an integer type. <tt>CST</tt> must be of pointer
2736 type. The <tt>CST</tt> value is zero extended, truncated, or unchanged to
2737 make it fit in <tt>TYPE</tt>.</dd>
2739 <dt><b><tt>inttoptr (CST to TYPE)</tt></b></dt>
2740 <dd>Convert a integer constant to a pointer constant. TYPE must be a pointer
2741 type. CST must be of integer type. The CST value is zero extended,
2742 truncated, or unchanged to make it fit in a pointer size. This one is
2743 <i>really</i> dangerous!</dd>
2745 <dt><b><tt>bitcast (CST to TYPE)</tt></b></dt>
2746 <dd>Convert a constant, CST, to another TYPE. The constraints of the operands
2747 are the same as those for the <a href="#i_bitcast">bitcast
2748 instruction</a>.</dd>
2750 <dt><b><tt>getelementptr (CSTPTR, IDX0, IDX1, ...)</tt></b></dt>
2751 <dt><b><tt>getelementptr inbounds (CSTPTR, IDX0, IDX1, ...)</tt></b></dt>
2752 <dd>Perform the <a href="#i_getelementptr">getelementptr operation</a> on
2753 constants. As with the <a href="#i_getelementptr">getelementptr</a>
2754 instruction, the index list may have zero or more indexes, which are
2755 required to make sense for the type of "CSTPTR".</dd>
2757 <dt><b><tt>select (COND, VAL1, VAL2)</tt></b></dt>
2758 <dd>Perform the <a href="#i_select">select operation</a> on constants.</dd>
2760 <dt><b><tt>icmp COND (VAL1, VAL2)</tt></b></dt>
2761 <dd>Performs the <a href="#i_icmp">icmp operation</a> on constants.</dd>
2763 <dt><b><tt>fcmp COND (VAL1, VAL2)</tt></b></dt>
2764 <dd>Performs the <a href="#i_fcmp">fcmp operation</a> on constants.</dd>
2766 <dt><b><tt>extractelement (VAL, IDX)</tt></b></dt>
2767 <dd>Perform the <a href="#i_extractelement">extractelement operation</a> on
2770 <dt><b><tt>insertelement (VAL, ELT, IDX)</tt></b></dt>
2771 <dd>Perform the <a href="#i_insertelement">insertelement operation</a> on
2774 <dt><b><tt>shufflevector (VEC1, VEC2, IDXMASK)</tt></b></dt>
2775 <dd>Perform the <a href="#i_shufflevector">shufflevector operation</a> on
2778 <dt><b><tt>extractvalue (VAL, IDX0, IDX1, ...)</tt></b></dt>
2779 <dd>Perform the <a href="#i_extractvalue">extractvalue operation</a> on
2780 constants. The index list is interpreted in a similar manner as indices in
2781 a '<a href="#i_getelementptr">getelementptr</a>' operation. At least one
2782 index value must be specified.</dd>
2784 <dt><b><tt>insertvalue (VAL, ELT, IDX0, IDX1, ...)</tt></b></dt>
2785 <dd>Perform the <a href="#i_insertvalue">insertvalue operation</a> on
2786 constants. The index list is interpreted in a similar manner as indices in
2787 a '<a href="#i_getelementptr">getelementptr</a>' operation. At least one
2788 index value must be specified.</dd>
2790 <dt><b><tt>OPCODE (LHS, RHS)</tt></b></dt>
2791 <dd>Perform the specified operation of the LHS and RHS constants. OPCODE may
2792 be any of the <a href="#binaryops">binary</a>
2793 or <a href="#bitwiseops">bitwise binary</a> operations. The constraints
2794 on operands are the same as those for the corresponding instruction
2795 (e.g. no bitwise operations on floating point values are allowed).</dd>
2802 <!-- *********************************************************************** -->
2803 <h2><a name="othervalues">Other Values</a></h2>
2804 <!-- *********************************************************************** -->
2806 <!-- ======================================================================= -->
2808 <a name="inlineasm">Inline Assembler Expressions</a>
2813 <p>LLVM supports inline assembler expressions (as opposed
2814 to <a href="#moduleasm">Module-Level Inline Assembly</a>) through the use of
2815 a special value. This value represents the inline assembler as a string
2816 (containing the instructions to emit), a list of operand constraints (stored
2817 as a string), a flag that indicates whether or not the inline asm
2818 expression has side effects, and a flag indicating whether the function
2819 containing the asm needs to align its stack conservatively. An example
2820 inline assembler expression is:</p>
2822 <pre class="doc_code">
2823 i32 (i32) asm "bswap $0", "=r,r"
2826 <p>Inline assembler expressions may <b>only</b> be used as the callee operand of
2827 a <a href="#i_call"><tt>call</tt> instruction</a>. Thus, typically we
2830 <pre class="doc_code">
2831 %X = call i32 asm "<a href="#int_bswap">bswap</a> $0", "=r,r"(i32 %Y)
2834 <p>Inline asms with side effects not visible in the constraint list must be
2835 marked as having side effects. This is done through the use of the
2836 '<tt>sideeffect</tt>' keyword, like so:</p>
2838 <pre class="doc_code">
2839 call void asm sideeffect "eieio", ""()
2842 <p>In some cases inline asms will contain code that will not work unless the
2843 stack is aligned in some way, such as calls or SSE instructions on x86,
2844 yet will not contain code that does that alignment within the asm.
2845 The compiler should make conservative assumptions about what the asm might
2846 contain and should generate its usual stack alignment code in the prologue
2847 if the '<tt>alignstack</tt>' keyword is present:</p>
2849 <pre class="doc_code">
2850 call void asm alignstack "eieio", ""()
2853 <p>If both keywords appear the '<tt>sideeffect</tt>' keyword must come
2857 <p>TODO: The format of the asm and constraints string still need to be
2858 documented here. Constraints on what can be done (e.g. duplication, moving,
2859 etc need to be documented). This is probably best done by reference to
2860 another document that covers inline asm from a holistic perspective.</p>
2863 <!-- _______________________________________________________________________ -->
2865 <a name="inlineasm_md">Inline Asm Metadata</a>
2870 <p>The call instructions that wrap inline asm nodes may have a
2871 "<tt>!srcloc</tt>" MDNode attached to it that contains a list of constant
2872 integers. If present, the code generator will use the integer as the
2873 location cookie value when report errors through the <tt>LLVMContext</tt>
2874 error reporting mechanisms. This allows a front-end to correlate backend
2875 errors that occur with inline asm back to the source code that produced it.
2878 <pre class="doc_code">
2879 call void asm sideeffect "something bad", ""()<b>, !srcloc !42</b>
2881 !42 = !{ i32 1234567 }
2884 <p>It is up to the front-end to make sense of the magic numbers it places in the
2885 IR. If the MDNode contains multiple constants, the code generator will use
2886 the one that corresponds to the line of the asm that the error occurs on.</p>
2892 <!-- ======================================================================= -->
2894 <a name="metadata">Metadata Nodes and Metadata Strings</a>
2899 <p>LLVM IR allows metadata to be attached to instructions in the program that
2900 can convey extra information about the code to the optimizers and code
2901 generator. One example application of metadata is source-level debug
2902 information. There are two metadata primitives: strings and nodes. All
2903 metadata has the <tt>metadata</tt> type and is identified in syntax by a
2904 preceding exclamation point ('<tt>!</tt>').</p>
2906 <p>A metadata string is a string surrounded by double quotes. It can contain
2907 any character by escaping non-printable characters with "<tt>\xx</tt>" where
2908 "<tt>xx</tt>" is the two digit hex code. For example:
2909 "<tt>!"test\00"</tt>".</p>
2911 <p>Metadata nodes are represented with notation similar to structure constants
2912 (a comma separated list of elements, surrounded by braces and preceded by an
2913 exclamation point). Metadata nodes can have any values as their operand. For
2916 <div class="doc_code">
2918 !{ metadata !"test\00", i32 10}
2922 <p>A <a href="#namedmetadatastructure">named metadata</a> is a collection of
2923 metadata nodes, which can be looked up in the module symbol table. For
2926 <div class="doc_code">
2928 !foo = metadata !{!4, !3}
2932 <p>Metadata can be used as function arguments. Here <tt>llvm.dbg.value</tt>
2933 function is using two metadata arguments:</p>
2935 <div class="doc_code">
2937 call void @llvm.dbg.value(metadata !24, i64 0, metadata !25)
2941 <p>Metadata can be attached with an instruction. Here metadata <tt>!21</tt> is
2942 attached to the <tt>add</tt> instruction using the <tt>!dbg</tt>
2945 <div class="doc_code">
2947 %indvar.next = add i64 %indvar, 1, !dbg !21
2951 <p>More information about specific metadata nodes recognized by the optimizers
2952 and code generator is found below.</p>
2954 <!-- _______________________________________________________________________ -->
2956 <a name="tbaa">'<tt>tbaa</tt>' Metadata</a>
2961 <p>In LLVM IR, memory does not have types, so LLVM's own type system is not
2962 suitable for doing TBAA. Instead, metadata is added to the IR to describe
2963 a type system of a higher level language. This can be used to implement
2964 typical C/C++ TBAA, but it can also be used to implement custom alias
2965 analysis behavior for other languages.</p>
2967 <p>The current metadata format is very simple. TBAA metadata nodes have up to
2968 three fields, e.g.:</p>
2970 <div class="doc_code">
2972 !0 = metadata !{ metadata !"an example type tree" }
2973 !1 = metadata !{ metadata !"int", metadata !0 }
2974 !2 = metadata !{ metadata !"float", metadata !0 }
2975 !3 = metadata !{ metadata !"const float", metadata !2, i64 1 }
2979 <p>The first field is an identity field. It can be any value, usually
2980 a metadata string, which uniquely identifies the type. The most important
2981 name in the tree is the name of the root node. Two trees with
2982 different root node names are entirely disjoint, even if they
2983 have leaves with common names.</p>
2985 <p>The second field identifies the type's parent node in the tree, or
2986 is null or omitted for a root node. A type is considered to alias
2987 all of its descendants and all of its ancestors in the tree. Also,
2988 a type is considered to alias all types in other trees, so that
2989 bitcode produced from multiple front-ends is handled conservatively.</p>
2991 <p>If the third field is present, it's an integer which if equal to 1
2992 indicates that the type is "constant" (meaning
2993 <tt>pointsToConstantMemory</tt> should return true; see
2994 <a href="AliasAnalysis.html#OtherItfs">other useful
2995 <tt>AliasAnalysis</tt> methods</a>).</p>
2999 <!-- _______________________________________________________________________ -->
3001 <a name="fpaccuracy">'<tt>fpaccuracy</tt>' Metadata</a>
3006 <p><tt>fpaccuracy</tt> metadata may be attached to any instruction of floating
3007 point type. It expresses the maximum relative error of the result of
3008 that instruction, in ULPs. ULP is defined as follows:</p>
3012 <p>If <tt>x</tt> is a real number that lies between two finite consecutive
3013 floating-point numbers <tt>a</tt> and <tt>b</tt>, without being equal to one
3014 of them, then <tt>ulp(x) = |b - a|</tt>, otherwise <tt>ulp(x)</tt> is the
3015 distance between the two non-equal finite floating-point numbers nearest
3016 <tt>x</tt>. Moreover, <tt>ulp(NaN)</tt> is <tt>NaN</tt>.</p>
3020 <p>The maximum relative error may be any rational number. The metadata node
3021 shall consist of a pair of unsigned integers respectively representing
3022 the numerator and denominator. For example, 2.5 ULP:</p>
3024 <div class="doc_code">
3026 !0 = metadata !{ i32 5, i32 2 }
3034 <!-- ======================================================================= -->
3036 <a name="module_flags">Module Flags Metadata</a>
3041 <p>Occasionally, the front-end needs to transmit data to the linker which
3042 affects its behavior. The LLVM IR isn't sufficient to transmit this
3043 information, so one should use the <tt>llvm.module.flags</tt> named
3046 <p>The <tt>llvm.module.flags</tt> metadata is a named metadata, whose elements
3047 consist of metadata triplets. For example:</p>
3049 <pre class="doc_code">
3050 !0 = metadata !{ i32 0, metadata !"foo", i32 1 }
3051 !1 = metadata !{ i32 1, metadata !"bar", i32 37 }
3053 !llvm.module.flags = !{ !0, !1 }
3056 <p>The first field specifies the behavior of the linker upon encountering two of
3057 the same values. Behavior could range from: emitting an error if some of the
3058 modules' flags disagree, emitting a warning, etc. The second field is the
3059 name of the metadata. The third field is the value of the metadata.</p>
3061 <p>When two modules are linked together, the <tt>llvm.module.flags</tt> metadata
3062 are unioned together.</p>
3066 <!-- _______________________________________________________________________ -->
3068 <a name="objc_metadata">Objective-C Metadata</a>
3073 <p>The following module flags are used to convey Objective-C metadata to the
3076 <table border="1" cellspacing="0" cellpadding="4">
3086 <dt><tt>Error</tt></dt>
3087 <dd>Causes the linker to emit an error when two values disagree.</dd>
3095 <dt><tt>Require</tt></dt>
3096 <dd>Causes the linker to emit an error when the specified value is not
3105 <dt><tt>Override</tt></dt>
3106 <dd>Causes the linker to use the specified value if the two values
3107 disagree. It's an error if two pieces of the same metadata have
3108 the <tt>Override</tt> behavior but different values.</dd>
3115 <p>The names are:</p>
3118 <li><tt>Objective-C Version</tt></li>
3119 <li><tt>Objective-C Garbage Collection</tt></li>
3120 <li><tt>Objective-C GC Only</tt></li>
3121 <li><tt>Objective-C Image Info Section</tt></li>
3126 <p>Here is an example of how to use the Objective-C metadata:</p>
3128 <pre class="doc_code">
3130 !0 = metadata !{ i32 1, metadata !"Objective-C Version", i32 2 }
3131 !1 = metadata !{ i32 1, metadata !"Objective-C Garbage Collection", i32 2 }
3132 !2 = metadata !{ i32 1, metadata !"Objective-C Image Info Section",
3133 metadata !"__DATA, __objc_imageinfo, regular, no_dead_strip" }
3134 !llvm.module.flags = !{ !0, !1, !2 }
3137 !0 = metadata !{ i32 1, metadata !"Objective-C Version", i32 2 }
3138 !1 = metadata !{ i32 1, metadata !"Objective-C Garbage Collection", i32 2 }
3139 !2 = metadata !{ i32 1, metadata !"Objective-C GC Only", i32 4 }
3140 !3 = metadata !{ i32 1, metadata !"Objective-C Image Info Section",
3141 metadata !"__DATA, __objc_imageinfo, regular, no_dead_strip" }
3142 !4 = metadata !{ i32 2, metadata !"Objective-C GC Only",
3144 metadata !"Objective-C Garbage Collection", i32 2
3147 !llvm.module.flags = !{ !0, !1, !2, !3, !4 }
3149 <u>Linked Module</u>
3150 !0 = metadata !{ i32 1, metadata !"Objective-C Version", i32 2 }
3151 !1 = metadata !{ i32 3, metadata !"Objective-C Garbage Collection", i32 2 }
3152 !2 = metadata !{ i32 1, metadata !"Objective-C GC Only", i32 4 }
3153 !3 = metadata !{ i32 1, metadata !"Objective-C Image Info Section",
3154 metadata !"__DATA, __objc_imageinfo, regular, no_dead_strip" }
3155 !4 = metadata !{ i32 2, metadata !"Objective-C GC Only",
3157 metadata !"Objective-C Garbage Collection", i32 2
3160 !llvm.module.flags = !{ !0, !1, !2, !3, !4 }
3168 <!-- *********************************************************************** -->
3170 <a name="intrinsic_globals">Intrinsic Global Variables</a>
3172 <!-- *********************************************************************** -->
3174 <p>LLVM has a number of "magic" global variables that contain data that affect
3175 code generation or other IR semantics. These are documented here. All globals
3176 of this sort should have a section specified as "<tt>llvm.metadata</tt>". This
3177 section and all globals that start with "<tt>llvm.</tt>" are reserved for use
3180 <!-- ======================================================================= -->
3182 <a name="intg_used">The '<tt>llvm.used</tt>' Global Variable</a>
3187 <p>The <tt>@llvm.used</tt> global is an array with i8* element type which has <a
3188 href="#linkage_appending">appending linkage</a>. This array contains a list of
3189 pointers to global variables and functions which may optionally have a pointer
3190 cast formed of bitcast or getelementptr. For example, a legal use of it is:</p>
3192 <div class="doc_code">
3197 @llvm.used = appending global [2 x i8*] [
3199 i8* bitcast (i32* @Y to i8*)
3200 ], section "llvm.metadata"
3204 <p>If a global variable appears in the <tt>@llvm.used</tt> list, then the
3205 compiler, assembler, and linker are required to treat the symbol as if there
3206 is a reference to the global that it cannot see. For example, if a variable
3207 has internal linkage and no references other than that from
3208 the <tt>@llvm.used</tt> list, it cannot be deleted. This is commonly used to
3209 represent references from inline asms and other things the compiler cannot
3210 "see", and corresponds to "<tt>attribute((used))</tt>" in GNU C.</p>
3212 <p>On some targets, the code generator must emit a directive to the assembler or
3213 object file to prevent the assembler and linker from molesting the
3218 <!-- ======================================================================= -->
3220 <a name="intg_compiler_used">
3221 The '<tt>llvm.compiler.used</tt>' Global Variable
3227 <p>The <tt>@llvm.compiler.used</tt> directive is the same as the
3228 <tt>@llvm.used</tt> directive, except that it only prevents the compiler from
3229 touching the symbol. On targets that support it, this allows an intelligent
3230 linker to optimize references to the symbol without being impeded as it would
3231 be by <tt>@llvm.used</tt>.</p>
3233 <p>This is a rare construct that should only be used in rare circumstances, and
3234 should not be exposed to source languages.</p>
3238 <!-- ======================================================================= -->
3240 <a name="intg_global_ctors">The '<tt>llvm.global_ctors</tt>' Global Variable</a>
3245 <div class="doc_code">
3247 %0 = type { i32, void ()* }
3248 @llvm.global_ctors = appending global [1 x %0] [%0 { i32 65535, void ()* @ctor }]
3252 <p>The <tt>@llvm.global_ctors</tt> array contains a list of constructor
3253 functions and associated priorities. The functions referenced by this array
3254 will be called in ascending order of priority (i.e. lowest first) when the
3255 module is loaded. The order of functions with the same priority is not
3260 <!-- ======================================================================= -->
3262 <a name="intg_global_dtors">The '<tt>llvm.global_dtors</tt>' Global Variable</a>
3267 <div class="doc_code">
3269 %0 = type { i32, void ()* }
3270 @llvm.global_dtors = appending global [1 x %0] [%0 { i32 65535, void ()* @dtor }]
3274 <p>The <tt>@llvm.global_dtors</tt> array contains a list of destructor functions
3275 and associated priorities. The functions referenced by this array will be
3276 called in descending order of priority (i.e. highest first) when the module
3277 is loaded. The order of functions with the same priority is not defined.</p>
3283 <!-- *********************************************************************** -->
3284 <h2><a name="instref">Instruction Reference</a></h2>
3285 <!-- *********************************************************************** -->
3289 <p>The LLVM instruction set consists of several different classifications of
3290 instructions: <a href="#terminators">terminator
3291 instructions</a>, <a href="#binaryops">binary instructions</a>,
3292 <a href="#bitwiseops">bitwise binary instructions</a>,
3293 <a href="#memoryops">memory instructions</a>, and
3294 <a href="#otherops">other instructions</a>.</p>
3296 <!-- ======================================================================= -->
3298 <a name="terminators">Terminator Instructions</a>
3303 <p>As mentioned <a href="#functionstructure">previously</a>, every basic block
3304 in a program ends with a "Terminator" instruction, which indicates which
3305 block should be executed after the current block is finished. These
3306 terminator instructions typically yield a '<tt>void</tt>' value: they produce
3307 control flow, not values (the one exception being the
3308 '<a href="#i_invoke"><tt>invoke</tt></a>' instruction).</p>
3310 <p>The terminator instructions are:
3311 '<a href="#i_ret"><tt>ret</tt></a>',
3312 '<a href="#i_br"><tt>br</tt></a>',
3313 '<a href="#i_switch"><tt>switch</tt></a>',
3314 '<a href="#i_indirectbr"><tt>indirectbr</tt></a>',
3315 '<a href="#i_invoke"><tt>invoke</tt></a>',
3316 '<a href="#i_unwind"><tt>unwind</tt></a>',
3317 '<a href="#i_resume"><tt>resume</tt></a>', and
3318 '<a href="#i_unreachable"><tt>unreachable</tt></a>'.</p>
3320 <!-- _______________________________________________________________________ -->
3322 <a name="i_ret">'<tt>ret</tt>' Instruction</a>
3329 ret <type> <value> <i>; Return a value from a non-void function</i>
3330 ret void <i>; Return from void function</i>
3334 <p>The '<tt>ret</tt>' instruction is used to return control flow (and optionally
3335 a value) from a function back to the caller.</p>
3337 <p>There are two forms of the '<tt>ret</tt>' instruction: one that returns a
3338 value and then causes control flow, and one that just causes control flow to
3342 <p>The '<tt>ret</tt>' instruction optionally accepts a single argument, the
3343 return value. The type of the return value must be a
3344 '<a href="#t_firstclass">first class</a>' type.</p>
3346 <p>A function is not <a href="#wellformed">well formed</a> if it it has a
3347 non-void return type and contains a '<tt>ret</tt>' instruction with no return
3348 value or a return value with a type that does not match its type, or if it
3349 has a void return type and contains a '<tt>ret</tt>' instruction with a
3353 <p>When the '<tt>ret</tt>' instruction is executed, control flow returns back to
3354 the calling function's context. If the caller is a
3355 "<a href="#i_call"><tt>call</tt></a>" instruction, execution continues at the
3356 instruction after the call. If the caller was an
3357 "<a href="#i_invoke"><tt>invoke</tt></a>" instruction, execution continues at
3358 the beginning of the "normal" destination block. If the instruction returns
3359 a value, that value shall set the call or invoke instruction's return
3364 ret i32 5 <i>; Return an integer value of 5</i>
3365 ret void <i>; Return from a void function</i>
3366 ret { i32, i8 } { i32 4, i8 2 } <i>; Return a struct of values 4 and 2</i>
3370 <!-- _______________________________________________________________________ -->
3372 <a name="i_br">'<tt>br</tt>' Instruction</a>
3379 br i1 <cond>, label <iftrue>, label <iffalse>
3380 br label <dest> <i>; Unconditional branch</i>
3384 <p>The '<tt>br</tt>' instruction is used to cause control flow to transfer to a
3385 different basic block in the current function. There are two forms of this
3386 instruction, corresponding to a conditional branch and an unconditional
3390 <p>The conditional branch form of the '<tt>br</tt>' instruction takes a single
3391 '<tt>i1</tt>' value and two '<tt>label</tt>' values. The unconditional form
3392 of the '<tt>br</tt>' instruction takes a single '<tt>label</tt>' value as a
3396 <p>Upon execution of a conditional '<tt>br</tt>' instruction, the '<tt>i1</tt>'
3397 argument is evaluated. If the value is <tt>true</tt>, control flows to the
3398 '<tt>iftrue</tt>' <tt>label</tt> argument. If "cond" is <tt>false</tt>,
3399 control flows to the '<tt>iffalse</tt>' <tt>label</tt> argument.</p>
3404 %cond = <a href="#i_icmp">icmp</a> eq i32 %a, %b
3405 br i1 %cond, label %IfEqual, label %IfUnequal
3407 <a href="#i_ret">ret</a> i32 1
3409 <a href="#i_ret">ret</a> i32 0
3414 <!-- _______________________________________________________________________ -->
3416 <a name="i_switch">'<tt>switch</tt>' Instruction</a>
3423 switch <intty> <value>, label <defaultdest> [ <intty> <val>, label <dest> ... ]
3427 <p>The '<tt>switch</tt>' instruction is used to transfer control flow to one of
3428 several different places. It is a generalization of the '<tt>br</tt>'
3429 instruction, allowing a branch to occur to one of many possible
3433 <p>The '<tt>switch</tt>' instruction uses three parameters: an integer
3434 comparison value '<tt>value</tt>', a default '<tt>label</tt>' destination,
3435 and an array of pairs of comparison value constants and '<tt>label</tt>'s.
3436 The table is not allowed to contain duplicate constant entries.</p>
3439 <p>The <tt>switch</tt> instruction specifies a table of values and
3440 destinations. When the '<tt>switch</tt>' instruction is executed, this table
3441 is searched for the given value. If the value is found, control flow is
3442 transferred to the corresponding destination; otherwise, control flow is
3443 transferred to the default destination.</p>
3445 <h5>Implementation:</h5>
3446 <p>Depending on properties of the target machine and the particular
3447 <tt>switch</tt> instruction, this instruction may be code generated in
3448 different ways. For example, it could be generated as a series of chained
3449 conditional branches or with a lookup table.</p>
3453 <i>; Emulate a conditional br instruction</i>
3454 %Val = <a href="#i_zext">zext</a> i1 %value to i32
3455 switch i32 %Val, label %truedest [ i32 0, label %falsedest ]
3457 <i>; Emulate an unconditional br instruction</i>
3458 switch i32 0, label %dest [ ]
3460 <i>; Implement a jump table:</i>
3461 switch i32 %val, label %otherwise [ i32 0, label %onzero
3463 i32 2, label %ontwo ]
3469 <!-- _______________________________________________________________________ -->
3471 <a name="i_indirectbr">'<tt>indirectbr</tt>' Instruction</a>
3478 indirectbr <somety>* <address>, [ label <dest1>, label <dest2>, ... ]
3483 <p>The '<tt>indirectbr</tt>' instruction implements an indirect branch to a label
3484 within the current function, whose address is specified by
3485 "<tt>address</tt>". Address must be derived from a <a
3486 href="#blockaddress">blockaddress</a> constant.</p>
3490 <p>The '<tt>address</tt>' argument is the address of the label to jump to. The
3491 rest of the arguments indicate the full set of possible destinations that the
3492 address may point to. Blocks are allowed to occur multiple times in the
3493 destination list, though this isn't particularly useful.</p>
3495 <p>This destination list is required so that dataflow analysis has an accurate
3496 understanding of the CFG.</p>
3500 <p>Control transfers to the block specified in the address argument. All
3501 possible destination blocks must be listed in the label list, otherwise this
3502 instruction has undefined behavior. This implies that jumps to labels
3503 defined in other functions have undefined behavior as well.</p>
3505 <h5>Implementation:</h5>
3507 <p>This is typically implemented with a jump through a register.</p>
3511 indirectbr i8* %Addr, [ label %bb1, label %bb2, label %bb3 ]
3517 <!-- _______________________________________________________________________ -->
3519 <a name="i_invoke">'<tt>invoke</tt>' Instruction</a>
3526 <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>]
3527 to label <normal label> unwind label <exception label>
3531 <p>The '<tt>invoke</tt>' instruction causes control to transfer to a specified
3532 function, with the possibility of control flow transfer to either the
3533 '<tt>normal</tt>' label or the '<tt>exception</tt>' label. If the callee
3534 function returns with the "<tt><a href="#i_ret">ret</a></tt>" instruction,
3535 control flow will return to the "normal" label. If the callee (or any
3536 indirect callees) returns with the "<a href="#i_unwind"><tt>unwind</tt></a>"
3537 instruction, control is interrupted and continued at the dynamically nearest
3538 "exception" label.</p>
3540 <p>The '<tt>exception</tt>' label is a
3541 <i><a href="ExceptionHandling.html#overview">landing pad</a></i> for the
3542 exception. As such, '<tt>exception</tt>' label is required to have the
3543 "<a href="#i_landingpad"><tt>landingpad</tt></a>" instruction, which contains
3544 the information about the behavior of the program after unwinding
3545 happens, as its first non-PHI instruction. The restrictions on the
3546 "<tt>landingpad</tt>" instruction's tightly couples it to the
3547 "<tt>invoke</tt>" instruction, so that the important information contained
3548 within the "<tt>landingpad</tt>" instruction can't be lost through normal
3552 <p>This instruction requires several arguments:</p>
3555 <li>The optional "cconv" marker indicates which <a href="#callingconv">calling
3556 convention</a> the call should use. If none is specified, the call
3557 defaults to using C calling conventions.</li>
3559 <li>The optional <a href="#paramattrs">Parameter Attributes</a> list for
3560 return values. Only '<tt>zeroext</tt>', '<tt>signext</tt>', and
3561 '<tt>inreg</tt>' attributes are valid here.</li>
3563 <li>'<tt>ptr to function ty</tt>': shall be the signature of the pointer to
3564 function value being invoked. In most cases, this is a direct function
3565 invocation, but indirect <tt>invoke</tt>s are just as possible, branching
3566 off an arbitrary pointer to function value.</li>
3568 <li>'<tt>function ptr val</tt>': An LLVM value containing a pointer to a
3569 function to be invoked. </li>
3571 <li>'<tt>function args</tt>': argument list whose types match the function
3572 signature argument types and parameter attributes. All arguments must be
3573 of <a href="#t_firstclass">first class</a> type. If the function
3574 signature indicates the function accepts a variable number of arguments,
3575 the extra arguments can be specified.</li>
3577 <li>'<tt>normal label</tt>': the label reached when the called function
3578 executes a '<tt><a href="#i_ret">ret</a></tt>' instruction. </li>
3580 <li>'<tt>exception label</tt>': the label reached when a callee returns with
3581 the <a href="#i_unwind"><tt>unwind</tt></a> instruction. </li>
3583 <li>The optional <a href="#fnattrs">function attributes</a> list. Only
3584 '<tt>noreturn</tt>', '<tt>nounwind</tt>', '<tt>readonly</tt>' and
3585 '<tt>readnone</tt>' attributes are valid here.</li>
3589 <p>This instruction is designed to operate as a standard
3590 '<tt><a href="#i_call">call</a></tt>' instruction in most regards. The
3591 primary difference is that it establishes an association with a label, which
3592 is used by the runtime library to unwind the stack.</p>
3594 <p>This instruction is used in languages with destructors to ensure that proper
3595 cleanup is performed in the case of either a <tt>longjmp</tt> or a thrown
3596 exception. Additionally, this is important for implementation of
3597 '<tt>catch</tt>' clauses in high-level languages that support them.</p>
3599 <p>For the purposes of the SSA form, the definition of the value returned by the
3600 '<tt>invoke</tt>' instruction is deemed to occur on the edge from the current
3601 block to the "normal" label. If the callee unwinds then no return value is
3604 <p>Note that the code generator does not yet completely support unwind, and
3605 that the invoke/unwind semantics are likely to change in future versions.</p>
3609 %retval = invoke i32 @Test(i32 15) to label %Continue
3610 unwind label %TestCleanup <i>; {i32}:retval set</i>
3611 %retval = invoke <a href="#callingconv">coldcc</a> i32 %Testfnptr(i32 15) to label %Continue
3612 unwind label %TestCleanup <i>; {i32}:retval set</i>
3617 <!-- _______________________________________________________________________ -->
3620 <a name="i_unwind">'<tt>unwind</tt>' Instruction</a>
3631 <p>The '<tt>unwind</tt>' instruction unwinds the stack, continuing control flow
3632 at the first callee in the dynamic call stack which used
3633 an <a href="#i_invoke"><tt>invoke</tt></a> instruction to perform the call.
3634 This is primarily used to implement exception handling.</p>
3637 <p>The '<tt>unwind</tt>' instruction causes execution of the current function to
3638 immediately halt. The dynamic call stack is then searched for the
3639 first <a href="#i_invoke"><tt>invoke</tt></a> instruction on the call stack.
3640 Once found, execution continues at the "exceptional" destination block
3641 specified by the <tt>invoke</tt> instruction. If there is no <tt>invoke</tt>
3642 instruction in the dynamic call chain, undefined behavior results.</p>
3644 <p>Note that the code generator does not yet completely support unwind, and
3645 that the invoke/unwind semantics are likely to change in future versions.</p>
3649 <!-- _______________________________________________________________________ -->
3652 <a name="i_resume">'<tt>resume</tt>' Instruction</a>
3659 resume <type> <value>
3663 <p>The '<tt>resume</tt>' instruction is a terminator instruction that has no
3667 <p>The '<tt>resume</tt>' instruction requires one argument, which must have the
3668 same type as the result of any '<tt>landingpad</tt>' instruction in the same
3672 <p>The '<tt>resume</tt>' instruction resumes propagation of an existing
3673 (in-flight) exception whose unwinding was interrupted with
3674 a <a href="#i_landingpad"><tt>landingpad</tt></a> instruction.</p>
3678 resume { i8*, i32 } %exn
3683 <!-- _______________________________________________________________________ -->
3686 <a name="i_unreachable">'<tt>unreachable</tt>' Instruction</a>
3697 <p>The '<tt>unreachable</tt>' instruction has no defined semantics. This
3698 instruction is used to inform the optimizer that a particular portion of the
3699 code is not reachable. This can be used to indicate that the code after a
3700 no-return function cannot be reached, and other facts.</p>
3703 <p>The '<tt>unreachable</tt>' instruction has no defined semantics.</p>
3709 <!-- ======================================================================= -->
3711 <a name="binaryops">Binary Operations</a>
3716 <p>Binary operators are used to do most of the computation in a program. They
3717 require two operands of the same type, execute an operation on them, and
3718 produce a single value. The operands might represent multiple data, as is
3719 the case with the <a href="#t_vector">vector</a> data type. The result value
3720 has the same type as its operands.</p>
3722 <p>There are several different binary operators:</p>
3724 <!-- _______________________________________________________________________ -->
3726 <a name="i_add">'<tt>add</tt>' Instruction</a>
3733 <result> = add <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3734 <result> = add nuw <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3735 <result> = add nsw <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3736 <result> = add nuw nsw <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3740 <p>The '<tt>add</tt>' instruction returns the sum of its two operands.</p>
3743 <p>The two arguments to the '<tt>add</tt>' instruction must
3744 be <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of
3745 integer values. Both arguments must have identical types.</p>
3748 <p>The value produced is the integer sum of the two operands.</p>
3750 <p>If the sum has unsigned overflow, the result returned is the mathematical
3751 result modulo 2<sup>n</sup>, where n is the bit width of the result.</p>
3753 <p>Because LLVM integers use a two's complement representation, this instruction
3754 is appropriate for both signed and unsigned integers.</p>
3756 <p><tt>nuw</tt> and <tt>nsw</tt> stand for "No Unsigned Wrap"
3757 and "No Signed Wrap", respectively. If the <tt>nuw</tt> and/or
3758 <tt>nsw</tt> keywords are present, the result value of the <tt>add</tt>
3759 is a <a href="#poisonvalues">poison value</a> if unsigned and/or signed overflow,
3760 respectively, occurs.</p>
3764 <result> = add i32 4, %var <i>; yields {i32}:result = 4 + %var</i>
3769 <!-- _______________________________________________________________________ -->
3771 <a name="i_fadd">'<tt>fadd</tt>' Instruction</a>
3778 <result> = fadd <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3782 <p>The '<tt>fadd</tt>' instruction returns the sum of its two operands.</p>
3785 <p>The two arguments to the '<tt>fadd</tt>' instruction must be
3786 <a href="#t_floating">floating point</a> or <a href="#t_vector">vector</a> of
3787 floating point values. Both arguments must have identical types.</p>
3790 <p>The value produced is the floating point sum of the two operands.</p>
3794 <result> = fadd float 4.0, %var <i>; yields {float}:result = 4.0 + %var</i>
3799 <!-- _______________________________________________________________________ -->
3801 <a name="i_sub">'<tt>sub</tt>' Instruction</a>
3808 <result> = sub <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3809 <result> = sub nuw <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3810 <result> = sub nsw <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3811 <result> = sub nuw nsw <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3815 <p>The '<tt>sub</tt>' instruction returns the difference of its two
3818 <p>Note that the '<tt>sub</tt>' instruction is used to represent the
3819 '<tt>neg</tt>' instruction present in most other intermediate
3820 representations.</p>
3823 <p>The two arguments to the '<tt>sub</tt>' instruction must
3824 be <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of
3825 integer values. Both arguments must have identical types.</p>
3828 <p>The value produced is the integer difference of the two operands.</p>
3830 <p>If the difference has unsigned overflow, the result returned is the
3831 mathematical result modulo 2<sup>n</sup>, where n is the bit width of the
3834 <p>Because LLVM integers use a two's complement representation, this instruction
3835 is appropriate for both signed and unsigned integers.</p>
3837 <p><tt>nuw</tt> and <tt>nsw</tt> stand for "No Unsigned Wrap"
3838 and "No Signed Wrap", respectively. If the <tt>nuw</tt> and/or
3839 <tt>nsw</tt> keywords are present, the result value of the <tt>sub</tt>
3840 is a <a href="#poisonvalues">poison value</a> if unsigned and/or signed overflow,
3841 respectively, occurs.</p>
3845 <result> = sub i32 4, %var <i>; yields {i32}:result = 4 - %var</i>
3846 <result> = sub i32 0, %val <i>; yields {i32}:result = -%var</i>
3851 <!-- _______________________________________________________________________ -->
3853 <a name="i_fsub">'<tt>fsub</tt>' Instruction</a>
3860 <result> = fsub <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3864 <p>The '<tt>fsub</tt>' instruction returns the difference of its two
3867 <p>Note that the '<tt>fsub</tt>' instruction is used to represent the
3868 '<tt>fneg</tt>' instruction present in most other intermediate
3869 representations.</p>
3872 <p>The two arguments to the '<tt>fsub</tt>' instruction must be
3873 <a href="#t_floating">floating point</a> or <a href="#t_vector">vector</a> of
3874 floating point values. Both arguments must have identical types.</p>
3877 <p>The value produced is the floating point difference of the two operands.</p>
3881 <result> = fsub float 4.0, %var <i>; yields {float}:result = 4.0 - %var</i>
3882 <result> = fsub float -0.0, %val <i>; yields {float}:result = -%var</i>
3887 <!-- _______________________________________________________________________ -->
3889 <a name="i_mul">'<tt>mul</tt>' Instruction</a>
3896 <result> = mul <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3897 <result> = mul nuw <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3898 <result> = mul nsw <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3899 <result> = mul nuw nsw <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3903 <p>The '<tt>mul</tt>' instruction returns the product of its two operands.</p>
3906 <p>The two arguments to the '<tt>mul</tt>' instruction must
3907 be <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of
3908 integer values. Both arguments must have identical types.</p>
3911 <p>The value produced is the integer product of the two operands.</p>
3913 <p>If the result of the multiplication has unsigned overflow, the result
3914 returned is the mathematical result modulo 2<sup>n</sup>, where n is the bit
3915 width of the result.</p>
3917 <p>Because LLVM integers use a two's complement representation, and the result
3918 is the same width as the operands, this instruction returns the correct
3919 result for both signed and unsigned integers. If a full product
3920 (e.g. <tt>i32</tt>x<tt>i32</tt>-><tt>i64</tt>) is needed, the operands should
3921 be sign-extended or zero-extended as appropriate to the width of the full
3924 <p><tt>nuw</tt> and <tt>nsw</tt> stand for "No Unsigned Wrap"
3925 and "No Signed Wrap", respectively. If the <tt>nuw</tt> and/or
3926 <tt>nsw</tt> keywords are present, the result value of the <tt>mul</tt>
3927 is a <a href="#poisonvalues">poison value</a> if unsigned and/or signed overflow,
3928 respectively, occurs.</p>
3932 <result> = mul i32 4, %var <i>; yields {i32}:result = 4 * %var</i>
3937 <!-- _______________________________________________________________________ -->
3939 <a name="i_fmul">'<tt>fmul</tt>' Instruction</a>
3946 <result> = fmul <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3950 <p>The '<tt>fmul</tt>' instruction returns the product of its two operands.</p>
3953 <p>The two arguments to the '<tt>fmul</tt>' instruction must be
3954 <a href="#t_floating">floating point</a> or <a href="#t_vector">vector</a> of
3955 floating point values. Both arguments must have identical types.</p>
3958 <p>The value produced is the floating point product of the two operands.</p>
3962 <result> = fmul float 4.0, %var <i>; yields {float}:result = 4.0 * %var</i>
3967 <!-- _______________________________________________________________________ -->
3969 <a name="i_udiv">'<tt>udiv</tt>' Instruction</a>
3976 <result> = udiv <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3977 <result> = udiv exact <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3981 <p>The '<tt>udiv</tt>' instruction returns the quotient of its two operands.</p>
3984 <p>The two arguments to the '<tt>udiv</tt>' instruction must be
3985 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
3986 values. Both arguments must have identical types.</p>
3989 <p>The value produced is the unsigned integer quotient of the two operands.</p>
3991 <p>Note that unsigned integer division and signed integer division are distinct
3992 operations; for signed integer division, use '<tt>sdiv</tt>'.</p>
3994 <p>Division by zero leads to undefined behavior.</p>
3996 <p>If the <tt>exact</tt> keyword is present, the result value of the
3997 <tt>udiv</tt> is a <a href="#poisonvalues">poison value</a> if %op1 is not a
3998 multiple of %op2 (as such, "((a udiv exact b) mul b) == a").</p>
4003 <result> = udiv i32 4, %var <i>; yields {i32}:result = 4 / %var</i>
4008 <!-- _______________________________________________________________________ -->
4010 <a name="i_sdiv">'<tt>sdiv</tt>' Instruction</a>
4017 <result> = sdiv <ty> <op1>, <op2> <i>; yields {ty}:result</i>
4018 <result> = sdiv exact <ty> <op1>, <op2> <i>; yields {ty}:result</i>
4022 <p>The '<tt>sdiv</tt>' instruction returns the quotient of its two operands.</p>
4025 <p>The two arguments to the '<tt>sdiv</tt>' instruction must be
4026 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
4027 values. Both arguments must have identical types.</p>
4030 <p>The value produced is the signed integer quotient of the two operands rounded
4033 <p>Note that signed integer division and unsigned integer division are distinct
4034 operations; for unsigned integer division, use '<tt>udiv</tt>'.</p>
4036 <p>Division by zero leads to undefined behavior. Overflow also leads to
4037 undefined behavior; this is a rare case, but can occur, for example, by doing
4038 a 32-bit division of -2147483648 by -1.</p>
4040 <p>If the <tt>exact</tt> keyword is present, the result value of the
4041 <tt>sdiv</tt> is a <a href="#poisonvalues">poison value</a> if the result would
4046 <result> = sdiv i32 4, %var <i>; yields {i32}:result = 4 / %var</i>
4051 <!-- _______________________________________________________________________ -->
4053 <a name="i_fdiv">'<tt>fdiv</tt>' Instruction</a>
4060 <result> = fdiv <ty> <op1>, <op2> <i>; yields {ty}:result</i>
4064 <p>The '<tt>fdiv</tt>' instruction returns the quotient of its two operands.</p>
4067 <p>The two arguments to the '<tt>fdiv</tt>' instruction must be
4068 <a href="#t_floating">floating point</a> or <a href="#t_vector">vector</a> of
4069 floating point values. Both arguments must have identical types.</p>
4072 <p>The value produced is the floating point quotient of the two operands.</p>
4076 <result> = fdiv float 4.0, %var <i>; yields {float}:result = 4.0 / %var</i>
4081 <!-- _______________________________________________________________________ -->
4083 <a name="i_urem">'<tt>urem</tt>' Instruction</a>
4090 <result> = urem <ty> <op1>, <op2> <i>; yields {ty}:result</i>
4094 <p>The '<tt>urem</tt>' instruction returns the remainder from the unsigned
4095 division of its two arguments.</p>
4098 <p>The two arguments to the '<tt>urem</tt>' instruction must be
4099 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
4100 values. Both arguments must have identical types.</p>
4103 <p>This instruction returns the unsigned integer <i>remainder</i> of a division.
4104 This instruction always performs an unsigned division to get the
4107 <p>Note that unsigned integer remainder and signed integer remainder are
4108 distinct operations; for signed integer remainder, use '<tt>srem</tt>'.</p>
4110 <p>Taking the remainder of a division by zero leads to undefined behavior.</p>
4114 <result> = urem i32 4, %var <i>; yields {i32}:result = 4 % %var</i>
4119 <!-- _______________________________________________________________________ -->
4121 <a name="i_srem">'<tt>srem</tt>' Instruction</a>
4128 <result> = srem <ty> <op1>, <op2> <i>; yields {ty}:result</i>
4132 <p>The '<tt>srem</tt>' instruction returns the remainder from the signed
4133 division of its two operands. This instruction can also take
4134 <a href="#t_vector">vector</a> versions of the values in which case the
4135 elements must be integers.</p>
4138 <p>The two arguments to the '<tt>srem</tt>' instruction must be
4139 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
4140 values. Both arguments must have identical types.</p>
4143 <p>This instruction returns the <i>remainder</i> of a division (where the result
4144 is either zero or has the same sign as the dividend, <tt>op1</tt>), not the
4145 <i>modulo</i> operator (where the result is either zero or has the same sign
4146 as the divisor, <tt>op2</tt>) of a value.
4147 For more information about the difference,
4148 see <a href="http://mathforum.org/dr.math/problems/anne.4.28.99.html">The
4149 Math Forum</a>. For a table of how this is implemented in various languages,
4150 please see <a href="http://en.wikipedia.org/wiki/Modulo_operation">
4151 Wikipedia: modulo operation</a>.</p>
4153 <p>Note that signed integer remainder and unsigned integer remainder are
4154 distinct operations; for unsigned integer remainder, use '<tt>urem</tt>'.</p>
4156 <p>Taking the remainder of a division by zero leads to undefined behavior.
4157 Overflow also leads to undefined behavior; this is a rare case, but can
4158 occur, for example, by taking the remainder of a 32-bit division of
4159 -2147483648 by -1. (The remainder doesn't actually overflow, but this rule
4160 lets srem be implemented using instructions that return both the result of
4161 the division and the remainder.)</p>
4165 <result> = srem i32 4, %var <i>; yields {i32}:result = 4 % %var</i>
4170 <!-- _______________________________________________________________________ -->
4172 <a name="i_frem">'<tt>frem</tt>' Instruction</a>
4179 <result> = frem <ty> <op1>, <op2> <i>; yields {ty}:result</i>
4183 <p>The '<tt>frem</tt>' instruction returns the remainder from the division of
4184 its two operands.</p>
4187 <p>The two arguments to the '<tt>frem</tt>' instruction must be
4188 <a href="#t_floating">floating point</a> or <a href="#t_vector">vector</a> of
4189 floating point values. Both arguments must have identical types.</p>
4192 <p>This instruction returns the <i>remainder</i> of a division. The remainder
4193 has the same sign as the dividend.</p>
4197 <result> = frem float 4.0, %var <i>; yields {float}:result = 4.0 % %var</i>
4204 <!-- ======================================================================= -->
4206 <a name="bitwiseops">Bitwise Binary Operations</a>
4211 <p>Bitwise binary operators are used to do various forms of bit-twiddling in a
4212 program. They are generally very efficient instructions and can commonly be
4213 strength reduced from other instructions. They require two operands of the
4214 same type, execute an operation on them, and produce a single value. The
4215 resulting value is the same type as its operands.</p>
4217 <!-- _______________________________________________________________________ -->
4219 <a name="i_shl">'<tt>shl</tt>' Instruction</a>
4226 <result> = shl <ty> <op1>, <op2> <i>; yields {ty}:result</i>
4227 <result> = shl nuw <ty> <op1>, <op2> <i>; yields {ty}:result</i>
4228 <result> = shl nsw <ty> <op1>, <op2> <i>; yields {ty}:result</i>
4229 <result> = shl nuw nsw <ty> <op1>, <op2> <i>; yields {ty}:result</i>
4233 <p>The '<tt>shl</tt>' instruction returns the first operand shifted to the left
4234 a specified number of bits.</p>
4237 <p>Both arguments to the '<tt>shl</tt>' instruction must be the
4238 same <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of
4239 integer type. '<tt>op2</tt>' is treated as an unsigned value.</p>
4242 <p>The value produced is <tt>op1</tt> * 2<sup><tt>op2</tt></sup> mod
4243 2<sup>n</sup>, where <tt>n</tt> is the width of the result. If <tt>op2</tt>
4244 is (statically or dynamically) negative or equal to or larger than the number
4245 of bits in <tt>op1</tt>, the result is undefined. If the arguments are
4246 vectors, each vector element of <tt>op1</tt> is shifted by the corresponding
4247 shift amount in <tt>op2</tt>.</p>
4249 <p>If the <tt>nuw</tt> keyword is present, then the shift produces a
4250 <a href="#poisonvalues">poison value</a> if it shifts out any non-zero bits. If
4251 the <tt>nsw</tt> keyword is present, then the shift produces a
4252 <a href="#poisonvalues">poison value</a> if it shifts out any bits that disagree
4253 with the resultant sign bit. As such, NUW/NSW have the same semantics as
4254 they would if the shift were expressed as a mul instruction with the same
4255 nsw/nuw bits in (mul %op1, (shl 1, %op2)).</p>
4259 <result> = shl i32 4, %var <i>; yields {i32}: 4 << %var</i>
4260 <result> = shl i32 4, 2 <i>; yields {i32}: 16</i>
4261 <result> = shl i32 1, 10 <i>; yields {i32}: 1024</i>
4262 <result> = shl i32 1, 32 <i>; undefined</i>
4263 <result> = shl <2 x i32> < i32 1, i32 1>, < i32 1, i32 2> <i>; yields: result=<2 x i32> < i32 2, i32 4></i>
4268 <!-- _______________________________________________________________________ -->
4270 <a name="i_lshr">'<tt>lshr</tt>' Instruction</a>
4277 <result> = lshr <ty> <op1>, <op2> <i>; yields {ty}:result</i>
4278 <result> = lshr exact <ty> <op1>, <op2> <i>; yields {ty}:result</i>
4282 <p>The '<tt>lshr</tt>' instruction (logical shift right) returns the first
4283 operand shifted to the right a specified number of bits with zero fill.</p>
4286 <p>Both arguments to the '<tt>lshr</tt>' instruction must be the same
4287 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
4288 type. '<tt>op2</tt>' is treated as an unsigned value.</p>
4291 <p>This instruction always performs a logical shift right operation. The most
4292 significant bits of the result will be filled with zero bits after the shift.
4293 If <tt>op2</tt> is (statically or dynamically) equal to or larger than the
4294 number of bits in <tt>op1</tt>, the result is undefined. If the arguments are
4295 vectors, each vector element of <tt>op1</tt> is shifted by the corresponding
4296 shift amount in <tt>op2</tt>.</p>
4298 <p>If the <tt>exact</tt> keyword is present, the result value of the
4299 <tt>lshr</tt> is a <a href="#poisonvalues">poison value</a> if any of the bits
4300 shifted out are non-zero.</p>
4305 <result> = lshr i32 4, 1 <i>; yields {i32}:result = 2</i>
4306 <result> = lshr i32 4, 2 <i>; yields {i32}:result = 1</i>
4307 <result> = lshr i8 4, 3 <i>; yields {i8}:result = 0</i>
4308 <result> = lshr i8 -2, 1 <i>; yields {i8}:result = 0x7FFFFFFF </i>
4309 <result> = lshr i32 1, 32 <i>; undefined</i>
4310 <result> = lshr <2 x i32> < i32 -2, i32 4>, < i32 1, i32 2> <i>; yields: result=<2 x i32> < i32 0x7FFFFFFF, i32 1></i>
4315 <!-- _______________________________________________________________________ -->
4317 <a name="i_ashr">'<tt>ashr</tt>' Instruction</a>
4324 <result> = ashr <ty> <op1>, <op2> <i>; yields {ty}:result</i>
4325 <result> = ashr exact <ty> <op1>, <op2> <i>; yields {ty}:result</i>
4329 <p>The '<tt>ashr</tt>' instruction (arithmetic shift right) returns the first
4330 operand shifted to the right a specified number of bits with sign
4334 <p>Both arguments to the '<tt>ashr</tt>' instruction must be the same
4335 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
4336 type. '<tt>op2</tt>' is treated as an unsigned value.</p>
4339 <p>This instruction always performs an arithmetic shift right operation, The
4340 most significant bits of the result will be filled with the sign bit
4341 of <tt>op1</tt>. If <tt>op2</tt> is (statically or dynamically) equal to or
4342 larger than the number of bits in <tt>op1</tt>, the result is undefined. If
4343 the arguments are vectors, each vector element of <tt>op1</tt> is shifted by
4344 the corresponding shift amount in <tt>op2</tt>.</p>
4346 <p>If the <tt>exact</tt> keyword is present, the result value of the
4347 <tt>ashr</tt> is a <a href="#poisonvalues">poison value</a> if any of the bits
4348 shifted out are non-zero.</p>
4352 <result> = ashr i32 4, 1 <i>; yields {i32}:result = 2</i>
4353 <result> = ashr i32 4, 2 <i>; yields {i32}:result = 1</i>
4354 <result> = ashr i8 4, 3 <i>; yields {i8}:result = 0</i>
4355 <result> = ashr i8 -2, 1 <i>; yields {i8}:result = -1</i>
4356 <result> = ashr i32 1, 32 <i>; undefined</i>
4357 <result> = ashr <2 x i32> < i32 -2, i32 4>, < i32 1, i32 3> <i>; yields: result=<2 x i32> < i32 -1, i32 0></i>
4362 <!-- _______________________________________________________________________ -->
4364 <a name="i_and">'<tt>and</tt>' Instruction</a>
4371 <result> = and <ty> <op1>, <op2> <i>; yields {ty}:result</i>
4375 <p>The '<tt>and</tt>' instruction returns the bitwise logical and of its two
4379 <p>The two arguments to the '<tt>and</tt>' instruction must be
4380 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
4381 values. Both arguments must have identical types.</p>
4384 <p>The truth table used for the '<tt>and</tt>' instruction is:</p>
4386 <table border="1" cellspacing="0" cellpadding="4">
4418 <result> = and i32 4, %var <i>; yields {i32}:result = 4 & %var</i>
4419 <result> = and i32 15, 40 <i>; yields {i32}:result = 8</i>
4420 <result> = and i32 4, 8 <i>; yields {i32}:result = 0</i>
4423 <!-- _______________________________________________________________________ -->
4425 <a name="i_or">'<tt>or</tt>' Instruction</a>
4432 <result> = or <ty> <op1>, <op2> <i>; yields {ty}:result</i>
4436 <p>The '<tt>or</tt>' instruction returns the bitwise logical inclusive or of its
4440 <p>The two arguments to the '<tt>or</tt>' instruction must be
4441 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
4442 values. Both arguments must have identical types.</p>
4445 <p>The truth table used for the '<tt>or</tt>' instruction is:</p>
4447 <table border="1" cellspacing="0" cellpadding="4">
4479 <result> = or i32 4, %var <i>; yields {i32}:result = 4 | %var</i>
4480 <result> = or i32 15, 40 <i>; yields {i32}:result = 47</i>
4481 <result> = or i32 4, 8 <i>; yields {i32}:result = 12</i>
4486 <!-- _______________________________________________________________________ -->
4488 <a name="i_xor">'<tt>xor</tt>' Instruction</a>
4495 <result> = xor <ty> <op1>, <op2> <i>; yields {ty}:result</i>
4499 <p>The '<tt>xor</tt>' instruction returns the bitwise logical exclusive or of
4500 its two operands. The <tt>xor</tt> is used to implement the "one's
4501 complement" operation, which is the "~" operator in C.</p>
4504 <p>The two arguments to the '<tt>xor</tt>' instruction must be
4505 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
4506 values. Both arguments must have identical types.</p>
4509 <p>The truth table used for the '<tt>xor</tt>' instruction is:</p>
4511 <table border="1" cellspacing="0" cellpadding="4">
4543 <result> = xor i32 4, %var <i>; yields {i32}:result = 4 ^ %var</i>
4544 <result> = xor i32 15, 40 <i>; yields {i32}:result = 39</i>
4545 <result> = xor i32 4, 8 <i>; yields {i32}:result = 12</i>
4546 <result> = xor i32 %V, -1 <i>; yields {i32}:result = ~%V</i>
4553 <!-- ======================================================================= -->
4555 <a name="vectorops">Vector Operations</a>
4560 <p>LLVM supports several instructions to represent vector operations in a
4561 target-independent manner. These instructions cover the element-access and
4562 vector-specific operations needed to process vectors effectively. While LLVM
4563 does directly support these vector operations, many sophisticated algorithms
4564 will want to use target-specific intrinsics to take full advantage of a
4565 specific target.</p>
4567 <!-- _______________________________________________________________________ -->
4569 <a name="i_extractelement">'<tt>extractelement</tt>' Instruction</a>
4576 <result> = extractelement <n x <ty>> <val>, i32 <idx> <i>; yields <ty></i>
4580 <p>The '<tt>extractelement</tt>' instruction extracts a single scalar element
4581 from a vector at a specified index.</p>
4585 <p>The first operand of an '<tt>extractelement</tt>' instruction is a value
4586 of <a href="#t_vector">vector</a> type. The second operand is an index
4587 indicating the position from which to extract the element. The index may be
4591 <p>The result is a scalar of the same type as the element type of
4592 <tt>val</tt>. Its value is the value at position <tt>idx</tt> of
4593 <tt>val</tt>. If <tt>idx</tt> exceeds the length of <tt>val</tt>, the
4594 results are undefined.</p>
4598 <result> = extractelement <4 x i32> %vec, i32 0 <i>; yields i32</i>
4603 <!-- _______________________________________________________________________ -->
4605 <a name="i_insertelement">'<tt>insertelement</tt>' Instruction</a>
4612 <result> = insertelement <n x <ty>> <val>, <ty> <elt>, i32 <idx> <i>; yields <n x <ty>></i>
4616 <p>The '<tt>insertelement</tt>' instruction inserts a scalar element into a
4617 vector at a specified index.</p>
4620 <p>The first operand of an '<tt>insertelement</tt>' instruction is a value
4621 of <a href="#t_vector">vector</a> type. The second operand is a scalar value
4622 whose type must equal the element type of the first operand. The third
4623 operand is an index indicating the position at which to insert the value.
4624 The index may be a variable.</p>
4627 <p>The result is a vector of the same type as <tt>val</tt>. Its element values
4628 are those of <tt>val</tt> except at position <tt>idx</tt>, where it gets the
4629 value <tt>elt</tt>. If <tt>idx</tt> exceeds the length of <tt>val</tt>, the
4630 results are undefined.</p>
4634 <result> = insertelement <4 x i32> %vec, i32 1, i32 0 <i>; yields <4 x i32></i>
4639 <!-- _______________________________________________________________________ -->
4641 <a name="i_shufflevector">'<tt>shufflevector</tt>' Instruction</a>
4648 <result> = shufflevector <n x <ty>> <v1>, <n x <ty>> <v2>, <m x i32> <mask> <i>; yields <m x <ty>></i>
4652 <p>The '<tt>shufflevector</tt>' instruction constructs a permutation of elements
4653 from two input vectors, returning a vector with the same element type as the
4654 input and length that is the same as the shuffle mask.</p>
4657 <p>The first two operands of a '<tt>shufflevector</tt>' instruction are vectors
4658 with types that match each other. The third argument is a shuffle mask whose
4659 element type is always 'i32'. The result of the instruction is a vector
4660 whose length is the same as the shuffle mask and whose element type is the
4661 same as the element type of the first two operands.</p>
4663 <p>The shuffle mask operand is required to be a constant vector with either
4664 constant integer or undef values.</p>
4667 <p>The elements of the two input vectors are numbered from left to right across
4668 both of the vectors. The shuffle mask operand specifies, for each element of
4669 the result vector, which element of the two input vectors the result element
4670 gets. The element selector may be undef (meaning "don't care") and the
4671 second operand may be undef if performing a shuffle from only one vector.</p>
4675 <result> = shufflevector <4 x i32> %v1, <4 x i32> %v2,
4676 <4 x i32> <i32 0, i32 4, i32 1, i32 5> <i>; yields <4 x i32></i>
4677 <result> = shufflevector <4 x i32> %v1, <4 x i32> undef,
4678 <4 x i32> <i32 0, i32 1, i32 2, i32 3> <i>; yields <4 x i32></i> - Identity shuffle.
4679 <result> = shufflevector <8 x i32> %v1, <8 x i32> undef,
4680 <4 x i32> <i32 0, i32 1, i32 2, i32 3> <i>; yields <4 x i32></i>
4681 <result> = shufflevector <4 x i32> %v1, <4 x i32> %v2,
4682 <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>
4689 <!-- ======================================================================= -->
4691 <a name="aggregateops">Aggregate Operations</a>
4696 <p>LLVM supports several instructions for working with
4697 <a href="#t_aggregate">aggregate</a> values.</p>
4699 <!-- _______________________________________________________________________ -->
4701 <a name="i_extractvalue">'<tt>extractvalue</tt>' Instruction</a>
4708 <result> = extractvalue <aggregate type> <val>, <idx>{, <idx>}*
4712 <p>The '<tt>extractvalue</tt>' instruction extracts the value of a member field
4713 from an <a href="#t_aggregate">aggregate</a> value.</p>
4716 <p>The first operand of an '<tt>extractvalue</tt>' instruction is a value
4717 of <a href="#t_struct">struct</a> or
4718 <a href="#t_array">array</a> type. The operands are constant indices to
4719 specify which value to extract in a similar manner as indices in a
4720 '<tt><a href="#i_getelementptr">getelementptr</a></tt>' instruction.</p>
4721 <p>The major differences to <tt>getelementptr</tt> indexing are:</p>
4723 <li>Since the value being indexed is not a pointer, the first index is
4724 omitted and assumed to be zero.</li>
4725 <li>At least one index must be specified.</li>
4726 <li>Not only struct indices but also array indices must be in
4731 <p>The result is the value at the position in the aggregate specified by the
4736 <result> = extractvalue {i32, float} %agg, 0 <i>; yields i32</i>
4741 <!-- _______________________________________________________________________ -->
4743 <a name="i_insertvalue">'<tt>insertvalue</tt>' Instruction</a>
4750 <result> = insertvalue <aggregate type> <val>, <ty> <elt>, <idx>{, <idx>}* <i>; yields <aggregate type></i>
4754 <p>The '<tt>insertvalue</tt>' instruction inserts a value into a member field
4755 in an <a href="#t_aggregate">aggregate</a> value.</p>
4758 <p>The first operand of an '<tt>insertvalue</tt>' instruction is a value
4759 of <a href="#t_struct">struct</a> or
4760 <a href="#t_array">array</a> type. The second operand is a first-class
4761 value to insert. The following operands are constant indices indicating
4762 the position at which to insert the value in a similar manner as indices in a
4763 '<tt><a href="#i_extractvalue">extractvalue</a></tt>' instruction. The
4764 value to insert must have the same type as the value identified by the
4768 <p>The result is an aggregate of the same type as <tt>val</tt>. Its value is
4769 that of <tt>val</tt> except that the value at the position specified by the
4770 indices is that of <tt>elt</tt>.</p>
4774 %agg1 = insertvalue {i32, float} undef, i32 1, 0 <i>; yields {i32 1, float undef}</i>
4775 %agg2 = insertvalue {i32, float} %agg1, float %val, 1 <i>; yields {i32 1, float %val}</i>
4776 %agg3 = insertvalue {i32, {float}} %agg1, float %val, 1, 0 <i>; yields {i32 1, float %val}</i>
4783 <!-- ======================================================================= -->
4785 <a name="memoryops">Memory Access and Addressing Operations</a>
4790 <p>A key design point of an SSA-based representation is how it represents
4791 memory. In LLVM, no memory locations are in SSA form, which makes things
4792 very simple. This section describes how to read, write, and allocate
4795 <!-- _______________________________________________________________________ -->
4797 <a name="i_alloca">'<tt>alloca</tt>' Instruction</a>
4804 <result> = alloca <type>[, <ty> <NumElements>][, align <alignment>] <i>; yields {type*}:result</i>
4808 <p>The '<tt>alloca</tt>' instruction allocates memory on the stack frame of the
4809 currently executing function, to be automatically released when this function
4810 returns to its caller. The object is always allocated in the generic address
4811 space (address space zero).</p>
4814 <p>The '<tt>alloca</tt>' instruction
4815 allocates <tt>sizeof(<type>)*NumElements</tt> bytes of memory on the
4816 runtime stack, returning a pointer of the appropriate type to the program.
4817 If "NumElements" is specified, it is the number of elements allocated,
4818 otherwise "NumElements" is defaulted to be one. If a constant alignment is
4819 specified, the value result of the allocation is guaranteed to be aligned to
4820 at least that boundary. If not specified, or if zero, the target can choose
4821 to align the allocation on any convenient boundary compatible with the
4824 <p>'<tt>type</tt>' may be any sized type.</p>
4827 <p>Memory is allocated; a pointer is returned. The operation is undefined if
4828 there is insufficient stack space for the allocation. '<tt>alloca</tt>'d
4829 memory is automatically released when the function returns. The
4830 '<tt>alloca</tt>' instruction is commonly used to represent automatic
4831 variables that must have an address available. When the function returns
4832 (either with the <tt><a href="#i_ret">ret</a></tt>
4833 or <tt><a href="#i_unwind">unwind</a></tt> instructions), the memory is
4834 reclaimed. Allocating zero bytes is legal, but the result is undefined.</p>
4838 %ptr = alloca i32 <i>; yields {i32*}:ptr</i>
4839 %ptr = alloca i32, i32 4 <i>; yields {i32*}:ptr</i>
4840 %ptr = alloca i32, i32 4, align 1024 <i>; yields {i32*}:ptr</i>
4841 %ptr = alloca i32, align 1024 <i>; yields {i32*}:ptr</i>
4846 <!-- _______________________________________________________________________ -->
4848 <a name="i_load">'<tt>load</tt>' Instruction</a>
4855 <result> = load [volatile] <ty>* <pointer>[, align <alignment>][, !nontemporal !<index>]
4856 <result> = load atomic [volatile] <ty>* <pointer> [singlethread] <ordering>, align <alignment>
4857 !<index> = !{ i32 1 }
4861 <p>The '<tt>load</tt>' instruction is used to read from memory.</p>
4864 <p>The argument to the '<tt>load</tt>' instruction specifies the memory address
4865 from which to load. The pointer must point to
4866 a <a href="#t_firstclass">first class</a> type. If the <tt>load</tt> is
4867 marked as <tt>volatile</tt>, then the optimizer is not allowed to modify the
4868 number or order of execution of this <tt>load</tt> with other <a
4869 href="#volatile">volatile operations</a>.</p>
4871 <p>If the <code>load</code> is marked as <code>atomic</code>, it takes an extra
4872 <a href="#ordering">ordering</a> and optional <code>singlethread</code>
4873 argument. The <code>release</code> and <code>acq_rel</code> orderings are
4874 not valid on <code>load</code> instructions. Atomic loads produce <a
4875 href="#memorymodel">defined</a> results when they may see multiple atomic
4876 stores. The type of the pointee must be an integer type whose bit width
4877 is a power of two greater than or equal to eight and less than or equal
4878 to a target-specific size limit. <code>align</code> must be explicitly
4879 specified on atomic loads, and the load has undefined behavior if the
4880 alignment is not set to a value which is at least the size in bytes of
4881 the pointee. <code>!nontemporal</code> does not have any defined semantics
4882 for atomic loads.</p>
4884 <p>The optional constant <tt>align</tt> argument specifies the alignment of the
4885 operation (that is, the alignment of the memory address). A value of 0 or an
4886 omitted <tt>align</tt> argument means that the operation has the preferential
4887 alignment for the target. It is the responsibility of the code emitter to
4888 ensure that the alignment information is correct. Overestimating the
4889 alignment results in undefined behavior. Underestimating the alignment may
4890 produce less efficient code. An alignment of 1 is always safe.</p>
4892 <p>The optional <tt>!nontemporal</tt> metadata must reference a single
4893 metatadata name <index> corresponding to a metadata node with
4894 one <tt>i32</tt> entry of value 1. The existence of
4895 the <tt>!nontemporal</tt> metatadata on the instruction tells the optimizer
4896 and code generator that this load is not expected to be reused in the cache.
4897 The code generator may select special instructions to save cache bandwidth,
4898 such as the <tt>MOVNT</tt> instruction on x86.</p>
4901 <p>The location of memory pointed to is loaded. If the value being loaded is of
4902 scalar type then the number of bytes read does not exceed the minimum number
4903 of bytes needed to hold all bits of the type. For example, loading an
4904 <tt>i24</tt> reads at most three bytes. When loading a value of a type like
4905 <tt>i20</tt> with a size that is not an integral number of bytes, the result
4906 is undefined if the value was not originally written using a store of the
4911 %ptr = <a href="#i_alloca">alloca</a> i32 <i>; yields {i32*}:ptr</i>
4912 <a href="#i_store">store</a> i32 3, i32* %ptr <i>; yields {void}</i>
4913 %val = load i32* %ptr <i>; yields {i32}:val = i32 3</i>
4918 <!-- _______________________________________________________________________ -->
4920 <a name="i_store">'<tt>store</tt>' Instruction</a>
4927 store [volatile] <ty> <value>, <ty>* <pointer>[, align <alignment>][, !nontemporal !<index>] <i>; yields {void}</i>
4928 store atomic [volatile] <ty> <value>, <ty>* <pointer> [singlethread] <ordering>, align <alignment> <i>; yields {void}</i>
4932 <p>The '<tt>store</tt>' instruction is used to write to memory.</p>
4935 <p>There are two arguments to the '<tt>store</tt>' instruction: a value to store
4936 and an address at which to store it. The type of the
4937 '<tt><pointer></tt>' operand must be a pointer to
4938 the <a href="#t_firstclass">first class</a> type of the
4939 '<tt><value></tt>' operand. If the <tt>store</tt> is marked as
4940 <tt>volatile</tt>, then the optimizer is not allowed to modify the number or
4941 order of execution of this <tt>store</tt> with other <a
4942 href="#volatile">volatile operations</a>.</p>
4944 <p>If the <code>store</code> is marked as <code>atomic</code>, it takes an extra
4945 <a href="#ordering">ordering</a> and optional <code>singlethread</code>
4946 argument. The <code>acquire</code> and <code>acq_rel</code> orderings aren't
4947 valid on <code>store</code> instructions. Atomic loads produce <a
4948 href="#memorymodel">defined</a> results when they may see multiple atomic
4949 stores. The type of the pointee must be an integer type whose bit width
4950 is a power of two greater than or equal to eight and less than or equal
4951 to a target-specific size limit. <code>align</code> must be explicitly
4952 specified on atomic stores, and the store has undefined behavior if the
4953 alignment is not set to a value which is at least the size in bytes of
4954 the pointee. <code>!nontemporal</code> does not have any defined semantics
4955 for atomic stores.</p>
4957 <p>The optional constant "align" argument specifies the alignment of the
4958 operation (that is, the alignment of the memory address). A value of 0 or an
4959 omitted "align" argument means that the operation has the preferential
4960 alignment for the target. It is the responsibility of the code emitter to
4961 ensure that the alignment information is correct. Overestimating the
4962 alignment results in an undefined behavior. Underestimating the alignment may
4963 produce less efficient code. An alignment of 1 is always safe.</p>
4965 <p>The optional !nontemporal metadata must reference a single metatadata
4966 name <index> corresponding to a metadata node with one i32 entry of
4967 value 1. The existence of the !nontemporal metatadata on the
4968 instruction tells the optimizer and code generator that this load is
4969 not expected to be reused in the cache. The code generator may
4970 select special instructions to save cache bandwidth, such as the
4971 MOVNT instruction on x86.</p>
4975 <p>The contents of memory are updated to contain '<tt><value></tt>' at the
4976 location specified by the '<tt><pointer></tt>' operand. If
4977 '<tt><value></tt>' is of scalar type then the number of bytes written
4978 does not exceed the minimum number of bytes needed to hold all bits of the
4979 type. For example, storing an <tt>i24</tt> writes at most three bytes. When
4980 writing a value of a type like <tt>i20</tt> with a size that is not an
4981 integral number of bytes, it is unspecified what happens to the extra bits
4982 that do not belong to the type, but they will typically be overwritten.</p>
4986 %ptr = <a href="#i_alloca">alloca</a> i32 <i>; yields {i32*}:ptr</i>
4987 store i32 3, i32* %ptr <i>; yields {void}</i>
4988 %val = <a href="#i_load">load</a> i32* %ptr <i>; yields {i32}:val = i32 3</i>
4993 <!-- _______________________________________________________________________ -->
4995 <a name="i_fence">'<tt>fence</tt>' Instruction</a>
5002 fence [singlethread] <ordering> <i>; yields {void}</i>
5006 <p>The '<tt>fence</tt>' instruction is used to introduce happens-before edges
5007 between operations.</p>
5009 <h5>Arguments:</h5> <p>'<code>fence</code>' instructions take an <a
5010 href="#ordering">ordering</a> argument which defines what
5011 <i>synchronizes-with</i> edges they add. They can only be given
5012 <code>acquire</code>, <code>release</code>, <code>acq_rel</code>, and
5013 <code>seq_cst</code> orderings.</p>
5016 <p>A fence <var>A</var> which has (at least) <code>release</code> ordering
5017 semantics <i>synchronizes with</i> a fence <var>B</var> with (at least)
5018 <code>acquire</code> ordering semantics if and only if there exist atomic
5019 operations <var>X</var> and <var>Y</var>, both operating on some atomic object
5020 <var>M</var>, such that <var>A</var> is sequenced before <var>X</var>,
5021 <var>X</var> modifies <var>M</var> (either directly or through some side effect
5022 of a sequence headed by <var>X</var>), <var>Y</var> is sequenced before
5023 <var>B</var>, and <var>Y</var> observes <var>M</var>. This provides a
5024 <i>happens-before</i> dependency between <var>A</var> and <var>B</var>. Rather
5025 than an explicit <code>fence</code>, one (but not both) of the atomic operations
5026 <var>X</var> or <var>Y</var> might provide a <code>release</code> or
5027 <code>acquire</code> (resp.) ordering constraint and still
5028 <i>synchronize-with</i> the explicit <code>fence</code> and establish the
5029 <i>happens-before</i> edge.</p>
5031 <p>A <code>fence</code> which has <code>seq_cst</code> ordering, in addition to
5032 having both <code>acquire</code> and <code>release</code> semantics specified
5033 above, participates in the global program order of other <code>seq_cst</code>
5034 operations and/or fences.</p>
5036 <p>The optional "<a href="#singlethread"><code>singlethread</code></a>" argument
5037 specifies that the fence only synchronizes with other fences in the same
5038 thread. (This is useful for interacting with signal handlers.)</p>
5042 fence acquire <i>; yields {void}</i>
5043 fence singlethread seq_cst <i>; yields {void}</i>
5048 <!-- _______________________________________________________________________ -->
5050 <a name="i_cmpxchg">'<tt>cmpxchg</tt>' Instruction</a>
5057 cmpxchg [volatile] <ty>* <pointer>, <ty> <cmp>, <ty> <new> [singlethread] <ordering> <i>; yields {ty}</i>
5061 <p>The '<tt>cmpxchg</tt>' instruction is used to atomically modify memory.
5062 It loads a value in memory and compares it to a given value. If they are
5063 equal, it stores a new value into the memory.</p>
5066 <p>There are three arguments to the '<code>cmpxchg</code>' instruction: an
5067 address to operate on, a value to compare to the value currently be at that
5068 address, and a new value to place at that address if the compared values are
5069 equal. The type of '<var><cmp></var>' must be an integer type whose
5070 bit width is a power of two greater than or equal to eight and less than
5071 or equal to a target-specific size limit. '<var><cmp></var>' and
5072 '<var><new></var>' must have the same type, and the type of
5073 '<var><pointer></var>' must be a pointer to that type. If the
5074 <code>cmpxchg</code> is marked as <code>volatile</code>, then the
5075 optimizer is not allowed to modify the number or order of execution
5076 of this <code>cmpxchg</code> with other <a href="#volatile">volatile
5079 <!-- FIXME: Extend allowed types. -->
5081 <p>The <a href="#ordering"><var>ordering</var></a> argument specifies how this
5082 <code>cmpxchg</code> synchronizes with other atomic operations.</p>
5084 <p>The optional "<code>singlethread</code>" argument declares that the
5085 <code>cmpxchg</code> is only atomic with respect to code (usually signal
5086 handlers) running in the same thread as the <code>cmpxchg</code>. Otherwise the
5087 cmpxchg is atomic with respect to all other code in the system.</p>
5089 <p>The pointer passed into cmpxchg must have alignment greater than or equal to
5090 the size in memory of the operand.
5093 <p>The contents of memory at the location specified by the
5094 '<tt><pointer></tt>' operand is read and compared to
5095 '<tt><cmp></tt>'; if the read value is the equal,
5096 '<tt><new></tt>' is written. The original value at the location
5099 <p>A successful <code>cmpxchg</code> is a read-modify-write instruction for the
5100 purpose of identifying <a href="#release_sequence">release sequences</a>. A
5101 failed <code>cmpxchg</code> is equivalent to an atomic load with an ordering
5102 parameter determined by dropping any <code>release</code> part of the
5103 <code>cmpxchg</code>'s ordering.</p>
5106 FIXME: Is compare_exchange_weak() necessary? (Consider after we've done
5107 optimization work on ARM.)
5109 FIXME: Is a weaker ordering constraint on failure helpful in practice?
5115 %orig = atomic <a href="#i_load">load</a> i32* %ptr unordered <i>; yields {i32}</i>
5116 <a href="#i_br">br</a> label %loop
5119 %cmp = <a href="#i_phi">phi</a> i32 [ %orig, %entry ], [%old, %loop]
5120 %squared = <a href="#i_mul">mul</a> i32 %cmp, %cmp
5121 %old = cmpxchg i32* %ptr, i32 %cmp, i32 %squared <i>; yields {i32}</i>
5122 %success = <a href="#i_icmp">icmp</a> eq i32 %cmp, %old
5123 <a href="#i_br">br</a> i1 %success, label %done, label %loop
5131 <!-- _______________________________________________________________________ -->
5133 <a name="i_atomicrmw">'<tt>atomicrmw</tt>' Instruction</a>
5140 atomicrmw [volatile] <operation> <ty>* <pointer>, <ty> <value> [singlethread] <ordering> <i>; yields {ty}</i>
5144 <p>The '<tt>atomicrmw</tt>' instruction is used to atomically modify memory.</p>
5147 <p>There are three arguments to the '<code>atomicrmw</code>' instruction: an
5148 operation to apply, an address whose value to modify, an argument to the
5149 operation. The operation must be one of the following keywords:</p>
5164 <p>The type of '<var><value></var>' must be an integer type whose
5165 bit width is a power of two greater than or equal to eight and less than
5166 or equal to a target-specific size limit. The type of the
5167 '<code><pointer></code>' operand must be a pointer to that type.
5168 If the <code>atomicrmw</code> is marked as <code>volatile</code>, then the
5169 optimizer is not allowed to modify the number or order of execution of this
5170 <code>atomicrmw</code> with other <a href="#volatile">volatile
5173 <!-- FIXME: Extend allowed types. -->
5176 <p>The contents of memory at the location specified by the
5177 '<tt><pointer></tt>' operand are atomically read, modified, and written
5178 back. The original value at the location is returned. The modification is
5179 specified by the <var>operation</var> argument:</p>
5182 <li>xchg: <code>*ptr = val</code></li>
5183 <li>add: <code>*ptr = *ptr + val</code></li>
5184 <li>sub: <code>*ptr = *ptr - val</code></li>
5185 <li>and: <code>*ptr = *ptr & val</code></li>
5186 <li>nand: <code>*ptr = ~(*ptr & val)</code></li>
5187 <li>or: <code>*ptr = *ptr | val</code></li>
5188 <li>xor: <code>*ptr = *ptr ^ val</code></li>
5189 <li>max: <code>*ptr = *ptr > val ? *ptr : val</code> (using a signed comparison)</li>
5190 <li>min: <code>*ptr = *ptr < val ? *ptr : val</code> (using a signed comparison)</li>
5191 <li>umax: <code>*ptr = *ptr > val ? *ptr : val</code> (using an unsigned comparison)</li>
5192 <li>umin: <code>*ptr = *ptr < val ? *ptr : val</code> (using an unsigned comparison)</li>
5197 %old = atomicrmw add i32* %ptr, i32 1 acquire <i>; yields {i32}</i>
5202 <!-- _______________________________________________________________________ -->
5204 <a name="i_getelementptr">'<tt>getelementptr</tt>' Instruction</a>
5211 <result> = getelementptr <pty>* <ptrval>{, <ty> <idx>}*
5212 <result> = getelementptr inbounds <pty>* <ptrval>{, <ty> <idx>}*
5213 <result> = getelementptr <ptr vector> ptrval, <vector index type> idx
5217 <p>The '<tt>getelementptr</tt>' instruction is used to get the address of a
5218 subelement of an <a href="#t_aggregate">aggregate</a> data structure.
5219 It performs address calculation only and does not access memory.</p>
5222 <p>The first argument is always a pointer or a vector of pointers,
5223 and forms the basis of the
5224 calculation. The remaining arguments are indices that indicate which of the
5225 elements of the aggregate object are indexed. The interpretation of each
5226 index is dependent on the type being indexed into. The first index always
5227 indexes the pointer value given as the first argument, the second index
5228 indexes a value of the type pointed to (not necessarily the value directly
5229 pointed to, since the first index can be non-zero), etc. The first type
5230 indexed into must be a pointer value, subsequent types can be arrays,
5231 vectors, and structs. Note that subsequent types being indexed into
5232 can never be pointers, since that would require loading the pointer before
5233 continuing calculation.</p>
5235 <p>The type of each index argument depends on the type it is indexing into.
5236 When indexing into a (optionally packed) structure, only <tt>i32</tt>
5237 integer <b>constants</b> are allowed. When indexing into an array, pointer
5238 or vector, integers of any width are allowed, and they are not required to be
5239 constant. These integers are treated as signed values where relevant.</p>
5241 <p>For example, let's consider a C code fragment and how it gets compiled to
5244 <pre class="doc_code">
5256 int *foo(struct ST *s) {
5257 return &s[1].Z.B[5][13];
5261 <p>The LLVM code generated by Clang is:</p>
5263 <pre class="doc_code">
5264 %struct.RT = <a href="#namedtypes">type</a> { i8, [10 x [20 x i32]], i8 }
5265 %struct.ST = <a href="#namedtypes">type</a> { i32, double, %struct.RT }
5267 define i32* @foo(%struct.ST* %s) nounwind uwtable readnone optsize ssp {
5269 %arrayidx = getelementptr inbounds %struct.ST* %s, i64 1, i32 2, i32 1, i64 5, i64 13
5275 <p>In the example above, the first index is indexing into the
5276 '<tt>%struct.ST*</tt>' type, which is a pointer, yielding a
5277 '<tt>%struct.ST</tt>' = '<tt>{ i32, double, %struct.RT }</tt>' type, a
5278 structure. The second index indexes into the third element of the structure,
5279 yielding a '<tt>%struct.RT</tt>' = '<tt>{ i8 , [10 x [20 x i32]], i8 }</tt>'
5280 type, another structure. The third index indexes into the second element of
5281 the structure, yielding a '<tt>[10 x [20 x i32]]</tt>' type, an array. The
5282 two dimensions of the array are subscripted into, yielding an '<tt>i32</tt>'
5283 type. The '<tt>getelementptr</tt>' instruction returns a pointer to this
5284 element, thus computing a value of '<tt>i32*</tt>' type.</p>
5286 <p>Note that it is perfectly legal to index partially through a structure,
5287 returning a pointer to an inner element. Because of this, the LLVM code for
5288 the given testcase is equivalent to:</p>
5290 <pre class="doc_code">
5291 define i32* @foo(%struct.ST* %s) {
5292 %t1 = getelementptr %struct.ST* %s, i32 1 <i>; yields %struct.ST*:%t1</i>
5293 %t2 = getelementptr %struct.ST* %t1, i32 0, i32 2 <i>; yields %struct.RT*:%t2</i>
5294 %t3 = getelementptr %struct.RT* %t2, i32 0, i32 1 <i>; yields [10 x [20 x i32]]*:%t3</i>
5295 %t4 = getelementptr [10 x [20 x i32]]* %t3, i32 0, i32 5 <i>; yields [20 x i32]*:%t4</i>
5296 %t5 = getelementptr [20 x i32]* %t4, i32 0, i32 13 <i>; yields i32*:%t5</i>
5301 <p>If the <tt>inbounds</tt> keyword is present, the result value of the
5302 <tt>getelementptr</tt> is a <a href="#poisonvalues">poison value</a> if the
5303 base pointer is not an <i>in bounds</i> address of an allocated object,
5304 or if any of the addresses that would be formed by successive addition of
5305 the offsets implied by the indices to the base address with infinitely
5306 precise signed arithmetic are not an <i>in bounds</i> address of that
5307 allocated object. The <i>in bounds</i> addresses for an allocated object
5308 are all the addresses that point into the object, plus the address one
5310 In cases where the base is a vector of pointers the <tt>inbounds</tt> keyword
5311 applies to each of the computations element-wise. </p>
5313 <p>If the <tt>inbounds</tt> keyword is not present, the offsets are added to
5314 the base address with silently-wrapping two's complement arithmetic. If the
5315 offsets have a different width from the pointer, they are sign-extended or
5316 truncated to the width of the pointer. The result value of the
5317 <tt>getelementptr</tt> may be outside the object pointed to by the base
5318 pointer. The result value may not necessarily be used to access memory
5319 though, even if it happens to point into allocated storage. See the
5320 <a href="#pointeraliasing">Pointer Aliasing Rules</a> section for more
5323 <p>The getelementptr instruction is often confusing. For some more insight into
5324 how it works, see <a href="GetElementPtr.html">the getelementptr FAQ</a>.</p>
5328 <i>; yields [12 x i8]*:aptr</i>
5329 %aptr = getelementptr {i32, [12 x i8]}* %saptr, i64 0, i32 1
5330 <i>; yields i8*:vptr</i>
5331 %vptr = getelementptr {i32, <2 x i8>}* %svptr, i64 0, i32 1, i32 1
5332 <i>; yields i8*:eptr</i>
5333 %eptr = getelementptr [12 x i8]* %aptr, i64 0, i32 1
5334 <i>; yields i32*:iptr</i>
5335 %iptr = getelementptr [10 x i32]* @arr, i16 0, i16 0
5338 <p>In cases where the pointer argument is a vector of pointers, only a
5339 single index may be used, and the number of vector elements has to be
5340 the same. For example: </p>
5341 <pre class="doc_code">
5342 %A = getelementptr <4 x i8*> %ptrs, <4 x i64> %offsets,
5349 <!-- ======================================================================= -->
5351 <a name="convertops">Conversion Operations</a>
5356 <p>The instructions in this category are the conversion instructions (casting)
5357 which all take a single operand and a type. They perform various bit
5358 conversions on the operand.</p>
5360 <!-- _______________________________________________________________________ -->
5362 <a name="i_trunc">'<tt>trunc .. to</tt>' Instruction</a>
5369 <result> = trunc <ty> <value> to <ty2> <i>; yields ty2</i>
5373 <p>The '<tt>trunc</tt>' instruction truncates its operand to the
5374 type <tt>ty2</tt>.</p>
5377 <p>The '<tt>trunc</tt>' instruction takes a value to trunc, and a type to trunc it to.
5378 Both types must be of <a href="#t_integer">integer</a> types, or vectors
5379 of the same number of integers.
5380 The bit size of the <tt>value</tt> must be larger than
5381 the bit size of the destination type, <tt>ty2</tt>.
5382 Equal sized types are not allowed.</p>
5385 <p>The '<tt>trunc</tt>' instruction truncates the high order bits
5386 in <tt>value</tt> and converts the remaining bits to <tt>ty2</tt>. Since the
5387 source size must be larger than the destination size, <tt>trunc</tt> cannot
5388 be a <i>no-op cast</i>. It will always truncate bits.</p>
5392 %X = trunc i32 257 to i8 <i>; yields i8:1</i>
5393 %Y = trunc i32 123 to i1 <i>; yields i1:true</i>
5394 %Z = trunc i32 122 to i1 <i>; yields i1:false</i>
5395 %W = trunc <2 x i16> <i16 8, i16 7> to <2 x i8> <i>; yields <i8 8, i8 7></i>
5400 <!-- _______________________________________________________________________ -->
5402 <a name="i_zext">'<tt>zext .. to</tt>' Instruction</a>
5409 <result> = zext <ty> <value> to <ty2> <i>; yields ty2</i>
5413 <p>The '<tt>zext</tt>' instruction zero extends its operand to type
5418 <p>The '<tt>zext</tt>' instruction takes a value to cast, and a type to cast it to.
5419 Both types must be of <a href="#t_integer">integer</a> types, or vectors
5420 of the same number of integers.
5421 The bit size of the <tt>value</tt> must be smaller than
5422 the bit size of the destination type,
5426 <p>The <tt>zext</tt> fills the high order bits of the <tt>value</tt> with zero
5427 bits until it reaches the size of the destination type, <tt>ty2</tt>.</p>
5429 <p>When zero extending from i1, the result will always be either 0 or 1.</p>
5433 %X = zext i32 257 to i64 <i>; yields i64:257</i>
5434 %Y = zext i1 true to i32 <i>; yields i32:1</i>
5435 %Z = zext <2 x i16> <i16 8, i16 7> to <2 x i32> <i>; yields <i32 8, i32 7></i>
5440 <!-- _______________________________________________________________________ -->
5442 <a name="i_sext">'<tt>sext .. to</tt>' Instruction</a>
5449 <result> = sext <ty> <value> to <ty2> <i>; yields ty2</i>
5453 <p>The '<tt>sext</tt>' sign extends <tt>value</tt> to the type <tt>ty2</tt>.</p>
5456 <p>The '<tt>sext</tt>' instruction takes a value to cast, and a type to cast it to.
5457 Both types must be of <a href="#t_integer">integer</a> types, or vectors
5458 of the same number of integers.
5459 The bit size of the <tt>value</tt> must be smaller than
5460 the bit size of the destination type,
5464 <p>The '<tt>sext</tt>' instruction performs a sign extension by copying the sign
5465 bit (highest order bit) of the <tt>value</tt> until it reaches the bit size
5466 of the type <tt>ty2</tt>.</p>
5468 <p>When sign extending from i1, the extension always results in -1 or 0.</p>
5472 %X = sext i8 -1 to i16 <i>; yields i16 :65535</i>
5473 %Y = sext i1 true to i32 <i>; yields i32:-1</i>
5474 %Z = sext <2 x i16> <i16 8, i16 7> to <2 x i32> <i>; yields <i32 8, i32 7></i>
5479 <!-- _______________________________________________________________________ -->
5481 <a name="i_fptrunc">'<tt>fptrunc .. to</tt>' Instruction</a>
5488 <result> = fptrunc <ty> <value> to <ty2> <i>; yields ty2</i>
5492 <p>The '<tt>fptrunc</tt>' instruction truncates <tt>value</tt> to type
5496 <p>The '<tt>fptrunc</tt>' instruction takes a <a href="#t_floating">floating
5497 point</a> value to cast and a <a href="#t_floating">floating point</a> type
5498 to cast it to. The size of <tt>value</tt> must be larger than the size of
5499 <tt>ty2</tt>. This implies that <tt>fptrunc</tt> cannot be used to make a
5500 <i>no-op cast</i>.</p>
5503 <p>The '<tt>fptrunc</tt>' instruction truncates a <tt>value</tt> from a larger
5504 <a href="#t_floating">floating point</a> type to a smaller
5505 <a href="#t_floating">floating point</a> type. If the value cannot fit
5506 within the destination type, <tt>ty2</tt>, then the results are
5511 %X = fptrunc double 123.0 to float <i>; yields float:123.0</i>
5512 %Y = fptrunc double 1.0E+300 to float <i>; yields undefined</i>
5517 <!-- _______________________________________________________________________ -->
5519 <a name="i_fpext">'<tt>fpext .. to</tt>' Instruction</a>
5526 <result> = fpext <ty> <value> to <ty2> <i>; yields ty2</i>
5530 <p>The '<tt>fpext</tt>' extends a floating point <tt>value</tt> to a larger
5531 floating point value.</p>
5534 <p>The '<tt>fpext</tt>' instruction takes a
5535 <a href="#t_floating">floating point</a> <tt>value</tt> to cast, and
5536 a <a href="#t_floating">floating point</a> type to cast it to. The source
5537 type must be smaller than the destination type.</p>
5540 <p>The '<tt>fpext</tt>' instruction extends the <tt>value</tt> from a smaller
5541 <a href="#t_floating">floating point</a> type to a larger
5542 <a href="#t_floating">floating point</a> type. The <tt>fpext</tt> cannot be
5543 used to make a <i>no-op cast</i> because it always changes bits. Use
5544 <tt>bitcast</tt> to make a <i>no-op cast</i> for a floating point cast.</p>
5548 %X = fpext float 3.125 to double <i>; yields double:3.125000e+00</i>
5549 %Y = fpext double %X to fp128 <i>; yields fp128:0xL00000000000000004000900000000000</i>
5554 <!-- _______________________________________________________________________ -->
5556 <a name="i_fptoui">'<tt>fptoui .. to</tt>' Instruction</a>
5563 <result> = fptoui <ty> <value> to <ty2> <i>; yields ty2</i>
5567 <p>The '<tt>fptoui</tt>' converts a floating point <tt>value</tt> to its
5568 unsigned integer equivalent of type <tt>ty2</tt>.</p>
5571 <p>The '<tt>fptoui</tt>' instruction takes a value to cast, which must be a
5572 scalar or vector <a href="#t_floating">floating point</a> value, and a type
5573 to cast it to <tt>ty2</tt>, which must be an <a href="#t_integer">integer</a>
5574 type. If <tt>ty</tt> is a vector floating point type, <tt>ty2</tt> must be a
5575 vector integer type with the same number of elements as <tt>ty</tt></p>
5578 <p>The '<tt>fptoui</tt>' instruction converts its
5579 <a href="#t_floating">floating point</a> operand into the nearest (rounding
5580 towards zero) unsigned integer value. If the value cannot fit
5581 in <tt>ty2</tt>, the results are undefined.</p>
5585 %X = fptoui double 123.0 to i32 <i>; yields i32:123</i>
5586 %Y = fptoui float 1.0E+300 to i1 <i>; yields undefined:1</i>
5587 %Z = fptoui float 1.04E+17 to i8 <i>; yields undefined:1</i>
5592 <!-- _______________________________________________________________________ -->
5594 <a name="i_fptosi">'<tt>fptosi .. to</tt>' Instruction</a>
5601 <result> = fptosi <ty> <value> to <ty2> <i>; yields ty2</i>
5605 <p>The '<tt>fptosi</tt>' instruction converts
5606 <a href="#t_floating">floating point</a> <tt>value</tt> to
5607 type <tt>ty2</tt>.</p>
5610 <p>The '<tt>fptosi</tt>' instruction takes a value to cast, which must be a
5611 scalar or vector <a href="#t_floating">floating point</a> value, and a type
5612 to cast it to <tt>ty2</tt>, which must be an <a href="#t_integer">integer</a>
5613 type. If <tt>ty</tt> is a vector floating point type, <tt>ty2</tt> must be a
5614 vector integer type with the same number of elements as <tt>ty</tt></p>
5617 <p>The '<tt>fptosi</tt>' instruction converts its
5618 <a href="#t_floating">floating point</a> operand into the nearest (rounding
5619 towards zero) signed integer value. If the value cannot fit in <tt>ty2</tt>,
5620 the results are undefined.</p>
5624 %X = fptosi double -123.0 to i32 <i>; yields i32:-123</i>
5625 %Y = fptosi float 1.0E-247 to i1 <i>; yields undefined:1</i>
5626 %Z = fptosi float 1.04E+17 to i8 <i>; yields undefined:1</i>
5631 <!-- _______________________________________________________________________ -->
5633 <a name="i_uitofp">'<tt>uitofp .. to</tt>' Instruction</a>
5640 <result> = uitofp <ty> <value> to <ty2> <i>; yields ty2</i>
5644 <p>The '<tt>uitofp</tt>' instruction regards <tt>value</tt> as an unsigned
5645 integer and converts that value to the <tt>ty2</tt> type.</p>
5648 <p>The '<tt>uitofp</tt>' instruction takes a value to cast, which must be a
5649 scalar or vector <a href="#t_integer">integer</a> value, and a type to cast
5650 it to <tt>ty2</tt>, which must be an <a href="#t_floating">floating point</a>
5651 type. If <tt>ty</tt> is a vector integer type, <tt>ty2</tt> must be a vector
5652 floating point type with the same number of elements as <tt>ty</tt></p>
5655 <p>The '<tt>uitofp</tt>' instruction interprets its operand as an unsigned
5656 integer quantity and converts it to the corresponding floating point
5657 value. If the value cannot fit in the floating point value, the results are
5662 %X = uitofp i32 257 to float <i>; yields float:257.0</i>
5663 %Y = uitofp i8 -1 to double <i>; yields double:255.0</i>
5668 <!-- _______________________________________________________________________ -->
5670 <a name="i_sitofp">'<tt>sitofp .. to</tt>' Instruction</a>
5677 <result> = sitofp <ty> <value> to <ty2> <i>; yields ty2</i>
5681 <p>The '<tt>sitofp</tt>' instruction regards <tt>value</tt> as a signed integer
5682 and converts that value to the <tt>ty2</tt> type.</p>
5685 <p>The '<tt>sitofp</tt>' instruction takes a value to cast, which must be a
5686 scalar or vector <a href="#t_integer">integer</a> value, and a type to cast
5687 it to <tt>ty2</tt>, which must be an <a href="#t_floating">floating point</a>
5688 type. If <tt>ty</tt> is a vector integer type, <tt>ty2</tt> must be a vector
5689 floating point type with the same number of elements as <tt>ty</tt></p>
5692 <p>The '<tt>sitofp</tt>' instruction interprets its operand as a signed integer
5693 quantity and converts it to the corresponding floating point value. If the
5694 value cannot fit in the floating point value, the results are undefined.</p>
5698 %X = sitofp i32 257 to float <i>; yields float:257.0</i>
5699 %Y = sitofp i8 -1 to double <i>; yields double:-1.0</i>
5704 <!-- _______________________________________________________________________ -->
5706 <a name="i_ptrtoint">'<tt>ptrtoint .. to</tt>' Instruction</a>
5713 <result> = ptrtoint <ty> <value> to <ty2> <i>; yields ty2</i>
5717 <p>The '<tt>ptrtoint</tt>' instruction converts the pointer or a vector of
5718 pointers <tt>value</tt> to
5719 the integer (or vector of integers) type <tt>ty2</tt>.</p>
5722 <p>The '<tt>ptrtoint</tt>' instruction takes a <tt>value</tt> to cast, which
5723 must be a a value of type <a href="#t_pointer">pointer</a> or a vector of
5724 pointers, and a type to cast it to
5725 <tt>ty2</tt>, which must be an <a href="#t_integer">integer</a> or a vector
5726 of integers type.</p>
5729 <p>The '<tt>ptrtoint</tt>' instruction converts <tt>value</tt> to integer type
5730 <tt>ty2</tt> by interpreting the pointer value as an integer and either
5731 truncating or zero extending that value to the size of the integer type. If
5732 <tt>value</tt> is smaller than <tt>ty2</tt> then a zero extension is done. If
5733 <tt>value</tt> is larger than <tt>ty2</tt> then a truncation is done. If they
5734 are the same size, then nothing is done (<i>no-op cast</i>) other than a type
5739 %X = ptrtoint i32* %P to i8 <i>; yields truncation on 32-bit architecture</i>
5740 %Y = ptrtoint i32* %P to i64 <i>; yields zero extension on 32-bit architecture</i>
5741 %Z = ptrtoint <4 x i32*> %P to <4 x i64><i>; yields vector zero extension for a vector of addresses on 32-bit architecture</i>
5746 <!-- _______________________________________________________________________ -->
5748 <a name="i_inttoptr">'<tt>inttoptr .. to</tt>' Instruction</a>
5755 <result> = inttoptr <ty> <value> to <ty2> <i>; yields ty2</i>
5759 <p>The '<tt>inttoptr</tt>' instruction converts an integer <tt>value</tt> to a
5760 pointer type, <tt>ty2</tt>.</p>
5763 <p>The '<tt>inttoptr</tt>' instruction takes an <a href="#t_integer">integer</a>
5764 value to cast, and a type to cast it to, which must be a
5765 <a href="#t_pointer">pointer</a> type.</p>
5768 <p>The '<tt>inttoptr</tt>' instruction converts <tt>value</tt> to type
5769 <tt>ty2</tt> by applying either a zero extension or a truncation depending on
5770 the size of the integer <tt>value</tt>. If <tt>value</tt> is larger than the
5771 size of a pointer then a truncation is done. If <tt>value</tt> is smaller
5772 than the size of a pointer then a zero extension is done. If they are the
5773 same size, nothing is done (<i>no-op cast</i>).</p>
5777 %X = inttoptr i32 255 to i32* <i>; yields zero extension on 64-bit architecture</i>
5778 %Y = inttoptr i32 255 to i32* <i>; yields no-op on 32-bit architecture</i>
5779 %Z = inttoptr i64 0 to i32* <i>; yields truncation on 32-bit architecture</i>
5780 %Z = inttoptr <4 x i32> %G to <4 x i8*><i>; yields truncation of vector G to four pointers</i>
5785 <!-- _______________________________________________________________________ -->
5787 <a name="i_bitcast">'<tt>bitcast .. to</tt>' Instruction</a>
5794 <result> = bitcast <ty> <value> to <ty2> <i>; yields ty2</i>
5798 <p>The '<tt>bitcast</tt>' instruction converts <tt>value</tt> to type
5799 <tt>ty2</tt> without changing any bits.</p>
5802 <p>The '<tt>bitcast</tt>' instruction takes a value to cast, which must be a
5803 non-aggregate first class value, and a type to cast it to, which must also be
5804 a non-aggregate <a href="#t_firstclass">first class</a> type. The bit sizes
5805 of <tt>value</tt> and the destination type, <tt>ty2</tt>, must be
5806 identical. If the source type is a pointer, the destination type must also be
5807 a pointer. This instruction supports bitwise conversion of vectors to
5808 integers and to vectors of other types (as long as they have the same
5812 <p>The '<tt>bitcast</tt>' instruction converts <tt>value</tt> to type
5813 <tt>ty2</tt>. It is always a <i>no-op cast</i> because no bits change with
5814 this conversion. The conversion is done as if the <tt>value</tt> had been
5815 stored to memory and read back as type <tt>ty2</tt>.
5816 Pointer (or vector of pointers) types may only be converted to other pointer
5817 (or vector of pointers) types with this instruction. To convert
5818 pointers to other types, use the <a href="#i_inttoptr">inttoptr</a> or
5819 <a href="#i_ptrtoint">ptrtoint</a> instructions first.</p>
5823 %X = bitcast i8 255 to i8 <i>; yields i8 :-1</i>
5824 %Y = bitcast i32* %x to sint* <i>; yields sint*:%x</i>
5825 %Z = bitcast <2 x int> %V to i64; <i>; yields i64: %V</i>
5826 %Z = bitcast <2 x i32*> %V to <2 x i64*> <i>; yields <2 x i64*></i>
5833 <!-- ======================================================================= -->
5835 <a name="otherops">Other Operations</a>
5840 <p>The instructions in this category are the "miscellaneous" instructions, which
5841 defy better classification.</p>
5843 <!-- _______________________________________________________________________ -->
5845 <a name="i_icmp">'<tt>icmp</tt>' Instruction</a>
5852 <result> = icmp <cond> <ty> <op1>, <op2> <i>; yields {i1} or {<N x i1>}:result</i>
5856 <p>The '<tt>icmp</tt>' instruction returns a boolean value or a vector of
5857 boolean values based on comparison of its two integer, integer vector,
5858 pointer, or pointer vector operands.</p>
5861 <p>The '<tt>icmp</tt>' instruction takes three operands. The first operand is
5862 the condition code indicating the kind of comparison to perform. It is not a
5863 value, just a keyword. The possible condition code are:</p>
5866 <li><tt>eq</tt>: equal</li>
5867 <li><tt>ne</tt>: not equal </li>
5868 <li><tt>ugt</tt>: unsigned greater than</li>
5869 <li><tt>uge</tt>: unsigned greater or equal</li>
5870 <li><tt>ult</tt>: unsigned less than</li>
5871 <li><tt>ule</tt>: unsigned less or equal</li>
5872 <li><tt>sgt</tt>: signed greater than</li>
5873 <li><tt>sge</tt>: signed greater or equal</li>
5874 <li><tt>slt</tt>: signed less than</li>
5875 <li><tt>sle</tt>: signed less or equal</li>
5878 <p>The remaining two arguments must be <a href="#t_integer">integer</a> or
5879 <a href="#t_pointer">pointer</a> or integer <a href="#t_vector">vector</a>
5880 typed. They must also be identical types.</p>
5883 <p>The '<tt>icmp</tt>' compares <tt>op1</tt> and <tt>op2</tt> according to the
5884 condition code given as <tt>cond</tt>. The comparison performed always yields
5885 either an <a href="#t_integer"><tt>i1</tt></a> or vector of <tt>i1</tt>
5886 result, as follows:</p>
5889 <li><tt>eq</tt>: yields <tt>true</tt> if the operands are equal,
5890 <tt>false</tt> otherwise. No sign interpretation is necessary or
5893 <li><tt>ne</tt>: yields <tt>true</tt> if the operands are unequal,
5894 <tt>false</tt> otherwise. No sign interpretation is necessary or
5897 <li><tt>ugt</tt>: interprets the operands as unsigned values and yields
5898 <tt>true</tt> if <tt>op1</tt> is greater than <tt>op2</tt>.</li>
5900 <li><tt>uge</tt>: interprets the operands as unsigned values and yields
5901 <tt>true</tt> if <tt>op1</tt> is greater than or equal
5902 to <tt>op2</tt>.</li>
5904 <li><tt>ult</tt>: interprets the operands as unsigned values and yields
5905 <tt>true</tt> if <tt>op1</tt> is less than <tt>op2</tt>.</li>
5907 <li><tt>ule</tt>: interprets the operands as unsigned values and yields
5908 <tt>true</tt> if <tt>op1</tt> is less than or equal to <tt>op2</tt>.</li>
5910 <li><tt>sgt</tt>: interprets the operands as signed values and yields
5911 <tt>true</tt> if <tt>op1</tt> is greater than <tt>op2</tt>.</li>
5913 <li><tt>sge</tt>: interprets the operands as signed values and yields
5914 <tt>true</tt> if <tt>op1</tt> is greater than or equal
5915 to <tt>op2</tt>.</li>
5917 <li><tt>slt</tt>: interprets the operands as signed values and yields
5918 <tt>true</tt> if <tt>op1</tt> is less than <tt>op2</tt>.</li>
5920 <li><tt>sle</tt>: interprets the operands as signed values and yields
5921 <tt>true</tt> if <tt>op1</tt> is less than or equal to <tt>op2</tt>.</li>
5924 <p>If the operands are <a href="#t_pointer">pointer</a> typed, the pointer
5925 values are compared as if they were integers.</p>
5927 <p>If the operands are integer vectors, then they are compared element by
5928 element. The result is an <tt>i1</tt> vector with the same number of elements
5929 as the values being compared. Otherwise, the result is an <tt>i1</tt>.</p>
5933 <result> = icmp eq i32 4, 5 <i>; yields: result=false</i>
5934 <result> = icmp ne float* %X, %X <i>; yields: result=false</i>
5935 <result> = icmp ult i16 4, 5 <i>; yields: result=true</i>
5936 <result> = icmp sgt i16 4, 5 <i>; yields: result=false</i>
5937 <result> = icmp ule i16 -4, 5 <i>; yields: result=false</i>
5938 <result> = icmp sge i16 4, 5 <i>; yields: result=false</i>
5941 <p>Note that the code generator does not yet support vector types with
5942 the <tt>icmp</tt> instruction.</p>
5946 <!-- _______________________________________________________________________ -->
5948 <a name="i_fcmp">'<tt>fcmp</tt>' Instruction</a>
5955 <result> = fcmp <cond> <ty> <op1>, <op2> <i>; yields {i1} or {<N x i1>}:result</i>
5959 <p>The '<tt>fcmp</tt>' instruction returns a boolean value or vector of boolean
5960 values based on comparison of its operands.</p>
5962 <p>If the operands are floating point scalars, then the result type is a boolean
5963 (<a href="#t_integer"><tt>i1</tt></a>).</p>
5965 <p>If the operands are floating point vectors, then the result type is a vector
5966 of boolean with the same number of elements as the operands being
5970 <p>The '<tt>fcmp</tt>' instruction takes three operands. The first operand is
5971 the condition code indicating the kind of comparison to perform. It is not a
5972 value, just a keyword. The possible condition code are:</p>
5975 <li><tt>false</tt>: no comparison, always returns false</li>
5976 <li><tt>oeq</tt>: ordered and equal</li>
5977 <li><tt>ogt</tt>: ordered and greater than </li>
5978 <li><tt>oge</tt>: ordered and greater than or equal</li>
5979 <li><tt>olt</tt>: ordered and less than </li>
5980 <li><tt>ole</tt>: ordered and less than or equal</li>
5981 <li><tt>one</tt>: ordered and not equal</li>
5982 <li><tt>ord</tt>: ordered (no nans)</li>
5983 <li><tt>ueq</tt>: unordered or equal</li>
5984 <li><tt>ugt</tt>: unordered or greater than </li>
5985 <li><tt>uge</tt>: unordered or greater than or equal</li>
5986 <li><tt>ult</tt>: unordered or less than </li>
5987 <li><tt>ule</tt>: unordered or less than or equal</li>
5988 <li><tt>une</tt>: unordered or not equal</li>
5989 <li><tt>uno</tt>: unordered (either nans)</li>
5990 <li><tt>true</tt>: no comparison, always returns true</li>
5993 <p><i>Ordered</i> means that neither operand is a QNAN while
5994 <i>unordered</i> means that either operand may be a QNAN.</p>
5996 <p>Each of <tt>val1</tt> and <tt>val2</tt> arguments must be either
5997 a <a href="#t_floating">floating point</a> type or
5998 a <a href="#t_vector">vector</a> of floating point type. They must have
5999 identical types.</p>
6002 <p>The '<tt>fcmp</tt>' instruction compares <tt>op1</tt> and <tt>op2</tt>
6003 according to the condition code given as <tt>cond</tt>. If the operands are
6004 vectors, then the vectors are compared element by element. Each comparison
6005 performed always yields an <a href="#t_integer">i1</a> result, as
6009 <li><tt>false</tt>: always yields <tt>false</tt>, regardless of operands.</li>
6011 <li><tt>oeq</tt>: yields <tt>true</tt> if both operands are not a QNAN and
6012 <tt>op1</tt> is equal to <tt>op2</tt>.</li>
6014 <li><tt>ogt</tt>: yields <tt>true</tt> if both operands are not a QNAN and
6015 <tt>op1</tt> is greater than <tt>op2</tt>.</li>
6017 <li><tt>oge</tt>: yields <tt>true</tt> if both operands are not a QNAN and
6018 <tt>op1</tt> is greater than or equal to <tt>op2</tt>.</li>
6020 <li><tt>olt</tt>: yields <tt>true</tt> if both operands are not a QNAN and
6021 <tt>op1</tt> is less than <tt>op2</tt>.</li>
6023 <li><tt>ole</tt>: yields <tt>true</tt> if both operands are not a QNAN and
6024 <tt>op1</tt> is less than or equal to <tt>op2</tt>.</li>
6026 <li><tt>one</tt>: yields <tt>true</tt> if both operands are not a QNAN and
6027 <tt>op1</tt> is not equal to <tt>op2</tt>.</li>
6029 <li><tt>ord</tt>: yields <tt>true</tt> if both operands are not a QNAN.</li>
6031 <li><tt>ueq</tt>: yields <tt>true</tt> if either operand is a QNAN or
6032 <tt>op1</tt> is equal to <tt>op2</tt>.</li>
6034 <li><tt>ugt</tt>: yields <tt>true</tt> if either operand is a QNAN or
6035 <tt>op1</tt> is greater than <tt>op2</tt>.</li>
6037 <li><tt>uge</tt>: yields <tt>true</tt> if either operand is a QNAN or
6038 <tt>op1</tt> is greater than or equal to <tt>op2</tt>.</li>
6040 <li><tt>ult</tt>: yields <tt>true</tt> if either operand is a QNAN or
6041 <tt>op1</tt> is less than <tt>op2</tt>.</li>
6043 <li><tt>ule</tt>: yields <tt>true</tt> if either operand is a QNAN or
6044 <tt>op1</tt> is less than or equal to <tt>op2</tt>.</li>
6046 <li><tt>une</tt>: yields <tt>true</tt> if either operand is a QNAN or
6047 <tt>op1</tt> is not equal to <tt>op2</tt>.</li>
6049 <li><tt>uno</tt>: yields <tt>true</tt> if either operand is a QNAN.</li>
6051 <li><tt>true</tt>: always yields <tt>true</tt>, regardless of operands.</li>
6056 <result> = fcmp oeq float 4.0, 5.0 <i>; yields: result=false</i>
6057 <result> = fcmp one float 4.0, 5.0 <i>; yields: result=true</i>
6058 <result> = fcmp olt float 4.0, 5.0 <i>; yields: result=true</i>
6059 <result> = fcmp ueq double 1.0, 2.0 <i>; yields: result=false</i>
6062 <p>Note that the code generator does not yet support vector types with
6063 the <tt>fcmp</tt> instruction.</p>
6067 <!-- _______________________________________________________________________ -->
6069 <a name="i_phi">'<tt>phi</tt>' Instruction</a>
6076 <result> = phi <ty> [ <val0>, <label0>], ...
6080 <p>The '<tt>phi</tt>' instruction is used to implement the φ node in the
6081 SSA graph representing the function.</p>
6084 <p>The type of the incoming values is specified with the first type field. After
6085 this, the '<tt>phi</tt>' instruction takes a list of pairs as arguments, with
6086 one pair for each predecessor basic block of the current block. Only values
6087 of <a href="#t_firstclass">first class</a> type may be used as the value
6088 arguments to the PHI node. Only labels may be used as the label
6091 <p>There must be no non-phi instructions between the start of a basic block and
6092 the PHI instructions: i.e. PHI instructions must be first in a basic
6095 <p>For the purposes of the SSA form, the use of each incoming value is deemed to
6096 occur on the edge from the corresponding predecessor block to the current
6097 block (but after any definition of an '<tt>invoke</tt>' instruction's return
6098 value on the same edge).</p>
6101 <p>At runtime, the '<tt>phi</tt>' instruction logically takes on the value
6102 specified by the pair corresponding to the predecessor basic block that
6103 executed just prior to the current block.</p>
6107 Loop: ; Infinite loop that counts from 0 on up...
6108 %indvar = phi i32 [ 0, %LoopHeader ], [ %nextindvar, %Loop ]
6109 %nextindvar = add i32 %indvar, 1
6115 <!-- _______________________________________________________________________ -->
6117 <a name="i_select">'<tt>select</tt>' Instruction</a>
6124 <result> = select <i>selty</i> <cond>, <ty> <val1>, <ty> <val2> <i>; yields ty</i>
6126 <i>selty</i> is either i1 or {<N x i1>}
6130 <p>The '<tt>select</tt>' instruction is used to choose one value based on a
6131 condition, without branching.</p>
6135 <p>The '<tt>select</tt>' instruction requires an 'i1' value or a vector of 'i1'
6136 values indicating the condition, and two values of the
6137 same <a href="#t_firstclass">first class</a> type. If the val1/val2 are
6138 vectors and the condition is a scalar, then entire vectors are selected, not
6139 individual elements.</p>
6142 <p>If the condition is an i1 and it evaluates to 1, the instruction returns the
6143 first value argument; otherwise, it returns the second value argument.</p>
6145 <p>If the condition is a vector of i1, then the value arguments must be vectors
6146 of the same size, and the selection is done element by element.</p>
6150 %X = select i1 true, i8 17, i8 42 <i>; yields i8:17</i>
6155 <!-- _______________________________________________________________________ -->
6157 <a name="i_call">'<tt>call</tt>' Instruction</a>
6164 <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>]
6168 <p>The '<tt>call</tt>' instruction represents a simple function call.</p>
6171 <p>This instruction requires several arguments:</p>
6174 <li>The optional "tail" marker indicates that the callee function does not
6175 access any allocas or varargs in the caller. Note that calls may be
6176 marked "tail" even if they do not occur before
6177 a <a href="#i_ret"><tt>ret</tt></a> instruction. If the "tail" marker is
6178 present, the function call is eligible for tail call optimization,
6179 but <a href="CodeGenerator.html#tailcallopt">might not in fact be
6180 optimized into a jump</a>. The code generator may optimize calls marked
6181 "tail" with either 1) automatic <a href="CodeGenerator.html#sibcallopt">
6182 sibling call optimization</a> when the caller and callee have
6183 matching signatures, or 2) forced tail call optimization when the
6184 following extra requirements are met:
6186 <li>Caller and callee both have the calling
6187 convention <tt>fastcc</tt>.</li>
6188 <li>The call is in tail position (ret immediately follows call and ret
6189 uses value of call or is void).</li>
6190 <li>Option <tt>-tailcallopt</tt> is enabled,
6191 or <code>llvm::GuaranteedTailCallOpt</code> is <code>true</code>.</li>
6192 <li><a href="CodeGenerator.html#tailcallopt">Platform specific
6193 constraints are met.</a></li>
6197 <li>The optional "cconv" marker indicates which <a href="#callingconv">calling
6198 convention</a> the call should use. If none is specified, the call
6199 defaults to using C calling conventions. The calling convention of the
6200 call must match the calling convention of the target function, or else the
6201 behavior is undefined.</li>
6203 <li>The optional <a href="#paramattrs">Parameter Attributes</a> list for
6204 return values. Only '<tt>zeroext</tt>', '<tt>signext</tt>', and
6205 '<tt>inreg</tt>' attributes are valid here.</li>
6207 <li>'<tt>ty</tt>': the type of the call instruction itself which is also the
6208 type of the return value. Functions that return no value are marked
6209 <tt><a href="#t_void">void</a></tt>.</li>
6211 <li>'<tt>fnty</tt>': shall be the signature of the pointer to function value
6212 being invoked. The argument types must match the types implied by this
6213 signature. This type can be omitted if the function is not varargs and if
6214 the function type does not return a pointer to a function.</li>
6216 <li>'<tt>fnptrval</tt>': An LLVM value containing a pointer to a function to
6217 be invoked. In most cases, this is a direct function invocation, but
6218 indirect <tt>call</tt>s are just as possible, calling an arbitrary pointer
6219 to function value.</li>
6221 <li>'<tt>function args</tt>': argument list whose types match the function
6222 signature argument types and parameter attributes. All arguments must be
6223 of <a href="#t_firstclass">first class</a> type. If the function
6224 signature indicates the function accepts a variable number of arguments,
6225 the extra arguments can be specified.</li>
6227 <li>The optional <a href="#fnattrs">function attributes</a> list. Only
6228 '<tt>noreturn</tt>', '<tt>nounwind</tt>', '<tt>readonly</tt>' and
6229 '<tt>readnone</tt>' attributes are valid here.</li>
6233 <p>The '<tt>call</tt>' instruction is used to cause control flow to transfer to
6234 a specified function, with its incoming arguments bound to the specified
6235 values. Upon a '<tt><a href="#i_ret">ret</a></tt>' instruction in the called
6236 function, control flow continues with the instruction after the function
6237 call, and the return value of the function is bound to the result
6242 %retval = call i32 @test(i32 %argc)
6243 call i32 (i8*, ...)* @printf(i8* %msg, i32 12, i8 42) <i>; yields i32</i>
6244 %X = tail call i32 @foo() <i>; yields i32</i>
6245 %Y = tail call <a href="#callingconv">fastcc</a> i32 @foo() <i>; yields i32</i>
6246 call void %foo(i8 97 signext)
6248 %struct.A = type { i32, i8 }
6249 %r = call %struct.A @foo() <i>; yields { 32, i8 }</i>
6250 %gr = extractvalue %struct.A %r, 0 <i>; yields i32</i>
6251 %gr1 = extractvalue %struct.A %r, 1 <i>; yields i8</i>
6252 %Z = call void @foo() noreturn <i>; indicates that %foo never returns normally</i>
6253 %ZZ = call zeroext i32 @bar() <i>; Return value is %zero extended</i>
6256 <p>llvm treats calls to some functions with names and arguments that match the
6257 standard C99 library as being the C99 library functions, and may perform
6258 optimizations or generate code for them under that assumption. This is
6259 something we'd like to change in the future to provide better support for
6260 freestanding environments and non-C-based languages.</p>
6264 <!-- _______________________________________________________________________ -->
6266 <a name="i_va_arg">'<tt>va_arg</tt>' Instruction</a>
6273 <resultval> = va_arg <va_list*> <arglist>, <argty>
6277 <p>The '<tt>va_arg</tt>' instruction is used to access arguments passed through
6278 the "variable argument" area of a function call. It is used to implement the
6279 <tt>va_arg</tt> macro in C.</p>
6282 <p>This instruction takes a <tt>va_list*</tt> value and the type of the
6283 argument. It returns a value of the specified argument type and increments
6284 the <tt>va_list</tt> to point to the next argument. The actual type
6285 of <tt>va_list</tt> is target specific.</p>
6288 <p>The '<tt>va_arg</tt>' instruction loads an argument of the specified type
6289 from the specified <tt>va_list</tt> and causes the <tt>va_list</tt> to point
6290 to the next argument. For more information, see the variable argument
6291 handling <a href="#int_varargs">Intrinsic Functions</a>.</p>
6293 <p>It is legal for this instruction to be called in a function which does not
6294 take a variable number of arguments, for example, the <tt>vfprintf</tt>
6297 <p><tt>va_arg</tt> is an LLVM instruction instead of
6298 an <a href="#intrinsics">intrinsic function</a> because it takes a type as an
6302 <p>See the <a href="#int_varargs">variable argument processing</a> section.</p>
6304 <p>Note that the code generator does not yet fully support va_arg on many
6305 targets. Also, it does not currently support va_arg with aggregate types on
6310 <!-- _______________________________________________________________________ -->
6312 <a name="i_landingpad">'<tt>landingpad</tt>' Instruction</a>
6319 <resultval> = landingpad <somety> personality <type> <pers_fn> <clause>+
6320 <resultval> = landingpad <somety> personality <type> <pers_fn> cleanup <clause>*
6322 <clause> := catch <type> <value>
6323 <clause> := filter <array constant type> <array constant>
6327 <p>The '<tt>landingpad</tt>' instruction is used by
6328 <a href="ExceptionHandling.html#overview">LLVM's exception handling
6329 system</a> to specify that a basic block is a landing pad — one where
6330 the exception lands, and corresponds to the code found in the
6331 <i><tt>catch</tt></i> portion of a <i><tt>try/catch</tt></i> sequence. It
6332 defines values supplied by the personality function (<tt>pers_fn</tt>) upon
6333 re-entry to the function. The <tt>resultval</tt> has the
6334 type <tt>somety</tt>.</p>
6337 <p>This instruction takes a <tt>pers_fn</tt> value. This is the personality
6338 function associated with the unwinding mechanism. The optional
6339 <tt>cleanup</tt> flag indicates that the landing pad block is a cleanup.</p>
6341 <p>A <tt>clause</tt> begins with the clause type — <tt>catch</tt>
6342 or <tt>filter</tt> — and contains the global variable representing the
6343 "type" that may be caught or filtered respectively. Unlike the
6344 <tt>catch</tt> clause, the <tt>filter</tt> clause takes an array constant as
6345 its argument. Use "<tt>[0 x i8**] undef</tt>" for a filter which cannot
6346 throw. The '<tt>landingpad</tt>' instruction must contain <em>at least</em>
6347 one <tt>clause</tt> or the <tt>cleanup</tt> flag.</p>
6350 <p>The '<tt>landingpad</tt>' instruction defines the values which are set by the
6351 personality function (<tt>pers_fn</tt>) upon re-entry to the function, and
6352 therefore the "result type" of the <tt>landingpad</tt> instruction. As with
6353 calling conventions, how the personality function results are represented in
6354 LLVM IR is target specific.</p>
6356 <p>The clauses are applied in order from top to bottom. If two
6357 <tt>landingpad</tt> instructions are merged together through inlining, the
6358 clauses from the calling function are appended to the list of clauses.</p>
6360 <p>The <tt>landingpad</tt> instruction has several restrictions:</p>
6363 <li>A landing pad block is a basic block which is the unwind destination of an
6364 '<tt>invoke</tt>' instruction.</li>
6365 <li>A landing pad block must have a '<tt>landingpad</tt>' instruction as its
6366 first non-PHI instruction.</li>
6367 <li>There can be only one '<tt>landingpad</tt>' instruction within the landing
6369 <li>A basic block that is not a landing pad block may not include a
6370 '<tt>landingpad</tt>' instruction.</li>
6371 <li>All '<tt>landingpad</tt>' instructions in a function must have the same
6372 personality function.</li>
6377 ;; A landing pad which can catch an integer.
6378 %res = landingpad { i8*, i32 } personality i32 (...)* @__gxx_personality_v0
6380 ;; A landing pad that is a cleanup.
6381 %res = landingpad { i8*, i32 } personality i32 (...)* @__gxx_personality_v0
6383 ;; A landing pad which can catch an integer and can only throw a double.
6384 %res = landingpad { i8*, i32 } personality i32 (...)* @__gxx_personality_v0
6386 filter [1 x i8**] [@_ZTId]
6395 <!-- *********************************************************************** -->
6396 <h2><a name="intrinsics">Intrinsic Functions</a></h2>
6397 <!-- *********************************************************************** -->
6401 <p>LLVM supports the notion of an "intrinsic function". These functions have
6402 well known names and semantics and are required to follow certain
6403 restrictions. Overall, these intrinsics represent an extension mechanism for
6404 the LLVM language that does not require changing all of the transformations
6405 in LLVM when adding to the language (or the bitcode reader/writer, the
6406 parser, etc...).</p>
6408 <p>Intrinsic function names must all start with an "<tt>llvm.</tt>" prefix. This
6409 prefix is reserved in LLVM for intrinsic names; thus, function names may not
6410 begin with this prefix. Intrinsic functions must always be external
6411 functions: you cannot define the body of intrinsic functions. Intrinsic
6412 functions may only be used in call or invoke instructions: it is illegal to
6413 take the address of an intrinsic function. Additionally, because intrinsic
6414 functions are part of the LLVM language, it is required if any are added that
6415 they be documented here.</p>
6417 <p>Some intrinsic functions can be overloaded, i.e., the intrinsic represents a
6418 family of functions that perform the same operation but on different data
6419 types. Because LLVM can represent over 8 million different integer types,
6420 overloading is used commonly to allow an intrinsic function to operate on any
6421 integer type. One or more of the argument types or the result type can be
6422 overloaded to accept any integer type. Argument types may also be defined as
6423 exactly matching a previous argument's type or the result type. This allows
6424 an intrinsic function which accepts multiple arguments, but needs all of them
6425 to be of the same type, to only be overloaded with respect to a single
6426 argument or the result.</p>
6428 <p>Overloaded intrinsics will have the names of its overloaded argument types
6429 encoded into its function name, each preceded by a period. Only those types
6430 which are overloaded result in a name suffix. Arguments whose type is matched
6431 against another type do not. For example, the <tt>llvm.ctpop</tt> function
6432 can take an integer of any width and returns an integer of exactly the same
6433 integer width. This leads to a family of functions such as
6434 <tt>i8 @llvm.ctpop.i8(i8 %val)</tt> and <tt>i29 @llvm.ctpop.i29(i29
6435 %val)</tt>. Only one type, the return type, is overloaded, and only one type
6436 suffix is required. Because the argument's type is matched against the return
6437 type, it does not require its own name suffix.</p>
6439 <p>To learn how to add an intrinsic function, please see the
6440 <a href="ExtendingLLVM.html">Extending LLVM Guide</a>.</p>
6442 <!-- ======================================================================= -->
6444 <a name="int_varargs">Variable Argument Handling Intrinsics</a>
6449 <p>Variable argument support is defined in LLVM with
6450 the <a href="#i_va_arg"><tt>va_arg</tt></a> instruction and these three
6451 intrinsic functions. These functions are related to the similarly named
6452 macros defined in the <tt><stdarg.h></tt> header file.</p>
6454 <p>All of these functions operate on arguments that use a target-specific value
6455 type "<tt>va_list</tt>". The LLVM assembly language reference manual does
6456 not define what this type is, so all transformations should be prepared to
6457 handle these functions regardless of the type used.</p>
6459 <p>This example shows how the <a href="#i_va_arg"><tt>va_arg</tt></a>
6460 instruction and the variable argument handling intrinsic functions are
6463 <pre class="doc_code">
6464 define i32 @test(i32 %X, ...) {
6465 ; Initialize variable argument processing
6467 %ap2 = bitcast i8** %ap to i8*
6468 call void @llvm.va_start(i8* %ap2)
6470 ; Read a single integer argument
6471 %tmp = va_arg i8** %ap, i32
6473 ; Demonstrate usage of llvm.va_copy and llvm.va_end
6475 %aq2 = bitcast i8** %aq to i8*
6476 call void @llvm.va_copy(i8* %aq2, i8* %ap2)
6477 call void @llvm.va_end(i8* %aq2)
6479 ; Stop processing of arguments.
6480 call void @llvm.va_end(i8* %ap2)
6484 declare void @llvm.va_start(i8*)
6485 declare void @llvm.va_copy(i8*, i8*)
6486 declare void @llvm.va_end(i8*)
6489 <!-- _______________________________________________________________________ -->
6491 <a name="int_va_start">'<tt>llvm.va_start</tt>' Intrinsic</a>
6499 declare void %llvm.va_start(i8* <arglist>)
6503 <p>The '<tt>llvm.va_start</tt>' intrinsic initializes <tt>*<arglist></tt>
6504 for subsequent use by <tt><a href="#i_va_arg">va_arg</a></tt>.</p>
6507 <p>The argument is a pointer to a <tt>va_list</tt> element to initialize.</p>
6510 <p>The '<tt>llvm.va_start</tt>' intrinsic works just like the <tt>va_start</tt>
6511 macro available in C. In a target-dependent way, it initializes
6512 the <tt>va_list</tt> element to which the argument points, so that the next
6513 call to <tt>va_arg</tt> will produce the first variable argument passed to
6514 the function. Unlike the C <tt>va_start</tt> macro, this intrinsic does not
6515 need to know the last argument of the function as the compiler can figure
6520 <!-- _______________________________________________________________________ -->
6522 <a name="int_va_end">'<tt>llvm.va_end</tt>' Intrinsic</a>
6529 declare void @llvm.va_end(i8* <arglist>)
6533 <p>The '<tt>llvm.va_end</tt>' intrinsic destroys <tt>*<arglist></tt>,
6534 which has been initialized previously
6535 with <tt><a href="#int_va_start">llvm.va_start</a></tt>
6536 or <tt><a href="#i_va_copy">llvm.va_copy</a></tt>.</p>
6539 <p>The argument is a pointer to a <tt>va_list</tt> to destroy.</p>
6542 <p>The '<tt>llvm.va_end</tt>' intrinsic works just like the <tt>va_end</tt>
6543 macro available in C. In a target-dependent way, it destroys
6544 the <tt>va_list</tt> element to which the argument points. Calls
6545 to <a href="#int_va_start"><tt>llvm.va_start</tt></a>
6546 and <a href="#int_va_copy"> <tt>llvm.va_copy</tt></a> must be matched exactly
6547 with calls to <tt>llvm.va_end</tt>.</p>
6551 <!-- _______________________________________________________________________ -->
6553 <a name="int_va_copy">'<tt>llvm.va_copy</tt>' Intrinsic</a>
6560 declare void @llvm.va_copy(i8* <destarglist>, i8* <srcarglist>)
6564 <p>The '<tt>llvm.va_copy</tt>' intrinsic copies the current argument position
6565 from the source argument list to the destination argument list.</p>
6568 <p>The first argument is a pointer to a <tt>va_list</tt> element to initialize.
6569 The second argument is a pointer to a <tt>va_list</tt> element to copy
6573 <p>The '<tt>llvm.va_copy</tt>' intrinsic works just like the <tt>va_copy</tt>
6574 macro available in C. In a target-dependent way, it copies the
6575 source <tt>va_list</tt> element into the destination <tt>va_list</tt>
6576 element. This intrinsic is necessary because
6577 the <tt><a href="#int_va_start"> llvm.va_start</a></tt> intrinsic may be
6578 arbitrarily complex and require, for example, memory allocation.</p>
6584 <!-- ======================================================================= -->
6586 <a name="int_gc">Accurate Garbage Collection Intrinsics</a>
6591 <p>LLVM support for <a href="GarbageCollection.html">Accurate Garbage
6592 Collection</a> (GC) requires the implementation and generation of these
6593 intrinsics. These intrinsics allow identification of <a href="#int_gcroot">GC
6594 roots on the stack</a>, as well as garbage collector implementations that
6595 require <a href="#int_gcread">read</a> and <a href="#int_gcwrite">write</a>
6596 barriers. Front-ends for type-safe garbage collected languages should generate
6597 these intrinsics to make use of the LLVM garbage collectors. For more details,
6598 see <a href="GarbageCollection.html">Accurate Garbage Collection with
6601 <p>The garbage collection intrinsics only operate on objects in the generic
6602 address space (address space zero).</p>
6604 <!-- _______________________________________________________________________ -->
6606 <a name="int_gcroot">'<tt>llvm.gcroot</tt>' Intrinsic</a>
6613 declare void @llvm.gcroot(i8** %ptrloc, i8* %metadata)
6617 <p>The '<tt>llvm.gcroot</tt>' intrinsic declares the existence of a GC root to
6618 the code generator, and allows some metadata to be associated with it.</p>
6621 <p>The first argument specifies the address of a stack object that contains the
6622 root pointer. The second pointer (which must be either a constant or a
6623 global value address) contains the meta-data to be associated with the
6627 <p>At runtime, a call to this intrinsic stores a null pointer into the "ptrloc"
6628 location. At compile-time, the code generator generates information to allow
6629 the runtime to find the pointer at GC safe points. The '<tt>llvm.gcroot</tt>'
6630 intrinsic may only be used in a function which <a href="#gc">specifies a GC
6635 <!-- _______________________________________________________________________ -->
6637 <a name="int_gcread">'<tt>llvm.gcread</tt>' Intrinsic</a>
6644 declare i8* @llvm.gcread(i8* %ObjPtr, i8** %Ptr)
6648 <p>The '<tt>llvm.gcread</tt>' intrinsic identifies reads of references from heap
6649 locations, allowing garbage collector implementations that require read
6653 <p>The second argument is the address to read from, which should be an address
6654 allocated from the garbage collector. The first object is a pointer to the
6655 start of the referenced object, if needed by the language runtime (otherwise
6659 <p>The '<tt>llvm.gcread</tt>' intrinsic has the same semantics as a load
6660 instruction, but may be replaced with substantially more complex code by the
6661 garbage collector runtime, as needed. The '<tt>llvm.gcread</tt>' intrinsic
6662 may only be used in a function which <a href="#gc">specifies a GC
6667 <!-- _______________________________________________________________________ -->
6669 <a name="int_gcwrite">'<tt>llvm.gcwrite</tt>' Intrinsic</a>
6676 declare void @llvm.gcwrite(i8* %P1, i8* %Obj, i8** %P2)
6680 <p>The '<tt>llvm.gcwrite</tt>' intrinsic identifies writes of references to heap
6681 locations, allowing garbage collector implementations that require write
6682 barriers (such as generational or reference counting collectors).</p>
6685 <p>The first argument is the reference to store, the second is the start of the
6686 object to store it to, and the third is the address of the field of Obj to
6687 store to. If the runtime does not require a pointer to the object, Obj may
6691 <p>The '<tt>llvm.gcwrite</tt>' intrinsic has the same semantics as a store
6692 instruction, but may be replaced with substantially more complex code by the
6693 garbage collector runtime, as needed. The '<tt>llvm.gcwrite</tt>' intrinsic
6694 may only be used in a function which <a href="#gc">specifies a GC
6701 <!-- ======================================================================= -->
6703 <a name="int_codegen">Code Generator Intrinsics</a>
6708 <p>These intrinsics are provided by LLVM to expose special features that may
6709 only be implemented with code generator support.</p>
6711 <!-- _______________________________________________________________________ -->
6713 <a name="int_returnaddress">'<tt>llvm.returnaddress</tt>' Intrinsic</a>
6720 declare i8 *@llvm.returnaddress(i32 <level>)
6724 <p>The '<tt>llvm.returnaddress</tt>' intrinsic attempts to compute a
6725 target-specific value indicating the return address of the current function
6726 or one of its callers.</p>
6729 <p>The argument to this intrinsic indicates which function to return the address
6730 for. Zero indicates the calling function, one indicates its caller, etc.
6731 The argument is <b>required</b> to be a constant integer value.</p>
6734 <p>The '<tt>llvm.returnaddress</tt>' intrinsic either returns a pointer
6735 indicating the return address of the specified call frame, or zero if it
6736 cannot be identified. The value returned by this intrinsic is likely to be
6737 incorrect or 0 for arguments other than zero, so it should only be used for
6738 debugging purposes.</p>
6740 <p>Note that calling this intrinsic does not prevent function inlining or other
6741 aggressive transformations, so the value returned may not be that of the
6742 obvious source-language caller.</p>
6746 <!-- _______________________________________________________________________ -->
6748 <a name="int_frameaddress">'<tt>llvm.frameaddress</tt>' Intrinsic</a>
6755 declare i8* @llvm.frameaddress(i32 <level>)
6759 <p>The '<tt>llvm.frameaddress</tt>' intrinsic attempts to return the
6760 target-specific frame pointer value for the specified stack frame.</p>
6763 <p>The argument to this intrinsic indicates which function to return the frame
6764 pointer for. Zero indicates the calling function, one indicates its caller,
6765 etc. The argument is <b>required</b> to be a constant integer value.</p>
6768 <p>The '<tt>llvm.frameaddress</tt>' intrinsic either returns a pointer
6769 indicating the frame address of the specified call frame, or zero if it
6770 cannot be identified. The value returned by this intrinsic is likely to be
6771 incorrect or 0 for arguments other than zero, so it should only be used for
6772 debugging purposes.</p>
6774 <p>Note that calling this intrinsic does not prevent function inlining or other
6775 aggressive transformations, so the value returned may not be that of the
6776 obvious source-language caller.</p>
6780 <!-- _______________________________________________________________________ -->
6782 <a name="int_stacksave">'<tt>llvm.stacksave</tt>' Intrinsic</a>
6789 declare i8* @llvm.stacksave()
6793 <p>The '<tt>llvm.stacksave</tt>' intrinsic is used to remember the current state
6794 of the function stack, for use
6795 with <a href="#int_stackrestore"> <tt>llvm.stackrestore</tt></a>. This is
6796 useful for implementing language features like scoped automatic variable
6797 sized arrays in C99.</p>
6800 <p>This intrinsic returns a opaque pointer value that can be passed
6801 to <a href="#int_stackrestore"><tt>llvm.stackrestore</tt></a>. When
6802 an <tt>llvm.stackrestore</tt> intrinsic is executed with a value saved
6803 from <tt>llvm.stacksave</tt>, it effectively restores the state of the stack
6804 to the state it was in when the <tt>llvm.stacksave</tt> intrinsic executed.
6805 In practice, this pops any <a href="#i_alloca">alloca</a> blocks from the
6806 stack that were allocated after the <tt>llvm.stacksave</tt> was executed.</p>
6810 <!-- _______________________________________________________________________ -->
6812 <a name="int_stackrestore">'<tt>llvm.stackrestore</tt>' Intrinsic</a>
6819 declare void @llvm.stackrestore(i8* %ptr)
6823 <p>The '<tt>llvm.stackrestore</tt>' intrinsic is used to restore the state of
6824 the function stack to the state it was in when the
6825 corresponding <a href="#int_stacksave"><tt>llvm.stacksave</tt></a> intrinsic
6826 executed. This is useful for implementing language features like scoped
6827 automatic variable sized arrays in C99.</p>
6830 <p>See the description
6831 for <a href="#int_stacksave"><tt>llvm.stacksave</tt></a>.</p>
6835 <!-- _______________________________________________________________________ -->
6837 <a name="int_prefetch">'<tt>llvm.prefetch</tt>' Intrinsic</a>
6844 declare void @llvm.prefetch(i8* <address>, i32 <rw>, i32 <locality>, i32 <cache type>)
6848 <p>The '<tt>llvm.prefetch</tt>' intrinsic is a hint to the code generator to
6849 insert a prefetch instruction if supported; otherwise, it is a noop.
6850 Prefetches have no effect on the behavior of the program but can change its
6851 performance characteristics.</p>
6854 <p><tt>address</tt> is the address to be prefetched, <tt>rw</tt> is the
6855 specifier determining if the fetch should be for a read (0) or write (1),
6856 and <tt>locality</tt> is a temporal locality specifier ranging from (0) - no
6857 locality, to (3) - extremely local keep in cache. The <tt>cache type</tt>
6858 specifies whether the prefetch is performed on the data (1) or instruction (0)
6859 cache. The <tt>rw</tt>, <tt>locality</tt> and <tt>cache type</tt> arguments
6860 must be constant integers.</p>
6863 <p>This intrinsic does not modify the behavior of the program. In particular,
6864 prefetches cannot trap and do not produce a value. On targets that support
6865 this intrinsic, the prefetch can provide hints to the processor cache for
6866 better performance.</p>
6870 <!-- _______________________________________________________________________ -->
6872 <a name="int_pcmarker">'<tt>llvm.pcmarker</tt>' Intrinsic</a>
6879 declare void @llvm.pcmarker(i32 <id>)
6883 <p>The '<tt>llvm.pcmarker</tt>' intrinsic is a method to export a Program
6884 Counter (PC) in a region of code to simulators and other tools. The method
6885 is target specific, but it is expected that the marker will use exported
6886 symbols to transmit the PC of the marker. The marker makes no guarantees
6887 that it will remain with any specific instruction after optimizations. It is
6888 possible that the presence of a marker will inhibit optimizations. The
6889 intended use is to be inserted after optimizations to allow correlations of
6890 simulation runs.</p>
6893 <p><tt>id</tt> is a numerical id identifying the marker.</p>
6896 <p>This intrinsic does not modify the behavior of the program. Backends that do
6897 not support this intrinsic may ignore it.</p>
6901 <!-- _______________________________________________________________________ -->
6903 <a name="int_readcyclecounter">'<tt>llvm.readcyclecounter</tt>' Intrinsic</a>
6910 declare i64 @llvm.readcyclecounter()
6914 <p>The '<tt>llvm.readcyclecounter</tt>' intrinsic provides access to the cycle
6915 counter register (or similar low latency, high accuracy clocks) on those
6916 targets that support it. On X86, it should map to RDTSC. On Alpha, it
6917 should map to RPCC. As the backing counters overflow quickly (on the order
6918 of 9 seconds on alpha), this should only be used for small timings.</p>
6921 <p>When directly supported, reading the cycle counter should not modify any
6922 memory. Implementations are allowed to either return a application specific
6923 value or a system wide value. On backends without support, this is lowered
6924 to a constant 0.</p>
6930 <!-- ======================================================================= -->
6932 <a name="int_libc">Standard C Library Intrinsics</a>
6937 <p>LLVM provides intrinsics for a few important standard C library functions.
6938 These intrinsics allow source-language front-ends to pass information about
6939 the alignment of the pointer arguments to the code generator, providing
6940 opportunity for more efficient code generation.</p>
6942 <!-- _______________________________________________________________________ -->
6944 <a name="int_memcpy">'<tt>llvm.memcpy</tt>' Intrinsic</a>
6950 <p>This is an overloaded intrinsic. You can use <tt>llvm.memcpy</tt> on any
6951 integer bit width and for different address spaces. Not all targets support
6952 all bit widths however.</p>
6955 declare void @llvm.memcpy.p0i8.p0i8.i32(i8* <dest>, i8* <src>,
6956 i32 <len>, i32 <align>, i1 <isvolatile>)
6957 declare void @llvm.memcpy.p0i8.p0i8.i64(i8* <dest>, i8* <src>,
6958 i64 <len>, i32 <align>, i1 <isvolatile>)
6962 <p>The '<tt>llvm.memcpy.*</tt>' intrinsics copy a block of memory from the
6963 source location to the destination location.</p>
6965 <p>Note that, unlike the standard libc function, the <tt>llvm.memcpy.*</tt>
6966 intrinsics do not return a value, takes extra alignment/isvolatile arguments
6967 and the pointers can be in specified address spaces.</p>
6971 <p>The first argument is a pointer to the destination, the second is a pointer
6972 to the source. The third argument is an integer argument specifying the
6973 number of bytes to copy, the fourth argument is the alignment of the
6974 source and destination locations, and the fifth is a boolean indicating a
6975 volatile access.</p>
6977 <p>If the call to this intrinsic has an alignment value that is not 0 or 1,
6978 then the caller guarantees that both the source and destination pointers are
6979 aligned to that boundary.</p>
6981 <p>If the <tt>isvolatile</tt> parameter is <tt>true</tt>, the
6982 <tt>llvm.memcpy</tt> call is a <a href="#volatile">volatile operation</a>.
6983 The detailed access behavior is not very cleanly specified and it is unwise
6984 to depend on it.</p>
6988 <p>The '<tt>llvm.memcpy.*</tt>' intrinsics copy a block of memory from the
6989 source location to the destination location, which are not allowed to
6990 overlap. It copies "len" bytes of memory over. If the argument is known to
6991 be aligned to some boundary, this can be specified as the fourth argument,
6992 otherwise it should be set to 0 or 1.</p>
6996 <!-- _______________________________________________________________________ -->
6998 <a name="int_memmove">'<tt>llvm.memmove</tt>' Intrinsic</a>
7004 <p>This is an overloaded intrinsic. You can use llvm.memmove on any integer bit
7005 width and for different address space. Not all targets support all bit
7009 declare void @llvm.memmove.p0i8.p0i8.i32(i8* <dest>, i8* <src>,
7010 i32 <len>, i32 <align>, i1 <isvolatile>)
7011 declare void @llvm.memmove.p0i8.p0i8.i64(i8* <dest>, i8* <src>,
7012 i64 <len>, i32 <align>, i1 <isvolatile>)
7016 <p>The '<tt>llvm.memmove.*</tt>' intrinsics move a block of memory from the
7017 source location to the destination location. It is similar to the
7018 '<tt>llvm.memcpy</tt>' intrinsic but allows the two memory locations to
7021 <p>Note that, unlike the standard libc function, the <tt>llvm.memmove.*</tt>
7022 intrinsics do not return a value, takes extra alignment/isvolatile arguments
7023 and the pointers can be in specified address spaces.</p>
7027 <p>The first argument is a pointer to the destination, the second is a pointer
7028 to the source. The third argument is an integer argument specifying the
7029 number of bytes to copy, the fourth argument is the alignment of the
7030 source and destination locations, and the fifth is a boolean indicating a
7031 volatile access.</p>
7033 <p>If the call to this intrinsic has an alignment value that is not 0 or 1,
7034 then the caller guarantees that the source and destination pointers are
7035 aligned to that boundary.</p>
7037 <p>If the <tt>isvolatile</tt> parameter is <tt>true</tt>, the
7038 <tt>llvm.memmove</tt> call is a <a href="#volatile">volatile operation</a>.
7039 The detailed access behavior is not very cleanly specified and it is unwise
7040 to depend on it.</p>
7044 <p>The '<tt>llvm.memmove.*</tt>' intrinsics copy a block of memory from the
7045 source location to the destination location, which may overlap. It copies
7046 "len" bytes of memory over. If the argument is known to be aligned to some
7047 boundary, this can be specified as the fourth argument, otherwise it should
7048 be set to 0 or 1.</p>
7052 <!-- _______________________________________________________________________ -->
7054 <a name="int_memset">'<tt>llvm.memset.*</tt>' Intrinsics</a>
7060 <p>This is an overloaded intrinsic. You can use llvm.memset on any integer bit
7061 width and for different address spaces. However, not all targets support all
7065 declare void @llvm.memset.p0i8.i32(i8* <dest>, i8 <val>,
7066 i32 <len>, i32 <align>, i1 <isvolatile>)
7067 declare void @llvm.memset.p0i8.i64(i8* <dest>, i8 <val>,
7068 i64 <len>, i32 <align>, i1 <isvolatile>)
7072 <p>The '<tt>llvm.memset.*</tt>' intrinsics fill a block of memory with a
7073 particular byte value.</p>
7075 <p>Note that, unlike the standard libc function, the <tt>llvm.memset</tt>
7076 intrinsic does not return a value and takes extra alignment/volatile
7077 arguments. Also, the destination can be in an arbitrary address space.</p>
7080 <p>The first argument is a pointer to the destination to fill, the second is the
7081 byte value with which to fill it, the third argument is an integer argument
7082 specifying the number of bytes to fill, and the fourth argument is the known
7083 alignment of the destination location.</p>
7085 <p>If the call to this intrinsic has an alignment value that is not 0 or 1,
7086 then the caller guarantees that the destination pointer is aligned to that
7089 <p>If the <tt>isvolatile</tt> parameter is <tt>true</tt>, the
7090 <tt>llvm.memset</tt> call is a <a href="#volatile">volatile operation</a>.
7091 The detailed access behavior is not very cleanly specified and it is unwise
7092 to depend on it.</p>
7095 <p>The '<tt>llvm.memset.*</tt>' intrinsics fill "len" bytes of memory starting
7096 at the destination location. If the argument is known to be aligned to some
7097 boundary, this can be specified as the fourth argument, otherwise it should
7098 be set to 0 or 1.</p>
7102 <!-- _______________________________________________________________________ -->
7104 <a name="int_sqrt">'<tt>llvm.sqrt.*</tt>' Intrinsic</a>
7110 <p>This is an overloaded intrinsic. You can use <tt>llvm.sqrt</tt> on any
7111 floating point or vector of floating point type. Not all targets support all
7115 declare float @llvm.sqrt.f32(float %Val)
7116 declare double @llvm.sqrt.f64(double %Val)
7117 declare x86_fp80 @llvm.sqrt.f80(x86_fp80 %Val)
7118 declare fp128 @llvm.sqrt.f128(fp128 %Val)
7119 declare ppc_fp128 @llvm.sqrt.ppcf128(ppc_fp128 %Val)
7123 <p>The '<tt>llvm.sqrt</tt>' intrinsics return the sqrt of the specified operand,
7124 returning the same value as the libm '<tt>sqrt</tt>' functions would.
7125 Unlike <tt>sqrt</tt> in libm, however, <tt>llvm.sqrt</tt> has undefined
7126 behavior for negative numbers other than -0.0 (which allows for better
7127 optimization, because there is no need to worry about errno being
7128 set). <tt>llvm.sqrt(-0.0)</tt> is defined to return -0.0 like IEEE sqrt.</p>
7131 <p>The argument and return value are floating point numbers of the same
7135 <p>This function returns the sqrt of the specified operand if it is a
7136 nonnegative floating point number.</p>
7140 <!-- _______________________________________________________________________ -->
7142 <a name="int_powi">'<tt>llvm.powi.*</tt>' Intrinsic</a>
7148 <p>This is an overloaded intrinsic. You can use <tt>llvm.powi</tt> on any
7149 floating point or vector of floating point type. Not all targets support all
7153 declare float @llvm.powi.f32(float %Val, i32 %power)
7154 declare double @llvm.powi.f64(double %Val, i32 %power)
7155 declare x86_fp80 @llvm.powi.f80(x86_fp80 %Val, i32 %power)
7156 declare fp128 @llvm.powi.f128(fp128 %Val, i32 %power)
7157 declare ppc_fp128 @llvm.powi.ppcf128(ppc_fp128 %Val, i32 %power)
7161 <p>The '<tt>llvm.powi.*</tt>' intrinsics return the first operand raised to the
7162 specified (positive or negative) power. The order of evaluation of
7163 multiplications is not defined. When a vector of floating point type is
7164 used, the second argument remains a scalar integer value.</p>
7167 <p>The second argument is an integer power, and the first is a value to raise to
7171 <p>This function returns the first value raised to the second power with an
7172 unspecified sequence of rounding operations.</p>
7176 <!-- _______________________________________________________________________ -->
7178 <a name="int_sin">'<tt>llvm.sin.*</tt>' Intrinsic</a>
7184 <p>This is an overloaded intrinsic. You can use <tt>llvm.sin</tt> on any
7185 floating point or vector of floating point type. Not all targets support all
7189 declare float @llvm.sin.f32(float %Val)
7190 declare double @llvm.sin.f64(double %Val)
7191 declare x86_fp80 @llvm.sin.f80(x86_fp80 %Val)
7192 declare fp128 @llvm.sin.f128(fp128 %Val)
7193 declare ppc_fp128 @llvm.sin.ppcf128(ppc_fp128 %Val)
7197 <p>The '<tt>llvm.sin.*</tt>' intrinsics return the sine of the operand.</p>
7200 <p>The argument and return value are floating point numbers of the same
7204 <p>This function returns the sine of the specified operand, returning the same
7205 values as the libm <tt>sin</tt> functions would, and handles error conditions
7206 in the same way.</p>
7210 <!-- _______________________________________________________________________ -->
7212 <a name="int_cos">'<tt>llvm.cos.*</tt>' Intrinsic</a>
7218 <p>This is an overloaded intrinsic. You can use <tt>llvm.cos</tt> on any
7219 floating point or vector of floating point type. Not all targets support all
7223 declare float @llvm.cos.f32(float %Val)
7224 declare double @llvm.cos.f64(double %Val)
7225 declare x86_fp80 @llvm.cos.f80(x86_fp80 %Val)
7226 declare fp128 @llvm.cos.f128(fp128 %Val)
7227 declare ppc_fp128 @llvm.cos.ppcf128(ppc_fp128 %Val)
7231 <p>The '<tt>llvm.cos.*</tt>' intrinsics return the cosine of the operand.</p>
7234 <p>The argument and return value are floating point numbers of the same
7238 <p>This function returns the cosine of the specified operand, returning the same
7239 values as the libm <tt>cos</tt> functions would, and handles error conditions
7240 in the same way.</p>
7244 <!-- _______________________________________________________________________ -->
7246 <a name="int_pow">'<tt>llvm.pow.*</tt>' Intrinsic</a>
7252 <p>This is an overloaded intrinsic. You can use <tt>llvm.pow</tt> on any
7253 floating point or vector of floating point type. Not all targets support all
7257 declare float @llvm.pow.f32(float %Val, float %Power)
7258 declare double @llvm.pow.f64(double %Val, double %Power)
7259 declare x86_fp80 @llvm.pow.f80(x86_fp80 %Val, x86_fp80 %Power)
7260 declare fp128 @llvm.pow.f128(fp128 %Val, fp128 %Power)
7261 declare ppc_fp128 @llvm.pow.ppcf128(ppc_fp128 %Val, ppc_fp128 Power)
7265 <p>The '<tt>llvm.pow.*</tt>' intrinsics return the first operand raised to the
7266 specified (positive or negative) power.</p>
7269 <p>The second argument is a floating point power, and the first is a value to
7270 raise to that power.</p>
7273 <p>This function returns the first value raised to the second power, returning
7274 the same values as the libm <tt>pow</tt> functions would, and handles error
7275 conditions in the same way.</p>
7279 <!-- _______________________________________________________________________ -->
7281 <a name="int_exp">'<tt>llvm.exp.*</tt>' Intrinsic</a>
7287 <p>This is an overloaded intrinsic. You can use <tt>llvm.exp</tt> on any
7288 floating point or vector of floating point type. Not all targets support all
7292 declare float @llvm.exp.f32(float %Val)
7293 declare double @llvm.exp.f64(double %Val)
7294 declare x86_fp80 @llvm.exp.f80(x86_fp80 %Val)
7295 declare fp128 @llvm.exp.f128(fp128 %Val)
7296 declare ppc_fp128 @llvm.exp.ppcf128(ppc_fp128 %Val)
7300 <p>The '<tt>llvm.exp.*</tt>' intrinsics perform the exp function.</p>
7303 <p>The argument and return value are floating point numbers of the same
7307 <p>This function returns the same values as the libm <tt>exp</tt> functions
7308 would, and handles error conditions in the same way.</p>
7312 <!-- _______________________________________________________________________ -->
7314 <a name="int_log">'<tt>llvm.log.*</tt>' Intrinsic</a>
7320 <p>This is an overloaded intrinsic. You can use <tt>llvm.log</tt> on any
7321 floating point or vector of floating point type. Not all targets support all
7325 declare float @llvm.log.f32(float %Val)
7326 declare double @llvm.log.f64(double %Val)
7327 declare x86_fp80 @llvm.log.f80(x86_fp80 %Val)
7328 declare fp128 @llvm.log.f128(fp128 %Val)
7329 declare ppc_fp128 @llvm.log.ppcf128(ppc_fp128 %Val)
7333 <p>The '<tt>llvm.log.*</tt>' intrinsics perform the log function.</p>
7336 <p>The argument and return value are floating point numbers of the same
7340 <p>This function returns the same values as the libm <tt>log</tt> functions
7341 would, and handles error conditions in the same way.</p>
7345 <!-- _______________________________________________________________________ -->
7347 <a name="int_fma">'<tt>llvm.fma.*</tt>' Intrinsic</a>
7353 <p>This is an overloaded intrinsic. You can use <tt>llvm.fma</tt> on any
7354 floating point or vector of floating point type. Not all targets support all
7358 declare float @llvm.fma.f32(float %a, float %b, float %c)
7359 declare double @llvm.fma.f64(double %a, double %b, double %c)
7360 declare x86_fp80 @llvm.fma.f80(x86_fp80 %a, x86_fp80 %b, x86_fp80 %c)
7361 declare fp128 @llvm.fma.f128(fp128 %a, fp128 %b, fp128 %c)
7362 declare ppc_fp128 @llvm.fma.ppcf128(ppc_fp128 %a, ppc_fp128 %b, ppc_fp128 %c)
7366 <p>The '<tt>llvm.fma.*</tt>' intrinsics perform the fused multiply-add
7370 <p>The argument and return value are floating point numbers of the same
7374 <p>This function returns the same values as the libm <tt>fma</tt> functions
7381 <!-- ======================================================================= -->
7383 <a name="int_manip">Bit Manipulation Intrinsics</a>
7388 <p>LLVM provides intrinsics for a few important bit manipulation operations.
7389 These allow efficient code generation for some algorithms.</p>
7391 <!-- _______________________________________________________________________ -->
7393 <a name="int_bswap">'<tt>llvm.bswap.*</tt>' Intrinsics</a>
7399 <p>This is an overloaded intrinsic function. You can use bswap on any integer
7400 type that is an even number of bytes (i.e. BitWidth % 16 == 0).</p>
7403 declare i16 @llvm.bswap.i16(i16 <id>)
7404 declare i32 @llvm.bswap.i32(i32 <id>)
7405 declare i64 @llvm.bswap.i64(i64 <id>)
7409 <p>The '<tt>llvm.bswap</tt>' family of intrinsics is used to byte swap integer
7410 values with an even number of bytes (positive multiple of 16 bits). These
7411 are useful for performing operations on data that is not in the target's
7412 native byte order.</p>
7415 <p>The <tt>llvm.bswap.i16</tt> intrinsic returns an i16 value that has the high
7416 and low byte of the input i16 swapped. Similarly,
7417 the <tt>llvm.bswap.i32</tt> intrinsic returns an i32 value that has the four
7418 bytes of the input i32 swapped, so that if the input bytes are numbered 0, 1,
7419 2, 3 then the returned i32 will have its bytes in 3, 2, 1, 0 order.
7420 The <tt>llvm.bswap.i48</tt>, <tt>llvm.bswap.i64</tt> and other intrinsics
7421 extend this concept to additional even-byte lengths (6 bytes, 8 bytes and
7422 more, respectively).</p>
7426 <!-- _______________________________________________________________________ -->
7428 <a name="int_ctpop">'<tt>llvm.ctpop.*</tt>' Intrinsic</a>
7434 <p>This is an overloaded intrinsic. You can use llvm.ctpop on any integer bit
7435 width, or on any vector with integer elements. Not all targets support all
7436 bit widths or vector types, however.</p>
7439 declare i8 @llvm.ctpop.i8(i8 <src>)
7440 declare i16 @llvm.ctpop.i16(i16 <src>)
7441 declare i32 @llvm.ctpop.i32(i32 <src>)
7442 declare i64 @llvm.ctpop.i64(i64 <src>)
7443 declare i256 @llvm.ctpop.i256(i256 <src>)
7444 declare <2 x i32> @llvm.ctpop.v2i32(<2 x i32> <src>)
7448 <p>The '<tt>llvm.ctpop</tt>' family of intrinsics counts the number of bits set
7452 <p>The only argument is the value to be counted. The argument may be of any
7453 integer type, or a vector with integer elements.
7454 The return type must match the argument type.</p>
7457 <p>The '<tt>llvm.ctpop</tt>' intrinsic counts the 1's in a variable, or within each
7458 element of a vector.</p>
7462 <!-- _______________________________________________________________________ -->
7464 <a name="int_ctlz">'<tt>llvm.ctlz.*</tt>' Intrinsic</a>
7470 <p>This is an overloaded intrinsic. You can use <tt>llvm.ctlz</tt> on any
7471 integer bit width, or any vector whose elements are integers. Not all
7472 targets support all bit widths or vector types, however.</p>
7475 declare i8 @llvm.ctlz.i8 (i8 <src>, i1 <is_zero_undef>)
7476 declare i16 @llvm.ctlz.i16 (i16 <src>, i1 <is_zero_undef>)
7477 declare i32 @llvm.ctlz.i32 (i32 <src>, i1 <is_zero_undef>)
7478 declare i64 @llvm.ctlz.i64 (i64 <src>, i1 <is_zero_undef>)
7479 declare i256 @llvm.ctlz.i256(i256 <src>, i1 <is_zero_undef>)
7480 declase <2 x i32> @llvm.ctlz.v2i32(<2 x i32> <src>, i1 <is_zero_undef>)
7484 <p>The '<tt>llvm.ctlz</tt>' family of intrinsic functions counts the number of
7485 leading zeros in a variable.</p>
7488 <p>The first argument is the value to be counted. This argument may be of any
7489 integer type, or a vectory with integer element type. The return type
7490 must match the first argument type.</p>
7492 <p>The second argument must be a constant and is a flag to indicate whether the
7493 intrinsic should ensure that a zero as the first argument produces a defined
7494 result. Historically some architectures did not provide a defined result for
7495 zero values as efficiently, and many algorithms are now predicated on
7496 avoiding zero-value inputs.</p>
7499 <p>The '<tt>llvm.ctlz</tt>' intrinsic counts the leading (most significant)
7500 zeros in a variable, or within each element of the vector.
7501 If <tt>src == 0</tt> then the result is the size in bits of the type of
7502 <tt>src</tt> if <tt>is_zero_undef == 0</tt> and <tt>undef</tt> otherwise.
7503 For example, <tt>llvm.ctlz(i32 2) = 30</tt>.</p>
7507 <!-- _______________________________________________________________________ -->
7509 <a name="int_cttz">'<tt>llvm.cttz.*</tt>' Intrinsic</a>
7515 <p>This is an overloaded intrinsic. You can use <tt>llvm.cttz</tt> on any
7516 integer bit width, or any vector of integer elements. Not all targets
7517 support all bit widths or vector types, however.</p>
7520 declare i8 @llvm.cttz.i8 (i8 <src>, i1 <is_zero_undef>)
7521 declare i16 @llvm.cttz.i16 (i16 <src>, i1 <is_zero_undef>)
7522 declare i32 @llvm.cttz.i32 (i32 <src>, i1 <is_zero_undef>)
7523 declare i64 @llvm.cttz.i64 (i64 <src>, i1 <is_zero_undef>)
7524 declare i256 @llvm.cttz.i256(i256 <src>, i1 <is_zero_undef>)
7525 declase <2 x i32> @llvm.cttz.v2i32(<2 x i32> <src>, i1 <is_zero_undef>)
7529 <p>The '<tt>llvm.cttz</tt>' family of intrinsic functions counts the number of
7533 <p>The first argument is the value to be counted. This argument may be of any
7534 integer type, or a vectory with integer element type. The return type
7535 must match the first argument type.</p>
7537 <p>The second argument must be a constant and is a flag to indicate whether the
7538 intrinsic should ensure that a zero as the first argument produces a defined
7539 result. Historically some architectures did not provide a defined result for
7540 zero values as efficiently, and many algorithms are now predicated on
7541 avoiding zero-value inputs.</p>
7544 <p>The '<tt>llvm.cttz</tt>' intrinsic counts the trailing (least significant)
7545 zeros in a variable, or within each element of a vector.
7546 If <tt>src == 0</tt> then the result is the size in bits of the type of
7547 <tt>src</tt> if <tt>is_zero_undef == 0</tt> and <tt>undef</tt> otherwise.
7548 For example, <tt>llvm.cttz(2) = 1</tt>.</p>
7554 <!-- ======================================================================= -->
7556 <a name="int_overflow">Arithmetic with Overflow Intrinsics</a>
7561 <p>LLVM provides intrinsics for some arithmetic with overflow operations.</p>
7563 <!-- _______________________________________________________________________ -->
7565 <a name="int_sadd_overflow">
7566 '<tt>llvm.sadd.with.overflow.*</tt>' Intrinsics
7573 <p>This is an overloaded intrinsic. You can use <tt>llvm.sadd.with.overflow</tt>
7574 on any integer bit width.</p>
7577 declare {i16, i1} @llvm.sadd.with.overflow.i16(i16 %a, i16 %b)
7578 declare {i32, i1} @llvm.sadd.with.overflow.i32(i32 %a, i32 %b)
7579 declare {i64, i1} @llvm.sadd.with.overflow.i64(i64 %a, i64 %b)
7583 <p>The '<tt>llvm.sadd.with.overflow</tt>' family of intrinsic functions perform
7584 a signed addition of the two arguments, and indicate whether an overflow
7585 occurred during the signed summation.</p>
7588 <p>The arguments (%a and %b) and the first element of the result structure may
7589 be of integer types of any bit width, but they must have the same bit
7590 width. The second element of the result structure must be of
7591 type <tt>i1</tt>. <tt>%a</tt> and <tt>%b</tt> are the two values that will
7592 undergo signed addition.</p>
7595 <p>The '<tt>llvm.sadd.with.overflow</tt>' family of intrinsic functions perform
7596 a signed addition of the two variables. They return a structure — the
7597 first element of which is the signed summation, and the second element of
7598 which is a bit specifying if the signed summation resulted in an
7603 %res = call {i32, i1} @llvm.sadd.with.overflow.i32(i32 %a, i32 %b)
7604 %sum = extractvalue {i32, i1} %res, 0
7605 %obit = extractvalue {i32, i1} %res, 1
7606 br i1 %obit, label %overflow, label %normal
7611 <!-- _______________________________________________________________________ -->
7613 <a name="int_uadd_overflow">
7614 '<tt>llvm.uadd.with.overflow.*</tt>' Intrinsics
7621 <p>This is an overloaded intrinsic. You can use <tt>llvm.uadd.with.overflow</tt>
7622 on any integer bit width.</p>
7625 declare {i16, i1} @llvm.uadd.with.overflow.i16(i16 %a, i16 %b)
7626 declare {i32, i1} @llvm.uadd.with.overflow.i32(i32 %a, i32 %b)
7627 declare {i64, i1} @llvm.uadd.with.overflow.i64(i64 %a, i64 %b)
7631 <p>The '<tt>llvm.uadd.with.overflow</tt>' family of intrinsic functions perform
7632 an unsigned addition of the two arguments, and indicate whether a carry
7633 occurred during the unsigned summation.</p>
7636 <p>The arguments (%a and %b) and the first element of the result structure may
7637 be of integer types of any bit width, but they must have the same bit
7638 width. The second element of the result structure must be of
7639 type <tt>i1</tt>. <tt>%a</tt> and <tt>%b</tt> are the two values that will
7640 undergo unsigned addition.</p>
7643 <p>The '<tt>llvm.uadd.with.overflow</tt>' family of intrinsic functions perform
7644 an unsigned addition of the two arguments. They return a structure —
7645 the first element of which is the sum, and the second element of which is a
7646 bit specifying if the unsigned summation resulted in a carry.</p>
7650 %res = call {i32, i1} @llvm.uadd.with.overflow.i32(i32 %a, i32 %b)
7651 %sum = extractvalue {i32, i1} %res, 0
7652 %obit = extractvalue {i32, i1} %res, 1
7653 br i1 %obit, label %carry, label %normal
7658 <!-- _______________________________________________________________________ -->
7660 <a name="int_ssub_overflow">
7661 '<tt>llvm.ssub.with.overflow.*</tt>' Intrinsics
7668 <p>This is an overloaded intrinsic. You can use <tt>llvm.ssub.with.overflow</tt>
7669 on any integer bit width.</p>
7672 declare {i16, i1} @llvm.ssub.with.overflow.i16(i16 %a, i16 %b)
7673 declare {i32, i1} @llvm.ssub.with.overflow.i32(i32 %a, i32 %b)
7674 declare {i64, i1} @llvm.ssub.with.overflow.i64(i64 %a, i64 %b)
7678 <p>The '<tt>llvm.ssub.with.overflow</tt>' family of intrinsic functions perform
7679 a signed subtraction of the two arguments, and indicate whether an overflow
7680 occurred during the signed subtraction.</p>
7683 <p>The arguments (%a and %b) and the first element of the result structure may
7684 be of integer types of any bit width, but they must have the same bit
7685 width. The second element of the result structure must be of
7686 type <tt>i1</tt>. <tt>%a</tt> and <tt>%b</tt> are the two values that will
7687 undergo signed subtraction.</p>
7690 <p>The '<tt>llvm.ssub.with.overflow</tt>' family of intrinsic functions perform
7691 a signed subtraction of the two arguments. They return a structure —
7692 the first element of which is the subtraction, and the second element of
7693 which is a bit specifying if the signed subtraction resulted in an
7698 %res = call {i32, i1} @llvm.ssub.with.overflow.i32(i32 %a, i32 %b)
7699 %sum = extractvalue {i32, i1} %res, 0
7700 %obit = extractvalue {i32, i1} %res, 1
7701 br i1 %obit, label %overflow, label %normal
7706 <!-- _______________________________________________________________________ -->
7708 <a name="int_usub_overflow">
7709 '<tt>llvm.usub.with.overflow.*</tt>' Intrinsics
7716 <p>This is an overloaded intrinsic. You can use <tt>llvm.usub.with.overflow</tt>
7717 on any integer bit width.</p>
7720 declare {i16, i1} @llvm.usub.with.overflow.i16(i16 %a, i16 %b)
7721 declare {i32, i1} @llvm.usub.with.overflow.i32(i32 %a, i32 %b)
7722 declare {i64, i1} @llvm.usub.with.overflow.i64(i64 %a, i64 %b)
7726 <p>The '<tt>llvm.usub.with.overflow</tt>' family of intrinsic functions perform
7727 an unsigned subtraction of the two arguments, and indicate whether an
7728 overflow occurred during the unsigned subtraction.</p>
7731 <p>The arguments (%a and %b) and the first element of the result structure may
7732 be of integer types of any bit width, but they must have the same bit
7733 width. The second element of the result structure must be of
7734 type <tt>i1</tt>. <tt>%a</tt> and <tt>%b</tt> are the two values that will
7735 undergo unsigned subtraction.</p>
7738 <p>The '<tt>llvm.usub.with.overflow</tt>' family of intrinsic functions perform
7739 an unsigned subtraction of the two arguments. They return a structure —
7740 the first element of which is the subtraction, and the second element of
7741 which is a bit specifying if the unsigned subtraction resulted in an
7746 %res = call {i32, i1} @llvm.usub.with.overflow.i32(i32 %a, i32 %b)
7747 %sum = extractvalue {i32, i1} %res, 0
7748 %obit = extractvalue {i32, i1} %res, 1
7749 br i1 %obit, label %overflow, label %normal
7754 <!-- _______________________________________________________________________ -->
7756 <a name="int_smul_overflow">
7757 '<tt>llvm.smul.with.overflow.*</tt>' Intrinsics
7764 <p>This is an overloaded intrinsic. You can use <tt>llvm.smul.with.overflow</tt>
7765 on any integer bit width.</p>
7768 declare {i16, i1} @llvm.smul.with.overflow.i16(i16 %a, i16 %b)
7769 declare {i32, i1} @llvm.smul.with.overflow.i32(i32 %a, i32 %b)
7770 declare {i64, i1} @llvm.smul.with.overflow.i64(i64 %a, i64 %b)
7775 <p>The '<tt>llvm.smul.with.overflow</tt>' family of intrinsic functions perform
7776 a signed multiplication of the two arguments, and indicate whether an
7777 overflow occurred during the signed multiplication.</p>
7780 <p>The arguments (%a and %b) and the first element of the result structure may
7781 be of integer types of any bit width, but they must have the same bit
7782 width. The second element of the result structure must be of
7783 type <tt>i1</tt>. <tt>%a</tt> and <tt>%b</tt> are the two values that will
7784 undergo signed multiplication.</p>
7787 <p>The '<tt>llvm.smul.with.overflow</tt>' family of intrinsic functions perform
7788 a signed multiplication of the two arguments. They return a structure —
7789 the first element of which is the multiplication, and the second element of
7790 which is a bit specifying if the signed multiplication resulted in an
7795 %res = call {i32, i1} @llvm.smul.with.overflow.i32(i32 %a, i32 %b)
7796 %sum = extractvalue {i32, i1} %res, 0
7797 %obit = extractvalue {i32, i1} %res, 1
7798 br i1 %obit, label %overflow, label %normal
7803 <!-- _______________________________________________________________________ -->
7805 <a name="int_umul_overflow">
7806 '<tt>llvm.umul.with.overflow.*</tt>' Intrinsics
7813 <p>This is an overloaded intrinsic. You can use <tt>llvm.umul.with.overflow</tt>
7814 on any integer bit width.</p>
7817 declare {i16, i1} @llvm.umul.with.overflow.i16(i16 %a, i16 %b)
7818 declare {i32, i1} @llvm.umul.with.overflow.i32(i32 %a, i32 %b)
7819 declare {i64, i1} @llvm.umul.with.overflow.i64(i64 %a, i64 %b)
7823 <p>The '<tt>llvm.umul.with.overflow</tt>' family of intrinsic functions perform
7824 a unsigned multiplication of the two arguments, and indicate whether an
7825 overflow occurred during the unsigned multiplication.</p>
7828 <p>The arguments (%a and %b) and the first element of the result structure may
7829 be of integer types of any bit width, but they must have the same bit
7830 width. The second element of the result structure must be of
7831 type <tt>i1</tt>. <tt>%a</tt> and <tt>%b</tt> are the two values that will
7832 undergo unsigned multiplication.</p>
7835 <p>The '<tt>llvm.umul.with.overflow</tt>' family of intrinsic functions perform
7836 an unsigned multiplication of the two arguments. They return a structure
7837 — the first element of which is the multiplication, and the second
7838 element of which is a bit specifying if the unsigned multiplication resulted
7843 %res = call {i32, i1} @llvm.umul.with.overflow.i32(i32 %a, i32 %b)
7844 %sum = extractvalue {i32, i1} %res, 0
7845 %obit = extractvalue {i32, i1} %res, 1
7846 br i1 %obit, label %overflow, label %normal
7853 <!-- ======================================================================= -->
7855 <a name="int_fp16">Half Precision Floating Point Intrinsics</a>
7860 <p>Half precision floating point is a storage-only format. This means that it is
7861 a dense encoding (in memory) but does not support computation in the
7864 <p>This means that code must first load the half-precision floating point
7865 value as an i16, then convert it to float with <a
7866 href="#int_convert_from_fp16"><tt>llvm.convert.from.fp16</tt></a>.
7867 Computation can then be performed on the float value (including extending to
7868 double etc). To store the value back to memory, it is first converted to
7869 float if needed, then converted to i16 with
7870 <a href="#int_convert_to_fp16"><tt>llvm.convert.to.fp16</tt></a>, then
7871 storing as an i16 value.</p>
7873 <!-- _______________________________________________________________________ -->
7875 <a name="int_convert_to_fp16">
7876 '<tt>llvm.convert.to.fp16</tt>' Intrinsic
7884 declare i16 @llvm.convert.to.fp16(f32 %a)
7888 <p>The '<tt>llvm.convert.to.fp16</tt>' intrinsic function performs
7889 a conversion from single precision floating point format to half precision
7890 floating point format.</p>
7893 <p>The intrinsic function contains single argument - the value to be
7897 <p>The '<tt>llvm.convert.to.fp16</tt>' intrinsic function performs
7898 a conversion from single precision floating point format to half precision
7899 floating point format. The return value is an <tt>i16</tt> which
7900 contains the converted number.</p>
7904 %res = call i16 @llvm.convert.to.fp16(f32 %a)
7905 store i16 %res, i16* @x, align 2
7910 <!-- _______________________________________________________________________ -->
7912 <a name="int_convert_from_fp16">
7913 '<tt>llvm.convert.from.fp16</tt>' Intrinsic
7921 declare f32 @llvm.convert.from.fp16(i16 %a)
7925 <p>The '<tt>llvm.convert.from.fp16</tt>' intrinsic function performs
7926 a conversion from half precision floating point format to single precision
7927 floating point format.</p>
7930 <p>The intrinsic function contains single argument - the value to be
7934 <p>The '<tt>llvm.convert.from.fp16</tt>' intrinsic function performs a
7935 conversion from half single precision floating point format to single
7936 precision floating point format. The input half-float value is represented by
7937 an <tt>i16</tt> value.</p>
7941 %a = load i16* @x, align 2
7942 %res = call f32 @llvm.convert.from.fp16(i16 %a)
7949 <!-- ======================================================================= -->
7951 <a name="int_debugger">Debugger Intrinsics</a>
7956 <p>The LLVM debugger intrinsics (which all start with <tt>llvm.dbg.</tt>
7957 prefix), are described in
7958 the <a href="SourceLevelDebugging.html#format_common_intrinsics">LLVM Source
7959 Level Debugging</a> document.</p>
7963 <!-- ======================================================================= -->
7965 <a name="int_eh">Exception Handling Intrinsics</a>
7970 <p>The LLVM exception handling intrinsics (which all start with
7971 <tt>llvm.eh.</tt> prefix), are described in
7972 the <a href="ExceptionHandling.html#format_common_intrinsics">LLVM Exception
7973 Handling</a> document.</p>
7977 <!-- ======================================================================= -->
7979 <a name="int_trampoline">Trampoline Intrinsics</a>
7984 <p>These intrinsics make it possible to excise one parameter, marked with
7985 the <a href="#nest"><tt>nest</tt></a> attribute, from a function.
7986 The result is a callable
7987 function pointer lacking the nest parameter - the caller does not need to
7988 provide a value for it. Instead, the value to use is stored in advance in a
7989 "trampoline", a block of memory usually allocated on the stack, which also
7990 contains code to splice the nest value into the argument list. This is used
7991 to implement the GCC nested function address extension.</p>
7993 <p>For example, if the function is
7994 <tt>i32 f(i8* nest %c, i32 %x, i32 %y)</tt> then the resulting function
7995 pointer has signature <tt>i32 (i32, i32)*</tt>. It can be created as
7998 <pre class="doc_code">
7999 %tramp = alloca [10 x i8], align 4 ; size and alignment only correct for X86
8000 %tramp1 = getelementptr [10 x i8]* %tramp, i32 0, i32 0
8001 call i8* @llvm.init.trampoline(i8* %tramp1, i8* bitcast (i32 (i8*, i32, i32)* @f to i8*), i8* %nval)
8002 %p = call i8* @llvm.adjust.trampoline(i8* %tramp1)
8003 %fp = bitcast i8* %p to i32 (i32, i32)*
8006 <p>The call <tt>%val = call i32 %fp(i32 %x, i32 %y)</tt> is then equivalent
8007 to <tt>%val = call i32 %f(i8* %nval, i32 %x, i32 %y)</tt>.</p>
8009 <!-- _______________________________________________________________________ -->
8012 '<tt>llvm.init.trampoline</tt>' Intrinsic
8020 declare void @llvm.init.trampoline(i8* <tramp>, i8* <func>, i8* <nval>)
8024 <p>This fills the memory pointed to by <tt>tramp</tt> with executable code,
8025 turning it into a trampoline.</p>
8028 <p>The <tt>llvm.init.trampoline</tt> intrinsic takes three arguments, all
8029 pointers. The <tt>tramp</tt> argument must point to a sufficiently large and
8030 sufficiently aligned block of memory; this memory is written to by the
8031 intrinsic. Note that the size and the alignment are target-specific - LLVM
8032 currently provides no portable way of determining them, so a front-end that
8033 generates this intrinsic needs to have some target-specific knowledge.
8034 The <tt>func</tt> argument must hold a function bitcast to
8035 an <tt>i8*</tt>.</p>
8038 <p>The block of memory pointed to by <tt>tramp</tt> is filled with target
8039 dependent code, turning it into a function. Then <tt>tramp</tt> needs to be
8040 passed to <a href="#int_at">llvm.adjust.trampoline</a> to get a pointer
8041 which can be <a href="#int_trampoline">bitcast (to a new function) and
8042 called</a>. The new function's signature is the same as that of
8043 <tt>func</tt> with any arguments marked with the <tt>nest</tt> attribute
8044 removed. At most one such <tt>nest</tt> argument is allowed, and it must be of
8045 pointer type. Calling the new function is equivalent to calling <tt>func</tt>
8046 with the same argument list, but with <tt>nval</tt> used for the missing
8047 <tt>nest</tt> argument. If, after calling <tt>llvm.init.trampoline</tt>, the
8048 memory pointed to by <tt>tramp</tt> is modified, then the effect of any later call
8049 to the returned function pointer is undefined.</p>
8052 <!-- _______________________________________________________________________ -->
8055 '<tt>llvm.adjust.trampoline</tt>' Intrinsic
8063 declare i8* @llvm.adjust.trampoline(i8* <tramp>)
8067 <p>This performs any required machine-specific adjustment to the address of a
8068 trampoline (passed as <tt>tramp</tt>).</p>
8071 <p><tt>tramp</tt> must point to a block of memory which already has trampoline code
8072 filled in by a previous call to <a href="#int_it"><tt>llvm.init.trampoline</tt>
8076 <p>On some architectures the address of the code to be executed needs to be
8077 different to the address where the trampoline is actually stored. This
8078 intrinsic returns the executable address corresponding to <tt>tramp</tt>
8079 after performing the required machine specific adjustments.
8080 The pointer returned can then be <a href="#int_trampoline"> bitcast and
8088 <!-- ======================================================================= -->
8090 <a name="int_memorymarkers">Memory Use Markers</a>
8095 <p>This class of intrinsics exists to information about the lifetime of memory
8096 objects and ranges where variables are immutable.</p>
8098 <!-- _______________________________________________________________________ -->
8100 <a name="int_lifetime_start">'<tt>llvm.lifetime.start</tt>' Intrinsic</a>
8107 declare void @llvm.lifetime.start(i64 <size>, i8* nocapture <ptr>)
8111 <p>The '<tt>llvm.lifetime.start</tt>' intrinsic specifies the start of a memory
8112 object's lifetime.</p>
8115 <p>The first argument is a constant integer representing the size of the
8116 object, or -1 if it is variable sized. The second argument is a pointer to
8120 <p>This intrinsic indicates that before this point in the code, the value of the
8121 memory pointed to by <tt>ptr</tt> is dead. This means that it is known to
8122 never be used and has an undefined value. A load from the pointer that
8123 precedes this intrinsic can be replaced with
8124 <tt>'<a href="#undefvalues">undef</a>'</tt>.</p>
8128 <!-- _______________________________________________________________________ -->
8130 <a name="int_lifetime_end">'<tt>llvm.lifetime.end</tt>' Intrinsic</a>
8137 declare void @llvm.lifetime.end(i64 <size>, i8* nocapture <ptr>)
8141 <p>The '<tt>llvm.lifetime.end</tt>' intrinsic specifies the end of a memory
8142 object's lifetime.</p>
8145 <p>The first argument is a constant integer representing the size of the
8146 object, or -1 if it is variable sized. The second argument is a pointer to
8150 <p>This intrinsic indicates that after this point in the code, the value of the
8151 memory pointed to by <tt>ptr</tt> is dead. This means that it is known to
8152 never be used and has an undefined value. Any stores into the memory object
8153 following this intrinsic may be removed as dead.
8157 <!-- _______________________________________________________________________ -->
8159 <a name="int_invariant_start">'<tt>llvm.invariant.start</tt>' Intrinsic</a>
8166 declare {}* @llvm.invariant.start(i64 <size>, i8* nocapture <ptr>)
8170 <p>The '<tt>llvm.invariant.start</tt>' intrinsic specifies that the contents of
8171 a memory object will not change.</p>
8174 <p>The first argument is a constant integer representing the size of the
8175 object, or -1 if it is variable sized. The second argument is a pointer to
8179 <p>This intrinsic indicates that until an <tt>llvm.invariant.end</tt> that uses
8180 the return value, the referenced memory location is constant and
8185 <!-- _______________________________________________________________________ -->
8187 <a name="int_invariant_end">'<tt>llvm.invariant.end</tt>' Intrinsic</a>
8194 declare void @llvm.invariant.end({}* <start>, i64 <size>, i8* nocapture <ptr>)
8198 <p>The '<tt>llvm.invariant.end</tt>' intrinsic specifies that the contents of
8199 a memory object are mutable.</p>
8202 <p>The first argument is the matching <tt>llvm.invariant.start</tt> intrinsic.
8203 The second argument is a constant integer representing the size of the
8204 object, or -1 if it is variable sized and the third argument is a pointer
8208 <p>This intrinsic indicates that the memory is mutable again.</p>
8214 <!-- ======================================================================= -->
8216 <a name="int_general">General Intrinsics</a>
8221 <p>This class of intrinsics is designed to be generic and has no specific
8224 <!-- _______________________________________________________________________ -->
8226 <a name="int_var_annotation">'<tt>llvm.var.annotation</tt>' Intrinsic</a>
8233 declare void @llvm.var.annotation(i8* <val>, i8* <str>, i8* <str>, i32 <int>)
8237 <p>The '<tt>llvm.var.annotation</tt>' intrinsic.</p>
8240 <p>The first argument is a pointer to a value, the second is a pointer to a
8241 global string, the third is a pointer to a global string which is the source
8242 file name, and the last argument is the line number.</p>
8245 <p>This intrinsic allows annotation of local variables with arbitrary strings.
8246 This can be useful for special purpose optimizations that want to look for
8247 these annotations. These have no other defined use; they are ignored by code
8248 generation and optimization.</p>
8252 <!-- _______________________________________________________________________ -->
8254 <a name="int_annotation">'<tt>llvm.annotation.*</tt>' Intrinsic</a>
8260 <p>This is an overloaded intrinsic. You can use '<tt>llvm.annotation</tt>' on
8261 any integer bit width.</p>
8264 declare i8 @llvm.annotation.i8(i8 <val>, i8* <str>, i8* <str>, i32 <int>)
8265 declare i16 @llvm.annotation.i16(i16 <val>, i8* <str>, i8* <str>, i32 <int>)
8266 declare i32 @llvm.annotation.i32(i32 <val>, i8* <str>, i8* <str>, i32 <int>)
8267 declare i64 @llvm.annotation.i64(i64 <val>, i8* <str>, i8* <str>, i32 <int>)
8268 declare i256 @llvm.annotation.i256(i256 <val>, i8* <str>, i8* <str>, i32 <int>)
8272 <p>The '<tt>llvm.annotation</tt>' intrinsic.</p>
8275 <p>The first argument is an integer value (result of some expression), the
8276 second is a pointer to a global string, the third is a pointer to a global
8277 string which is the source file name, and the last argument is the line
8278 number. It returns the value of the first argument.</p>
8281 <p>This intrinsic allows annotations to be put on arbitrary expressions with
8282 arbitrary strings. This can be useful for special purpose optimizations that
8283 want to look for these annotations. These have no other defined use; they
8284 are ignored by code generation and optimization.</p>
8288 <!-- _______________________________________________________________________ -->
8290 <a name="int_trap">'<tt>llvm.trap</tt>' Intrinsic</a>
8297 declare void @llvm.trap()
8301 <p>The '<tt>llvm.trap</tt>' intrinsic.</p>
8307 <p>This intrinsics is lowered to the target dependent trap instruction. If the
8308 target does not have a trap instruction, this intrinsic will be lowered to
8309 the call of the <tt>abort()</tt> function.</p>
8313 <!-- _______________________________________________________________________ -->
8315 <a name="int_stackprotector">'<tt>llvm.stackprotector</tt>' Intrinsic</a>
8322 declare void @llvm.stackprotector(i8* <guard>, i8** <slot>)
8326 <p>The <tt>llvm.stackprotector</tt> intrinsic takes the <tt>guard</tt> and
8327 stores it onto the stack at <tt>slot</tt>. The stack slot is adjusted to
8328 ensure that it is placed on the stack before local variables.</p>
8331 <p>The <tt>llvm.stackprotector</tt> intrinsic requires two pointer
8332 arguments. The first argument is the value loaded from the stack
8333 guard <tt>@__stack_chk_guard</tt>. The second variable is an <tt>alloca</tt>
8334 that has enough space to hold the value of the guard.</p>
8337 <p>This intrinsic causes the prologue/epilogue inserter to force the position of
8338 the <tt>AllocaInst</tt> stack slot to be before local variables on the
8339 stack. This is to ensure that if a local variable on the stack is
8340 overwritten, it will destroy the value of the guard. When the function exits,
8341 the guard on the stack is checked against the original guard. If they are
8342 different, then the program aborts by calling the <tt>__stack_chk_fail()</tt>
8347 <!-- _______________________________________________________________________ -->
8349 <a name="int_objectsize">'<tt>llvm.objectsize</tt>' Intrinsic</a>
8356 declare i32 @llvm.objectsize.i32(i8* <object>, i1 <type>)
8357 declare i64 @llvm.objectsize.i64(i8* <object>, i1 <type>)
8361 <p>The <tt>llvm.objectsize</tt> intrinsic is designed to provide information to
8362 the optimizers to determine at compile time whether a) an operation (like
8363 memcpy) will overflow a buffer that corresponds to an object, or b) that a
8364 runtime check for overflow isn't necessary. An object in this context means
8365 an allocation of a specific class, structure, array, or other object.</p>
8368 <p>The <tt>llvm.objectsize</tt> intrinsic takes two arguments. The first
8369 argument is a pointer to or into the <tt>object</tt>. The second argument
8370 is a boolean 0 or 1. This argument determines whether you want the
8371 maximum (0) or minimum (1) bytes remaining. This needs to be a literal 0 or
8372 1, variables are not allowed.</p>
8375 <p>The <tt>llvm.objectsize</tt> intrinsic is lowered to either a constant
8376 representing the size of the object concerned, or <tt>i32/i64 -1 or 0</tt>,
8377 depending on the <tt>type</tt> argument, if the size cannot be determined at
8381 <!-- _______________________________________________________________________ -->
8383 <a name="int_expect">'<tt>llvm.expect</tt>' Intrinsic</a>
8390 declare i32 @llvm.expect.i32(i32 <val>, i32 <expected_val>)
8391 declare i64 @llvm.expect.i64(i64 <val>, i64 <expected_val>)
8395 <p>The <tt>llvm.expect</tt> intrinsic provides information about expected (the
8396 most probable) value of <tt>val</tt>, which can be used by optimizers.</p>
8399 <p>The <tt>llvm.expect</tt> intrinsic takes two arguments. The first
8400 argument is a value. The second argument is an expected value, this needs to
8401 be a constant value, variables are not allowed.</p>
8404 <p>This intrinsic is lowered to the <tt>val</tt>.</p>
8410 <!-- *********************************************************************** -->
8413 <a href="http://jigsaw.w3.org/css-validator/check/referer"><img
8414 src="http://jigsaw.w3.org/css-validator/images/vcss-blue" alt="Valid CSS"></a>
8415 <a href="http://validator.w3.org/check/referer"><img
8416 src="http://www.w3.org/Icons/valid-html401-blue" alt="Valid HTML 4.01"></a>
8418 <a href="mailto:sabre@nondot.org">Chris Lattner</a><br>
8419 <a href="http://llvm.org/">The LLVM Compiler Infrastructure</a><br>
8420 Last modified: $Date$