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="_static/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_internal">'<tt>internal</tt>' Linkage</a></li>
29 <li><a href="#linkage_available_externally">'<tt>available_externally</tt>' Linkage</a></li>
30 <li><a href="#linkage_linkonce">'<tt>linkonce</tt>' Linkage</a></li>
31 <li><a href="#linkage_common">'<tt>common</tt>' Linkage</a></li>
32 <li><a href="#linkage_weak">'<tt>weak</tt>' Linkage</a></li>
33 <li><a href="#linkage_appending">'<tt>appending</tt>' Linkage</a></li>
34 <li><a href="#linkage_externweak">'<tt>extern_weak</tt>' Linkage</a></li>
35 <li><a href="#linkage_linkonce_odr">'<tt>linkonce_odr</tt>' Linkage</a></li>
36 <li><a href="#linkage_linkonce_odr_auto_hide">'<tt>linkonce_odr_auto_hide</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="#tbaa.struct">'<tt>tbaa.struct</tt>' Metadata</a></li>
107 <li><a href="#fpmath">'<tt>fpmath</tt>' Metadata</a></li>
108 <li><a href="#range">'<tt>range</tt>' Metadata</a></li>
113 <li><a href="#module_flags">Module Flags Metadata</a>
115 <li><a href="#objc_gc_flags">Objective-C Garbage Collection Module Flags Metadata</a></li>
118 <li><a href="#intrinsic_globals">Intrinsic Global Variables</a>
120 <li><a href="#intg_used">The '<tt>llvm.used</tt>' Global Variable</a></li>
121 <li><a href="#intg_compiler_used">The '<tt>llvm.compiler.used</tt>'
122 Global Variable</a></li>
123 <li><a href="#intg_global_ctors">The '<tt>llvm.global_ctors</tt>'
124 Global Variable</a></li>
125 <li><a href="#intg_global_dtors">The '<tt>llvm.global_dtors</tt>'
126 Global Variable</a></li>
129 <li><a href="#instref">Instruction Reference</a>
131 <li><a href="#terminators">Terminator Instructions</a>
133 <li><a href="#i_ret">'<tt>ret</tt>' Instruction</a></li>
134 <li><a href="#i_br">'<tt>br</tt>' Instruction</a></li>
135 <li><a href="#i_switch">'<tt>switch</tt>' Instruction</a></li>
136 <li><a href="#i_indirectbr">'<tt>indirectbr</tt>' Instruction</a></li>
137 <li><a href="#i_invoke">'<tt>invoke</tt>' Instruction</a></li>
138 <li><a href="#i_resume">'<tt>resume</tt>' Instruction</a></li>
139 <li><a href="#i_unreachable">'<tt>unreachable</tt>' Instruction</a></li>
142 <li><a href="#binaryops">Binary Operations</a>
144 <li><a href="#i_add">'<tt>add</tt>' Instruction</a></li>
145 <li><a href="#i_fadd">'<tt>fadd</tt>' Instruction</a></li>
146 <li><a href="#i_sub">'<tt>sub</tt>' Instruction</a></li>
147 <li><a href="#i_fsub">'<tt>fsub</tt>' Instruction</a></li>
148 <li><a href="#i_mul">'<tt>mul</tt>' Instruction</a></li>
149 <li><a href="#i_fmul">'<tt>fmul</tt>' Instruction</a></li>
150 <li><a href="#i_udiv">'<tt>udiv</tt>' Instruction</a></li>
151 <li><a href="#i_sdiv">'<tt>sdiv</tt>' Instruction</a></li>
152 <li><a href="#i_fdiv">'<tt>fdiv</tt>' Instruction</a></li>
153 <li><a href="#i_urem">'<tt>urem</tt>' Instruction</a></li>
154 <li><a href="#i_srem">'<tt>srem</tt>' Instruction</a></li>
155 <li><a href="#i_frem">'<tt>frem</tt>' Instruction</a></li>
158 <li><a href="#bitwiseops">Bitwise Binary Operations</a>
160 <li><a href="#i_shl">'<tt>shl</tt>' Instruction</a></li>
161 <li><a href="#i_lshr">'<tt>lshr</tt>' Instruction</a></li>
162 <li><a href="#i_ashr">'<tt>ashr</tt>' Instruction</a></li>
163 <li><a href="#i_and">'<tt>and</tt>' Instruction</a></li>
164 <li><a href="#i_or">'<tt>or</tt>' Instruction</a></li>
165 <li><a href="#i_xor">'<tt>xor</tt>' Instruction</a></li>
168 <li><a href="#vectorops">Vector Operations</a>
170 <li><a href="#i_extractelement">'<tt>extractelement</tt>' Instruction</a></li>
171 <li><a href="#i_insertelement">'<tt>insertelement</tt>' Instruction</a></li>
172 <li><a href="#i_shufflevector">'<tt>shufflevector</tt>' Instruction</a></li>
175 <li><a href="#aggregateops">Aggregate Operations</a>
177 <li><a href="#i_extractvalue">'<tt>extractvalue</tt>' Instruction</a></li>
178 <li><a href="#i_insertvalue">'<tt>insertvalue</tt>' Instruction</a></li>
181 <li><a href="#memoryops">Memory Access and Addressing Operations</a>
183 <li><a href="#i_alloca">'<tt>alloca</tt>' Instruction</a></li>
184 <li><a href="#i_load">'<tt>load</tt>' Instruction</a></li>
185 <li><a href="#i_store">'<tt>store</tt>' Instruction</a></li>
186 <li><a href="#i_fence">'<tt>fence</tt>' Instruction</a></li>
187 <li><a href="#i_cmpxchg">'<tt>cmpxchg</tt>' Instruction</a></li>
188 <li><a href="#i_atomicrmw">'<tt>atomicrmw</tt>' Instruction</a></li>
189 <li><a href="#i_getelementptr">'<tt>getelementptr</tt>' Instruction</a></li>
192 <li><a href="#convertops">Conversion Operations</a>
194 <li><a href="#i_trunc">'<tt>trunc .. to</tt>' Instruction</a></li>
195 <li><a href="#i_zext">'<tt>zext .. to</tt>' Instruction</a></li>
196 <li><a href="#i_sext">'<tt>sext .. to</tt>' Instruction</a></li>
197 <li><a href="#i_fptrunc">'<tt>fptrunc .. to</tt>' Instruction</a></li>
198 <li><a href="#i_fpext">'<tt>fpext .. to</tt>' Instruction</a></li>
199 <li><a href="#i_fptoui">'<tt>fptoui .. to</tt>' Instruction</a></li>
200 <li><a href="#i_fptosi">'<tt>fptosi .. to</tt>' Instruction</a></li>
201 <li><a href="#i_uitofp">'<tt>uitofp .. to</tt>' Instruction</a></li>
202 <li><a href="#i_sitofp">'<tt>sitofp .. to</tt>' Instruction</a></li>
203 <li><a href="#i_ptrtoint">'<tt>ptrtoint .. to</tt>' Instruction</a></li>
204 <li><a href="#i_inttoptr">'<tt>inttoptr .. to</tt>' Instruction</a></li>
205 <li><a href="#i_bitcast">'<tt>bitcast .. to</tt>' Instruction</a></li>
208 <li><a href="#otherops">Other Operations</a>
210 <li><a href="#i_icmp">'<tt>icmp</tt>' Instruction</a></li>
211 <li><a href="#i_fcmp">'<tt>fcmp</tt>' Instruction</a></li>
212 <li><a href="#i_phi">'<tt>phi</tt>' Instruction</a></li>
213 <li><a href="#i_select">'<tt>select</tt>' Instruction</a></li>
214 <li><a href="#i_call">'<tt>call</tt>' Instruction</a></li>
215 <li><a href="#i_va_arg">'<tt>va_arg</tt>' Instruction</a></li>
216 <li><a href="#i_landingpad">'<tt>landingpad</tt>' Instruction</a></li>
221 <li><a href="#intrinsics">Intrinsic Functions</a>
223 <li><a href="#int_varargs">Variable Argument Handling Intrinsics</a>
225 <li><a href="#int_va_start">'<tt>llvm.va_start</tt>' Intrinsic</a></li>
226 <li><a href="#int_va_end">'<tt>llvm.va_end</tt>' Intrinsic</a></li>
227 <li><a href="#int_va_copy">'<tt>llvm.va_copy</tt>' Intrinsic</a></li>
230 <li><a href="#int_gc">Accurate Garbage Collection Intrinsics</a>
232 <li><a href="#int_gcroot">'<tt>llvm.gcroot</tt>' Intrinsic</a></li>
233 <li><a href="#int_gcread">'<tt>llvm.gcread</tt>' Intrinsic</a></li>
234 <li><a href="#int_gcwrite">'<tt>llvm.gcwrite</tt>' Intrinsic</a></li>
237 <li><a href="#int_codegen">Code Generator Intrinsics</a>
239 <li><a href="#int_returnaddress">'<tt>llvm.returnaddress</tt>' Intrinsic</a></li>
240 <li><a href="#int_frameaddress">'<tt>llvm.frameaddress</tt>' Intrinsic</a></li>
241 <li><a href="#int_stacksave">'<tt>llvm.stacksave</tt>' Intrinsic</a></li>
242 <li><a href="#int_stackrestore">'<tt>llvm.stackrestore</tt>' Intrinsic</a></li>
243 <li><a href="#int_prefetch">'<tt>llvm.prefetch</tt>' Intrinsic</a></li>
244 <li><a href="#int_pcmarker">'<tt>llvm.pcmarker</tt>' Intrinsic</a></li>
245 <li><a href="#int_readcyclecounter">'<tt>llvm.readcyclecounter</tt>' Intrinsic</a></li>
248 <li><a href="#int_libc">Standard C Library Intrinsics</a>
250 <li><a href="#int_memcpy">'<tt>llvm.memcpy.*</tt>' Intrinsic</a></li>
251 <li><a href="#int_memmove">'<tt>llvm.memmove.*</tt>' Intrinsic</a></li>
252 <li><a href="#int_memset">'<tt>llvm.memset.*</tt>' Intrinsic</a></li>
253 <li><a href="#int_sqrt">'<tt>llvm.sqrt.*</tt>' Intrinsic</a></li>
254 <li><a href="#int_powi">'<tt>llvm.powi.*</tt>' Intrinsic</a></li>
255 <li><a href="#int_sin">'<tt>llvm.sin.*</tt>' Intrinsic</a></li>
256 <li><a href="#int_cos">'<tt>llvm.cos.*</tt>' Intrinsic</a></li>
257 <li><a href="#int_pow">'<tt>llvm.pow.*</tt>' Intrinsic</a></li>
258 <li><a href="#int_exp">'<tt>llvm.exp.*</tt>' Intrinsic</a></li>
259 <li><a href="#int_log">'<tt>llvm.log.*</tt>' Intrinsic</a></li>
260 <li><a href="#int_fma">'<tt>llvm.fma.*</tt>' Intrinsic</a></li>
261 <li><a href="#int_fabs">'<tt>llvm.fabs.*</tt>' Intrinsic</a></li>
262 <li><a href="#int_floor">'<tt>llvm.floor.*</tt>' Intrinsic</a></li>
265 <li><a href="#int_manip">Bit Manipulation Intrinsics</a>
267 <li><a href="#int_bswap">'<tt>llvm.bswap.*</tt>' Intrinsics</a></li>
268 <li><a href="#int_ctpop">'<tt>llvm.ctpop.*</tt>' Intrinsic </a></li>
269 <li><a href="#int_ctlz">'<tt>llvm.ctlz.*</tt>' Intrinsic </a></li>
270 <li><a href="#int_cttz">'<tt>llvm.cttz.*</tt>' Intrinsic </a></li>
273 <li><a href="#int_overflow">Arithmetic with Overflow Intrinsics</a>
275 <li><a href="#int_sadd_overflow">'<tt>llvm.sadd.with.overflow.*</tt> Intrinsics</a></li>
276 <li><a href="#int_uadd_overflow">'<tt>llvm.uadd.with.overflow.*</tt> Intrinsics</a></li>
277 <li><a href="#int_ssub_overflow">'<tt>llvm.ssub.with.overflow.*</tt> Intrinsics</a></li>
278 <li><a href="#int_usub_overflow">'<tt>llvm.usub.with.overflow.*</tt> Intrinsics</a></li>
279 <li><a href="#int_smul_overflow">'<tt>llvm.smul.with.overflow.*</tt> Intrinsics</a></li>
280 <li><a href="#int_umul_overflow">'<tt>llvm.umul.with.overflow.*</tt> Intrinsics</a></li>
283 <li><a href="#spec_arithmetic">Specialised Arithmetic Intrinsics</a>
285 <li><a href="#fmuladd">'<tt>llvm.fmuladd</tt> Intrinsic</a></li>
288 <li><a href="#int_fp16">Half Precision Floating Point Intrinsics</a>
290 <li><a href="#int_convert_to_fp16">'<tt>llvm.convert.to.fp16</tt>' Intrinsic</a></li>
291 <li><a href="#int_convert_from_fp16">'<tt>llvm.convert.from.fp16</tt>' Intrinsic</a></li>
294 <li><a href="#int_debugger">Debugger intrinsics</a></li>
295 <li><a href="#int_eh">Exception Handling intrinsics</a></li>
296 <li><a href="#int_trampoline">Trampoline Intrinsics</a>
298 <li><a href="#int_it">'<tt>llvm.init.trampoline</tt>' Intrinsic</a></li>
299 <li><a href="#int_at">'<tt>llvm.adjust.trampoline</tt>' Intrinsic</a></li>
302 <li><a href="#int_memorymarkers">Memory Use Markers</a>
304 <li><a href="#int_lifetime_start">'<tt>llvm.lifetime.start</tt>' Intrinsic</a></li>
305 <li><a href="#int_lifetime_end">'<tt>llvm.lifetime.end</tt>' Intrinsic</a></li>
306 <li><a href="#int_invariant_start">'<tt>llvm.invariant.start</tt>' Intrinsic</a></li>
307 <li><a href="#int_invariant_end">'<tt>llvm.invariant.end</tt>' Intrinsic</a></li>
310 <li><a href="#int_general">General intrinsics</a>
312 <li><a href="#int_var_annotation">
313 '<tt>llvm.var.annotation</tt>' Intrinsic</a></li>
314 <li><a href="#int_annotation">
315 '<tt>llvm.annotation.*</tt>' Intrinsic</a></li>
316 <li><a href="#int_trap">
317 '<tt>llvm.trap</tt>' Intrinsic</a></li>
318 <li><a href="#int_debugtrap">
319 '<tt>llvm.debugtrap</tt>' Intrinsic</a></li>
320 <li><a href="#int_stackprotector">
321 '<tt>llvm.stackprotector</tt>' Intrinsic</a></li>
322 <li><a href="#int_objectsize">
323 '<tt>llvm.objectsize</tt>' Intrinsic</a></li>
324 <li><a href="#int_expect">
325 '<tt>llvm.expect</tt>' Intrinsic</a></li>
326 <li><a href="#int_donothing">
327 '<tt>llvm.donothing</tt>' Intrinsic</a></li>
334 <div class="doc_author">
335 <p>Written by <a href="mailto:sabre@nondot.org">Chris Lattner</a>
336 and <a href="mailto:vadve@cs.uiuc.edu">Vikram Adve</a></p>
339 <!-- *********************************************************************** -->
340 <h2><a name="abstract">Abstract</a></h2>
341 <!-- *********************************************************************** -->
345 <p>This document is a reference manual for the LLVM assembly language. LLVM is
346 a Static Single Assignment (SSA) based representation that provides type
347 safety, low-level operations, flexibility, and the capability of representing
348 'all' high-level languages cleanly. It is the common code representation
349 used throughout all phases of the LLVM compilation strategy.</p>
353 <!-- *********************************************************************** -->
354 <h2><a name="introduction">Introduction</a></h2>
355 <!-- *********************************************************************** -->
359 <p>The LLVM code representation is designed to be used in three different forms:
360 as an in-memory compiler IR, as an on-disk bitcode representation (suitable
361 for fast loading by a Just-In-Time compiler), and as a human readable
362 assembly language representation. This allows LLVM to provide a powerful
363 intermediate representation for efficient compiler transformations and
364 analysis, while providing a natural means to debug and visualize the
365 transformations. The three different forms of LLVM are all equivalent. This
366 document describes the human readable representation and notation.</p>
368 <p>The LLVM representation aims to be light-weight and low-level while being
369 expressive, typed, and extensible at the same time. It aims to be a
370 "universal IR" of sorts, by being at a low enough level that high-level ideas
371 may be cleanly mapped to it (similar to how microprocessors are "universal
372 IR's", allowing many source languages to be mapped to them). By providing
373 type information, LLVM can be used as the target of optimizations: for
374 example, through pointer analysis, it can be proven that a C automatic
375 variable is never accessed outside of the current function, allowing it to
376 be promoted to a simple SSA value instead of a memory location.</p>
378 <!-- _______________________________________________________________________ -->
380 <a name="wellformed">Well-Formedness</a>
385 <p>It is important to note that this document describes 'well formed' LLVM
386 assembly language. There is a difference between what the parser accepts and
387 what is considered 'well formed'. For example, the following instruction is
388 syntactically okay, but not well formed:</p>
390 <pre class="doc_code">
391 %x = <a href="#i_add">add</a> i32 1, %x
394 <p>because the definition of <tt>%x</tt> does not dominate all of its uses. The
395 LLVM infrastructure provides a verification pass that may be used to verify
396 that an LLVM module is well formed. This pass is automatically run by the
397 parser after parsing input assembly and by the optimizer before it outputs
398 bitcode. The violations pointed out by the verifier pass indicate bugs in
399 transformation passes or input to the parser.</p>
405 <!-- Describe the typesetting conventions here. -->
407 <!-- *********************************************************************** -->
408 <h2><a name="identifiers">Identifiers</a></h2>
409 <!-- *********************************************************************** -->
413 <p>LLVM identifiers come in two basic types: global and local. Global
414 identifiers (functions, global variables) begin with the <tt>'@'</tt>
415 character. Local identifiers (register names, types) begin with
416 the <tt>'%'</tt> character. Additionally, there are three different formats
417 for identifiers, for different purposes:</p>
420 <li>Named values are represented as a string of characters with their prefix.
421 For example, <tt>%foo</tt>, <tt>@DivisionByZero</tt>,
422 <tt>%a.really.long.identifier</tt>. The actual regular expression used is
423 '<tt>[%@][a-zA-Z$._][a-zA-Z$._0-9]*</tt>'. Identifiers which require
424 other characters in their names can be surrounded with quotes. Special
425 characters may be escaped using <tt>"\xx"</tt> where <tt>xx</tt> is the
426 ASCII code for the character in hexadecimal. In this way, any character
427 can be used in a name value, even quotes themselves.</li>
429 <li>Unnamed values are represented as an unsigned numeric value with their
430 prefix. For example, <tt>%12</tt>, <tt>@2</tt>, <tt>%44</tt>.</li>
432 <li>Constants, which are described in a <a href="#constants">section about
433 constants</a>, below.</li>
436 <p>LLVM requires that values start with a prefix for two reasons: Compilers
437 don't need to worry about name clashes with reserved words, and the set of
438 reserved words may be expanded in the future without penalty. Additionally,
439 unnamed identifiers allow a compiler to quickly come up with a temporary
440 variable without having to avoid symbol table conflicts.</p>
442 <p>Reserved words in LLVM are very similar to reserved words in other
443 languages. There are keywords for different opcodes
444 ('<tt><a href="#i_add">add</a></tt>',
445 '<tt><a href="#i_bitcast">bitcast</a></tt>',
446 '<tt><a href="#i_ret">ret</a></tt>', etc...), for primitive type names
447 ('<tt><a href="#t_void">void</a></tt>',
448 '<tt><a href="#t_primitive">i32</a></tt>', etc...), and others. These
449 reserved words cannot conflict with variable names, because none of them
450 start with a prefix character (<tt>'%'</tt> or <tt>'@'</tt>).</p>
452 <p>Here is an example of LLVM code to multiply the integer variable
453 '<tt>%X</tt>' by 8:</p>
457 <pre class="doc_code">
458 %result = <a href="#i_mul">mul</a> i32 %X, 8
461 <p>After strength reduction:</p>
463 <pre class="doc_code">
464 %result = <a href="#i_shl">shl</a> i32 %X, i8 3
467 <p>And the hard way:</p>
469 <pre class="doc_code">
470 %0 = <a href="#i_add">add</a> i32 %X, %X <i>; yields {i32}:%0</i>
471 %1 = <a href="#i_add">add</a> i32 %0, %0 <i>; yields {i32}:%1</i>
472 %result = <a href="#i_add">add</a> i32 %1, %1
475 <p>This last way of multiplying <tt>%X</tt> by 8 illustrates several important
476 lexical features of LLVM:</p>
479 <li>Comments are delimited with a '<tt>;</tt>' and go until the end of
482 <li>Unnamed temporaries are created when the result of a computation is not
483 assigned to a named value.</li>
485 <li>Unnamed temporaries are numbered sequentially</li>
488 <p>It also shows a convention that we follow in this document. When
489 demonstrating instructions, we will follow an instruction with a comment that
490 defines the type and name of value produced. Comments are shown in italic
495 <!-- *********************************************************************** -->
496 <h2><a name="highlevel">High Level Structure</a></h2>
497 <!-- *********************************************************************** -->
499 <!-- ======================================================================= -->
501 <a name="modulestructure">Module Structure</a>
506 <p>LLVM programs are composed of <tt>Module</tt>s, each of which is a
507 translation unit of the input programs. Each module consists of functions,
508 global variables, and symbol table entries. Modules may be combined together
509 with the LLVM linker, which merges function (and global variable)
510 definitions, resolves forward declarations, and merges symbol table
511 entries. Here is an example of the "hello world" module:</p>
513 <pre class="doc_code">
514 <i>; Declare the string constant as a global constant.</i>
515 <a href="#identifiers">@.str</a> = <a href="#linkage_private">private</a> <a href="#globalvars">unnamed_addr</a> <a href="#globalvars">constant</a> <a href="#t_array">[13 x i8]</a> c"hello world\0A\00"
517 <i>; External declaration of the puts function</i>
518 <a href="#functionstructure">declare</a> i32 @puts(i8* <a href="#nocapture">nocapture</a>) <a href="#fnattrs">nounwind</a>
520 <i>; Definition of main function</i>
521 define i32 @main() { <i>; i32()* </i>
522 <i>; Convert [13 x i8]* to i8 *...</i>
523 %cast210 = <a href="#i_getelementptr">getelementptr</a> [13 x i8]* @.str, i64 0, i64 0
525 <i>; Call puts function to write out the string to stdout.</i>
526 <a href="#i_call">call</a> i32 @puts(i8* %cast210)
527 <a href="#i_ret">ret</a> i32 0
530 <i>; Named metadata</i>
531 !1 = metadata !{i32 42}
535 <p>This example is made up of a <a href="#globalvars">global variable</a> named
536 "<tt>.str</tt>", an external declaration of the "<tt>puts</tt>" function,
537 a <a href="#functionstructure">function definition</a> for
538 "<tt>main</tt>" and <a href="#namedmetadatastructure">named metadata</a>
541 <p>In general, a module is made up of a list of global values (where both
542 functions and global variables are global values). Global values are
543 represented by a pointer to a memory location (in this case, a pointer to an
544 array of char, and a pointer to a function), and have one of the
545 following <a href="#linkage">linkage types</a>.</p>
549 <!-- ======================================================================= -->
551 <a name="linkage">Linkage Types</a>
556 <p>All Global Variables and Functions have one of the following types of
560 <dt><tt><b><a name="linkage_private">private</a></b></tt></dt>
561 <dd>Global values with "<tt>private</tt>" linkage are only directly accessible
562 by objects in the current module. In particular, linking code into a
563 module with an private global value may cause the private to be renamed as
564 necessary to avoid collisions. Because the symbol is private to the
565 module, all references can be updated. This doesn't show up in any symbol
566 table in the object file.</dd>
568 <dt><tt><b><a name="linkage_linker_private">linker_private</a></b></tt></dt>
569 <dd>Similar to <tt>private</tt>, but the symbol is passed through the
570 assembler and evaluated by the linker. Unlike normal strong symbols, they
571 are removed by the linker from the final linked image (executable or
572 dynamic library).</dd>
574 <dt><tt><b><a name="linkage_linker_private_weak">linker_private_weak</a></b></tt></dt>
575 <dd>Similar to "<tt>linker_private</tt>", but the symbol is weak. Note that
576 <tt>linker_private_weak</tt> symbols are subject to coalescing by the
577 linker. The symbols are removed by the linker from the final linked image
578 (executable or dynamic library).</dd>
580 <dt><tt><b><a name="linkage_internal">internal</a></b></tt></dt>
581 <dd>Similar to private, but the value shows as a local symbol
582 (<tt>STB_LOCAL</tt> in the case of ELF) in the object file. This
583 corresponds to the notion of the '<tt>static</tt>' keyword in C.</dd>
585 <dt><tt><b><a name="linkage_available_externally">available_externally</a></b></tt></dt>
586 <dd>Globals with "<tt>available_externally</tt>" linkage are never emitted
587 into the object file corresponding to the LLVM module. They exist to
588 allow inlining and other optimizations to take place given knowledge of
589 the definition of the global, which is known to be somewhere outside the
590 module. Globals with <tt>available_externally</tt> linkage are allowed to
591 be discarded at will, and are otherwise the same as <tt>linkonce_odr</tt>.
592 This linkage type is only allowed on definitions, not declarations.</dd>
594 <dt><tt><b><a name="linkage_linkonce">linkonce</a></b></tt></dt>
595 <dd>Globals with "<tt>linkonce</tt>" linkage are merged with other globals of
596 the same name when linkage occurs. This can be used to implement
597 some forms of inline functions, templates, or other code which must be
598 generated in each translation unit that uses it, but where the body may
599 be overridden with a more definitive definition later. Unreferenced
600 <tt>linkonce</tt> globals are allowed to be discarded. Note that
601 <tt>linkonce</tt> linkage does not actually allow the optimizer to
602 inline the body of this function into callers because it doesn't know if
603 this definition of the function is the definitive definition within the
604 program or whether it will be overridden by a stronger definition.
605 To enable inlining and other optimizations, use "<tt>linkonce_odr</tt>"
608 <dt><tt><b><a name="linkage_weak">weak</a></b></tt></dt>
609 <dd>"<tt>weak</tt>" linkage has the same merging semantics as
610 <tt>linkonce</tt> linkage, except that unreferenced globals with
611 <tt>weak</tt> linkage may not be discarded. This is used for globals that
612 are declared "weak" in C source code.</dd>
614 <dt><tt><b><a name="linkage_common">common</a></b></tt></dt>
615 <dd>"<tt>common</tt>" linkage is most similar to "<tt>weak</tt>" linkage, but
616 they are used for tentative definitions in C, such as "<tt>int X;</tt>" at
618 Symbols with "<tt>common</tt>" linkage are merged in the same way as
619 <tt>weak symbols</tt>, and they may not be deleted if unreferenced.
620 <tt>common</tt> symbols may not have an explicit section,
621 must have a zero initializer, and may not be marked '<a
622 href="#globalvars"><tt>constant</tt></a>'. Functions and aliases may not
623 have common linkage.</dd>
626 <dt><tt><b><a name="linkage_appending">appending</a></b></tt></dt>
627 <dd>"<tt>appending</tt>" linkage may only be applied to global variables of
628 pointer to array type. When two global variables with appending linkage
629 are linked together, the two global arrays are appended together. This is
630 the LLVM, typesafe, equivalent of having the system linker append together
631 "sections" with identical names when .o files are linked.</dd>
633 <dt><tt><b><a name="linkage_externweak">extern_weak</a></b></tt></dt>
634 <dd>The semantics of this linkage follow the ELF object file model: the symbol
635 is weak until linked, if not linked, the symbol becomes null instead of
636 being an undefined reference.</dd>
638 <dt><tt><b><a name="linkage_linkonce_odr">linkonce_odr</a></b></tt></dt>
639 <dt><tt><b><a name="linkage_weak_odr">weak_odr</a></b></tt></dt>
640 <dd>Some languages allow differing globals to be merged, such as two functions
641 with different semantics. Other languages, such as <tt>C++</tt>, ensure
642 that only equivalent globals are ever merged (the "one definition rule"
643 — "ODR"). Such languages can use the <tt>linkonce_odr</tt>
644 and <tt>weak_odr</tt> linkage types to indicate that the global will only
645 be merged with equivalent globals. These linkage types are otherwise the
646 same as their non-<tt>odr</tt> versions.</dd>
648 <dt><tt><b><a name="linkage_linkonce_odr_auto_hide">linkonce_odr_auto_hide</a></b></tt></dt>
649 <dd>Similar to "<tt>linkonce_odr</tt>", but nothing in the translation unit
650 takes the address of this definition. For instance, functions that had an
651 inline definition, but the compiler decided not to inline it.
652 <tt>linkonce_odr_auto_hide</tt> may have only <tt>default</tt> visibility.
653 The symbols are removed by the linker from the final linked image
654 (executable or dynamic library).</dd>
656 <dt><tt><b><a name="linkage_external">external</a></b></tt></dt>
657 <dd>If none of the above identifiers are used, the global is externally
658 visible, meaning that it participates in linkage and can be used to
659 resolve external symbol references.</dd>
662 <p>The next two types of linkage are targeted for Microsoft Windows platform
663 only. They are designed to support importing (exporting) symbols from (to)
664 DLLs (Dynamic Link Libraries).</p>
667 <dt><tt><b><a name="linkage_dllimport">dllimport</a></b></tt></dt>
668 <dd>"<tt>dllimport</tt>" linkage causes the compiler to reference a function
669 or variable via a global pointer to a pointer that is set up by the DLL
670 exporting the symbol. On Microsoft Windows targets, the pointer name is
671 formed by combining <code>__imp_</code> and the function or variable
674 <dt><tt><b><a name="linkage_dllexport">dllexport</a></b></tt></dt>
675 <dd>"<tt>dllexport</tt>" linkage causes the compiler to provide a global
676 pointer to a pointer in a DLL, so that it can be referenced with the
677 <tt>dllimport</tt> attribute. On Microsoft Windows targets, the pointer
678 name is formed by combining <code>__imp_</code> and the function or
682 <p>For example, since the "<tt>.LC0</tt>" variable is defined to be internal, if
683 another module defined a "<tt>.LC0</tt>" variable and was linked with this
684 one, one of the two would be renamed, preventing a collision. Since
685 "<tt>main</tt>" and "<tt>puts</tt>" are external (i.e., lacking any linkage
686 declarations), they are accessible outside of the current module.</p>
688 <p>It is illegal for a function <i>declaration</i> to have any linkage type
689 other than <tt>external</tt>, <tt>dllimport</tt>
690 or <tt>extern_weak</tt>.</p>
692 <p>Aliases can have only <tt>external</tt>, <tt>internal</tt>, <tt>weak</tt>
693 or <tt>weak_odr</tt> linkages.</p>
697 <!-- ======================================================================= -->
699 <a name="callingconv">Calling Conventions</a>
704 <p>LLVM <a href="#functionstructure">functions</a>, <a href="#i_call">calls</a>
705 and <a href="#i_invoke">invokes</a> can all have an optional calling
706 convention specified for the call. The calling convention of any pair of
707 dynamic caller/callee must match, or the behavior of the program is
708 undefined. The following calling conventions are supported by LLVM, and more
709 may be added in the future:</p>
712 <dt><b>"<tt>ccc</tt>" - The C calling convention</b>:</dt>
713 <dd>This calling convention (the default if no other calling convention is
714 specified) matches the target C calling conventions. This calling
715 convention supports varargs function calls and tolerates some mismatch in
716 the declared prototype and implemented declaration of the function (as
719 <dt><b>"<tt>fastcc</tt>" - The fast calling convention</b>:</dt>
720 <dd>This calling convention attempts to make calls as fast as possible
721 (e.g. by passing things in registers). This calling convention allows the
722 target to use whatever tricks it wants to produce fast code for the
723 target, without having to conform to an externally specified ABI
724 (Application Binary Interface).
725 <a href="CodeGenerator.html#tailcallopt">Tail calls can only be optimized
726 when this or the GHC convention is used.</a> This calling convention
727 does not support varargs and requires the prototype of all callees to
728 exactly match the prototype of the function definition.</dd>
730 <dt><b>"<tt>coldcc</tt>" - The cold calling convention</b>:</dt>
731 <dd>This calling convention attempts to make code in the caller as efficient
732 as possible under the assumption that the call is not commonly executed.
733 As such, these calls often preserve all registers so that the call does
734 not break any live ranges in the caller side. This calling convention
735 does not support varargs and requires the prototype of all callees to
736 exactly match the prototype of the function definition.</dd>
738 <dt><b>"<tt>cc <em>10</em></tt>" - GHC convention</b>:</dt>
739 <dd>This calling convention has been implemented specifically for use by the
740 <a href="http://www.haskell.org/ghc">Glasgow Haskell Compiler (GHC)</a>.
741 It passes everything in registers, going to extremes to achieve this by
742 disabling callee save registers. This calling convention should not be
743 used lightly but only for specific situations such as an alternative to
744 the <em>register pinning</em> performance technique often used when
745 implementing functional programming languages.At the moment only X86
746 supports this convention and it has the following limitations:
748 <li>On <em>X86-32</em> only supports up to 4 bit type parameters. No
749 floating point types are supported.</li>
750 <li>On <em>X86-64</em> only supports up to 10 bit type parameters and
751 6 floating point parameters.</li>
753 This calling convention supports
754 <a href="CodeGenerator.html#tailcallopt">tail call optimization</a> but
755 requires both the caller and callee are using it.
758 <dt><b>"<tt>cc <<em>n</em>></tt>" - Numbered convention</b>:</dt>
759 <dd>Any calling convention may be specified by number, allowing
760 target-specific calling conventions to be used. Target specific calling
761 conventions start at 64.</dd>
764 <p>More calling conventions can be added/defined on an as-needed basis, to
765 support Pascal conventions or any other well-known target-independent
770 <!-- ======================================================================= -->
772 <a name="visibility">Visibility Styles</a>
777 <p>All Global Variables and Functions have one of the following visibility
781 <dt><b>"<tt>default</tt>" - Default style</b>:</dt>
782 <dd>On targets that use the ELF object file format, default visibility means
783 that the declaration is visible to other modules and, in shared libraries,
784 means that the declared entity may be overridden. On Darwin, default
785 visibility means that the declaration is visible to other modules. Default
786 visibility corresponds to "external linkage" in the language.</dd>
788 <dt><b>"<tt>hidden</tt>" - Hidden style</b>:</dt>
789 <dd>Two declarations of an object with hidden visibility refer to the same
790 object if they are in the same shared object. Usually, hidden visibility
791 indicates that the symbol will not be placed into the dynamic symbol
792 table, so no other module (executable or shared library) can reference it
795 <dt><b>"<tt>protected</tt>" - Protected style</b>:</dt>
796 <dd>On ELF, protected visibility indicates that the symbol will be placed in
797 the dynamic symbol table, but that references within the defining module
798 will bind to the local symbol. That is, the symbol cannot be overridden by
804 <!-- ======================================================================= -->
806 <a name="namedtypes">Named Types</a>
811 <p>LLVM IR allows you to specify name aliases for certain types. This can make
812 it easier to read the IR and make the IR more condensed (particularly when
813 recursive types are involved). An example of a name specification is:</p>
815 <pre class="doc_code">
816 %mytype = type { %mytype*, i32 }
819 <p>You may give a name to any <a href="#typesystem">type</a> except
820 "<a href="#t_void">void</a>". Type name aliases may be used anywhere a type
821 is expected with the syntax "%mytype".</p>
823 <p>Note that type names are aliases for the structural type that they indicate,
824 and that you can therefore specify multiple names for the same type. This
825 often leads to confusing behavior when dumping out a .ll file. Since LLVM IR
826 uses structural typing, the name is not part of the type. When printing out
827 LLVM IR, the printer will pick <em>one name</em> to render all types of a
828 particular shape. This means that if you have code where two different
829 source types end up having the same LLVM type, that the dumper will sometimes
830 print the "wrong" or unexpected type. This is an important design point and
831 isn't going to change.</p>
835 <!-- ======================================================================= -->
837 <a name="globalvars">Global Variables</a>
842 <p>Global variables define regions of memory allocated at compilation time
843 instead of run-time. Global variables may optionally be initialized, may
844 have an explicit section to be placed in, and may have an optional explicit
845 alignment specified.</p>
847 <p>A variable may be defined as <tt>thread_local</tt>, which
848 means that it will not be shared by threads (each thread will have a
849 separated copy of the variable). Not all targets support thread-local
850 variables. Optionally, a TLS model may be specified:</p>
853 <dt><b><tt>localdynamic</tt></b>:</dt>
854 <dd>For variables that are only used within the current shared library.</dd>
856 <dt><b><tt>initialexec</tt></b>:</dt>
857 <dd>For variables in modules that will not be loaded dynamically.</dd>
859 <dt><b><tt>localexec</tt></b>:</dt>
860 <dd>For variables defined in the executable and only used within it.</dd>
863 <p>The models correspond to the ELF TLS models; see
864 <a href="http://people.redhat.com/drepper/tls.pdf">ELF
865 Handling For Thread-Local Storage</a> for more information on under which
866 circumstances the different models may be used. The target may choose a
867 different TLS model if the specified model is not supported, or if a better
868 choice of model can be made.</p>
870 <p>A variable may be defined as a global
871 "constant," which indicates that the contents of the variable
872 will <b>never</b> be modified (enabling better optimization, allowing the
873 global data to be placed in the read-only section of an executable, etc).
874 Note that variables that need runtime initialization cannot be marked
875 "constant" as there is a store to the variable.</p>
877 <p>LLVM explicitly allows <em>declarations</em> of global variables to be marked
878 constant, even if the final definition of the global is not. This capability
879 can be used to enable slightly better optimization of the program, but
880 requires the language definition to guarantee that optimizations based on the
881 'constantness' are valid for the translation units that do not include the
884 <p>As SSA values, global variables define pointer values that are in scope
885 (i.e. they dominate) all basic blocks in the program. Global variables
886 always define a pointer to their "content" type because they describe a
887 region of memory, and all memory objects in LLVM are accessed through
890 <p>Global variables can be marked with <tt>unnamed_addr</tt> which indicates
891 that the address is not significant, only the content. Constants marked
892 like this can be merged with other constants if they have the same
893 initializer. Note that a constant with significant address <em>can</em>
894 be merged with a <tt>unnamed_addr</tt> constant, the result being a
895 constant whose address is significant.</p>
897 <p>A global variable may be declared to reside in a target-specific numbered
898 address space. For targets that support them, address spaces may affect how
899 optimizations are performed and/or what target instructions are used to
900 access the variable. The default address space is zero. The address space
901 qualifier must precede any other attributes.</p>
903 <p>LLVM allows an explicit section to be specified for globals. If the target
904 supports it, it will emit globals to the section specified.</p>
906 <p>An explicit alignment may be specified for a global, which must be a power
907 of 2. If not present, or if the alignment is set to zero, the alignment of
908 the global is set by the target to whatever it feels convenient. If an
909 explicit alignment is specified, the global is forced to have exactly that
910 alignment. Targets and optimizers are not allowed to over-align the global
911 if the global has an assigned section. In this case, the extra alignment
912 could be observable: for example, code could assume that the globals are
913 densely packed in their section and try to iterate over them as an array,
914 alignment padding would break this iteration.</p>
916 <p>For example, the following defines a global in a numbered address space with
917 an initializer, section, and alignment:</p>
919 <pre class="doc_code">
920 @G = addrspace(5) constant float 1.0, section "foo", align 4
923 <p>The following example defines a thread-local global with
924 the <tt>initialexec</tt> TLS model:</p>
926 <pre class="doc_code">
927 @G = thread_local(initialexec) global i32 0, align 4
933 <!-- ======================================================================= -->
935 <a name="functionstructure">Functions</a>
940 <p>LLVM function definitions consist of the "<tt>define</tt>" keyword, an
941 optional <a href="#linkage">linkage type</a>, an optional
942 <a href="#visibility">visibility style</a>, an optional
943 <a href="#callingconv">calling convention</a>,
944 an optional <tt>unnamed_addr</tt> attribute, a return type, an optional
945 <a href="#paramattrs">parameter attribute</a> for the return type, a function
946 name, a (possibly empty) argument list (each with optional
947 <a href="#paramattrs">parameter attributes</a>), optional
948 <a href="#fnattrs">function attributes</a>, an optional section, an optional
949 alignment, an optional <a href="#gc">garbage collector name</a>, an opening
950 curly brace, a list of basic blocks, and a closing curly brace.</p>
952 <p>LLVM function declarations consist of the "<tt>declare</tt>" keyword, an
953 optional <a href="#linkage">linkage type</a>, an optional
954 <a href="#visibility">visibility style</a>, an optional
955 <a href="#callingconv">calling convention</a>,
956 an optional <tt>unnamed_addr</tt> attribute, a return type, an optional
957 <a href="#paramattrs">parameter attribute</a> for the return type, a function
958 name, a possibly empty list of arguments, an optional alignment, and an
959 optional <a href="#gc">garbage collector name</a>.</p>
961 <p>A function definition contains a list of basic blocks, forming the CFG
962 (Control Flow Graph) for the function. Each basic block may optionally start
963 with a label (giving the basic block a symbol table entry), contains a list
964 of instructions, and ends with a <a href="#terminators">terminator</a>
965 instruction (such as a branch or function return).</p>
967 <p>The first basic block in a function is special in two ways: it is immediately
968 executed on entrance to the function, and it is not allowed to have
969 predecessor basic blocks (i.e. there can not be any branches to the entry
970 block of a function). Because the block can have no predecessors, it also
971 cannot have any <a href="#i_phi">PHI nodes</a>.</p>
973 <p>LLVM allows an explicit section to be specified for functions. If the target
974 supports it, it will emit functions to the section specified.</p>
976 <p>An explicit alignment may be specified for a function. If not present, or if
977 the alignment is set to zero, the alignment of the function is set by the
978 target to whatever it feels convenient. If an explicit alignment is
979 specified, the function is forced to have at least that much alignment. All
980 alignments must be a power of 2.</p>
982 <p>If the <tt>unnamed_addr</tt> attribute is given, the address is know to not
983 be significant and two identical functions can be merged.</p>
986 <pre class="doc_code">
987 define [<a href="#linkage">linkage</a>] [<a href="#visibility">visibility</a>]
988 [<a href="#callingconv">cconv</a>] [<a href="#paramattrs">ret attrs</a>]
989 <ResultType> @<FunctionName> ([argument list])
990 [<a href="#fnattrs">fn Attrs</a>] [section "name"] [align N]
991 [<a href="#gc">gc</a>] { ... }
996 <!-- ======================================================================= -->
998 <a name="aliasstructure">Aliases</a>
1003 <p>Aliases act as "second name" for the aliasee value (which can be either
1004 function, global variable, another alias or bitcast of global value). Aliases
1005 may have an optional <a href="#linkage">linkage type</a>, and an
1006 optional <a href="#visibility">visibility style</a>.</p>
1009 <pre class="doc_code">
1010 @<Name> = alias [Linkage] [Visibility] <AliaseeTy> @<Aliasee>
1015 <!-- ======================================================================= -->
1017 <a name="namedmetadatastructure">Named Metadata</a>
1022 <p>Named metadata is a collection of metadata. <a href="#metadata">Metadata
1023 nodes</a> (but not metadata strings) are the only valid operands for
1024 a named metadata.</p>
1027 <pre class="doc_code">
1028 ; Some unnamed metadata nodes, which are referenced by the named metadata.
1029 !0 = metadata !{metadata !"zero"}
1030 !1 = metadata !{metadata !"one"}
1031 !2 = metadata !{metadata !"two"}
1033 !name = !{!0, !1, !2}
1038 <!-- ======================================================================= -->
1040 <a name="paramattrs">Parameter Attributes</a>
1045 <p>The return type and each parameter of a function type may have a set of
1046 <i>parameter attributes</i> associated with them. Parameter attributes are
1047 used to communicate additional information about the result or parameters of
1048 a function. Parameter attributes are considered to be part of the function,
1049 not of the function type, so functions with different parameter attributes
1050 can have the same function type.</p>
1052 <p>Parameter attributes are simple keywords that follow the type specified. If
1053 multiple parameter attributes are needed, they are space separated. For
1056 <pre class="doc_code">
1057 declare i32 @printf(i8* noalias nocapture, ...)
1058 declare i32 @atoi(i8 zeroext)
1059 declare signext i8 @returns_signed_char()
1062 <p>Note that any attributes for the function result (<tt>nounwind</tt>,
1063 <tt>readonly</tt>) come immediately after the argument list.</p>
1065 <p>Currently, only the following parameter attributes are defined:</p>
1068 <dt><tt><b>zeroext</b></tt></dt>
1069 <dd>This indicates to the code generator that the parameter or return value
1070 should be zero-extended to the extent required by the target's ABI (which
1071 is usually 32-bits, but is 8-bits for a i1 on x86-64) by the caller (for a
1072 parameter) or the callee (for a return value).</dd>
1074 <dt><tt><b>signext</b></tt></dt>
1075 <dd>This indicates to the code generator that the parameter or return value
1076 should be sign-extended to the extent required by the target's ABI (which
1077 is usually 32-bits) by the caller (for a parameter) or the callee (for a
1080 <dt><tt><b>inreg</b></tt></dt>
1081 <dd>This indicates that this parameter or return value should be treated in a
1082 special target-dependent fashion during while emitting code for a function
1083 call or return (usually, by putting it in a register as opposed to memory,
1084 though some targets use it to distinguish between two different kinds of
1085 registers). Use of this attribute is target-specific.</dd>
1087 <dt><tt><b><a name="byval">byval</a></b></tt></dt>
1088 <dd><p>This indicates that the pointer parameter should really be passed by
1089 value to the function. The attribute implies that a hidden copy of the
1091 is made between the caller and the callee, so the callee is unable to
1092 modify the value in the caller. This attribute is only valid on LLVM
1093 pointer arguments. It is generally used to pass structs and arrays by
1094 value, but is also valid on pointers to scalars. The copy is considered
1095 to belong to the caller not the callee (for example,
1096 <tt><a href="#readonly">readonly</a></tt> functions should not write to
1097 <tt>byval</tt> parameters). This is not a valid attribute for return
1100 <p>The byval attribute also supports specifying an alignment with
1101 the align attribute. It indicates the alignment of the stack slot to
1102 form and the known alignment of the pointer specified to the call site. If
1103 the alignment is not specified, then the code generator makes a
1104 target-specific assumption.</p></dd>
1106 <dt><tt><b><a name="sret">sret</a></b></tt></dt>
1107 <dd>This indicates that the pointer parameter specifies the address of a
1108 structure that is the return value of the function in the source program.
1109 This pointer must be guaranteed by the caller to be valid: loads and
1110 stores to the structure may be assumed by the callee to not to trap and
1111 to be properly aligned. This may only be applied to the first parameter.
1112 This is not a valid attribute for return values. </dd>
1114 <dt><tt><b><a name="noalias">noalias</a></b></tt></dt>
1115 <dd>This indicates that pointer values
1116 <a href="#pointeraliasing"><i>based</i></a> on the argument or return
1117 value do not alias pointer values which are not <i>based</i> on it,
1118 ignoring certain "irrelevant" dependencies.
1119 For a call to the parent function, dependencies between memory
1120 references from before or after the call and from those during the call
1121 are "irrelevant" to the <tt>noalias</tt> keyword for the arguments and
1122 return value used in that call.
1123 The caller shares the responsibility with the callee for ensuring that
1124 these requirements are met.
1125 For further details, please see the discussion of the NoAlias response in
1126 <a href="AliasAnalysis.html#MustMayNo">alias analysis</a>.<br>
1128 Note that this definition of <tt>noalias</tt> is intentionally
1129 similar to the definition of <tt>restrict</tt> in C99 for function
1130 arguments, though it is slightly weaker.
1132 For function return values, C99's <tt>restrict</tt> is not meaningful,
1133 while LLVM's <tt>noalias</tt> is.
1136 <dt><tt><b><a name="nocapture">nocapture</a></b></tt></dt>
1137 <dd>This indicates that the callee does not make any copies of the pointer
1138 that outlive the callee itself. This is not a valid attribute for return
1141 <dt><tt><b><a name="nest">nest</a></b></tt></dt>
1142 <dd>This indicates that the pointer parameter can be excised using the
1143 <a href="#int_trampoline">trampoline intrinsics</a>. This is not a valid
1144 attribute for return values.</dd>
1149 <!-- ======================================================================= -->
1151 <a name="gc">Garbage Collector Names</a>
1156 <p>Each function may specify a garbage collector name, which is simply a
1159 <pre class="doc_code">
1160 define void @f() gc "name" { ... }
1163 <p>The compiler declares the supported values of <i>name</i>. Specifying a
1164 collector which will cause the compiler to alter its output in order to
1165 support the named garbage collection algorithm.</p>
1169 <!-- ======================================================================= -->
1171 <a name="fnattrs">Function Attributes</a>
1176 <p>Function attributes are set to communicate additional information about a
1177 function. Function attributes are considered to be part of the function, not
1178 of the function type, so functions with different parameter attributes can
1179 have the same function type.</p>
1181 <p>Function attributes are simple keywords that follow the type specified. If
1182 multiple attributes are needed, they are space separated. For example:</p>
1184 <pre class="doc_code">
1185 define void @f() noinline { ... }
1186 define void @f() alwaysinline { ... }
1187 define void @f() alwaysinline optsize { ... }
1188 define void @f() optsize { ... }
1192 <dt><tt><b>address_safety</b></tt></dt>
1193 <dd>This attribute indicates that the address safety analysis
1194 is enabled for this function. </dd>
1196 <dt><tt><b>alignstack(<<em>n</em>>)</b></tt></dt>
1197 <dd>This attribute indicates that, when emitting the prologue and epilogue,
1198 the backend should forcibly align the stack pointer. Specify the
1199 desired alignment, which must be a power of two, in parentheses.
1201 <dt><tt><b>alwaysinline</b></tt></dt>
1202 <dd>This attribute indicates that the inliner should attempt to inline this
1203 function into callers whenever possible, ignoring any active inlining size
1204 threshold for this caller.</dd>
1206 <dt><tt><b>nonlazybind</b></tt></dt>
1207 <dd>This attribute suppresses lazy symbol binding for the function. This
1208 may make calls to the function faster, at the cost of extra program
1209 startup time if the function is not called during program startup.</dd>
1211 <dt><tt><b>inlinehint</b></tt></dt>
1212 <dd>This attribute indicates that the source code contained a hint that inlining
1213 this function is desirable (such as the "inline" keyword in C/C++). It
1214 is just a hint; it imposes no requirements on the inliner.</dd>
1216 <dt><tt><b>naked</b></tt></dt>
1217 <dd>This attribute disables prologue / epilogue emission for the function.
1218 This can have very system-specific consequences.</dd>
1220 <dt><tt><b>noimplicitfloat</b></tt></dt>
1221 <dd>This attributes disables implicit floating point instructions.</dd>
1223 <dt><tt><b>noinline</b></tt></dt>
1224 <dd>This attribute indicates that the inliner should never inline this
1225 function in any situation. This attribute may not be used together with
1226 the <tt>alwaysinline</tt> attribute.</dd>
1228 <dt><tt><b>noredzone</b></tt></dt>
1229 <dd>This attribute indicates that the code generator should not use a red
1230 zone, even if the target-specific ABI normally permits it.</dd>
1232 <dt><tt><b>noreturn</b></tt></dt>
1233 <dd>This function attribute indicates that the function never returns
1234 normally. This produces undefined behavior at runtime if the function
1235 ever does dynamically return.</dd>
1237 <dt><tt><b>nounwind</b></tt></dt>
1238 <dd>This function attribute indicates that the function never returns with an
1239 unwind or exceptional control flow. If the function does unwind, its
1240 runtime behavior is undefined.</dd>
1242 <dt><tt><b>optsize</b></tt></dt>
1243 <dd>This attribute suggests that optimization passes and code generator passes
1244 make choices that keep the code size of this function low, and otherwise
1245 do optimizations specifically to reduce code size.</dd>
1247 <dt><tt><b>readnone</b></tt></dt>
1248 <dd>This attribute indicates that the function computes its result (or decides
1249 to unwind an exception) based strictly on its arguments, without
1250 dereferencing any pointer arguments or otherwise accessing any mutable
1251 state (e.g. memory, control registers, etc) visible to caller functions.
1252 It does not write through any pointer arguments
1253 (including <tt><a href="#byval">byval</a></tt> arguments) and never
1254 changes any state visible to callers. This means that it cannot unwind
1255 exceptions by calling the <tt>C++</tt> exception throwing methods.</dd>
1257 <dt><tt><b><a name="readonly">readonly</a></b></tt></dt>
1258 <dd>This attribute indicates that the function does not write through any
1259 pointer arguments (including <tt><a href="#byval">byval</a></tt>
1260 arguments) or otherwise modify any state (e.g. memory, control registers,
1261 etc) visible to caller functions. It may dereference pointer arguments
1262 and read state that may be set in the caller. A readonly function always
1263 returns the same value (or unwinds an exception identically) when called
1264 with the same set of arguments and global state. It cannot unwind an
1265 exception by calling the <tt>C++</tt> exception throwing methods.</dd>
1267 <dt><tt><b><a name="returns_twice">returns_twice</a></b></tt></dt>
1268 <dd>This attribute indicates that this function can return twice. The
1269 C <code>setjmp</code> is an example of such a function. The compiler
1270 disables some optimizations (like tail calls) in the caller of these
1273 <dt><tt><b><a name="ssp">ssp</a></b></tt></dt>
1274 <dd>This attribute indicates that the function should emit a stack smashing
1275 protector. It is in the form of a "canary"—a random value placed on
1276 the stack before the local variables that's checked upon return from the
1277 function to see if it has been overwritten. A heuristic is used to
1278 determine if a function needs stack protectors or not.<br>
1280 If a function that has an <tt>ssp</tt> attribute is inlined into a
1281 function that doesn't have an <tt>ssp</tt> attribute, then the resulting
1282 function will have an <tt>ssp</tt> attribute.</dd>
1284 <dt><tt><b>sspreq</b></tt></dt>
1285 <dd>This attribute indicates that the function should <em>always</em> emit a
1286 stack smashing protector. This overrides
1287 the <tt><a href="#ssp">ssp</a></tt> function attribute.<br>
1289 If a function that has an <tt>sspreq</tt> attribute is inlined into a
1290 function that doesn't have an <tt>sspreq</tt> attribute or which has
1291 an <tt>ssp</tt> attribute, then the resulting function will have
1292 an <tt>sspreq</tt> attribute.</dd>
1294 <dt><tt><b><a name="uwtable">uwtable</a></b></tt></dt>
1295 <dd>This attribute indicates that the ABI being targeted requires that
1296 an unwind table entry be produce for this function even if we can
1297 show that no exceptions passes by it. This is normally the case for
1298 the ELF x86-64 abi, but it can be disabled for some compilation
1304 <!-- ======================================================================= -->
1306 <a name="moduleasm">Module-Level Inline Assembly</a>
1311 <p>Modules may contain "module-level inline asm" blocks, which corresponds to
1312 the GCC "file scope inline asm" blocks. These blocks are internally
1313 concatenated by LLVM and treated as a single unit, but may be separated in
1314 the <tt>.ll</tt> file if desired. The syntax is very simple:</p>
1316 <pre class="doc_code">
1317 module asm "inline asm code goes here"
1318 module asm "more can go here"
1321 <p>The strings can contain any character by escaping non-printable characters.
1322 The escape sequence used is simply "\xx" where "xx" is the two digit hex code
1325 <p>The inline asm code is simply printed to the machine code .s file when
1326 assembly code is generated.</p>
1330 <!-- ======================================================================= -->
1332 <a name="datalayout">Data Layout</a>
1337 <p>A module may specify a target specific data layout string that specifies how
1338 data is to be laid out in memory. The syntax for the data layout is
1341 <pre class="doc_code">
1342 target datalayout = "<i>layout specification</i>"
1345 <p>The <i>layout specification</i> consists of a list of specifications
1346 separated by the minus sign character ('-'). Each specification starts with
1347 a letter and may include other information after the letter to define some
1348 aspect of the data layout. The specifications accepted are as follows:</p>
1352 <dd>Specifies that the target lays out data in big-endian form. That is, the
1353 bits with the most significance have the lowest address location.</dd>
1356 <dd>Specifies that the target lays out data in little-endian form. That is,
1357 the bits with the least significance have the lowest address
1360 <dt><tt>S<i>size</i></tt></dt>
1361 <dd>Specifies the natural alignment of the stack in bits. Alignment promotion
1362 of stack variables is limited to the natural stack alignment to avoid
1363 dynamic stack realignment. The stack alignment must be a multiple of
1364 8-bits. If omitted, the natural stack alignment defaults to "unspecified",
1365 which does not prevent any alignment promotions.</dd>
1367 <dt><tt>p[n]:<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1368 <dd>This specifies the <i>size</i> of a pointer and its <i>abi</i> and
1369 <i>preferred</i> alignments for address space <i>n</i>. All sizes are in
1370 bits. Specifying the <i>pref</i> alignment is optional. If omitted, the
1371 preceding <tt>:</tt> should be omitted too. The address space,
1372 <i>n</i> is optional, and if not specified, denotes the default address
1373 space 0. The value of <i>n</i> must be in the range [1,2^23).</dd>
1375 <dt><tt>i<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1376 <dd>This specifies the alignment for an integer type of a given bit
1377 <i>size</i>. The value of <i>size</i> must be in the range [1,2^23).</dd>
1379 <dt><tt>v<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1380 <dd>This specifies the alignment for a vector type of a given bit
1383 <dt><tt>f<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1384 <dd>This specifies the alignment for a floating point type of a given bit
1385 <i>size</i>. Only values of <i>size</i> that are supported by the target
1386 will work. 32 (float) and 64 (double) are supported on all targets;
1387 80 or 128 (different flavors of long double) are also supported on some
1390 <dt><tt>a<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1391 <dd>This specifies the alignment for an aggregate type of a given bit
1394 <dt><tt>s<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1395 <dd>This specifies the alignment for a stack object of a given bit
1398 <dt><tt>n<i>size1</i>:<i>size2</i>:<i>size3</i>...</tt></dt>
1399 <dd>This specifies a set of native integer widths for the target CPU
1400 in bits. For example, it might contain "n32" for 32-bit PowerPC,
1401 "n32:64" for PowerPC 64, or "n8:16:32:64" for X86-64. Elements of
1402 this set are considered to support most general arithmetic
1403 operations efficiently.</dd>
1406 <p>When constructing the data layout for a given target, LLVM starts with a
1407 default set of specifications which are then (possibly) overridden by the
1408 specifications in the <tt>datalayout</tt> keyword. The default specifications
1409 are given in this list:</p>
1412 <li><tt>E</tt> - big endian</li>
1413 <li><tt>p:64:64:64</tt> - 64-bit pointers with 64-bit alignment</li>
1414 <li><tt>p1:32:32:32</tt> - 32-bit pointers with 32-bit alignment for
1415 address space 1</li>
1416 <li><tt>p2:16:32:32</tt> - 16-bit pointers with 32-bit alignment for
1417 address space 2</li>
1418 <li><tt>i1:8:8</tt> - i1 is 8-bit (byte) aligned</li>
1419 <li><tt>i8:8:8</tt> - i8 is 8-bit (byte) aligned</li>
1420 <li><tt>i16:16:16</tt> - i16 is 16-bit aligned</li>
1421 <li><tt>i32:32:32</tt> - i32 is 32-bit aligned</li>
1422 <li><tt>i64:32:64</tt> - i64 has ABI alignment of 32-bits but preferred
1423 alignment of 64-bits</li>
1424 <li><tt>f32:32:32</tt> - float is 32-bit aligned</li>
1425 <li><tt>f64:64:64</tt> - double is 64-bit aligned</li>
1426 <li><tt>v64:64:64</tt> - 64-bit vector is 64-bit aligned</li>
1427 <li><tt>v128:128:128</tt> - 128-bit vector is 128-bit aligned</li>
1428 <li><tt>a0:0:1</tt> - aggregates are 8-bit aligned</li>
1429 <li><tt>s0:64:64</tt> - stack objects are 64-bit aligned</li>
1432 <p>When LLVM is determining the alignment for a given type, it uses the
1433 following rules:</p>
1436 <li>If the type sought is an exact match for one of the specifications, that
1437 specification is used.</li>
1439 <li>If no match is found, and the type sought is an integer type, then the
1440 smallest integer type that is larger than the bitwidth of the sought type
1441 is used. If none of the specifications are larger than the bitwidth then
1442 the largest integer type is used. For example, given the default
1443 specifications above, the i7 type will use the alignment of i8 (next
1444 largest) while both i65 and i256 will use the alignment of i64 (largest
1447 <li>If no match is found, and the type sought is a vector type, then the
1448 largest vector type that is smaller than the sought vector type will be
1449 used as a fall back. This happens because <128 x double> can be
1450 implemented in terms of 64 <2 x double>, for example.</li>
1453 <p>The function of the data layout string may not be what you expect. Notably,
1454 this is not a specification from the frontend of what alignment the code
1455 generator should use.</p>
1457 <p>Instead, if specified, the target data layout is required to match what the
1458 ultimate <em>code generator</em> expects. This string is used by the
1459 mid-level optimizers to
1460 improve code, and this only works if it matches what the ultimate code
1461 generator uses. If you would like to generate IR that does not embed this
1462 target-specific detail into the IR, then you don't have to specify the
1463 string. This will disable some optimizations that require precise layout
1464 information, but this also prevents those optimizations from introducing
1465 target specificity into the IR.</p>
1471 <!-- ======================================================================= -->
1473 <a name="pointeraliasing">Pointer Aliasing Rules</a>
1478 <p>Any memory access must be done through a pointer value associated
1479 with an address range of the memory access, otherwise the behavior
1480 is undefined. Pointer values are associated with address ranges
1481 according to the following rules:</p>
1484 <li>A pointer value is associated with the addresses associated with
1485 any value it is <i>based</i> on.
1486 <li>An address of a global variable is associated with the address
1487 range of the variable's storage.</li>
1488 <li>The result value of an allocation instruction is associated with
1489 the address range of the allocated storage.</li>
1490 <li>A null pointer in the default address-space is associated with
1492 <li>An integer constant other than zero or a pointer value returned
1493 from a function not defined within LLVM may be associated with address
1494 ranges allocated through mechanisms other than those provided by
1495 LLVM. Such ranges shall not overlap with any ranges of addresses
1496 allocated by mechanisms provided by LLVM.</li>
1499 <p>A pointer value is <i>based</i> on another pointer value according
1500 to the following rules:</p>
1503 <li>A pointer value formed from a
1504 <tt><a href="#i_getelementptr">getelementptr</a></tt> operation
1505 is <i>based</i> on the first operand of the <tt>getelementptr</tt>.</li>
1506 <li>The result value of a
1507 <tt><a href="#i_bitcast">bitcast</a></tt> is <i>based</i> on the operand
1508 of the <tt>bitcast</tt>.</li>
1509 <li>A pointer value formed by an
1510 <tt><a href="#i_inttoptr">inttoptr</a></tt> is <i>based</i> on all
1511 pointer values that contribute (directly or indirectly) to the
1512 computation of the pointer's value.</li>
1513 <li>The "<i>based</i> on" relationship is transitive.</li>
1516 <p>Note that this definition of <i>"based"</i> is intentionally
1517 similar to the definition of <i>"based"</i> in C99, though it is
1518 slightly weaker.</p>
1520 <p>LLVM IR does not associate types with memory. The result type of a
1521 <tt><a href="#i_load">load</a></tt> merely indicates the size and
1522 alignment of the memory from which to load, as well as the
1523 interpretation of the value. The first operand type of a
1524 <tt><a href="#i_store">store</a></tt> similarly only indicates the size
1525 and alignment of the store.</p>
1527 <p>Consequently, type-based alias analysis, aka TBAA, aka
1528 <tt>-fstrict-aliasing</tt>, is not applicable to general unadorned
1529 LLVM IR. <a href="#metadata">Metadata</a> may be used to encode
1530 additional information which specialized optimization passes may use
1531 to implement type-based alias analysis.</p>
1535 <!-- ======================================================================= -->
1537 <a name="volatile">Volatile Memory Accesses</a>
1542 <p>Certain memory accesses, such as <a href="#i_load"><tt>load</tt></a>s, <a
1543 href="#i_store"><tt>store</tt></a>s, and <a
1544 href="#int_memcpy"><tt>llvm.memcpy</tt></a>s may be marked <tt>volatile</tt>.
1545 The optimizers must not change the number of volatile operations or change their
1546 order of execution relative to other volatile operations. The optimizers
1547 <i>may</i> change the order of volatile operations relative to non-volatile
1548 operations. This is not Java's "volatile" and has no cross-thread
1549 synchronization behavior.</p>
1553 <!-- ======================================================================= -->
1555 <a name="memmodel">Memory Model for Concurrent Operations</a>
1560 <p>The LLVM IR does not define any way to start parallel threads of execution
1561 or to register signal handlers. Nonetheless, there are platform-specific
1562 ways to create them, and we define LLVM IR's behavior in their presence. This
1563 model is inspired by the C++0x memory model.</p>
1565 <p>For a more informal introduction to this model, see the
1566 <a href="Atomics.html">LLVM Atomic Instructions and Concurrency Guide</a>.
1568 <p>We define a <i>happens-before</i> partial order as the least partial order
1571 <li>Is a superset of single-thread program order, and</li>
1572 <li>When a <i>synchronizes-with</i> <tt>b</tt>, includes an edge from
1573 <tt>a</tt> to <tt>b</tt>. <i>Synchronizes-with</i> pairs are introduced
1574 by platform-specific techniques, like pthread locks, thread
1575 creation, thread joining, etc., and by atomic instructions.
1576 (See also <a href="#ordering">Atomic Memory Ordering Constraints</a>).
1580 <p>Note that program order does not introduce <i>happens-before</i> edges
1581 between a thread and signals executing inside that thread.</p>
1583 <p>Every (defined) read operation (load instructions, memcpy, atomic
1584 loads/read-modify-writes, etc.) <var>R</var> reads a series of bytes written by
1585 (defined) write operations (store instructions, atomic
1586 stores/read-modify-writes, memcpy, etc.). For the purposes of this section,
1587 initialized globals are considered to have a write of the initializer which is
1588 atomic and happens before any other read or write of the memory in question.
1589 For each byte of a read <var>R</var>, <var>R<sub>byte</sub></var> may see
1590 any write to the same byte, except:</p>
1593 <li>If <var>write<sub>1</sub></var> happens before
1594 <var>write<sub>2</sub></var>, and <var>write<sub>2</sub></var> happens
1595 before <var>R<sub>byte</sub></var>, then <var>R<sub>byte</sub></var>
1596 does not see <var>write<sub>1</sub></var>.
1597 <li>If <var>R<sub>byte</sub></var> happens before
1598 <var>write<sub>3</sub></var>, then <var>R<sub>byte</sub></var> does not
1599 see <var>write<sub>3</sub></var>.
1602 <p>Given that definition, <var>R<sub>byte</sub></var> is defined as follows:
1604 <li>If <var>R</var> is volatile, the result is target-dependent. (Volatile
1605 is supposed to give guarantees which can support
1606 <code>sig_atomic_t</code> in C/C++, and may be used for accesses to
1607 addresses which do not behave like normal memory. It does not generally
1608 provide cross-thread synchronization.)
1609 <li>Otherwise, if there is no write to the same byte that happens before
1610 <var>R<sub>byte</sub></var>, <var>R<sub>byte</sub></var> returns
1611 <tt>undef</tt> for that byte.
1612 <li>Otherwise, if <var>R<sub>byte</sub></var> may see exactly one write,
1613 <var>R<sub>byte</sub></var> returns the value written by that
1615 <li>Otherwise, if <var>R</var> is atomic, and all the writes
1616 <var>R<sub>byte</sub></var> may see are atomic, it chooses one of the
1617 values written. See the <a href="#ordering">Atomic Memory Ordering
1618 Constraints</a> section for additional constraints on how the choice
1620 <li>Otherwise <var>R<sub>byte</sub></var> returns <tt>undef</tt>.</li>
1623 <p><var>R</var> returns the value composed of the series of bytes it read.
1624 This implies that some bytes within the value may be <tt>undef</tt>
1625 <b>without</b> the entire value being <tt>undef</tt>. Note that this only
1626 defines the semantics of the operation; it doesn't mean that targets will
1627 emit more than one instruction to read the series of bytes.</p>
1629 <p>Note that in cases where none of the atomic intrinsics are used, this model
1630 places only one restriction on IR transformations on top of what is required
1631 for single-threaded execution: introducing a store to a byte which might not
1632 otherwise be stored is not allowed in general. (Specifically, in the case
1633 where another thread might write to and read from an address, introducing a
1634 store can change a load that may see exactly one write into a load that may
1635 see multiple writes.)</p>
1637 <!-- FIXME: This model assumes all targets where concurrency is relevant have
1638 a byte-size store which doesn't affect adjacent bytes. As far as I can tell,
1639 none of the backends currently in the tree fall into this category; however,
1640 there might be targets which care. If there are, we want a paragraph
1643 Targets may specify that stores narrower than a certain width are not
1644 available; on such a target, for the purposes of this model, treat any
1645 non-atomic write with an alignment or width less than the minimum width
1646 as if it writes to the relevant surrounding bytes.
1651 <!-- ======================================================================= -->
1653 <a name="ordering">Atomic Memory Ordering Constraints</a>
1658 <p>Atomic instructions (<a href="#i_cmpxchg"><code>cmpxchg</code></a>,
1659 <a href="#i_atomicrmw"><code>atomicrmw</code></a>,
1660 <a href="#i_fence"><code>fence</code></a>,
1661 <a href="#i_load"><code>atomic load</code></a>, and
1662 <a href="#i_store"><code>atomic store</code></a>) take an ordering parameter
1663 that determines which other atomic instructions on the same address they
1664 <i>synchronize with</i>. These semantics are borrowed from Java and C++0x,
1665 but are somewhat more colloquial. If these descriptions aren't precise enough,
1666 check those specs (see spec references in the
1667 <a href="Atomics.html#introduction">atomics guide</a>).
1668 <a href="#i_fence"><code>fence</code></a> instructions
1669 treat these orderings somewhat differently since they don't take an address.
1670 See that instruction's documentation for details.</p>
1672 <p>For a simpler introduction to the ordering constraints, see the
1673 <a href="Atomics.html">LLVM Atomic Instructions and Concurrency Guide</a>.</p>
1676 <dt><code>unordered</code></dt>
1677 <dd>The set of values that can be read is governed by the happens-before
1678 partial order. A value cannot be read unless some operation wrote it.
1679 This is intended to provide a guarantee strong enough to model Java's
1680 non-volatile shared variables. This ordering cannot be specified for
1681 read-modify-write operations; it is not strong enough to make them atomic
1682 in any interesting way.</dd>
1683 <dt><code>monotonic</code></dt>
1684 <dd>In addition to the guarantees of <code>unordered</code>, there is a single
1685 total order for modifications by <code>monotonic</code> operations on each
1686 address. All modification orders must be compatible with the happens-before
1687 order. There is no guarantee that the modification orders can be combined to
1688 a global total order for the whole program (and this often will not be
1689 possible). The read in an atomic read-modify-write operation
1690 (<a href="#i_cmpxchg"><code>cmpxchg</code></a> and
1691 <a href="#i_atomicrmw"><code>atomicrmw</code></a>)
1692 reads the value in the modification order immediately before the value it
1693 writes. If one atomic read happens before another atomic read of the same
1694 address, the later read must see the same value or a later value in the
1695 address's modification order. This disallows reordering of
1696 <code>monotonic</code> (or stronger) operations on the same address. If an
1697 address is written <code>monotonic</code>ally by one thread, and other threads
1698 <code>monotonic</code>ally read that address repeatedly, the other threads must
1699 eventually see the write. This corresponds to the C++0x/C1x
1700 <code>memory_order_relaxed</code>.</dd>
1701 <dt><code>acquire</code></dt>
1702 <dd>In addition to the guarantees of <code>monotonic</code>,
1703 a <i>synchronizes-with</i> edge may be formed with a <code>release</code>
1704 operation. This is intended to model C++'s <code>memory_order_acquire</code>.</dd>
1705 <dt><code>release</code></dt>
1706 <dd>In addition to the guarantees of <code>monotonic</code>, if this operation
1707 writes a value which is subsequently read by an <code>acquire</code> operation,
1708 it <i>synchronizes-with</i> that operation. (This isn't a complete
1709 description; see the C++0x definition of a release sequence.) This corresponds
1710 to the C++0x/C1x <code>memory_order_release</code>.</dd>
1711 <dt><code>acq_rel</code> (acquire+release)</dt><dd>Acts as both an
1712 <code>acquire</code> and <code>release</code> operation on its address.
1713 This corresponds to the C++0x/C1x <code>memory_order_acq_rel</code>.</dd>
1714 <dt><code>seq_cst</code> (sequentially consistent)</dt><dd>
1715 <dd>In addition to the guarantees of <code>acq_rel</code>
1716 (<code>acquire</code> for an operation which only reads, <code>release</code>
1717 for an operation which only writes), there is a global total order on all
1718 sequentially-consistent operations on all addresses, which is consistent with
1719 the <i>happens-before</i> partial order and with the modification orders of
1720 all the affected addresses. Each sequentially-consistent read sees the last
1721 preceding write to the same address in this global order. This corresponds
1722 to the C++0x/C1x <code>memory_order_seq_cst</code> and Java volatile.</dd>
1725 <p id="singlethread">If an atomic operation is marked <code>singlethread</code>,
1726 it only <i>synchronizes with</i> or participates in modification and seq_cst
1727 total orderings with other operations running in the same thread (for example,
1728 in signal handlers).</p>
1734 <!-- *********************************************************************** -->
1735 <h2><a name="typesystem">Type System</a></h2>
1736 <!-- *********************************************************************** -->
1740 <p>The LLVM type system is one of the most important features of the
1741 intermediate representation. Being typed enables a number of optimizations
1742 to be performed on the intermediate representation directly, without having
1743 to do extra analyses on the side before the transformation. A strong type
1744 system makes it easier to read the generated code and enables novel analyses
1745 and transformations that are not feasible to perform on normal three address
1746 code representations.</p>
1748 <!-- ======================================================================= -->
1750 <a name="t_classifications">Type Classifications</a>
1755 <p>The types fall into a few useful classifications:</p>
1757 <table border="1" cellspacing="0" cellpadding="4">
1759 <tr><th>Classification</th><th>Types</th></tr>
1761 <td><a href="#t_integer">integer</a></td>
1762 <td><tt>i1, i2, i3, ... i8, ... i16, ... i32, ... i64, ... </tt></td>
1765 <td><a href="#t_floating">floating point</a></td>
1766 <td><tt>half, float, double, x86_fp80, fp128, ppc_fp128</tt></td>
1769 <td><a name="t_firstclass">first class</a></td>
1770 <td><a href="#t_integer">integer</a>,
1771 <a href="#t_floating">floating point</a>,
1772 <a href="#t_pointer">pointer</a>,
1773 <a href="#t_vector">vector</a>,
1774 <a href="#t_struct">structure</a>,
1775 <a href="#t_array">array</a>,
1776 <a href="#t_label">label</a>,
1777 <a href="#t_metadata">metadata</a>.
1781 <td><a href="#t_primitive">primitive</a></td>
1782 <td><a href="#t_label">label</a>,
1783 <a href="#t_void">void</a>,
1784 <a href="#t_integer">integer</a>,
1785 <a href="#t_floating">floating point</a>,
1786 <a href="#t_x86mmx">x86mmx</a>,
1787 <a href="#t_metadata">metadata</a>.</td>
1790 <td><a href="#t_derived">derived</a></td>
1791 <td><a href="#t_array">array</a>,
1792 <a href="#t_function">function</a>,
1793 <a href="#t_pointer">pointer</a>,
1794 <a href="#t_struct">structure</a>,
1795 <a href="#t_vector">vector</a>,
1796 <a href="#t_opaque">opaque</a>.
1802 <p>The <a href="#t_firstclass">first class</a> types are perhaps the most
1803 important. Values of these types are the only ones which can be produced by
1808 <!-- ======================================================================= -->
1810 <a name="t_primitive">Primitive Types</a>
1815 <p>The primitive types are the fundamental building blocks of the LLVM
1818 <!-- _______________________________________________________________________ -->
1820 <a name="t_integer">Integer Type</a>
1826 <p>The integer type is a very simple type that simply specifies an arbitrary
1827 bit width for the integer type desired. Any bit width from 1 bit to
1828 2<sup>23</sup>-1 (about 8 million) can be specified.</p>
1835 <p>The number of bits the integer will occupy is specified by the <tt>N</tt>
1839 <table class="layout">
1841 <td class="left"><tt>i1</tt></td>
1842 <td class="left">a single-bit integer.</td>
1845 <td class="left"><tt>i32</tt></td>
1846 <td class="left">a 32-bit integer.</td>
1849 <td class="left"><tt>i1942652</tt></td>
1850 <td class="left">a really big integer of over 1 million bits.</td>
1856 <!-- _______________________________________________________________________ -->
1858 <a name="t_floating">Floating Point Types</a>
1865 <tr><th>Type</th><th>Description</th></tr>
1866 <tr><td><tt>half</tt></td><td>16-bit floating point value</td></tr>
1867 <tr><td><tt>float</tt></td><td>32-bit floating point value</td></tr>
1868 <tr><td><tt>double</tt></td><td>64-bit floating point value</td></tr>
1869 <tr><td><tt>fp128</tt></td><td>128-bit floating point value (112-bit mantissa)</td></tr>
1870 <tr><td><tt>x86_fp80</tt></td><td>80-bit floating point value (X87)</td></tr>
1871 <tr><td><tt>ppc_fp128</tt></td><td>128-bit floating point value (two 64-bits)</td></tr>
1877 <!-- _______________________________________________________________________ -->
1879 <a name="t_x86mmx">X86mmx Type</a>
1885 <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>
1894 <!-- _______________________________________________________________________ -->
1896 <a name="t_void">Void Type</a>
1902 <p>The void type does not represent any value and has no size.</p>
1911 <!-- _______________________________________________________________________ -->
1913 <a name="t_label">Label Type</a>
1919 <p>The label type represents code labels.</p>
1928 <!-- _______________________________________________________________________ -->
1930 <a name="t_metadata">Metadata Type</a>
1936 <p>The metadata type represents embedded metadata. No derived types may be
1937 created from metadata except for <a href="#t_function">function</a>
1949 <!-- ======================================================================= -->
1951 <a name="t_derived">Derived Types</a>
1956 <p>The real power in LLVM comes from the derived types in the system. This is
1957 what allows a programmer to represent arrays, functions, pointers, and other
1958 useful types. Each of these types contain one or more element types which
1959 may be a primitive type, or another derived type. For example, it is
1960 possible to have a two dimensional array, using an array as the element type
1961 of another array.</p>
1963 <!-- _______________________________________________________________________ -->
1965 <a name="t_aggregate">Aggregate Types</a>
1970 <p>Aggregate Types are a subset of derived types that can contain multiple
1971 member types. <a href="#t_array">Arrays</a> and
1972 <a href="#t_struct">structs</a> are aggregate types.
1973 <a href="#t_vector">Vectors</a> are not considered to be aggregate types.</p>
1977 <!-- _______________________________________________________________________ -->
1979 <a name="t_array">Array Type</a>
1985 <p>The array type is a very simple derived type that arranges elements
1986 sequentially in memory. The array type requires a size (number of elements)
1987 and an underlying data type.</p>
1991 [<# elements> x <elementtype>]
1994 <p>The number of elements is a constant integer value; <tt>elementtype</tt> may
1995 be any type with a size.</p>
1998 <table class="layout">
2000 <td class="left"><tt>[40 x i32]</tt></td>
2001 <td class="left">Array of 40 32-bit integer values.</td>
2004 <td class="left"><tt>[41 x i32]</tt></td>
2005 <td class="left">Array of 41 32-bit integer values.</td>
2008 <td class="left"><tt>[4 x i8]</tt></td>
2009 <td class="left">Array of 4 8-bit integer values.</td>
2012 <p>Here are some examples of multidimensional arrays:</p>
2013 <table class="layout">
2015 <td class="left"><tt>[3 x [4 x i32]]</tt></td>
2016 <td class="left">3x4 array of 32-bit integer values.</td>
2019 <td class="left"><tt>[12 x [10 x float]]</tt></td>
2020 <td class="left">12x10 array of single precision floating point values.</td>
2023 <td class="left"><tt>[2 x [3 x [4 x i16]]]</tt></td>
2024 <td class="left">2x3x4 array of 16-bit integer values.</td>
2028 <p>There is no restriction on indexing beyond the end of the array implied by
2029 a static type (though there are restrictions on indexing beyond the bounds
2030 of an allocated object in some cases). This means that single-dimension
2031 'variable sized array' addressing can be implemented in LLVM with a zero
2032 length array type. An implementation of 'pascal style arrays' in LLVM could
2033 use the type "<tt>{ i32, [0 x float]}</tt>", for example.</p>
2037 <!-- _______________________________________________________________________ -->
2039 <a name="t_function">Function Type</a>
2045 <p>The function type can be thought of as a function signature. It consists of
2046 a return type and a list of formal parameter types. The return type of a
2047 function type is a first class type or a void type.</p>
2051 <returntype> (<parameter list>)
2054 <p>...where '<tt><parameter list></tt>' is a comma-separated list of type
2055 specifiers. Optionally, the parameter list may include a type <tt>...</tt>,
2056 which indicates that the function takes a variable number of arguments.
2057 Variable argument functions can access their arguments with
2058 the <a href="#int_varargs">variable argument handling intrinsic</a>
2059 functions. '<tt><returntype></tt>' is any type except
2060 <a href="#t_label">label</a>.</p>
2063 <table class="layout">
2065 <td class="left"><tt>i32 (i32)</tt></td>
2066 <td class="left">function taking an <tt>i32</tt>, returning an <tt>i32</tt>
2068 </tr><tr class="layout">
2069 <td class="left"><tt>float (i16, i32 *) *
2071 <td class="left"><a href="#t_pointer">Pointer</a> to a function that takes
2072 an <tt>i16</tt> and a <a href="#t_pointer">pointer</a> to <tt>i32</tt>,
2073 returning <tt>float</tt>.
2075 </tr><tr class="layout">
2076 <td class="left"><tt>i32 (i8*, ...)</tt></td>
2077 <td class="left">A vararg function that takes at least one
2078 <a href="#t_pointer">pointer</a> to <tt>i8 </tt> (char in C),
2079 which returns an integer. This is the signature for <tt>printf</tt> in
2082 </tr><tr class="layout">
2083 <td class="left"><tt>{i32, i32} (i32)</tt></td>
2084 <td class="left">A function taking an <tt>i32</tt>, returning a
2085 <a href="#t_struct">structure</a> containing two <tt>i32</tt> values
2092 <!-- _______________________________________________________________________ -->
2094 <a name="t_struct">Structure Type</a>
2100 <p>The structure type is used to represent a collection of data members together
2101 in memory. The elements of a structure may be any type that has a size.</p>
2103 <p>Structures in memory are accessed using '<tt><a href="#i_load">load</a></tt>'
2104 and '<tt><a href="#i_store">store</a></tt>' by getting a pointer to a field
2105 with the '<tt><a href="#i_getelementptr">getelementptr</a></tt>' instruction.
2106 Structures in registers are accessed using the
2107 '<tt><a href="#i_extractvalue">extractvalue</a></tt>' and
2108 '<tt><a href="#i_insertvalue">insertvalue</a></tt>' instructions.</p>
2110 <p>Structures may optionally be "packed" structures, which indicate that the
2111 alignment of the struct is one byte, and that there is no padding between
2112 the elements. In non-packed structs, padding between field types is inserted
2113 as defined by the DataLayout string in the module, which is required to match
2114 what the underlying code generator expects.</p>
2116 <p>Structures can either be "literal" or "identified". A literal structure is
2117 defined inline with other types (e.g. <tt>{i32, i32}*</tt>) whereas identified
2118 types are always defined at the top level with a name. Literal types are
2119 uniqued by their contents and can never be recursive or opaque since there is
2120 no way to write one. Identified types can be recursive, can be opaqued, and are
2126 %T1 = type { <type list> } <i>; Identified normal struct type</i>
2127 %T2 = type <{ <type list> }> <i>; Identified packed struct type</i>
2131 <table class="layout">
2133 <td class="left"><tt>{ i32, i32, i32 }</tt></td>
2134 <td class="left">A triple of three <tt>i32</tt> values</td>
2137 <td class="left"><tt>{ float, i32 (i32) * }</tt></td>
2138 <td class="left">A pair, where the first element is a <tt>float</tt> and the
2139 second element is a <a href="#t_pointer">pointer</a> to a
2140 <a href="#t_function">function</a> that takes an <tt>i32</tt>, returning
2141 an <tt>i32</tt>.</td>
2144 <td class="left"><tt><{ i8, i32 }></tt></td>
2145 <td class="left">A packed struct known to be 5 bytes in size.</td>
2151 <!-- _______________________________________________________________________ -->
2153 <a name="t_opaque">Opaque Structure Types</a>
2159 <p>Opaque structure types are used to represent named structure types that do
2160 not have a body specified. This corresponds (for example) to the C notion of
2161 a forward declared structure.</p>
2170 <table class="layout">
2172 <td class="left"><tt>opaque</tt></td>
2173 <td class="left">An opaque type.</td>
2181 <!-- _______________________________________________________________________ -->
2183 <a name="t_pointer">Pointer Type</a>
2189 <p>The pointer type is used to specify memory locations.
2190 Pointers are commonly used to reference objects in memory.</p>
2192 <p>Pointer types may have an optional address space attribute defining the
2193 numbered address space where the pointed-to object resides. The default
2194 address space is number zero. The semantics of non-zero address
2195 spaces are target-specific.</p>
2197 <p>Note that LLVM does not permit pointers to void (<tt>void*</tt>) nor does it
2198 permit pointers to labels (<tt>label*</tt>). Use <tt>i8*</tt> instead.</p>
2206 <table class="layout">
2208 <td class="left"><tt>[4 x i32]*</tt></td>
2209 <td class="left">A <a href="#t_pointer">pointer</a> to <a
2210 href="#t_array">array</a> of four <tt>i32</tt> values.</td>
2213 <td class="left"><tt>i32 (i32*) *</tt></td>
2214 <td class="left"> A <a href="#t_pointer">pointer</a> to a <a
2215 href="#t_function">function</a> that takes an <tt>i32*</tt>, returning an
2219 <td class="left"><tt>i32 addrspace(5)*</tt></td>
2220 <td class="left">A <a href="#t_pointer">pointer</a> to an <tt>i32</tt> value
2221 that resides in address space #5.</td>
2227 <!-- _______________________________________________________________________ -->
2229 <a name="t_vector">Vector Type</a>
2235 <p>A vector type is a simple derived type that represents a vector of elements.
2236 Vector types are used when multiple primitive data are operated in parallel
2237 using a single instruction (SIMD). A vector type requires a size (number of
2238 elements) and an underlying primitive data type. Vector types are considered
2239 <a href="#t_firstclass">first class</a>.</p>
2243 < <# elements> x <elementtype> >
2246 <p>The number of elements is a constant integer value larger than 0; elementtype
2247 may be any integer or floating point type, or a pointer to these types.
2248 Vectors of size zero are not allowed. </p>
2251 <table class="layout">
2253 <td class="left"><tt><4 x i32></tt></td>
2254 <td class="left">Vector of 4 32-bit integer values.</td>
2257 <td class="left"><tt><8 x float></tt></td>
2258 <td class="left">Vector of 8 32-bit floating-point values.</td>
2261 <td class="left"><tt><2 x i64></tt></td>
2262 <td class="left">Vector of 2 64-bit integer values.</td>
2265 <td class="left"><tt><4 x i64*></tt></td>
2266 <td class="left">Vector of 4 pointers to 64-bit integer values.</td>
2276 <!-- *********************************************************************** -->
2277 <h2><a name="constants">Constants</a></h2>
2278 <!-- *********************************************************************** -->
2282 <p>LLVM has several different basic types of constants. This section describes
2283 them all and their syntax.</p>
2285 <!-- ======================================================================= -->
2287 <a name="simpleconstants">Simple Constants</a>
2293 <dt><b>Boolean constants</b></dt>
2294 <dd>The two strings '<tt>true</tt>' and '<tt>false</tt>' are both valid
2295 constants of the <tt><a href="#t_integer">i1</a></tt> type.</dd>
2297 <dt><b>Integer constants</b></dt>
2298 <dd>Standard integers (such as '4') are constants of
2299 the <a href="#t_integer">integer</a> type. Negative numbers may be used
2300 with integer types.</dd>
2302 <dt><b>Floating point constants</b></dt>
2303 <dd>Floating point constants use standard decimal notation (e.g. 123.421),
2304 exponential notation (e.g. 1.23421e+2), or a more precise hexadecimal
2305 notation (see below). The assembler requires the exact decimal value of a
2306 floating-point constant. For example, the assembler accepts 1.25 but
2307 rejects 1.3 because 1.3 is a repeating decimal in binary. Floating point
2308 constants must have a <a href="#t_floating">floating point</a> type. </dd>
2310 <dt><b>Null pointer constants</b></dt>
2311 <dd>The identifier '<tt>null</tt>' is recognized as a null pointer constant
2312 and must be of <a href="#t_pointer">pointer type</a>.</dd>
2315 <p>The one non-intuitive notation for constants is the hexadecimal form of
2316 floating point constants. For example, the form '<tt>double
2317 0x432ff973cafa8000</tt>' is equivalent to (but harder to read than)
2318 '<tt>double 4.5e+15</tt>'. The only time hexadecimal floating point
2319 constants are required (and the only time that they are generated by the
2320 disassembler) is when a floating point constant must be emitted but it cannot
2321 be represented as a decimal floating point number in a reasonable number of
2322 digits. For example, NaN's, infinities, and other special values are
2323 represented in their IEEE hexadecimal format so that assembly and disassembly
2324 do not cause any bits to change in the constants.</p>
2326 <p>When using the hexadecimal form, constants of types half, float, and double are
2327 represented using the 16-digit form shown above (which matches the IEEE754
2328 representation for double); half and float values must, however, be exactly
2329 representable as IEE754 half and single precision, respectively.
2330 Hexadecimal format is always used
2331 for long double, and there are three forms of long double. The 80-bit format
2332 used by x86 is represented as <tt>0xK</tt> followed by 20 hexadecimal digits.
2333 The 128-bit format used by PowerPC (two adjacent doubles) is represented
2334 by <tt>0xM</tt> followed by 32 hexadecimal digits. The IEEE 128-bit format
2335 is represented by <tt>0xL</tt> followed by 32 hexadecimal digits; no
2336 currently supported target uses this format. Long doubles will only work if
2337 they match the long double format on your target. The IEEE 16-bit format
2338 (half precision) is represented by <tt>0xH</tt> followed by 4 hexadecimal
2339 digits. All hexadecimal formats are big-endian (sign bit at the left).</p>
2341 <p>There are no constants of type x86mmx.</p>
2344 <!-- ======================================================================= -->
2346 <a name="aggregateconstants"></a> <!-- old anchor -->
2347 <a name="complexconstants">Complex Constants</a>
2352 <p>Complex constants are a (potentially recursive) combination of simple
2353 constants and smaller complex constants.</p>
2356 <dt><b>Structure constants</b></dt>
2357 <dd>Structure constants are represented with notation similar to structure
2358 type definitions (a comma separated list of elements, surrounded by braces
2359 (<tt>{}</tt>)). For example: "<tt>{ i32 4, float 17.0, i32* @G }</tt>",
2360 where "<tt>@G</tt>" is declared as "<tt>@G = external global i32</tt>".
2361 Structure constants must have <a href="#t_struct">structure type</a>, and
2362 the number and types of elements must match those specified by the
2365 <dt><b>Array constants</b></dt>
2366 <dd>Array constants are represented with notation similar to array type
2367 definitions (a comma separated list of elements, surrounded by square
2368 brackets (<tt>[]</tt>)). For example: "<tt>[ i32 42, i32 11, i32 74
2369 ]</tt>". Array constants must have <a href="#t_array">array type</a>, and
2370 the number and types of elements must match those specified by the
2373 <dt><b>Vector constants</b></dt>
2374 <dd>Vector constants are represented with notation similar to vector type
2375 definitions (a comma separated list of elements, surrounded by
2376 less-than/greater-than's (<tt><></tt>)). For example: "<tt>< i32
2377 42, i32 11, i32 74, i32 100 ></tt>". Vector constants must
2378 have <a href="#t_vector">vector type</a>, and the number and types of
2379 elements must match those specified by the type.</dd>
2381 <dt><b>Zero initialization</b></dt>
2382 <dd>The string '<tt>zeroinitializer</tt>' can be used to zero initialize a
2383 value to zero of <em>any</em> type, including scalar and
2384 <a href="#t_aggregate">aggregate</a> types.
2385 This is often used to avoid having to print large zero initializers
2386 (e.g. for large arrays) and is always exactly equivalent to using explicit
2387 zero initializers.</dd>
2389 <dt><b>Metadata node</b></dt>
2390 <dd>A metadata node is a structure-like constant with
2391 <a href="#t_metadata">metadata type</a>. For example: "<tt>metadata !{
2392 i32 0, metadata !"test" }</tt>". Unlike other constants that are meant to
2393 be interpreted as part of the instruction stream, metadata is a place to
2394 attach additional information such as debug info.</dd>
2399 <!-- ======================================================================= -->
2401 <a name="globalconstants">Global Variable and Function Addresses</a>
2406 <p>The addresses of <a href="#globalvars">global variables</a>
2407 and <a href="#functionstructure">functions</a> are always implicitly valid
2408 (link-time) constants. These constants are explicitly referenced when
2409 the <a href="#identifiers">identifier for the global</a> is used and always
2410 have <a href="#t_pointer">pointer</a> type. For example, the following is a
2411 legal LLVM file:</p>
2413 <pre class="doc_code">
2416 @Z = global [2 x i32*] [ i32* @X, i32* @Y ]
2421 <!-- ======================================================================= -->
2423 <a name="undefvalues">Undefined Values</a>
2428 <p>The string '<tt>undef</tt>' can be used anywhere a constant is expected, and
2429 indicates that the user of the value may receive an unspecified bit-pattern.
2430 Undefined values may be of any type (other than '<tt>label</tt>'
2431 or '<tt>void</tt>') and be used anywhere a constant is permitted.</p>
2433 <p>Undefined values are useful because they indicate to the compiler that the
2434 program is well defined no matter what value is used. This gives the
2435 compiler more freedom to optimize. Here are some examples of (potentially
2436 surprising) transformations that are valid (in pseudo IR):</p>
2439 <pre class="doc_code">
2449 <p>This is safe because all of the output bits are affected by the undef bits.
2450 Any output bit can have a zero or one depending on the input bits.</p>
2452 <pre class="doc_code">
2463 <p>These logical operations have bits that are not always affected by the input.
2464 For example, if <tt>%X</tt> has a zero bit, then the output of the
2465 '<tt>and</tt>' operation will always be a zero for that bit, no matter what
2466 the corresponding bit from the '<tt>undef</tt>' is. As such, it is unsafe to
2467 optimize or assume that the result of the '<tt>and</tt>' is '<tt>undef</tt>'.
2468 However, it is safe to assume that all bits of the '<tt>undef</tt>' could be
2469 0, and optimize the '<tt>and</tt>' to 0. Likewise, it is safe to assume that
2470 all the bits of the '<tt>undef</tt>' operand to the '<tt>or</tt>' could be
2471 set, allowing the '<tt>or</tt>' to be folded to -1.</p>
2473 <pre class="doc_code">
2474 %A = select undef, %X, %Y
2475 %B = select undef, 42, %Y
2476 %C = select %X, %Y, undef
2487 <p>This set of examples shows that undefined '<tt>select</tt>' (and conditional
2488 branch) conditions can go <em>either way</em>, but they have to come from one
2489 of the two operands. In the <tt>%A</tt> example, if <tt>%X</tt> and
2490 <tt>%Y</tt> were both known to have a clear low bit, then <tt>%A</tt> would
2491 have to have a cleared low bit. However, in the <tt>%C</tt> example, the
2492 optimizer is allowed to assume that the '<tt>undef</tt>' operand could be the
2493 same as <tt>%Y</tt>, allowing the whole '<tt>select</tt>' to be
2496 <pre class="doc_code">
2497 %A = xor undef, undef
2515 <p>This example points out that two '<tt>undef</tt>' operands are not
2516 necessarily the same. This can be surprising to people (and also matches C
2517 semantics) where they assume that "<tt>X^X</tt>" is always zero, even
2518 if <tt>X</tt> is undefined. This isn't true for a number of reasons, but the
2519 short answer is that an '<tt>undef</tt>' "variable" can arbitrarily change
2520 its value over its "live range". This is true because the variable doesn't
2521 actually <em>have a live range</em>. Instead, the value is logically read
2522 from arbitrary registers that happen to be around when needed, so the value
2523 is not necessarily consistent over time. In fact, <tt>%A</tt> and <tt>%C</tt>
2524 need to have the same semantics or the core LLVM "replace all uses with"
2525 concept would not hold.</p>
2527 <pre class="doc_code">
2535 <p>These examples show the crucial difference between an <em>undefined
2536 value</em> and <em>undefined behavior</em>. An undefined value (like
2537 '<tt>undef</tt>') is allowed to have an arbitrary bit-pattern. This means that
2538 the <tt>%A</tt> operation can be constant folded to '<tt>undef</tt>', because
2539 the '<tt>undef</tt>' could be an SNaN, and <tt>fdiv</tt> is not (currently)
2540 defined on SNaN's. However, in the second example, we can make a more
2541 aggressive assumption: because the <tt>undef</tt> is allowed to be an
2542 arbitrary value, we are allowed to assume that it could be zero. Since a
2543 divide by zero has <em>undefined behavior</em>, we are allowed to assume that
2544 the operation does not execute at all. This allows us to delete the divide and
2545 all code after it. Because the undefined operation "can't happen", the
2546 optimizer can assume that it occurs in dead code.</p>
2548 <pre class="doc_code">
2549 a: store undef -> %X
2550 b: store %X -> undef
2556 <p>These examples reiterate the <tt>fdiv</tt> example: a store <em>of</em> an
2557 undefined value can be assumed to not have any effect; we can assume that the
2558 value is overwritten with bits that happen to match what was already there.
2559 However, a store <em>to</em> an undefined location could clobber arbitrary
2560 memory, therefore, it has undefined behavior.</p>
2564 <!-- ======================================================================= -->
2566 <a name="poisonvalues">Poison Values</a>
2571 <p>Poison values are similar to <a href="#undefvalues">undef values</a>, however
2572 they also represent the fact that an instruction or constant expression which
2573 cannot evoke side effects has nevertheless detected a condition which results
2574 in undefined behavior.</p>
2576 <p>There is currently no way of representing a poison value in the IR; they
2577 only exist when produced by operations such as
2578 <a href="#i_add"><tt>add</tt></a> with the <tt>nsw</tt> flag.</p>
2580 <p>Poison value behavior is defined in terms of value <i>dependence</i>:</p>
2583 <li>Values other than <a href="#i_phi"><tt>phi</tt></a> nodes depend on
2584 their operands.</li>
2586 <li><a href="#i_phi"><tt>Phi</tt></a> nodes depend on the operand corresponding
2587 to their dynamic predecessor basic block.</li>
2589 <li>Function arguments depend on the corresponding actual argument values in
2590 the dynamic callers of their functions.</li>
2592 <li><a href="#i_call"><tt>Call</tt></a> instructions depend on the
2593 <a href="#i_ret"><tt>ret</tt></a> instructions that dynamically transfer
2594 control back to them.</li>
2596 <li><a href="#i_invoke"><tt>Invoke</tt></a> instructions depend on the
2597 <a href="#i_ret"><tt>ret</tt></a>, <a href="#i_resume"><tt>resume</tt></a>,
2598 or exception-throwing call instructions that dynamically transfer control
2601 <li>Non-volatile loads and stores depend on the most recent stores to all of the
2602 referenced memory addresses, following the order in the IR
2603 (including loads and stores implied by intrinsics such as
2604 <a href="#int_memcpy"><tt>@llvm.memcpy</tt></a>.)</li>
2606 <!-- TODO: In the case of multiple threads, this only applies if the store
2607 "happens-before" the load or store. -->
2609 <!-- TODO: floating-point exception state -->
2611 <li>An instruction with externally visible side effects depends on the most
2612 recent preceding instruction with externally visible side effects, following
2613 the order in the IR. (This includes
2614 <a href="#volatile">volatile operations</a>.)</li>
2616 <li>An instruction <i>control-depends</i> on a
2617 <a href="#terminators">terminator instruction</a>
2618 if the terminator instruction has multiple successors and the instruction
2619 is always executed when control transfers to one of the successors, and
2620 may not be executed when control is transferred to another.</li>
2622 <li>Additionally, an instruction also <i>control-depends</i> on a terminator
2623 instruction if the set of instructions it otherwise depends on would be
2624 different if the terminator had transferred control to a different
2627 <li>Dependence is transitive.</li>
2631 <p>Poison Values have the same behavior as <a href="#undefvalues">undef values</a>,
2632 with the additional affect that any instruction which has a <i>dependence</i>
2633 on a poison value has undefined behavior.</p>
2635 <p>Here are some examples:</p>
2637 <pre class="doc_code">
2639 %poison = sub nuw i32 0, 1 ; Results in a poison value.
2640 %still_poison = and i32 %poison, 0 ; 0, but also poison.
2641 %poison_yet_again = getelementptr i32* @h, i32 %still_poison
2642 store i32 0, i32* %poison_yet_again ; memory at @h[0] is poisoned
2644 store i32 %poison, i32* @g ; Poison value stored to memory.
2645 %poison2 = load i32* @g ; Poison value loaded back from memory.
2647 store volatile i32 %poison, i32* @g ; External observation; undefined behavior.
2649 %narrowaddr = bitcast i32* @g to i16*
2650 %wideaddr = bitcast i32* @g to i64*
2651 %poison3 = load i16* %narrowaddr ; Returns a poison value.
2652 %poison4 = load i64* %wideaddr ; Returns a poison value.
2654 %cmp = icmp slt i32 %poison, 0 ; Returns a poison value.
2655 br i1 %cmp, label %true, label %end ; Branch to either destination.
2658 store volatile i32 0, i32* @g ; This is control-dependent on %cmp, so
2659 ; it has undefined behavior.
2663 %p = phi i32 [ 0, %entry ], [ 1, %true ]
2664 ; Both edges into this PHI are
2665 ; control-dependent on %cmp, so this
2666 ; always results in a poison value.
2668 store volatile i32 0, i32* @g ; This would depend on the store in %true
2669 ; if %cmp is true, or the store in %entry
2670 ; otherwise, so this is undefined behavior.
2672 br i1 %cmp, label %second_true, label %second_end
2673 ; The same branch again, but this time the
2674 ; true block doesn't have side effects.
2681 store volatile i32 0, i32* @g ; This time, the instruction always depends
2682 ; on the store in %end. Also, it is
2683 ; control-equivalent to %end, so this is
2684 ; well-defined (ignoring earlier undefined
2685 ; behavior in this example).
2690 <!-- ======================================================================= -->
2692 <a name="blockaddress">Addresses of Basic Blocks</a>
2697 <p><b><tt>blockaddress(@function, %block)</tt></b></p>
2699 <p>The '<tt>blockaddress</tt>' constant computes the address of the specified
2700 basic block in the specified function, and always has an i8* type. Taking
2701 the address of the entry block is illegal.</p>
2703 <p>This value only has defined behavior when used as an operand to the
2704 '<a href="#i_indirectbr"><tt>indirectbr</tt></a>' instruction, or for
2705 comparisons against null. Pointer equality tests between labels addresses
2706 results in undefined behavior — though, again, comparison against null
2707 is ok, and no label is equal to the null pointer. This may be passed around
2708 as an opaque pointer sized value as long as the bits are not inspected. This
2709 allows <tt>ptrtoint</tt> and arithmetic to be performed on these values so
2710 long as the original value is reconstituted before the <tt>indirectbr</tt>
2713 <p>Finally, some targets may provide defined semantics when using the value as
2714 the operand to an inline assembly, but that is target specific.</p>
2719 <!-- ======================================================================= -->
2721 <a name="constantexprs">Constant Expressions</a>
2726 <p>Constant expressions are used to allow expressions involving other constants
2727 to be used as constants. Constant expressions may be of
2728 any <a href="#t_firstclass">first class</a> type and may involve any LLVM
2729 operation that does not have side effects (e.g. load and call are not
2730 supported). The following is the syntax for constant expressions:</p>
2733 <dt><b><tt>trunc (CST to TYPE)</tt></b></dt>
2734 <dd>Truncate a constant to another type. The bit size of CST must be larger
2735 than the bit size of TYPE. Both types must be integers.</dd>
2737 <dt><b><tt>zext (CST to TYPE)</tt></b></dt>
2738 <dd>Zero extend a constant to another type. The bit size of CST must be
2739 smaller than the bit size of TYPE. Both types must be integers.</dd>
2741 <dt><b><tt>sext (CST to TYPE)</tt></b></dt>
2742 <dd>Sign extend a constant to another type. The bit size of CST must be
2743 smaller than the bit size of TYPE. Both types must be integers.</dd>
2745 <dt><b><tt>fptrunc (CST to TYPE)</tt></b></dt>
2746 <dd>Truncate a floating point constant to another floating point type. The
2747 size of CST must be larger than the size of TYPE. Both types must be
2748 floating point.</dd>
2750 <dt><b><tt>fpext (CST to TYPE)</tt></b></dt>
2751 <dd>Floating point extend a constant to another type. The size of CST must be
2752 smaller or equal to the size of TYPE. Both types must be floating
2755 <dt><b><tt>fptoui (CST to TYPE)</tt></b></dt>
2756 <dd>Convert a floating point constant to the corresponding unsigned integer
2757 constant. TYPE must be a scalar or vector integer type. CST must be of
2758 scalar or vector floating point type. Both CST and TYPE must be scalars,
2759 or vectors of the same number of elements. If the value won't fit in the
2760 integer type, the results are undefined.</dd>
2762 <dt><b><tt>fptosi (CST to TYPE)</tt></b></dt>
2763 <dd>Convert a floating point constant to the corresponding signed integer
2764 constant. TYPE must be a scalar or vector integer type. CST must be of
2765 scalar or vector floating point type. Both CST and TYPE must be scalars,
2766 or vectors of the same number of elements. If the value won't fit in the
2767 integer type, the results are undefined.</dd>
2769 <dt><b><tt>uitofp (CST to TYPE)</tt></b></dt>
2770 <dd>Convert an unsigned integer constant to the corresponding floating point
2771 constant. TYPE must be a scalar or vector floating point type. CST must be
2772 of scalar or vector integer type. Both CST and TYPE must be scalars, or
2773 vectors of the same number of elements. If the value won't fit in the
2774 floating point type, the results are undefined.</dd>
2776 <dt><b><tt>sitofp (CST to TYPE)</tt></b></dt>
2777 <dd>Convert a signed integer constant to the corresponding floating point
2778 constant. TYPE must be a scalar or vector floating point type. CST must be
2779 of scalar or vector integer type. Both CST and TYPE must be scalars, or
2780 vectors of the same number of elements. If the value won't fit in the
2781 floating point type, the results are undefined.</dd>
2783 <dt><b><tt>ptrtoint (CST to TYPE)</tt></b></dt>
2784 <dd>Convert a pointer typed constant to the corresponding integer constant
2785 <tt>TYPE</tt> must be an integer type. <tt>CST</tt> must be of pointer
2786 type. The <tt>CST</tt> value is zero extended, truncated, or unchanged to
2787 make it fit in <tt>TYPE</tt>.</dd>
2789 <dt><b><tt>inttoptr (CST to TYPE)</tt></b></dt>
2790 <dd>Convert an integer constant to a pointer constant. TYPE must be a pointer
2791 type. CST must be of integer type. The CST value is zero extended,
2792 truncated, or unchanged to make it fit in a pointer size. This one is
2793 <i>really</i> dangerous!</dd>
2795 <dt><b><tt>bitcast (CST to TYPE)</tt></b></dt>
2796 <dd>Convert a constant, CST, to another TYPE. The constraints of the operands
2797 are the same as those for the <a href="#i_bitcast">bitcast
2798 instruction</a>.</dd>
2800 <dt><b><tt>getelementptr (CSTPTR, IDX0, IDX1, ...)</tt></b></dt>
2801 <dt><b><tt>getelementptr inbounds (CSTPTR, IDX0, IDX1, ...)</tt></b></dt>
2802 <dd>Perform the <a href="#i_getelementptr">getelementptr operation</a> on
2803 constants. As with the <a href="#i_getelementptr">getelementptr</a>
2804 instruction, the index list may have zero or more indexes, which are
2805 required to make sense for the type of "CSTPTR".</dd>
2807 <dt><b><tt>select (COND, VAL1, VAL2)</tt></b></dt>
2808 <dd>Perform the <a href="#i_select">select operation</a> on constants.</dd>
2810 <dt><b><tt>icmp COND (VAL1, VAL2)</tt></b></dt>
2811 <dd>Performs the <a href="#i_icmp">icmp operation</a> on constants.</dd>
2813 <dt><b><tt>fcmp COND (VAL1, VAL2)</tt></b></dt>
2814 <dd>Performs the <a href="#i_fcmp">fcmp operation</a> on constants.</dd>
2816 <dt><b><tt>extractelement (VAL, IDX)</tt></b></dt>
2817 <dd>Perform the <a href="#i_extractelement">extractelement operation</a> on
2820 <dt><b><tt>insertelement (VAL, ELT, IDX)</tt></b></dt>
2821 <dd>Perform the <a href="#i_insertelement">insertelement operation</a> on
2824 <dt><b><tt>shufflevector (VEC1, VEC2, IDXMASK)</tt></b></dt>
2825 <dd>Perform the <a href="#i_shufflevector">shufflevector operation</a> on
2828 <dt><b><tt>extractvalue (VAL, IDX0, IDX1, ...)</tt></b></dt>
2829 <dd>Perform the <a href="#i_extractvalue">extractvalue operation</a> on
2830 constants. The index list is interpreted in a similar manner as indices in
2831 a '<a href="#i_getelementptr">getelementptr</a>' operation. At least one
2832 index value must be specified.</dd>
2834 <dt><b><tt>insertvalue (VAL, ELT, IDX0, IDX1, ...)</tt></b></dt>
2835 <dd>Perform the <a href="#i_insertvalue">insertvalue operation</a> on
2836 constants. The index list is interpreted in a similar manner as indices in
2837 a '<a href="#i_getelementptr">getelementptr</a>' operation. At least one
2838 index value must be specified.</dd>
2840 <dt><b><tt>OPCODE (LHS, RHS)</tt></b></dt>
2841 <dd>Perform the specified operation of the LHS and RHS constants. OPCODE may
2842 be any of the <a href="#binaryops">binary</a>
2843 or <a href="#bitwiseops">bitwise binary</a> operations. The constraints
2844 on operands are the same as those for the corresponding instruction
2845 (e.g. no bitwise operations on floating point values are allowed).</dd>
2852 <!-- *********************************************************************** -->
2853 <h2><a name="othervalues">Other Values</a></h2>
2854 <!-- *********************************************************************** -->
2856 <!-- ======================================================================= -->
2858 <a name="inlineasm">Inline Assembler Expressions</a>
2863 <p>LLVM supports inline assembler expressions (as opposed
2864 to <a href="#moduleasm">Module-Level Inline Assembly</a>) through the use of
2865 a special value. This value represents the inline assembler as a string
2866 (containing the instructions to emit), a list of operand constraints (stored
2867 as a string), a flag that indicates whether or not the inline asm
2868 expression has side effects, and a flag indicating whether the function
2869 containing the asm needs to align its stack conservatively. An example
2870 inline assembler expression is:</p>
2872 <pre class="doc_code">
2873 i32 (i32) asm "bswap $0", "=r,r"
2876 <p>Inline assembler expressions may <b>only</b> be used as the callee operand of
2877 a <a href="#i_call"><tt>call</tt></a> or an
2878 <a href="#i_invoke"><tt>invoke</tt></a> instruction.
2879 Thus, typically we have:</p>
2881 <pre class="doc_code">
2882 %X = call i32 asm "<a href="#int_bswap">bswap</a> $0", "=r,r"(i32 %Y)
2885 <p>Inline asms with side effects not visible in the constraint list must be
2886 marked as having side effects. This is done through the use of the
2887 '<tt>sideeffect</tt>' keyword, like so:</p>
2889 <pre class="doc_code">
2890 call void asm sideeffect "eieio", ""()
2893 <p>In some cases inline asms will contain code that will not work unless the
2894 stack is aligned in some way, such as calls or SSE instructions on x86,
2895 yet will not contain code that does that alignment within the asm.
2896 The compiler should make conservative assumptions about what the asm might
2897 contain and should generate its usual stack alignment code in the prologue
2898 if the '<tt>alignstack</tt>' keyword is present:</p>
2900 <pre class="doc_code">
2901 call void asm alignstack "eieio", ""()
2904 <p>Inline asms also support using non-standard assembly dialects. The assumed
2905 dialect is ATT. When the '<tt>inteldialect</tt>' keyword is present, the
2906 inline asm is using the Intel dialect. Currently, ATT and Intel are the
2907 only supported dialects. An example is:</p>
2909 <pre class="doc_code">
2910 call void asm inteldialect "eieio", ""()
2913 <p>If multiple keywords appear the '<tt>sideeffect</tt>' keyword must come
2914 first, the '<tt>alignstack</tt>' keyword second and the
2915 '<tt>inteldialect</tt>' keyword last.</p>
2918 <p>TODO: The format of the asm and constraints string still need to be
2919 documented here. Constraints on what can be done (e.g. duplication, moving,
2920 etc need to be documented). This is probably best done by reference to
2921 another document that covers inline asm from a holistic perspective.</p>
2924 <!-- _______________________________________________________________________ -->
2926 <a name="inlineasm_md">Inline Asm Metadata</a>
2931 <p>The call instructions that wrap inline asm nodes may have a
2932 "<tt>!srcloc</tt>" MDNode attached to it that contains a list of constant
2933 integers. If present, the code generator will use the integer as the
2934 location cookie value when report errors through the <tt>LLVMContext</tt>
2935 error reporting mechanisms. This allows a front-end to correlate backend
2936 errors that occur with inline asm back to the source code that produced it.
2939 <pre class="doc_code">
2940 call void asm sideeffect "something bad", ""()<b>, !srcloc !42</b>
2942 !42 = !{ i32 1234567 }
2945 <p>It is up to the front-end to make sense of the magic numbers it places in the
2946 IR. If the MDNode contains multiple constants, the code generator will use
2947 the one that corresponds to the line of the asm that the error occurs on.</p>
2953 <!-- ======================================================================= -->
2955 <a name="metadata">Metadata Nodes and Metadata Strings</a>
2960 <p>LLVM IR allows metadata to be attached to instructions in the program that
2961 can convey extra information about the code to the optimizers and code
2962 generator. One example application of metadata is source-level debug
2963 information. There are two metadata primitives: strings and nodes. All
2964 metadata has the <tt>metadata</tt> type and is identified in syntax by a
2965 preceding exclamation point ('<tt>!</tt>').</p>
2967 <p>A metadata string is a string surrounded by double quotes. It can contain
2968 any character by escaping non-printable characters with "<tt>\xx</tt>" where
2969 "<tt>xx</tt>" is the two digit hex code. For example:
2970 "<tt>!"test\00"</tt>".</p>
2972 <p>Metadata nodes are represented with notation similar to structure constants
2973 (a comma separated list of elements, surrounded by braces and preceded by an
2974 exclamation point). Metadata nodes can have any values as their operand. For
2977 <div class="doc_code">
2979 !{ metadata !"test\00", i32 10}
2983 <p>A <a href="#namedmetadatastructure">named metadata</a> is a collection of
2984 metadata nodes, which can be looked up in the module symbol table. For
2987 <div class="doc_code">
2989 !foo = metadata !{!4, !3}
2993 <p>Metadata can be used as function arguments. Here <tt>llvm.dbg.value</tt>
2994 function is using two metadata arguments:</p>
2996 <div class="doc_code">
2998 call void @llvm.dbg.value(metadata !24, i64 0, metadata !25)
3002 <p>Metadata can be attached with an instruction. Here metadata <tt>!21</tt> is
3003 attached to the <tt>add</tt> instruction using the <tt>!dbg</tt>
3006 <div class="doc_code">
3008 %indvar.next = add i64 %indvar, 1, !dbg !21
3012 <p>More information about specific metadata nodes recognized by the optimizers
3013 and code generator is found below.</p>
3015 <!-- _______________________________________________________________________ -->
3017 <a name="tbaa">'<tt>tbaa</tt>' Metadata</a>
3022 <p>In LLVM IR, memory does not have types, so LLVM's own type system is not
3023 suitable for doing TBAA. Instead, metadata is added to the IR to describe
3024 a type system of a higher level language. This can be used to implement
3025 typical C/C++ TBAA, but it can also be used to implement custom alias
3026 analysis behavior for other languages.</p>
3028 <p>The current metadata format is very simple. TBAA metadata nodes have up to
3029 three fields, e.g.:</p>
3031 <div class="doc_code">
3033 !0 = metadata !{ metadata !"an example type tree" }
3034 !1 = metadata !{ metadata !"int", metadata !0 }
3035 !2 = metadata !{ metadata !"float", metadata !0 }
3036 !3 = metadata !{ metadata !"const float", metadata !2, i64 1 }
3040 <p>The first field is an identity field. It can be any value, usually
3041 a metadata string, which uniquely identifies the type. The most important
3042 name in the tree is the name of the root node. Two trees with
3043 different root node names are entirely disjoint, even if they
3044 have leaves with common names.</p>
3046 <p>The second field identifies the type's parent node in the tree, or
3047 is null or omitted for a root node. A type is considered to alias
3048 all of its descendants and all of its ancestors in the tree. Also,
3049 a type is considered to alias all types in other trees, so that
3050 bitcode produced from multiple front-ends is handled conservatively.</p>
3052 <p>If the third field is present, it's an integer which if equal to 1
3053 indicates that the type is "constant" (meaning
3054 <tt>pointsToConstantMemory</tt> should return true; see
3055 <a href="AliasAnalysis.html#OtherItfs">other useful
3056 <tt>AliasAnalysis</tt> methods</a>).</p>
3060 <!-- _______________________________________________________________________ -->
3062 <a name="tbaa.struct">'<tt>tbaa.struct</tt>' Metadata</a>
3067 <p>The <a href="#int_memcpy"><tt>llvm.memcpy</tt></a> is often used to implement
3068 aggregate assignment operations in C and similar languages, however it is
3069 defined to copy a contiguous region of memory, which is more than strictly
3070 necessary for aggregate types which contain holes due to padding. Also, it
3071 doesn't contain any TBAA information about the fields of the aggregate.</p>
3073 <p><tt>!tbaa.struct</tt> metadata can describe which memory subregions in a memcpy
3074 are padding and what the TBAA tags of the struct are.</p>
3076 <p>The current metadata format is very simple. <tt>!tbaa.struct</tt> metadata nodes
3077 are a list of operands which are in conceptual groups of three. For each
3078 group of three, the first operand gives the byte offset of a field in bytes,
3079 the second gives its size in bytes, and the third gives its
3082 <div class="doc_code">
3084 !4 = metadata !{ i64 0, i64 4, metadata !1, i64 8, i64 4, metadata !2 }
3088 <p>This describes a struct with two fields. The first is at offset 0 bytes
3089 with size 4 bytes, and has tbaa tag !1. The second is at offset 8 bytes
3090 and has size 4 bytes and has tbaa tag !2.</p>
3092 <p>Note that the fields need not be contiguous. In this example, there is a
3093 4 byte gap between the two fields. This gap represents padding which
3094 does not carry useful data and need not be preserved.</p>
3098 <!-- _______________________________________________________________________ -->
3100 <a name="fpmath">'<tt>fpmath</tt>' Metadata</a>
3105 <p><tt>fpmath</tt> metadata may be attached to any instruction of floating point
3106 type. It can be used to express the maximum acceptable error in the result of
3107 that instruction, in ULPs, thus potentially allowing the compiler to use a
3108 more efficient but less accurate method of computing it. ULP is defined as
3113 <p>If <tt>x</tt> is a real number that lies between two finite consecutive
3114 floating-point numbers <tt>a</tt> and <tt>b</tt>, without being equal to one
3115 of them, then <tt>ulp(x) = |b - a|</tt>, otherwise <tt>ulp(x)</tt> is the
3116 distance between the two non-equal finite floating-point numbers nearest
3117 <tt>x</tt>. Moreover, <tt>ulp(NaN)</tt> is <tt>NaN</tt>.</p>
3121 <p>The metadata node shall consist of a single positive floating point number
3122 representing the maximum relative error, for example:</p>
3124 <div class="doc_code">
3126 !0 = metadata !{ float 2.5 } ; maximum acceptable inaccuracy is 2.5 ULPs
3132 <!-- _______________________________________________________________________ -->
3134 <a name="range">'<tt>range</tt>' Metadata</a>
3138 <p><tt>range</tt> metadata may be attached only to loads of integer types. It
3139 expresses the possible ranges the loaded value is in. The ranges are
3140 represented with a flattened list of integers. The loaded value is known to
3141 be in the union of the ranges defined by each consecutive pair. Each pair
3142 has the following properties:</p>
3144 <li>The type must match the type loaded by the instruction.</li>
3145 <li>The pair <tt>a,b</tt> represents the range <tt>[a,b)</tt>.</li>
3146 <li>Both <tt>a</tt> and <tt>b</tt> are constants.</li>
3147 <li>The range is allowed to wrap.</li>
3148 <li>The range should not represent the full or empty set. That is,
3149 <tt>a!=b</tt>. </li>
3151 <p> In addition, the pairs must be in signed order of the lower bound and
3152 they must be non-contiguous.</p>
3155 <div class="doc_code">
3157 %a = load i8* %x, align 1, !range !0 ; Can only be 0 or 1
3158 %b = load i8* %y, align 1, !range !1 ; Can only be 255 (-1), 0 or 1
3159 %c = load i8* %z, align 1, !range !2 ; Can only be 0, 1, 3, 4 or 5
3160 %d = load i8* %z, align 1, !range !3 ; Can only be -2, -1, 3, 4 or 5
3162 !0 = metadata !{ i8 0, i8 2 }
3163 !1 = metadata !{ i8 255, i8 2 }
3164 !2 = metadata !{ i8 0, i8 2, i8 3, i8 6 }
3165 !3 = metadata !{ i8 -2, i8 0, i8 3, i8 6 }
3173 <!-- *********************************************************************** -->
3175 <a name="module_flags">Module Flags Metadata</a>
3177 <!-- *********************************************************************** -->
3181 <p>Information about the module as a whole is difficult to convey to LLVM's
3182 subsystems. The LLVM IR isn't sufficient to transmit this
3183 information. The <tt>llvm.module.flags</tt> named metadata exists in order to
3184 facilitate this. These flags are in the form of key / value pairs —
3185 much like a dictionary — making it easy for any subsystem who cares
3186 about a flag to look it up.</p>
3188 <p>The <tt>llvm.module.flags</tt> metadata contains a list of metadata
3189 triplets. Each triplet has the following form:</p>
3192 <li>The first element is a <i>behavior</i> flag, which specifies the behavior
3193 when two (or more) modules are merged together, and it encounters two (or
3194 more) metadata with the same ID. The supported behaviors are described
3197 <li>The second element is a metadata string that is a unique ID for the
3198 metadata. How each ID is interpreted is documented below.</li>
3200 <li>The third element is the value of the flag.</li>
3203 <p>When two (or more) modules are merged together, the resulting
3204 <tt>llvm.module.flags</tt> metadata is the union of the
3205 modules' <tt>llvm.module.flags</tt> metadata. The only exception being a flag
3206 with the <i>Override</i> behavior, which may override another flag's value
3209 <p>The following behaviors are supported:</p>
3211 <table border="1" cellspacing="0" cellpadding="4">
3221 <dt><b>Error</b></dt>
3222 <dd>Emits an error if two values disagree. It is an error to have an ID
3223 with both an Error and a Warning behavior.</dd>
3231 <dt><b>Warning</b></dt>
3232 <dd>Emits a warning if two values disagree.</dd>
3240 <dt><b>Require</b></dt>
3241 <dd>Emits an error when the specified value is not present or doesn't
3242 have the specified value. It is an error for two (or more)
3243 <tt>llvm.module.flags</tt> with the same ID to have the Require
3244 behavior but different values. There may be multiple Require flags
3253 <dt><b>Override</b></dt>
3254 <dd>Uses the specified value if the two values disagree. It is an
3255 error for two (or more) <tt>llvm.module.flags</tt> with the same
3256 ID to have the Override behavior but different values.</dd>
3263 <p>An example of module flags:</p>
3265 <pre class="doc_code">
3266 !0 = metadata !{ i32 1, metadata !"foo", i32 1 }
3267 !1 = metadata !{ i32 4, metadata !"bar", i32 37 }
3268 !2 = metadata !{ i32 2, metadata !"qux", i32 42 }
3269 !3 = metadata !{ i32 3, metadata !"qux",
3271 metadata !"foo", i32 1
3274 !llvm.module.flags = !{ !0, !1, !2, !3 }
3278 <li><p>Metadata <tt>!0</tt> has the ID <tt>!"foo"</tt> and the value '1'. The
3279 behavior if two or more <tt>!"foo"</tt> flags are seen is to emit an
3280 error if their values are not equal.</p></li>
3282 <li><p>Metadata <tt>!1</tt> has the ID <tt>!"bar"</tt> and the value '37'. The
3283 behavior if two or more <tt>!"bar"</tt> flags are seen is to use the
3284 value '37' if their values are not equal.</p></li>
3286 <li><p>Metadata <tt>!2</tt> has the ID <tt>!"qux"</tt> and the value '42'. The
3287 behavior if two or more <tt>!"qux"</tt> flags are seen is to emit a
3288 warning if their values are not equal.</p></li>
3290 <li><p>Metadata <tt>!3</tt> has the ID <tt>!"qux"</tt> and the value:</p>
3292 <pre class="doc_code">
3293 metadata !{ metadata !"foo", i32 1 }
3296 <p>The behavior is to emit an error if the <tt>llvm.module.flags</tt> does
3297 not contain a flag with the ID <tt>!"foo"</tt> that has the value
3298 '1'. If two or more <tt>!"qux"</tt> flags exist, then they must have
3299 the same value or an error will be issued.</p></li>
3303 <!-- ======================================================================= -->
3305 <a name="objc_gc_flags">Objective-C Garbage Collection Module Flags Metadata</a>
3310 <p>On the Mach-O platform, Objective-C stores metadata about garbage collection
3311 in a special section called "image info". The metadata consists of a version
3312 number and a bitmask specifying what types of garbage collection are
3313 supported (if any) by the file. If two or more modules are linked together
3314 their garbage collection metadata needs to be merged rather than appended
3317 <p>The Objective-C garbage collection module flags metadata consists of the
3318 following key-value pairs:</p>
3320 <table border="1" cellspacing="0" cellpadding="4">
3328 <td><tt>Objective-C Version</tt></td>
3329 <td align="left"><b>[Required]</b> — The Objective-C ABI
3330 version. Valid values are 1 and 2.</td>
3333 <td><tt>Objective-C Image Info Version</tt></td>
3334 <td align="left"><b>[Required]</b> — The version of the image info
3335 section. Currently always 0.</td>
3338 <td><tt>Objective-C Image Info Section</tt></td>
3339 <td align="left"><b>[Required]</b> — The section to place the
3340 metadata. Valid values are <tt>"__OBJC, __image_info, regular"</tt> for
3341 Objective-C ABI version 1, and <tt>"__DATA,__objc_imageinfo, regular,
3342 no_dead_strip"</tt> for Objective-C ABI version 2.</td>
3345 <td><tt>Objective-C Garbage Collection</tt></td>
3346 <td align="left"><b>[Required]</b> — Specifies whether garbage
3347 collection is supported or not. Valid values are 0, for no garbage
3348 collection, and 2, for garbage collection supported.</td>
3351 <td><tt>Objective-C GC Only</tt></td>
3352 <td align="left"><b>[Optional]</b> — Specifies that only garbage
3353 collection is supported. If present, its value must be 6. This flag
3354 requires that the <tt>Objective-C Garbage Collection</tt> flag have the
3360 <p>Some important flag interactions:</p>
3363 <li>If a module with <tt>Objective-C Garbage Collection</tt> set to 0 is
3364 merged with a module with <tt>Objective-C Garbage Collection</tt> set to
3365 2, then the resulting module has the <tt>Objective-C Garbage
3366 Collection</tt> flag set to 0.</li>
3368 <li>A module with <tt>Objective-C Garbage Collection</tt> set to 0 cannot be
3369 merged with a module with <tt>Objective-C GC Only</tt> set to 6.</li>
3376 <!-- *********************************************************************** -->
3378 <a name="intrinsic_globals">Intrinsic Global Variables</a>
3380 <!-- *********************************************************************** -->
3382 <p>LLVM has a number of "magic" global variables that contain data that affect
3383 code generation or other IR semantics. These are documented here. All globals
3384 of this sort should have a section specified as "<tt>llvm.metadata</tt>". This
3385 section and all globals that start with "<tt>llvm.</tt>" are reserved for use
3388 <!-- ======================================================================= -->
3390 <a name="intg_used">The '<tt>llvm.used</tt>' Global Variable</a>
3395 <p>The <tt>@llvm.used</tt> global is an array with i8* element type which has <a
3396 href="#linkage_appending">appending linkage</a>. This array contains a list of
3397 pointers to global variables and functions which may optionally have a pointer
3398 cast formed of bitcast or getelementptr. For example, a legal use of it is:</p>
3400 <div class="doc_code">
3405 @llvm.used = appending global [2 x i8*] [
3407 i8* bitcast (i32* @Y to i8*)
3408 ], section "llvm.metadata"
3412 <p>If a global variable appears in the <tt>@llvm.used</tt> list, then the
3413 compiler, assembler, and linker are required to treat the symbol as if there
3414 is a reference to the global that it cannot see. For example, if a variable
3415 has internal linkage and no references other than that from
3416 the <tt>@llvm.used</tt> list, it cannot be deleted. This is commonly used to
3417 represent references from inline asms and other things the compiler cannot
3418 "see", and corresponds to "<tt>attribute((used))</tt>" in GNU C.</p>
3420 <p>On some targets, the code generator must emit a directive to the assembler or
3421 object file to prevent the assembler and linker from molesting the
3426 <!-- ======================================================================= -->
3428 <a name="intg_compiler_used">
3429 The '<tt>llvm.compiler.used</tt>' Global Variable
3435 <p>The <tt>@llvm.compiler.used</tt> directive is the same as the
3436 <tt>@llvm.used</tt> directive, except that it only prevents the compiler from
3437 touching the symbol. On targets that support it, this allows an intelligent
3438 linker to optimize references to the symbol without being impeded as it would
3439 be by <tt>@llvm.used</tt>.</p>
3441 <p>This is a rare construct that should only be used in rare circumstances, and
3442 should not be exposed to source languages.</p>
3446 <!-- ======================================================================= -->
3448 <a name="intg_global_ctors">The '<tt>llvm.global_ctors</tt>' Global Variable</a>
3453 <div class="doc_code">
3455 %0 = type { i32, void ()* }
3456 @llvm.global_ctors = appending global [1 x %0] [%0 { i32 65535, void ()* @ctor }]
3460 <p>The <tt>@llvm.global_ctors</tt> array contains a list of constructor
3461 functions and associated priorities. The functions referenced by this array
3462 will be called in ascending order of priority (i.e. lowest first) when the
3463 module is loaded. The order of functions with the same priority is not
3468 <!-- ======================================================================= -->
3470 <a name="intg_global_dtors">The '<tt>llvm.global_dtors</tt>' Global Variable</a>
3475 <div class="doc_code">
3477 %0 = type { i32, void ()* }
3478 @llvm.global_dtors = appending global [1 x %0] [%0 { i32 65535, void ()* @dtor }]
3482 <p>The <tt>@llvm.global_dtors</tt> array contains a list of destructor functions
3483 and associated priorities. The functions referenced by this array will be
3484 called in descending order of priority (i.e. highest first) when the module
3485 is loaded. The order of functions with the same priority is not defined.</p>
3491 <!-- *********************************************************************** -->
3492 <h2><a name="instref">Instruction Reference</a></h2>
3493 <!-- *********************************************************************** -->
3497 <p>The LLVM instruction set consists of several different classifications of
3498 instructions: <a href="#terminators">terminator
3499 instructions</a>, <a href="#binaryops">binary instructions</a>,
3500 <a href="#bitwiseops">bitwise binary instructions</a>,
3501 <a href="#memoryops">memory instructions</a>, and
3502 <a href="#otherops">other instructions</a>.</p>
3504 <!-- ======================================================================= -->
3506 <a name="terminators">Terminator Instructions</a>
3511 <p>As mentioned <a href="#functionstructure">previously</a>, every basic block
3512 in a program ends with a "Terminator" instruction, which indicates which
3513 block should be executed after the current block is finished. These
3514 terminator instructions typically yield a '<tt>void</tt>' value: they produce
3515 control flow, not values (the one exception being the
3516 '<a href="#i_invoke"><tt>invoke</tt></a>' instruction).</p>
3518 <p>The terminator instructions are:
3519 '<a href="#i_ret"><tt>ret</tt></a>',
3520 '<a href="#i_br"><tt>br</tt></a>',
3521 '<a href="#i_switch"><tt>switch</tt></a>',
3522 '<a href="#i_indirectbr"><tt>indirectbr</tt></a>',
3523 '<a href="#i_invoke"><tt>invoke</tt></a>',
3524 '<a href="#i_resume"><tt>resume</tt></a>', and
3525 '<a href="#i_unreachable"><tt>unreachable</tt></a>'.</p>
3527 <!-- _______________________________________________________________________ -->
3529 <a name="i_ret">'<tt>ret</tt>' Instruction</a>
3536 ret <type> <value> <i>; Return a value from a non-void function</i>
3537 ret void <i>; Return from void function</i>
3541 <p>The '<tt>ret</tt>' instruction is used to return control flow (and optionally
3542 a value) from a function back to the caller.</p>
3544 <p>There are two forms of the '<tt>ret</tt>' instruction: one that returns a
3545 value and then causes control flow, and one that just causes control flow to
3549 <p>The '<tt>ret</tt>' instruction optionally accepts a single argument, the
3550 return value. The type of the return value must be a
3551 '<a href="#t_firstclass">first class</a>' type.</p>
3553 <p>A function is not <a href="#wellformed">well formed</a> if it it has a
3554 non-void return type and contains a '<tt>ret</tt>' instruction with no return
3555 value or a return value with a type that does not match its type, or if it
3556 has a void return type and contains a '<tt>ret</tt>' instruction with a
3560 <p>When the '<tt>ret</tt>' instruction is executed, control flow returns back to
3561 the calling function's context. If the caller is a
3562 "<a href="#i_call"><tt>call</tt></a>" instruction, execution continues at the
3563 instruction after the call. If the caller was an
3564 "<a href="#i_invoke"><tt>invoke</tt></a>" instruction, execution continues at
3565 the beginning of the "normal" destination block. If the instruction returns
3566 a value, that value shall set the call or invoke instruction's return
3571 ret i32 5 <i>; Return an integer value of 5</i>
3572 ret void <i>; Return from a void function</i>
3573 ret { i32, i8 } { i32 4, i8 2 } <i>; Return a struct of values 4 and 2</i>
3577 <!-- _______________________________________________________________________ -->
3579 <a name="i_br">'<tt>br</tt>' Instruction</a>
3586 br i1 <cond>, label <iftrue>, label <iffalse>
3587 br label <dest> <i>; Unconditional branch</i>
3591 <p>The '<tt>br</tt>' instruction is used to cause control flow to transfer to a
3592 different basic block in the current function. There are two forms of this
3593 instruction, corresponding to a conditional branch and an unconditional
3597 <p>The conditional branch form of the '<tt>br</tt>' instruction takes a single
3598 '<tt>i1</tt>' value and two '<tt>label</tt>' values. The unconditional form
3599 of the '<tt>br</tt>' instruction takes a single '<tt>label</tt>' value as a
3603 <p>Upon execution of a conditional '<tt>br</tt>' instruction, the '<tt>i1</tt>'
3604 argument is evaluated. If the value is <tt>true</tt>, control flows to the
3605 '<tt>iftrue</tt>' <tt>label</tt> argument. If "cond" is <tt>false</tt>,
3606 control flows to the '<tt>iffalse</tt>' <tt>label</tt> argument.</p>
3611 %cond = <a href="#i_icmp">icmp</a> eq i32 %a, %b
3612 br i1 %cond, label %IfEqual, label %IfUnequal
3614 <a href="#i_ret">ret</a> i32 1
3616 <a href="#i_ret">ret</a> i32 0
3621 <!-- _______________________________________________________________________ -->
3623 <a name="i_switch">'<tt>switch</tt>' Instruction</a>
3630 switch <intty> <value>, label <defaultdest> [ <intty> <val>, label <dest> ... ]
3634 <p>The '<tt>switch</tt>' instruction is used to transfer control flow to one of
3635 several different places. It is a generalization of the '<tt>br</tt>'
3636 instruction, allowing a branch to occur to one of many possible
3640 <p>The '<tt>switch</tt>' instruction uses three parameters: an integer
3641 comparison value '<tt>value</tt>', a default '<tt>label</tt>' destination,
3642 and an array of pairs of comparison value constants and '<tt>label</tt>'s.
3643 The table is not allowed to contain duplicate constant entries.</p>
3646 <p>The <tt>switch</tt> instruction specifies a table of values and
3647 destinations. When the '<tt>switch</tt>' instruction is executed, this table
3648 is searched for the given value. If the value is found, control flow is
3649 transferred to the corresponding destination; otherwise, control flow is
3650 transferred to the default destination.</p>
3652 <h5>Implementation:</h5>
3653 <p>Depending on properties of the target machine and the particular
3654 <tt>switch</tt> instruction, this instruction may be code generated in
3655 different ways. For example, it could be generated as a series of chained
3656 conditional branches or with a lookup table.</p>
3660 <i>; Emulate a conditional br instruction</i>
3661 %Val = <a href="#i_zext">zext</a> i1 %value to i32
3662 switch i32 %Val, label %truedest [ i32 0, label %falsedest ]
3664 <i>; Emulate an unconditional br instruction</i>
3665 switch i32 0, label %dest [ ]
3667 <i>; Implement a jump table:</i>
3668 switch i32 %val, label %otherwise [ i32 0, label %onzero
3670 i32 2, label %ontwo ]
3676 <!-- _______________________________________________________________________ -->
3678 <a name="i_indirectbr">'<tt>indirectbr</tt>' Instruction</a>
3685 indirectbr <somety>* <address>, [ label <dest1>, label <dest2>, ... ]
3690 <p>The '<tt>indirectbr</tt>' instruction implements an indirect branch to a label
3691 within the current function, whose address is specified by
3692 "<tt>address</tt>". Address must be derived from a <a
3693 href="#blockaddress">blockaddress</a> constant.</p>
3697 <p>The '<tt>address</tt>' argument is the address of the label to jump to. The
3698 rest of the arguments indicate the full set of possible destinations that the
3699 address may point to. Blocks are allowed to occur multiple times in the
3700 destination list, though this isn't particularly useful.</p>
3702 <p>This destination list is required so that dataflow analysis has an accurate
3703 understanding of the CFG.</p>
3707 <p>Control transfers to the block specified in the address argument. All
3708 possible destination blocks must be listed in the label list, otherwise this
3709 instruction has undefined behavior. This implies that jumps to labels
3710 defined in other functions have undefined behavior as well.</p>
3712 <h5>Implementation:</h5>
3714 <p>This is typically implemented with a jump through a register.</p>
3718 indirectbr i8* %Addr, [ label %bb1, label %bb2, label %bb3 ]
3724 <!-- _______________________________________________________________________ -->
3726 <a name="i_invoke">'<tt>invoke</tt>' Instruction</a>
3733 <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>]
3734 to label <normal label> unwind label <exception label>
3738 <p>The '<tt>invoke</tt>' instruction causes control to transfer to a specified
3739 function, with the possibility of control flow transfer to either the
3740 '<tt>normal</tt>' label or the '<tt>exception</tt>' label. If the callee
3741 function returns with the "<tt><a href="#i_ret">ret</a></tt>" instruction,
3742 control flow will return to the "normal" label. If the callee (or any
3743 indirect callees) returns via the "<a href="#i_resume"><tt>resume</tt></a>"
3744 instruction or other exception handling mechanism, control is interrupted and
3745 continued at the dynamically nearest "exception" label.</p>
3747 <p>The '<tt>exception</tt>' label is a
3748 <i><a href="ExceptionHandling.html#overview">landing pad</a></i> for the
3749 exception. As such, '<tt>exception</tt>' label is required to have the
3750 "<a href="#i_landingpad"><tt>landingpad</tt></a>" instruction, which contains
3751 the information about the behavior of the program after unwinding
3752 happens, as its first non-PHI instruction. The restrictions on the
3753 "<tt>landingpad</tt>" instruction's tightly couples it to the
3754 "<tt>invoke</tt>" instruction, so that the important information contained
3755 within the "<tt>landingpad</tt>" instruction can't be lost through normal
3759 <p>This instruction requires several arguments:</p>
3762 <li>The optional "cconv" marker indicates which <a href="#callingconv">calling
3763 convention</a> the call should use. If none is specified, the call
3764 defaults to using C calling conventions.</li>
3766 <li>The optional <a href="#paramattrs">Parameter Attributes</a> list for
3767 return values. Only '<tt>zeroext</tt>', '<tt>signext</tt>', and
3768 '<tt>inreg</tt>' attributes are valid here.</li>
3770 <li>'<tt>ptr to function ty</tt>': shall be the signature of the pointer to
3771 function value being invoked. In most cases, this is a direct function
3772 invocation, but indirect <tt>invoke</tt>s are just as possible, branching
3773 off an arbitrary pointer to function value.</li>
3775 <li>'<tt>function ptr val</tt>': An LLVM value containing a pointer to a
3776 function to be invoked. </li>
3778 <li>'<tt>function args</tt>': argument list whose types match the function
3779 signature argument types and parameter attributes. All arguments must be
3780 of <a href="#t_firstclass">first class</a> type. If the function
3781 signature indicates the function accepts a variable number of arguments,
3782 the extra arguments can be specified.</li>
3784 <li>'<tt>normal label</tt>': the label reached when the called function
3785 executes a '<tt><a href="#i_ret">ret</a></tt>' instruction. </li>
3787 <li>'<tt>exception label</tt>': the label reached when a callee returns via
3788 the <a href="#i_resume"><tt>resume</tt></a> instruction or other exception
3789 handling mechanism.</li>
3791 <li>The optional <a href="#fnattrs">function attributes</a> list. Only
3792 '<tt>noreturn</tt>', '<tt>nounwind</tt>', '<tt>readonly</tt>' and
3793 '<tt>readnone</tt>' attributes are valid here.</li>
3797 <p>This instruction is designed to operate as a standard
3798 '<tt><a href="#i_call">call</a></tt>' instruction in most regards. The
3799 primary difference is that it establishes an association with a label, which
3800 is used by the runtime library to unwind the stack.</p>
3802 <p>This instruction is used in languages with destructors to ensure that proper
3803 cleanup is performed in the case of either a <tt>longjmp</tt> or a thrown
3804 exception. Additionally, this is important for implementation of
3805 '<tt>catch</tt>' clauses in high-level languages that support them.</p>
3807 <p>For the purposes of the SSA form, the definition of the value returned by the
3808 '<tt>invoke</tt>' instruction is deemed to occur on the edge from the current
3809 block to the "normal" label. If the callee unwinds then no return value is
3814 %retval = invoke i32 @Test(i32 15) to label %Continue
3815 unwind label %TestCleanup <i>; {i32}:retval set</i>
3816 %retval = invoke <a href="#callingconv">coldcc</a> i32 %Testfnptr(i32 15) to label %Continue
3817 unwind label %TestCleanup <i>; {i32}:retval set</i>
3822 <!-- _______________________________________________________________________ -->
3825 <a name="i_resume">'<tt>resume</tt>' Instruction</a>
3832 resume <type> <value>
3836 <p>The '<tt>resume</tt>' instruction is a terminator instruction that has no
3840 <p>The '<tt>resume</tt>' instruction requires one argument, which must have the
3841 same type as the result of any '<tt>landingpad</tt>' instruction in the same
3845 <p>The '<tt>resume</tt>' instruction resumes propagation of an existing
3846 (in-flight) exception whose unwinding was interrupted with
3847 a <a href="#i_landingpad"><tt>landingpad</tt></a> instruction.</p>
3851 resume { i8*, i32 } %exn
3856 <!-- _______________________________________________________________________ -->
3859 <a name="i_unreachable">'<tt>unreachable</tt>' Instruction</a>
3870 <p>The '<tt>unreachable</tt>' instruction has no defined semantics. This
3871 instruction is used to inform the optimizer that a particular portion of the
3872 code is not reachable. This can be used to indicate that the code after a
3873 no-return function cannot be reached, and other facts.</p>
3876 <p>The '<tt>unreachable</tt>' instruction has no defined semantics.</p>
3882 <!-- ======================================================================= -->
3884 <a name="binaryops">Binary Operations</a>
3889 <p>Binary operators are used to do most of the computation in a program. They
3890 require two operands of the same type, execute an operation on them, and
3891 produce a single value. The operands might represent multiple data, as is
3892 the case with the <a href="#t_vector">vector</a> data type. The result value
3893 has the same type as its operands.</p>
3895 <p>There are several different binary operators:</p>
3897 <!-- _______________________________________________________________________ -->
3899 <a name="i_add">'<tt>add</tt>' Instruction</a>
3906 <result> = add <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3907 <result> = add nuw <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3908 <result> = add nsw <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3909 <result> = add nuw nsw <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3913 <p>The '<tt>add</tt>' instruction returns the sum of its two operands.</p>
3916 <p>The two arguments to the '<tt>add</tt>' instruction must
3917 be <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of
3918 integer values. Both arguments must have identical types.</p>
3921 <p>The value produced is the integer sum of the two operands.</p>
3923 <p>If the sum has unsigned overflow, the result returned is the mathematical
3924 result modulo 2<sup>n</sup>, where n is the bit width of the result.</p>
3926 <p>Because LLVM integers use a two's complement representation, this instruction
3927 is appropriate for both signed and unsigned integers.</p>
3929 <p><tt>nuw</tt> and <tt>nsw</tt> stand for "No Unsigned Wrap"
3930 and "No Signed Wrap", respectively. If the <tt>nuw</tt> and/or
3931 <tt>nsw</tt> keywords are present, the result value of the <tt>add</tt>
3932 is a <a href="#poisonvalues">poison value</a> if unsigned and/or signed overflow,
3933 respectively, occurs.</p>
3937 <result> = add i32 4, %var <i>; yields {i32}:result = 4 + %var</i>
3942 <!-- _______________________________________________________________________ -->
3944 <a name="i_fadd">'<tt>fadd</tt>' Instruction</a>
3951 <result> = fadd <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3955 <p>The '<tt>fadd</tt>' instruction returns the sum of its two operands.</p>
3958 <p>The two arguments to the '<tt>fadd</tt>' instruction must be
3959 <a href="#t_floating">floating point</a> or <a href="#t_vector">vector</a> of
3960 floating point values. Both arguments must have identical types.</p>
3963 <p>The value produced is the floating point sum of the two operands.</p>
3967 <result> = fadd float 4.0, %var <i>; yields {float}:result = 4.0 + %var</i>
3972 <!-- _______________________________________________________________________ -->
3974 <a name="i_sub">'<tt>sub</tt>' Instruction</a>
3981 <result> = sub <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3982 <result> = sub nuw <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3983 <result> = sub nsw <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3984 <result> = sub nuw nsw <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3988 <p>The '<tt>sub</tt>' instruction returns the difference of its two
3991 <p>Note that the '<tt>sub</tt>' instruction is used to represent the
3992 '<tt>neg</tt>' instruction present in most other intermediate
3993 representations.</p>
3996 <p>The two arguments to the '<tt>sub</tt>' instruction must
3997 be <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of
3998 integer values. Both arguments must have identical types.</p>
4001 <p>The value produced is the integer difference of the two operands.</p>
4003 <p>If the difference has unsigned overflow, the result returned is the
4004 mathematical result modulo 2<sup>n</sup>, where n is the bit width of the
4007 <p>Because LLVM integers use a two's complement representation, this instruction
4008 is appropriate for both signed and unsigned integers.</p>
4010 <p><tt>nuw</tt> and <tt>nsw</tt> stand for "No Unsigned Wrap"
4011 and "No Signed Wrap", respectively. If the <tt>nuw</tt> and/or
4012 <tt>nsw</tt> keywords are present, the result value of the <tt>sub</tt>
4013 is a <a href="#poisonvalues">poison value</a> if unsigned and/or signed overflow,
4014 respectively, occurs.</p>
4018 <result> = sub i32 4, %var <i>; yields {i32}:result = 4 - %var</i>
4019 <result> = sub i32 0, %val <i>; yields {i32}:result = -%var</i>
4024 <!-- _______________________________________________________________________ -->
4026 <a name="i_fsub">'<tt>fsub</tt>' Instruction</a>
4033 <result> = fsub <ty> <op1>, <op2> <i>; yields {ty}:result</i>
4037 <p>The '<tt>fsub</tt>' instruction returns the difference of its two
4040 <p>Note that the '<tt>fsub</tt>' instruction is used to represent the
4041 '<tt>fneg</tt>' instruction present in most other intermediate
4042 representations.</p>
4045 <p>The two arguments to the '<tt>fsub</tt>' instruction must be
4046 <a href="#t_floating">floating point</a> or <a href="#t_vector">vector</a> of
4047 floating point values. Both arguments must have identical types.</p>
4050 <p>The value produced is the floating point difference of the two operands.</p>
4054 <result> = fsub float 4.0, %var <i>; yields {float}:result = 4.0 - %var</i>
4055 <result> = fsub float -0.0, %val <i>; yields {float}:result = -%var</i>
4060 <!-- _______________________________________________________________________ -->
4062 <a name="i_mul">'<tt>mul</tt>' Instruction</a>
4069 <result> = mul <ty> <op1>, <op2> <i>; yields {ty}:result</i>
4070 <result> = mul nuw <ty> <op1>, <op2> <i>; yields {ty}:result</i>
4071 <result> = mul nsw <ty> <op1>, <op2> <i>; yields {ty}:result</i>
4072 <result> = mul nuw nsw <ty> <op1>, <op2> <i>; yields {ty}:result</i>
4076 <p>The '<tt>mul</tt>' instruction returns the product of its two operands.</p>
4079 <p>The two arguments to the '<tt>mul</tt>' instruction must
4080 be <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of
4081 integer values. Both arguments must have identical types.</p>
4084 <p>The value produced is the integer product of the two operands.</p>
4086 <p>If the result of the multiplication has unsigned overflow, the result
4087 returned is the mathematical result modulo 2<sup>n</sup>, where n is the bit
4088 width of the result.</p>
4090 <p>Because LLVM integers use a two's complement representation, and the result
4091 is the same width as the operands, this instruction returns the correct
4092 result for both signed and unsigned integers. If a full product
4093 (e.g. <tt>i32</tt>x<tt>i32</tt>-><tt>i64</tt>) is needed, the operands should
4094 be sign-extended or zero-extended as appropriate to the width of the full
4097 <p><tt>nuw</tt> and <tt>nsw</tt> stand for "No Unsigned Wrap"
4098 and "No Signed Wrap", respectively. If the <tt>nuw</tt> and/or
4099 <tt>nsw</tt> keywords are present, the result value of the <tt>mul</tt>
4100 is a <a href="#poisonvalues">poison value</a> if unsigned and/or signed overflow,
4101 respectively, occurs.</p>
4105 <result> = mul i32 4, %var <i>; yields {i32}:result = 4 * %var</i>
4110 <!-- _______________________________________________________________________ -->
4112 <a name="i_fmul">'<tt>fmul</tt>' Instruction</a>
4119 <result> = fmul <ty> <op1>, <op2> <i>; yields {ty}:result</i>
4123 <p>The '<tt>fmul</tt>' instruction returns the product of its two operands.</p>
4126 <p>The two arguments to the '<tt>fmul</tt>' instruction must be
4127 <a href="#t_floating">floating point</a> or <a href="#t_vector">vector</a> of
4128 floating point values. Both arguments must have identical types.</p>
4131 <p>The value produced is the floating point product of the two operands.</p>
4135 <result> = fmul float 4.0, %var <i>; yields {float}:result = 4.0 * %var</i>
4140 <!-- _______________________________________________________________________ -->
4142 <a name="i_udiv">'<tt>udiv</tt>' Instruction</a>
4149 <result> = udiv <ty> <op1>, <op2> <i>; yields {ty}:result</i>
4150 <result> = udiv exact <ty> <op1>, <op2> <i>; yields {ty}:result</i>
4154 <p>The '<tt>udiv</tt>' instruction returns the quotient of its two operands.</p>
4157 <p>The two arguments to the '<tt>udiv</tt>' instruction must be
4158 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
4159 values. Both arguments must have identical types.</p>
4162 <p>The value produced is the unsigned integer quotient of the two operands.</p>
4164 <p>Note that unsigned integer division and signed integer division are distinct
4165 operations; for signed integer division, use '<tt>sdiv</tt>'.</p>
4167 <p>Division by zero leads to undefined behavior.</p>
4169 <p>If the <tt>exact</tt> keyword is present, the result value of the
4170 <tt>udiv</tt> is a <a href="#poisonvalues">poison value</a> if %op1 is not a
4171 multiple of %op2 (as such, "((a udiv exact b) mul b) == a").</p>
4176 <result> = udiv i32 4, %var <i>; yields {i32}:result = 4 / %var</i>
4181 <!-- _______________________________________________________________________ -->
4183 <a name="i_sdiv">'<tt>sdiv</tt>' Instruction</a>
4190 <result> = sdiv <ty> <op1>, <op2> <i>; yields {ty}:result</i>
4191 <result> = sdiv exact <ty> <op1>, <op2> <i>; yields {ty}:result</i>
4195 <p>The '<tt>sdiv</tt>' instruction returns the quotient of its two operands.</p>
4198 <p>The two arguments to the '<tt>sdiv</tt>' instruction must be
4199 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
4200 values. Both arguments must have identical types.</p>
4203 <p>The value produced is the signed integer quotient of the two operands rounded
4206 <p>Note that signed integer division and unsigned integer division are distinct
4207 operations; for unsigned integer division, use '<tt>udiv</tt>'.</p>
4209 <p>Division by zero leads to undefined behavior. Overflow also leads to
4210 undefined behavior; this is a rare case, but can occur, for example, by doing
4211 a 32-bit division of -2147483648 by -1.</p>
4213 <p>If the <tt>exact</tt> keyword is present, the result value of the
4214 <tt>sdiv</tt> is a <a href="#poisonvalues">poison value</a> if the result would
4219 <result> = sdiv i32 4, %var <i>; yields {i32}:result = 4 / %var</i>
4224 <!-- _______________________________________________________________________ -->
4226 <a name="i_fdiv">'<tt>fdiv</tt>' Instruction</a>
4233 <result> = fdiv <ty> <op1>, <op2> <i>; yields {ty}:result</i>
4237 <p>The '<tt>fdiv</tt>' instruction returns the quotient of its two operands.</p>
4240 <p>The two arguments to the '<tt>fdiv</tt>' instruction must be
4241 <a href="#t_floating">floating point</a> or <a href="#t_vector">vector</a> of
4242 floating point values. Both arguments must have identical types.</p>
4245 <p>The value produced is the floating point quotient of the two operands.</p>
4249 <result> = fdiv float 4.0, %var <i>; yields {float}:result = 4.0 / %var</i>
4254 <!-- _______________________________________________________________________ -->
4256 <a name="i_urem">'<tt>urem</tt>' Instruction</a>
4263 <result> = urem <ty> <op1>, <op2> <i>; yields {ty}:result</i>
4267 <p>The '<tt>urem</tt>' instruction returns the remainder from the unsigned
4268 division of its two arguments.</p>
4271 <p>The two arguments to the '<tt>urem</tt>' instruction must be
4272 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
4273 values. Both arguments must have identical types.</p>
4276 <p>This instruction returns the unsigned integer <i>remainder</i> of a division.
4277 This instruction always performs an unsigned division to get the
4280 <p>Note that unsigned integer remainder and signed integer remainder are
4281 distinct operations; for signed integer remainder, use '<tt>srem</tt>'.</p>
4283 <p>Taking the remainder of a division by zero leads to undefined behavior.</p>
4287 <result> = urem i32 4, %var <i>; yields {i32}:result = 4 % %var</i>
4292 <!-- _______________________________________________________________________ -->
4294 <a name="i_srem">'<tt>srem</tt>' Instruction</a>
4301 <result> = srem <ty> <op1>, <op2> <i>; yields {ty}:result</i>
4305 <p>The '<tt>srem</tt>' instruction returns the remainder from the signed
4306 division of its two operands. This instruction can also take
4307 <a href="#t_vector">vector</a> versions of the values in which case the
4308 elements must be integers.</p>
4311 <p>The two arguments to the '<tt>srem</tt>' instruction must be
4312 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
4313 values. Both arguments must have identical types.</p>
4316 <p>This instruction returns the <i>remainder</i> of a division (where the result
4317 is either zero or has the same sign as the dividend, <tt>op1</tt>), not the
4318 <i>modulo</i> operator (where the result is either zero or has the same sign
4319 as the divisor, <tt>op2</tt>) of a value.
4320 For more information about the difference,
4321 see <a href="http://mathforum.org/dr.math/problems/anne.4.28.99.html">The
4322 Math Forum</a>. For a table of how this is implemented in various languages,
4323 please see <a href="http://en.wikipedia.org/wiki/Modulo_operation">
4324 Wikipedia: modulo operation</a>.</p>
4326 <p>Note that signed integer remainder and unsigned integer remainder are
4327 distinct operations; for unsigned integer remainder, use '<tt>urem</tt>'.</p>
4329 <p>Taking the remainder of a division by zero leads to undefined behavior.
4330 Overflow also leads to undefined behavior; this is a rare case, but can
4331 occur, for example, by taking the remainder of a 32-bit division of
4332 -2147483648 by -1. (The remainder doesn't actually overflow, but this rule
4333 lets srem be implemented using instructions that return both the result of
4334 the division and the remainder.)</p>
4338 <result> = srem i32 4, %var <i>; yields {i32}:result = 4 % %var</i>
4343 <!-- _______________________________________________________________________ -->
4345 <a name="i_frem">'<tt>frem</tt>' Instruction</a>
4352 <result> = frem <ty> <op1>, <op2> <i>; yields {ty}:result</i>
4356 <p>The '<tt>frem</tt>' instruction returns the remainder from the division of
4357 its two operands.</p>
4360 <p>The two arguments to the '<tt>frem</tt>' instruction must be
4361 <a href="#t_floating">floating point</a> or <a href="#t_vector">vector</a> of
4362 floating point values. Both arguments must have identical types.</p>
4365 <p>This instruction returns the <i>remainder</i> of a division. The remainder
4366 has the same sign as the dividend.</p>
4370 <result> = frem float 4.0, %var <i>; yields {float}:result = 4.0 % %var</i>
4377 <!-- ======================================================================= -->
4379 <a name="bitwiseops">Bitwise Binary Operations</a>
4384 <p>Bitwise binary operators are used to do various forms of bit-twiddling in a
4385 program. They are generally very efficient instructions and can commonly be
4386 strength reduced from other instructions. They require two operands of the
4387 same type, execute an operation on them, and produce a single value. The
4388 resulting value is the same type as its operands.</p>
4390 <!-- _______________________________________________________________________ -->
4392 <a name="i_shl">'<tt>shl</tt>' Instruction</a>
4399 <result> = shl <ty> <op1>, <op2> <i>; yields {ty}:result</i>
4400 <result> = shl nuw <ty> <op1>, <op2> <i>; yields {ty}:result</i>
4401 <result> = shl nsw <ty> <op1>, <op2> <i>; yields {ty}:result</i>
4402 <result> = shl nuw nsw <ty> <op1>, <op2> <i>; yields {ty}:result</i>
4406 <p>The '<tt>shl</tt>' instruction returns the first operand shifted to the left
4407 a specified number of bits.</p>
4410 <p>Both arguments to the '<tt>shl</tt>' instruction must be the
4411 same <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of
4412 integer type. '<tt>op2</tt>' is treated as an unsigned value.</p>
4415 <p>The value produced is <tt>op1</tt> * 2<sup><tt>op2</tt></sup> mod
4416 2<sup>n</sup>, where <tt>n</tt> is the width of the result. If <tt>op2</tt>
4417 is (statically or dynamically) negative or equal to or larger than the number
4418 of bits in <tt>op1</tt>, the result is undefined. If the arguments are
4419 vectors, each vector element of <tt>op1</tt> is shifted by the corresponding
4420 shift amount in <tt>op2</tt>.</p>
4422 <p>If the <tt>nuw</tt> keyword is present, then the shift produces a
4423 <a href="#poisonvalues">poison value</a> if it shifts out any non-zero bits. If
4424 the <tt>nsw</tt> keyword is present, then the shift produces a
4425 <a href="#poisonvalues">poison value</a> if it shifts out any bits that disagree
4426 with the resultant sign bit. As such, NUW/NSW have the same semantics as
4427 they would if the shift were expressed as a mul instruction with the same
4428 nsw/nuw bits in (mul %op1, (shl 1, %op2)).</p>
4432 <result> = shl i32 4, %var <i>; yields {i32}: 4 << %var</i>
4433 <result> = shl i32 4, 2 <i>; yields {i32}: 16</i>
4434 <result> = shl i32 1, 10 <i>; yields {i32}: 1024</i>
4435 <result> = shl i32 1, 32 <i>; undefined</i>
4436 <result> = shl <2 x i32> < i32 1, i32 1>, < i32 1, i32 2> <i>; yields: result=<2 x i32> < i32 2, i32 4></i>
4441 <!-- _______________________________________________________________________ -->
4443 <a name="i_lshr">'<tt>lshr</tt>' Instruction</a>
4450 <result> = lshr <ty> <op1>, <op2> <i>; yields {ty}:result</i>
4451 <result> = lshr exact <ty> <op1>, <op2> <i>; yields {ty}:result</i>
4455 <p>The '<tt>lshr</tt>' instruction (logical shift right) returns the first
4456 operand shifted to the right a specified number of bits with zero fill.</p>
4459 <p>Both arguments to the '<tt>lshr</tt>' instruction must be the same
4460 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
4461 type. '<tt>op2</tt>' is treated as an unsigned value.</p>
4464 <p>This instruction always performs a logical shift right operation. The most
4465 significant bits of the result will be filled with zero bits after the shift.
4466 If <tt>op2</tt> is (statically or dynamically) equal to or larger than the
4467 number of bits in <tt>op1</tt>, the result is undefined. If the arguments are
4468 vectors, each vector element of <tt>op1</tt> is shifted by the corresponding
4469 shift amount in <tt>op2</tt>.</p>
4471 <p>If the <tt>exact</tt> keyword is present, the result value of the
4472 <tt>lshr</tt> is a <a href="#poisonvalues">poison value</a> if any of the bits
4473 shifted out are non-zero.</p>
4478 <result> = lshr i32 4, 1 <i>; yields {i32}:result = 2</i>
4479 <result> = lshr i32 4, 2 <i>; yields {i32}:result = 1</i>
4480 <result> = lshr i8 4, 3 <i>; yields {i8}:result = 0</i>
4481 <result> = lshr i8 -2, 1 <i>; yields {i8}:result = 0x7FFFFFFF </i>
4482 <result> = lshr i32 1, 32 <i>; undefined</i>
4483 <result> = lshr <2 x i32> < i32 -2, i32 4>, < i32 1, i32 2> <i>; yields: result=<2 x i32> < i32 0x7FFFFFFF, i32 1></i>
4488 <!-- _______________________________________________________________________ -->
4490 <a name="i_ashr">'<tt>ashr</tt>' Instruction</a>
4497 <result> = ashr <ty> <op1>, <op2> <i>; yields {ty}:result</i>
4498 <result> = ashr exact <ty> <op1>, <op2> <i>; yields {ty}:result</i>
4502 <p>The '<tt>ashr</tt>' instruction (arithmetic shift right) returns the first
4503 operand shifted to the right a specified number of bits with sign
4507 <p>Both arguments to the '<tt>ashr</tt>' instruction must be the same
4508 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
4509 type. '<tt>op2</tt>' is treated as an unsigned value.</p>
4512 <p>This instruction always performs an arithmetic shift right operation, The
4513 most significant bits of the result will be filled with the sign bit
4514 of <tt>op1</tt>. If <tt>op2</tt> is (statically or dynamically) equal to or
4515 larger than the number of bits in <tt>op1</tt>, the result is undefined. If
4516 the arguments are vectors, each vector element of <tt>op1</tt> is shifted by
4517 the corresponding shift amount in <tt>op2</tt>.</p>
4519 <p>If the <tt>exact</tt> keyword is present, the result value of the
4520 <tt>ashr</tt> is a <a href="#poisonvalues">poison value</a> if any of the bits
4521 shifted out are non-zero.</p>
4525 <result> = ashr i32 4, 1 <i>; yields {i32}:result = 2</i>
4526 <result> = ashr i32 4, 2 <i>; yields {i32}:result = 1</i>
4527 <result> = ashr i8 4, 3 <i>; yields {i8}:result = 0</i>
4528 <result> = ashr i8 -2, 1 <i>; yields {i8}:result = -1</i>
4529 <result> = ashr i32 1, 32 <i>; undefined</i>
4530 <result> = ashr <2 x i32> < i32 -2, i32 4>, < i32 1, i32 3> <i>; yields: result=<2 x i32> < i32 -1, i32 0></i>
4535 <!-- _______________________________________________________________________ -->
4537 <a name="i_and">'<tt>and</tt>' Instruction</a>
4544 <result> = and <ty> <op1>, <op2> <i>; yields {ty}:result</i>
4548 <p>The '<tt>and</tt>' instruction returns the bitwise logical and of its two
4552 <p>The two arguments to the '<tt>and</tt>' instruction must be
4553 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
4554 values. Both arguments must have identical types.</p>
4557 <p>The truth table used for the '<tt>and</tt>' instruction is:</p>
4559 <table border="1" cellspacing="0" cellpadding="4">
4591 <result> = and i32 4, %var <i>; yields {i32}:result = 4 & %var</i>
4592 <result> = and i32 15, 40 <i>; yields {i32}:result = 8</i>
4593 <result> = and i32 4, 8 <i>; yields {i32}:result = 0</i>
4596 <!-- _______________________________________________________________________ -->
4598 <a name="i_or">'<tt>or</tt>' Instruction</a>
4605 <result> = or <ty> <op1>, <op2> <i>; yields {ty}:result</i>
4609 <p>The '<tt>or</tt>' instruction returns the bitwise logical inclusive or of its
4613 <p>The two arguments to the '<tt>or</tt>' instruction must be
4614 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
4615 values. Both arguments must have identical types.</p>
4618 <p>The truth table used for the '<tt>or</tt>' instruction is:</p>
4620 <table border="1" cellspacing="0" cellpadding="4">
4652 <result> = or i32 4, %var <i>; yields {i32}:result = 4 | %var</i>
4653 <result> = or i32 15, 40 <i>; yields {i32}:result = 47</i>
4654 <result> = or i32 4, 8 <i>; yields {i32}:result = 12</i>
4659 <!-- _______________________________________________________________________ -->
4661 <a name="i_xor">'<tt>xor</tt>' Instruction</a>
4668 <result> = xor <ty> <op1>, <op2> <i>; yields {ty}:result</i>
4672 <p>The '<tt>xor</tt>' instruction returns the bitwise logical exclusive or of
4673 its two operands. The <tt>xor</tt> is used to implement the "one's
4674 complement" operation, which is the "~" operator in C.</p>
4677 <p>The two arguments to the '<tt>xor</tt>' instruction must be
4678 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
4679 values. Both arguments must have identical types.</p>
4682 <p>The truth table used for the '<tt>xor</tt>' instruction is:</p>
4684 <table border="1" cellspacing="0" cellpadding="4">
4716 <result> = xor i32 4, %var <i>; yields {i32}:result = 4 ^ %var</i>
4717 <result> = xor i32 15, 40 <i>; yields {i32}:result = 39</i>
4718 <result> = xor i32 4, 8 <i>; yields {i32}:result = 12</i>
4719 <result> = xor i32 %V, -1 <i>; yields {i32}:result = ~%V</i>
4726 <!-- ======================================================================= -->
4728 <a name="vectorops">Vector Operations</a>
4733 <p>LLVM supports several instructions to represent vector operations in a
4734 target-independent manner. These instructions cover the element-access and
4735 vector-specific operations needed to process vectors effectively. While LLVM
4736 does directly support these vector operations, many sophisticated algorithms
4737 will want to use target-specific intrinsics to take full advantage of a
4738 specific target.</p>
4740 <!-- _______________________________________________________________________ -->
4742 <a name="i_extractelement">'<tt>extractelement</tt>' Instruction</a>
4749 <result> = extractelement <n x <ty>> <val>, i32 <idx> <i>; yields <ty></i>
4753 <p>The '<tt>extractelement</tt>' instruction extracts a single scalar element
4754 from a vector at a specified index.</p>
4758 <p>The first operand of an '<tt>extractelement</tt>' instruction is a value
4759 of <a href="#t_vector">vector</a> type. The second operand is an index
4760 indicating the position from which to extract the element. The index may be
4764 <p>The result is a scalar of the same type as the element type of
4765 <tt>val</tt>. Its value is the value at position <tt>idx</tt> of
4766 <tt>val</tt>. If <tt>idx</tt> exceeds the length of <tt>val</tt>, the
4767 results are undefined.</p>
4771 <result> = extractelement <4 x i32> %vec, i32 0 <i>; yields i32</i>
4776 <!-- _______________________________________________________________________ -->
4778 <a name="i_insertelement">'<tt>insertelement</tt>' Instruction</a>
4785 <result> = insertelement <n x <ty>> <val>, <ty> <elt>, i32 <idx> <i>; yields <n x <ty>></i>
4789 <p>The '<tt>insertelement</tt>' instruction inserts a scalar element into a
4790 vector at a specified index.</p>
4793 <p>The first operand of an '<tt>insertelement</tt>' instruction is a value
4794 of <a href="#t_vector">vector</a> type. The second operand is a scalar value
4795 whose type must equal the element type of the first operand. The third
4796 operand is an index indicating the position at which to insert the value.
4797 The index may be a variable.</p>
4800 <p>The result is a vector of the same type as <tt>val</tt>. Its element values
4801 are those of <tt>val</tt> except at position <tt>idx</tt>, where it gets the
4802 value <tt>elt</tt>. If <tt>idx</tt> exceeds the length of <tt>val</tt>, the
4803 results are undefined.</p>
4807 <result> = insertelement <4 x i32> %vec, i32 1, i32 0 <i>; yields <4 x i32></i>
4812 <!-- _______________________________________________________________________ -->
4814 <a name="i_shufflevector">'<tt>shufflevector</tt>' Instruction</a>
4821 <result> = shufflevector <n x <ty>> <v1>, <n x <ty>> <v2>, <m x i32> <mask> <i>; yields <m x <ty>></i>
4825 <p>The '<tt>shufflevector</tt>' instruction constructs a permutation of elements
4826 from two input vectors, returning a vector with the same element type as the
4827 input and length that is the same as the shuffle mask.</p>
4830 <p>The first two operands of a '<tt>shufflevector</tt>' instruction are vectors
4831 with the same type. The third argument is a shuffle mask whose
4832 element type is always 'i32'. The result of the instruction is a vector
4833 whose length is the same as the shuffle mask and whose element type is the
4834 same as the element type of the first two operands.</p>
4836 <p>The shuffle mask operand is required to be a constant vector with either
4837 constant integer or undef values.</p>
4840 <p>The elements of the two input vectors are numbered from left to right across
4841 both of the vectors. The shuffle mask operand specifies, for each element of
4842 the result vector, which element of the two input vectors the result element
4843 gets. The element selector may be undef (meaning "don't care") and the
4844 second operand may be undef if performing a shuffle from only one vector.</p>
4848 <result> = shufflevector <4 x i32> %v1, <4 x i32> %v2,
4849 <4 x i32> <i32 0, i32 4, i32 1, i32 5> <i>; yields <4 x i32></i>
4850 <result> = shufflevector <4 x i32> %v1, <4 x i32> undef,
4851 <4 x i32> <i32 0, i32 1, i32 2, i32 3> <i>; yields <4 x i32></i> - Identity shuffle.
4852 <result> = shufflevector <8 x i32> %v1, <8 x i32> undef,
4853 <4 x i32> <i32 0, i32 1, i32 2, i32 3> <i>; yields <4 x i32></i>
4854 <result> = shufflevector <4 x i32> %v1, <4 x i32> %v2,
4855 <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>
4862 <!-- ======================================================================= -->
4864 <a name="aggregateops">Aggregate Operations</a>
4869 <p>LLVM supports several instructions for working with
4870 <a href="#t_aggregate">aggregate</a> values.</p>
4872 <!-- _______________________________________________________________________ -->
4874 <a name="i_extractvalue">'<tt>extractvalue</tt>' Instruction</a>
4881 <result> = extractvalue <aggregate type> <val>, <idx>{, <idx>}*
4885 <p>The '<tt>extractvalue</tt>' instruction extracts the value of a member field
4886 from an <a href="#t_aggregate">aggregate</a> value.</p>
4889 <p>The first operand of an '<tt>extractvalue</tt>' instruction is a value
4890 of <a href="#t_struct">struct</a> or
4891 <a href="#t_array">array</a> type. The operands are constant indices to
4892 specify which value to extract in a similar manner as indices in a
4893 '<tt><a href="#i_getelementptr">getelementptr</a></tt>' instruction.</p>
4894 <p>The major differences to <tt>getelementptr</tt> indexing are:</p>
4896 <li>Since the value being indexed is not a pointer, the first index is
4897 omitted and assumed to be zero.</li>
4898 <li>At least one index must be specified.</li>
4899 <li>Not only struct indices but also array indices must be in
4904 <p>The result is the value at the position in the aggregate specified by the
4909 <result> = extractvalue {i32, float} %agg, 0 <i>; yields i32</i>
4914 <!-- _______________________________________________________________________ -->
4916 <a name="i_insertvalue">'<tt>insertvalue</tt>' Instruction</a>
4923 <result> = insertvalue <aggregate type> <val>, <ty> <elt>, <idx>{, <idx>}* <i>; yields <aggregate type></i>
4927 <p>The '<tt>insertvalue</tt>' instruction inserts a value into a member field
4928 in an <a href="#t_aggregate">aggregate</a> value.</p>
4931 <p>The first operand of an '<tt>insertvalue</tt>' instruction is a value
4932 of <a href="#t_struct">struct</a> or
4933 <a href="#t_array">array</a> type. The second operand is a first-class
4934 value to insert. The following operands are constant indices indicating
4935 the position at which to insert the value in a similar manner as indices in a
4936 '<tt><a href="#i_extractvalue">extractvalue</a></tt>' instruction. The
4937 value to insert must have the same type as the value identified by the
4941 <p>The result is an aggregate of the same type as <tt>val</tt>. Its value is
4942 that of <tt>val</tt> except that the value at the position specified by the
4943 indices is that of <tt>elt</tt>.</p>
4947 %agg1 = insertvalue {i32, float} undef, i32 1, 0 <i>; yields {i32 1, float undef}</i>
4948 %agg2 = insertvalue {i32, float} %agg1, float %val, 1 <i>; yields {i32 1, float %val}</i>
4949 %agg3 = insertvalue {i32, {float}} %agg1, float %val, 1, 0 <i>; yields {i32 1, float %val}</i>
4956 <!-- ======================================================================= -->
4958 <a name="memoryops">Memory Access and Addressing Operations</a>
4963 <p>A key design point of an SSA-based representation is how it represents
4964 memory. In LLVM, no memory locations are in SSA form, which makes things
4965 very simple. This section describes how to read, write, and allocate
4968 <!-- _______________________________________________________________________ -->
4970 <a name="i_alloca">'<tt>alloca</tt>' Instruction</a>
4977 <result> = alloca <type>[, <ty> <NumElements>][, align <alignment>] <i>; yields {type*}:result</i>
4981 <p>The '<tt>alloca</tt>' instruction allocates memory on the stack frame of the
4982 currently executing function, to be automatically released when this function
4983 returns to its caller. The object is always allocated in the generic address
4984 space (address space zero).</p>
4987 <p>The '<tt>alloca</tt>' instruction
4988 allocates <tt>sizeof(<type>)*NumElements</tt> bytes of memory on the
4989 runtime stack, returning a pointer of the appropriate type to the program.
4990 If "NumElements" is specified, it is the number of elements allocated,
4991 otherwise "NumElements" is defaulted to be one. If a constant alignment is
4992 specified, the value result of the allocation is guaranteed to be aligned to
4993 at least that boundary. If not specified, or if zero, the target can choose
4994 to align the allocation on any convenient boundary compatible with the
4997 <p>'<tt>type</tt>' may be any sized type.</p>
5000 <p>Memory is allocated; a pointer is returned. The operation is undefined if
5001 there is insufficient stack space for the allocation. '<tt>alloca</tt>'d
5002 memory is automatically released when the function returns. The
5003 '<tt>alloca</tt>' instruction is commonly used to represent automatic
5004 variables that must have an address available. When the function returns
5005 (either with the <tt><a href="#i_ret">ret</a></tt>
5006 or <tt><a href="#i_resume">resume</a></tt> instructions), the memory is
5007 reclaimed. Allocating zero bytes is legal, but the result is undefined.
5008 The order in which memory is allocated (ie., which way the stack grows) is
5015 %ptr = alloca i32 <i>; yields {i32*}:ptr</i>
5016 %ptr = alloca i32, i32 4 <i>; yields {i32*}:ptr</i>
5017 %ptr = alloca i32, i32 4, align 1024 <i>; yields {i32*}:ptr</i>
5018 %ptr = alloca i32, align 1024 <i>; yields {i32*}:ptr</i>
5023 <!-- _______________________________________________________________________ -->
5025 <a name="i_load">'<tt>load</tt>' Instruction</a>
5032 <result> = load [volatile] <ty>* <pointer>[, align <alignment>][, !nontemporal !<index>][, !invariant.load !<index>]
5033 <result> = load atomic [volatile] <ty>* <pointer> [singlethread] <ordering>, align <alignment>
5034 !<index> = !{ i32 1 }
5038 <p>The '<tt>load</tt>' instruction is used to read from memory.</p>
5041 <p>The argument to the '<tt>load</tt>' instruction specifies the memory address
5042 from which to load. The pointer must point to
5043 a <a href="#t_firstclass">first class</a> type. If the <tt>load</tt> is
5044 marked as <tt>volatile</tt>, then the optimizer is not allowed to modify the
5045 number or order of execution of this <tt>load</tt> with other <a
5046 href="#volatile">volatile operations</a>.</p>
5048 <p>If the <code>load</code> is marked as <code>atomic</code>, it takes an extra
5049 <a href="#ordering">ordering</a> and optional <code>singlethread</code>
5050 argument. The <code>release</code> and <code>acq_rel</code> orderings are
5051 not valid on <code>load</code> instructions. Atomic loads produce <a
5052 href="#memorymodel">defined</a> results when they may see multiple atomic
5053 stores. The type of the pointee must be an integer type whose bit width
5054 is a power of two greater than or equal to eight and less than or equal
5055 to a target-specific size limit. <code>align</code> must be explicitly
5056 specified on atomic loads, and the load has undefined behavior if the
5057 alignment is not set to a value which is at least the size in bytes of
5058 the pointee. <code>!nontemporal</code> does not have any defined semantics
5059 for atomic loads.</p>
5061 <p>The optional constant <tt>align</tt> argument specifies the alignment of the
5062 operation (that is, the alignment of the memory address). A value of 0 or an
5063 omitted <tt>align</tt> argument means that the operation has the abi
5064 alignment for the target. It is the responsibility of the code emitter to
5065 ensure that the alignment information is correct. Overestimating the
5066 alignment results in undefined behavior. Underestimating the alignment may
5067 produce less efficient code. An alignment of 1 is always safe.</p>
5069 <p>The optional <tt>!nontemporal</tt> metadata must reference a single
5070 metatadata name <index> corresponding to a metadata node with
5071 one <tt>i32</tt> entry of value 1. The existence of
5072 the <tt>!nontemporal</tt> metatadata on the instruction tells the optimizer
5073 and code generator that this load is not expected to be reused in the cache.
5074 The code generator may select special instructions to save cache bandwidth,
5075 such as the <tt>MOVNT</tt> instruction on x86.</p>
5077 <p>The optional <tt>!invariant.load</tt> metadata must reference a single
5078 metatadata name <index> corresponding to a metadata node with no
5079 entries. The existence of the <tt>!invariant.load</tt> metatadata on the
5080 instruction tells the optimizer and code generator that this load address
5081 points to memory which does not change value during program execution.
5082 The optimizer may then move this load around, for example, by hoisting it
5083 out of loops using loop invariant code motion.</p>
5086 <p>The location of memory pointed to is loaded. If the value being loaded is of
5087 scalar type then the number of bytes read does not exceed the minimum number
5088 of bytes needed to hold all bits of the type. For example, loading an
5089 <tt>i24</tt> reads at most three bytes. When loading a value of a type like
5090 <tt>i20</tt> with a size that is not an integral number of bytes, the result
5091 is undefined if the value was not originally written using a store of the
5096 %ptr = <a href="#i_alloca">alloca</a> i32 <i>; yields {i32*}:ptr</i>
5097 <a href="#i_store">store</a> i32 3, i32* %ptr <i>; yields {void}</i>
5098 %val = load i32* %ptr <i>; yields {i32}:val = i32 3</i>
5103 <!-- _______________________________________________________________________ -->
5105 <a name="i_store">'<tt>store</tt>' Instruction</a>
5112 store [volatile] <ty> <value>, <ty>* <pointer>[, align <alignment>][, !nontemporal !<index>] <i>; yields {void}</i>
5113 store atomic [volatile] <ty> <value>, <ty>* <pointer> [singlethread] <ordering>, align <alignment> <i>; yields {void}</i>
5117 <p>The '<tt>store</tt>' instruction is used to write to memory.</p>
5120 <p>There are two arguments to the '<tt>store</tt>' instruction: a value to store
5121 and an address at which to store it. The type of the
5122 '<tt><pointer></tt>' operand must be a pointer to
5123 the <a href="#t_firstclass">first class</a> type of the
5124 '<tt><value></tt>' operand. If the <tt>store</tt> is marked as
5125 <tt>volatile</tt>, then the optimizer is not allowed to modify the number or
5126 order of execution of this <tt>store</tt> with other <a
5127 href="#volatile">volatile operations</a>.</p>
5129 <p>If the <code>store</code> is marked as <code>atomic</code>, it takes an extra
5130 <a href="#ordering">ordering</a> and optional <code>singlethread</code>
5131 argument. The <code>acquire</code> and <code>acq_rel</code> orderings aren't
5132 valid on <code>store</code> instructions. Atomic loads produce <a
5133 href="#memorymodel">defined</a> results when they may see multiple atomic
5134 stores. The type of the pointee must be an integer type whose bit width
5135 is a power of two greater than or equal to eight and less than or equal
5136 to a target-specific size limit. <code>align</code> must be explicitly
5137 specified on atomic stores, and the store has undefined behavior if the
5138 alignment is not set to a value which is at least the size in bytes of
5139 the pointee. <code>!nontemporal</code> does not have any defined semantics
5140 for atomic stores.</p>
5142 <p>The optional constant "align" argument specifies the alignment of the
5143 operation (that is, the alignment of the memory address). A value of 0 or an
5144 omitted "align" argument means that the operation has the abi
5145 alignment for the target. It is the responsibility of the code emitter to
5146 ensure that the alignment information is correct. Overestimating the
5147 alignment results in an undefined behavior. Underestimating the alignment may
5148 produce less efficient code. An alignment of 1 is always safe.</p>
5150 <p>The optional !nontemporal metadata must reference a single metatadata
5151 name <index> corresponding to a metadata node with one i32 entry of
5152 value 1. The existence of the !nontemporal metatadata on the
5153 instruction tells the optimizer and code generator that this load is
5154 not expected to be reused in the cache. The code generator may
5155 select special instructions to save cache bandwidth, such as the
5156 MOVNT instruction on x86.</p>
5160 <p>The contents of memory are updated to contain '<tt><value></tt>' at the
5161 location specified by the '<tt><pointer></tt>' operand. If
5162 '<tt><value></tt>' is of scalar type then the number of bytes written
5163 does not exceed the minimum number of bytes needed to hold all bits of the
5164 type. For example, storing an <tt>i24</tt> writes at most three bytes. When
5165 writing a value of a type like <tt>i20</tt> with a size that is not an
5166 integral number of bytes, it is unspecified what happens to the extra bits
5167 that do not belong to the type, but they will typically be overwritten.</p>
5171 %ptr = <a href="#i_alloca">alloca</a> i32 <i>; yields {i32*}:ptr</i>
5172 store i32 3, i32* %ptr <i>; yields {void}</i>
5173 %val = <a href="#i_load">load</a> i32* %ptr <i>; yields {i32}:val = i32 3</i>
5178 <!-- _______________________________________________________________________ -->
5180 <a name="i_fence">'<tt>fence</tt>' Instruction</a>
5187 fence [singlethread] <ordering> <i>; yields {void}</i>
5191 <p>The '<tt>fence</tt>' instruction is used to introduce happens-before edges
5192 between operations.</p>
5194 <h5>Arguments:</h5> <p>'<code>fence</code>' instructions take an <a
5195 href="#ordering">ordering</a> argument which defines what
5196 <i>synchronizes-with</i> edges they add. They can only be given
5197 <code>acquire</code>, <code>release</code>, <code>acq_rel</code>, and
5198 <code>seq_cst</code> orderings.</p>
5201 <p>A fence <var>A</var> which has (at least) <code>release</code> ordering
5202 semantics <i>synchronizes with</i> a fence <var>B</var> with (at least)
5203 <code>acquire</code> ordering semantics if and only if there exist atomic
5204 operations <var>X</var> and <var>Y</var>, both operating on some atomic object
5205 <var>M</var>, such that <var>A</var> is sequenced before <var>X</var>,
5206 <var>X</var> modifies <var>M</var> (either directly or through some side effect
5207 of a sequence headed by <var>X</var>), <var>Y</var> is sequenced before
5208 <var>B</var>, and <var>Y</var> observes <var>M</var>. This provides a
5209 <i>happens-before</i> dependency between <var>A</var> and <var>B</var>. Rather
5210 than an explicit <code>fence</code>, one (but not both) of the atomic operations
5211 <var>X</var> or <var>Y</var> might provide a <code>release</code> or
5212 <code>acquire</code> (resp.) ordering constraint and still
5213 <i>synchronize-with</i> the explicit <code>fence</code> and establish the
5214 <i>happens-before</i> edge.</p>
5216 <p>A <code>fence</code> which has <code>seq_cst</code> ordering, in addition to
5217 having both <code>acquire</code> and <code>release</code> semantics specified
5218 above, participates in the global program order of other <code>seq_cst</code>
5219 operations and/or fences.</p>
5221 <p>The optional "<a href="#singlethread"><code>singlethread</code></a>" argument
5222 specifies that the fence only synchronizes with other fences in the same
5223 thread. (This is useful for interacting with signal handlers.)</p>
5227 fence acquire <i>; yields {void}</i>
5228 fence singlethread seq_cst <i>; yields {void}</i>
5233 <!-- _______________________________________________________________________ -->
5235 <a name="i_cmpxchg">'<tt>cmpxchg</tt>' Instruction</a>
5242 cmpxchg [volatile] <ty>* <pointer>, <ty> <cmp>, <ty> <new> [singlethread] <ordering> <i>; yields {ty}</i>
5246 <p>The '<tt>cmpxchg</tt>' instruction is used to atomically modify memory.
5247 It loads a value in memory and compares it to a given value. If they are
5248 equal, it stores a new value into the memory.</p>
5251 <p>There are three arguments to the '<code>cmpxchg</code>' instruction: an
5252 address to operate on, a value to compare to the value currently be at that
5253 address, and a new value to place at that address if the compared values are
5254 equal. The type of '<var><cmp></var>' must be an integer type whose
5255 bit width is a power of two greater than or equal to eight and less than
5256 or equal to a target-specific size limit. '<var><cmp></var>' and
5257 '<var><new></var>' must have the same type, and the type of
5258 '<var><pointer></var>' must be a pointer to that type. If the
5259 <code>cmpxchg</code> is marked as <code>volatile</code>, then the
5260 optimizer is not allowed to modify the number or order of execution
5261 of this <code>cmpxchg</code> with other <a href="#volatile">volatile
5264 <!-- FIXME: Extend allowed types. -->
5266 <p>The <a href="#ordering"><var>ordering</var></a> argument specifies how this
5267 <code>cmpxchg</code> synchronizes with other atomic operations.</p>
5269 <p>The optional "<code>singlethread</code>" argument declares that the
5270 <code>cmpxchg</code> is only atomic with respect to code (usually signal
5271 handlers) running in the same thread as the <code>cmpxchg</code>. Otherwise the
5272 cmpxchg is atomic with respect to all other code in the system.</p>
5274 <p>The pointer passed into cmpxchg must have alignment greater than or equal to
5275 the size in memory of the operand.
5278 <p>The contents of memory at the location specified by the
5279 '<tt><pointer></tt>' operand is read and compared to
5280 '<tt><cmp></tt>'; if the read value is the equal,
5281 '<tt><new></tt>' is written. The original value at the location
5284 <p>A successful <code>cmpxchg</code> is a read-modify-write instruction for the
5285 purpose of identifying <a href="#release_sequence">release sequences</a>. A
5286 failed <code>cmpxchg</code> is equivalent to an atomic load with an ordering
5287 parameter determined by dropping any <code>release</code> part of the
5288 <code>cmpxchg</code>'s ordering.</p>
5291 FIXME: Is compare_exchange_weak() necessary? (Consider after we've done
5292 optimization work on ARM.)
5294 FIXME: Is a weaker ordering constraint on failure helpful in practice?
5300 %orig = atomic <a href="#i_load">load</a> i32* %ptr unordered <i>; yields {i32}</i>
5301 <a href="#i_br">br</a> label %loop
5304 %cmp = <a href="#i_phi">phi</a> i32 [ %orig, %entry ], [%old, %loop]
5305 %squared = <a href="#i_mul">mul</a> i32 %cmp, %cmp
5306 %old = cmpxchg i32* %ptr, i32 %cmp, i32 %squared <i>; yields {i32}</i>
5307 %success = <a href="#i_icmp">icmp</a> eq i32 %cmp, %old
5308 <a href="#i_br">br</a> i1 %success, label %done, label %loop
5316 <!-- _______________________________________________________________________ -->
5318 <a name="i_atomicrmw">'<tt>atomicrmw</tt>' Instruction</a>
5325 atomicrmw [volatile] <operation> <ty>* <pointer>, <ty> <value> [singlethread] <ordering> <i>; yields {ty}</i>
5329 <p>The '<tt>atomicrmw</tt>' instruction is used to atomically modify memory.</p>
5332 <p>There are three arguments to the '<code>atomicrmw</code>' instruction: an
5333 operation to apply, an address whose value to modify, an argument to the
5334 operation. The operation must be one of the following keywords:</p>
5349 <p>The type of '<var><value></var>' must be an integer type whose
5350 bit width is a power of two greater than or equal to eight and less than
5351 or equal to a target-specific size limit. The type of the
5352 '<code><pointer></code>' operand must be a pointer to that type.
5353 If the <code>atomicrmw</code> is marked as <code>volatile</code>, then the
5354 optimizer is not allowed to modify the number or order of execution of this
5355 <code>atomicrmw</code> with other <a href="#volatile">volatile
5358 <!-- FIXME: Extend allowed types. -->
5361 <p>The contents of memory at the location specified by the
5362 '<tt><pointer></tt>' operand are atomically read, modified, and written
5363 back. The original value at the location is returned. The modification is
5364 specified by the <var>operation</var> argument:</p>
5367 <li>xchg: <code>*ptr = val</code></li>
5368 <li>add: <code>*ptr = *ptr + val</code></li>
5369 <li>sub: <code>*ptr = *ptr - val</code></li>
5370 <li>and: <code>*ptr = *ptr & val</code></li>
5371 <li>nand: <code>*ptr = ~(*ptr & val)</code></li>
5372 <li>or: <code>*ptr = *ptr | val</code></li>
5373 <li>xor: <code>*ptr = *ptr ^ val</code></li>
5374 <li>max: <code>*ptr = *ptr > val ? *ptr : val</code> (using a signed comparison)</li>
5375 <li>min: <code>*ptr = *ptr < val ? *ptr : val</code> (using a signed comparison)</li>
5376 <li>umax: <code>*ptr = *ptr > val ? *ptr : val</code> (using an unsigned comparison)</li>
5377 <li>umin: <code>*ptr = *ptr < val ? *ptr : val</code> (using an unsigned comparison)</li>
5382 %old = atomicrmw add i32* %ptr, i32 1 acquire <i>; yields {i32}</i>
5387 <!-- _______________________________________________________________________ -->
5389 <a name="i_getelementptr">'<tt>getelementptr</tt>' Instruction</a>
5396 <result> = getelementptr <pty>* <ptrval>{, <ty> <idx>}*
5397 <result> = getelementptr inbounds <pty>* <ptrval>{, <ty> <idx>}*
5398 <result> = getelementptr <ptr vector> ptrval, <vector index type> idx
5402 <p>The '<tt>getelementptr</tt>' instruction is used to get the address of a
5403 subelement of an <a href="#t_aggregate">aggregate</a> data structure.
5404 It performs address calculation only and does not access memory.</p>
5407 <p>The first argument is always a pointer or a vector of pointers,
5408 and forms the basis of the
5409 calculation. The remaining arguments are indices that indicate which of the
5410 elements of the aggregate object are indexed. The interpretation of each
5411 index is dependent on the type being indexed into. The first index always
5412 indexes the pointer value given as the first argument, the second index
5413 indexes a value of the type pointed to (not necessarily the value directly
5414 pointed to, since the first index can be non-zero), etc. The first type
5415 indexed into must be a pointer value, subsequent types can be arrays,
5416 vectors, and structs. Note that subsequent types being indexed into
5417 can never be pointers, since that would require loading the pointer before
5418 continuing calculation.</p>
5420 <p>The type of each index argument depends on the type it is indexing into.
5421 When indexing into a (optionally packed) structure, only <tt>i32</tt>
5422 integer <b>constants</b> are allowed. When indexing into an array, pointer
5423 or vector, integers of any width are allowed, and they are not required to be
5424 constant. These integers are treated as signed values where relevant.</p>
5426 <p>For example, let's consider a C code fragment and how it gets compiled to
5429 <pre class="doc_code">
5441 int *foo(struct ST *s) {
5442 return &s[1].Z.B[5][13];
5446 <p>The LLVM code generated by Clang is:</p>
5448 <pre class="doc_code">
5449 %struct.RT = <a href="#namedtypes">type</a> { i8, [10 x [20 x i32]], i8 }
5450 %struct.ST = <a href="#namedtypes">type</a> { i32, double, %struct.RT }
5452 define i32* @foo(%struct.ST* %s) nounwind uwtable readnone optsize ssp {
5454 %arrayidx = getelementptr inbounds %struct.ST* %s, i64 1, i32 2, i32 1, i64 5, i64 13
5460 <p>In the example above, the first index is indexing into the
5461 '<tt>%struct.ST*</tt>' type, which is a pointer, yielding a
5462 '<tt>%struct.ST</tt>' = '<tt>{ i32, double, %struct.RT }</tt>' type, a
5463 structure. The second index indexes into the third element of the structure,
5464 yielding a '<tt>%struct.RT</tt>' = '<tt>{ i8 , [10 x [20 x i32]], i8 }</tt>'
5465 type, another structure. The third index indexes into the second element of
5466 the structure, yielding a '<tt>[10 x [20 x i32]]</tt>' type, an array. The
5467 two dimensions of the array are subscripted into, yielding an '<tt>i32</tt>'
5468 type. The '<tt>getelementptr</tt>' instruction returns a pointer to this
5469 element, thus computing a value of '<tt>i32*</tt>' type.</p>
5471 <p>Note that it is perfectly legal to index partially through a structure,
5472 returning a pointer to an inner element. Because of this, the LLVM code for
5473 the given testcase is equivalent to:</p>
5475 <pre class="doc_code">
5476 define i32* @foo(%struct.ST* %s) {
5477 %t1 = getelementptr %struct.ST* %s, i32 1 <i>; yields %struct.ST*:%t1</i>
5478 %t2 = getelementptr %struct.ST* %t1, i32 0, i32 2 <i>; yields %struct.RT*:%t2</i>
5479 %t3 = getelementptr %struct.RT* %t2, i32 0, i32 1 <i>; yields [10 x [20 x i32]]*:%t3</i>
5480 %t4 = getelementptr [10 x [20 x i32]]* %t3, i32 0, i32 5 <i>; yields [20 x i32]*:%t4</i>
5481 %t5 = getelementptr [20 x i32]* %t4, i32 0, i32 13 <i>; yields i32*:%t5</i>
5486 <p>If the <tt>inbounds</tt> keyword is present, the result value of the
5487 <tt>getelementptr</tt> is a <a href="#poisonvalues">poison value</a> if the
5488 base pointer is not an <i>in bounds</i> address of an allocated object,
5489 or if any of the addresses that would be formed by successive addition of
5490 the offsets implied by the indices to the base address with infinitely
5491 precise signed arithmetic are not an <i>in bounds</i> address of that
5492 allocated object. The <i>in bounds</i> addresses for an allocated object
5493 are all the addresses that point into the object, plus the address one
5495 In cases where the base is a vector of pointers the <tt>inbounds</tt> keyword
5496 applies to each of the computations element-wise. </p>
5498 <p>If the <tt>inbounds</tt> keyword is not present, the offsets are added to
5499 the base address with silently-wrapping two's complement arithmetic. If the
5500 offsets have a different width from the pointer, they are sign-extended or
5501 truncated to the width of the pointer. The result value of the
5502 <tt>getelementptr</tt> may be outside the object pointed to by the base
5503 pointer. The result value may not necessarily be used to access memory
5504 though, even if it happens to point into allocated storage. See the
5505 <a href="#pointeraliasing">Pointer Aliasing Rules</a> section for more
5508 <p>The getelementptr instruction is often confusing. For some more insight into
5509 how it works, see <a href="GetElementPtr.html">the getelementptr FAQ</a>.</p>
5513 <i>; yields [12 x i8]*:aptr</i>
5514 %aptr = getelementptr {i32, [12 x i8]}* %saptr, i64 0, i32 1
5515 <i>; yields i8*:vptr</i>
5516 %vptr = getelementptr {i32, <2 x i8>}* %svptr, i64 0, i32 1, i32 1
5517 <i>; yields i8*:eptr</i>
5518 %eptr = getelementptr [12 x i8]* %aptr, i64 0, i32 1
5519 <i>; yields i32*:iptr</i>
5520 %iptr = getelementptr [10 x i32]* @arr, i16 0, i16 0
5523 <p>In cases where the pointer argument is a vector of pointers, only a
5524 single index may be used, and the number of vector elements has to be
5525 the same. For example: </p>
5526 <pre class="doc_code">
5527 %A = getelementptr <4 x i8*> %ptrs, <4 x i64> %offsets,
5534 <!-- ======================================================================= -->
5536 <a name="convertops">Conversion Operations</a>
5541 <p>The instructions in this category are the conversion instructions (casting)
5542 which all take a single operand and a type. They perform various bit
5543 conversions on the operand.</p>
5545 <!-- _______________________________________________________________________ -->
5547 <a name="i_trunc">'<tt>trunc .. to</tt>' Instruction</a>
5554 <result> = trunc <ty> <value> to <ty2> <i>; yields ty2</i>
5558 <p>The '<tt>trunc</tt>' instruction truncates its operand to the
5559 type <tt>ty2</tt>.</p>
5562 <p>The '<tt>trunc</tt>' instruction takes a value to trunc, and a type to trunc it to.
5563 Both types must be of <a href="#t_integer">integer</a> types, or vectors
5564 of the same number of integers.
5565 The bit size of the <tt>value</tt> must be larger than
5566 the bit size of the destination type, <tt>ty2</tt>.
5567 Equal sized types are not allowed.</p>
5570 <p>The '<tt>trunc</tt>' instruction truncates the high order bits
5571 in <tt>value</tt> and converts the remaining bits to <tt>ty2</tt>. Since the
5572 source size must be larger than the destination size, <tt>trunc</tt> cannot
5573 be a <i>no-op cast</i>. It will always truncate bits.</p>
5577 %X = trunc i32 257 to i8 <i>; yields i8:1</i>
5578 %Y = trunc i32 123 to i1 <i>; yields i1:true</i>
5579 %Z = trunc i32 122 to i1 <i>; yields i1:false</i>
5580 %W = trunc <2 x i16> <i16 8, i16 7> to <2 x i8> <i>; yields <i8 8, i8 7></i>
5585 <!-- _______________________________________________________________________ -->
5587 <a name="i_zext">'<tt>zext .. to</tt>' Instruction</a>
5594 <result> = zext <ty> <value> to <ty2> <i>; yields ty2</i>
5598 <p>The '<tt>zext</tt>' instruction zero extends its operand to type
5603 <p>The '<tt>zext</tt>' instruction takes a value to cast, and a type to cast it to.
5604 Both types must be of <a href="#t_integer">integer</a> types, or vectors
5605 of the same number of integers.
5606 The bit size of the <tt>value</tt> must be smaller than
5607 the bit size of the destination type,
5611 <p>The <tt>zext</tt> fills the high order bits of the <tt>value</tt> with zero
5612 bits until it reaches the size of the destination type, <tt>ty2</tt>.</p>
5614 <p>When zero extending from i1, the result will always be either 0 or 1.</p>
5618 %X = zext i32 257 to i64 <i>; yields i64:257</i>
5619 %Y = zext i1 true to i32 <i>; yields i32:1</i>
5620 %Z = zext <2 x i16> <i16 8, i16 7> to <2 x i32> <i>; yields <i32 8, i32 7></i>
5625 <!-- _______________________________________________________________________ -->
5627 <a name="i_sext">'<tt>sext .. to</tt>' Instruction</a>
5634 <result> = sext <ty> <value> to <ty2> <i>; yields ty2</i>
5638 <p>The '<tt>sext</tt>' sign extends <tt>value</tt> to the type <tt>ty2</tt>.</p>
5641 <p>The '<tt>sext</tt>' instruction takes a value to cast, and a type to cast it to.
5642 Both types must be of <a href="#t_integer">integer</a> types, or vectors
5643 of the same number of integers.
5644 The bit size of the <tt>value</tt> must be smaller than
5645 the bit size of the destination type,
5649 <p>The '<tt>sext</tt>' instruction performs a sign extension by copying the sign
5650 bit (highest order bit) of the <tt>value</tt> until it reaches the bit size
5651 of the type <tt>ty2</tt>.</p>
5653 <p>When sign extending from i1, the extension always results in -1 or 0.</p>
5657 %X = sext i8 -1 to i16 <i>; yields i16 :65535</i>
5658 %Y = sext i1 true to i32 <i>; yields i32:-1</i>
5659 %Z = sext <2 x i16> <i16 8, i16 7> to <2 x i32> <i>; yields <i32 8, i32 7></i>
5664 <!-- _______________________________________________________________________ -->
5666 <a name="i_fptrunc">'<tt>fptrunc .. to</tt>' Instruction</a>
5673 <result> = fptrunc <ty> <value> to <ty2> <i>; yields ty2</i>
5677 <p>The '<tt>fptrunc</tt>' instruction truncates <tt>value</tt> to type
5681 <p>The '<tt>fptrunc</tt>' instruction takes a <a href="#t_floating">floating
5682 point</a> value to cast and a <a href="#t_floating">floating point</a> type
5683 to cast it to. The size of <tt>value</tt> must be larger than the size of
5684 <tt>ty2</tt>. This implies that <tt>fptrunc</tt> cannot be used to make a
5685 <i>no-op cast</i>.</p>
5688 <p>The '<tt>fptrunc</tt>' instruction truncates a <tt>value</tt> from a larger
5689 <a href="#t_floating">floating point</a> type to a smaller
5690 <a href="#t_floating">floating point</a> type. If the value cannot fit
5691 within the destination type, <tt>ty2</tt>, then the results are
5696 %X = fptrunc double 123.0 to float <i>; yields float:123.0</i>
5697 %Y = fptrunc double 1.0E+300 to float <i>; yields undefined</i>
5702 <!-- _______________________________________________________________________ -->
5704 <a name="i_fpext">'<tt>fpext .. to</tt>' Instruction</a>
5711 <result> = fpext <ty> <value> to <ty2> <i>; yields ty2</i>
5715 <p>The '<tt>fpext</tt>' extends a floating point <tt>value</tt> to a larger
5716 floating point value.</p>
5719 <p>The '<tt>fpext</tt>' instruction takes a
5720 <a href="#t_floating">floating point</a> <tt>value</tt> to cast, and
5721 a <a href="#t_floating">floating point</a> type to cast it to. The source
5722 type must be smaller than the destination type.</p>
5725 <p>The '<tt>fpext</tt>' instruction extends the <tt>value</tt> from a smaller
5726 <a href="#t_floating">floating point</a> type to a larger
5727 <a href="#t_floating">floating point</a> type. The <tt>fpext</tt> cannot be
5728 used to make a <i>no-op cast</i> because it always changes bits. Use
5729 <tt>bitcast</tt> to make a <i>no-op cast</i> for a floating point cast.</p>
5733 %X = fpext float 3.125 to double <i>; yields double:3.125000e+00</i>
5734 %Y = fpext double %X to fp128 <i>; yields fp128:0xL00000000000000004000900000000000</i>
5739 <!-- _______________________________________________________________________ -->
5741 <a name="i_fptoui">'<tt>fptoui .. to</tt>' Instruction</a>
5748 <result> = fptoui <ty> <value> to <ty2> <i>; yields ty2</i>
5752 <p>The '<tt>fptoui</tt>' converts a floating point <tt>value</tt> to its
5753 unsigned integer equivalent of type <tt>ty2</tt>.</p>
5756 <p>The '<tt>fptoui</tt>' instruction takes a value to cast, which must be a
5757 scalar or vector <a href="#t_floating">floating point</a> value, and a type
5758 to cast it to <tt>ty2</tt>, which must be an <a href="#t_integer">integer</a>
5759 type. If <tt>ty</tt> is a vector floating point type, <tt>ty2</tt> must be a
5760 vector integer type with the same number of elements as <tt>ty</tt></p>
5763 <p>The '<tt>fptoui</tt>' instruction converts its
5764 <a href="#t_floating">floating point</a> operand into the nearest (rounding
5765 towards zero) unsigned integer value. If the value cannot fit
5766 in <tt>ty2</tt>, the results are undefined.</p>
5770 %X = fptoui double 123.0 to i32 <i>; yields i32:123</i>
5771 %Y = fptoui float 1.0E+300 to i1 <i>; yields undefined:1</i>
5772 %Z = fptoui float 1.04E+17 to i8 <i>; yields undefined:1</i>
5777 <!-- _______________________________________________________________________ -->
5779 <a name="i_fptosi">'<tt>fptosi .. to</tt>' Instruction</a>
5786 <result> = fptosi <ty> <value> to <ty2> <i>; yields ty2</i>
5790 <p>The '<tt>fptosi</tt>' instruction converts
5791 <a href="#t_floating">floating point</a> <tt>value</tt> to
5792 type <tt>ty2</tt>.</p>
5795 <p>The '<tt>fptosi</tt>' instruction takes a value to cast, which must be a
5796 scalar or vector <a href="#t_floating">floating point</a> value, and a type
5797 to cast it to <tt>ty2</tt>, which must be an <a href="#t_integer">integer</a>
5798 type. If <tt>ty</tt> is a vector floating point type, <tt>ty2</tt> must be a
5799 vector integer type with the same number of elements as <tt>ty</tt></p>
5802 <p>The '<tt>fptosi</tt>' instruction converts its
5803 <a href="#t_floating">floating point</a> operand into the nearest (rounding
5804 towards zero) signed integer value. If the value cannot fit in <tt>ty2</tt>,
5805 the results are undefined.</p>
5809 %X = fptosi double -123.0 to i32 <i>; yields i32:-123</i>
5810 %Y = fptosi float 1.0E-247 to i1 <i>; yields undefined:1</i>
5811 %Z = fptosi float 1.04E+17 to i8 <i>; yields undefined:1</i>
5816 <!-- _______________________________________________________________________ -->
5818 <a name="i_uitofp">'<tt>uitofp .. to</tt>' Instruction</a>
5825 <result> = uitofp <ty> <value> to <ty2> <i>; yields ty2</i>
5829 <p>The '<tt>uitofp</tt>' instruction regards <tt>value</tt> as an unsigned
5830 integer and converts that value to the <tt>ty2</tt> type.</p>
5833 <p>The '<tt>uitofp</tt>' instruction takes a value to cast, which must be a
5834 scalar or vector <a href="#t_integer">integer</a> value, and a type to cast
5835 it to <tt>ty2</tt>, which must be an <a href="#t_floating">floating point</a>
5836 type. If <tt>ty</tt> is a vector integer type, <tt>ty2</tt> must be a vector
5837 floating point type with the same number of elements as <tt>ty</tt></p>
5840 <p>The '<tt>uitofp</tt>' instruction interprets its operand as an unsigned
5841 integer quantity and converts it to the corresponding floating point
5842 value. If the value cannot fit in the floating point value, the results are
5847 %X = uitofp i32 257 to float <i>; yields float:257.0</i>
5848 %Y = uitofp i8 -1 to double <i>; yields double:255.0</i>
5853 <!-- _______________________________________________________________________ -->
5855 <a name="i_sitofp">'<tt>sitofp .. to</tt>' Instruction</a>
5862 <result> = sitofp <ty> <value> to <ty2> <i>; yields ty2</i>
5866 <p>The '<tt>sitofp</tt>' instruction regards <tt>value</tt> as a signed integer
5867 and converts that value to the <tt>ty2</tt> type.</p>
5870 <p>The '<tt>sitofp</tt>' instruction takes a value to cast, which must be a
5871 scalar or vector <a href="#t_integer">integer</a> value, and a type to cast
5872 it to <tt>ty2</tt>, which must be an <a href="#t_floating">floating point</a>
5873 type. If <tt>ty</tt> is a vector integer type, <tt>ty2</tt> must be a vector
5874 floating point type with the same number of elements as <tt>ty</tt></p>
5877 <p>The '<tt>sitofp</tt>' instruction interprets its operand as a signed integer
5878 quantity and converts it to the corresponding floating point value. If the
5879 value cannot fit in the floating point value, the results are undefined.</p>
5883 %X = sitofp i32 257 to float <i>; yields float:257.0</i>
5884 %Y = sitofp i8 -1 to double <i>; yields double:-1.0</i>
5889 <!-- _______________________________________________________________________ -->
5891 <a name="i_ptrtoint">'<tt>ptrtoint .. to</tt>' Instruction</a>
5898 <result> = ptrtoint <ty> <value> to <ty2> <i>; yields ty2</i>
5902 <p>The '<tt>ptrtoint</tt>' instruction converts the pointer or a vector of
5903 pointers <tt>value</tt> to
5904 the integer (or vector of integers) type <tt>ty2</tt>.</p>
5907 <p>The '<tt>ptrtoint</tt>' instruction takes a <tt>value</tt> to cast, which
5908 must be a a value of type <a href="#t_pointer">pointer</a> or a vector of
5909 pointers, and a type to cast it to
5910 <tt>ty2</tt>, which must be an <a href="#t_integer">integer</a> or a vector
5911 of integers type.</p>
5914 <p>The '<tt>ptrtoint</tt>' instruction converts <tt>value</tt> to integer type
5915 <tt>ty2</tt> by interpreting the pointer value as an integer and either
5916 truncating or zero extending that value to the size of the integer type. If
5917 <tt>value</tt> is smaller than <tt>ty2</tt> then a zero extension is done. If
5918 <tt>value</tt> is larger than <tt>ty2</tt> then a truncation is done. If they
5919 are the same size, then nothing is done (<i>no-op cast</i>) other than a type
5924 %X = ptrtoint i32* %P to i8 <i>; yields truncation on 32-bit architecture</i>
5925 %Y = ptrtoint i32* %P to i64 <i>; yields zero extension on 32-bit architecture</i>
5926 %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>
5931 <!-- _______________________________________________________________________ -->
5933 <a name="i_inttoptr">'<tt>inttoptr .. to</tt>' Instruction</a>
5940 <result> = inttoptr <ty> <value> to <ty2> <i>; yields ty2</i>
5944 <p>The '<tt>inttoptr</tt>' instruction converts an integer <tt>value</tt> to a
5945 pointer type, <tt>ty2</tt>.</p>
5948 <p>The '<tt>inttoptr</tt>' instruction takes an <a href="#t_integer">integer</a>
5949 value to cast, and a type to cast it to, which must be a
5950 <a href="#t_pointer">pointer</a> type.</p>
5953 <p>The '<tt>inttoptr</tt>' instruction converts <tt>value</tt> to type
5954 <tt>ty2</tt> by applying either a zero extension or a truncation depending on
5955 the size of the integer <tt>value</tt>. If <tt>value</tt> is larger than the
5956 size of a pointer then a truncation is done. If <tt>value</tt> is smaller
5957 than the size of a pointer then a zero extension is done. If they are the
5958 same size, nothing is done (<i>no-op cast</i>).</p>
5962 %X = inttoptr i32 255 to i32* <i>; yields zero extension on 64-bit architecture</i>
5963 %Y = inttoptr i32 255 to i32* <i>; yields no-op on 32-bit architecture</i>
5964 %Z = inttoptr i64 0 to i32* <i>; yields truncation on 32-bit architecture</i>
5965 %Z = inttoptr <4 x i32> %G to <4 x i8*><i>; yields truncation of vector G to four pointers</i>
5970 <!-- _______________________________________________________________________ -->
5972 <a name="i_bitcast">'<tt>bitcast .. to</tt>' Instruction</a>
5979 <result> = bitcast <ty> <value> to <ty2> <i>; yields ty2</i>
5983 <p>The '<tt>bitcast</tt>' instruction converts <tt>value</tt> to type
5984 <tt>ty2</tt> without changing any bits.</p>
5987 <p>The '<tt>bitcast</tt>' instruction takes a value to cast, which must be a
5988 non-aggregate first class value, and a type to cast it to, which must also be
5989 a non-aggregate <a href="#t_firstclass">first class</a> type. The bit sizes
5990 of <tt>value</tt> and the destination type, <tt>ty2</tt>, must be
5991 identical. If the source type is a pointer, the destination type must also be
5992 a pointer. This instruction supports bitwise conversion of vectors to
5993 integers and to vectors of other types (as long as they have the same
5997 <p>The '<tt>bitcast</tt>' instruction converts <tt>value</tt> to type
5998 <tt>ty2</tt>. It is always a <i>no-op cast</i> because no bits change with
5999 this conversion. The conversion is done as if the <tt>value</tt> had been
6000 stored to memory and read back as type <tt>ty2</tt>.
6001 Pointer (or vector of pointers) types may only be converted to other pointer
6002 (or vector of pointers) types with this instruction. To convert
6003 pointers to other types, use the <a href="#i_inttoptr">inttoptr</a> or
6004 <a href="#i_ptrtoint">ptrtoint</a> instructions first.</p>
6008 %X = bitcast i8 255 to i8 <i>; yields i8 :-1</i>
6009 %Y = bitcast i32* %x to sint* <i>; yields sint*:%x</i>
6010 %Z = bitcast <2 x int> %V to i64; <i>; yields i64: %V</i>
6011 %Z = bitcast <2 x i32*> %V to <2 x i64*> <i>; yields <2 x i64*></i>
6018 <!-- ======================================================================= -->
6020 <a name="otherops">Other Operations</a>
6025 <p>The instructions in this category are the "miscellaneous" instructions, which
6026 defy better classification.</p>
6028 <!-- _______________________________________________________________________ -->
6030 <a name="i_icmp">'<tt>icmp</tt>' Instruction</a>
6037 <result> = icmp <cond> <ty> <op1>, <op2> <i>; yields {i1} or {<N x i1>}:result</i>
6041 <p>The '<tt>icmp</tt>' instruction returns a boolean value or a vector of
6042 boolean values based on comparison of its two integer, integer vector,
6043 pointer, or pointer vector operands.</p>
6046 <p>The '<tt>icmp</tt>' instruction takes three operands. The first operand is
6047 the condition code indicating the kind of comparison to perform. It is not a
6048 value, just a keyword. The possible condition code are:</p>
6051 <li><tt>eq</tt>: equal</li>
6052 <li><tt>ne</tt>: not equal </li>
6053 <li><tt>ugt</tt>: unsigned greater than</li>
6054 <li><tt>uge</tt>: unsigned greater or equal</li>
6055 <li><tt>ult</tt>: unsigned less than</li>
6056 <li><tt>ule</tt>: unsigned less or equal</li>
6057 <li><tt>sgt</tt>: signed greater than</li>
6058 <li><tt>sge</tt>: signed greater or equal</li>
6059 <li><tt>slt</tt>: signed less than</li>
6060 <li><tt>sle</tt>: signed less or equal</li>
6063 <p>The remaining two arguments must be <a href="#t_integer">integer</a> or
6064 <a href="#t_pointer">pointer</a> or integer <a href="#t_vector">vector</a>
6065 typed. They must also be identical types.</p>
6068 <p>The '<tt>icmp</tt>' compares <tt>op1</tt> and <tt>op2</tt> according to the
6069 condition code given as <tt>cond</tt>. The comparison performed always yields
6070 either an <a href="#t_integer"><tt>i1</tt></a> or vector of <tt>i1</tt>
6071 result, as follows:</p>
6074 <li><tt>eq</tt>: yields <tt>true</tt> if the operands are equal,
6075 <tt>false</tt> otherwise. No sign interpretation is necessary or
6078 <li><tt>ne</tt>: yields <tt>true</tt> if the operands are unequal,
6079 <tt>false</tt> otherwise. No sign interpretation is necessary or
6082 <li><tt>ugt</tt>: interprets the operands as unsigned values and yields
6083 <tt>true</tt> if <tt>op1</tt> is greater than <tt>op2</tt>.</li>
6085 <li><tt>uge</tt>: interprets the operands as unsigned values and yields
6086 <tt>true</tt> if <tt>op1</tt> is greater than or equal
6087 to <tt>op2</tt>.</li>
6089 <li><tt>ult</tt>: interprets the operands as unsigned values and yields
6090 <tt>true</tt> if <tt>op1</tt> is less than <tt>op2</tt>.</li>
6092 <li><tt>ule</tt>: interprets the operands as unsigned values and yields
6093 <tt>true</tt> if <tt>op1</tt> is less than or equal to <tt>op2</tt>.</li>
6095 <li><tt>sgt</tt>: interprets the operands as signed values and yields
6096 <tt>true</tt> if <tt>op1</tt> is greater than <tt>op2</tt>.</li>
6098 <li><tt>sge</tt>: interprets the operands as signed values and yields
6099 <tt>true</tt> if <tt>op1</tt> is greater than or equal
6100 to <tt>op2</tt>.</li>
6102 <li><tt>slt</tt>: interprets the operands as signed values and yields
6103 <tt>true</tt> if <tt>op1</tt> is less than <tt>op2</tt>.</li>
6105 <li><tt>sle</tt>: interprets the operands as signed values and yields
6106 <tt>true</tt> if <tt>op1</tt> is less than or equal to <tt>op2</tt>.</li>
6109 <p>If the operands are <a href="#t_pointer">pointer</a> typed, the pointer
6110 values are compared as if they were integers.</p>
6112 <p>If the operands are integer vectors, then they are compared element by
6113 element. The result is an <tt>i1</tt> vector with the same number of elements
6114 as the values being compared. Otherwise, the result is an <tt>i1</tt>.</p>
6118 <result> = icmp eq i32 4, 5 <i>; yields: result=false</i>
6119 <result> = icmp ne float* %X, %X <i>; yields: result=false</i>
6120 <result> = icmp ult i16 4, 5 <i>; yields: result=true</i>
6121 <result> = icmp sgt i16 4, 5 <i>; yields: result=false</i>
6122 <result> = icmp ule i16 -4, 5 <i>; yields: result=false</i>
6123 <result> = icmp sge i16 4, 5 <i>; yields: result=false</i>
6126 <p>Note that the code generator does not yet support vector types with
6127 the <tt>icmp</tt> instruction.</p>
6131 <!-- _______________________________________________________________________ -->
6133 <a name="i_fcmp">'<tt>fcmp</tt>' Instruction</a>
6140 <result> = fcmp <cond> <ty> <op1>, <op2> <i>; yields {i1} or {<N x i1>}:result</i>
6144 <p>The '<tt>fcmp</tt>' instruction returns a boolean value or vector of boolean
6145 values based on comparison of its operands.</p>
6147 <p>If the operands are floating point scalars, then the result type is a boolean
6148 (<a href="#t_integer"><tt>i1</tt></a>).</p>
6150 <p>If the operands are floating point vectors, then the result type is a vector
6151 of boolean with the same number of elements as the operands being
6155 <p>The '<tt>fcmp</tt>' instruction takes three operands. The first operand is
6156 the condition code indicating the kind of comparison to perform. It is not a
6157 value, just a keyword. The possible condition code are:</p>
6160 <li><tt>false</tt>: no comparison, always returns false</li>
6161 <li><tt>oeq</tt>: ordered and equal</li>
6162 <li><tt>ogt</tt>: ordered and greater than </li>
6163 <li><tt>oge</tt>: ordered and greater than or equal</li>
6164 <li><tt>olt</tt>: ordered and less than </li>
6165 <li><tt>ole</tt>: ordered and less than or equal</li>
6166 <li><tt>one</tt>: ordered and not equal</li>
6167 <li><tt>ord</tt>: ordered (no nans)</li>
6168 <li><tt>ueq</tt>: unordered or equal</li>
6169 <li><tt>ugt</tt>: unordered or greater than </li>
6170 <li><tt>uge</tt>: unordered or greater than or equal</li>
6171 <li><tt>ult</tt>: unordered or less than </li>
6172 <li><tt>ule</tt>: unordered or less than or equal</li>
6173 <li><tt>une</tt>: unordered or not equal</li>
6174 <li><tt>uno</tt>: unordered (either nans)</li>
6175 <li><tt>true</tt>: no comparison, always returns true</li>
6178 <p><i>Ordered</i> means that neither operand is a QNAN while
6179 <i>unordered</i> means that either operand may be a QNAN.</p>
6181 <p>Each of <tt>val1</tt> and <tt>val2</tt> arguments must be either
6182 a <a href="#t_floating">floating point</a> type or
6183 a <a href="#t_vector">vector</a> of floating point type. They must have
6184 identical types.</p>
6187 <p>The '<tt>fcmp</tt>' instruction compares <tt>op1</tt> and <tt>op2</tt>
6188 according to the condition code given as <tt>cond</tt>. If the operands are
6189 vectors, then the vectors are compared element by element. Each comparison
6190 performed always yields an <a href="#t_integer">i1</a> result, as
6194 <li><tt>false</tt>: always yields <tt>false</tt>, regardless of operands.</li>
6196 <li><tt>oeq</tt>: yields <tt>true</tt> if both operands are not a QNAN and
6197 <tt>op1</tt> is equal to <tt>op2</tt>.</li>
6199 <li><tt>ogt</tt>: yields <tt>true</tt> if both operands are not a QNAN and
6200 <tt>op1</tt> is greater than <tt>op2</tt>.</li>
6202 <li><tt>oge</tt>: yields <tt>true</tt> if both operands are not a QNAN and
6203 <tt>op1</tt> is greater than or equal to <tt>op2</tt>.</li>
6205 <li><tt>olt</tt>: yields <tt>true</tt> if both operands are not a QNAN and
6206 <tt>op1</tt> is less than <tt>op2</tt>.</li>
6208 <li><tt>ole</tt>: yields <tt>true</tt> if both operands are not a QNAN and
6209 <tt>op1</tt> is less than or equal to <tt>op2</tt>.</li>
6211 <li><tt>one</tt>: yields <tt>true</tt> if both operands are not a QNAN and
6212 <tt>op1</tt> is not equal to <tt>op2</tt>.</li>
6214 <li><tt>ord</tt>: yields <tt>true</tt> if both operands are not a QNAN.</li>
6216 <li><tt>ueq</tt>: yields <tt>true</tt> if either operand is a QNAN or
6217 <tt>op1</tt> is equal to <tt>op2</tt>.</li>
6219 <li><tt>ugt</tt>: yields <tt>true</tt> if either operand is a QNAN or
6220 <tt>op1</tt> is greater than <tt>op2</tt>.</li>
6222 <li><tt>uge</tt>: yields <tt>true</tt> if either operand is a QNAN or
6223 <tt>op1</tt> is greater than or equal to <tt>op2</tt>.</li>
6225 <li><tt>ult</tt>: yields <tt>true</tt> if either operand is a QNAN or
6226 <tt>op1</tt> is less than <tt>op2</tt>.</li>
6228 <li><tt>ule</tt>: yields <tt>true</tt> if either operand is a QNAN or
6229 <tt>op1</tt> is less than or equal to <tt>op2</tt>.</li>
6231 <li><tt>une</tt>: yields <tt>true</tt> if either operand is a QNAN or
6232 <tt>op1</tt> is not equal to <tt>op2</tt>.</li>
6234 <li><tt>uno</tt>: yields <tt>true</tt> if either operand is a QNAN.</li>
6236 <li><tt>true</tt>: always yields <tt>true</tt>, regardless of operands.</li>
6241 <result> = fcmp oeq float 4.0, 5.0 <i>; yields: result=false</i>
6242 <result> = fcmp one float 4.0, 5.0 <i>; yields: result=true</i>
6243 <result> = fcmp olt float 4.0, 5.0 <i>; yields: result=true</i>
6244 <result> = fcmp ueq double 1.0, 2.0 <i>; yields: result=false</i>
6247 <p>Note that the code generator does not yet support vector types with
6248 the <tt>fcmp</tt> instruction.</p>
6252 <!-- _______________________________________________________________________ -->
6254 <a name="i_phi">'<tt>phi</tt>' Instruction</a>
6261 <result> = phi <ty> [ <val0>, <label0>], ...
6265 <p>The '<tt>phi</tt>' instruction is used to implement the φ node in the
6266 SSA graph representing the function.</p>
6269 <p>The type of the incoming values is specified with the first type field. After
6270 this, the '<tt>phi</tt>' instruction takes a list of pairs as arguments, with
6271 one pair for each predecessor basic block of the current block. Only values
6272 of <a href="#t_firstclass">first class</a> type may be used as the value
6273 arguments to the PHI node. Only labels may be used as the label
6276 <p>There must be no non-phi instructions between the start of a basic block and
6277 the PHI instructions: i.e. PHI instructions must be first in a basic
6280 <p>For the purposes of the SSA form, the use of each incoming value is deemed to
6281 occur on the edge from the corresponding predecessor block to the current
6282 block (but after any definition of an '<tt>invoke</tt>' instruction's return
6283 value on the same edge).</p>
6286 <p>At runtime, the '<tt>phi</tt>' instruction logically takes on the value
6287 specified by the pair corresponding to the predecessor basic block that
6288 executed just prior to the current block.</p>
6292 Loop: ; Infinite loop that counts from 0 on up...
6293 %indvar = phi i32 [ 0, %LoopHeader ], [ %nextindvar, %Loop ]
6294 %nextindvar = add i32 %indvar, 1
6300 <!-- _______________________________________________________________________ -->
6302 <a name="i_select">'<tt>select</tt>' Instruction</a>
6309 <result> = select <i>selty</i> <cond>, <ty> <val1>, <ty> <val2> <i>; yields ty</i>
6311 <i>selty</i> is either i1 or {<N x i1>}
6315 <p>The '<tt>select</tt>' instruction is used to choose one value based on a
6316 condition, without branching.</p>
6320 <p>The '<tt>select</tt>' instruction requires an 'i1' value or a vector of 'i1'
6321 values indicating the condition, and two values of the
6322 same <a href="#t_firstclass">first class</a> type. If the val1/val2 are
6323 vectors and the condition is a scalar, then entire vectors are selected, not
6324 individual elements.</p>
6327 <p>If the condition is an i1 and it evaluates to 1, the instruction returns the
6328 first value argument; otherwise, it returns the second value argument.</p>
6330 <p>If the condition is a vector of i1, then the value arguments must be vectors
6331 of the same size, and the selection is done element by element.</p>
6335 %X = select i1 true, i8 17, i8 42 <i>; yields i8:17</i>
6340 <!-- _______________________________________________________________________ -->
6342 <a name="i_call">'<tt>call</tt>' Instruction</a>
6349 <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>]
6353 <p>The '<tt>call</tt>' instruction represents a simple function call.</p>
6356 <p>This instruction requires several arguments:</p>
6359 <li>The optional "tail" marker indicates that the callee function does not
6360 access any allocas or varargs in the caller. Note that calls may be
6361 marked "tail" even if they do not occur before
6362 a <a href="#i_ret"><tt>ret</tt></a> instruction. If the "tail" marker is
6363 present, the function call is eligible for tail call optimization,
6364 but <a href="CodeGenerator.html#tailcallopt">might not in fact be
6365 optimized into a jump</a>. The code generator may optimize calls marked
6366 "tail" with either 1) automatic <a href="CodeGenerator.html#sibcallopt">
6367 sibling call optimization</a> when the caller and callee have
6368 matching signatures, or 2) forced tail call optimization when the
6369 following extra requirements are met:
6371 <li>Caller and callee both have the calling
6372 convention <tt>fastcc</tt>.</li>
6373 <li>The call is in tail position (ret immediately follows call and ret
6374 uses value of call or is void).</li>
6375 <li>Option <tt>-tailcallopt</tt> is enabled,
6376 or <code>llvm::GuaranteedTailCallOpt</code> is <code>true</code>.</li>
6377 <li><a href="CodeGenerator.html#tailcallopt">Platform specific
6378 constraints are met.</a></li>
6382 <li>The optional "cconv" marker indicates which <a href="#callingconv">calling
6383 convention</a> the call should use. If none is specified, the call
6384 defaults to using C calling conventions. The calling convention of the
6385 call must match the calling convention of the target function, or else the
6386 behavior is undefined.</li>
6388 <li>The optional <a href="#paramattrs">Parameter Attributes</a> list for
6389 return values. Only '<tt>zeroext</tt>', '<tt>signext</tt>', and
6390 '<tt>inreg</tt>' attributes are valid here.</li>
6392 <li>'<tt>ty</tt>': the type of the call instruction itself which is also the
6393 type of the return value. Functions that return no value are marked
6394 <tt><a href="#t_void">void</a></tt>.</li>
6396 <li>'<tt>fnty</tt>': shall be the signature of the pointer to function value
6397 being invoked. The argument types must match the types implied by this
6398 signature. This type can be omitted if the function is not varargs and if
6399 the function type does not return a pointer to a function.</li>
6401 <li>'<tt>fnptrval</tt>': An LLVM value containing a pointer to a function to
6402 be invoked. In most cases, this is a direct function invocation, but
6403 indirect <tt>call</tt>s are just as possible, calling an arbitrary pointer
6404 to function value.</li>
6406 <li>'<tt>function args</tt>': argument list whose types match the function
6407 signature argument types and parameter attributes. All arguments must be
6408 of <a href="#t_firstclass">first class</a> type. If the function
6409 signature indicates the function accepts a variable number of arguments,
6410 the extra arguments can be specified.</li>
6412 <li>The optional <a href="#fnattrs">function attributes</a> list. Only
6413 '<tt>noreturn</tt>', '<tt>nounwind</tt>', '<tt>readonly</tt>' and
6414 '<tt>readnone</tt>' attributes are valid here.</li>
6418 <p>The '<tt>call</tt>' instruction is used to cause control flow to transfer to
6419 a specified function, with its incoming arguments bound to the specified
6420 values. Upon a '<tt><a href="#i_ret">ret</a></tt>' instruction in the called
6421 function, control flow continues with the instruction after the function
6422 call, and the return value of the function is bound to the result
6427 %retval = call i32 @test(i32 %argc)
6428 call i32 (i8*, ...)* @printf(i8* %msg, i32 12, i8 42) <i>; yields i32</i>
6429 %X = tail call i32 @foo() <i>; yields i32</i>
6430 %Y = tail call <a href="#callingconv">fastcc</a> i32 @foo() <i>; yields i32</i>
6431 call void %foo(i8 97 signext)
6433 %struct.A = type { i32, i8 }
6434 %r = call %struct.A @foo() <i>; yields { 32, i8 }</i>
6435 %gr = extractvalue %struct.A %r, 0 <i>; yields i32</i>
6436 %gr1 = extractvalue %struct.A %r, 1 <i>; yields i8</i>
6437 %Z = call void @foo() noreturn <i>; indicates that %foo never returns normally</i>
6438 %ZZ = call zeroext i32 @bar() <i>; Return value is %zero extended</i>
6441 <p>llvm treats calls to some functions with names and arguments that match the
6442 standard C99 library as being the C99 library functions, and may perform
6443 optimizations or generate code for them under that assumption. This is
6444 something we'd like to change in the future to provide better support for
6445 freestanding environments and non-C-based languages.</p>
6449 <!-- _______________________________________________________________________ -->
6451 <a name="i_va_arg">'<tt>va_arg</tt>' Instruction</a>
6458 <resultval> = va_arg <va_list*> <arglist>, <argty>
6462 <p>The '<tt>va_arg</tt>' instruction is used to access arguments passed through
6463 the "variable argument" area of a function call. It is used to implement the
6464 <tt>va_arg</tt> macro in C.</p>
6467 <p>This instruction takes a <tt>va_list*</tt> value and the type of the
6468 argument. It returns a value of the specified argument type and increments
6469 the <tt>va_list</tt> to point to the next argument. The actual type
6470 of <tt>va_list</tt> is target specific.</p>
6473 <p>The '<tt>va_arg</tt>' instruction loads an argument of the specified type
6474 from the specified <tt>va_list</tt> and causes the <tt>va_list</tt> to point
6475 to the next argument. For more information, see the variable argument
6476 handling <a href="#int_varargs">Intrinsic Functions</a>.</p>
6478 <p>It is legal for this instruction to be called in a function which does not
6479 take a variable number of arguments, for example, the <tt>vfprintf</tt>
6482 <p><tt>va_arg</tt> is an LLVM instruction instead of
6483 an <a href="#intrinsics">intrinsic function</a> because it takes a type as an
6487 <p>See the <a href="#int_varargs">variable argument processing</a> section.</p>
6489 <p>Note that the code generator does not yet fully support va_arg on many
6490 targets. Also, it does not currently support va_arg with aggregate types on
6495 <!-- _______________________________________________________________________ -->
6497 <a name="i_landingpad">'<tt>landingpad</tt>' Instruction</a>
6504 <resultval> = landingpad <resultty> personality <type> <pers_fn> <clause>+
6505 <resultval> = landingpad <resultty> personality <type> <pers_fn> cleanup <clause>*
6507 <clause> := catch <type> <value>
6508 <clause> := filter <array constant type> <array constant>
6512 <p>The '<tt>landingpad</tt>' instruction is used by
6513 <a href="ExceptionHandling.html#overview">LLVM's exception handling
6514 system</a> to specify that a basic block is a landing pad — one where
6515 the exception lands, and corresponds to the code found in the
6516 <i><tt>catch</tt></i> portion of a <i><tt>try/catch</tt></i> sequence. It
6517 defines values supplied by the personality function (<tt>pers_fn</tt>) upon
6518 re-entry to the function. The <tt>resultval</tt> has the
6519 type <tt>resultty</tt>.</p>
6522 <p>This instruction takes a <tt>pers_fn</tt> value. This is the personality
6523 function associated with the unwinding mechanism. The optional
6524 <tt>cleanup</tt> flag indicates that the landing pad block is a cleanup.</p>
6526 <p>A <tt>clause</tt> begins with the clause type — <tt>catch</tt>
6527 or <tt>filter</tt> — and contains the global variable representing the
6528 "type" that may be caught or filtered respectively. Unlike the
6529 <tt>catch</tt> clause, the <tt>filter</tt> clause takes an array constant as
6530 its argument. Use "<tt>[0 x i8**] undef</tt>" for a filter which cannot
6531 throw. The '<tt>landingpad</tt>' instruction must contain <em>at least</em>
6532 one <tt>clause</tt> or the <tt>cleanup</tt> flag.</p>
6535 <p>The '<tt>landingpad</tt>' instruction defines the values which are set by the
6536 personality function (<tt>pers_fn</tt>) upon re-entry to the function, and
6537 therefore the "result type" of the <tt>landingpad</tt> instruction. As with
6538 calling conventions, how the personality function results are represented in
6539 LLVM IR is target specific.</p>
6541 <p>The clauses are applied in order from top to bottom. If two
6542 <tt>landingpad</tt> instructions are merged together through inlining, the
6543 clauses from the calling function are appended to the list of clauses.
6544 When the call stack is being unwound due to an exception being thrown, the
6545 exception is compared against each <tt>clause</tt> in turn. If it doesn't
6546 match any of the clauses, and the <tt>cleanup</tt> flag is not set, then
6547 unwinding continues further up the call stack.</p>
6549 <p>The <tt>landingpad</tt> instruction has several restrictions:</p>
6552 <li>A landing pad block is a basic block which is the unwind destination of an
6553 '<tt>invoke</tt>' instruction.</li>
6554 <li>A landing pad block must have a '<tt>landingpad</tt>' instruction as its
6555 first non-PHI instruction.</li>
6556 <li>There can be only one '<tt>landingpad</tt>' instruction within the landing
6558 <li>A basic block that is not a landing pad block may not include a
6559 '<tt>landingpad</tt>' instruction.</li>
6560 <li>All '<tt>landingpad</tt>' instructions in a function must have the same
6561 personality function.</li>
6566 ;; A landing pad which can catch an integer.
6567 %res = landingpad { i8*, i32 } personality i32 (...)* @__gxx_personality_v0
6569 ;; A landing pad that is a cleanup.
6570 %res = landingpad { i8*, i32 } personality i32 (...)* @__gxx_personality_v0
6572 ;; A landing pad which can catch an integer and can only throw a double.
6573 %res = landingpad { i8*, i32 } personality i32 (...)* @__gxx_personality_v0
6575 filter [1 x i8**] [@_ZTId]
6584 <!-- *********************************************************************** -->
6585 <h2><a name="intrinsics">Intrinsic Functions</a></h2>
6586 <!-- *********************************************************************** -->
6590 <p>LLVM supports the notion of an "intrinsic function". These functions have
6591 well known names and semantics and are required to follow certain
6592 restrictions. Overall, these intrinsics represent an extension mechanism for
6593 the LLVM language that does not require changing all of the transformations
6594 in LLVM when adding to the language (or the bitcode reader/writer, the
6595 parser, etc...).</p>
6597 <p>Intrinsic function names must all start with an "<tt>llvm.</tt>" prefix. This
6598 prefix is reserved in LLVM for intrinsic names; thus, function names may not
6599 begin with this prefix. Intrinsic functions must always be external
6600 functions: you cannot define the body of intrinsic functions. Intrinsic
6601 functions may only be used in call or invoke instructions: it is illegal to
6602 take the address of an intrinsic function. Additionally, because intrinsic
6603 functions are part of the LLVM language, it is required if any are added that
6604 they be documented here.</p>
6606 <p>Some intrinsic functions can be overloaded, i.e., the intrinsic represents a
6607 family of functions that perform the same operation but on different data
6608 types. Because LLVM can represent over 8 million different integer types,
6609 overloading is used commonly to allow an intrinsic function to operate on any
6610 integer type. One or more of the argument types or the result type can be
6611 overloaded to accept any integer type. Argument types may also be defined as
6612 exactly matching a previous argument's type or the result type. This allows
6613 an intrinsic function which accepts multiple arguments, but needs all of them
6614 to be of the same type, to only be overloaded with respect to a single
6615 argument or the result.</p>
6617 <p>Overloaded intrinsics will have the names of its overloaded argument types
6618 encoded into its function name, each preceded by a period. Only those types
6619 which are overloaded result in a name suffix. Arguments whose type is matched
6620 against another type do not. For example, the <tt>llvm.ctpop</tt> function
6621 can take an integer of any width and returns an integer of exactly the same
6622 integer width. This leads to a family of functions such as
6623 <tt>i8 @llvm.ctpop.i8(i8 %val)</tt> and <tt>i29 @llvm.ctpop.i29(i29
6624 %val)</tt>. Only one type, the return type, is overloaded, and only one type
6625 suffix is required. Because the argument's type is matched against the return
6626 type, it does not require its own name suffix.</p>
6628 <p>To learn how to add an intrinsic function, please see the
6629 <a href="ExtendingLLVM.html">Extending LLVM Guide</a>.</p>
6631 <!-- ======================================================================= -->
6633 <a name="int_varargs">Variable Argument Handling Intrinsics</a>
6638 <p>Variable argument support is defined in LLVM with
6639 the <a href="#i_va_arg"><tt>va_arg</tt></a> instruction and these three
6640 intrinsic functions. These functions are related to the similarly named
6641 macros defined in the <tt><stdarg.h></tt> header file.</p>
6643 <p>All of these functions operate on arguments that use a target-specific value
6644 type "<tt>va_list</tt>". The LLVM assembly language reference manual does
6645 not define what this type is, so all transformations should be prepared to
6646 handle these functions regardless of the type used.</p>
6648 <p>This example shows how the <a href="#i_va_arg"><tt>va_arg</tt></a>
6649 instruction and the variable argument handling intrinsic functions are
6652 <pre class="doc_code">
6653 define i32 @test(i32 %X, ...) {
6654 ; Initialize variable argument processing
6656 %ap2 = bitcast i8** %ap to i8*
6657 call void @llvm.va_start(i8* %ap2)
6659 ; Read a single integer argument
6660 %tmp = va_arg i8** %ap, i32
6662 ; Demonstrate usage of llvm.va_copy and llvm.va_end
6664 %aq2 = bitcast i8** %aq to i8*
6665 call void @llvm.va_copy(i8* %aq2, i8* %ap2)
6666 call void @llvm.va_end(i8* %aq2)
6668 ; Stop processing of arguments.
6669 call void @llvm.va_end(i8* %ap2)
6673 declare void @llvm.va_start(i8*)
6674 declare void @llvm.va_copy(i8*, i8*)
6675 declare void @llvm.va_end(i8*)
6678 <!-- _______________________________________________________________________ -->
6680 <a name="int_va_start">'<tt>llvm.va_start</tt>' Intrinsic</a>
6688 declare void %llvm.va_start(i8* <arglist>)
6692 <p>The '<tt>llvm.va_start</tt>' intrinsic initializes <tt>*<arglist></tt>
6693 for subsequent use by <tt><a href="#i_va_arg">va_arg</a></tt>.</p>
6696 <p>The argument is a pointer to a <tt>va_list</tt> element to initialize.</p>
6699 <p>The '<tt>llvm.va_start</tt>' intrinsic works just like the <tt>va_start</tt>
6700 macro available in C. In a target-dependent way, it initializes
6701 the <tt>va_list</tt> element to which the argument points, so that the next
6702 call to <tt>va_arg</tt> will produce the first variable argument passed to
6703 the function. Unlike the C <tt>va_start</tt> macro, this intrinsic does not
6704 need to know the last argument of the function as the compiler can figure
6709 <!-- _______________________________________________________________________ -->
6711 <a name="int_va_end">'<tt>llvm.va_end</tt>' Intrinsic</a>
6718 declare void @llvm.va_end(i8* <arglist>)
6722 <p>The '<tt>llvm.va_end</tt>' intrinsic destroys <tt>*<arglist></tt>,
6723 which has been initialized previously
6724 with <tt><a href="#int_va_start">llvm.va_start</a></tt>
6725 or <tt><a href="#i_va_copy">llvm.va_copy</a></tt>.</p>
6728 <p>The argument is a pointer to a <tt>va_list</tt> to destroy.</p>
6731 <p>The '<tt>llvm.va_end</tt>' intrinsic works just like the <tt>va_end</tt>
6732 macro available in C. In a target-dependent way, it destroys
6733 the <tt>va_list</tt> element to which the argument points. Calls
6734 to <a href="#int_va_start"><tt>llvm.va_start</tt></a>
6735 and <a href="#int_va_copy"> <tt>llvm.va_copy</tt></a> must be matched exactly
6736 with calls to <tt>llvm.va_end</tt>.</p>
6740 <!-- _______________________________________________________________________ -->
6742 <a name="int_va_copy">'<tt>llvm.va_copy</tt>' Intrinsic</a>
6749 declare void @llvm.va_copy(i8* <destarglist>, i8* <srcarglist>)
6753 <p>The '<tt>llvm.va_copy</tt>' intrinsic copies the current argument position
6754 from the source argument list to the destination argument list.</p>
6757 <p>The first argument is a pointer to a <tt>va_list</tt> element to initialize.
6758 The second argument is a pointer to a <tt>va_list</tt> element to copy
6762 <p>The '<tt>llvm.va_copy</tt>' intrinsic works just like the <tt>va_copy</tt>
6763 macro available in C. In a target-dependent way, it copies the
6764 source <tt>va_list</tt> element into the destination <tt>va_list</tt>
6765 element. This intrinsic is necessary because
6766 the <tt><a href="#int_va_start"> llvm.va_start</a></tt> intrinsic may be
6767 arbitrarily complex and require, for example, memory allocation.</p>
6773 <!-- ======================================================================= -->
6775 <a name="int_gc">Accurate Garbage Collection Intrinsics</a>
6780 <p>LLVM support for <a href="GarbageCollection.html">Accurate Garbage
6781 Collection</a> (GC) requires the implementation and generation of these
6782 intrinsics. These intrinsics allow identification of <a href="#int_gcroot">GC
6783 roots on the stack</a>, as well as garbage collector implementations that
6784 require <a href="#int_gcread">read</a> and <a href="#int_gcwrite">write</a>
6785 barriers. Front-ends for type-safe garbage collected languages should generate
6786 these intrinsics to make use of the LLVM garbage collectors. For more details,
6787 see <a href="GarbageCollection.html">Accurate Garbage Collection with
6790 <p>The garbage collection intrinsics only operate on objects in the generic
6791 address space (address space zero).</p>
6793 <!-- _______________________________________________________________________ -->
6795 <a name="int_gcroot">'<tt>llvm.gcroot</tt>' Intrinsic</a>
6802 declare void @llvm.gcroot(i8** %ptrloc, i8* %metadata)
6806 <p>The '<tt>llvm.gcroot</tt>' intrinsic declares the existence of a GC root to
6807 the code generator, and allows some metadata to be associated with it.</p>
6810 <p>The first argument specifies the address of a stack object that contains the
6811 root pointer. The second pointer (which must be either a constant or a
6812 global value address) contains the meta-data to be associated with the
6816 <p>At runtime, a call to this intrinsic stores a null pointer into the "ptrloc"
6817 location. At compile-time, the code generator generates information to allow
6818 the runtime to find the pointer at GC safe points. The '<tt>llvm.gcroot</tt>'
6819 intrinsic may only be used in a function which <a href="#gc">specifies a GC
6824 <!-- _______________________________________________________________________ -->
6826 <a name="int_gcread">'<tt>llvm.gcread</tt>' Intrinsic</a>
6833 declare i8* @llvm.gcread(i8* %ObjPtr, i8** %Ptr)
6837 <p>The '<tt>llvm.gcread</tt>' intrinsic identifies reads of references from heap
6838 locations, allowing garbage collector implementations that require read
6842 <p>The second argument is the address to read from, which should be an address
6843 allocated from the garbage collector. The first object is a pointer to the
6844 start of the referenced object, if needed by the language runtime (otherwise
6848 <p>The '<tt>llvm.gcread</tt>' intrinsic has the same semantics as a load
6849 instruction, but may be replaced with substantially more complex code by the
6850 garbage collector runtime, as needed. The '<tt>llvm.gcread</tt>' intrinsic
6851 may only be used in a function which <a href="#gc">specifies a GC
6856 <!-- _______________________________________________________________________ -->
6858 <a name="int_gcwrite">'<tt>llvm.gcwrite</tt>' Intrinsic</a>
6865 declare void @llvm.gcwrite(i8* %P1, i8* %Obj, i8** %P2)
6869 <p>The '<tt>llvm.gcwrite</tt>' intrinsic identifies writes of references to heap
6870 locations, allowing garbage collector implementations that require write
6871 barriers (such as generational or reference counting collectors).</p>
6874 <p>The first argument is the reference to store, the second is the start of the
6875 object to store it to, and the third is the address of the field of Obj to
6876 store to. If the runtime does not require a pointer to the object, Obj may
6880 <p>The '<tt>llvm.gcwrite</tt>' intrinsic has the same semantics as a store
6881 instruction, but may be replaced with substantially more complex code by the
6882 garbage collector runtime, as needed. The '<tt>llvm.gcwrite</tt>' intrinsic
6883 may only be used in a function which <a href="#gc">specifies a GC
6890 <!-- ======================================================================= -->
6892 <a name="int_codegen">Code Generator Intrinsics</a>
6897 <p>These intrinsics are provided by LLVM to expose special features that may
6898 only be implemented with code generator support.</p>
6900 <!-- _______________________________________________________________________ -->
6902 <a name="int_returnaddress">'<tt>llvm.returnaddress</tt>' Intrinsic</a>
6909 declare i8 *@llvm.returnaddress(i32 <level>)
6913 <p>The '<tt>llvm.returnaddress</tt>' intrinsic attempts to compute a
6914 target-specific value indicating the return address of the current function
6915 or one of its callers.</p>
6918 <p>The argument to this intrinsic indicates which function to return the address
6919 for. Zero indicates the calling function, one indicates its caller, etc.
6920 The argument is <b>required</b> to be a constant integer value.</p>
6923 <p>The '<tt>llvm.returnaddress</tt>' intrinsic either returns a pointer
6924 indicating the return address of the specified call frame, or zero if it
6925 cannot be identified. The value returned by this intrinsic is likely to be
6926 incorrect or 0 for arguments other than zero, so it should only be used for
6927 debugging purposes.</p>
6929 <p>Note that calling this intrinsic does not prevent function inlining or other
6930 aggressive transformations, so the value returned may not be that of the
6931 obvious source-language caller.</p>
6935 <!-- _______________________________________________________________________ -->
6937 <a name="int_frameaddress">'<tt>llvm.frameaddress</tt>' Intrinsic</a>
6944 declare i8* @llvm.frameaddress(i32 <level>)
6948 <p>The '<tt>llvm.frameaddress</tt>' intrinsic attempts to return the
6949 target-specific frame pointer value for the specified stack frame.</p>
6952 <p>The argument to this intrinsic indicates which function to return the frame
6953 pointer for. Zero indicates the calling function, one indicates its caller,
6954 etc. The argument is <b>required</b> to be a constant integer value.</p>
6957 <p>The '<tt>llvm.frameaddress</tt>' intrinsic either returns a pointer
6958 indicating the frame address of the specified call frame, or zero if it
6959 cannot be identified. The value returned by this intrinsic is likely to be
6960 incorrect or 0 for arguments other than zero, so it should only be used for
6961 debugging purposes.</p>
6963 <p>Note that calling this intrinsic does not prevent function inlining or other
6964 aggressive transformations, so the value returned may not be that of the
6965 obvious source-language caller.</p>
6969 <!-- _______________________________________________________________________ -->
6971 <a name="int_stacksave">'<tt>llvm.stacksave</tt>' Intrinsic</a>
6978 declare i8* @llvm.stacksave()
6982 <p>The '<tt>llvm.stacksave</tt>' intrinsic is used to remember the current state
6983 of the function stack, for use
6984 with <a href="#int_stackrestore"> <tt>llvm.stackrestore</tt></a>. This is
6985 useful for implementing language features like scoped automatic variable
6986 sized arrays in C99.</p>
6989 <p>This intrinsic returns a opaque pointer value that can be passed
6990 to <a href="#int_stackrestore"><tt>llvm.stackrestore</tt></a>. When
6991 an <tt>llvm.stackrestore</tt> intrinsic is executed with a value saved
6992 from <tt>llvm.stacksave</tt>, it effectively restores the state of the stack
6993 to the state it was in when the <tt>llvm.stacksave</tt> intrinsic executed.
6994 In practice, this pops any <a href="#i_alloca">alloca</a> blocks from the
6995 stack that were allocated after the <tt>llvm.stacksave</tt> was executed.</p>
6999 <!-- _______________________________________________________________________ -->
7001 <a name="int_stackrestore">'<tt>llvm.stackrestore</tt>' Intrinsic</a>
7008 declare void @llvm.stackrestore(i8* %ptr)
7012 <p>The '<tt>llvm.stackrestore</tt>' intrinsic is used to restore the state of
7013 the function stack to the state it was in when the
7014 corresponding <a href="#int_stacksave"><tt>llvm.stacksave</tt></a> intrinsic
7015 executed. This is useful for implementing language features like scoped
7016 automatic variable sized arrays in C99.</p>
7019 <p>See the description
7020 for <a href="#int_stacksave"><tt>llvm.stacksave</tt></a>.</p>
7024 <!-- _______________________________________________________________________ -->
7026 <a name="int_prefetch">'<tt>llvm.prefetch</tt>' Intrinsic</a>
7033 declare void @llvm.prefetch(i8* <address>, i32 <rw>, i32 <locality>, i32 <cache type>)
7037 <p>The '<tt>llvm.prefetch</tt>' intrinsic is a hint to the code generator to
7038 insert a prefetch instruction if supported; otherwise, it is a noop.
7039 Prefetches have no effect on the behavior of the program but can change its
7040 performance characteristics.</p>
7043 <p><tt>address</tt> is the address to be prefetched, <tt>rw</tt> is the
7044 specifier determining if the fetch should be for a read (0) or write (1),
7045 and <tt>locality</tt> is a temporal locality specifier ranging from (0) - no
7046 locality, to (3) - extremely local keep in cache. The <tt>cache type</tt>
7047 specifies whether the prefetch is performed on the data (1) or instruction (0)
7048 cache. The <tt>rw</tt>, <tt>locality</tt> and <tt>cache type</tt> arguments
7049 must be constant integers.</p>
7052 <p>This intrinsic does not modify the behavior of the program. In particular,
7053 prefetches cannot trap and do not produce a value. On targets that support
7054 this intrinsic, the prefetch can provide hints to the processor cache for
7055 better performance.</p>
7059 <!-- _______________________________________________________________________ -->
7061 <a name="int_pcmarker">'<tt>llvm.pcmarker</tt>' Intrinsic</a>
7068 declare void @llvm.pcmarker(i32 <id>)
7072 <p>The '<tt>llvm.pcmarker</tt>' intrinsic is a method to export a Program
7073 Counter (PC) in a region of code to simulators and other tools. The method
7074 is target specific, but it is expected that the marker will use exported
7075 symbols to transmit the PC of the marker. The marker makes no guarantees
7076 that it will remain with any specific instruction after optimizations. It is
7077 possible that the presence of a marker will inhibit optimizations. The
7078 intended use is to be inserted after optimizations to allow correlations of
7079 simulation runs.</p>
7082 <p><tt>id</tt> is a numerical id identifying the marker.</p>
7085 <p>This intrinsic does not modify the behavior of the program. Backends that do
7086 not support this intrinsic may ignore it.</p>
7090 <!-- _______________________________________________________________________ -->
7092 <a name="int_readcyclecounter">'<tt>llvm.readcyclecounter</tt>' Intrinsic</a>
7099 declare i64 @llvm.readcyclecounter()
7103 <p>The '<tt>llvm.readcyclecounter</tt>' intrinsic provides access to the cycle
7104 counter register (or similar low latency, high accuracy clocks) on those
7105 targets that support it. On X86, it should map to RDTSC. On Alpha, it
7106 should map to RPCC. As the backing counters overflow quickly (on the order
7107 of 9 seconds on alpha), this should only be used for small timings.</p>
7110 <p>When directly supported, reading the cycle counter should not modify any
7111 memory. Implementations are allowed to either return a application specific
7112 value or a system wide value. On backends without support, this is lowered
7113 to a constant 0.</p>
7119 <!-- ======================================================================= -->
7121 <a name="int_libc">Standard C Library Intrinsics</a>
7126 <p>LLVM provides intrinsics for a few important standard C library functions.
7127 These intrinsics allow source-language front-ends to pass information about
7128 the alignment of the pointer arguments to the code generator, providing
7129 opportunity for more efficient code generation.</p>
7131 <!-- _______________________________________________________________________ -->
7133 <a name="int_memcpy">'<tt>llvm.memcpy</tt>' Intrinsic</a>
7139 <p>This is an overloaded intrinsic. You can use <tt>llvm.memcpy</tt> on any
7140 integer bit width and for different address spaces. Not all targets support
7141 all bit widths however.</p>
7144 declare void @llvm.memcpy.p0i8.p0i8.i32(i8* <dest>, i8* <src>,
7145 i32 <len>, i32 <align>, i1 <isvolatile>)
7146 declare void @llvm.memcpy.p0i8.p0i8.i64(i8* <dest>, i8* <src>,
7147 i64 <len>, i32 <align>, i1 <isvolatile>)
7151 <p>The '<tt>llvm.memcpy.*</tt>' intrinsics copy a block of memory from the
7152 source location to the destination location.</p>
7154 <p>Note that, unlike the standard libc function, the <tt>llvm.memcpy.*</tt>
7155 intrinsics do not return a value, takes extra alignment/isvolatile arguments
7156 and the pointers can be in specified address spaces.</p>
7160 <p>The first argument is a pointer to the destination, the second is a pointer
7161 to the source. The third argument is an integer argument specifying the
7162 number of bytes to copy, the fourth argument is the alignment of the
7163 source and destination locations, and the fifth is a boolean indicating a
7164 volatile access.</p>
7166 <p>If the call to this intrinsic has an alignment value that is not 0 or 1,
7167 then the caller guarantees that both the source and destination pointers are
7168 aligned to that boundary.</p>
7170 <p>If the <tt>isvolatile</tt> parameter is <tt>true</tt>, the
7171 <tt>llvm.memcpy</tt> call is a <a href="#volatile">volatile operation</a>.
7172 The detailed access behavior is not very cleanly specified and it is unwise
7173 to depend on it.</p>
7177 <p>The '<tt>llvm.memcpy.*</tt>' intrinsics copy a block of memory from the
7178 source location to the destination location, which are not allowed to
7179 overlap. It copies "len" bytes of memory over. If the argument is known to
7180 be aligned to some boundary, this can be specified as the fourth argument,
7181 otherwise it should be set to 0 or 1.</p>
7185 <!-- _______________________________________________________________________ -->
7187 <a name="int_memmove">'<tt>llvm.memmove</tt>' Intrinsic</a>
7193 <p>This is an overloaded intrinsic. You can use llvm.memmove on any integer bit
7194 width and for different address space. Not all targets support all bit
7198 declare void @llvm.memmove.p0i8.p0i8.i32(i8* <dest>, i8* <src>,
7199 i32 <len>, i32 <align>, i1 <isvolatile>)
7200 declare void @llvm.memmove.p0i8.p0i8.i64(i8* <dest>, i8* <src>,
7201 i64 <len>, i32 <align>, i1 <isvolatile>)
7205 <p>The '<tt>llvm.memmove.*</tt>' intrinsics move a block of memory from the
7206 source location to the destination location. It is similar to the
7207 '<tt>llvm.memcpy</tt>' intrinsic but allows the two memory locations to
7210 <p>Note that, unlike the standard libc function, the <tt>llvm.memmove.*</tt>
7211 intrinsics do not return a value, takes extra alignment/isvolatile arguments
7212 and the pointers can be in specified address spaces.</p>
7216 <p>The first argument is a pointer to the destination, the second is a pointer
7217 to the source. The third argument is an integer argument specifying the
7218 number of bytes to copy, the fourth argument is the alignment of the
7219 source and destination locations, and the fifth is a boolean indicating a
7220 volatile access.</p>
7222 <p>If the call to this intrinsic has an alignment value that is not 0 or 1,
7223 then the caller guarantees that the source and destination pointers are
7224 aligned to that boundary.</p>
7226 <p>If the <tt>isvolatile</tt> parameter is <tt>true</tt>, the
7227 <tt>llvm.memmove</tt> call is a <a href="#volatile">volatile operation</a>.
7228 The detailed access behavior is not very cleanly specified and it is unwise
7229 to depend on it.</p>
7233 <p>The '<tt>llvm.memmove.*</tt>' intrinsics copy a block of memory from the
7234 source location to the destination location, which may overlap. It copies
7235 "len" bytes of memory over. If the argument is known to be aligned to some
7236 boundary, this can be specified as the fourth argument, otherwise it should
7237 be set to 0 or 1.</p>
7241 <!-- _______________________________________________________________________ -->
7243 <a name="int_memset">'<tt>llvm.memset.*</tt>' Intrinsics</a>
7249 <p>This is an overloaded intrinsic. You can use llvm.memset on any integer bit
7250 width and for different address spaces. However, not all targets support all
7254 declare void @llvm.memset.p0i8.i32(i8* <dest>, i8 <val>,
7255 i32 <len>, i32 <align>, i1 <isvolatile>)
7256 declare void @llvm.memset.p0i8.i64(i8* <dest>, i8 <val>,
7257 i64 <len>, i32 <align>, i1 <isvolatile>)
7261 <p>The '<tt>llvm.memset.*</tt>' intrinsics fill a block of memory with a
7262 particular byte value.</p>
7264 <p>Note that, unlike the standard libc function, the <tt>llvm.memset</tt>
7265 intrinsic does not return a value and takes extra alignment/volatile
7266 arguments. Also, the destination can be in an arbitrary address space.</p>
7269 <p>The first argument is a pointer to the destination to fill, the second is the
7270 byte value with which to fill it, the third argument is an integer argument
7271 specifying the number of bytes to fill, and the fourth argument is the known
7272 alignment of the destination location.</p>
7274 <p>If the call to this intrinsic has an alignment value that is not 0 or 1,
7275 then the caller guarantees that the destination pointer is aligned to that
7278 <p>If the <tt>isvolatile</tt> parameter is <tt>true</tt>, the
7279 <tt>llvm.memset</tt> call is a <a href="#volatile">volatile operation</a>.
7280 The detailed access behavior is not very cleanly specified and it is unwise
7281 to depend on it.</p>
7284 <p>The '<tt>llvm.memset.*</tt>' intrinsics fill "len" bytes of memory starting
7285 at the destination location. If the argument is known to be aligned to some
7286 boundary, this can be specified as the fourth argument, otherwise it should
7287 be set to 0 or 1.</p>
7291 <!-- _______________________________________________________________________ -->
7293 <a name="int_sqrt">'<tt>llvm.sqrt.*</tt>' Intrinsic</a>
7299 <p>This is an overloaded intrinsic. You can use <tt>llvm.sqrt</tt> on any
7300 floating point or vector of floating point type. Not all targets support all
7304 declare float @llvm.sqrt.f32(float %Val)
7305 declare double @llvm.sqrt.f64(double %Val)
7306 declare x86_fp80 @llvm.sqrt.f80(x86_fp80 %Val)
7307 declare fp128 @llvm.sqrt.f128(fp128 %Val)
7308 declare ppc_fp128 @llvm.sqrt.ppcf128(ppc_fp128 %Val)
7312 <p>The '<tt>llvm.sqrt</tt>' intrinsics return the sqrt of the specified operand,
7313 returning the same value as the libm '<tt>sqrt</tt>' functions would.
7314 Unlike <tt>sqrt</tt> in libm, however, <tt>llvm.sqrt</tt> has undefined
7315 behavior for negative numbers other than -0.0 (which allows for better
7316 optimization, because there is no need to worry about errno being
7317 set). <tt>llvm.sqrt(-0.0)</tt> is defined to return -0.0 like IEEE sqrt.</p>
7320 <p>The argument and return value are floating point numbers of the same
7324 <p>This function returns the sqrt of the specified operand if it is a
7325 nonnegative floating point number.</p>
7329 <!-- _______________________________________________________________________ -->
7331 <a name="int_powi">'<tt>llvm.powi.*</tt>' Intrinsic</a>
7337 <p>This is an overloaded intrinsic. You can use <tt>llvm.powi</tt> on any
7338 floating point or vector of floating point type. Not all targets support all
7342 declare float @llvm.powi.f32(float %Val, i32 %power)
7343 declare double @llvm.powi.f64(double %Val, i32 %power)
7344 declare x86_fp80 @llvm.powi.f80(x86_fp80 %Val, i32 %power)
7345 declare fp128 @llvm.powi.f128(fp128 %Val, i32 %power)
7346 declare ppc_fp128 @llvm.powi.ppcf128(ppc_fp128 %Val, i32 %power)
7350 <p>The '<tt>llvm.powi.*</tt>' intrinsics return the first operand raised to the
7351 specified (positive or negative) power. The order of evaluation of
7352 multiplications is not defined. When a vector of floating point type is
7353 used, the second argument remains a scalar integer value.</p>
7356 <p>The second argument is an integer power, and the first is a value to raise to
7360 <p>This function returns the first value raised to the second power with an
7361 unspecified sequence of rounding operations.</p>
7365 <!-- _______________________________________________________________________ -->
7367 <a name="int_sin">'<tt>llvm.sin.*</tt>' Intrinsic</a>
7373 <p>This is an overloaded intrinsic. You can use <tt>llvm.sin</tt> on any
7374 floating point or vector of floating point type. Not all targets support all
7378 declare float @llvm.sin.f32(float %Val)
7379 declare double @llvm.sin.f64(double %Val)
7380 declare x86_fp80 @llvm.sin.f80(x86_fp80 %Val)
7381 declare fp128 @llvm.sin.f128(fp128 %Val)
7382 declare ppc_fp128 @llvm.sin.ppcf128(ppc_fp128 %Val)
7386 <p>The '<tt>llvm.sin.*</tt>' intrinsics return the sine of the operand.</p>
7389 <p>The argument and return value are floating point numbers of the same
7393 <p>This function returns the sine of the specified operand, returning the same
7394 values as the libm <tt>sin</tt> functions would, and handles error conditions
7395 in the same way.</p>
7399 <!-- _______________________________________________________________________ -->
7401 <a name="int_cos">'<tt>llvm.cos.*</tt>' Intrinsic</a>
7407 <p>This is an overloaded intrinsic. You can use <tt>llvm.cos</tt> on any
7408 floating point or vector of floating point type. Not all targets support all
7412 declare float @llvm.cos.f32(float %Val)
7413 declare double @llvm.cos.f64(double %Val)
7414 declare x86_fp80 @llvm.cos.f80(x86_fp80 %Val)
7415 declare fp128 @llvm.cos.f128(fp128 %Val)
7416 declare ppc_fp128 @llvm.cos.ppcf128(ppc_fp128 %Val)
7420 <p>The '<tt>llvm.cos.*</tt>' intrinsics return the cosine of the operand.</p>
7423 <p>The argument and return value are floating point numbers of the same
7427 <p>This function returns the cosine of the specified operand, returning the same
7428 values as the libm <tt>cos</tt> functions would, and handles error conditions
7429 in the same way.</p>
7433 <!-- _______________________________________________________________________ -->
7435 <a name="int_pow">'<tt>llvm.pow.*</tt>' Intrinsic</a>
7441 <p>This is an overloaded intrinsic. You can use <tt>llvm.pow</tt> on any
7442 floating point or vector of floating point type. Not all targets support all
7446 declare float @llvm.pow.f32(float %Val, float %Power)
7447 declare double @llvm.pow.f64(double %Val, double %Power)
7448 declare x86_fp80 @llvm.pow.f80(x86_fp80 %Val, x86_fp80 %Power)
7449 declare fp128 @llvm.pow.f128(fp128 %Val, fp128 %Power)
7450 declare ppc_fp128 @llvm.pow.ppcf128(ppc_fp128 %Val, ppc_fp128 Power)
7454 <p>The '<tt>llvm.pow.*</tt>' intrinsics return the first operand raised to the
7455 specified (positive or negative) power.</p>
7458 <p>The second argument is a floating point power, and the first is a value to
7459 raise to that power.</p>
7462 <p>This function returns the first value raised to the second power, returning
7463 the same values as the libm <tt>pow</tt> functions would, and handles error
7464 conditions in the same way.</p>
7468 <!-- _______________________________________________________________________ -->
7470 <a name="int_exp">'<tt>llvm.exp.*</tt>' Intrinsic</a>
7476 <p>This is an overloaded intrinsic. You can use <tt>llvm.exp</tt> on any
7477 floating point or vector of floating point type. Not all targets support all
7481 declare float @llvm.exp.f32(float %Val)
7482 declare double @llvm.exp.f64(double %Val)
7483 declare x86_fp80 @llvm.exp.f80(x86_fp80 %Val)
7484 declare fp128 @llvm.exp.f128(fp128 %Val)
7485 declare ppc_fp128 @llvm.exp.ppcf128(ppc_fp128 %Val)
7489 <p>The '<tt>llvm.exp.*</tt>' intrinsics perform the exp function.</p>
7492 <p>The argument and return value are floating point numbers of the same
7496 <p>This function returns the same values as the libm <tt>exp</tt> functions
7497 would, and handles error conditions in the same way.</p>
7501 <!-- _______________________________________________________________________ -->
7503 <a name="int_log">'<tt>llvm.log.*</tt>' Intrinsic</a>
7509 <p>This is an overloaded intrinsic. You can use <tt>llvm.log</tt> on any
7510 floating point or vector of floating point type. Not all targets support all
7514 declare float @llvm.log.f32(float %Val)
7515 declare double @llvm.log.f64(double %Val)
7516 declare x86_fp80 @llvm.log.f80(x86_fp80 %Val)
7517 declare fp128 @llvm.log.f128(fp128 %Val)
7518 declare ppc_fp128 @llvm.log.ppcf128(ppc_fp128 %Val)
7522 <p>The '<tt>llvm.log.*</tt>' intrinsics perform the log function.</p>
7525 <p>The argument and return value are floating point numbers of the same
7529 <p>This function returns the same values as the libm <tt>log</tt> functions
7530 would, and handles error conditions in the same way.</p>
7534 <!-- _______________________________________________________________________ -->
7536 <a name="int_fma">'<tt>llvm.fma.*</tt>' Intrinsic</a>
7542 <p>This is an overloaded intrinsic. You can use <tt>llvm.fma</tt> on any
7543 floating point or vector of floating point type. Not all targets support all
7547 declare float @llvm.fma.f32(float %a, float %b, float %c)
7548 declare double @llvm.fma.f64(double %a, double %b, double %c)
7549 declare x86_fp80 @llvm.fma.f80(x86_fp80 %a, x86_fp80 %b, x86_fp80 %c)
7550 declare fp128 @llvm.fma.f128(fp128 %a, fp128 %b, fp128 %c)
7551 declare ppc_fp128 @llvm.fma.ppcf128(ppc_fp128 %a, ppc_fp128 %b, ppc_fp128 %c)
7555 <p>The '<tt>llvm.fma.*</tt>' intrinsics perform the fused multiply-add
7559 <p>The argument and return value are floating point numbers of the same
7563 <p>This function returns the same values as the libm <tt>fma</tt> functions
7568 <!-- _______________________________________________________________________ -->
7570 <a name="int_fabs">'<tt>llvm.fabs.*</tt>' Intrinsic</a>
7576 <p>This is an overloaded intrinsic. You can use <tt>llvm.fabs</tt> on any
7577 floating point or vector of floating point type. Not all targets support all
7581 declare float @llvm.fabs.f32(float %Val)
7582 declare double @llvm.fabs.f64(double %Val)
7583 declare x86_fp80 @llvm.fabs.f80(x86_fp80 %Val)
7584 declare fp128 @llvm.fabs.f128(fp128 %Val)
7585 declare ppc_fp128 @llvm.fabs.ppcf128(ppc_fp128 %Val)
7589 <p>The '<tt>llvm.fabs.*</tt>' intrinsics return the absolute value of
7593 <p>The argument and return value are floating point numbers of the same
7597 <p>This function returns the same values as the libm <tt>fabs</tt> functions
7598 would, and handles error conditions in the same way.</p>
7602 <!-- _______________________________________________________________________ -->
7604 <a name="int_floor">'<tt>llvm.floor.*</tt>' Intrinsic</a>
7610 <p>This is an overloaded intrinsic. You can use <tt>llvm.floor</tt> on any
7611 floating point or vector of floating point type. Not all targets support all
7615 declare float @llvm.floor.f32(float %Val)
7616 declare double @llvm.floor.f64(double %Val)
7617 declare x86_fp80 @llvm.floor.f80(x86_fp80 %Val)
7618 declare fp128 @llvm.floor.f128(fp128 %Val)
7619 declare ppc_fp128 @llvm.floor.ppcf128(ppc_fp128 %Val)
7623 <p>The '<tt>llvm.floor.*</tt>' intrinsics return the floor of
7627 <p>The argument and return value are floating point numbers of the same
7631 <p>This function returns the same values as the libm <tt>floor</tt> functions
7632 would, and handles error conditions in the same way.</p>
7638 <!-- ======================================================================= -->
7640 <a name="int_manip">Bit Manipulation Intrinsics</a>
7645 <p>LLVM provides intrinsics for a few important bit manipulation operations.
7646 These allow efficient code generation for some algorithms.</p>
7648 <!-- _______________________________________________________________________ -->
7650 <a name="int_bswap">'<tt>llvm.bswap.*</tt>' Intrinsics</a>
7656 <p>This is an overloaded intrinsic function. You can use bswap on any integer
7657 type that is an even number of bytes (i.e. BitWidth % 16 == 0).</p>
7660 declare i16 @llvm.bswap.i16(i16 <id>)
7661 declare i32 @llvm.bswap.i32(i32 <id>)
7662 declare i64 @llvm.bswap.i64(i64 <id>)
7666 <p>The '<tt>llvm.bswap</tt>' family of intrinsics is used to byte swap integer
7667 values with an even number of bytes (positive multiple of 16 bits). These
7668 are useful for performing operations on data that is not in the target's
7669 native byte order.</p>
7672 <p>The <tt>llvm.bswap.i16</tt> intrinsic returns an i16 value that has the high
7673 and low byte of the input i16 swapped. Similarly,
7674 the <tt>llvm.bswap.i32</tt> intrinsic returns an i32 value that has the four
7675 bytes of the input i32 swapped, so that if the input bytes are numbered 0, 1,
7676 2, 3 then the returned i32 will have its bytes in 3, 2, 1, 0 order.
7677 The <tt>llvm.bswap.i48</tt>, <tt>llvm.bswap.i64</tt> and other intrinsics
7678 extend this concept to additional even-byte lengths (6 bytes, 8 bytes and
7679 more, respectively).</p>
7683 <!-- _______________________________________________________________________ -->
7685 <a name="int_ctpop">'<tt>llvm.ctpop.*</tt>' Intrinsic</a>
7691 <p>This is an overloaded intrinsic. You can use llvm.ctpop on any integer bit
7692 width, or on any vector with integer elements. Not all targets support all
7693 bit widths or vector types, however.</p>
7696 declare i8 @llvm.ctpop.i8(i8 <src>)
7697 declare i16 @llvm.ctpop.i16(i16 <src>)
7698 declare i32 @llvm.ctpop.i32(i32 <src>)
7699 declare i64 @llvm.ctpop.i64(i64 <src>)
7700 declare i256 @llvm.ctpop.i256(i256 <src>)
7701 declare <2 x i32> @llvm.ctpop.v2i32(<2 x i32> <src>)
7705 <p>The '<tt>llvm.ctpop</tt>' family of intrinsics counts the number of bits set
7709 <p>The only argument is the value to be counted. The argument may be of any
7710 integer type, or a vector with integer elements.
7711 The return type must match the argument type.</p>
7714 <p>The '<tt>llvm.ctpop</tt>' intrinsic counts the 1's in a variable, or within each
7715 element of a vector.</p>
7719 <!-- _______________________________________________________________________ -->
7721 <a name="int_ctlz">'<tt>llvm.ctlz.*</tt>' Intrinsic</a>
7727 <p>This is an overloaded intrinsic. You can use <tt>llvm.ctlz</tt> on any
7728 integer bit width, or any vector whose elements are integers. Not all
7729 targets support all bit widths or vector types, however.</p>
7732 declare i8 @llvm.ctlz.i8 (i8 <src>, i1 <is_zero_undef>)
7733 declare i16 @llvm.ctlz.i16 (i16 <src>, i1 <is_zero_undef>)
7734 declare i32 @llvm.ctlz.i32 (i32 <src>, i1 <is_zero_undef>)
7735 declare i64 @llvm.ctlz.i64 (i64 <src>, i1 <is_zero_undef>)
7736 declare i256 @llvm.ctlz.i256(i256 <src>, i1 <is_zero_undef>)
7737 declase <2 x i32> @llvm.ctlz.v2i32(<2 x i32> <src>, i1 <is_zero_undef>)
7741 <p>The '<tt>llvm.ctlz</tt>' family of intrinsic functions counts the number of
7742 leading zeros in a variable.</p>
7745 <p>The first argument is the value to be counted. This argument may be of any
7746 integer type, or a vectory with integer element type. The return type
7747 must match the first argument type.</p>
7749 <p>The second argument must be a constant and is a flag to indicate whether the
7750 intrinsic should ensure that a zero as the first argument produces a defined
7751 result. Historically some architectures did not provide a defined result for
7752 zero values as efficiently, and many algorithms are now predicated on
7753 avoiding zero-value inputs.</p>
7756 <p>The '<tt>llvm.ctlz</tt>' intrinsic counts the leading (most significant)
7757 zeros in a variable, or within each element of the vector.
7758 If <tt>src == 0</tt> then the result is the size in bits of the type of
7759 <tt>src</tt> if <tt>is_zero_undef == 0</tt> and <tt>undef</tt> otherwise.
7760 For example, <tt>llvm.ctlz(i32 2) = 30</tt>.</p>
7764 <!-- _______________________________________________________________________ -->
7766 <a name="int_cttz">'<tt>llvm.cttz.*</tt>' Intrinsic</a>
7772 <p>This is an overloaded intrinsic. You can use <tt>llvm.cttz</tt> on any
7773 integer bit width, or any vector of integer elements. Not all targets
7774 support all bit widths or vector types, however.</p>
7777 declare i8 @llvm.cttz.i8 (i8 <src>, i1 <is_zero_undef>)
7778 declare i16 @llvm.cttz.i16 (i16 <src>, i1 <is_zero_undef>)
7779 declare i32 @llvm.cttz.i32 (i32 <src>, i1 <is_zero_undef>)
7780 declare i64 @llvm.cttz.i64 (i64 <src>, i1 <is_zero_undef>)
7781 declare i256 @llvm.cttz.i256(i256 <src>, i1 <is_zero_undef>)
7782 declase <2 x i32> @llvm.cttz.v2i32(<2 x i32> <src>, i1 <is_zero_undef>)
7786 <p>The '<tt>llvm.cttz</tt>' family of intrinsic functions counts the number of
7790 <p>The first argument is the value to be counted. This argument may be of any
7791 integer type, or a vectory with integer element type. The return type
7792 must match the first argument type.</p>
7794 <p>The second argument must be a constant and is a flag to indicate whether the
7795 intrinsic should ensure that a zero as the first argument produces a defined
7796 result. Historically some architectures did not provide a defined result for
7797 zero values as efficiently, and many algorithms are now predicated on
7798 avoiding zero-value inputs.</p>
7801 <p>The '<tt>llvm.cttz</tt>' intrinsic counts the trailing (least significant)
7802 zeros in a variable, or within each element of a vector.
7803 If <tt>src == 0</tt> then the result is the size in bits of the type of
7804 <tt>src</tt> if <tt>is_zero_undef == 0</tt> and <tt>undef</tt> otherwise.
7805 For example, <tt>llvm.cttz(2) = 1</tt>.</p>
7811 <!-- ======================================================================= -->
7813 <a name="int_overflow">Arithmetic with Overflow Intrinsics</a>
7818 <p>LLVM provides intrinsics for some arithmetic with overflow operations.</p>
7820 <!-- _______________________________________________________________________ -->
7822 <a name="int_sadd_overflow">
7823 '<tt>llvm.sadd.with.overflow.*</tt>' Intrinsics
7830 <p>This is an overloaded intrinsic. You can use <tt>llvm.sadd.with.overflow</tt>
7831 on any integer bit width.</p>
7834 declare {i16, i1} @llvm.sadd.with.overflow.i16(i16 %a, i16 %b)
7835 declare {i32, i1} @llvm.sadd.with.overflow.i32(i32 %a, i32 %b)
7836 declare {i64, i1} @llvm.sadd.with.overflow.i64(i64 %a, i64 %b)
7840 <p>The '<tt>llvm.sadd.with.overflow</tt>' family of intrinsic functions perform
7841 a signed addition of the two arguments, and indicate whether an overflow
7842 occurred during the signed summation.</p>
7845 <p>The arguments (%a and %b) and the first element of the result structure may
7846 be of integer types of any bit width, but they must have the same bit
7847 width. The second element of the result structure must be of
7848 type <tt>i1</tt>. <tt>%a</tt> and <tt>%b</tt> are the two values that will
7849 undergo signed addition.</p>
7852 <p>The '<tt>llvm.sadd.with.overflow</tt>' family of intrinsic functions perform
7853 a signed addition of the two variables. They return a structure — the
7854 first element of which is the signed summation, and the second element of
7855 which is a bit specifying if the signed summation resulted in an
7860 %res = call {i32, i1} @llvm.sadd.with.overflow.i32(i32 %a, i32 %b)
7861 %sum = extractvalue {i32, i1} %res, 0
7862 %obit = extractvalue {i32, i1} %res, 1
7863 br i1 %obit, label %overflow, label %normal
7868 <!-- _______________________________________________________________________ -->
7870 <a name="int_uadd_overflow">
7871 '<tt>llvm.uadd.with.overflow.*</tt>' Intrinsics
7878 <p>This is an overloaded intrinsic. You can use <tt>llvm.uadd.with.overflow</tt>
7879 on any integer bit width.</p>
7882 declare {i16, i1} @llvm.uadd.with.overflow.i16(i16 %a, i16 %b)
7883 declare {i32, i1} @llvm.uadd.with.overflow.i32(i32 %a, i32 %b)
7884 declare {i64, i1} @llvm.uadd.with.overflow.i64(i64 %a, i64 %b)
7888 <p>The '<tt>llvm.uadd.with.overflow</tt>' family of intrinsic functions perform
7889 an unsigned addition of the two arguments, and indicate whether a carry
7890 occurred during the unsigned summation.</p>
7893 <p>The arguments (%a and %b) and the first element of the result structure may
7894 be of integer types of any bit width, but they must have the same bit
7895 width. The second element of the result structure must be of
7896 type <tt>i1</tt>. <tt>%a</tt> and <tt>%b</tt> are the two values that will
7897 undergo unsigned addition.</p>
7900 <p>The '<tt>llvm.uadd.with.overflow</tt>' family of intrinsic functions perform
7901 an unsigned addition of the two arguments. They return a structure —
7902 the first element of which is the sum, and the second element of which is a
7903 bit specifying if the unsigned summation resulted in a carry.</p>
7907 %res = call {i32, i1} @llvm.uadd.with.overflow.i32(i32 %a, i32 %b)
7908 %sum = extractvalue {i32, i1} %res, 0
7909 %obit = extractvalue {i32, i1} %res, 1
7910 br i1 %obit, label %carry, label %normal
7915 <!-- _______________________________________________________________________ -->
7917 <a name="int_ssub_overflow">
7918 '<tt>llvm.ssub.with.overflow.*</tt>' Intrinsics
7925 <p>This is an overloaded intrinsic. You can use <tt>llvm.ssub.with.overflow</tt>
7926 on any integer bit width.</p>
7929 declare {i16, i1} @llvm.ssub.with.overflow.i16(i16 %a, i16 %b)
7930 declare {i32, i1} @llvm.ssub.with.overflow.i32(i32 %a, i32 %b)
7931 declare {i64, i1} @llvm.ssub.with.overflow.i64(i64 %a, i64 %b)
7935 <p>The '<tt>llvm.ssub.with.overflow</tt>' family of intrinsic functions perform
7936 a signed subtraction of the two arguments, and indicate whether an overflow
7937 occurred during the signed subtraction.</p>
7940 <p>The arguments (%a and %b) and the first element of the result structure may
7941 be of integer types of any bit width, but they must have the same bit
7942 width. The second element of the result structure must be of
7943 type <tt>i1</tt>. <tt>%a</tt> and <tt>%b</tt> are the two values that will
7944 undergo signed subtraction.</p>
7947 <p>The '<tt>llvm.ssub.with.overflow</tt>' family of intrinsic functions perform
7948 a signed subtraction of the two arguments. They return a structure —
7949 the first element of which is the subtraction, and the second element of
7950 which is a bit specifying if the signed subtraction resulted in an
7955 %res = call {i32, i1} @llvm.ssub.with.overflow.i32(i32 %a, i32 %b)
7956 %sum = extractvalue {i32, i1} %res, 0
7957 %obit = extractvalue {i32, i1} %res, 1
7958 br i1 %obit, label %overflow, label %normal
7963 <!-- _______________________________________________________________________ -->
7965 <a name="int_usub_overflow">
7966 '<tt>llvm.usub.with.overflow.*</tt>' Intrinsics
7973 <p>This is an overloaded intrinsic. You can use <tt>llvm.usub.with.overflow</tt>
7974 on any integer bit width.</p>
7977 declare {i16, i1} @llvm.usub.with.overflow.i16(i16 %a, i16 %b)
7978 declare {i32, i1} @llvm.usub.with.overflow.i32(i32 %a, i32 %b)
7979 declare {i64, i1} @llvm.usub.with.overflow.i64(i64 %a, i64 %b)
7983 <p>The '<tt>llvm.usub.with.overflow</tt>' family of intrinsic functions perform
7984 an unsigned subtraction of the two arguments, and indicate whether an
7985 overflow occurred during the unsigned subtraction.</p>
7988 <p>The arguments (%a and %b) and the first element of the result structure may
7989 be of integer types of any bit width, but they must have the same bit
7990 width. The second element of the result structure must be of
7991 type <tt>i1</tt>. <tt>%a</tt> and <tt>%b</tt> are the two values that will
7992 undergo unsigned subtraction.</p>
7995 <p>The '<tt>llvm.usub.with.overflow</tt>' family of intrinsic functions perform
7996 an unsigned subtraction of the two arguments. They return a structure —
7997 the first element of which is the subtraction, and the second element of
7998 which is a bit specifying if the unsigned subtraction resulted in an
8003 %res = call {i32, i1} @llvm.usub.with.overflow.i32(i32 %a, i32 %b)
8004 %sum = extractvalue {i32, i1} %res, 0
8005 %obit = extractvalue {i32, i1} %res, 1
8006 br i1 %obit, label %overflow, label %normal
8011 <!-- _______________________________________________________________________ -->
8013 <a name="int_smul_overflow">
8014 '<tt>llvm.smul.with.overflow.*</tt>' Intrinsics
8021 <p>This is an overloaded intrinsic. You can use <tt>llvm.smul.with.overflow</tt>
8022 on any integer bit width.</p>
8025 declare {i16, i1} @llvm.smul.with.overflow.i16(i16 %a, i16 %b)
8026 declare {i32, i1} @llvm.smul.with.overflow.i32(i32 %a, i32 %b)
8027 declare {i64, i1} @llvm.smul.with.overflow.i64(i64 %a, i64 %b)
8032 <p>The '<tt>llvm.smul.with.overflow</tt>' family of intrinsic functions perform
8033 a signed multiplication of the two arguments, and indicate whether an
8034 overflow occurred during the signed multiplication.</p>
8037 <p>The arguments (%a and %b) and the first element of the result structure may
8038 be of integer types of any bit width, but they must have the same bit
8039 width. The second element of the result structure must be of
8040 type <tt>i1</tt>. <tt>%a</tt> and <tt>%b</tt> are the two values that will
8041 undergo signed multiplication.</p>
8044 <p>The '<tt>llvm.smul.with.overflow</tt>' family of intrinsic functions perform
8045 a signed multiplication of the two arguments. They return a structure —
8046 the first element of which is the multiplication, and the second element of
8047 which is a bit specifying if the signed multiplication resulted in an
8052 %res = call {i32, i1} @llvm.smul.with.overflow.i32(i32 %a, i32 %b)
8053 %sum = extractvalue {i32, i1} %res, 0
8054 %obit = extractvalue {i32, i1} %res, 1
8055 br i1 %obit, label %overflow, label %normal
8060 <!-- _______________________________________________________________________ -->
8062 <a name="int_umul_overflow">
8063 '<tt>llvm.umul.with.overflow.*</tt>' Intrinsics
8070 <p>This is an overloaded intrinsic. You can use <tt>llvm.umul.with.overflow</tt>
8071 on any integer bit width.</p>
8074 declare {i16, i1} @llvm.umul.with.overflow.i16(i16 %a, i16 %b)
8075 declare {i32, i1} @llvm.umul.with.overflow.i32(i32 %a, i32 %b)
8076 declare {i64, i1} @llvm.umul.with.overflow.i64(i64 %a, i64 %b)
8080 <p>The '<tt>llvm.umul.with.overflow</tt>' family of intrinsic functions perform
8081 a unsigned multiplication of the two arguments, and indicate whether an
8082 overflow occurred during the unsigned multiplication.</p>
8085 <p>The arguments (%a and %b) and the first element of the result structure may
8086 be of integer types of any bit width, but they must have the same bit
8087 width. The second element of the result structure must be of
8088 type <tt>i1</tt>. <tt>%a</tt> and <tt>%b</tt> are the two values that will
8089 undergo unsigned multiplication.</p>
8092 <p>The '<tt>llvm.umul.with.overflow</tt>' family of intrinsic functions perform
8093 an unsigned multiplication of the two arguments. They return a structure
8094 — the first element of which is the multiplication, and the second
8095 element of which is a bit specifying if the unsigned multiplication resulted
8100 %res = call {i32, i1} @llvm.umul.with.overflow.i32(i32 %a, i32 %b)
8101 %sum = extractvalue {i32, i1} %res, 0
8102 %obit = extractvalue {i32, i1} %res, 1
8103 br i1 %obit, label %overflow, label %normal
8110 <!-- ======================================================================= -->
8112 <a name="spec_arithmetic">Specialised Arithmetic Intrinsics</a>
8115 <!-- _______________________________________________________________________ -->
8118 <a name="fmuladd">'<tt>llvm.fmuladd.*</tt>' Intrinsic</a>
8125 declare float @llvm.fmuladd.f32(float %a, float %b, float %c)
8126 declare double @llvm.fmuladd.f64(double %a, double %b, double %c)
8130 <p>The '<tt>llvm.fmuladd.*</tt>' intrinsic functions represent multiply-add
8131 expressions that can be fused if the code generator determines that the fused
8132 expression would be legal and efficient.</p>
8135 <p>The '<tt>llvm.fmuladd.*</tt>' intrinsics each take three arguments: two
8136 multiplicands, a and b, and an addend c.</p>
8139 <p>The expression:</p>
8141 %0 = call float @llvm.fmuladd.f32(%a, %b, %c)
8143 <p>is equivalent to the expression a * b + c, except that rounding will not be
8144 performed between the multiplication and addition steps if the code generator
8145 fuses the operations. Fusion is not guaranteed, even if the target platform
8146 supports it. If a fused multiply-add is required the corresponding llvm.fma.*
8147 intrinsic function should be used instead.</p>
8151 %r2 = call float @llvm.fmuladd.f32(float %a, float %b, float %c) ; yields {float}:r2 = (a * b) + c
8156 <!-- ======================================================================= -->
8158 <a name="int_fp16">Half Precision Floating Point Intrinsics</a>
8163 <p>For most target platforms, half precision floating point is a storage-only
8164 format. This means that it is
8165 a dense encoding (in memory) but does not support computation in the
8168 <p>This means that code must first load the half-precision floating point
8169 value as an i16, then convert it to float with <a
8170 href="#int_convert_from_fp16"><tt>llvm.convert.from.fp16</tt></a>.
8171 Computation can then be performed on the float value (including extending to
8172 double etc). To store the value back to memory, it is first converted to
8173 float if needed, then converted to i16 with
8174 <a href="#int_convert_to_fp16"><tt>llvm.convert.to.fp16</tt></a>, then
8175 storing as an i16 value.</p>
8177 <!-- _______________________________________________________________________ -->
8179 <a name="int_convert_to_fp16">
8180 '<tt>llvm.convert.to.fp16</tt>' Intrinsic
8188 declare i16 @llvm.convert.to.fp16(f32 %a)
8192 <p>The '<tt>llvm.convert.to.fp16</tt>' intrinsic function performs
8193 a conversion from single precision floating point format to half precision
8194 floating point format.</p>
8197 <p>The intrinsic function contains single argument - the value to be
8201 <p>The '<tt>llvm.convert.to.fp16</tt>' intrinsic function performs
8202 a conversion from single precision floating point format to half precision
8203 floating point format. The return value is an <tt>i16</tt> which
8204 contains the converted number.</p>
8208 %res = call i16 @llvm.convert.to.fp16(f32 %a)
8209 store i16 %res, i16* @x, align 2
8214 <!-- _______________________________________________________________________ -->
8216 <a name="int_convert_from_fp16">
8217 '<tt>llvm.convert.from.fp16</tt>' Intrinsic
8225 declare f32 @llvm.convert.from.fp16(i16 %a)
8229 <p>The '<tt>llvm.convert.from.fp16</tt>' intrinsic function performs
8230 a conversion from half precision floating point format to single precision
8231 floating point format.</p>
8234 <p>The intrinsic function contains single argument - the value to be
8238 <p>The '<tt>llvm.convert.from.fp16</tt>' intrinsic function performs a
8239 conversion from half single precision floating point format to single
8240 precision floating point format. The input half-float value is represented by
8241 an <tt>i16</tt> value.</p>
8245 %a = load i16* @x, align 2
8246 %res = call f32 @llvm.convert.from.fp16(i16 %a)
8253 <!-- ======================================================================= -->
8255 <a name="int_debugger">Debugger Intrinsics</a>
8260 <p>The LLVM debugger intrinsics (which all start with <tt>llvm.dbg.</tt>
8261 prefix), are described in
8262 the <a href="SourceLevelDebugging.html#format_common_intrinsics">LLVM Source
8263 Level Debugging</a> document.</p>
8267 <!-- ======================================================================= -->
8269 <a name="int_eh">Exception Handling Intrinsics</a>
8274 <p>The LLVM exception handling intrinsics (which all start with
8275 <tt>llvm.eh.</tt> prefix), are described in
8276 the <a href="ExceptionHandling.html#format_common_intrinsics">LLVM Exception
8277 Handling</a> document.</p>
8281 <!-- ======================================================================= -->
8283 <a name="int_trampoline">Trampoline Intrinsics</a>
8288 <p>These intrinsics make it possible to excise one parameter, marked with
8289 the <a href="#nest"><tt>nest</tt></a> attribute, from a function.
8290 The result is a callable
8291 function pointer lacking the nest parameter - the caller does not need to
8292 provide a value for it. Instead, the value to use is stored in advance in a
8293 "trampoline", a block of memory usually allocated on the stack, which also
8294 contains code to splice the nest value into the argument list. This is used
8295 to implement the GCC nested function address extension.</p>
8297 <p>For example, if the function is
8298 <tt>i32 f(i8* nest %c, i32 %x, i32 %y)</tt> then the resulting function
8299 pointer has signature <tt>i32 (i32, i32)*</tt>. It can be created as
8302 <pre class="doc_code">
8303 %tramp = alloca [10 x i8], align 4 ; size and alignment only correct for X86
8304 %tramp1 = getelementptr [10 x i8]* %tramp, i32 0, i32 0
8305 call i8* @llvm.init.trampoline(i8* %tramp1, i8* bitcast (i32 (i8*, i32, i32)* @f to i8*), i8* %nval)
8306 %p = call i8* @llvm.adjust.trampoline(i8* %tramp1)
8307 %fp = bitcast i8* %p to i32 (i32, i32)*
8310 <p>The call <tt>%val = call i32 %fp(i32 %x, i32 %y)</tt> is then equivalent
8311 to <tt>%val = call i32 %f(i8* %nval, i32 %x, i32 %y)</tt>.</p>
8313 <!-- _______________________________________________________________________ -->
8316 '<tt>llvm.init.trampoline</tt>' Intrinsic
8324 declare void @llvm.init.trampoline(i8* <tramp>, i8* <func>, i8* <nval>)
8328 <p>This fills the memory pointed to by <tt>tramp</tt> with executable code,
8329 turning it into a trampoline.</p>
8332 <p>The <tt>llvm.init.trampoline</tt> intrinsic takes three arguments, all
8333 pointers. The <tt>tramp</tt> argument must point to a sufficiently large and
8334 sufficiently aligned block of memory; this memory is written to by the
8335 intrinsic. Note that the size and the alignment are target-specific - LLVM
8336 currently provides no portable way of determining them, so a front-end that
8337 generates this intrinsic needs to have some target-specific knowledge.
8338 The <tt>func</tt> argument must hold a function bitcast to
8339 an <tt>i8*</tt>.</p>
8342 <p>The block of memory pointed to by <tt>tramp</tt> is filled with target
8343 dependent code, turning it into a function. Then <tt>tramp</tt> needs to be
8344 passed to <a href="#int_at">llvm.adjust.trampoline</a> to get a pointer
8345 which can be <a href="#int_trampoline">bitcast (to a new function) and
8346 called</a>. The new function's signature is the same as that of
8347 <tt>func</tt> with any arguments marked with the <tt>nest</tt> attribute
8348 removed. At most one such <tt>nest</tt> argument is allowed, and it must be of
8349 pointer type. Calling the new function is equivalent to calling <tt>func</tt>
8350 with the same argument list, but with <tt>nval</tt> used for the missing
8351 <tt>nest</tt> argument. If, after calling <tt>llvm.init.trampoline</tt>, the
8352 memory pointed to by <tt>tramp</tt> is modified, then the effect of any later call
8353 to the returned function pointer is undefined.</p>
8356 <!-- _______________________________________________________________________ -->
8359 '<tt>llvm.adjust.trampoline</tt>' Intrinsic
8367 declare i8* @llvm.adjust.trampoline(i8* <tramp>)
8371 <p>This performs any required machine-specific adjustment to the address of a
8372 trampoline (passed as <tt>tramp</tt>).</p>
8375 <p><tt>tramp</tt> must point to a block of memory which already has trampoline code
8376 filled in by a previous call to <a href="#int_it"><tt>llvm.init.trampoline</tt>
8380 <p>On some architectures the address of the code to be executed needs to be
8381 different to the address where the trampoline is actually stored. This
8382 intrinsic returns the executable address corresponding to <tt>tramp</tt>
8383 after performing the required machine specific adjustments.
8384 The pointer returned can then be <a href="#int_trampoline"> bitcast and
8392 <!-- ======================================================================= -->
8394 <a name="int_memorymarkers">Memory Use Markers</a>
8399 <p>This class of intrinsics exists to information about the lifetime of memory
8400 objects and ranges where variables are immutable.</p>
8402 <!-- _______________________________________________________________________ -->
8404 <a name="int_lifetime_start">'<tt>llvm.lifetime.start</tt>' Intrinsic</a>
8411 declare void @llvm.lifetime.start(i64 <size>, i8* nocapture <ptr>)
8415 <p>The '<tt>llvm.lifetime.start</tt>' intrinsic specifies the start of a memory
8416 object's lifetime.</p>
8419 <p>The first argument is a constant integer representing the size of the
8420 object, or -1 if it is variable sized. The second argument is a pointer to
8424 <p>This intrinsic indicates that before this point in the code, the value of the
8425 memory pointed to by <tt>ptr</tt> is dead. This means that it is known to
8426 never be used and has an undefined value. A load from the pointer that
8427 precedes this intrinsic can be replaced with
8428 <tt>'<a href="#undefvalues">undef</a>'</tt>.</p>
8432 <!-- _______________________________________________________________________ -->
8434 <a name="int_lifetime_end">'<tt>llvm.lifetime.end</tt>' Intrinsic</a>
8441 declare void @llvm.lifetime.end(i64 <size>, i8* nocapture <ptr>)
8445 <p>The '<tt>llvm.lifetime.end</tt>' intrinsic specifies the end of a memory
8446 object's lifetime.</p>
8449 <p>The first argument is a constant integer representing the size of the
8450 object, or -1 if it is variable sized. The second argument is a pointer to
8454 <p>This intrinsic indicates that after this point in the code, the value of the
8455 memory pointed to by <tt>ptr</tt> is dead. This means that it is known to
8456 never be used and has an undefined value. Any stores into the memory object
8457 following this intrinsic may be removed as dead.
8461 <!-- _______________________________________________________________________ -->
8463 <a name="int_invariant_start">'<tt>llvm.invariant.start</tt>' Intrinsic</a>
8470 declare {}* @llvm.invariant.start(i64 <size>, i8* nocapture <ptr>)
8474 <p>The '<tt>llvm.invariant.start</tt>' intrinsic specifies that the contents of
8475 a memory object will not change.</p>
8478 <p>The first argument is a constant integer representing the size of the
8479 object, or -1 if it is variable sized. The second argument is a pointer to
8483 <p>This intrinsic indicates that until an <tt>llvm.invariant.end</tt> that uses
8484 the return value, the referenced memory location is constant and
8489 <!-- _______________________________________________________________________ -->
8491 <a name="int_invariant_end">'<tt>llvm.invariant.end</tt>' Intrinsic</a>
8498 declare void @llvm.invariant.end({}* <start>, i64 <size>, i8* nocapture <ptr>)
8502 <p>The '<tt>llvm.invariant.end</tt>' intrinsic specifies that the contents of
8503 a memory object are mutable.</p>
8506 <p>The first argument is the matching <tt>llvm.invariant.start</tt> intrinsic.
8507 The second argument is a constant integer representing the size of the
8508 object, or -1 if it is variable sized and the third argument is a pointer
8512 <p>This intrinsic indicates that the memory is mutable again.</p>
8518 <!-- ======================================================================= -->
8520 <a name="int_general">General Intrinsics</a>
8525 <p>This class of intrinsics is designed to be generic and has no specific
8528 <!-- _______________________________________________________________________ -->
8530 <a name="int_var_annotation">'<tt>llvm.var.annotation</tt>' Intrinsic</a>
8537 declare void @llvm.var.annotation(i8* <val>, i8* <str>, i8* <str>, i32 <int>)
8541 <p>The '<tt>llvm.var.annotation</tt>' intrinsic.</p>
8544 <p>The first argument is a pointer to a value, the second is a pointer to a
8545 global string, the third is a pointer to a global string which is the source
8546 file name, and the last argument is the line number.</p>
8549 <p>This intrinsic allows annotation of local variables with arbitrary strings.
8550 This can be useful for special purpose optimizations that want to look for
8551 these annotations. These have no other defined use; they are ignored by code
8552 generation and optimization.</p>
8556 <!-- _______________________________________________________________________ -->
8558 <a name="int_annotation">'<tt>llvm.annotation.*</tt>' Intrinsic</a>
8564 <p>This is an overloaded intrinsic. You can use '<tt>llvm.annotation</tt>' on
8565 any integer bit width.</p>
8568 declare i8 @llvm.annotation.i8(i8 <val>, i8* <str>, i8* <str>, i32 <int>)
8569 declare i16 @llvm.annotation.i16(i16 <val>, i8* <str>, i8* <str>, i32 <int>)
8570 declare i32 @llvm.annotation.i32(i32 <val>, i8* <str>, i8* <str>, i32 <int>)
8571 declare i64 @llvm.annotation.i64(i64 <val>, i8* <str>, i8* <str>, i32 <int>)
8572 declare i256 @llvm.annotation.i256(i256 <val>, i8* <str>, i8* <str>, i32 <int>)
8576 <p>The '<tt>llvm.annotation</tt>' intrinsic.</p>
8579 <p>The first argument is an integer value (result of some expression), the
8580 second is a pointer to a global string, the third is a pointer to a global
8581 string which is the source file name, and the last argument is the line
8582 number. It returns the value of the first argument.</p>
8585 <p>This intrinsic allows annotations to be put on arbitrary expressions with
8586 arbitrary strings. This can be useful for special purpose optimizations that
8587 want to look for these annotations. These have no other defined use; they
8588 are ignored by code generation and optimization.</p>
8592 <!-- _______________________________________________________________________ -->
8594 <a name="int_trap">'<tt>llvm.trap</tt>' Intrinsic</a>
8601 declare void @llvm.trap() noreturn nounwind
8605 <p>The '<tt>llvm.trap</tt>' intrinsic.</p>
8611 <p>This intrinsic is lowered to the target dependent trap instruction. If the
8612 target does not have a trap instruction, this intrinsic will be lowered to
8613 a call of the <tt>abort()</tt> function.</p>
8617 <!-- _______________________________________________________________________ -->
8619 <a name="int_debugtrap">'<tt>llvm.debugtrap</tt>' Intrinsic</a>
8626 declare void @llvm.debugtrap() nounwind
8630 <p>The '<tt>llvm.debugtrap</tt>' intrinsic.</p>
8636 <p>This intrinsic is lowered to code which is intended to cause an execution
8637 trap with the intention of requesting the attention of a debugger.</p>
8641 <!-- _______________________________________________________________________ -->
8643 <a name="int_stackprotector">'<tt>llvm.stackprotector</tt>' Intrinsic</a>
8650 declare void @llvm.stackprotector(i8* <guard>, i8** <slot>)
8654 <p>The <tt>llvm.stackprotector</tt> intrinsic takes the <tt>guard</tt> and
8655 stores it onto the stack at <tt>slot</tt>. The stack slot is adjusted to
8656 ensure that it is placed on the stack before local variables.</p>
8659 <p>The <tt>llvm.stackprotector</tt> intrinsic requires two pointer
8660 arguments. The first argument is the value loaded from the stack
8661 guard <tt>@__stack_chk_guard</tt>. The second variable is an <tt>alloca</tt>
8662 that has enough space to hold the value of the guard.</p>
8665 <p>This intrinsic causes the prologue/epilogue inserter to force the position of
8666 the <tt>AllocaInst</tt> stack slot to be before local variables on the
8667 stack. This is to ensure that if a local variable on the stack is
8668 overwritten, it will destroy the value of the guard. When the function exits,
8669 the guard on the stack is checked against the original guard. If they are
8670 different, then the program aborts by calling the <tt>__stack_chk_fail()</tt>
8675 <!-- _______________________________________________________________________ -->
8677 <a name="int_objectsize">'<tt>llvm.objectsize</tt>' Intrinsic</a>
8684 declare i32 @llvm.objectsize.i32(i8* <object>, i1 <min>)
8685 declare i64 @llvm.objectsize.i64(i8* <object>, i1 <min>)
8689 <p>The <tt>llvm.objectsize</tt> intrinsic is designed to provide information to
8690 the optimizers to determine at compile time whether a) an operation (like
8691 memcpy) will overflow a buffer that corresponds to an object, or b) that a
8692 runtime check for overflow isn't necessary. An object in this context means
8693 an allocation of a specific class, structure, array, or other object.</p>
8696 <p>The <tt>llvm.objectsize</tt> intrinsic takes two arguments. The first
8697 argument is a pointer to or into the <tt>object</tt>. The second argument
8698 is a boolean and determines whether <tt>llvm.objectsize</tt> returns 0 (if
8699 true) or -1 (if false) when the object size is unknown.
8700 The second argument only accepts constants.</p>
8703 <p>The <tt>llvm.objectsize</tt> intrinsic is lowered to a constant representing
8704 the size of the object concerned. If the size cannot be determined at compile
8705 time, <tt>llvm.objectsize</tt> returns <tt>i32/i64 -1 or 0</tt>
8706 (depending on the <tt>min</tt> argument).</p>
8709 <!-- _______________________________________________________________________ -->
8711 <a name="int_expect">'<tt>llvm.expect</tt>' Intrinsic</a>
8718 declare i32 @llvm.expect.i32(i32 <val>, i32 <expected_val>)
8719 declare i64 @llvm.expect.i64(i64 <val>, i64 <expected_val>)
8723 <p>The <tt>llvm.expect</tt> intrinsic provides information about expected (the
8724 most probable) value of <tt>val</tt>, which can be used by optimizers.</p>
8727 <p>The <tt>llvm.expect</tt> intrinsic takes two arguments. The first
8728 argument is a value. The second argument is an expected value, this needs to
8729 be a constant value, variables are not allowed.</p>
8732 <p>This intrinsic is lowered to the <tt>val</tt>.</p>
8735 <!-- _______________________________________________________________________ -->
8737 <a name="int_donothing">'<tt>llvm.donothing</tt>' Intrinsic</a>
8744 declare void @llvm.donothing() nounwind readnone
8748 <p>The <tt>llvm.donothing</tt> intrinsic doesn't perform any operation. It's the
8749 only intrinsic that can be called with an invoke instruction.</p>
8755 <p>This intrinsic does nothing, and it's removed by optimizers and ignored by
8762 <!-- *********************************************************************** -->
8765 <a href="http://jigsaw.w3.org/css-validator/check/referer"><img
8766 src="http://jigsaw.w3.org/css-validator/images/vcss-blue" alt="Valid CSS"></a>
8767 <a href="http://validator.w3.org/check/referer"><img
8768 src="http://www.w3.org/Icons/valid-html401-blue" alt="Valid HTML 4.01"></a>
8770 <a href="mailto:sabre@nondot.org">Chris Lattner</a><br>
8771 <a href="http://llvm.org/">The LLVM Compiler Infrastructure</a><br>
8772 Last modified: $Date$