1 <!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01//EN"
2 "http://www.w3.org/TR/html4/strict.dtd">
5 <title>LLVM Assembly Language Reference Manual</title>
6 <meta http-equiv="Content-Type" content="text/html; charset=utf-8">
7 <meta name="author" content="Chris Lattner">
8 <meta name="description"
9 content="LLVM Assembly Language Reference Manual.">
10 <link rel="stylesheet" href="llvm.css" type="text/css">
15 <h1>LLVM Language Reference Manual</h1>
17 <li><a href="#abstract">Abstract</a></li>
18 <li><a href="#introduction">Introduction</a></li>
19 <li><a href="#identifiers">Identifiers</a></li>
20 <li><a href="#highlevel">High Level Structure</a>
22 <li><a href="#modulestructure">Module Structure</a></li>
23 <li><a href="#linkage">Linkage Types</a>
25 <li><a href="#linkage_private">'<tt>private</tt>' Linkage</a></li>
26 <li><a href="#linkage_linker_private">'<tt>linker_private</tt>' Linkage</a></li>
27 <li><a href="#linkage_linker_private_weak">'<tt>linker_private_weak</tt>' Linkage</a></li>
28 <li><a href="#linkage_linker_private_weak_def_auto">'<tt>linker_private_weak_def_auto</tt>' Linkage</a></li>
29 <li><a href="#linkage_internal">'<tt>internal</tt>' Linkage</a></li>
30 <li><a href="#linkage_available_externally">'<tt>available_externally</tt>' Linkage</a></li>
31 <li><a href="#linkage_linkonce">'<tt>linkonce</tt>' Linkage</a></li>
32 <li><a href="#linkage_common">'<tt>common</tt>' Linkage</a></li>
33 <li><a href="#linkage_weak">'<tt>weak</tt>' Linkage</a></li>
34 <li><a href="#linkage_appending">'<tt>appending</tt>' Linkage</a></li>
35 <li><a href="#linkage_externweak">'<tt>extern_weak</tt>' Linkage</a></li>
36 <li><a href="#linkage_linkonce_odr">'<tt>linkonce_odr</tt>' Linkage</a></li>
37 <li><a href="#linkage_weak">'<tt>weak_odr</tt>' Linkage</a></li>
38 <li><a href="#linkage_external">'<tt>externally visible</tt>' Linkage</a></li>
39 <li><a href="#linkage_dllimport">'<tt>dllimport</tt>' Linkage</a></li>
40 <li><a href="#linkage_dllexport">'<tt>dllexport</tt>' Linkage</a></li>
43 <li><a href="#callingconv">Calling Conventions</a></li>
44 <li><a href="#namedtypes">Named Types</a></li>
45 <li><a href="#globalvars">Global Variables</a></li>
46 <li><a href="#functionstructure">Functions</a></li>
47 <li><a href="#aliasstructure">Aliases</a></li>
48 <li><a href="#namedmetadatastructure">Named Metadata</a></li>
49 <li><a href="#paramattrs">Parameter Attributes</a></li>
50 <li><a href="#fnattrs">Function Attributes</a></li>
51 <li><a href="#gc">Garbage Collector Names</a></li>
52 <li><a href="#moduleasm">Module-Level Inline Assembly</a></li>
53 <li><a href="#datalayout">Data Layout</a></li>
54 <li><a href="#pointeraliasing">Pointer Aliasing Rules</a></li>
55 <li><a href="#volatile">Volatile Memory Accesses</a></li>
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="#trapvalues">Trap Values</a></li>
96 <li><a href="#blockaddress">Addresses of Basic Blocks</a></li>
97 <li><a href="#constantexprs">Constant Expressions</a></li>
100 <li><a href="#othervalues">Other Values</a>
102 <li><a href="#inlineasm">Inline Assembler Expressions</a></li>
103 <li><a href="#metadata">Metadata Nodes and Metadata Strings</a></li>
106 <li><a href="#intrinsic_globals">Intrinsic Global Variables</a>
108 <li><a href="#intg_used">The '<tt>llvm.used</tt>' Global Variable</a></li>
109 <li><a href="#intg_compiler_used">The '<tt>llvm.compiler.used</tt>'
110 Global Variable</a></li>
111 <li><a href="#intg_global_ctors">The '<tt>llvm.global_ctors</tt>'
112 Global Variable</a></li>
113 <li><a href="#intg_global_dtors">The '<tt>llvm.global_dtors</tt>'
114 Global Variable</a></li>
117 <li><a href="#instref">Instruction Reference</a>
119 <li><a href="#terminators">Terminator Instructions</a>
121 <li><a href="#i_ret">'<tt>ret</tt>' Instruction</a></li>
122 <li><a href="#i_br">'<tt>br</tt>' Instruction</a></li>
123 <li><a href="#i_switch">'<tt>switch</tt>' Instruction</a></li>
124 <li><a href="#i_indirectbr">'<tt>indirectbr</tt>' Instruction</a></li>
125 <li><a href="#i_invoke">'<tt>invoke</tt>' Instruction</a></li>
126 <li><a href="#i_unwind">'<tt>unwind</tt>' Instruction</a></li>
127 <li><a href="#i_resume">'<tt>resume</tt>' Instruction</a></li>
128 <li><a href="#i_unreachable">'<tt>unreachable</tt>' Instruction</a></li>
131 <li><a href="#binaryops">Binary Operations</a>
133 <li><a href="#i_add">'<tt>add</tt>' Instruction</a></li>
134 <li><a href="#i_fadd">'<tt>fadd</tt>' Instruction</a></li>
135 <li><a href="#i_sub">'<tt>sub</tt>' Instruction</a></li>
136 <li><a href="#i_fsub">'<tt>fsub</tt>' Instruction</a></li>
137 <li><a href="#i_mul">'<tt>mul</tt>' Instruction</a></li>
138 <li><a href="#i_fmul">'<tt>fmul</tt>' Instruction</a></li>
139 <li><a href="#i_udiv">'<tt>udiv</tt>' Instruction</a></li>
140 <li><a href="#i_sdiv">'<tt>sdiv</tt>' Instruction</a></li>
141 <li><a href="#i_fdiv">'<tt>fdiv</tt>' Instruction</a></li>
142 <li><a href="#i_urem">'<tt>urem</tt>' Instruction</a></li>
143 <li><a href="#i_srem">'<tt>srem</tt>' Instruction</a></li>
144 <li><a href="#i_frem">'<tt>frem</tt>' Instruction</a></li>
147 <li><a href="#bitwiseops">Bitwise Binary Operations</a>
149 <li><a href="#i_shl">'<tt>shl</tt>' Instruction</a></li>
150 <li><a href="#i_lshr">'<tt>lshr</tt>' Instruction</a></li>
151 <li><a href="#i_ashr">'<tt>ashr</tt>' Instruction</a></li>
152 <li><a href="#i_and">'<tt>and</tt>' Instruction</a></li>
153 <li><a href="#i_or">'<tt>or</tt>' Instruction</a></li>
154 <li><a href="#i_xor">'<tt>xor</tt>' Instruction</a></li>
157 <li><a href="#vectorops">Vector Operations</a>
159 <li><a href="#i_extractelement">'<tt>extractelement</tt>' Instruction</a></li>
160 <li><a href="#i_insertelement">'<tt>insertelement</tt>' Instruction</a></li>
161 <li><a href="#i_shufflevector">'<tt>shufflevector</tt>' Instruction</a></li>
164 <li><a href="#aggregateops">Aggregate Operations</a>
166 <li><a href="#i_extractvalue">'<tt>extractvalue</tt>' Instruction</a></li>
167 <li><a href="#i_insertvalue">'<tt>insertvalue</tt>' Instruction</a></li>
170 <li><a href="#memoryops">Memory Access and Addressing Operations</a>
172 <li><a href="#i_alloca">'<tt>alloca</tt>' Instruction</a></li>
173 <li><a href="#i_load">'<tt>load</tt>' Instruction</a></li>
174 <li><a href="#i_store">'<tt>store</tt>' Instruction</a></li>
175 <li><a href="#i_fence">'<tt>fence</tt>' Instruction</a></li>
176 <li><a href="#i_cmpxchg">'<tt>cmpxchg</tt>' Instruction</a></li>
177 <li><a href="#i_atomicrmw">'<tt>atomicrmw</tt>' Instruction</a></li>
178 <li><a href="#i_getelementptr">'<tt>getelementptr</tt>' Instruction</a></li>
181 <li><a href="#convertops">Conversion Operations</a>
183 <li><a href="#i_trunc">'<tt>trunc .. to</tt>' Instruction</a></li>
184 <li><a href="#i_zext">'<tt>zext .. to</tt>' Instruction</a></li>
185 <li><a href="#i_sext">'<tt>sext .. to</tt>' Instruction</a></li>
186 <li><a href="#i_fptrunc">'<tt>fptrunc .. to</tt>' Instruction</a></li>
187 <li><a href="#i_fpext">'<tt>fpext .. to</tt>' Instruction</a></li>
188 <li><a href="#i_fptoui">'<tt>fptoui .. to</tt>' Instruction</a></li>
189 <li><a href="#i_fptosi">'<tt>fptosi .. to</tt>' Instruction</a></li>
190 <li><a href="#i_uitofp">'<tt>uitofp .. to</tt>' Instruction</a></li>
191 <li><a href="#i_sitofp">'<tt>sitofp .. to</tt>' Instruction</a></li>
192 <li><a href="#i_ptrtoint">'<tt>ptrtoint .. to</tt>' Instruction</a></li>
193 <li><a href="#i_inttoptr">'<tt>inttoptr .. to</tt>' Instruction</a></li>
194 <li><a href="#i_bitcast">'<tt>bitcast .. to</tt>' Instruction</a></li>
197 <li><a href="#otherops">Other Operations</a>
199 <li><a href="#i_icmp">'<tt>icmp</tt>' Instruction</a></li>
200 <li><a href="#i_fcmp">'<tt>fcmp</tt>' Instruction</a></li>
201 <li><a href="#i_phi">'<tt>phi</tt>' Instruction</a></li>
202 <li><a href="#i_select">'<tt>select</tt>' Instruction</a></li>
203 <li><a href="#i_call">'<tt>call</tt>' Instruction</a></li>
204 <li><a href="#i_va_arg">'<tt>va_arg</tt>' Instruction</a></li>
205 <li><a href="#i_landingpad">'<tt>landingpad</tt>' Instruction</a></li>
210 <li><a href="#intrinsics">Intrinsic Functions</a>
212 <li><a href="#int_varargs">Variable Argument Handling Intrinsics</a>
214 <li><a href="#int_va_start">'<tt>llvm.va_start</tt>' Intrinsic</a></li>
215 <li><a href="#int_va_end">'<tt>llvm.va_end</tt>' Intrinsic</a></li>
216 <li><a href="#int_va_copy">'<tt>llvm.va_copy</tt>' Intrinsic</a></li>
219 <li><a href="#int_gc">Accurate Garbage Collection Intrinsics</a>
221 <li><a href="#int_gcroot">'<tt>llvm.gcroot</tt>' Intrinsic</a></li>
222 <li><a href="#int_gcread">'<tt>llvm.gcread</tt>' Intrinsic</a></li>
223 <li><a href="#int_gcwrite">'<tt>llvm.gcwrite</tt>' Intrinsic</a></li>
226 <li><a href="#int_codegen">Code Generator Intrinsics</a>
228 <li><a href="#int_returnaddress">'<tt>llvm.returnaddress</tt>' Intrinsic</a></li>
229 <li><a href="#int_frameaddress">'<tt>llvm.frameaddress</tt>' Intrinsic</a></li>
230 <li><a href="#int_stacksave">'<tt>llvm.stacksave</tt>' Intrinsic</a></li>
231 <li><a href="#int_stackrestore">'<tt>llvm.stackrestore</tt>' Intrinsic</a></li>
232 <li><a href="#int_prefetch">'<tt>llvm.prefetch</tt>' Intrinsic</a></li>
233 <li><a href="#int_pcmarker">'<tt>llvm.pcmarker</tt>' Intrinsic</a></li>
234 <li><a href="#int_readcyclecounter">'<tt>llvm.readcyclecounter</tt>' Intrinsic</a></li>
237 <li><a href="#int_libc">Standard C Library Intrinsics</a>
239 <li><a href="#int_memcpy">'<tt>llvm.memcpy.*</tt>' Intrinsic</a></li>
240 <li><a href="#int_memmove">'<tt>llvm.memmove.*</tt>' Intrinsic</a></li>
241 <li><a href="#int_memset">'<tt>llvm.memset.*</tt>' Intrinsic</a></li>
242 <li><a href="#int_sqrt">'<tt>llvm.sqrt.*</tt>' Intrinsic</a></li>
243 <li><a href="#int_powi">'<tt>llvm.powi.*</tt>' Intrinsic</a></li>
244 <li><a href="#int_sin">'<tt>llvm.sin.*</tt>' Intrinsic</a></li>
245 <li><a href="#int_cos">'<tt>llvm.cos.*</tt>' Intrinsic</a></li>
246 <li><a href="#int_pow">'<tt>llvm.pow.*</tt>' Intrinsic</a></li>
247 <li><a href="#int_exp">'<tt>llvm.exp.*</tt>' Intrinsic</a></li>
248 <li><a href="#int_log">'<tt>llvm.log.*</tt>' Intrinsic</a></li>
249 <li><a href="#int_fma">'<tt>llvm.fma.*</tt>' Intrinsic</a></li>
252 <li><a href="#int_manip">Bit Manipulation Intrinsics</a>
254 <li><a href="#int_bswap">'<tt>llvm.bswap.*</tt>' Intrinsics</a></li>
255 <li><a href="#int_ctpop">'<tt>llvm.ctpop.*</tt>' Intrinsic </a></li>
256 <li><a href="#int_ctlz">'<tt>llvm.ctlz.*</tt>' Intrinsic </a></li>
257 <li><a href="#int_cttz">'<tt>llvm.cttz.*</tt>' Intrinsic </a></li>
260 <li><a href="#int_overflow">Arithmetic with Overflow Intrinsics</a>
262 <li><a href="#int_sadd_overflow">'<tt>llvm.sadd.with.overflow.*</tt> Intrinsics</a></li>
263 <li><a href="#int_uadd_overflow">'<tt>llvm.uadd.with.overflow.*</tt> Intrinsics</a></li>
264 <li><a href="#int_ssub_overflow">'<tt>llvm.ssub.with.overflow.*</tt> Intrinsics</a></li>
265 <li><a href="#int_usub_overflow">'<tt>llvm.usub.with.overflow.*</tt> Intrinsics</a></li>
266 <li><a href="#int_smul_overflow">'<tt>llvm.smul.with.overflow.*</tt> Intrinsics</a></li>
267 <li><a href="#int_umul_overflow">'<tt>llvm.umul.with.overflow.*</tt> Intrinsics</a></li>
270 <li><a href="#int_fp16">Half Precision Floating Point Intrinsics</a>
272 <li><a href="#int_convert_to_fp16">'<tt>llvm.convert.to.fp16</tt>' Intrinsic</a></li>
273 <li><a href="#int_convert_from_fp16">'<tt>llvm.convert.from.fp16</tt>' Intrinsic</a></li>
276 <li><a href="#int_debugger">Debugger intrinsics</a></li>
277 <li><a href="#int_eh">Exception Handling intrinsics</a></li>
278 <li><a href="#int_trampoline">Trampoline Intrinsics</a>
280 <li><a href="#int_it">'<tt>llvm.init.trampoline</tt>' Intrinsic</a></li>
281 <li><a href="#int_at">'<tt>llvm.adjust.trampoline</tt>' Intrinsic</a></li>
284 <li><a href="#int_atomics">Atomic intrinsics</a>
286 <li><a href="#int_memory_barrier"><tt>llvm.memory_barrier</tt></a></li>
287 <li><a href="#int_atomic_cmp_swap"><tt>llvm.atomic.cmp.swap</tt></a></li>
288 <li><a href="#int_atomic_swap"><tt>llvm.atomic.swap</tt></a></li>
289 <li><a href="#int_atomic_load_add"><tt>llvm.atomic.load.add</tt></a></li>
290 <li><a href="#int_atomic_load_sub"><tt>llvm.atomic.load.sub</tt></a></li>
291 <li><a href="#int_atomic_load_and"><tt>llvm.atomic.load.and</tt></a></li>
292 <li><a href="#int_atomic_load_nand"><tt>llvm.atomic.load.nand</tt></a></li>
293 <li><a href="#int_atomic_load_or"><tt>llvm.atomic.load.or</tt></a></li>
294 <li><a href="#int_atomic_load_xor"><tt>llvm.atomic.load.xor</tt></a></li>
295 <li><a href="#int_atomic_load_max"><tt>llvm.atomic.load.max</tt></a></li>
296 <li><a href="#int_atomic_load_min"><tt>llvm.atomic.load.min</tt></a></li>
297 <li><a href="#int_atomic_load_umax"><tt>llvm.atomic.load.umax</tt></a></li>
298 <li><a href="#int_atomic_load_umin"><tt>llvm.atomic.load.umin</tt></a></li>
301 <li><a href="#int_memorymarkers">Memory Use Markers</a>
303 <li><a href="#int_lifetime_start"><tt>llvm.lifetime.start</tt></a></li>
304 <li><a href="#int_lifetime_end"><tt>llvm.lifetime.end</tt></a></li>
305 <li><a href="#int_invariant_start"><tt>llvm.invariant.start</tt></a></li>
306 <li><a href="#int_invariant_end"><tt>llvm.invariant.end</tt></a></li>
309 <li><a href="#int_general">General intrinsics</a>
311 <li><a href="#int_var_annotation">
312 '<tt>llvm.var.annotation</tt>' Intrinsic</a></li>
313 <li><a href="#int_annotation">
314 '<tt>llvm.annotation.*</tt>' Intrinsic</a></li>
315 <li><a href="#int_trap">
316 '<tt>llvm.trap</tt>' Intrinsic</a></li>
317 <li><a href="#int_stackprotector">
318 '<tt>llvm.stackprotector</tt>' Intrinsic</a></li>
319 <li><a href="#int_objectsize">
320 '<tt>llvm.objectsize</tt>' Intrinsic</a></li>
327 <div class="doc_author">
328 <p>Written by <a href="mailto:sabre@nondot.org">Chris Lattner</a>
329 and <a href="mailto:vadve@cs.uiuc.edu">Vikram Adve</a></p>
332 <!-- *********************************************************************** -->
333 <h2><a name="abstract">Abstract</a></h2>
334 <!-- *********************************************************************** -->
338 <p>This document is a reference manual for the LLVM assembly language. LLVM is
339 a Static Single Assignment (SSA) based representation that provides type
340 safety, low-level operations, flexibility, and the capability of representing
341 'all' high-level languages cleanly. It is the common code representation
342 used throughout all phases of the LLVM compilation strategy.</p>
346 <!-- *********************************************************************** -->
347 <h2><a name="introduction">Introduction</a></h2>
348 <!-- *********************************************************************** -->
352 <p>The LLVM code representation is designed to be used in three different forms:
353 as an in-memory compiler IR, as an on-disk bitcode representation (suitable
354 for fast loading by a Just-In-Time compiler), and as a human readable
355 assembly language representation. This allows LLVM to provide a powerful
356 intermediate representation for efficient compiler transformations and
357 analysis, while providing a natural means to debug and visualize the
358 transformations. The three different forms of LLVM are all equivalent. This
359 document describes the human readable representation and notation.</p>
361 <p>The LLVM representation aims to be light-weight and low-level while being
362 expressive, typed, and extensible at the same time. It aims to be a
363 "universal IR" of sorts, by being at a low enough level that high-level ideas
364 may be cleanly mapped to it (similar to how microprocessors are "universal
365 IR's", allowing many source languages to be mapped to them). By providing
366 type information, LLVM can be used as the target of optimizations: for
367 example, through pointer analysis, it can be proven that a C automatic
368 variable is never accessed outside of the current function, allowing it to
369 be promoted to a simple SSA value instead of a memory location.</p>
371 <!-- _______________________________________________________________________ -->
373 <a name="wellformed">Well-Formedness</a>
378 <p>It is important to note that this document describes 'well formed' LLVM
379 assembly language. There is a difference between what the parser accepts and
380 what is considered 'well formed'. For example, the following instruction is
381 syntactically okay, but not well formed:</p>
383 <pre class="doc_code">
384 %x = <a href="#i_add">add</a> i32 1, %x
387 <p>because the definition of <tt>%x</tt> does not dominate all of its uses. The
388 LLVM infrastructure provides a verification pass that may be used to verify
389 that an LLVM module is well formed. This pass is automatically run by the
390 parser after parsing input assembly and by the optimizer before it outputs
391 bitcode. The violations pointed out by the verifier pass indicate bugs in
392 transformation passes or input to the parser.</p>
398 <!-- Describe the typesetting conventions here. -->
400 <!-- *********************************************************************** -->
401 <h2><a name="identifiers">Identifiers</a></h2>
402 <!-- *********************************************************************** -->
406 <p>LLVM identifiers come in two basic types: global and local. Global
407 identifiers (functions, global variables) begin with the <tt>'@'</tt>
408 character. Local identifiers (register names, types) begin with
409 the <tt>'%'</tt> character. Additionally, there are three different formats
410 for identifiers, for different purposes:</p>
413 <li>Named values are represented as a string of characters with their prefix.
414 For example, <tt>%foo</tt>, <tt>@DivisionByZero</tt>,
415 <tt>%a.really.long.identifier</tt>. The actual regular expression used is
416 '<tt>[%@][a-zA-Z$._][a-zA-Z$._0-9]*</tt>'. Identifiers which require
417 other characters in their names can be surrounded with quotes. Special
418 characters may be escaped using <tt>"\xx"</tt> where <tt>xx</tt> is the
419 ASCII code for the character in hexadecimal. In this way, any character
420 can be used in a name value, even quotes themselves.</li>
422 <li>Unnamed values are represented as an unsigned numeric value with their
423 prefix. For example, <tt>%12</tt>, <tt>@2</tt>, <tt>%44</tt>.</li>
425 <li>Constants, which are described in a <a href="#constants">section about
426 constants</a>, below.</li>
429 <p>LLVM requires that values start with a prefix for two reasons: Compilers
430 don't need to worry about name clashes with reserved words, and the set of
431 reserved words may be expanded in the future without penalty. Additionally,
432 unnamed identifiers allow a compiler to quickly come up with a temporary
433 variable without having to avoid symbol table conflicts.</p>
435 <p>Reserved words in LLVM are very similar to reserved words in other
436 languages. There are keywords for different opcodes
437 ('<tt><a href="#i_add">add</a></tt>',
438 '<tt><a href="#i_bitcast">bitcast</a></tt>',
439 '<tt><a href="#i_ret">ret</a></tt>', etc...), for primitive type names
440 ('<tt><a href="#t_void">void</a></tt>',
441 '<tt><a href="#t_primitive">i32</a></tt>', etc...), and others. These
442 reserved words cannot conflict with variable names, because none of them
443 start with a prefix character (<tt>'%'</tt> or <tt>'@'</tt>).</p>
445 <p>Here is an example of LLVM code to multiply the integer variable
446 '<tt>%X</tt>' by 8:</p>
450 <pre class="doc_code">
451 %result = <a href="#i_mul">mul</a> i32 %X, 8
454 <p>After strength reduction:</p>
456 <pre class="doc_code">
457 %result = <a href="#i_shl">shl</a> i32 %X, i8 3
460 <p>And the hard way:</p>
462 <pre class="doc_code">
463 %0 = <a href="#i_add">add</a> i32 %X, %X <i>; yields {i32}:%0</i>
464 %1 = <a href="#i_add">add</a> i32 %0, %0 <i>; yields {i32}:%1</i>
465 %result = <a href="#i_add">add</a> i32 %1, %1
468 <p>This last way of multiplying <tt>%X</tt> by 8 illustrates several important
469 lexical features of LLVM:</p>
472 <li>Comments are delimited with a '<tt>;</tt>' and go until the end of
475 <li>Unnamed temporaries are created when the result of a computation is not
476 assigned to a named value.</li>
478 <li>Unnamed temporaries are numbered sequentially</li>
481 <p>It also shows a convention that we follow in this document. When
482 demonstrating instructions, we will follow an instruction with a comment that
483 defines the type and name of value produced. Comments are shown in italic
488 <!-- *********************************************************************** -->
489 <h2><a name="highlevel">High Level Structure</a></h2>
490 <!-- *********************************************************************** -->
492 <!-- ======================================================================= -->
494 <a name="modulestructure">Module Structure</a>
499 <p>LLVM programs are composed of "Module"s, each of which is a translation unit
500 of the input programs. Each module consists of functions, global variables,
501 and symbol table entries. Modules may be combined together with the LLVM
502 linker, which merges function (and global variable) definitions, resolves
503 forward declarations, and merges symbol table entries. Here is an example of
504 the "hello world" module:</p>
506 <pre class="doc_code">
507 <i>; Declare the string constant as a global constant.</i>
508 <a href="#identifiers">@.LC0</a> = <a href="#linkage_internal">internal</a> <a href="#globalvars">constant</a> <a href="#t_array">[13 x i8]</a> c"hello world\0A\00" <i>; [13 x i8]*</i>
510 <i>; External declaration of the puts function</i>
511 <a href="#functionstructure">declare</a> i32 @puts(i8*) <i>; i32 (i8*)* </i>
513 <i>; Definition of main function</i>
514 define i32 @main() { <i>; i32()* </i>
515 <i>; Convert [13 x i8]* to i8 *...</i>
516 %cast210 = <a href="#i_getelementptr">getelementptr</a> [13 x i8]* @.LC0, i64 0, i64 0 <i>; i8*</i>
518 <i>; Call puts function to write out the string to stdout.</i>
519 <a href="#i_call">call</a> i32 @puts(i8* %cast210) <i>; i32</i>
520 <a href="#i_ret">ret</a> i32 0
523 <i>; Named metadata</i>
524 !1 = metadata !{i32 41}
528 <p>This example is made up of a <a href="#globalvars">global variable</a> named
529 "<tt>.LC0</tt>", an external declaration of the "<tt>puts</tt>" function,
530 a <a href="#functionstructure">function definition</a> for
531 "<tt>main</tt>" and <a href="#namedmetadatastructure">named metadata</a>
534 <p>In general, a module is made up of a list of global values, where both
535 functions and global variables are global values. Global values are
536 represented by a pointer to a memory location (in this case, a pointer to an
537 array of char, and a pointer to a function), and have one of the
538 following <a href="#linkage">linkage types</a>.</p>
542 <!-- ======================================================================= -->
544 <a name="linkage">Linkage Types</a>
549 <p>All Global Variables and Functions have one of the following types of
553 <dt><tt><b><a name="linkage_private">private</a></b></tt></dt>
554 <dd>Global values with "<tt>private</tt>" linkage are only directly accessible
555 by objects in the current module. In particular, linking code into a
556 module with an private global value may cause the private to be renamed as
557 necessary to avoid collisions. Because the symbol is private to the
558 module, all references can be updated. This doesn't show up in any symbol
559 table in the object file.</dd>
561 <dt><tt><b><a name="linkage_linker_private">linker_private</a></b></tt></dt>
562 <dd>Similar to <tt>private</tt>, but the symbol is passed through the
563 assembler and evaluated by the linker. Unlike normal strong symbols, they
564 are removed by the linker from the final linked image (executable or
565 dynamic library).</dd>
567 <dt><tt><b><a name="linkage_linker_private_weak">linker_private_weak</a></b></tt></dt>
568 <dd>Similar to "<tt>linker_private</tt>", but the symbol is weak. Note that
569 <tt>linker_private_weak</tt> symbols are subject to coalescing by the
570 linker. The symbols are removed by the linker from the final linked image
571 (executable or dynamic library).</dd>
573 <dt><tt><b><a name="linkage_linker_private_weak_def_auto">linker_private_weak_def_auto</a></b></tt></dt>
574 <dd>Similar to "<tt>linker_private_weak</tt>", but it's known that the address
575 of the object is not taken. For instance, functions that had an inline
576 definition, but the compiler decided not to inline it. Note,
577 unlike <tt>linker_private</tt> and <tt>linker_private_weak</tt>,
578 <tt>linker_private_weak_def_auto</tt> may have only <tt>default</tt>
579 visibility. The symbols are removed by the linker from the final linked
580 image (executable or dynamic library).</dd>
582 <dt><tt><b><a name="linkage_internal">internal</a></b></tt></dt>
583 <dd>Similar to private, but the value shows as a local symbol
584 (<tt>STB_LOCAL</tt> in the case of ELF) in the object file. This
585 corresponds to the notion of the '<tt>static</tt>' keyword in C.</dd>
587 <dt><tt><b><a name="linkage_available_externally">available_externally</a></b></tt></dt>
588 <dd>Globals with "<tt>available_externally</tt>" linkage are never emitted
589 into the object file corresponding to the LLVM module. They exist to
590 allow inlining and other optimizations to take place given knowledge of
591 the definition of the global, which is known to be somewhere outside the
592 module. Globals with <tt>available_externally</tt> linkage are allowed to
593 be discarded at will, and are otherwise the same as <tt>linkonce_odr</tt>.
594 This linkage type is only allowed on definitions, not declarations.</dd>
596 <dt><tt><b><a name="linkage_linkonce">linkonce</a></b></tt></dt>
597 <dd>Globals with "<tt>linkonce</tt>" linkage are merged with other globals of
598 the same name when linkage occurs. This can be used to implement
599 some forms of inline functions, templates, or other code which must be
600 generated in each translation unit that uses it, but where the body may
601 be overridden with a more definitive definition later. Unreferenced
602 <tt>linkonce</tt> globals are allowed to be discarded. Note that
603 <tt>linkonce</tt> linkage does not actually allow the optimizer to
604 inline the body of this function into callers because it doesn't know if
605 this definition of the function is the definitive definition within the
606 program or whether it will be overridden by a stronger definition.
607 To enable inlining and other optimizations, use "<tt>linkonce_odr</tt>"
610 <dt><tt><b><a name="linkage_weak">weak</a></b></tt></dt>
611 <dd>"<tt>weak</tt>" linkage has the same merging semantics as
612 <tt>linkonce</tt> linkage, except that unreferenced globals with
613 <tt>weak</tt> linkage may not be discarded. This is used for globals that
614 are declared "weak" in C source code.</dd>
616 <dt><tt><b><a name="linkage_common">common</a></b></tt></dt>
617 <dd>"<tt>common</tt>" linkage is most similar to "<tt>weak</tt>" linkage, but
618 they are used for tentative definitions in C, such as "<tt>int X;</tt>" at
620 Symbols with "<tt>common</tt>" linkage are merged in the same way as
621 <tt>weak symbols</tt>, and they may not be deleted if unreferenced.
622 <tt>common</tt> symbols may not have an explicit section,
623 must have a zero initializer, and may not be marked '<a
624 href="#globalvars"><tt>constant</tt></a>'. Functions and aliases may not
625 have common linkage.</dd>
628 <dt><tt><b><a name="linkage_appending">appending</a></b></tt></dt>
629 <dd>"<tt>appending</tt>" linkage may only be applied to global variables of
630 pointer to array type. When two global variables with appending linkage
631 are linked together, the two global arrays are appended together. This is
632 the LLVM, typesafe, equivalent of having the system linker append together
633 "sections" with identical names when .o files are linked.</dd>
635 <dt><tt><b><a name="linkage_externweak">extern_weak</a></b></tt></dt>
636 <dd>The semantics of this linkage follow the ELF object file model: the symbol
637 is weak until linked, if not linked, the symbol becomes null instead of
638 being an undefined reference.</dd>
640 <dt><tt><b><a name="linkage_linkonce_odr">linkonce_odr</a></b></tt></dt>
641 <dt><tt><b><a name="linkage_weak_odr">weak_odr</a></b></tt></dt>
642 <dd>Some languages allow differing globals to be merged, such as two functions
643 with different semantics. Other languages, such as <tt>C++</tt>, ensure
644 that only equivalent globals are ever merged (the "one definition rule"
645 — "ODR"). Such languages can use the <tt>linkonce_odr</tt>
646 and <tt>weak_odr</tt> linkage types to indicate that the global will only
647 be merged with equivalent globals. These linkage types are otherwise the
648 same as their non-<tt>odr</tt> versions.</dd>
650 <dt><tt><b><a name="linkage_external">externally visible</a></b></tt>:</dt>
651 <dd>If none of the above identifiers are used, the global is externally
652 visible, meaning that it participates in linkage and can be used to
653 resolve external symbol references.</dd>
656 <p>The next two types of linkage are targeted for Microsoft Windows platform
657 only. They are designed to support importing (exporting) symbols from (to)
658 DLLs (Dynamic Link Libraries).</p>
661 <dt><tt><b><a name="linkage_dllimport">dllimport</a></b></tt></dt>
662 <dd>"<tt>dllimport</tt>" linkage causes the compiler to reference a function
663 or variable via a global pointer to a pointer that is set up by the DLL
664 exporting the symbol. On Microsoft Windows targets, the pointer name is
665 formed by combining <code>__imp_</code> and the function or variable
668 <dt><tt><b><a name="linkage_dllexport">dllexport</a></b></tt></dt>
669 <dd>"<tt>dllexport</tt>" linkage causes the compiler to provide a global
670 pointer to a pointer in a DLL, so that it can be referenced with the
671 <tt>dllimport</tt> attribute. On Microsoft Windows targets, the pointer
672 name is formed by combining <code>__imp_</code> and the function or
676 <p>For example, since the "<tt>.LC0</tt>" variable is defined to be internal, if
677 another module defined a "<tt>.LC0</tt>" variable and was linked with this
678 one, one of the two would be renamed, preventing a collision. Since
679 "<tt>main</tt>" and "<tt>puts</tt>" are external (i.e., lacking any linkage
680 declarations), they are accessible outside of the current module.</p>
682 <p>It is illegal for a function <i>declaration</i> to have any linkage type
683 other than "externally visible", <tt>dllimport</tt>
684 or <tt>extern_weak</tt>.</p>
686 <p>Aliases can have only <tt>external</tt>, <tt>internal</tt>, <tt>weak</tt>
687 or <tt>weak_odr</tt> linkages.</p>
691 <!-- ======================================================================= -->
693 <a name="callingconv">Calling Conventions</a>
698 <p>LLVM <a href="#functionstructure">functions</a>, <a href="#i_call">calls</a>
699 and <a href="#i_invoke">invokes</a> can all have an optional calling
700 convention specified for the call. The calling convention of any pair of
701 dynamic caller/callee must match, or the behavior of the program is
702 undefined. The following calling conventions are supported by LLVM, and more
703 may be added in the future:</p>
706 <dt><b>"<tt>ccc</tt>" - The C calling convention</b>:</dt>
707 <dd>This calling convention (the default if no other calling convention is
708 specified) matches the target C calling conventions. This calling
709 convention supports varargs function calls and tolerates some mismatch in
710 the declared prototype and implemented declaration of the function (as
713 <dt><b>"<tt>fastcc</tt>" - The fast calling convention</b>:</dt>
714 <dd>This calling convention attempts to make calls as fast as possible
715 (e.g. by passing things in registers). This calling convention allows the
716 target to use whatever tricks it wants to produce fast code for the
717 target, without having to conform to an externally specified ABI
718 (Application Binary Interface).
719 <a href="CodeGenerator.html#tailcallopt">Tail calls can only be optimized
720 when this or the GHC convention is used.</a> This calling convention
721 does not support varargs and requires the prototype of all callees to
722 exactly match the prototype of the function definition.</dd>
724 <dt><b>"<tt>coldcc</tt>" - The cold calling convention</b>:</dt>
725 <dd>This calling convention attempts to make code in the caller as efficient
726 as possible under the assumption that the call is not commonly executed.
727 As such, these calls often preserve all registers so that the call does
728 not break any live ranges in the caller side. This calling convention
729 does not support varargs and requires the prototype of all callees to
730 exactly match the prototype of the function definition.</dd>
732 <dt><b>"<tt>cc <em>10</em></tt>" - GHC convention</b>:</dt>
733 <dd>This calling convention has been implemented specifically for use by the
734 <a href="http://www.haskell.org/ghc">Glasgow Haskell Compiler (GHC)</a>.
735 It passes everything in registers, going to extremes to achieve this by
736 disabling callee save registers. This calling convention should not be
737 used lightly but only for specific situations such as an alternative to
738 the <em>register pinning</em> performance technique often used when
739 implementing functional programming languages.At the moment only X86
740 supports this convention and it has the following limitations:
742 <li>On <em>X86-32</em> only supports up to 4 bit type parameters. No
743 floating point types are supported.</li>
744 <li>On <em>X86-64</em> only supports up to 10 bit type parameters and
745 6 floating point parameters.</li>
747 This calling convention supports
748 <a href="CodeGenerator.html#tailcallopt">tail call optimization</a> but
749 requires both the caller and callee are using it.
752 <dt><b>"<tt>cc <<em>n</em>></tt>" - Numbered convention</b>:</dt>
753 <dd>Any calling convention may be specified by number, allowing
754 target-specific calling conventions to be used. Target specific calling
755 conventions start at 64.</dd>
758 <p>More calling conventions can be added/defined on an as-needed basis, to
759 support Pascal conventions or any other well-known target-independent
764 <!-- ======================================================================= -->
766 <a name="visibility">Visibility Styles</a>
771 <p>All Global Variables and Functions have one of the following visibility
775 <dt><b>"<tt>default</tt>" - Default style</b>:</dt>
776 <dd>On targets that use the ELF object file format, default visibility means
777 that the declaration is visible to other modules and, in shared libraries,
778 means that the declared entity may be overridden. On Darwin, default
779 visibility means that the declaration is visible to other modules. Default
780 visibility corresponds to "external linkage" in the language.</dd>
782 <dt><b>"<tt>hidden</tt>" - Hidden style</b>:</dt>
783 <dd>Two declarations of an object with hidden visibility refer to the same
784 object if they are in the same shared object. Usually, hidden visibility
785 indicates that the symbol will not be placed into the dynamic symbol
786 table, so no other module (executable or shared library) can reference it
789 <dt><b>"<tt>protected</tt>" - Protected style</b>:</dt>
790 <dd>On ELF, protected visibility indicates that the symbol will be placed in
791 the dynamic symbol table, but that references within the defining module
792 will bind to the local symbol. That is, the symbol cannot be overridden by
798 <!-- ======================================================================= -->
800 <a name="namedtypes">Named Types</a>
805 <p>LLVM IR allows you to specify name aliases for certain types. This can make
806 it easier to read the IR and make the IR more condensed (particularly when
807 recursive types are involved). An example of a name specification is:</p>
809 <pre class="doc_code">
810 %mytype = type { %mytype*, i32 }
813 <p>You may give a name to any <a href="#typesystem">type</a> except
814 "<a href="#t_void">void</a>". Type name aliases may be used anywhere a type
815 is expected with the syntax "%mytype".</p>
817 <p>Note that type names are aliases for the structural type that they indicate,
818 and that you can therefore specify multiple names for the same type. This
819 often leads to confusing behavior when dumping out a .ll file. Since LLVM IR
820 uses structural typing, the name is not part of the type. When printing out
821 LLVM IR, the printer will pick <em>one name</em> to render all types of a
822 particular shape. This means that if you have code where two different
823 source types end up having the same LLVM type, that the dumper will sometimes
824 print the "wrong" or unexpected type. This is an important design point and
825 isn't going to change.</p>
829 <!-- ======================================================================= -->
831 <a name="globalvars">Global Variables</a>
836 <p>Global variables define regions of memory allocated at compilation time
837 instead of run-time. Global variables may optionally be initialized, may
838 have an explicit section to be placed in, and may have an optional explicit
839 alignment specified. A variable may be defined as "thread_local", which
840 means that it will not be shared by threads (each thread will have a
841 separated copy of the variable). A variable may be defined as a global
842 "constant," which indicates that the contents of the variable
843 will <b>never</b> be modified (enabling better optimization, allowing the
844 global data to be placed in the read-only section of an executable, etc).
845 Note that variables that need runtime initialization cannot be marked
846 "constant" as there is a store to the variable.</p>
848 <p>LLVM explicitly allows <em>declarations</em> of global variables to be marked
849 constant, even if the final definition of the global is not. This capability
850 can be used to enable slightly better optimization of the program, but
851 requires the language definition to guarantee that optimizations based on the
852 'constantness' are valid for the translation units that do not include the
855 <p>As SSA values, global variables define pointer values that are in scope
856 (i.e. they dominate) all basic blocks in the program. Global variables
857 always define a pointer to their "content" type because they describe a
858 region of memory, and all memory objects in LLVM are accessed through
861 <p>Global variables can be marked with <tt>unnamed_addr</tt> which indicates
862 that the address is not significant, only the content. Constants marked
863 like this can be merged with other constants if they have the same
864 initializer. Note that a constant with significant address <em>can</em>
865 be merged with a <tt>unnamed_addr</tt> constant, the result being a
866 constant whose address is significant.</p>
868 <p>A global variable may be declared to reside in a target-specific numbered
869 address space. For targets that support them, address spaces may affect how
870 optimizations are performed and/or what target instructions are used to
871 access the variable. The default address space is zero. The address space
872 qualifier must precede any other attributes.</p>
874 <p>LLVM allows an explicit section to be specified for globals. If the target
875 supports it, it will emit globals to the section specified.</p>
877 <p>An explicit alignment may be specified for a global, which must be a power
878 of 2. If not present, or if the alignment is set to zero, the alignment of
879 the global is set by the target to whatever it feels convenient. If an
880 explicit alignment is specified, the global is forced to have exactly that
881 alignment. Targets and optimizers are not allowed to over-align the global
882 if the global has an assigned section. In this case, the extra alignment
883 could be observable: for example, code could assume that the globals are
884 densely packed in their section and try to iterate over them as an array,
885 alignment padding would break this iteration.</p>
887 <p>For example, the following defines a global in a numbered address space with
888 an initializer, section, and alignment:</p>
890 <pre class="doc_code">
891 @G = addrspace(5) constant float 1.0, section "foo", align 4
897 <!-- ======================================================================= -->
899 <a name="functionstructure">Functions</a>
904 <p>LLVM function definitions consist of the "<tt>define</tt>" keyword, an
905 optional <a href="#linkage">linkage type</a>, an optional
906 <a href="#visibility">visibility style</a>, an optional
907 <a href="#callingconv">calling convention</a>,
908 an optional <tt>unnamed_addr</tt> attribute, a return type, an optional
909 <a href="#paramattrs">parameter attribute</a> for the return type, a function
910 name, a (possibly empty) argument list (each with optional
911 <a href="#paramattrs">parameter attributes</a>), optional
912 <a href="#fnattrs">function attributes</a>, an optional section, an optional
913 alignment, an optional <a href="#gc">garbage collector name</a>, an opening
914 curly brace, a list of basic blocks, and a closing curly brace.</p>
916 <p>LLVM function declarations consist of the "<tt>declare</tt>" keyword, an
917 optional <a href="#linkage">linkage type</a>, an optional
918 <a href="#visibility">visibility style</a>, an optional
919 <a href="#callingconv">calling convention</a>,
920 an optional <tt>unnamed_addr</tt> attribute, a return type, an optional
921 <a href="#paramattrs">parameter attribute</a> for the return type, a function
922 name, a possibly empty list of arguments, an optional alignment, and an
923 optional <a href="#gc">garbage collector name</a>.</p>
925 <p>A function definition contains a list of basic blocks, forming the CFG
926 (Control Flow Graph) for the function. Each basic block may optionally start
927 with a label (giving the basic block a symbol table entry), contains a list
928 of instructions, and ends with a <a href="#terminators">terminator</a>
929 instruction (such as a branch or function return).</p>
931 <p>The first basic block in a function is special in two ways: it is immediately
932 executed on entrance to the function, and it is not allowed to have
933 predecessor basic blocks (i.e. there can not be any branches to the entry
934 block of a function). Because the block can have no predecessors, it also
935 cannot have any <a href="#i_phi">PHI nodes</a>.</p>
937 <p>LLVM allows an explicit section to be specified for functions. If the target
938 supports it, it will emit functions to the section specified.</p>
940 <p>An explicit alignment may be specified for a function. If not present, or if
941 the alignment is set to zero, the alignment of the function is set by the
942 target to whatever it feels convenient. If an explicit alignment is
943 specified, the function is forced to have at least that much alignment. All
944 alignments must be a power of 2.</p>
946 <p>If the <tt>unnamed_addr</tt> attribute is given, the address is know to not
947 be significant and two identical functions can be merged</p>.
950 <pre class="doc_code">
951 define [<a href="#linkage">linkage</a>] [<a href="#visibility">visibility</a>]
952 [<a href="#callingconv">cconv</a>] [<a href="#paramattrs">ret attrs</a>]
953 <ResultType> @<FunctionName> ([argument list])
954 [<a href="#fnattrs">fn Attrs</a>] [section "name"] [align N]
955 [<a href="#gc">gc</a>] { ... }
960 <!-- ======================================================================= -->
962 <a name="aliasstructure">Aliases</a>
967 <p>Aliases act as "second name" for the aliasee value (which can be either
968 function, global variable, another alias or bitcast of global value). Aliases
969 may have an optional <a href="#linkage">linkage type</a>, and an
970 optional <a href="#visibility">visibility style</a>.</p>
973 <pre class="doc_code">
974 @<Name> = alias [Linkage] [Visibility] <AliaseeTy> @<Aliasee>
979 <!-- ======================================================================= -->
981 <a name="namedmetadatastructure">Named Metadata</a>
986 <p>Named metadata is a collection of metadata. <a href="#metadata">Metadata
987 nodes</a> (but not metadata strings) are the only valid operands for
988 a named metadata.</p>
991 <pre class="doc_code">
992 ; Some unnamed metadata nodes, which are referenced by the named metadata.
993 !0 = metadata !{metadata !"zero"}
994 !1 = metadata !{metadata !"one"}
995 !2 = metadata !{metadata !"two"}
997 !name = !{!0, !1, !2}
1002 <!-- ======================================================================= -->
1004 <a name="paramattrs">Parameter Attributes</a>
1009 <p>The return type and each parameter of a function type may have a set of
1010 <i>parameter attributes</i> associated with them. Parameter attributes are
1011 used to communicate additional information about the result or parameters of
1012 a function. Parameter attributes are considered to be part of the function,
1013 not of the function type, so functions with different parameter attributes
1014 can have the same function type.</p>
1016 <p>Parameter attributes are simple keywords that follow the type specified. If
1017 multiple parameter attributes are needed, they are space separated. For
1020 <pre class="doc_code">
1021 declare i32 @printf(i8* noalias nocapture, ...)
1022 declare i32 @atoi(i8 zeroext)
1023 declare signext i8 @returns_signed_char()
1026 <p>Note that any attributes for the function result (<tt>nounwind</tt>,
1027 <tt>readonly</tt>) come immediately after the argument list.</p>
1029 <p>Currently, only the following parameter attributes are defined:</p>
1032 <dt><tt><b>zeroext</b></tt></dt>
1033 <dd>This indicates to the code generator that the parameter or return value
1034 should be zero-extended to the extent required by the target's ABI (which
1035 is usually 32-bits, but is 8-bits for a i1 on x86-64) by the caller (for a
1036 parameter) or the callee (for a return value).</dd>
1038 <dt><tt><b>signext</b></tt></dt>
1039 <dd>This indicates to the code generator that the parameter or return value
1040 should be sign-extended to the extent required by the target's ABI (which
1041 is usually 32-bits) by the caller (for a parameter) or the callee (for a
1044 <dt><tt><b>inreg</b></tt></dt>
1045 <dd>This indicates that this parameter or return value should be treated in a
1046 special target-dependent fashion during while emitting code for a function
1047 call or return (usually, by putting it in a register as opposed to memory,
1048 though some targets use it to distinguish between two different kinds of
1049 registers). Use of this attribute is target-specific.</dd>
1051 <dt><tt><b><a name="byval">byval</a></b></tt></dt>
1052 <dd><p>This indicates that the pointer parameter should really be passed by
1053 value to the function. The attribute implies that a hidden copy of the
1055 is made between the caller and the callee, so the callee is unable to
1056 modify the value in the callee. This attribute is only valid on LLVM
1057 pointer arguments. It is generally used to pass structs and arrays by
1058 value, but is also valid on pointers to scalars. The copy is considered
1059 to belong to the caller not the callee (for example,
1060 <tt><a href="#readonly">readonly</a></tt> functions should not write to
1061 <tt>byval</tt> parameters). This is not a valid attribute for return
1064 <p>The byval attribute also supports specifying an alignment with
1065 the align attribute. It indicates the alignment of the stack slot to
1066 form and the known alignment of the pointer specified to the call site. If
1067 the alignment is not specified, then the code generator makes a
1068 target-specific assumption.</p></dd>
1070 <dt><tt><b><a name="sret">sret</a></b></tt></dt>
1071 <dd>This indicates that the pointer parameter specifies the address of a
1072 structure that is the return value of the function in the source program.
1073 This pointer must be guaranteed by the caller to be valid: loads and
1074 stores to the structure may be assumed by the callee to not to trap. This
1075 may only be applied to the first parameter. This is not a valid attribute
1076 for return values. </dd>
1078 <dt><tt><b><a name="noalias">noalias</a></b></tt></dt>
1079 <dd>This indicates that pointer values
1080 <a href="#pointeraliasing"><i>based</i></a> on the argument or return
1081 value do not alias pointer values which are not <i>based</i> on it,
1082 ignoring certain "irrelevant" dependencies.
1083 For a call to the parent function, dependencies between memory
1084 references from before or after the call and from those during the call
1085 are "irrelevant" to the <tt>noalias</tt> keyword for the arguments and
1086 return value used in that call.
1087 The caller shares the responsibility with the callee for ensuring that
1088 these requirements are met.
1089 For further details, please see the discussion of the NoAlias response in
1090 <a href="AliasAnalysis.html#MustMayNo">alias analysis</a>.<br>
1092 Note that this definition of <tt>noalias</tt> is intentionally
1093 similar to the definition of <tt>restrict</tt> in C99 for function
1094 arguments, though it is slightly weaker.
1096 For function return values, C99's <tt>restrict</tt> is not meaningful,
1097 while LLVM's <tt>noalias</tt> is.
1100 <dt><tt><b><a name="nocapture">nocapture</a></b></tt></dt>
1101 <dd>This indicates that the callee does not make any copies of the pointer
1102 that outlive the callee itself. This is not a valid attribute for return
1105 <dt><tt><b><a name="nest">nest</a></b></tt></dt>
1106 <dd>This indicates that the pointer parameter can be excised using the
1107 <a href="#int_trampoline">trampoline intrinsics</a>. This is not a valid
1108 attribute for return values.</dd>
1113 <!-- ======================================================================= -->
1115 <a name="gc">Garbage Collector Names</a>
1120 <p>Each function may specify a garbage collector name, which is simply a
1123 <pre class="doc_code">
1124 define void @f() gc "name" { ... }
1127 <p>The compiler declares the supported values of <i>name</i>. Specifying a
1128 collector which will cause the compiler to alter its output in order to
1129 support the named garbage collection algorithm.</p>
1133 <!-- ======================================================================= -->
1135 <a name="fnattrs">Function Attributes</a>
1140 <p>Function attributes are set to communicate additional information about a
1141 function. Function attributes are considered to be part of the function, not
1142 of the function type, so functions with different parameter attributes can
1143 have the same function type.</p>
1145 <p>Function attributes are simple keywords that follow the type specified. If
1146 multiple attributes are needed, they are space separated. For example:</p>
1148 <pre class="doc_code">
1149 define void @f() noinline { ... }
1150 define void @f() alwaysinline { ... }
1151 define void @f() alwaysinline optsize { ... }
1152 define void @f() optsize { ... }
1156 <dt><tt><b>alignstack(<<em>n</em>>)</b></tt></dt>
1157 <dd>This attribute indicates that, when emitting the prologue and epilogue,
1158 the backend should forcibly align the stack pointer. Specify the
1159 desired alignment, which must be a power of two, in parentheses.
1161 <dt><tt><b>alwaysinline</b></tt></dt>
1162 <dd>This attribute indicates that the inliner should attempt to inline this
1163 function into callers whenever possible, ignoring any active inlining size
1164 threshold for this caller.</dd>
1166 <dt><tt><b>nonlazybind</b></tt></dt>
1167 <dd>This attribute suppresses lazy symbol binding for the function. This
1168 may make calls to the function faster, at the cost of extra program
1169 startup time if the function is not called during program startup.</dd>
1171 <dt><tt><b>inlinehint</b></tt></dt>
1172 <dd>This attribute indicates that the source code contained a hint that inlining
1173 this function is desirable (such as the "inline" keyword in C/C++). It
1174 is just a hint; it imposes no requirements on the inliner.</dd>
1176 <dt><tt><b>naked</b></tt></dt>
1177 <dd>This attribute disables prologue / epilogue emission for the function.
1178 This can have very system-specific consequences.</dd>
1180 <dt><tt><b>noimplicitfloat</b></tt></dt>
1181 <dd>This attributes disables implicit floating point instructions.</dd>
1183 <dt><tt><b>noinline</b></tt></dt>
1184 <dd>This attribute indicates that the inliner should never inline this
1185 function in any situation. This attribute may not be used together with
1186 the <tt>alwaysinline</tt> attribute.</dd>
1188 <dt><tt><b>noredzone</b></tt></dt>
1189 <dd>This attribute indicates that the code generator should not use a red
1190 zone, even if the target-specific ABI normally permits it.</dd>
1192 <dt><tt><b>noreturn</b></tt></dt>
1193 <dd>This function attribute indicates that the function never returns
1194 normally. This produces undefined behavior at runtime if the function
1195 ever does dynamically return.</dd>
1197 <dt><tt><b>nounwind</b></tt></dt>
1198 <dd>This function attribute indicates that the function never returns with an
1199 unwind or exceptional control flow. If the function does unwind, its
1200 runtime behavior is undefined.</dd>
1202 <dt><tt><b>optsize</b></tt></dt>
1203 <dd>This attribute suggests that optimization passes and code generator passes
1204 make choices that keep the code size of this function low, and otherwise
1205 do optimizations specifically to reduce code size.</dd>
1207 <dt><tt><b>readnone</b></tt></dt>
1208 <dd>This attribute indicates that the function computes its result (or decides
1209 to unwind an exception) based strictly on its arguments, without
1210 dereferencing any pointer arguments or otherwise accessing any mutable
1211 state (e.g. memory, control registers, etc) visible to caller functions.
1212 It does not write through any pointer arguments
1213 (including <tt><a href="#byval">byval</a></tt> arguments) and never
1214 changes any state visible to callers. This means that it cannot unwind
1215 exceptions by calling the <tt>C++</tt> exception throwing methods, but
1216 could use the <tt>unwind</tt> instruction.</dd>
1218 <dt><tt><b><a name="readonly">readonly</a></b></tt></dt>
1219 <dd>This attribute indicates that the function does not write through any
1220 pointer arguments (including <tt><a href="#byval">byval</a></tt>
1221 arguments) or otherwise modify any state (e.g. memory, control registers,
1222 etc) visible to caller functions. It may dereference pointer arguments
1223 and read state that may be set in the caller. A readonly function always
1224 returns the same value (or unwinds an exception identically) when called
1225 with the same set of arguments and global state. It cannot unwind an
1226 exception by calling the <tt>C++</tt> exception throwing methods, but may
1227 use the <tt>unwind</tt> instruction.</dd>
1229 <dt><tt><b><a name="ssp">ssp</a></b></tt></dt>
1230 <dd>This attribute indicates that the function should emit a stack smashing
1231 protector. It is in the form of a "canary"—a random value placed on
1232 the stack before the local variables that's checked upon return from the
1233 function to see if it has been overwritten. A heuristic is used to
1234 determine if a function needs stack protectors or not.<br>
1236 If a function that has an <tt>ssp</tt> attribute is inlined into a
1237 function that doesn't have an <tt>ssp</tt> attribute, then the resulting
1238 function will have an <tt>ssp</tt> attribute.</dd>
1240 <dt><tt><b>sspreq</b></tt></dt>
1241 <dd>This attribute indicates that the function should <em>always</em> emit a
1242 stack smashing protector. This overrides
1243 the <tt><a href="#ssp">ssp</a></tt> function attribute.<br>
1245 If a function that has an <tt>sspreq</tt> attribute is inlined into a
1246 function that doesn't have an <tt>sspreq</tt> attribute or which has
1247 an <tt>ssp</tt> attribute, then the resulting function will have
1248 an <tt>sspreq</tt> attribute.</dd>
1250 <dt><tt><b><a name="uwtable">uwtable</a></b></tt></dt>
1251 <dd>This attribute indicates that the ABI being targeted requires that
1252 an unwind table entry be produce for this function even if we can
1253 show that no exceptions passes by it. This is normally the case for
1254 the ELF x86-64 abi, but it can be disabled for some compilation
1257 <dt><tt><b><a name="returns_twice">returns_twice</a></b></tt></dt>
1258 <dd>This attribute indicates that this function can return
1259 twice. The C <code>setjmp</code> is an example of such a function.
1260 The compiler disables some optimizations (like tail calls) in the caller of
1261 these functions.</dd>
1266 <!-- ======================================================================= -->
1268 <a name="moduleasm">Module-Level Inline Assembly</a>
1273 <p>Modules may contain "module-level inline asm" blocks, which corresponds to
1274 the GCC "file scope inline asm" blocks. These blocks are internally
1275 concatenated by LLVM and treated as a single unit, but may be separated in
1276 the <tt>.ll</tt> file if desired. The syntax is very simple:</p>
1278 <pre class="doc_code">
1279 module asm "inline asm code goes here"
1280 module asm "more can go here"
1283 <p>The strings can contain any character by escaping non-printable characters.
1284 The escape sequence used is simply "\xx" where "xx" is the two digit hex code
1287 <p>The inline asm code is simply printed to the machine code .s file when
1288 assembly code is generated.</p>
1292 <!-- ======================================================================= -->
1294 <a name="datalayout">Data Layout</a>
1299 <p>A module may specify a target specific data layout string that specifies how
1300 data is to be laid out in memory. The syntax for the data layout is
1303 <pre class="doc_code">
1304 target datalayout = "<i>layout specification</i>"
1307 <p>The <i>layout specification</i> consists of a list of specifications
1308 separated by the minus sign character ('-'). Each specification starts with
1309 a letter and may include other information after the letter to define some
1310 aspect of the data layout. The specifications accepted are as follows:</p>
1314 <dd>Specifies that the target lays out data in big-endian form. That is, the
1315 bits with the most significance have the lowest address location.</dd>
1318 <dd>Specifies that the target lays out data in little-endian form. That is,
1319 the bits with the least significance have the lowest address
1322 <dt><tt>p:<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1323 <dd>This specifies the <i>size</i> of a pointer and its <i>abi</i> and
1324 <i>preferred</i> alignments. All sizes are in bits. Specifying
1325 the <i>pref</i> alignment is optional. If omitted, the
1326 preceding <tt>:</tt> should be omitted too.</dd>
1328 <dt><tt>i<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1329 <dd>This specifies the alignment for an integer type of a given bit
1330 <i>size</i>. The value of <i>size</i> must be in the range [1,2^23).</dd>
1332 <dt><tt>v<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1333 <dd>This specifies the alignment for a vector type of a given bit
1336 <dt><tt>f<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1337 <dd>This specifies the alignment for a floating point type of a given bit
1338 <i>size</i>. Only values of <i>size</i> that are supported by the target
1339 will work. 32 (float) and 64 (double) are supported on all targets;
1340 80 or 128 (different flavors of long double) are also supported on some
1343 <dt><tt>a<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1344 <dd>This specifies the alignment for an aggregate type of a given bit
1347 <dt><tt>s<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1348 <dd>This specifies the alignment for a stack object of a given bit
1351 <dt><tt>n<i>size1</i>:<i>size2</i>:<i>size3</i>...</tt></dt>
1352 <dd>This specifies a set of native integer widths for the target CPU
1353 in bits. For example, it might contain "n32" for 32-bit PowerPC,
1354 "n32:64" for PowerPC 64, or "n8:16:32:64" for X86-64. Elements of
1355 this set are considered to support most general arithmetic
1356 operations efficiently.</dd>
1359 <p>When constructing the data layout for a given target, LLVM starts with a
1360 default set of specifications which are then (possibly) overridden by the
1361 specifications in the <tt>datalayout</tt> keyword. The default specifications
1362 are given in this list:</p>
1365 <li><tt>E</tt> - big endian</li>
1366 <li><tt>p:64:64:64</tt> - 64-bit pointers with 64-bit alignment</li>
1367 <li><tt>i1:8:8</tt> - i1 is 8-bit (byte) aligned</li>
1368 <li><tt>i8:8:8</tt> - i8 is 8-bit (byte) aligned</li>
1369 <li><tt>i16:16:16</tt> - i16 is 16-bit aligned</li>
1370 <li><tt>i32:32:32</tt> - i32 is 32-bit aligned</li>
1371 <li><tt>i64:32:64</tt> - i64 has ABI alignment of 32-bits but preferred
1372 alignment of 64-bits</li>
1373 <li><tt>f32:32:32</tt> - float is 32-bit aligned</li>
1374 <li><tt>f64:64:64</tt> - double is 64-bit aligned</li>
1375 <li><tt>v64:64:64</tt> - 64-bit vector is 64-bit aligned</li>
1376 <li><tt>v128:128:128</tt> - 128-bit vector is 128-bit aligned</li>
1377 <li><tt>a0:0:1</tt> - aggregates are 8-bit aligned</li>
1378 <li><tt>s0:64:64</tt> - stack objects are 64-bit aligned</li>
1381 <p>When LLVM is determining the alignment for a given type, it uses the
1382 following rules:</p>
1385 <li>If the type sought is an exact match for one of the specifications, that
1386 specification is used.</li>
1388 <li>If no match is found, and the type sought is an integer type, then the
1389 smallest integer type that is larger than the bitwidth of the sought type
1390 is used. If none of the specifications are larger than the bitwidth then
1391 the the largest integer type is used. For example, given the default
1392 specifications above, the i7 type will use the alignment of i8 (next
1393 largest) while both i65 and i256 will use the alignment of i64 (largest
1396 <li>If no match is found, and the type sought is a vector type, then the
1397 largest vector type that is smaller than the sought vector type will be
1398 used as a fall back. This happens because <128 x double> can be
1399 implemented in terms of 64 <2 x double>, for example.</li>
1404 <!-- ======================================================================= -->
1406 <a name="pointeraliasing">Pointer Aliasing Rules</a>
1411 <p>Any memory access must be done through a pointer value associated
1412 with an address range of the memory access, otherwise the behavior
1413 is undefined. Pointer values are associated with address ranges
1414 according to the following rules:</p>
1417 <li>A pointer value is associated with the addresses associated with
1418 any value it is <i>based</i> on.
1419 <li>An address of a global variable is associated with the address
1420 range of the variable's storage.</li>
1421 <li>The result value of an allocation instruction is associated with
1422 the address range of the allocated storage.</li>
1423 <li>A null pointer in the default address-space is associated with
1425 <li>An integer constant other than zero or a pointer value returned
1426 from a function not defined within LLVM may be associated with address
1427 ranges allocated through mechanisms other than those provided by
1428 LLVM. Such ranges shall not overlap with any ranges of addresses
1429 allocated by mechanisms provided by LLVM.</li>
1432 <p>A pointer value is <i>based</i> on another pointer value according
1433 to the following rules:</p>
1436 <li>A pointer value formed from a
1437 <tt><a href="#i_getelementptr">getelementptr</a></tt> operation
1438 is <i>based</i> on the first operand of the <tt>getelementptr</tt>.</li>
1439 <li>The result value of a
1440 <tt><a href="#i_bitcast">bitcast</a></tt> is <i>based</i> on the operand
1441 of the <tt>bitcast</tt>.</li>
1442 <li>A pointer value formed by an
1443 <tt><a href="#i_inttoptr">inttoptr</a></tt> is <i>based</i> on all
1444 pointer values that contribute (directly or indirectly) to the
1445 computation of the pointer's value.</li>
1446 <li>The "<i>based</i> on" relationship is transitive.</li>
1449 <p>Note that this definition of <i>"based"</i> is intentionally
1450 similar to the definition of <i>"based"</i> in C99, though it is
1451 slightly weaker.</p>
1453 <p>LLVM IR does not associate types with memory. The result type of a
1454 <tt><a href="#i_load">load</a></tt> merely indicates the size and
1455 alignment of the memory from which to load, as well as the
1456 interpretation of the value. The first operand type of a
1457 <tt><a href="#i_store">store</a></tt> similarly only indicates the size
1458 and alignment of the store.</p>
1460 <p>Consequently, type-based alias analysis, aka TBAA, aka
1461 <tt>-fstrict-aliasing</tt>, is not applicable to general unadorned
1462 LLVM IR. <a href="#metadata">Metadata</a> may be used to encode
1463 additional information which specialized optimization passes may use
1464 to implement type-based alias analysis.</p>
1468 <!-- ======================================================================= -->
1470 <a name="volatile">Volatile Memory Accesses</a>
1475 <p>Certain memory accesses, such as <a href="#i_load"><tt>load</tt></a>s, <a
1476 href="#i_store"><tt>store</tt></a>s, and <a
1477 href="#int_memcpy"><tt>llvm.memcpy</tt></a>s may be marked <tt>volatile</tt>.
1478 The optimizers must not change the number of volatile operations or change their
1479 order of execution relative to other volatile operations. The optimizers
1480 <i>may</i> change the order of volatile operations relative to non-volatile
1481 operations. This is not Java's "volatile" and has no cross-thread
1482 synchronization behavior.</p>
1486 <!-- ======================================================================= -->
1488 <a name="memmodel">Memory Model for Concurrent Operations</a>
1493 <p>The LLVM IR does not define any way to start parallel threads of execution
1494 or to register signal handlers. Nonetheless, there are platform-specific
1495 ways to create them, and we define LLVM IR's behavior in their presence. This
1496 model is inspired by the C++0x memory model.</p>
1498 <p>For a more informal introduction to this model, see the
1499 <a href="Atomics.html">LLVM Atomic Instructions and Concurrency Guide</a>.
1501 <p>We define a <i>happens-before</i> partial order as the least partial order
1504 <li>Is a superset of single-thread program order, and</li>
1505 <li>When a <i>synchronizes-with</i> <tt>b</tt>, includes an edge from
1506 <tt>a</tt> to <tt>b</tt>. <i>Synchronizes-with</i> pairs are introduced
1507 by platform-specific techniques, like pthread locks, thread
1508 creation, thread joining, etc., and by atomic instructions.
1509 (See also <a href="#ordering">Atomic Memory Ordering Constraints</a>).
1513 <p>Note that program order does not introduce <i>happens-before</i> edges
1514 between a thread and signals executing inside that thread.</p>
1516 <p>Every (defined) read operation (load instructions, memcpy, atomic
1517 loads/read-modify-writes, etc.) <var>R</var> reads a series of bytes written by
1518 (defined) write operations (store instructions, atomic
1519 stores/read-modify-writes, memcpy, etc.). For the purposes of this section,
1520 initialized globals are considered to have a write of the initializer which is
1521 atomic and happens before any other read or write of the memory in question.
1522 For each byte of a read <var>R</var>, <var>R<sub>byte</sub></var> may see
1523 any write to the same byte, except:</p>
1526 <li>If <var>write<sub>1</sub></var> happens before
1527 <var>write<sub>2</sub></var>, and <var>write<sub>2</sub></var> happens
1528 before <var>R<sub>byte</sub></var>, then <var>R<sub>byte</sub></var>
1529 does not see <var>write<sub>1</sub></var>.
1530 <li>If <var>R<sub>byte</sub></var> happens before
1531 <var>write<sub>3</sub></var>, then <var>R<sub>byte</sub></var> does not
1532 see <var>write<sub>3</sub></var>.
1535 <p>Given that definition, <var>R<sub>byte</sub></var> is defined as follows:
1537 <li>If <var>R</var> is volatile, the result is target-dependent. (Volatile
1538 is supposed to give guarantees which can support
1539 <code>sig_atomic_t</code> in C/C++, and may be used for accesses to
1540 addresses which do not behave like normal memory. It does not generally
1541 provide cross-thread synchronization.)
1542 <li>Otherwise, if there is no write to the same byte that happens before
1543 <var>R<sub>byte</sub></var>, <var>R<sub>byte</sub></var> returns
1544 <tt>undef</tt> for that byte.
1545 <li>Otherwise, if <var>R<sub>byte</sub></var> may see exactly one write,
1546 <var>R<sub>byte</sub></var> returns the value written by that
1548 <li>Otherwise, if <var>R</var> is atomic, and all the writes
1549 <var>R<sub>byte</sub></var> may see are atomic, it chooses one of the
1550 values written. See the <a href="#ordering">Atomic Memory Ordering
1551 Constraints</a> section for additional constraints on how the choice
1553 <li>Otherwise <var>R<sub>byte</sub></var> returns <tt>undef</tt>.</li>
1556 <p><var>R</var> returns the value composed of the series of bytes it read.
1557 This implies that some bytes within the value may be <tt>undef</tt>
1558 <b>without</b> the entire value being <tt>undef</tt>. Note that this only
1559 defines the semantics of the operation; it doesn't mean that targets will
1560 emit more than one instruction to read the series of bytes.</p>
1562 <p>Note that in cases where none of the atomic intrinsics are used, this model
1563 places only one restriction on IR transformations on top of what is required
1564 for single-threaded execution: introducing a store to a byte which might not
1565 otherwise be stored is not allowed in general. (Specifically, in the case
1566 where another thread might write to and read from an address, introducing a
1567 store can change a load that may see exactly one write into a load that may
1568 see multiple writes.)</p>
1570 <!-- FIXME: This model assumes all targets where concurrency is relevant have
1571 a byte-size store which doesn't affect adjacent bytes. As far as I can tell,
1572 none of the backends currently in the tree fall into this category; however,
1573 there might be targets which care. If there are, we want a paragraph
1576 Targets may specify that stores narrower than a certain width are not
1577 available; on such a target, for the purposes of this model, treat any
1578 non-atomic write with an alignment or width less than the minimum width
1579 as if it writes to the relevant surrounding bytes.
1584 <!-- ======================================================================= -->
1586 <a name="ordering">Atomic Memory Ordering Constraints</a>
1591 <p>Atomic instructions (<a href="#i_cmpxchg"><code>cmpxchg</code></a>,
1592 <a href="#i_atomicrmw"><code>atomicrmw</code></a>,
1593 <a href="#i_fence"><code>fence</code></a>,
1594 <a href="#i_load"><code>atomic load</code></a>, and
1595 <a href="#i_store"><code>atomic store</code></a>) take an ordering parameter
1596 that determines which other atomic instructions on the same address they
1597 <i>synchronize with</i>. These semantics are borrowed from Java and C++0x,
1598 but are somewhat more colloquial. If these descriptions aren't precise enough,
1599 check those specs (see spec references in the
1600 <a href="Atomic.html#introduction">atomics guide</a>).
1601 <a href="#i_fence"><code>fence</code></a> instructions
1602 treat these orderings somewhat differently since they don't take an address.
1603 See that instruction's documentation for details.</p>
1605 <p>For a simpler introduction to the ordering constraints, see the
1606 <a href="Atomics.html">LLVM Atomic Instructions and Concurrency Guide</a>.</p>
1609 <dt><code>unordered</code></dt>
1610 <dd>The set of values that can be read is governed by the happens-before
1611 partial order. A value cannot be read unless some operation wrote it.
1612 This is intended to provide a guarantee strong enough to model Java's
1613 non-volatile shared variables. This ordering cannot be specified for
1614 read-modify-write operations; it is not strong enough to make them atomic
1615 in any interesting way.</dd>
1616 <dt><code>monotonic</code></dt>
1617 <dd>In addition to the guarantees of <code>unordered</code>, there is a single
1618 total order for modifications by <code>monotonic</code> operations on each
1619 address. All modification orders must be compatible with the happens-before
1620 order. There is no guarantee that the modification orders can be combined to
1621 a global total order for the whole program (and this often will not be
1622 possible). The read in an atomic read-modify-write operation
1623 (<a href="#i_cmpxchg"><code>cmpxchg</code></a> and
1624 <a href="#i_atomicrmw"><code>atomicrmw</code></a>)
1625 reads the value in the modification order immediately before the value it
1626 writes. If one atomic read happens before another atomic read of the same
1627 address, the later read must see the same value or a later value in the
1628 address's modification order. This disallows reordering of
1629 <code>monotonic</code> (or stronger) operations on the same address. If an
1630 address is written <code>monotonic</code>ally by one thread, and other threads
1631 <code>monotonic</code>ally read that address repeatedly, the other threads must
1632 eventually see the write. This corresponds to the C++0x/C1x
1633 <code>memory_order_relaxed</code>.</dd>
1634 <dt><code>acquire</code></dt>
1635 <dd>In addition to the guarantees of <code>monotonic</code>,
1636 a <i>synchronizes-with</i> edge may be formed with a <code>release</code>
1637 operation. This is intended to model C++'s <code>memory_order_acquire</code>.</dd>
1638 <dt><code>release</code></dt>
1639 <dd>In addition to the guarantees of <code>monotonic</code>, if this operation
1640 writes a value which is subsequently read by an <code>acquire</code> operation,
1641 it <i>synchronizes-with</i> that operation. (This isn't a complete
1642 description; see the C++0x definition of a release sequence.) This corresponds
1643 to the C++0x/C1x <code>memory_order_release</code>.</dd>
1644 <dt><code>acq_rel</code> (acquire+release)</dt><dd>Acts as both an
1645 <code>acquire</code> and <code>release</code> operation on its address.
1646 This corresponds to the C++0x/C1x <code>memory_order_acq_rel</code>.</dd>
1647 <dt><code>seq_cst</code> (sequentially consistent)</dt><dd>
1648 <dd>In addition to the guarantees of <code>acq_rel</code>
1649 (<code>acquire</code> for an operation which only reads, <code>release</code>
1650 for an operation which only writes), there is a global total order on all
1651 sequentially-consistent operations on all addresses, which is consistent with
1652 the <i>happens-before</i> partial order and with the modification orders of
1653 all the affected addresses. Each sequentially-consistent read sees the last
1654 preceding write to the same address in this global order. This corresponds
1655 to the C++0x/C1x <code>memory_order_seq_cst</code> and Java volatile.</dd>
1658 <p id="singlethread">If an atomic operation is marked <code>singlethread</code>,
1659 it only <i>synchronizes with</i> or participates in modification and seq_cst
1660 total orderings with other operations running in the same thread (for example,
1661 in signal handlers).</p>
1667 <!-- *********************************************************************** -->
1668 <h2><a name="typesystem">Type System</a></h2>
1669 <!-- *********************************************************************** -->
1673 <p>The LLVM type system is one of the most important features of the
1674 intermediate representation. Being typed enables a number of optimizations
1675 to be performed on the intermediate representation directly, without having
1676 to do extra analyses on the side before the transformation. A strong type
1677 system makes it easier to read the generated code and enables novel analyses
1678 and transformations that are not feasible to perform on normal three address
1679 code representations.</p>
1681 <!-- ======================================================================= -->
1683 <a name="t_classifications">Type Classifications</a>
1688 <p>The types fall into a few useful classifications:</p>
1690 <table border="1" cellspacing="0" cellpadding="4">
1692 <tr><th>Classification</th><th>Types</th></tr>
1694 <td><a href="#t_integer">integer</a></td>
1695 <td><tt>i1, i2, i3, ... i8, ... i16, ... i32, ... i64, ... </tt></td>
1698 <td><a href="#t_floating">floating point</a></td>
1699 <td><tt>float, double, x86_fp80, fp128, ppc_fp128</tt></td>
1702 <td><a name="t_firstclass">first class</a></td>
1703 <td><a href="#t_integer">integer</a>,
1704 <a href="#t_floating">floating point</a>,
1705 <a href="#t_pointer">pointer</a>,
1706 <a href="#t_vector">vector</a>,
1707 <a href="#t_struct">structure</a>,
1708 <a href="#t_array">array</a>,
1709 <a href="#t_label">label</a>,
1710 <a href="#t_metadata">metadata</a>.
1714 <td><a href="#t_primitive">primitive</a></td>
1715 <td><a href="#t_label">label</a>,
1716 <a href="#t_void">void</a>,
1717 <a href="#t_integer">integer</a>,
1718 <a href="#t_floating">floating point</a>,
1719 <a href="#t_x86mmx">x86mmx</a>,
1720 <a href="#t_metadata">metadata</a>.</td>
1723 <td><a href="#t_derived">derived</a></td>
1724 <td><a href="#t_array">array</a>,
1725 <a href="#t_function">function</a>,
1726 <a href="#t_pointer">pointer</a>,
1727 <a href="#t_struct">structure</a>,
1728 <a href="#t_vector">vector</a>,
1729 <a href="#t_opaque">opaque</a>.
1735 <p>The <a href="#t_firstclass">first class</a> types are perhaps the most
1736 important. Values of these types are the only ones which can be produced by
1741 <!-- ======================================================================= -->
1743 <a name="t_primitive">Primitive Types</a>
1748 <p>The primitive types are the fundamental building blocks of the LLVM
1751 <!-- _______________________________________________________________________ -->
1753 <a name="t_integer">Integer Type</a>
1759 <p>The integer type is a very simple type that simply specifies an arbitrary
1760 bit width for the integer type desired. Any bit width from 1 bit to
1761 2<sup>23</sup>-1 (about 8 million) can be specified.</p>
1768 <p>The number of bits the integer will occupy is specified by the <tt>N</tt>
1772 <table class="layout">
1774 <td class="left"><tt>i1</tt></td>
1775 <td class="left">a single-bit integer.</td>
1778 <td class="left"><tt>i32</tt></td>
1779 <td class="left">a 32-bit integer.</td>
1782 <td class="left"><tt>i1942652</tt></td>
1783 <td class="left">a really big integer of over 1 million bits.</td>
1789 <!-- _______________________________________________________________________ -->
1791 <a name="t_floating">Floating Point Types</a>
1798 <tr><th>Type</th><th>Description</th></tr>
1799 <tr><td><tt>float</tt></td><td>32-bit floating point value</td></tr>
1800 <tr><td><tt>double</tt></td><td>64-bit floating point value</td></tr>
1801 <tr><td><tt>fp128</tt></td><td>128-bit floating point value (112-bit mantissa)</td></tr>
1802 <tr><td><tt>x86_fp80</tt></td><td>80-bit floating point value (X87)</td></tr>
1803 <tr><td><tt>ppc_fp128</tt></td><td>128-bit floating point value (two 64-bits)</td></tr>
1809 <!-- _______________________________________________________________________ -->
1811 <a name="t_x86mmx">X86mmx Type</a>
1817 <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>
1826 <!-- _______________________________________________________________________ -->
1828 <a name="t_void">Void Type</a>
1834 <p>The void type does not represent any value and has no size.</p>
1843 <!-- _______________________________________________________________________ -->
1845 <a name="t_label">Label Type</a>
1851 <p>The label type represents code labels.</p>
1860 <!-- _______________________________________________________________________ -->
1862 <a name="t_metadata">Metadata Type</a>
1868 <p>The metadata type represents embedded metadata. No derived types may be
1869 created from metadata except for <a href="#t_function">function</a>
1881 <!-- ======================================================================= -->
1883 <a name="t_derived">Derived Types</a>
1888 <p>The real power in LLVM comes from the derived types in the system. This is
1889 what allows a programmer to represent arrays, functions, pointers, and other
1890 useful types. Each of these types contain one or more element types which
1891 may be a primitive type, or another derived type. For example, it is
1892 possible to have a two dimensional array, using an array as the element type
1893 of another array.</p>
1898 <!-- _______________________________________________________________________ -->
1900 <a name="t_aggregate">Aggregate Types</a>
1905 <p>Aggregate Types are a subset of derived types that can contain multiple
1906 member types. <a href="#t_array">Arrays</a>,
1907 <a href="#t_struct">structs</a>, and <a href="#t_vector">vectors</a> are
1908 aggregate types.</p>
1912 <!-- _______________________________________________________________________ -->
1914 <a name="t_array">Array Type</a>
1920 <p>The array type is a very simple derived type that arranges elements
1921 sequentially in memory. The array type requires a size (number of elements)
1922 and an underlying data type.</p>
1926 [<# elements> x <elementtype>]
1929 <p>The number of elements is a constant integer value; <tt>elementtype</tt> may
1930 be any type with a size.</p>
1933 <table class="layout">
1935 <td class="left"><tt>[40 x i32]</tt></td>
1936 <td class="left">Array of 40 32-bit integer values.</td>
1939 <td class="left"><tt>[41 x i32]</tt></td>
1940 <td class="left">Array of 41 32-bit integer values.</td>
1943 <td class="left"><tt>[4 x i8]</tt></td>
1944 <td class="left">Array of 4 8-bit integer values.</td>
1947 <p>Here are some examples of multidimensional arrays:</p>
1948 <table class="layout">
1950 <td class="left"><tt>[3 x [4 x i32]]</tt></td>
1951 <td class="left">3x4 array of 32-bit integer values.</td>
1954 <td class="left"><tt>[12 x [10 x float]]</tt></td>
1955 <td class="left">12x10 array of single precision floating point values.</td>
1958 <td class="left"><tt>[2 x [3 x [4 x i16]]]</tt></td>
1959 <td class="left">2x3x4 array of 16-bit integer values.</td>
1963 <p>There is no restriction on indexing beyond the end of the array implied by
1964 a static type (though there are restrictions on indexing beyond the bounds
1965 of an allocated object in some cases). This means that single-dimension
1966 'variable sized array' addressing can be implemented in LLVM with a zero
1967 length array type. An implementation of 'pascal style arrays' in LLVM could
1968 use the type "<tt>{ i32, [0 x float]}</tt>", for example.</p>
1972 <!-- _______________________________________________________________________ -->
1974 <a name="t_function">Function Type</a>
1980 <p>The function type can be thought of as a function signature. It consists of
1981 a return type and a list of formal parameter types. The return type of a
1982 function type is a first class type or a void type.</p>
1986 <returntype> (<parameter list>)
1989 <p>...where '<tt><parameter list></tt>' is a comma-separated list of type
1990 specifiers. Optionally, the parameter list may include a type <tt>...</tt>,
1991 which indicates that the function takes a variable number of arguments.
1992 Variable argument functions can access their arguments with
1993 the <a href="#int_varargs">variable argument handling intrinsic</a>
1994 functions. '<tt><returntype></tt>' is any type except
1995 <a href="#t_label">label</a>.</p>
1998 <table class="layout">
2000 <td class="left"><tt>i32 (i32)</tt></td>
2001 <td class="left">function taking an <tt>i32</tt>, returning an <tt>i32</tt>
2003 </tr><tr class="layout">
2004 <td class="left"><tt>float (i16, i32 *) *
2006 <td class="left"><a href="#t_pointer">Pointer</a> to a function that takes
2007 an <tt>i16</tt> and a <a href="#t_pointer">pointer</a> to <tt>i32</tt>,
2008 returning <tt>float</tt>.
2010 </tr><tr class="layout">
2011 <td class="left"><tt>i32 (i8*, ...)</tt></td>
2012 <td class="left">A vararg function that takes at least one
2013 <a href="#t_pointer">pointer</a> to <tt>i8 </tt> (char in C),
2014 which returns an integer. This is the signature for <tt>printf</tt> in
2017 </tr><tr class="layout">
2018 <td class="left"><tt>{i32, i32} (i32)</tt></td>
2019 <td class="left">A function taking an <tt>i32</tt>, returning a
2020 <a href="#t_struct">structure</a> containing two <tt>i32</tt> values
2027 <!-- _______________________________________________________________________ -->
2029 <a name="t_struct">Structure Type</a>
2035 <p>The structure type is used to represent a collection of data members together
2036 in memory. The elements of a structure may be any type that has a size.</p>
2038 <p>Structures in memory are accessed using '<tt><a href="#i_load">load</a></tt>'
2039 and '<tt><a href="#i_store">store</a></tt>' by getting a pointer to a field
2040 with the '<tt><a href="#i_getelementptr">getelementptr</a></tt>' instruction.
2041 Structures in registers are accessed using the
2042 '<tt><a href="#i_extractvalue">extractvalue</a></tt>' and
2043 '<tt><a href="#i_insertvalue">insertvalue</a></tt>' instructions.</p>
2045 <p>Structures may optionally be "packed" structures, which indicate that the
2046 alignment of the struct is one byte, and that there is no padding between
2047 the elements. In non-packed structs, padding between field types is inserted
2048 as defined by the TargetData string in the module, which is required to match
2049 what the underlying processor expects.</p>
2051 <p>Structures can either be "literal" or "identified". A literal structure is
2052 defined inline with other types (e.g. <tt>{i32, i32}*</tt>) whereas identified
2053 types are always defined at the top level with a name. Literal types are
2054 uniqued by their contents and can never be recursive or opaque since there is
2055 no way to write one. Identified types can be recursive, can be opaqued, and are
2061 %T1 = type { <type list> } <i>; Identified normal struct type</i>
2062 %T2 = type <{ <type list> }> <i>; Identified packed struct type</i>
2066 <table class="layout">
2068 <td class="left"><tt>{ i32, i32, i32 }</tt></td>
2069 <td class="left">A triple of three <tt>i32</tt> values</td>
2072 <td class="left"><tt>{ float, i32 (i32) * }</tt></td>
2073 <td class="left">A pair, where the first element is a <tt>float</tt> and the
2074 second element is a <a href="#t_pointer">pointer</a> to a
2075 <a href="#t_function">function</a> that takes an <tt>i32</tt>, returning
2076 an <tt>i32</tt>.</td>
2079 <td class="left"><tt><{ i8, i32 }></tt></td>
2080 <td class="left">A packed struct known to be 5 bytes in size.</td>
2086 <!-- _______________________________________________________________________ -->
2088 <a name="t_opaque">Opaque Structure Types</a>
2094 <p>Opaque structure types are used to represent named structure types that do
2095 not have a body specified. This corresponds (for example) to the C notion of
2096 a forward declared structure.</p>
2105 <table class="layout">
2107 <td class="left"><tt>opaque</tt></td>
2108 <td class="left">An opaque type.</td>
2116 <!-- _______________________________________________________________________ -->
2118 <a name="t_pointer">Pointer Type</a>
2124 <p>The pointer type is used to specify memory locations.
2125 Pointers are commonly used to reference objects in memory.</p>
2127 <p>Pointer types may have an optional address space attribute defining the
2128 numbered address space where the pointed-to object resides. The default
2129 address space is number zero. The semantics of non-zero address
2130 spaces are target-specific.</p>
2132 <p>Note that LLVM does not permit pointers to void (<tt>void*</tt>) nor does it
2133 permit pointers to labels (<tt>label*</tt>). Use <tt>i8*</tt> instead.</p>
2141 <table class="layout">
2143 <td class="left"><tt>[4 x i32]*</tt></td>
2144 <td class="left">A <a href="#t_pointer">pointer</a> to <a
2145 href="#t_array">array</a> of four <tt>i32</tt> values.</td>
2148 <td class="left"><tt>i32 (i32*) *</tt></td>
2149 <td class="left"> A <a href="#t_pointer">pointer</a> to a <a
2150 href="#t_function">function</a> that takes an <tt>i32*</tt>, returning an
2154 <td class="left"><tt>i32 addrspace(5)*</tt></td>
2155 <td class="left">A <a href="#t_pointer">pointer</a> to an <tt>i32</tt> value
2156 that resides in address space #5.</td>
2162 <!-- _______________________________________________________________________ -->
2164 <a name="t_vector">Vector Type</a>
2170 <p>A vector type is a simple derived type that represents a vector of elements.
2171 Vector types are used when multiple primitive data are operated in parallel
2172 using a single instruction (SIMD). A vector type requires a size (number of
2173 elements) and an underlying primitive data type. Vector types are considered
2174 <a href="#t_firstclass">first class</a>.</p>
2178 < <# elements> x <elementtype> >
2181 <p>The number of elements is a constant integer value larger than 0; elementtype
2182 may be any integer or floating point type. Vectors of size zero are not
2183 allowed, and pointers are not allowed as the element type.</p>
2186 <table class="layout">
2188 <td class="left"><tt><4 x i32></tt></td>
2189 <td class="left">Vector of 4 32-bit integer values.</td>
2192 <td class="left"><tt><8 x float></tt></td>
2193 <td class="left">Vector of 8 32-bit floating-point values.</td>
2196 <td class="left"><tt><2 x i64></tt></td>
2197 <td class="left">Vector of 2 64-bit integer values.</td>
2205 <!-- *********************************************************************** -->
2206 <h2><a name="constants">Constants</a></h2>
2207 <!-- *********************************************************************** -->
2211 <p>LLVM has several different basic types of constants. This section describes
2212 them all and their syntax.</p>
2214 <!-- ======================================================================= -->
2216 <a name="simpleconstants">Simple Constants</a>
2222 <dt><b>Boolean constants</b></dt>
2223 <dd>The two strings '<tt>true</tt>' and '<tt>false</tt>' are both valid
2224 constants of the <tt><a href="#t_integer">i1</a></tt> type.</dd>
2226 <dt><b>Integer constants</b></dt>
2227 <dd>Standard integers (such as '4') are constants of
2228 the <a href="#t_integer">integer</a> type. Negative numbers may be used
2229 with integer types.</dd>
2231 <dt><b>Floating point constants</b></dt>
2232 <dd>Floating point constants use standard decimal notation (e.g. 123.421),
2233 exponential notation (e.g. 1.23421e+2), or a more precise hexadecimal
2234 notation (see below). The assembler requires the exact decimal value of a
2235 floating-point constant. For example, the assembler accepts 1.25 but
2236 rejects 1.3 because 1.3 is a repeating decimal in binary. Floating point
2237 constants must have a <a href="#t_floating">floating point</a> type. </dd>
2239 <dt><b>Null pointer constants</b></dt>
2240 <dd>The identifier '<tt>null</tt>' is recognized as a null pointer constant
2241 and must be of <a href="#t_pointer">pointer type</a>.</dd>
2244 <p>The one non-intuitive notation for constants is the hexadecimal form of
2245 floating point constants. For example, the form '<tt>double
2246 0x432ff973cafa8000</tt>' is equivalent to (but harder to read than)
2247 '<tt>double 4.5e+15</tt>'. The only time hexadecimal floating point
2248 constants are required (and the only time that they are generated by the
2249 disassembler) is when a floating point constant must be emitted but it cannot
2250 be represented as a decimal floating point number in a reasonable number of
2251 digits. For example, NaN's, infinities, and other special values are
2252 represented in their IEEE hexadecimal format so that assembly and disassembly
2253 do not cause any bits to change in the constants.</p>
2255 <p>When using the hexadecimal form, constants of types float and double are
2256 represented using the 16-digit form shown above (which matches the IEEE754
2257 representation for double); float values must, however, be exactly
2258 representable as IEE754 single precision. Hexadecimal format is always used
2259 for long double, and there are three forms of long double. The 80-bit format
2260 used by x86 is represented as <tt>0xK</tt> followed by 20 hexadecimal digits.
2261 The 128-bit format used by PowerPC (two adjacent doubles) is represented
2262 by <tt>0xM</tt> followed by 32 hexadecimal digits. The IEEE 128-bit format
2263 is represented by <tt>0xL</tt> followed by 32 hexadecimal digits; no
2264 currently supported target uses this format. Long doubles will only work if
2265 they match the long double format on your target. All hexadecimal formats
2266 are big-endian (sign bit at the left).</p>
2268 <p>There are no constants of type x86mmx.</p>
2271 <!-- ======================================================================= -->
2273 <a name="aggregateconstants"></a> <!-- old anchor -->
2274 <a name="complexconstants">Complex Constants</a>
2279 <p>Complex constants are a (potentially recursive) combination of simple
2280 constants and smaller complex constants.</p>
2283 <dt><b>Structure constants</b></dt>
2284 <dd>Structure constants are represented with notation similar to structure
2285 type definitions (a comma separated list of elements, surrounded by braces
2286 (<tt>{}</tt>)). For example: "<tt>{ i32 4, float 17.0, i32* @G }</tt>",
2287 where "<tt>@G</tt>" is declared as "<tt>@G = external global i32</tt>".
2288 Structure constants must have <a href="#t_struct">structure type</a>, and
2289 the number and types of elements must match those specified by the
2292 <dt><b>Array constants</b></dt>
2293 <dd>Array constants are represented with notation similar to array type
2294 definitions (a comma separated list of elements, surrounded by square
2295 brackets (<tt>[]</tt>)). For example: "<tt>[ i32 42, i32 11, i32 74
2296 ]</tt>". Array constants must have <a href="#t_array">array type</a>, and
2297 the number and types of elements must match those specified by the
2300 <dt><b>Vector constants</b></dt>
2301 <dd>Vector constants are represented with notation similar to vector type
2302 definitions (a comma separated list of elements, surrounded by
2303 less-than/greater-than's (<tt><></tt>)). For example: "<tt>< i32
2304 42, i32 11, i32 74, i32 100 ></tt>". Vector constants must
2305 have <a href="#t_vector">vector type</a>, and the number and types of
2306 elements must match those specified by the type.</dd>
2308 <dt><b>Zero initialization</b></dt>
2309 <dd>The string '<tt>zeroinitializer</tt>' can be used to zero initialize a
2310 value to zero of <em>any</em> type, including scalar and
2311 <a href="#t_aggregate">aggregate</a> types.
2312 This is often used to avoid having to print large zero initializers
2313 (e.g. for large arrays) and is always exactly equivalent to using explicit
2314 zero initializers.</dd>
2316 <dt><b>Metadata node</b></dt>
2317 <dd>A metadata node is a structure-like constant with
2318 <a href="#t_metadata">metadata type</a>. For example: "<tt>metadata !{
2319 i32 0, metadata !"test" }</tt>". Unlike other constants that are meant to
2320 be interpreted as part of the instruction stream, metadata is a place to
2321 attach additional information such as debug info.</dd>
2326 <!-- ======================================================================= -->
2328 <a name="globalconstants">Global Variable and Function Addresses</a>
2333 <p>The addresses of <a href="#globalvars">global variables</a>
2334 and <a href="#functionstructure">functions</a> are always implicitly valid
2335 (link-time) constants. These constants are explicitly referenced when
2336 the <a href="#identifiers">identifier for the global</a> is used and always
2337 have <a href="#t_pointer">pointer</a> type. For example, the following is a
2338 legal LLVM file:</p>
2340 <pre class="doc_code">
2343 @Z = global [2 x i32*] [ i32* @X, i32* @Y ]
2348 <!-- ======================================================================= -->
2350 <a name="undefvalues">Undefined Values</a>
2355 <p>The string '<tt>undef</tt>' can be used anywhere a constant is expected, and
2356 indicates that the user of the value may receive an unspecified bit-pattern.
2357 Undefined values may be of any type (other than '<tt>label</tt>'
2358 or '<tt>void</tt>') and be used anywhere a constant is permitted.</p>
2360 <p>Undefined values are useful because they indicate to the compiler that the
2361 program is well defined no matter what value is used. This gives the
2362 compiler more freedom to optimize. Here are some examples of (potentially
2363 surprising) transformations that are valid (in pseudo IR):</p>
2366 <pre class="doc_code">
2376 <p>This is safe because all of the output bits are affected by the undef bits.
2377 Any output bit can have a zero or one depending on the input bits.</p>
2379 <pre class="doc_code">
2390 <p>These logical operations have bits that are not always affected by the input.
2391 For example, if <tt>%X</tt> has a zero bit, then the output of the
2392 '<tt>and</tt>' operation will always be a zero for that bit, no matter what
2393 the corresponding bit from the '<tt>undef</tt>' is. As such, it is unsafe to
2394 optimize or assume that the result of the '<tt>and</tt>' is '<tt>undef</tt>'.
2395 However, it is safe to assume that all bits of the '<tt>undef</tt>' could be
2396 0, and optimize the '<tt>and</tt>' to 0. Likewise, it is safe to assume that
2397 all the bits of the '<tt>undef</tt>' operand to the '<tt>or</tt>' could be
2398 set, allowing the '<tt>or</tt>' to be folded to -1.</p>
2400 <pre class="doc_code">
2401 %A = select undef, %X, %Y
2402 %B = select undef, 42, %Y
2403 %C = select %X, %Y, undef
2414 <p>This set of examples shows that undefined '<tt>select</tt>' (and conditional
2415 branch) conditions can go <em>either way</em>, but they have to come from one
2416 of the two operands. In the <tt>%A</tt> example, if <tt>%X</tt> and
2417 <tt>%Y</tt> were both known to have a clear low bit, then <tt>%A</tt> would
2418 have to have a cleared low bit. However, in the <tt>%C</tt> example, the
2419 optimizer is allowed to assume that the '<tt>undef</tt>' operand could be the
2420 same as <tt>%Y</tt>, allowing the whole '<tt>select</tt>' to be
2423 <pre class="doc_code">
2424 %A = xor undef, undef
2442 <p>This example points out that two '<tt>undef</tt>' operands are not
2443 necessarily the same. This can be surprising to people (and also matches C
2444 semantics) where they assume that "<tt>X^X</tt>" is always zero, even
2445 if <tt>X</tt> is undefined. This isn't true for a number of reasons, but the
2446 short answer is that an '<tt>undef</tt>' "variable" can arbitrarily change
2447 its value over its "live range". This is true because the variable doesn't
2448 actually <em>have a live range</em>. Instead, the value is logically read
2449 from arbitrary registers that happen to be around when needed, so the value
2450 is not necessarily consistent over time. In fact, <tt>%A</tt> and <tt>%C</tt>
2451 need to have the same semantics or the core LLVM "replace all uses with"
2452 concept would not hold.</p>
2454 <pre class="doc_code">
2462 <p>These examples show the crucial difference between an <em>undefined
2463 value</em> and <em>undefined behavior</em>. An undefined value (like
2464 '<tt>undef</tt>') is allowed to have an arbitrary bit-pattern. This means that
2465 the <tt>%A</tt> operation can be constant folded to '<tt>undef</tt>', because
2466 the '<tt>undef</tt>' could be an SNaN, and <tt>fdiv</tt> is not (currently)
2467 defined on SNaN's. However, in the second example, we can make a more
2468 aggressive assumption: because the <tt>undef</tt> is allowed to be an
2469 arbitrary value, we are allowed to assume that it could be zero. Since a
2470 divide by zero has <em>undefined behavior</em>, we are allowed to assume that
2471 the operation does not execute at all. This allows us to delete the divide and
2472 all code after it. Because the undefined operation "can't happen", the
2473 optimizer can assume that it occurs in dead code.</p>
2475 <pre class="doc_code">
2476 a: store undef -> %X
2477 b: store %X -> undef
2483 <p>These examples reiterate the <tt>fdiv</tt> example: a store <em>of</em> an
2484 undefined value can be assumed to not have any effect; we can assume that the
2485 value is overwritten with bits that happen to match what was already there.
2486 However, a store <em>to</em> an undefined location could clobber arbitrary
2487 memory, therefore, it has undefined behavior.</p>
2491 <!-- ======================================================================= -->
2493 <a name="trapvalues">Trap Values</a>
2498 <p>Trap values are similar to <a href="#undefvalues">undef values</a>, however
2499 instead of representing an unspecified bit pattern, they represent the
2500 fact that an instruction or constant expression which cannot evoke side
2501 effects has nevertheless detected a condition which results in undefined
2504 <p>There is currently no way of representing a trap value in the IR; they
2505 only exist when produced by operations such as
2506 <a href="#i_add"><tt>add</tt></a> with the <tt>nsw</tt> flag.</p>
2508 <p>Trap value behavior is defined in terms of value <i>dependence</i>:</p>
2511 <li>Values other than <a href="#i_phi"><tt>phi</tt></a> nodes depend on
2512 their operands.</li>
2514 <li><a href="#i_phi"><tt>Phi</tt></a> nodes depend on the operand corresponding
2515 to their dynamic predecessor basic block.</li>
2517 <li>Function arguments depend on the corresponding actual argument values in
2518 the dynamic callers of their functions.</li>
2520 <li><a href="#i_call"><tt>Call</tt></a> instructions depend on the
2521 <a href="#i_ret"><tt>ret</tt></a> instructions that dynamically transfer
2522 control back to them.</li>
2524 <li><a href="#i_invoke"><tt>Invoke</tt></a> instructions depend on the
2525 <a href="#i_ret"><tt>ret</tt></a>, <a href="#i_unwind"><tt>unwind</tt></a>,
2526 or exception-throwing call instructions that dynamically transfer control
2529 <li>Non-volatile loads and stores depend on the most recent stores to all of the
2530 referenced memory addresses, following the order in the IR
2531 (including loads and stores implied by intrinsics such as
2532 <a href="#int_memcpy"><tt>@llvm.memcpy</tt></a>.)</li>
2534 <!-- TODO: In the case of multiple threads, this only applies if the store
2535 "happens-before" the load or store. -->
2537 <!-- TODO: floating-point exception state -->
2539 <li>An instruction with externally visible side effects depends on the most
2540 recent preceding instruction with externally visible side effects, following
2541 the order in the IR. (This includes
2542 <a href="#volatile">volatile operations</a>.)</li>
2544 <li>An instruction <i>control-depends</i> on a
2545 <a href="#terminators">terminator instruction</a>
2546 if the terminator instruction has multiple successors and the instruction
2547 is always executed when control transfers to one of the successors, and
2548 may not be executed when control is transferred to another.</li>
2550 <li>Additionally, an instruction also <i>control-depends</i> on a terminator
2551 instruction if the set of instructions it otherwise depends on would be
2552 different if the terminator had transferred control to a different
2555 <li>Dependence is transitive.</li>
2559 <p>Whenever a trap value is generated, all values which depend on it evaluate
2560 to trap. If they have side effects, the evoke their side effects as if each
2561 operand with a trap value were undef. If they have externally-visible side
2562 effects, the behavior is undefined.</p>
2564 <p>Here are some examples:</p>
2566 <pre class="doc_code">
2568 %trap = sub nuw i32 0, 1 ; Results in a trap value.
2569 %still_trap = and i32 %trap, 0 ; Whereas (and i32 undef, 0) would return 0.
2570 %trap_yet_again = getelementptr i32* @h, i32 %still_trap
2571 store i32 0, i32* %trap_yet_again ; undefined behavior
2573 store i32 %trap, i32* @g ; Trap value conceptually stored to memory.
2574 %trap2 = load i32* @g ; Returns a trap value, not just undef.
2576 volatile store i32 %trap, i32* @g ; External observation; undefined behavior.
2578 %narrowaddr = bitcast i32* @g to i16*
2579 %wideaddr = bitcast i32* @g to i64*
2580 %trap3 = load i16* %narrowaddr ; Returns a trap value.
2581 %trap4 = load i64* %wideaddr ; Returns a trap value.
2583 %cmp = icmp slt i32 %trap, 0 ; Returns a trap value.
2584 br i1 %cmp, label %true, label %end ; Branch to either destination.
2587 volatile store i32 0, i32* @g ; This is control-dependent on %cmp, so
2588 ; it has undefined behavior.
2592 %p = phi i32 [ 0, %entry ], [ 1, %true ]
2593 ; Both edges into this PHI are
2594 ; control-dependent on %cmp, so this
2595 ; always results in a trap value.
2597 volatile store i32 0, i32* @g ; This would depend on the store in %true
2598 ; if %cmp is true, or the store in %entry
2599 ; otherwise, so this is undefined behavior.
2601 br i1 %cmp, label %second_true, label %second_end
2602 ; The same branch again, but this time the
2603 ; true block doesn't have side effects.
2610 volatile store i32 0, i32* @g ; This time, the instruction always depends
2611 ; on the store in %end. Also, it is
2612 ; control-equivalent to %end, so this is
2613 ; well-defined (again, ignoring earlier
2614 ; undefined behavior in this example).
2619 <!-- ======================================================================= -->
2621 <a name="blockaddress">Addresses of Basic Blocks</a>
2626 <p><b><tt>blockaddress(@function, %block)</tt></b></p>
2628 <p>The '<tt>blockaddress</tt>' constant computes the address of the specified
2629 basic block in the specified function, and always has an i8* type. Taking
2630 the address of the entry block is illegal.</p>
2632 <p>This value only has defined behavior when used as an operand to the
2633 '<a href="#i_indirectbr"><tt>indirectbr</tt></a>' instruction, or for
2634 comparisons against null. Pointer equality tests between labels addresses
2635 results in undefined behavior — though, again, comparison against null
2636 is ok, and no label is equal to the null pointer. This may be passed around
2637 as an opaque pointer sized value as long as the bits are not inspected. This
2638 allows <tt>ptrtoint</tt> and arithmetic to be performed on these values so
2639 long as the original value is reconstituted before the <tt>indirectbr</tt>
2642 <p>Finally, some targets may provide defined semantics when using the value as
2643 the operand to an inline assembly, but that is target specific.</p>
2648 <!-- ======================================================================= -->
2650 <a name="constantexprs">Constant Expressions</a>
2655 <p>Constant expressions are used to allow expressions involving other constants
2656 to be used as constants. Constant expressions may be of
2657 any <a href="#t_firstclass">first class</a> type and may involve any LLVM
2658 operation that does not have side effects (e.g. load and call are not
2659 supported). The following is the syntax for constant expressions:</p>
2662 <dt><b><tt>trunc (CST to TYPE)</tt></b></dt>
2663 <dd>Truncate a constant to another type. The bit size of CST must be larger
2664 than the bit size of TYPE. Both types must be integers.</dd>
2666 <dt><b><tt>zext (CST to TYPE)</tt></b></dt>
2667 <dd>Zero extend a constant to another type. The bit size of CST must be
2668 smaller than the bit size of TYPE. Both types must be integers.</dd>
2670 <dt><b><tt>sext (CST to TYPE)</tt></b></dt>
2671 <dd>Sign extend a constant to another type. The bit size of CST must be
2672 smaller than the bit size of TYPE. Both types must be integers.</dd>
2674 <dt><b><tt>fptrunc (CST to TYPE)</tt></b></dt>
2675 <dd>Truncate a floating point constant to another floating point type. The
2676 size of CST must be larger than the size of TYPE. Both types must be
2677 floating point.</dd>
2679 <dt><b><tt>fpext (CST to TYPE)</tt></b></dt>
2680 <dd>Floating point extend a constant to another type. The size of CST must be
2681 smaller or equal to the size of TYPE. Both types must be floating
2684 <dt><b><tt>fptoui (CST to TYPE)</tt></b></dt>
2685 <dd>Convert a floating point constant to the corresponding unsigned integer
2686 constant. TYPE must be a scalar or vector integer type. CST must be of
2687 scalar or vector floating point type. Both CST and TYPE must be scalars,
2688 or vectors of the same number of elements. If the value won't fit in the
2689 integer type, the results are undefined.</dd>
2691 <dt><b><tt>fptosi (CST to TYPE)</tt></b></dt>
2692 <dd>Convert a floating point constant to the corresponding signed integer
2693 constant. TYPE must be a scalar or vector integer type. CST must be of
2694 scalar or vector floating point type. Both CST and TYPE must be scalars,
2695 or vectors of the same number of elements. If the value won't fit in the
2696 integer type, the results are undefined.</dd>
2698 <dt><b><tt>uitofp (CST to TYPE)</tt></b></dt>
2699 <dd>Convert an unsigned integer constant to the corresponding floating point
2700 constant. TYPE must be a scalar or vector floating point type. CST must be
2701 of scalar or vector integer type. Both CST and TYPE must be scalars, or
2702 vectors of the same number of elements. If the value won't fit in the
2703 floating point type, the results are undefined.</dd>
2705 <dt><b><tt>sitofp (CST to TYPE)</tt></b></dt>
2706 <dd>Convert a signed integer constant to the corresponding floating point
2707 constant. TYPE must be a scalar or vector floating point type. CST must be
2708 of scalar or vector integer type. Both CST and TYPE must be scalars, or
2709 vectors of the same number of elements. If the value won't fit in the
2710 floating point type, the results are undefined.</dd>
2712 <dt><b><tt>ptrtoint (CST to TYPE)</tt></b></dt>
2713 <dd>Convert a pointer typed constant to the corresponding integer constant
2714 <tt>TYPE</tt> must be an integer type. <tt>CST</tt> must be of pointer
2715 type. The <tt>CST</tt> value is zero extended, truncated, or unchanged to
2716 make it fit in <tt>TYPE</tt>.</dd>
2718 <dt><b><tt>inttoptr (CST to TYPE)</tt></b></dt>
2719 <dd>Convert a integer constant to a pointer constant. TYPE must be a pointer
2720 type. CST must be of integer type. The CST value is zero extended,
2721 truncated, or unchanged to make it fit in a pointer size. This one is
2722 <i>really</i> dangerous!</dd>
2724 <dt><b><tt>bitcast (CST to TYPE)</tt></b></dt>
2725 <dd>Convert a constant, CST, to another TYPE. The constraints of the operands
2726 are the same as those for the <a href="#i_bitcast">bitcast
2727 instruction</a>.</dd>
2729 <dt><b><tt>getelementptr (CSTPTR, IDX0, IDX1, ...)</tt></b></dt>
2730 <dt><b><tt>getelementptr inbounds (CSTPTR, IDX0, IDX1, ...)</tt></b></dt>
2731 <dd>Perform the <a href="#i_getelementptr">getelementptr operation</a> on
2732 constants. As with the <a href="#i_getelementptr">getelementptr</a>
2733 instruction, the index list may have zero or more indexes, which are
2734 required to make sense for the type of "CSTPTR".</dd>
2736 <dt><b><tt>select (COND, VAL1, VAL2)</tt></b></dt>
2737 <dd>Perform the <a href="#i_select">select operation</a> on constants.</dd>
2739 <dt><b><tt>icmp COND (VAL1, VAL2)</tt></b></dt>
2740 <dd>Performs the <a href="#i_icmp">icmp operation</a> on constants.</dd>
2742 <dt><b><tt>fcmp COND (VAL1, VAL2)</tt></b></dt>
2743 <dd>Performs the <a href="#i_fcmp">fcmp operation</a> on constants.</dd>
2745 <dt><b><tt>extractelement (VAL, IDX)</tt></b></dt>
2746 <dd>Perform the <a href="#i_extractelement">extractelement operation</a> on
2749 <dt><b><tt>insertelement (VAL, ELT, IDX)</tt></b></dt>
2750 <dd>Perform the <a href="#i_insertelement">insertelement operation</a> on
2753 <dt><b><tt>shufflevector (VEC1, VEC2, IDXMASK)</tt></b></dt>
2754 <dd>Perform the <a href="#i_shufflevector">shufflevector operation</a> on
2757 <dt><b><tt>extractvalue (VAL, IDX0, IDX1, ...)</tt></b></dt>
2758 <dd>Perform the <a href="#i_extractvalue">extractvalue operation</a> on
2759 constants. The index list is interpreted in a similar manner as indices in
2760 a '<a href="#i_getelementptr">getelementptr</a>' operation. At least one
2761 index value must be specified.</dd>
2763 <dt><b><tt>insertvalue (VAL, ELT, IDX0, IDX1, ...)</tt></b></dt>
2764 <dd>Perform the <a href="#i_insertvalue">insertvalue operation</a> on
2765 constants. The index list is interpreted in a similar manner as indices in
2766 a '<a href="#i_getelementptr">getelementptr</a>' operation. At least one
2767 index value must be specified.</dd>
2769 <dt><b><tt>OPCODE (LHS, RHS)</tt></b></dt>
2770 <dd>Perform the specified operation of the LHS and RHS constants. OPCODE may
2771 be any of the <a href="#binaryops">binary</a>
2772 or <a href="#bitwiseops">bitwise binary</a> operations. The constraints
2773 on operands are the same as those for the corresponding instruction
2774 (e.g. no bitwise operations on floating point values are allowed).</dd>
2781 <!-- *********************************************************************** -->
2782 <h2><a name="othervalues">Other Values</a></h2>
2783 <!-- *********************************************************************** -->
2785 <!-- ======================================================================= -->
2787 <a name="inlineasm">Inline Assembler Expressions</a>
2792 <p>LLVM supports inline assembler expressions (as opposed
2793 to <a href="#moduleasm"> Module-Level Inline Assembly</a>) through the use of
2794 a special value. This value represents the inline assembler as a string
2795 (containing the instructions to emit), a list of operand constraints (stored
2796 as a string), a flag that indicates whether or not the inline asm
2797 expression has side effects, and a flag indicating whether the function
2798 containing the asm needs to align its stack conservatively. An example
2799 inline assembler expression is:</p>
2801 <pre class="doc_code">
2802 i32 (i32) asm "bswap $0", "=r,r"
2805 <p>Inline assembler expressions may <b>only</b> be used as the callee operand of
2806 a <a href="#i_call"><tt>call</tt> instruction</a>. Thus, typically we
2809 <pre class="doc_code">
2810 %X = call i32 asm "<a href="#int_bswap">bswap</a> $0", "=r,r"(i32 %Y)
2813 <p>Inline asms with side effects not visible in the constraint list must be
2814 marked as having side effects. This is done through the use of the
2815 '<tt>sideeffect</tt>' keyword, like so:</p>
2817 <pre class="doc_code">
2818 call void asm sideeffect "eieio", ""()
2821 <p>In some cases inline asms will contain code that will not work unless the
2822 stack is aligned in some way, such as calls or SSE instructions on x86,
2823 yet will not contain code that does that alignment within the asm.
2824 The compiler should make conservative assumptions about what the asm might
2825 contain and should generate its usual stack alignment code in the prologue
2826 if the '<tt>alignstack</tt>' keyword is present:</p>
2828 <pre class="doc_code">
2829 call void asm alignstack "eieio", ""()
2832 <p>If both keywords appear the '<tt>sideeffect</tt>' keyword must come
2835 <p>TODO: The format of the asm and constraints string still need to be
2836 documented here. Constraints on what can be done (e.g. duplication, moving,
2837 etc need to be documented). This is probably best done by reference to
2838 another document that covers inline asm from a holistic perspective.</p>
2841 <a name="inlineasm_md">Inline Asm Metadata</a>
2846 <p>The call instructions that wrap inline asm nodes may have a "!srcloc" MDNode
2847 attached to it that contains a list of constant integers. If present, the
2848 code generator will use the integer as the location cookie value when report
2849 errors through the LLVMContext error reporting mechanisms. This allows a
2850 front-end to correlate backend errors that occur with inline asm back to the
2851 source code that produced it. For example:</p>
2853 <pre class="doc_code">
2854 call void asm sideeffect "something bad", ""()<b>, !srcloc !42</b>
2856 !42 = !{ i32 1234567 }
2859 <p>It is up to the front-end to make sense of the magic numbers it places in the
2860 IR. If the MDNode contains multiple constants, the code generator will use
2861 the one that corresponds to the line of the asm that the error occurs on.</p>
2867 <!-- ======================================================================= -->
2869 <a name="metadata">Metadata Nodes and Metadata Strings</a>
2874 <p>LLVM IR allows metadata to be attached to instructions in the program that
2875 can convey extra information about the code to the optimizers and code
2876 generator. One example application of metadata is source-level debug
2877 information. There are two metadata primitives: strings and nodes. All
2878 metadata has the <tt>metadata</tt> type and is identified in syntax by a
2879 preceding exclamation point ('<tt>!</tt>').</p>
2881 <p>A metadata string is a string surrounded by double quotes. It can contain
2882 any character by escaping non-printable characters with "\xx" where "xx" is
2883 the two digit hex code. For example: "<tt>!"test\00"</tt>".</p>
2885 <p>Metadata nodes are represented with notation similar to structure constants
2886 (a comma separated list of elements, surrounded by braces and preceded by an
2887 exclamation point). For example: "<tt>!{ metadata !"test\00", i32
2888 10}</tt>". Metadata nodes can have any values as their operand.</p>
2890 <p>A <a href="#namedmetadatastructure">named metadata</a> is a collection of
2891 metadata nodes, which can be looked up in the module symbol table. For
2892 example: "<tt>!foo = metadata !{!4, !3}</tt>".
2894 <p>Metadata can be used as function arguments. Here <tt>llvm.dbg.value</tt>
2895 function is using two metadata arguments.</p>
2897 <div class="doc_code">
2899 call void @llvm.dbg.value(metadata !24, i64 0, metadata !25)
2903 <p>Metadata can be attached with an instruction. Here metadata <tt>!21</tt> is
2904 attached with <tt>add</tt> instruction using <tt>!dbg</tt> identifier.</p>
2906 <div class="doc_code">
2908 %indvar.next = add i64 %indvar, 1, !dbg !21
2916 <!-- *********************************************************************** -->
2918 <a name="intrinsic_globals">Intrinsic Global Variables</a>
2920 <!-- *********************************************************************** -->
2922 <p>LLVM has a number of "magic" global variables that contain data that affect
2923 code generation or other IR semantics. These are documented here. All globals
2924 of this sort should have a section specified as "<tt>llvm.metadata</tt>". This
2925 section and all globals that start with "<tt>llvm.</tt>" are reserved for use
2928 <!-- ======================================================================= -->
2930 <a name="intg_used">The '<tt>llvm.used</tt>' Global Variable</a>
2935 <p>The <tt>@llvm.used</tt> global is an array with i8* element type which has <a
2936 href="#linkage_appending">appending linkage</a>. This array contains a list of
2937 pointers to global variables and functions which may optionally have a pointer
2938 cast formed of bitcast or getelementptr. For example, a legal use of it is:</p>
2944 @llvm.used = appending global [2 x i8*] [
2946 i8* bitcast (i32* @Y to i8*)
2947 ], section "llvm.metadata"
2950 <p>If a global variable appears in the <tt>@llvm.used</tt> list, then the
2951 compiler, assembler, and linker are required to treat the symbol as if there is
2952 a reference to the global that it cannot see. For example, if a variable has
2953 internal linkage and no references other than that from the <tt>@llvm.used</tt>
2954 list, it cannot be deleted. This is commonly used to represent references from
2955 inline asms and other things the compiler cannot "see", and corresponds to
2956 "attribute((used))" in GNU C.</p>
2958 <p>On some targets, the code generator must emit a directive to the assembler or
2959 object file to prevent the assembler and linker from molesting the symbol.</p>
2963 <!-- ======================================================================= -->
2965 <a name="intg_compiler_used">
2966 The '<tt>llvm.compiler.used</tt>' Global Variable
2972 <p>The <tt>@llvm.compiler.used</tt> directive is the same as the
2973 <tt>@llvm.used</tt> directive, except that it only prevents the compiler from
2974 touching the symbol. On targets that support it, this allows an intelligent
2975 linker to optimize references to the symbol without being impeded as it would be
2976 by <tt>@llvm.used</tt>.</p>
2978 <p>This is a rare construct that should only be used in rare circumstances, and
2979 should not be exposed to source languages.</p>
2983 <!-- ======================================================================= -->
2985 <a name="intg_global_ctors">The '<tt>llvm.global_ctors</tt>' Global Variable</a>
2990 %0 = type { i32, void ()* }
2991 @llvm.global_ctors = appending global [1 x %0] [%0 { i32 65535, void ()* @ctor }]
2993 <p>The <tt>@llvm.global_ctors</tt> array contains a list of constructor functions and associated priorities. The functions referenced by this array will be called in ascending order of priority (i.e. lowest first) when the module is loaded. The order of functions with the same priority is not defined.
2998 <!-- ======================================================================= -->
3000 <a name="intg_global_dtors">The '<tt>llvm.global_dtors</tt>' Global Variable</a>
3005 %0 = type { i32, void ()* }
3006 @llvm.global_dtors = appending global [1 x %0] [%0 { i32 65535, void ()* @dtor }]
3009 <p>The <tt>@llvm.global_dtors</tt> array contains a list of destructor functions and associated priorities. The functions referenced by this array will be called in descending order of priority (i.e. highest first) when the module is loaded. The order of functions with the same priority is not defined.
3016 <!-- *********************************************************************** -->
3017 <h2><a name="instref">Instruction Reference</a></h2>
3018 <!-- *********************************************************************** -->
3022 <p>The LLVM instruction set consists of several different classifications of
3023 instructions: <a href="#terminators">terminator
3024 instructions</a>, <a href="#binaryops">binary instructions</a>,
3025 <a href="#bitwiseops">bitwise binary instructions</a>,
3026 <a href="#memoryops">memory instructions</a>, and
3027 <a href="#otherops">other instructions</a>.</p>
3029 <!-- ======================================================================= -->
3031 <a name="terminators">Terminator Instructions</a>
3036 <p>As mentioned <a href="#functionstructure">previously</a>, every basic block
3037 in a program ends with a "Terminator" instruction, which indicates which
3038 block should be executed after the current block is finished. These
3039 terminator instructions typically yield a '<tt>void</tt>' value: they produce
3040 control flow, not values (the one exception being the
3041 '<a href="#i_invoke"><tt>invoke</tt></a>' instruction).</p>
3043 <p>The terminator instructions are:
3044 '<a href="#i_ret"><tt>ret</tt></a>',
3045 '<a href="#i_br"><tt>br</tt></a>',
3046 '<a href="#i_switch"><tt>switch</tt></a>',
3047 '<a href="#i_indirectbr"><tt>indirectbr</tt></a>',
3048 '<a href="#i_invoke"><tt>invoke</tt></a>',
3049 '<a href="#i_unwind"><tt>unwind</tt></a>',
3050 '<a href="#i_resume"><tt>resume</tt></a>', and
3051 '<a href="#i_unreachable"><tt>unreachable</tt></a>'.</p>
3053 <!-- _______________________________________________________________________ -->
3055 <a name="i_ret">'<tt>ret</tt>' Instruction</a>
3062 ret <type> <value> <i>; Return a value from a non-void function</i>
3063 ret void <i>; Return from void function</i>
3067 <p>The '<tt>ret</tt>' instruction is used to return control flow (and optionally
3068 a value) from a function back to the caller.</p>
3070 <p>There are two forms of the '<tt>ret</tt>' instruction: one that returns a
3071 value and then causes control flow, and one that just causes control flow to
3075 <p>The '<tt>ret</tt>' instruction optionally accepts a single argument, the
3076 return value. The type of the return value must be a
3077 '<a href="#t_firstclass">first class</a>' type.</p>
3079 <p>A function is not <a href="#wellformed">well formed</a> if it it has a
3080 non-void return type and contains a '<tt>ret</tt>' instruction with no return
3081 value or a return value with a type that does not match its type, or if it
3082 has a void return type and contains a '<tt>ret</tt>' instruction with a
3086 <p>When the '<tt>ret</tt>' instruction is executed, control flow returns back to
3087 the calling function's context. If the caller is a
3088 "<a href="#i_call"><tt>call</tt></a>" instruction, execution continues at the
3089 instruction after the call. If the caller was an
3090 "<a href="#i_invoke"><tt>invoke</tt></a>" instruction, execution continues at
3091 the beginning of the "normal" destination block. If the instruction returns
3092 a value, that value shall set the call or invoke instruction's return
3097 ret i32 5 <i>; Return an integer value of 5</i>
3098 ret void <i>; Return from a void function</i>
3099 ret { i32, i8 } { i32 4, i8 2 } <i>; Return a struct of values 4 and 2</i>
3103 <!-- _______________________________________________________________________ -->
3105 <a name="i_br">'<tt>br</tt>' Instruction</a>
3112 br i1 <cond>, label <iftrue>, label <iffalse>
3113 br label <dest> <i>; Unconditional branch</i>
3117 <p>The '<tt>br</tt>' instruction is used to cause control flow to transfer to a
3118 different basic block in the current function. There are two forms of this
3119 instruction, corresponding to a conditional branch and an unconditional
3123 <p>The conditional branch form of the '<tt>br</tt>' instruction takes a single
3124 '<tt>i1</tt>' value and two '<tt>label</tt>' values. The unconditional form
3125 of the '<tt>br</tt>' instruction takes a single '<tt>label</tt>' value as a
3129 <p>Upon execution of a conditional '<tt>br</tt>' instruction, the '<tt>i1</tt>'
3130 argument is evaluated. If the value is <tt>true</tt>, control flows to the
3131 '<tt>iftrue</tt>' <tt>label</tt> argument. If "cond" is <tt>false</tt>,
3132 control flows to the '<tt>iffalse</tt>' <tt>label</tt> argument.</p>
3137 %cond = <a href="#i_icmp">icmp</a> eq i32 %a, %b
3138 br i1 %cond, label %IfEqual, label %IfUnequal
3140 <a href="#i_ret">ret</a> i32 1
3142 <a href="#i_ret">ret</a> i32 0
3147 <!-- _______________________________________________________________________ -->
3149 <a name="i_switch">'<tt>switch</tt>' Instruction</a>
3156 switch <intty> <value>, label <defaultdest> [ <intty> <val>, label <dest> ... ]
3160 <p>The '<tt>switch</tt>' instruction is used to transfer control flow to one of
3161 several different places. It is a generalization of the '<tt>br</tt>'
3162 instruction, allowing a branch to occur to one of many possible
3166 <p>The '<tt>switch</tt>' instruction uses three parameters: an integer
3167 comparison value '<tt>value</tt>', a default '<tt>label</tt>' destination,
3168 and an array of pairs of comparison value constants and '<tt>label</tt>'s.
3169 The table is not allowed to contain duplicate constant entries.</p>
3172 <p>The <tt>switch</tt> instruction specifies a table of values and
3173 destinations. When the '<tt>switch</tt>' instruction is executed, this table
3174 is searched for the given value. If the value is found, control flow is
3175 transferred to the corresponding destination; otherwise, control flow is
3176 transferred to the default destination.</p>
3178 <h5>Implementation:</h5>
3179 <p>Depending on properties of the target machine and the particular
3180 <tt>switch</tt> instruction, this instruction may be code generated in
3181 different ways. For example, it could be generated as a series of chained
3182 conditional branches or with a lookup table.</p>
3186 <i>; Emulate a conditional br instruction</i>
3187 %Val = <a href="#i_zext">zext</a> i1 %value to i32
3188 switch i32 %Val, label %truedest [ i32 0, label %falsedest ]
3190 <i>; Emulate an unconditional br instruction</i>
3191 switch i32 0, label %dest [ ]
3193 <i>; Implement a jump table:</i>
3194 switch i32 %val, label %otherwise [ i32 0, label %onzero
3196 i32 2, label %ontwo ]
3202 <!-- _______________________________________________________________________ -->
3204 <a name="i_indirectbr">'<tt>indirectbr</tt>' Instruction</a>
3211 indirectbr <somety>* <address>, [ label <dest1>, label <dest2>, ... ]
3216 <p>The '<tt>indirectbr</tt>' instruction implements an indirect branch to a label
3217 within the current function, whose address is specified by
3218 "<tt>address</tt>". Address must be derived from a <a
3219 href="#blockaddress">blockaddress</a> constant.</p>
3223 <p>The '<tt>address</tt>' argument is the address of the label to jump to. The
3224 rest of the arguments indicate the full set of possible destinations that the
3225 address may point to. Blocks are allowed to occur multiple times in the
3226 destination list, though this isn't particularly useful.</p>
3228 <p>This destination list is required so that dataflow analysis has an accurate
3229 understanding of the CFG.</p>
3233 <p>Control transfers to the block specified in the address argument. All
3234 possible destination blocks must be listed in the label list, otherwise this
3235 instruction has undefined behavior. This implies that jumps to labels
3236 defined in other functions have undefined behavior as well.</p>
3238 <h5>Implementation:</h5>
3240 <p>This is typically implemented with a jump through a register.</p>
3244 indirectbr i8* %Addr, [ label %bb1, label %bb2, label %bb3 ]
3250 <!-- _______________________________________________________________________ -->
3252 <a name="i_invoke">'<tt>invoke</tt>' Instruction</a>
3259 <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>]
3260 to label <normal label> unwind label <exception label>
3264 <p>The '<tt>invoke</tt>' instruction causes control to transfer to a specified
3265 function, with the possibility of control flow transfer to either the
3266 '<tt>normal</tt>' label or the '<tt>exception</tt>' label. If the callee
3267 function returns with the "<tt><a href="#i_ret">ret</a></tt>" instruction,
3268 control flow will return to the "normal" label. If the callee (or any
3269 indirect callees) returns with the "<a href="#i_unwind"><tt>unwind</tt></a>"
3270 instruction, control is interrupted and continued at the dynamically nearest
3271 "exception" label.</p>
3273 <p>The '<tt>exception</tt>' label is a
3274 <i><a href="ExceptionHandling.html#overview">landing pad</a></i> for the
3275 exception. As such, '<tt>exception</tt>' label is required to have the
3276 "<a href="#i_landingpad"><tt>landingpad</tt></a>" instruction, which contains
3277 the information about about the behavior of the program after unwinding
3278 happens, as its first non-PHI instruction. The restrictions on the
3279 "<tt>landingpad</tt>" instruction's tightly couples it to the
3280 "<tt>invoke</tt>" instruction, so that the important information contained
3281 within the "<tt>landingpad</tt>" instruction can't be lost through normal
3285 <p>This instruction requires several arguments:</p>
3288 <li>The optional "cconv" marker indicates which <a href="#callingconv">calling
3289 convention</a> the call should use. If none is specified, the call
3290 defaults to using C calling conventions.</li>
3292 <li>The optional <a href="#paramattrs">Parameter Attributes</a> list for
3293 return values. Only '<tt>zeroext</tt>', '<tt>signext</tt>', and
3294 '<tt>inreg</tt>' attributes are valid here.</li>
3296 <li>'<tt>ptr to function ty</tt>': shall be the signature of the pointer to
3297 function value being invoked. In most cases, this is a direct function
3298 invocation, but indirect <tt>invoke</tt>s are just as possible, branching
3299 off an arbitrary pointer to function value.</li>
3301 <li>'<tt>function ptr val</tt>': An LLVM value containing a pointer to a
3302 function to be invoked. </li>
3304 <li>'<tt>function args</tt>': argument list whose types match the function
3305 signature argument types and parameter attributes. All arguments must be
3306 of <a href="#t_firstclass">first class</a> type. If the function
3307 signature indicates the function accepts a variable number of arguments,
3308 the extra arguments can be specified.</li>
3310 <li>'<tt>normal label</tt>': the label reached when the called function
3311 executes a '<tt><a href="#i_ret">ret</a></tt>' instruction. </li>
3313 <li>'<tt>exception label</tt>': the label reached when a callee returns with
3314 the <a href="#i_unwind"><tt>unwind</tt></a> instruction. </li>
3316 <li>The optional <a href="#fnattrs">function attributes</a> list. Only
3317 '<tt>noreturn</tt>', '<tt>nounwind</tt>', '<tt>readonly</tt>' and
3318 '<tt>readnone</tt>' attributes are valid here.</li>
3322 <p>This instruction is designed to operate as a standard
3323 '<tt><a href="#i_call">call</a></tt>' instruction in most regards. The
3324 primary difference is that it establishes an association with a label, which
3325 is used by the runtime library to unwind the stack.</p>
3327 <p>This instruction is used in languages with destructors to ensure that proper
3328 cleanup is performed in the case of either a <tt>longjmp</tt> or a thrown
3329 exception. Additionally, this is important for implementation of
3330 '<tt>catch</tt>' clauses in high-level languages that support them.</p>
3332 <p>For the purposes of the SSA form, the definition of the value returned by the
3333 '<tt>invoke</tt>' instruction is deemed to occur on the edge from the current
3334 block to the "normal" label. If the callee unwinds then no return value is
3337 <p>Note that the code generator does not yet completely support unwind, and
3338 that the invoke/unwind semantics are likely to change in future versions.</p>
3342 %retval = invoke i32 @Test(i32 15) to label %Continue
3343 unwind label %TestCleanup <i>; {i32}:retval set</i>
3344 %retval = invoke <a href="#callingconv">coldcc</a> i32 %Testfnptr(i32 15) to label %Continue
3345 unwind label %TestCleanup <i>; {i32}:retval set</i>
3350 <!-- _______________________________________________________________________ -->
3353 <a name="i_unwind">'<tt>unwind</tt>' Instruction</a>
3364 <p>The '<tt>unwind</tt>' instruction unwinds the stack, continuing control flow
3365 at the first callee in the dynamic call stack which used
3366 an <a href="#i_invoke"><tt>invoke</tt></a> instruction to perform the call.
3367 This is primarily used to implement exception handling.</p>
3370 <p>The '<tt>unwind</tt>' instruction causes execution of the current function to
3371 immediately halt. The dynamic call stack is then searched for the
3372 first <a href="#i_invoke"><tt>invoke</tt></a> instruction on the call stack.
3373 Once found, execution continues at the "exceptional" destination block
3374 specified by the <tt>invoke</tt> instruction. If there is no <tt>invoke</tt>
3375 instruction in the dynamic call chain, undefined behavior results.</p>
3377 <p>Note that the code generator does not yet completely support unwind, and
3378 that the invoke/unwind semantics are likely to change in future versions.</p>
3382 <!-- _______________________________________________________________________ -->
3385 <a name="i_resume">'<tt>resume</tt>' Instruction</a>
3392 resume <type> <value>
3396 <p>The '<tt>resume</tt>' instruction is a terminator instruction that has no
3400 <p>The '<tt>resume</tt>' instruction requires one argument, which must have the
3401 same type as the result of any '<tt>landingpad</tt>' instruction in the same
3405 <p>The '<tt>resume</tt>' instruction resumes propagation of an existing
3406 (in-flight) exception whose unwinding was interrupted with
3407 a <a href="#i_landingpad"><tt>landingpad</tt></a> instruction.</p>
3411 resume { i8*, i32 } %exn
3416 <!-- _______________________________________________________________________ -->
3419 <a name="i_unreachable">'<tt>unreachable</tt>' Instruction</a>
3430 <p>The '<tt>unreachable</tt>' instruction has no defined semantics. This
3431 instruction is used to inform the optimizer that a particular portion of the
3432 code is not reachable. This can be used to indicate that the code after a
3433 no-return function cannot be reached, and other facts.</p>
3436 <p>The '<tt>unreachable</tt>' instruction has no defined semantics.</p>
3442 <!-- ======================================================================= -->
3444 <a name="binaryops">Binary Operations</a>
3449 <p>Binary operators are used to do most of the computation in a program. They
3450 require two operands of the same type, execute an operation on them, and
3451 produce a single value. The operands might represent multiple data, as is
3452 the case with the <a href="#t_vector">vector</a> data type. The result value
3453 has the same type as its operands.</p>
3455 <p>There are several different binary operators:</p>
3457 <!-- _______________________________________________________________________ -->
3459 <a name="i_add">'<tt>add</tt>' Instruction</a>
3466 <result> = add <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3467 <result> = add nuw <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3468 <result> = add nsw <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3469 <result> = add nuw nsw <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3473 <p>The '<tt>add</tt>' instruction returns the sum of its two operands.</p>
3476 <p>The two arguments to the '<tt>add</tt>' instruction must
3477 be <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of
3478 integer values. Both arguments must have identical types.</p>
3481 <p>The value produced is the integer sum of the two operands.</p>
3483 <p>If the sum has unsigned overflow, the result returned is the mathematical
3484 result modulo 2<sup>n</sup>, where n is the bit width of the result.</p>
3486 <p>Because LLVM integers use a two's complement representation, this instruction
3487 is appropriate for both signed and unsigned integers.</p>
3489 <p><tt>nuw</tt> and <tt>nsw</tt> stand for "No Unsigned Wrap"
3490 and "No Signed Wrap", respectively. If the <tt>nuw</tt> and/or
3491 <tt>nsw</tt> keywords are present, the result value of the <tt>add</tt>
3492 is a <a href="#trapvalues">trap value</a> if unsigned and/or signed overflow,
3493 respectively, occurs.</p>
3497 <result> = add i32 4, %var <i>; yields {i32}:result = 4 + %var</i>
3502 <!-- _______________________________________________________________________ -->
3504 <a name="i_fadd">'<tt>fadd</tt>' Instruction</a>
3511 <result> = fadd <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3515 <p>The '<tt>fadd</tt>' instruction returns the sum of its two operands.</p>
3518 <p>The two arguments to the '<tt>fadd</tt>' instruction must be
3519 <a href="#t_floating">floating point</a> or <a href="#t_vector">vector</a> of
3520 floating point values. Both arguments must have identical types.</p>
3523 <p>The value produced is the floating point sum of the two operands.</p>
3527 <result> = fadd float 4.0, %var <i>; yields {float}:result = 4.0 + %var</i>
3532 <!-- _______________________________________________________________________ -->
3534 <a name="i_sub">'<tt>sub</tt>' Instruction</a>
3541 <result> = sub <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3542 <result> = sub nuw <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3543 <result> = sub nsw <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3544 <result> = sub nuw nsw <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3548 <p>The '<tt>sub</tt>' instruction returns the difference of its two
3551 <p>Note that the '<tt>sub</tt>' instruction is used to represent the
3552 '<tt>neg</tt>' instruction present in most other intermediate
3553 representations.</p>
3556 <p>The two arguments to the '<tt>sub</tt>' instruction must
3557 be <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of
3558 integer values. Both arguments must have identical types.</p>
3561 <p>The value produced is the integer difference of the two operands.</p>
3563 <p>If the difference has unsigned overflow, the result returned is the
3564 mathematical result modulo 2<sup>n</sup>, where n is the bit width of the
3567 <p>Because LLVM integers use a two's complement representation, this instruction
3568 is appropriate for both signed and unsigned integers.</p>
3570 <p><tt>nuw</tt> and <tt>nsw</tt> stand for "No Unsigned Wrap"
3571 and "No Signed Wrap", respectively. If the <tt>nuw</tt> and/or
3572 <tt>nsw</tt> keywords are present, the result value of the <tt>sub</tt>
3573 is a <a href="#trapvalues">trap value</a> if unsigned and/or signed overflow,
3574 respectively, occurs.</p>
3578 <result> = sub i32 4, %var <i>; yields {i32}:result = 4 - %var</i>
3579 <result> = sub i32 0, %val <i>; yields {i32}:result = -%var</i>
3584 <!-- _______________________________________________________________________ -->
3586 <a name="i_fsub">'<tt>fsub</tt>' Instruction</a>
3593 <result> = fsub <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3597 <p>The '<tt>fsub</tt>' instruction returns the difference of its two
3600 <p>Note that the '<tt>fsub</tt>' instruction is used to represent the
3601 '<tt>fneg</tt>' instruction present in most other intermediate
3602 representations.</p>
3605 <p>The two arguments to the '<tt>fsub</tt>' instruction must be
3606 <a href="#t_floating">floating point</a> or <a href="#t_vector">vector</a> of
3607 floating point values. Both arguments must have identical types.</p>
3610 <p>The value produced is the floating point difference of the two operands.</p>
3614 <result> = fsub float 4.0, %var <i>; yields {float}:result = 4.0 - %var</i>
3615 <result> = fsub float -0.0, %val <i>; yields {float}:result = -%var</i>
3620 <!-- _______________________________________________________________________ -->
3622 <a name="i_mul">'<tt>mul</tt>' Instruction</a>
3629 <result> = mul <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3630 <result> = mul nuw <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3631 <result> = mul nsw <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3632 <result> = mul nuw nsw <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3636 <p>The '<tt>mul</tt>' instruction returns the product of its two operands.</p>
3639 <p>The two arguments to the '<tt>mul</tt>' instruction must
3640 be <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of
3641 integer values. Both arguments must have identical types.</p>
3644 <p>The value produced is the integer product of the two operands.</p>
3646 <p>If the result of the multiplication has unsigned overflow, the result
3647 returned is the mathematical result modulo 2<sup>n</sup>, where n is the bit
3648 width of the result.</p>
3650 <p>Because LLVM integers use a two's complement representation, and the result
3651 is the same width as the operands, this instruction returns the correct
3652 result for both signed and unsigned integers. If a full product
3653 (e.g. <tt>i32</tt>x<tt>i32</tt>-><tt>i64</tt>) is needed, the operands should
3654 be sign-extended or zero-extended as appropriate to the width of the full
3657 <p><tt>nuw</tt> and <tt>nsw</tt> stand for "No Unsigned Wrap"
3658 and "No Signed Wrap", respectively. If the <tt>nuw</tt> and/or
3659 <tt>nsw</tt> keywords are present, the result value of the <tt>mul</tt>
3660 is a <a href="#trapvalues">trap value</a> if unsigned and/or signed overflow,
3661 respectively, occurs.</p>
3665 <result> = mul i32 4, %var <i>; yields {i32}:result = 4 * %var</i>
3670 <!-- _______________________________________________________________________ -->
3672 <a name="i_fmul">'<tt>fmul</tt>' Instruction</a>
3679 <result> = fmul <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3683 <p>The '<tt>fmul</tt>' instruction returns the product of its two operands.</p>
3686 <p>The two arguments to the '<tt>fmul</tt>' instruction must be
3687 <a href="#t_floating">floating point</a> or <a href="#t_vector">vector</a> of
3688 floating point values. Both arguments must have identical types.</p>
3691 <p>The value produced is the floating point product of the two operands.</p>
3695 <result> = fmul float 4.0, %var <i>; yields {float}:result = 4.0 * %var</i>
3700 <!-- _______________________________________________________________________ -->
3702 <a name="i_udiv">'<tt>udiv</tt>' Instruction</a>
3709 <result> = udiv <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3710 <result> = udiv exact <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3714 <p>The '<tt>udiv</tt>' instruction returns the quotient of its two operands.</p>
3717 <p>The two arguments to the '<tt>udiv</tt>' instruction must be
3718 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
3719 values. Both arguments must have identical types.</p>
3722 <p>The value produced is the unsigned integer quotient of the two operands.</p>
3724 <p>Note that unsigned integer division and signed integer division are distinct
3725 operations; for signed integer division, use '<tt>sdiv</tt>'.</p>
3727 <p>Division by zero leads to undefined behavior.</p>
3729 <p>If the <tt>exact</tt> keyword is present, the result value of the
3730 <tt>udiv</tt> is a <a href="#trapvalues">trap value</a> if %op1 is not a
3731 multiple of %op2 (as such, "((a udiv exact b) mul b) == a").</p>
3736 <result> = udiv i32 4, %var <i>; yields {i32}:result = 4 / %var</i>
3741 <!-- _______________________________________________________________________ -->
3743 <a name="i_sdiv">'<tt>sdiv</tt>' Instruction</a>
3750 <result> = sdiv <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3751 <result> = sdiv exact <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3755 <p>The '<tt>sdiv</tt>' instruction returns the quotient of its two operands.</p>
3758 <p>The two arguments to the '<tt>sdiv</tt>' instruction must be
3759 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
3760 values. Both arguments must have identical types.</p>
3763 <p>The value produced is the signed integer quotient of the two operands rounded
3766 <p>Note that signed integer division and unsigned integer division are distinct
3767 operations; for unsigned integer division, use '<tt>udiv</tt>'.</p>
3769 <p>Division by zero leads to undefined behavior. Overflow also leads to
3770 undefined behavior; this is a rare case, but can occur, for example, by doing
3771 a 32-bit division of -2147483648 by -1.</p>
3773 <p>If the <tt>exact</tt> keyword is present, the result value of the
3774 <tt>sdiv</tt> is a <a href="#trapvalues">trap value</a> if the result would
3779 <result> = sdiv i32 4, %var <i>; yields {i32}:result = 4 / %var</i>
3784 <!-- _______________________________________________________________________ -->
3786 <a name="i_fdiv">'<tt>fdiv</tt>' Instruction</a>
3793 <result> = fdiv <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3797 <p>The '<tt>fdiv</tt>' instruction returns the quotient of its two operands.</p>
3800 <p>The two arguments to the '<tt>fdiv</tt>' instruction must be
3801 <a href="#t_floating">floating point</a> or <a href="#t_vector">vector</a> of
3802 floating point values. Both arguments must have identical types.</p>
3805 <p>The value produced is the floating point quotient of the two operands.</p>
3809 <result> = fdiv float 4.0, %var <i>; yields {float}:result = 4.0 / %var</i>
3814 <!-- _______________________________________________________________________ -->
3816 <a name="i_urem">'<tt>urem</tt>' Instruction</a>
3823 <result> = urem <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3827 <p>The '<tt>urem</tt>' instruction returns the remainder from the unsigned
3828 division of its two arguments.</p>
3831 <p>The two arguments to the '<tt>urem</tt>' instruction must be
3832 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
3833 values. Both arguments must have identical types.</p>
3836 <p>This instruction returns the unsigned integer <i>remainder</i> of a division.
3837 This instruction always performs an unsigned division to get the
3840 <p>Note that unsigned integer remainder and signed integer remainder are
3841 distinct operations; for signed integer remainder, use '<tt>srem</tt>'.</p>
3843 <p>Taking the remainder of a division by zero leads to undefined behavior.</p>
3847 <result> = urem i32 4, %var <i>; yields {i32}:result = 4 % %var</i>
3852 <!-- _______________________________________________________________________ -->
3854 <a name="i_srem">'<tt>srem</tt>' Instruction</a>
3861 <result> = srem <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3865 <p>The '<tt>srem</tt>' instruction returns the remainder from the signed
3866 division of its two operands. This instruction can also take
3867 <a href="#t_vector">vector</a> versions of the values in which case the
3868 elements must be integers.</p>
3871 <p>The two arguments to the '<tt>srem</tt>' instruction must be
3872 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
3873 values. Both arguments must have identical types.</p>
3876 <p>This instruction returns the <i>remainder</i> of a division (where the result
3877 is either zero or has the same sign as the dividend, <tt>op1</tt>), not the
3878 <i>modulo</i> operator (where the result is either zero or has the same sign
3879 as the divisor, <tt>op2</tt>) of a value.
3880 For more information about the difference,
3881 see <a href="http://mathforum.org/dr.math/problems/anne.4.28.99.html">The
3882 Math Forum</a>. For a table of how this is implemented in various languages,
3883 please see <a href="http://en.wikipedia.org/wiki/Modulo_operation">
3884 Wikipedia: modulo operation</a>.</p>
3886 <p>Note that signed integer remainder and unsigned integer remainder are
3887 distinct operations; for unsigned integer remainder, use '<tt>urem</tt>'.</p>
3889 <p>Taking the remainder of a division by zero leads to undefined behavior.
3890 Overflow also leads to undefined behavior; this is a rare case, but can
3891 occur, for example, by taking the remainder of a 32-bit division of
3892 -2147483648 by -1. (The remainder doesn't actually overflow, but this rule
3893 lets srem be implemented using instructions that return both the result of
3894 the division and the remainder.)</p>
3898 <result> = srem i32 4, %var <i>; yields {i32}:result = 4 % %var</i>
3903 <!-- _______________________________________________________________________ -->
3905 <a name="i_frem">'<tt>frem</tt>' Instruction</a>
3912 <result> = frem <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3916 <p>The '<tt>frem</tt>' instruction returns the remainder from the division of
3917 its two operands.</p>
3920 <p>The two arguments to the '<tt>frem</tt>' instruction must be
3921 <a href="#t_floating">floating point</a> or <a href="#t_vector">vector</a> of
3922 floating point values. Both arguments must have identical types.</p>
3925 <p>This instruction returns the <i>remainder</i> of a division. The remainder
3926 has the same sign as the dividend.</p>
3930 <result> = frem float 4.0, %var <i>; yields {float}:result = 4.0 % %var</i>
3937 <!-- ======================================================================= -->
3939 <a name="bitwiseops">Bitwise Binary Operations</a>
3944 <p>Bitwise binary operators are used to do various forms of bit-twiddling in a
3945 program. They are generally very efficient instructions and can commonly be
3946 strength reduced from other instructions. They require two operands of the
3947 same type, execute an operation on them, and produce a single value. The
3948 resulting value is the same type as its operands.</p>
3950 <!-- _______________________________________________________________________ -->
3952 <a name="i_shl">'<tt>shl</tt>' Instruction</a>
3959 <result> = shl <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3960 <result> = shl nuw <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3961 <result> = shl nsw <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3962 <result> = shl nuw nsw <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3966 <p>The '<tt>shl</tt>' instruction returns the first operand shifted to the left
3967 a specified number of bits.</p>
3970 <p>Both arguments to the '<tt>shl</tt>' instruction must be the
3971 same <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of
3972 integer type. '<tt>op2</tt>' is treated as an unsigned value.</p>
3975 <p>The value produced is <tt>op1</tt> * 2<sup><tt>op2</tt></sup> mod
3976 2<sup>n</sup>, where <tt>n</tt> is the width of the result. If <tt>op2</tt>
3977 is (statically or dynamically) negative or equal to or larger than the number
3978 of bits in <tt>op1</tt>, the result is undefined. If the arguments are
3979 vectors, each vector element of <tt>op1</tt> is shifted by the corresponding
3980 shift amount in <tt>op2</tt>.</p>
3982 <p>If the <tt>nuw</tt> keyword is present, then the shift produces a
3983 <a href="#trapvalues">trap value</a> if it shifts out any non-zero bits. If
3984 the <tt>nsw</tt> keyword is present, then the shift produces a
3985 <a href="#trapvalues">trap value</a> if it shifts out any bits that disagree
3986 with the resultant sign bit. As such, NUW/NSW have the same semantics as
3987 they would if the shift were expressed as a mul instruction with the same
3988 nsw/nuw bits in (mul %op1, (shl 1, %op2)).</p>
3992 <result> = shl i32 4, %var <i>; yields {i32}: 4 << %var</i>
3993 <result> = shl i32 4, 2 <i>; yields {i32}: 16</i>
3994 <result> = shl i32 1, 10 <i>; yields {i32}: 1024</i>
3995 <result> = shl i32 1, 32 <i>; undefined</i>
3996 <result> = shl <2 x i32> < i32 1, i32 1>, < i32 1, i32 2> <i>; yields: result=<2 x i32> < i32 2, i32 4></i>
4001 <!-- _______________________________________________________________________ -->
4003 <a name="i_lshr">'<tt>lshr</tt>' Instruction</a>
4010 <result> = lshr <ty> <op1>, <op2> <i>; yields {ty}:result</i>
4011 <result> = lshr exact <ty> <op1>, <op2> <i>; yields {ty}:result</i>
4015 <p>The '<tt>lshr</tt>' instruction (logical shift right) returns the first
4016 operand shifted to the right a specified number of bits with zero fill.</p>
4019 <p>Both arguments to the '<tt>lshr</tt>' instruction must be the same
4020 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
4021 type. '<tt>op2</tt>' is treated as an unsigned value.</p>
4024 <p>This instruction always performs a logical shift right operation. The most
4025 significant bits of the result will be filled with zero bits after the shift.
4026 If <tt>op2</tt> is (statically or dynamically) equal to or larger than the
4027 number of bits in <tt>op1</tt>, the result is undefined. If the arguments are
4028 vectors, each vector element of <tt>op1</tt> is shifted by the corresponding
4029 shift amount in <tt>op2</tt>.</p>
4031 <p>If the <tt>exact</tt> keyword is present, the result value of the
4032 <tt>lshr</tt> is a <a href="#trapvalues">trap value</a> if any of the bits
4033 shifted out are non-zero.</p>
4038 <result> = lshr i32 4, 1 <i>; yields {i32}:result = 2</i>
4039 <result> = lshr i32 4, 2 <i>; yields {i32}:result = 1</i>
4040 <result> = lshr i8 4, 3 <i>; yields {i8}:result = 0</i>
4041 <result> = lshr i8 -2, 1 <i>; yields {i8}:result = 0x7FFFFFFF </i>
4042 <result> = lshr i32 1, 32 <i>; undefined</i>
4043 <result> = lshr <2 x i32> < i32 -2, i32 4>, < i32 1, i32 2> <i>; yields: result=<2 x i32> < i32 0x7FFFFFFF, i32 1></i>
4048 <!-- _______________________________________________________________________ -->
4050 <a name="i_ashr">'<tt>ashr</tt>' Instruction</a>
4057 <result> = ashr <ty> <op1>, <op2> <i>; yields {ty}:result</i>
4058 <result> = ashr exact <ty> <op1>, <op2> <i>; yields {ty}:result</i>
4062 <p>The '<tt>ashr</tt>' instruction (arithmetic shift right) returns the first
4063 operand shifted to the right a specified number of bits with sign
4067 <p>Both arguments to the '<tt>ashr</tt>' instruction must be the same
4068 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
4069 type. '<tt>op2</tt>' is treated as an unsigned value.</p>
4072 <p>This instruction always performs an arithmetic shift right operation, The
4073 most significant bits of the result will be filled with the sign bit
4074 of <tt>op1</tt>. If <tt>op2</tt> is (statically or dynamically) equal to or
4075 larger than the number of bits in <tt>op1</tt>, the result is undefined. If
4076 the arguments are vectors, each vector element of <tt>op1</tt> is shifted by
4077 the corresponding shift amount in <tt>op2</tt>.</p>
4079 <p>If the <tt>exact</tt> keyword is present, the result value of the
4080 <tt>ashr</tt> is a <a href="#trapvalues">trap value</a> if any of the bits
4081 shifted out are non-zero.</p>
4085 <result> = ashr i32 4, 1 <i>; yields {i32}:result = 2</i>
4086 <result> = ashr i32 4, 2 <i>; yields {i32}:result = 1</i>
4087 <result> = ashr i8 4, 3 <i>; yields {i8}:result = 0</i>
4088 <result> = ashr i8 -2, 1 <i>; yields {i8}:result = -1</i>
4089 <result> = ashr i32 1, 32 <i>; undefined</i>
4090 <result> = ashr <2 x i32> < i32 -2, i32 4>, < i32 1, i32 3> <i>; yields: result=<2 x i32> < i32 -1, i32 0></i>
4095 <!-- _______________________________________________________________________ -->
4097 <a name="i_and">'<tt>and</tt>' Instruction</a>
4104 <result> = and <ty> <op1>, <op2> <i>; yields {ty}:result</i>
4108 <p>The '<tt>and</tt>' instruction returns the bitwise logical and of its two
4112 <p>The two arguments to the '<tt>and</tt>' instruction must be
4113 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
4114 values. Both arguments must have identical types.</p>
4117 <p>The truth table used for the '<tt>and</tt>' instruction is:</p>
4119 <table border="1" cellspacing="0" cellpadding="4">
4151 <result> = and i32 4, %var <i>; yields {i32}:result = 4 & %var</i>
4152 <result> = and i32 15, 40 <i>; yields {i32}:result = 8</i>
4153 <result> = and i32 4, 8 <i>; yields {i32}:result = 0</i>
4156 <!-- _______________________________________________________________________ -->
4158 <a name="i_or">'<tt>or</tt>' Instruction</a>
4165 <result> = or <ty> <op1>, <op2> <i>; yields {ty}:result</i>
4169 <p>The '<tt>or</tt>' instruction returns the bitwise logical inclusive or of its
4173 <p>The two arguments to the '<tt>or</tt>' instruction must be
4174 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
4175 values. Both arguments must have identical types.</p>
4178 <p>The truth table used for the '<tt>or</tt>' instruction is:</p>
4180 <table border="1" cellspacing="0" cellpadding="4">
4212 <result> = or i32 4, %var <i>; yields {i32}:result = 4 | %var</i>
4213 <result> = or i32 15, 40 <i>; yields {i32}:result = 47</i>
4214 <result> = or i32 4, 8 <i>; yields {i32}:result = 12</i>
4219 <!-- _______________________________________________________________________ -->
4221 <a name="i_xor">'<tt>xor</tt>' Instruction</a>
4228 <result> = xor <ty> <op1>, <op2> <i>; yields {ty}:result</i>
4232 <p>The '<tt>xor</tt>' instruction returns the bitwise logical exclusive or of
4233 its two operands. The <tt>xor</tt> is used to implement the "one's
4234 complement" operation, which is the "~" operator in C.</p>
4237 <p>The two arguments to the '<tt>xor</tt>' instruction must be
4238 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
4239 values. Both arguments must have identical types.</p>
4242 <p>The truth table used for the '<tt>xor</tt>' instruction is:</p>
4244 <table border="1" cellspacing="0" cellpadding="4">
4276 <result> = xor i32 4, %var <i>; yields {i32}:result = 4 ^ %var</i>
4277 <result> = xor i32 15, 40 <i>; yields {i32}:result = 39</i>
4278 <result> = xor i32 4, 8 <i>; yields {i32}:result = 12</i>
4279 <result> = xor i32 %V, -1 <i>; yields {i32}:result = ~%V</i>
4286 <!-- ======================================================================= -->
4288 <a name="vectorops">Vector Operations</a>
4293 <p>LLVM supports several instructions to represent vector operations in a
4294 target-independent manner. These instructions cover the element-access and
4295 vector-specific operations needed to process vectors effectively. While LLVM
4296 does directly support these vector operations, many sophisticated algorithms
4297 will want to use target-specific intrinsics to take full advantage of a
4298 specific target.</p>
4300 <!-- _______________________________________________________________________ -->
4302 <a name="i_extractelement">'<tt>extractelement</tt>' Instruction</a>
4309 <result> = extractelement <n x <ty>> <val>, i32 <idx> <i>; yields <ty></i>
4313 <p>The '<tt>extractelement</tt>' instruction extracts a single scalar element
4314 from a vector at a specified index.</p>
4318 <p>The first operand of an '<tt>extractelement</tt>' instruction is a value
4319 of <a href="#t_vector">vector</a> type. The second operand is an index
4320 indicating the position from which to extract the element. The index may be
4324 <p>The result is a scalar of the same type as the element type of
4325 <tt>val</tt>. Its value is the value at position <tt>idx</tt> of
4326 <tt>val</tt>. If <tt>idx</tt> exceeds the length of <tt>val</tt>, the
4327 results are undefined.</p>
4331 <result> = extractelement <4 x i32> %vec, i32 0 <i>; yields i32</i>
4336 <!-- _______________________________________________________________________ -->
4338 <a name="i_insertelement">'<tt>insertelement</tt>' Instruction</a>
4345 <result> = insertelement <n x <ty>> <val>, <ty> <elt>, i32 <idx> <i>; yields <n x <ty>></i>
4349 <p>The '<tt>insertelement</tt>' instruction inserts a scalar element into a
4350 vector at a specified index.</p>
4353 <p>The first operand of an '<tt>insertelement</tt>' instruction is a value
4354 of <a href="#t_vector">vector</a> type. The second operand is a scalar value
4355 whose type must equal the element type of the first operand. The third
4356 operand is an index indicating the position at which to insert the value.
4357 The index may be a variable.</p>
4360 <p>The result is a vector of the same type as <tt>val</tt>. Its element values
4361 are those of <tt>val</tt> except at position <tt>idx</tt>, where it gets the
4362 value <tt>elt</tt>. If <tt>idx</tt> exceeds the length of <tt>val</tt>, the
4363 results are undefined.</p>
4367 <result> = insertelement <4 x i32> %vec, i32 1, i32 0 <i>; yields <4 x i32></i>
4372 <!-- _______________________________________________________________________ -->
4374 <a name="i_shufflevector">'<tt>shufflevector</tt>' Instruction</a>
4381 <result> = shufflevector <n x <ty>> <v1>, <n x <ty>> <v2>, <m x i32> <mask> <i>; yields <m x <ty>></i>
4385 <p>The '<tt>shufflevector</tt>' instruction constructs a permutation of elements
4386 from two input vectors, returning a vector with the same element type as the
4387 input and length that is the same as the shuffle mask.</p>
4390 <p>The first two operands of a '<tt>shufflevector</tt>' instruction are vectors
4391 with types that match each other. The third argument is a shuffle mask whose
4392 element type is always 'i32'. The result of the instruction is a vector
4393 whose length is the same as the shuffle mask and whose element type is the
4394 same as the element type of the first two operands.</p>
4396 <p>The shuffle mask operand is required to be a constant vector with either
4397 constant integer or undef values.</p>
4400 <p>The elements of the two input vectors are numbered from left to right across
4401 both of the vectors. The shuffle mask operand specifies, for each element of
4402 the result vector, which element of the two input vectors the result element
4403 gets. The element selector may be undef (meaning "don't care") and the
4404 second operand may be undef if performing a shuffle from only one vector.</p>
4408 <result> = shufflevector <4 x i32> %v1, <4 x i32> %v2,
4409 <4 x i32> <i32 0, i32 4, i32 1, i32 5> <i>; yields <4 x i32></i>
4410 <result> = shufflevector <4 x i32> %v1, <4 x i32> undef,
4411 <4 x i32> <i32 0, i32 1, i32 2, i32 3> <i>; yields <4 x i32></i> - Identity shuffle.
4412 <result> = shufflevector <8 x i32> %v1, <8 x i32> undef,
4413 <4 x i32> <i32 0, i32 1, i32 2, i32 3> <i>; yields <4 x i32></i>
4414 <result> = shufflevector <4 x i32> %v1, <4 x i32> %v2,
4415 <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>
4422 <!-- ======================================================================= -->
4424 <a name="aggregateops">Aggregate Operations</a>
4429 <p>LLVM supports several instructions for working with
4430 <a href="#t_aggregate">aggregate</a> values.</p>
4432 <!-- _______________________________________________________________________ -->
4434 <a name="i_extractvalue">'<tt>extractvalue</tt>' Instruction</a>
4441 <result> = extractvalue <aggregate type> <val>, <idx>{, <idx>}*
4445 <p>The '<tt>extractvalue</tt>' instruction extracts the value of a member field
4446 from an <a href="#t_aggregate">aggregate</a> value.</p>
4449 <p>The first operand of an '<tt>extractvalue</tt>' instruction is a value
4450 of <a href="#t_struct">struct</a> or
4451 <a href="#t_array">array</a> type. The operands are constant indices to
4452 specify which value to extract in a similar manner as indices in a
4453 '<tt><a href="#i_getelementptr">getelementptr</a></tt>' instruction.</p>
4454 <p>The major differences to <tt>getelementptr</tt> indexing are:</p>
4456 <li>Since the value being indexed is not a pointer, the first index is
4457 omitted and assumed to be zero.</li>
4458 <li>At least one index must be specified.</li>
4459 <li>Not only struct indices but also array indices must be in
4464 <p>The result is the value at the position in the aggregate specified by the
4469 <result> = extractvalue {i32, float} %agg, 0 <i>; yields i32</i>
4474 <!-- _______________________________________________________________________ -->
4476 <a name="i_insertvalue">'<tt>insertvalue</tt>' Instruction</a>
4483 <result> = insertvalue <aggregate type> <val>, <ty> <elt>, <idx>{, <idx>}* <i>; yields <aggregate type></i>
4487 <p>The '<tt>insertvalue</tt>' instruction inserts a value into a member field
4488 in an <a href="#t_aggregate">aggregate</a> value.</p>
4491 <p>The first operand of an '<tt>insertvalue</tt>' instruction is a value
4492 of <a href="#t_struct">struct</a> or
4493 <a href="#t_array">array</a> type. The second operand is a first-class
4494 value to insert. The following operands are constant indices indicating
4495 the position at which to insert the value in a similar manner as indices in a
4496 '<tt><a href="#i_extractvalue">extractvalue</a></tt>' instruction. The
4497 value to insert must have the same type as the value identified by the
4501 <p>The result is an aggregate of the same type as <tt>val</tt>. Its value is
4502 that of <tt>val</tt> except that the value at the position specified by the
4503 indices is that of <tt>elt</tt>.</p>
4507 %agg1 = insertvalue {i32, float} undef, i32 1, 0 <i>; yields {i32 1, float undef}</i>
4508 %agg2 = insertvalue {i32, float} %agg1, float %val, 1 <i>; yields {i32 1, float %val}</i>
4509 %agg3 = insertvalue {i32, {float}} %agg1, float %val, 1, 0 <i>; yields {i32 1, float %val}</i>
4516 <!-- ======================================================================= -->
4518 <a name="memoryops">Memory Access and Addressing Operations</a>
4523 <p>A key design point of an SSA-based representation is how it represents
4524 memory. In LLVM, no memory locations are in SSA form, which makes things
4525 very simple. This section describes how to read, write, and allocate
4528 <!-- _______________________________________________________________________ -->
4530 <a name="i_alloca">'<tt>alloca</tt>' Instruction</a>
4537 <result> = alloca <type>[, <ty> <NumElements>][, align <alignment>] <i>; yields {type*}:result</i>
4541 <p>The '<tt>alloca</tt>' instruction allocates memory on the stack frame of the
4542 currently executing function, to be automatically released when this function
4543 returns to its caller. The object is always allocated in the generic address
4544 space (address space zero).</p>
4547 <p>The '<tt>alloca</tt>' instruction
4548 allocates <tt>sizeof(<type>)*NumElements</tt> bytes of memory on the
4549 runtime stack, returning a pointer of the appropriate type to the program.
4550 If "NumElements" is specified, it is the number of elements allocated,
4551 otherwise "NumElements" is defaulted to be one. If a constant alignment is
4552 specified, the value result of the allocation is guaranteed to be aligned to
4553 at least that boundary. If not specified, or if zero, the target can choose
4554 to align the allocation on any convenient boundary compatible with the
4557 <p>'<tt>type</tt>' may be any sized type.</p>
4560 <p>Memory is allocated; a pointer is returned. The operation is undefined if
4561 there is insufficient stack space for the allocation. '<tt>alloca</tt>'d
4562 memory is automatically released when the function returns. The
4563 '<tt>alloca</tt>' instruction is commonly used to represent automatic
4564 variables that must have an address available. When the function returns
4565 (either with the <tt><a href="#i_ret">ret</a></tt>
4566 or <tt><a href="#i_unwind">unwind</a></tt> instructions), the memory is
4567 reclaimed. Allocating zero bytes is legal, but the result is undefined.</p>
4571 %ptr = alloca i32 <i>; yields {i32*}:ptr</i>
4572 %ptr = alloca i32, i32 4 <i>; yields {i32*}:ptr</i>
4573 %ptr = alloca i32, i32 4, align 1024 <i>; yields {i32*}:ptr</i>
4574 %ptr = alloca i32, align 1024 <i>; yields {i32*}:ptr</i>
4579 <!-- _______________________________________________________________________ -->
4581 <a name="i_load">'<tt>load</tt>' Instruction</a>
4588 <result> = load [volatile] <ty>* <pointer>[, align <alignment>][, !nontemporal !<index>]
4589 <result> = load atomic [volatile] <ty>* <pointer> [singlethread] <ordering>, align <alignment>
4590 !<index> = !{ i32 1 }
4594 <p>The '<tt>load</tt>' instruction is used to read from memory.</p>
4597 <p>The argument to the '<tt>load</tt>' instruction specifies the memory address
4598 from which to load. The pointer must point to
4599 a <a href="#t_firstclass">first class</a> type. If the <tt>load</tt> is
4600 marked as <tt>volatile</tt>, then the optimizer is not allowed to modify the
4601 number or order of execution of this <tt>load</tt> with other <a
4602 href="#volatile">volatile operations</a>.</p>
4604 <p>If the <code>load</code> is marked as <code>atomic</code>, it takes an extra
4605 <a href="#ordering">ordering</a> and optional <code>singlethread</code>
4606 argument. The <code>release</code> and <code>acq_rel</code> orderings are
4607 not valid on <code>load</code> instructions. Atomic loads produce <a
4608 href="#memorymodel">defined</a> results when they may see multiple atomic
4609 stores. The type of the pointee must be an integer type whose bit width
4610 is a power of two greater than or equal to eight and less than or equal
4611 to a target-specific size limit. <code>align</code> must be explicitly
4612 specified on atomic loads, and the load has undefined behavior if the
4613 alignment is not set to a value which is at least the size in bytes of
4614 the pointee. <code>!nontemporal</code> does not have any defined semantics
4615 for atomic loads.</p>
4617 <p>The optional constant <tt>align</tt> argument specifies the alignment of the
4618 operation (that is, the alignment of the memory address). A value of 0 or an
4619 omitted <tt>align</tt> argument means that the operation has the preferential
4620 alignment for the target. It is the responsibility of the code emitter to
4621 ensure that the alignment information is correct. Overestimating the
4622 alignment results in undefined behavior. Underestimating the alignment may
4623 produce less efficient code. An alignment of 1 is always safe.</p>
4625 <p>The optional <tt>!nontemporal</tt> metadata must reference a single
4626 metatadata name <index> corresponding to a metadata node with
4627 one <tt>i32</tt> entry of value 1. The existence of
4628 the <tt>!nontemporal</tt> metatadata on the instruction tells the optimizer
4629 and code generator that this load is not expected to be reused in the cache.
4630 The code generator may select special instructions to save cache bandwidth,
4631 such as the <tt>MOVNT</tt> instruction on x86.</p>
4634 <p>The location of memory pointed to is loaded. If the value being loaded is of
4635 scalar type then the number of bytes read does not exceed the minimum number
4636 of bytes needed to hold all bits of the type. For example, loading an
4637 <tt>i24</tt> reads at most three bytes. When loading a value of a type like
4638 <tt>i20</tt> with a size that is not an integral number of bytes, the result
4639 is undefined if the value was not originally written using a store of the
4644 %ptr = <a href="#i_alloca">alloca</a> i32 <i>; yields {i32*}:ptr</i>
4645 <a href="#i_store">store</a> i32 3, i32* %ptr <i>; yields {void}</i>
4646 %val = load i32* %ptr <i>; yields {i32}:val = i32 3</i>
4651 <!-- _______________________________________________________________________ -->
4653 <a name="i_store">'<tt>store</tt>' Instruction</a>
4660 store [volatile] <ty> <value>, <ty>* <pointer>[, align <alignment>][, !nontemporal !<index>] <i>; yields {void}</i>
4661 store atomic [volatile] <ty> <value>, <ty>* <pointer> [singlethread] <ordering>, align <alignment> <i>; yields {void}</i>
4665 <p>The '<tt>store</tt>' instruction is used to write to memory.</p>
4668 <p>There are two arguments to the '<tt>store</tt>' instruction: a value to store
4669 and an address at which to store it. The type of the
4670 '<tt><pointer></tt>' operand must be a pointer to
4671 the <a href="#t_firstclass">first class</a> type of the
4672 '<tt><value></tt>' operand. If the <tt>store</tt> is marked as
4673 <tt>volatile</tt>, then the optimizer is not allowed to modify the number or
4674 order of execution of this <tt>store</tt> with other <a
4675 href="#volatile">volatile operations</a>.</p>
4677 <p>If the <code>store</code> is marked as <code>atomic</code>, it takes an extra
4678 <a href="#ordering">ordering</a> and optional <code>singlethread</code>
4679 argument. The <code>acquire</code> and <code>acq_rel</code> orderings aren't
4680 valid on <code>store</code> instructions. Atomic loads produce <a
4681 href="#memorymodel">defined</a> results when they may see multiple atomic
4682 stores. The type of the pointee must be an integer type whose bit width
4683 is a power of two greater than or equal to eight and less than or equal
4684 to a target-specific size limit. <code>align</code> must be explicitly
4685 specified on atomic stores, and the store has undefined behavior if the
4686 alignment is not set to a value which is at least the size in bytes of
4687 the pointee. <code>!nontemporal</code> does not have any defined semantics
4688 for atomic stores.</p>
4690 <p>The optional constant "align" argument specifies the alignment of the
4691 operation (that is, the alignment of the memory address). A value of 0 or an
4692 omitted "align" argument means that the operation has the preferential
4693 alignment for the target. It is the responsibility of the code emitter to
4694 ensure that the alignment information is correct. Overestimating the
4695 alignment results in an undefined behavior. Underestimating the alignment may
4696 produce less efficient code. An alignment of 1 is always safe.</p>
4698 <p>The optional !nontemporal metadata must reference a single metatadata
4699 name <index> corresponding to a metadata node with one i32 entry of
4700 value 1. The existence of the !nontemporal metatadata on the
4701 instruction tells the optimizer and code generator that this load is
4702 not expected to be reused in the cache. The code generator may
4703 select special instructions to save cache bandwidth, such as the
4704 MOVNT instruction on x86.</p>
4708 <p>The contents of memory are updated to contain '<tt><value></tt>' at the
4709 location specified by the '<tt><pointer></tt>' operand. If
4710 '<tt><value></tt>' is of scalar type then the number of bytes written
4711 does not exceed the minimum number of bytes needed to hold all bits of the
4712 type. For example, storing an <tt>i24</tt> writes at most three bytes. When
4713 writing a value of a type like <tt>i20</tt> with a size that is not an
4714 integral number of bytes, it is unspecified what happens to the extra bits
4715 that do not belong to the type, but they will typically be overwritten.</p>
4719 %ptr = <a href="#i_alloca">alloca</a> i32 <i>; yields {i32*}:ptr</i>
4720 store i32 3, i32* %ptr <i>; yields {void}</i>
4721 %val = <a href="#i_load">load</a> i32* %ptr <i>; yields {i32}:val = i32 3</i>
4726 <!-- _______________________________________________________________________ -->
4728 <a name="i_fence">'<tt>fence</tt>' Instruction</a>
4735 fence [singlethread] <ordering> <i>; yields {void}</i>
4739 <p>The '<tt>fence</tt>' instruction is used to introduce happens-before edges
4740 between operations.</p>
4742 <h5>Arguments:</h5> <p>'<code>fence</code>' instructions take an <a
4743 href="#ordering">ordering</a> argument which defines what
4744 <i>synchronizes-with</i> edges they add. They can only be given
4745 <code>acquire</code>, <code>release</code>, <code>acq_rel</code>, and
4746 <code>seq_cst</code> orderings.</p>
4749 <p>A fence <var>A</var> which has (at least) <code>release</code> ordering
4750 semantics <i>synchronizes with</i> a fence <var>B</var> with (at least)
4751 <code>acquire</code> ordering semantics if and only if there exist atomic
4752 operations <var>X</var> and <var>Y</var>, both operating on some atomic object
4753 <var>M</var>, such that <var>A</var> is sequenced before <var>X</var>,
4754 <var>X</var> modifies <var>M</var> (either directly or through some side effect
4755 of a sequence headed by <var>X</var>), <var>Y</var> is sequenced before
4756 <var>B</var>, and <var>Y</var> observes <var>M</var>. This provides a
4757 <i>happens-before</i> dependency between <var>A</var> and <var>B</var>. Rather
4758 than an explicit <code>fence</code>, one (but not both) of the atomic operations
4759 <var>X</var> or <var>Y</var> might provide a <code>release</code> or
4760 <code>acquire</code> (resp.) ordering constraint and still
4761 <i>synchronize-with</i> the explicit <code>fence</code> and establish the
4762 <i>happens-before</i> edge.</p>
4764 <p>A <code>fence</code> which has <code>seq_cst</code> ordering, in addition to
4765 having both <code>acquire</code> and <code>release</code> semantics specified
4766 above, participates in the global program order of other <code>seq_cst</code>
4767 operations and/or fences.</p>
4769 <p>The optional "<a href="#singlethread"><code>singlethread</code></a>" argument
4770 specifies that the fence only synchronizes with other fences in the same
4771 thread. (This is useful for interacting with signal handlers.)</p>
4775 fence acquire <i>; yields {void}</i>
4776 fence singlethread seq_cst <i>; yields {void}</i>
4781 <!-- _______________________________________________________________________ -->
4783 <a name="i_cmpxchg">'<tt>cmpxchg</tt>' Instruction</a>
4790 cmpxchg [volatile] <ty>* <pointer>, <ty> <cmp>, <ty> <new> [singlethread] <ordering> <i>; yields {ty}</i>
4794 <p>The '<tt>cmpxchg</tt>' instruction is used to atomically modify memory.
4795 It loads a value in memory and compares it to a given value. If they are
4796 equal, it stores a new value into the memory.</p>
4799 <p>There are three arguments to the '<code>cmpxchg</code>' instruction: an
4800 address to operate on, a value to compare to the value currently be at that
4801 address, and a new value to place at that address if the compared values are
4802 equal. The type of '<var><cmp></var>' must be an integer type whose
4803 bit width is a power of two greater than or equal to eight and less than
4804 or equal to a target-specific size limit. '<var><cmp></var>' and
4805 '<var><new></var>' must have the same type, and the type of
4806 '<var><pointer></var>' must be a pointer to that type. If the
4807 <code>cmpxchg</code> is marked as <code>volatile</code>, then the
4808 optimizer is not allowed to modify the number or order of execution
4809 of this <code>cmpxchg</code> with other <a href="#volatile">volatile
4812 <!-- FIXME: Extend allowed types. -->
4814 <p>The <a href="#ordering"><var>ordering</var></a> argument specifies how this
4815 <code>cmpxchg</code> synchronizes with other atomic operations.</p>
4817 <p>The optional "<code>singlethread</code>" argument declares that the
4818 <code>cmpxchg</code> is only atomic with respect to code (usually signal
4819 handlers) running in the same thread as the <code>cmpxchg</code>. Otherwise the
4820 cmpxchg is atomic with respect to all other code in the system.</p>
4822 <p>The pointer passed into cmpxchg must have alignment greater than or equal to
4823 the size in memory of the operand.
4826 <p>The contents of memory at the location specified by the
4827 '<tt><pointer></tt>' operand is read and compared to
4828 '<tt><cmp></tt>'; if the read value is the equal,
4829 '<tt><new></tt>' is written. The original value at the location
4832 <p>A successful <code>cmpxchg</code> is a read-modify-write instruction for the
4833 purpose of identifying <a href="#release_sequence">release sequences</a>. A
4834 failed <code>cmpxchg</code> is equivalent to an atomic load with an ordering
4835 parameter determined by dropping any <code>release</code> part of the
4836 <code>cmpxchg</code>'s ordering.</p>
4839 FIXME: Is compare_exchange_weak() necessary? (Consider after we've done
4840 optimization work on ARM.)
4842 FIXME: Is a weaker ordering constraint on failure helpful in practice?
4848 %orig = atomic <a href="#i_load">load</a> i32* %ptr unordered <i>; yields {i32}</i>
4849 <a href="#i_br">br</a> label %loop
4852 %cmp = <a href="#i_phi">phi</a> i32 [ %orig, %entry ], [%old, %loop]
4853 %squared = <a href="#i_mul">mul</a> i32 %cmp, %cmp
4854 %old = cmpxchg i32* %ptr, i32 %cmp, i32 %squared <i>; yields {i32}</i>
4855 %success = <a href="#i_icmp">icmp</a> eq i32 %cmp, %old
4856 <a href="#i_br">br</a> i1 %success, label %done, label %loop
4864 <!-- _______________________________________________________________________ -->
4866 <a name="i_atomicrmw">'<tt>atomicrmw</tt>' Instruction</a>
4873 atomicrmw [volatile] <operation> <ty>* <pointer>, <ty> <value> [singlethread] <ordering> <i>; yields {ty}</i>
4877 <p>The '<tt>atomicrmw</tt>' instruction is used to atomically modify memory.</p>
4880 <p>There are three arguments to the '<code>atomicrmw</code>' instruction: an
4881 operation to apply, an address whose value to modify, an argument to the
4882 operation. The operation must be one of the following keywords:</p>
4897 <p>The type of '<var><value></var>' must be an integer type whose
4898 bit width is a power of two greater than or equal to eight and less than
4899 or equal to a target-specific size limit. The type of the
4900 '<code><pointer></code>' operand must be a pointer to that type.
4901 If the <code>atomicrmw</code> is marked as <code>volatile</code>, then the
4902 optimizer is not allowed to modify the number or order of execution of this
4903 <code>atomicrmw</code> with other <a href="#volatile">volatile
4906 <!-- FIXME: Extend allowed types. -->
4909 <p>The contents of memory at the location specified by the
4910 '<tt><pointer></tt>' operand are atomically read, modified, and written
4911 back. The original value at the location is returned. The modification is
4912 specified by the <var>operation</var> argument:</p>
4915 <li>xchg: <code>*ptr = val</code></li>
4916 <li>add: <code>*ptr = *ptr + val</code></li>
4917 <li>sub: <code>*ptr = *ptr - val</code></li>
4918 <li>and: <code>*ptr = *ptr & val</code></li>
4919 <li>nand: <code>*ptr = ~(*ptr & val)</code></li>
4920 <li>or: <code>*ptr = *ptr | val</code></li>
4921 <li>xor: <code>*ptr = *ptr ^ val</code></li>
4922 <li>max: <code>*ptr = *ptr > val ? *ptr : val</code> (using a signed comparison)</li>
4923 <li>min: <code>*ptr = *ptr < val ? *ptr : val</code> (using a signed comparison)</li>
4924 <li>umax: <code>*ptr = *ptr > val ? *ptr : val</code> (using an unsigned comparison)</li>
4925 <li>umin: <code>*ptr = *ptr < val ? *ptr : val</code> (using an unsigned comparison)</li>
4930 %old = atomicrmw add i32* %ptr, i32 1 acquire <i>; yields {i32}</i>
4935 <!-- _______________________________________________________________________ -->
4937 <a name="i_getelementptr">'<tt>getelementptr</tt>' Instruction</a>
4944 <result> = getelementptr <pty>* <ptrval>{, <ty> <idx>}*
4945 <result> = getelementptr inbounds <pty>* <ptrval>{, <ty> <idx>}*
4949 <p>The '<tt>getelementptr</tt>' instruction is used to get the address of a
4950 subelement of an <a href="#t_aggregate">aggregate</a> data structure.
4951 It performs address calculation only and does not access memory.</p>
4954 <p>The first argument is always a pointer, and forms the basis of the
4955 calculation. The remaining arguments are indices that indicate which of the
4956 elements of the aggregate object are indexed. The interpretation of each
4957 index is dependent on the type being indexed into. The first index always
4958 indexes the pointer value given as the first argument, the second index
4959 indexes a value of the type pointed to (not necessarily the value directly
4960 pointed to, since the first index can be non-zero), etc. The first type
4961 indexed into must be a pointer value, subsequent types can be arrays,
4962 vectors, and structs. Note that subsequent types being indexed into
4963 can never be pointers, since that would require loading the pointer before
4964 continuing calculation.</p>
4966 <p>The type of each index argument depends on the type it is indexing into.
4967 When indexing into a (optionally packed) structure, only <tt>i32</tt>
4968 integer <b>constants</b> are allowed. When indexing into an array, pointer
4969 or vector, integers of any width are allowed, and they are not required to be
4970 constant. These integers are treated as signed values where relevant.</p>
4972 <p>For example, let's consider a C code fragment and how it gets compiled to
4975 <pre class="doc_code">
4987 int *foo(struct ST *s) {
4988 return &s[1].Z.B[5][13];
4992 <p>The LLVM code generated by the GCC frontend is:</p>
4994 <pre class="doc_code">
4995 %RT = <a href="#namedtypes">type</a> { i8 , [10 x [20 x i32]], i8 }
4996 %ST = <a href="#namedtypes">type</a> { i32, double, %RT }
4998 define i32* @foo(%ST* %s) {
5000 %reg = getelementptr %ST* %s, i32 1, i32 2, i32 1, i32 5, i32 13
5006 <p>In the example above, the first index is indexing into the '<tt>%ST*</tt>'
5007 type, which is a pointer, yielding a '<tt>%ST</tt>' = '<tt>{ i32, double, %RT
5008 }</tt>' type, a structure. The second index indexes into the third element
5009 of the structure, yielding a '<tt>%RT</tt>' = '<tt>{ i8 , [10 x [20 x i32]],
5010 i8 }</tt>' type, another structure. The third index indexes into the second
5011 element of the structure, yielding a '<tt>[10 x [20 x i32]]</tt>' type, an
5012 array. The two dimensions of the array are subscripted into, yielding an
5013 '<tt>i32</tt>' type. The '<tt>getelementptr</tt>' instruction returns a
5014 pointer to this element, thus computing a value of '<tt>i32*</tt>' type.</p>
5016 <p>Note that it is perfectly legal to index partially through a structure,
5017 returning a pointer to an inner element. Because of this, the LLVM code for
5018 the given testcase is equivalent to:</p>
5021 define i32* @foo(%ST* %s) {
5022 %t1 = getelementptr %ST* %s, i32 1 <i>; yields %ST*:%t1</i>
5023 %t2 = getelementptr %ST* %t1, i32 0, i32 2 <i>; yields %RT*:%t2</i>
5024 %t3 = getelementptr %RT* %t2, i32 0, i32 1 <i>; yields [10 x [20 x i32]]*:%t3</i>
5025 %t4 = getelementptr [10 x [20 x i32]]* %t3, i32 0, i32 5 <i>; yields [20 x i32]*:%t4</i>
5026 %t5 = getelementptr [20 x i32]* %t4, i32 0, i32 13 <i>; yields i32*:%t5</i>
5031 <p>If the <tt>inbounds</tt> keyword is present, the result value of the
5032 <tt>getelementptr</tt> is a <a href="#trapvalues">trap value</a> if the
5033 base pointer is not an <i>in bounds</i> address of an allocated object,
5034 or if any of the addresses that would be formed by successive addition of
5035 the offsets implied by the indices to the base address with infinitely
5036 precise signed arithmetic are not an <i>in bounds</i> address of that
5037 allocated object. The <i>in bounds</i> addresses for an allocated object
5038 are all the addresses that point into the object, plus the address one
5039 byte past the end.</p>
5041 <p>If the <tt>inbounds</tt> keyword is not present, the offsets are added to
5042 the base address with silently-wrapping two's complement arithmetic. If the
5043 offsets have a different width from the pointer, they are sign-extended or
5044 truncated to the width of the pointer. The result value of the
5045 <tt>getelementptr</tt> may be outside the object pointed to by the base
5046 pointer. The result value may not necessarily be used to access memory
5047 though, even if it happens to point into allocated storage. See the
5048 <a href="#pointeraliasing">Pointer Aliasing Rules</a> section for more
5051 <p>The getelementptr instruction is often confusing. For some more insight into
5052 how it works, see <a href="GetElementPtr.html">the getelementptr FAQ</a>.</p>
5056 <i>; yields [12 x i8]*:aptr</i>
5057 %aptr = getelementptr {i32, [12 x i8]}* %saptr, i64 0, i32 1
5058 <i>; yields i8*:vptr</i>
5059 %vptr = getelementptr {i32, <2 x i8>}* %svptr, i64 0, i32 1, i32 1
5060 <i>; yields i8*:eptr</i>
5061 %eptr = getelementptr [12 x i8]* %aptr, i64 0, i32 1
5062 <i>; yields i32*:iptr</i>
5063 %iptr = getelementptr [10 x i32]* @arr, i16 0, i16 0
5070 <!-- ======================================================================= -->
5072 <a name="convertops">Conversion Operations</a>
5077 <p>The instructions in this category are the conversion instructions (casting)
5078 which all take a single operand and a type. They perform various bit
5079 conversions on the operand.</p>
5081 <!-- _______________________________________________________________________ -->
5083 <a name="i_trunc">'<tt>trunc .. to</tt>' Instruction</a>
5090 <result> = trunc <ty> <value> to <ty2> <i>; yields ty2</i>
5094 <p>The '<tt>trunc</tt>' instruction truncates its operand to the
5095 type <tt>ty2</tt>.</p>
5098 <p>The '<tt>trunc</tt>' instruction takes a value to trunc, and a type to trunc it to.
5099 Both types must be of <a href="#t_integer">integer</a> types, or vectors
5100 of the same number of integers.
5101 The bit size of the <tt>value</tt> must be larger than
5102 the bit size of the destination type, <tt>ty2</tt>.
5103 Equal sized types are not allowed.</p>
5106 <p>The '<tt>trunc</tt>' instruction truncates the high order bits
5107 in <tt>value</tt> and converts the remaining bits to <tt>ty2</tt>. Since the
5108 source size must be larger than the destination size, <tt>trunc</tt> cannot
5109 be a <i>no-op cast</i>. It will always truncate bits.</p>
5113 %X = trunc i32 257 to i8 <i>; yields i8:1</i>
5114 %Y = trunc i32 123 to i1 <i>; yields i1:true</i>
5115 %Z = trunc i32 122 to i1 <i>; yields i1:false</i>
5116 %W = trunc <2 x i16> <i16 8, i16 7> to <2 x i8> <i>; yields <i8 8, i8 7></i>
5121 <!-- _______________________________________________________________________ -->
5123 <a name="i_zext">'<tt>zext .. to</tt>' Instruction</a>
5130 <result> = zext <ty> <value> to <ty2> <i>; yields ty2</i>
5134 <p>The '<tt>zext</tt>' instruction zero extends its operand to type
5139 <p>The '<tt>zext</tt>' instruction takes a value to cast, and a type to cast it to.
5140 Both types must be of <a href="#t_integer">integer</a> types, or vectors
5141 of the same number of integers.
5142 The bit size of the <tt>value</tt> must be smaller than
5143 the bit size of the destination type,
5147 <p>The <tt>zext</tt> fills the high order bits of the <tt>value</tt> with zero
5148 bits until it reaches the size of the destination type, <tt>ty2</tt>.</p>
5150 <p>When zero extending from i1, the result will always be either 0 or 1.</p>
5154 %X = zext i32 257 to i64 <i>; yields i64:257</i>
5155 %Y = zext i1 true to i32 <i>; yields i32:1</i>
5156 %Z = zext <2 x i16> <i16 8, i16 7> to <2 x i32> <i>; yields <i32 8, i32 7></i>
5161 <!-- _______________________________________________________________________ -->
5163 <a name="i_sext">'<tt>sext .. to</tt>' Instruction</a>
5170 <result> = sext <ty> <value> to <ty2> <i>; yields ty2</i>
5174 <p>The '<tt>sext</tt>' sign extends <tt>value</tt> to the type <tt>ty2</tt>.</p>
5177 <p>The '<tt>sext</tt>' instruction takes a value to cast, and a type to cast it to.
5178 Both types must be of <a href="#t_integer">integer</a> types, or vectors
5179 of the same number of integers.
5180 The bit size of the <tt>value</tt> must be smaller than
5181 the bit size of the destination type,
5185 <p>The '<tt>sext</tt>' instruction performs a sign extension by copying the sign
5186 bit (highest order bit) of the <tt>value</tt> until it reaches the bit size
5187 of the type <tt>ty2</tt>.</p>
5189 <p>When sign extending from i1, the extension always results in -1 or 0.</p>
5193 %X = sext i8 -1 to i16 <i>; yields i16 :65535</i>
5194 %Y = sext i1 true to i32 <i>; yields i32:-1</i>
5195 %Z = sext <2 x i16> <i16 8, i16 7> to <2 x i32> <i>; yields <i32 8, i32 7></i>
5200 <!-- _______________________________________________________________________ -->
5202 <a name="i_fptrunc">'<tt>fptrunc .. to</tt>' Instruction</a>
5209 <result> = fptrunc <ty> <value> to <ty2> <i>; yields ty2</i>
5213 <p>The '<tt>fptrunc</tt>' instruction truncates <tt>value</tt> to type
5217 <p>The '<tt>fptrunc</tt>' instruction takes a <a href="#t_floating">floating
5218 point</a> value to cast and a <a href="#t_floating">floating point</a> type
5219 to cast it to. The size of <tt>value</tt> must be larger than the size of
5220 <tt>ty2</tt>. This implies that <tt>fptrunc</tt> cannot be used to make a
5221 <i>no-op cast</i>.</p>
5224 <p>The '<tt>fptrunc</tt>' instruction truncates a <tt>value</tt> from a larger
5225 <a href="#t_floating">floating point</a> type to a smaller
5226 <a href="#t_floating">floating point</a> type. If the value cannot fit
5227 within the destination type, <tt>ty2</tt>, then the results are
5232 %X = fptrunc double 123.0 to float <i>; yields float:123.0</i>
5233 %Y = fptrunc double 1.0E+300 to float <i>; yields undefined</i>
5238 <!-- _______________________________________________________________________ -->
5240 <a name="i_fpext">'<tt>fpext .. to</tt>' Instruction</a>
5247 <result> = fpext <ty> <value> to <ty2> <i>; yields ty2</i>
5251 <p>The '<tt>fpext</tt>' extends a floating point <tt>value</tt> to a larger
5252 floating point value.</p>
5255 <p>The '<tt>fpext</tt>' instruction takes a
5256 <a href="#t_floating">floating point</a> <tt>value</tt> to cast, and
5257 a <a href="#t_floating">floating point</a> type to cast it to. The source
5258 type must be smaller than the destination type.</p>
5261 <p>The '<tt>fpext</tt>' instruction extends the <tt>value</tt> from a smaller
5262 <a href="#t_floating">floating point</a> type to a larger
5263 <a href="#t_floating">floating point</a> type. The <tt>fpext</tt> cannot be
5264 used to make a <i>no-op cast</i> because it always changes bits. Use
5265 <tt>bitcast</tt> to make a <i>no-op cast</i> for a floating point cast.</p>
5269 %X = fpext float 3.125 to double <i>; yields double:3.125000e+00</i>
5270 %Y = fpext double %X to fp128 <i>; yields fp128:0xL00000000000000004000900000000000</i>
5275 <!-- _______________________________________________________________________ -->
5277 <a name="i_fptoui">'<tt>fptoui .. to</tt>' Instruction</a>
5284 <result> = fptoui <ty> <value> to <ty2> <i>; yields ty2</i>
5288 <p>The '<tt>fptoui</tt>' converts a floating point <tt>value</tt> to its
5289 unsigned integer equivalent of type <tt>ty2</tt>.</p>
5292 <p>The '<tt>fptoui</tt>' instruction takes a value to cast, which must be a
5293 scalar or vector <a href="#t_floating">floating point</a> value, and a type
5294 to cast it to <tt>ty2</tt>, which must be an <a href="#t_integer">integer</a>
5295 type. If <tt>ty</tt> is a vector floating point type, <tt>ty2</tt> must be a
5296 vector integer type with the same number of elements as <tt>ty</tt></p>
5299 <p>The '<tt>fptoui</tt>' instruction converts its
5300 <a href="#t_floating">floating point</a> operand into the nearest (rounding
5301 towards zero) unsigned integer value. If the value cannot fit
5302 in <tt>ty2</tt>, the results are undefined.</p>
5306 %X = fptoui double 123.0 to i32 <i>; yields i32:123</i>
5307 %Y = fptoui float 1.0E+300 to i1 <i>; yields undefined:1</i>
5308 %Z = fptoui float 1.04E+17 to i8 <i>; yields undefined:1</i>
5313 <!-- _______________________________________________________________________ -->
5315 <a name="i_fptosi">'<tt>fptosi .. to</tt>' Instruction</a>
5322 <result> = fptosi <ty> <value> to <ty2> <i>; yields ty2</i>
5326 <p>The '<tt>fptosi</tt>' instruction converts
5327 <a href="#t_floating">floating point</a> <tt>value</tt> to
5328 type <tt>ty2</tt>.</p>
5331 <p>The '<tt>fptosi</tt>' instruction takes a value to cast, which must be a
5332 scalar or vector <a href="#t_floating">floating point</a> value, and a type
5333 to cast it to <tt>ty2</tt>, which must be an <a href="#t_integer">integer</a>
5334 type. If <tt>ty</tt> is a vector floating point type, <tt>ty2</tt> must be a
5335 vector integer type with the same number of elements as <tt>ty</tt></p>
5338 <p>The '<tt>fptosi</tt>' instruction converts its
5339 <a href="#t_floating">floating point</a> operand into the nearest (rounding
5340 towards zero) signed integer value. If the value cannot fit in <tt>ty2</tt>,
5341 the results are undefined.</p>
5345 %X = fptosi double -123.0 to i32 <i>; yields i32:-123</i>
5346 %Y = fptosi float 1.0E-247 to i1 <i>; yields undefined:1</i>
5347 %Z = fptosi float 1.04E+17 to i8 <i>; yields undefined:1</i>
5352 <!-- _______________________________________________________________________ -->
5354 <a name="i_uitofp">'<tt>uitofp .. to</tt>' Instruction</a>
5361 <result> = uitofp <ty> <value> to <ty2> <i>; yields ty2</i>
5365 <p>The '<tt>uitofp</tt>' instruction regards <tt>value</tt> as an unsigned
5366 integer and converts that value to the <tt>ty2</tt> type.</p>
5369 <p>The '<tt>uitofp</tt>' instruction takes a value to cast, which must be a
5370 scalar or vector <a href="#t_integer">integer</a> value, and a type to cast
5371 it to <tt>ty2</tt>, which must be an <a href="#t_floating">floating point</a>
5372 type. If <tt>ty</tt> is a vector integer type, <tt>ty2</tt> must be a vector
5373 floating point type with the same number of elements as <tt>ty</tt></p>
5376 <p>The '<tt>uitofp</tt>' instruction interprets its operand as an unsigned
5377 integer quantity and converts it to the corresponding floating point
5378 value. If the value cannot fit in the floating point value, the results are
5383 %X = uitofp i32 257 to float <i>; yields float:257.0</i>
5384 %Y = uitofp i8 -1 to double <i>; yields double:255.0</i>
5389 <!-- _______________________________________________________________________ -->
5391 <a name="i_sitofp">'<tt>sitofp .. to</tt>' Instruction</a>
5398 <result> = sitofp <ty> <value> to <ty2> <i>; yields ty2</i>
5402 <p>The '<tt>sitofp</tt>' instruction regards <tt>value</tt> as a signed integer
5403 and converts that value to the <tt>ty2</tt> type.</p>
5406 <p>The '<tt>sitofp</tt>' instruction takes a value to cast, which must be a
5407 scalar or vector <a href="#t_integer">integer</a> value, and a type to cast
5408 it to <tt>ty2</tt>, which must be an <a href="#t_floating">floating point</a>
5409 type. If <tt>ty</tt> is a vector integer type, <tt>ty2</tt> must be a vector
5410 floating point type with the same number of elements as <tt>ty</tt></p>
5413 <p>The '<tt>sitofp</tt>' instruction interprets its operand as a signed integer
5414 quantity and converts it to the corresponding floating point value. If the
5415 value cannot fit in the floating point value, the results are undefined.</p>
5419 %X = sitofp i32 257 to float <i>; yields float:257.0</i>
5420 %Y = sitofp i8 -1 to double <i>; yields double:-1.0</i>
5425 <!-- _______________________________________________________________________ -->
5427 <a name="i_ptrtoint">'<tt>ptrtoint .. to</tt>' Instruction</a>
5434 <result> = ptrtoint <ty> <value> to <ty2> <i>; yields ty2</i>
5438 <p>The '<tt>ptrtoint</tt>' instruction converts the pointer <tt>value</tt> to
5439 the integer type <tt>ty2</tt>.</p>
5442 <p>The '<tt>ptrtoint</tt>' instruction takes a <tt>value</tt> to cast, which
5443 must be a <a href="#t_pointer">pointer</a> value, and a type to cast it to
5444 <tt>ty2</tt>, which must be an <a href="#t_integer">integer</a> type.</p>
5447 <p>The '<tt>ptrtoint</tt>' instruction converts <tt>value</tt> to integer type
5448 <tt>ty2</tt> by interpreting the pointer value as an integer and either
5449 truncating or zero extending that value to the size of the integer type. If
5450 <tt>value</tt> is smaller than <tt>ty2</tt> then a zero extension is done. If
5451 <tt>value</tt> is larger than <tt>ty2</tt> then a truncation is done. If they
5452 are the same size, then nothing is done (<i>no-op cast</i>) other than a type
5457 %X = ptrtoint i32* %X to i8 <i>; yields truncation on 32-bit architecture</i>
5458 %Y = ptrtoint i32* %x to i64 <i>; yields zero extension on 32-bit architecture</i>
5463 <!-- _______________________________________________________________________ -->
5465 <a name="i_inttoptr">'<tt>inttoptr .. to</tt>' Instruction</a>
5472 <result> = inttoptr <ty> <value> to <ty2> <i>; yields ty2</i>
5476 <p>The '<tt>inttoptr</tt>' instruction converts an integer <tt>value</tt> to a
5477 pointer type, <tt>ty2</tt>.</p>
5480 <p>The '<tt>inttoptr</tt>' instruction takes an <a href="#t_integer">integer</a>
5481 value to cast, and a type to cast it to, which must be a
5482 <a href="#t_pointer">pointer</a> type.</p>
5485 <p>The '<tt>inttoptr</tt>' instruction converts <tt>value</tt> to type
5486 <tt>ty2</tt> by applying either a zero extension or a truncation depending on
5487 the size of the integer <tt>value</tt>. If <tt>value</tt> is larger than the
5488 size of a pointer then a truncation is done. If <tt>value</tt> is smaller
5489 than the size of a pointer then a zero extension is done. If they are the
5490 same size, nothing is done (<i>no-op cast</i>).</p>
5494 %X = inttoptr i32 255 to i32* <i>; yields zero extension on 64-bit architecture</i>
5495 %Y = inttoptr i32 255 to i32* <i>; yields no-op on 32-bit architecture</i>
5496 %Z = inttoptr i64 0 to i32* <i>; yields truncation on 32-bit architecture</i>
5501 <!-- _______________________________________________________________________ -->
5503 <a name="i_bitcast">'<tt>bitcast .. to</tt>' Instruction</a>
5510 <result> = bitcast <ty> <value> to <ty2> <i>; yields ty2</i>
5514 <p>The '<tt>bitcast</tt>' instruction converts <tt>value</tt> to type
5515 <tt>ty2</tt> without changing any bits.</p>
5518 <p>The '<tt>bitcast</tt>' instruction takes a value to cast, which must be a
5519 non-aggregate first class value, and a type to cast it to, which must also be
5520 a non-aggregate <a href="#t_firstclass">first class</a> type. The bit sizes
5521 of <tt>value</tt> and the destination type, <tt>ty2</tt>, must be
5522 identical. If the source type is a pointer, the destination type must also be
5523 a pointer. This instruction supports bitwise conversion of vectors to
5524 integers and to vectors of other types (as long as they have the same
5528 <p>The '<tt>bitcast</tt>' instruction converts <tt>value</tt> to type
5529 <tt>ty2</tt>. It is always a <i>no-op cast</i> because no bits change with
5530 this conversion. The conversion is done as if the <tt>value</tt> had been
5531 stored to memory and read back as type <tt>ty2</tt>. Pointer types may only
5532 be converted to other pointer types with this instruction. To convert
5533 pointers to other types, use the <a href="#i_inttoptr">inttoptr</a> or
5534 <a href="#i_ptrtoint">ptrtoint</a> instructions first.</p>
5538 %X = bitcast i8 255 to i8 <i>; yields i8 :-1</i>
5539 %Y = bitcast i32* %x to sint* <i>; yields sint*:%x</i>
5540 %Z = bitcast <2 x int> %V to i64; <i>; yields i64: %V</i>
5547 <!-- ======================================================================= -->
5549 <a name="otherops">Other Operations</a>
5554 <p>The instructions in this category are the "miscellaneous" instructions, which
5555 defy better classification.</p>
5557 <!-- _______________________________________________________________________ -->
5559 <a name="i_icmp">'<tt>icmp</tt>' Instruction</a>
5566 <result> = icmp <cond> <ty> <op1>, <op2> <i>; yields {i1} or {<N x i1>}:result</i>
5570 <p>The '<tt>icmp</tt>' instruction returns a boolean value or a vector of
5571 boolean values based on comparison of its two integer, integer vector, or
5572 pointer operands.</p>
5575 <p>The '<tt>icmp</tt>' instruction takes three operands. The first operand is
5576 the condition code indicating the kind of comparison to perform. It is not a
5577 value, just a keyword. The possible condition code are:</p>
5580 <li><tt>eq</tt>: equal</li>
5581 <li><tt>ne</tt>: not equal </li>
5582 <li><tt>ugt</tt>: unsigned greater than</li>
5583 <li><tt>uge</tt>: unsigned greater or equal</li>
5584 <li><tt>ult</tt>: unsigned less than</li>
5585 <li><tt>ule</tt>: unsigned less or equal</li>
5586 <li><tt>sgt</tt>: signed greater than</li>
5587 <li><tt>sge</tt>: signed greater or equal</li>
5588 <li><tt>slt</tt>: signed less than</li>
5589 <li><tt>sle</tt>: signed less or equal</li>
5592 <p>The remaining two arguments must be <a href="#t_integer">integer</a> or
5593 <a href="#t_pointer">pointer</a> or integer <a href="#t_vector">vector</a>
5594 typed. They must also be identical types.</p>
5597 <p>The '<tt>icmp</tt>' compares <tt>op1</tt> and <tt>op2</tt> according to the
5598 condition code given as <tt>cond</tt>. The comparison performed always yields
5599 either an <a href="#t_integer"><tt>i1</tt></a> or vector of <tt>i1</tt>
5600 result, as follows:</p>
5603 <li><tt>eq</tt>: yields <tt>true</tt> if the operands are equal,
5604 <tt>false</tt> otherwise. No sign interpretation is necessary or
5607 <li><tt>ne</tt>: yields <tt>true</tt> if the operands are unequal,
5608 <tt>false</tt> otherwise. No sign interpretation is necessary or
5611 <li><tt>ugt</tt>: interprets the operands as unsigned values and yields
5612 <tt>true</tt> if <tt>op1</tt> is greater than <tt>op2</tt>.</li>
5614 <li><tt>uge</tt>: interprets the operands as unsigned values and yields
5615 <tt>true</tt> if <tt>op1</tt> is greater than or equal
5616 to <tt>op2</tt>.</li>
5618 <li><tt>ult</tt>: interprets the operands as unsigned values and yields
5619 <tt>true</tt> if <tt>op1</tt> is less than <tt>op2</tt>.</li>
5621 <li><tt>ule</tt>: interprets the operands as unsigned values and yields
5622 <tt>true</tt> if <tt>op1</tt> is less than or equal to <tt>op2</tt>.</li>
5624 <li><tt>sgt</tt>: interprets the operands as signed values and yields
5625 <tt>true</tt> if <tt>op1</tt> is greater than <tt>op2</tt>.</li>
5627 <li><tt>sge</tt>: interprets the operands as signed values and yields
5628 <tt>true</tt> if <tt>op1</tt> is greater than or equal
5629 to <tt>op2</tt>.</li>
5631 <li><tt>slt</tt>: interprets the operands as signed values and yields
5632 <tt>true</tt> if <tt>op1</tt> is less than <tt>op2</tt>.</li>
5634 <li><tt>sle</tt>: interprets the operands as signed values and yields
5635 <tt>true</tt> if <tt>op1</tt> is less than or equal to <tt>op2</tt>.</li>
5638 <p>If the operands are <a href="#t_pointer">pointer</a> typed, the pointer
5639 values are compared as if they were integers.</p>
5641 <p>If the operands are integer vectors, then they are compared element by
5642 element. The result is an <tt>i1</tt> vector with the same number of elements
5643 as the values being compared. Otherwise, the result is an <tt>i1</tt>.</p>
5647 <result> = icmp eq i32 4, 5 <i>; yields: result=false</i>
5648 <result> = icmp ne float* %X, %X <i>; yields: result=false</i>
5649 <result> = icmp ult i16 4, 5 <i>; yields: result=true</i>
5650 <result> = icmp sgt i16 4, 5 <i>; yields: result=false</i>
5651 <result> = icmp ule i16 -4, 5 <i>; yields: result=false</i>
5652 <result> = icmp sge i16 4, 5 <i>; yields: result=false</i>
5655 <p>Note that the code generator does not yet support vector types with
5656 the <tt>icmp</tt> instruction.</p>
5660 <!-- _______________________________________________________________________ -->
5662 <a name="i_fcmp">'<tt>fcmp</tt>' Instruction</a>
5669 <result> = fcmp <cond> <ty> <op1>, <op2> <i>; yields {i1} or {<N x i1>}:result</i>
5673 <p>The '<tt>fcmp</tt>' instruction returns a boolean value or vector of boolean
5674 values based on comparison of its operands.</p>
5676 <p>If the operands are floating point scalars, then the result type is a boolean
5677 (<a href="#t_integer"><tt>i1</tt></a>).</p>
5679 <p>If the operands are floating point vectors, then the result type is a vector
5680 of boolean with the same number of elements as the operands being
5684 <p>The '<tt>fcmp</tt>' instruction takes three operands. The first operand is
5685 the condition code indicating the kind of comparison to perform. It is not a
5686 value, just a keyword. The possible condition code are:</p>
5689 <li><tt>false</tt>: no comparison, always returns false</li>
5690 <li><tt>oeq</tt>: ordered and equal</li>
5691 <li><tt>ogt</tt>: ordered and greater than </li>
5692 <li><tt>oge</tt>: ordered and greater than or equal</li>
5693 <li><tt>olt</tt>: ordered and less than </li>
5694 <li><tt>ole</tt>: ordered and less than or equal</li>
5695 <li><tt>one</tt>: ordered and not equal</li>
5696 <li><tt>ord</tt>: ordered (no nans)</li>
5697 <li><tt>ueq</tt>: unordered or equal</li>
5698 <li><tt>ugt</tt>: unordered or greater than </li>
5699 <li><tt>uge</tt>: unordered or greater than or equal</li>
5700 <li><tt>ult</tt>: unordered or less than </li>
5701 <li><tt>ule</tt>: unordered or less than or equal</li>
5702 <li><tt>une</tt>: unordered or not equal</li>
5703 <li><tt>uno</tt>: unordered (either nans)</li>
5704 <li><tt>true</tt>: no comparison, always returns true</li>
5707 <p><i>Ordered</i> means that neither operand is a QNAN while
5708 <i>unordered</i> means that either operand may be a QNAN.</p>
5710 <p>Each of <tt>val1</tt> and <tt>val2</tt> arguments must be either
5711 a <a href="#t_floating">floating point</a> type or
5712 a <a href="#t_vector">vector</a> of floating point type. They must have
5713 identical types.</p>
5716 <p>The '<tt>fcmp</tt>' instruction compares <tt>op1</tt> and <tt>op2</tt>
5717 according to the condition code given as <tt>cond</tt>. If the operands are
5718 vectors, then the vectors are compared element by element. Each comparison
5719 performed always yields an <a href="#t_integer">i1</a> result, as
5723 <li><tt>false</tt>: always yields <tt>false</tt>, regardless of operands.</li>
5725 <li><tt>oeq</tt>: yields <tt>true</tt> if both operands are not a QNAN and
5726 <tt>op1</tt> is equal to <tt>op2</tt>.</li>
5728 <li><tt>ogt</tt>: yields <tt>true</tt> if both operands are not a QNAN and
5729 <tt>op1</tt> is greater than <tt>op2</tt>.</li>
5731 <li><tt>oge</tt>: yields <tt>true</tt> if both operands are not a QNAN and
5732 <tt>op1</tt> is greater than or equal to <tt>op2</tt>.</li>
5734 <li><tt>olt</tt>: yields <tt>true</tt> if both operands are not a QNAN and
5735 <tt>op1</tt> is less than <tt>op2</tt>.</li>
5737 <li><tt>ole</tt>: yields <tt>true</tt> if both operands are not a QNAN and
5738 <tt>op1</tt> is less than or equal to <tt>op2</tt>.</li>
5740 <li><tt>one</tt>: yields <tt>true</tt> if both operands are not a QNAN and
5741 <tt>op1</tt> is not equal to <tt>op2</tt>.</li>
5743 <li><tt>ord</tt>: yields <tt>true</tt> if both operands are not a QNAN.</li>
5745 <li><tt>ueq</tt>: yields <tt>true</tt> if either operand is a QNAN or
5746 <tt>op1</tt> is equal to <tt>op2</tt>.</li>
5748 <li><tt>ugt</tt>: yields <tt>true</tt> if either operand is a QNAN or
5749 <tt>op1</tt> is greater than <tt>op2</tt>.</li>
5751 <li><tt>uge</tt>: yields <tt>true</tt> if either operand is a QNAN or
5752 <tt>op1</tt> is greater than or equal to <tt>op2</tt>.</li>
5754 <li><tt>ult</tt>: yields <tt>true</tt> if either operand is a QNAN or
5755 <tt>op1</tt> is less than <tt>op2</tt>.</li>
5757 <li><tt>ule</tt>: yields <tt>true</tt> if either operand is a QNAN or
5758 <tt>op1</tt> is less than or equal to <tt>op2</tt>.</li>
5760 <li><tt>une</tt>: yields <tt>true</tt> if either operand is a QNAN or
5761 <tt>op1</tt> is not equal to <tt>op2</tt>.</li>
5763 <li><tt>uno</tt>: yields <tt>true</tt> if either operand is a QNAN.</li>
5765 <li><tt>true</tt>: always yields <tt>true</tt>, regardless of operands.</li>
5770 <result> = fcmp oeq float 4.0, 5.0 <i>; yields: result=false</i>
5771 <result> = fcmp one float 4.0, 5.0 <i>; yields: result=true</i>
5772 <result> = fcmp olt float 4.0, 5.0 <i>; yields: result=true</i>
5773 <result> = fcmp ueq double 1.0, 2.0 <i>; yields: result=false</i>
5776 <p>Note that the code generator does not yet support vector types with
5777 the <tt>fcmp</tt> instruction.</p>
5781 <!-- _______________________________________________________________________ -->
5783 <a name="i_phi">'<tt>phi</tt>' Instruction</a>
5790 <result> = phi <ty> [ <val0>, <label0>], ...
5794 <p>The '<tt>phi</tt>' instruction is used to implement the φ node in the
5795 SSA graph representing the function.</p>
5798 <p>The type of the incoming values is specified with the first type field. After
5799 this, the '<tt>phi</tt>' instruction takes a list of pairs as arguments, with
5800 one pair for each predecessor basic block of the current block. Only values
5801 of <a href="#t_firstclass">first class</a> type may be used as the value
5802 arguments to the PHI node. Only labels may be used as the label
5805 <p>There must be no non-phi instructions between the start of a basic block and
5806 the PHI instructions: i.e. PHI instructions must be first in a basic
5809 <p>For the purposes of the SSA form, the use of each incoming value is deemed to
5810 occur on the edge from the corresponding predecessor block to the current
5811 block (but after any definition of an '<tt>invoke</tt>' instruction's return
5812 value on the same edge).</p>
5815 <p>At runtime, the '<tt>phi</tt>' instruction logically takes on the value
5816 specified by the pair corresponding to the predecessor basic block that
5817 executed just prior to the current block.</p>
5821 Loop: ; Infinite loop that counts from 0 on up...
5822 %indvar = phi i32 [ 0, %LoopHeader ], [ %nextindvar, %Loop ]
5823 %nextindvar = add i32 %indvar, 1
5829 <!-- _______________________________________________________________________ -->
5831 <a name="i_select">'<tt>select</tt>' Instruction</a>
5838 <result> = select <i>selty</i> <cond>, <ty> <val1>, <ty> <val2> <i>; yields ty</i>
5840 <i>selty</i> is either i1 or {<N x i1>}
5844 <p>The '<tt>select</tt>' instruction is used to choose one value based on a
5845 condition, without branching.</p>
5849 <p>The '<tt>select</tt>' instruction requires an 'i1' value or a vector of 'i1'
5850 values indicating the condition, and two values of the
5851 same <a href="#t_firstclass">first class</a> type. If the val1/val2 are
5852 vectors and the condition is a scalar, then entire vectors are selected, not
5853 individual elements.</p>
5856 <p>If the condition is an i1 and it evaluates to 1, the instruction returns the
5857 first value argument; otherwise, it returns the second value argument.</p>
5859 <p>If the condition is a vector of i1, then the value arguments must be vectors
5860 of the same size, and the selection is done element by element.</p>
5864 %X = select i1 true, i8 17, i8 42 <i>; yields i8:17</i>
5867 <p>Note that the code generator does not yet support conditions
5868 with vector type.</p>
5872 <!-- _______________________________________________________________________ -->
5874 <a name="i_call">'<tt>call</tt>' Instruction</a>
5881 <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>]
5885 <p>The '<tt>call</tt>' instruction represents a simple function call.</p>
5888 <p>This instruction requires several arguments:</p>
5891 <li>The optional "tail" marker indicates that the callee function does not
5892 access any allocas or varargs in the caller. Note that calls may be
5893 marked "tail" even if they do not occur before
5894 a <a href="#i_ret"><tt>ret</tt></a> instruction. If the "tail" marker is
5895 present, the function call is eligible for tail call optimization,
5896 but <a href="CodeGenerator.html#tailcallopt">might not in fact be
5897 optimized into a jump</a>. The code generator may optimize calls marked
5898 "tail" with either 1) automatic <a href="CodeGenerator.html#sibcallopt">
5899 sibling call optimization</a> when the caller and callee have
5900 matching signatures, or 2) forced tail call optimization when the
5901 following extra requirements are met:
5903 <li>Caller and callee both have the calling
5904 convention <tt>fastcc</tt>.</li>
5905 <li>The call is in tail position (ret immediately follows call and ret
5906 uses value of call or is void).</li>
5907 <li>Option <tt>-tailcallopt</tt> is enabled,
5908 or <code>llvm::GuaranteedTailCallOpt</code> is <code>true</code>.</li>
5909 <li><a href="CodeGenerator.html#tailcallopt">Platform specific
5910 constraints are met.</a></li>
5914 <li>The optional "cconv" marker indicates which <a href="#callingconv">calling
5915 convention</a> the call should use. If none is specified, the call
5916 defaults to using C calling conventions. The calling convention of the
5917 call must match the calling convention of the target function, or else the
5918 behavior is undefined.</li>
5920 <li>The optional <a href="#paramattrs">Parameter Attributes</a> list for
5921 return values. Only '<tt>zeroext</tt>', '<tt>signext</tt>', and
5922 '<tt>inreg</tt>' attributes are valid here.</li>
5924 <li>'<tt>ty</tt>': the type of the call instruction itself which is also the
5925 type of the return value. Functions that return no value are marked
5926 <tt><a href="#t_void">void</a></tt>.</li>
5928 <li>'<tt>fnty</tt>': shall be the signature of the pointer to function value
5929 being invoked. The argument types must match the types implied by this
5930 signature. This type can be omitted if the function is not varargs and if
5931 the function type does not return a pointer to a function.</li>
5933 <li>'<tt>fnptrval</tt>': An LLVM value containing a pointer to a function to
5934 be invoked. In most cases, this is a direct function invocation, but
5935 indirect <tt>call</tt>s are just as possible, calling an arbitrary pointer
5936 to function value.</li>
5938 <li>'<tt>function args</tt>': argument list whose types match the function
5939 signature argument types and parameter attributes. All arguments must be
5940 of <a href="#t_firstclass">first class</a> type. If the function
5941 signature indicates the function accepts a variable number of arguments,
5942 the extra arguments can be specified.</li>
5944 <li>The optional <a href="#fnattrs">function attributes</a> list. Only
5945 '<tt>noreturn</tt>', '<tt>nounwind</tt>', '<tt>readonly</tt>' and
5946 '<tt>readnone</tt>' attributes are valid here.</li>
5950 <p>The '<tt>call</tt>' instruction is used to cause control flow to transfer to
5951 a specified function, with its incoming arguments bound to the specified
5952 values. Upon a '<tt><a href="#i_ret">ret</a></tt>' instruction in the called
5953 function, control flow continues with the instruction after the function
5954 call, and the return value of the function is bound to the result
5959 %retval = call i32 @test(i32 %argc)
5960 call i32 (i8*, ...)* @printf(i8* %msg, i32 12, i8 42) <i>; yields i32</i>
5961 %X = tail call i32 @foo() <i>; yields i32</i>
5962 %Y = tail call <a href="#callingconv">fastcc</a> i32 @foo() <i>; yields i32</i>
5963 call void %foo(i8 97 signext)
5965 %struct.A = type { i32, i8 }
5966 %r = call %struct.A @foo() <i>; yields { 32, i8 }</i>
5967 %gr = extractvalue %struct.A %r, 0 <i>; yields i32</i>
5968 %gr1 = extractvalue %struct.A %r, 1 <i>; yields i8</i>
5969 %Z = call void @foo() noreturn <i>; indicates that %foo never returns normally</i>
5970 %ZZ = call zeroext i32 @bar() <i>; Return value is %zero extended</i>
5973 <p>llvm treats calls to some functions with names and arguments that match the
5974 standard C99 library as being the C99 library functions, and may perform
5975 optimizations or generate code for them under that assumption. This is
5976 something we'd like to change in the future to provide better support for
5977 freestanding environments and non-C-based languages.</p>
5981 <!-- _______________________________________________________________________ -->
5983 <a name="i_va_arg">'<tt>va_arg</tt>' Instruction</a>
5990 <resultval> = va_arg <va_list*> <arglist>, <argty>
5994 <p>The '<tt>va_arg</tt>' instruction is used to access arguments passed through
5995 the "variable argument" area of a function call. It is used to implement the
5996 <tt>va_arg</tt> macro in C.</p>
5999 <p>This instruction takes a <tt>va_list*</tt> value and the type of the
6000 argument. It returns a value of the specified argument type and increments
6001 the <tt>va_list</tt> to point to the next argument. The actual type
6002 of <tt>va_list</tt> is target specific.</p>
6005 <p>The '<tt>va_arg</tt>' instruction loads an argument of the specified type
6006 from the specified <tt>va_list</tt> and causes the <tt>va_list</tt> to point
6007 to the next argument. For more information, see the variable argument
6008 handling <a href="#int_varargs">Intrinsic Functions</a>.</p>
6010 <p>It is legal for this instruction to be called in a function which does not
6011 take a variable number of arguments, for example, the <tt>vfprintf</tt>
6014 <p><tt>va_arg</tt> is an LLVM instruction instead of
6015 an <a href="#intrinsics">intrinsic function</a> because it takes a type as an
6019 <p>See the <a href="#int_varargs">variable argument processing</a> section.</p>
6021 <p>Note that the code generator does not yet fully support va_arg on many
6022 targets. Also, it does not currently support va_arg with aggregate types on
6027 <!-- _______________________________________________________________________ -->
6029 <a name="i_landingpad">'<tt>landingpad</tt>' Instruction</a>
6036 <resultval> = landingpad <somety> personality <type> <pers_fn> <clause>+
6037 <resultval> = landingpad <somety> personality <type> <pers_fn> cleanup <clause>*
6039 <clause> := catch <type> <value>
6040 <clause> := filter <array constant type> <array constant>
6044 <p>The '<tt>landingpad</tt>' instruction is used by
6045 <a href="ExceptionHandling.html#overview">LLVM's exception handling
6046 system</a> to specify that a basic block is a landing pad — one where
6047 the exception lands, and corresponds to the code found in the
6048 <i><tt>catch</tt></i> portion of a <i><tt>try/catch</tt></i> sequence. It
6049 defines values supplied by the personality function (<tt>pers_fn</tt>) upon
6050 re-entry to the function. The <tt>resultval</tt> has the
6051 type <tt>somety</tt>.</p>
6054 <p>This instruction takes a <tt>pers_fn</tt> value. This is the personality
6055 function associated with the unwinding mechanism. The optional
6056 <tt>cleanup</tt> flag indicates that the landing pad block is a cleanup.</p>
6058 <p>A <tt>clause</tt> begins with the clause type — <tt>catch</tt>
6059 or <tt>filter</tt> — and contains the global variable representing the
6060 "type" that may be caught or filtered respectively. Unlike the
6061 <tt>catch</tt> clause, the <tt>filter</tt> clause takes an array constant as
6062 its argument. Use "<tt>[0 x i8**] undef</tt>" for a filter which cannot
6063 throw. The '<tt>landingpad</tt>' instruction must contain <em>at least</em>
6064 one <tt>clause</tt> or the <tt>cleanup</tt> flag.</p>
6067 <p>The '<tt>landingpad</tt>' instruction defines the values which are set by the
6068 personality function (<tt>pers_fn</tt>) upon re-entry to the function, and
6069 therefore the "result type" of the <tt>landingpad</tt> instruction. As with
6070 calling conventions, how the personality function results are represented in
6071 LLVM IR is target specific.</p>
6073 <p>The clauses are applied in order from top to bottom. If two
6074 <tt>landingpad</tt> instructions are merged together through inlining, the
6075 clauses from the calling function are appended to the list of clauses.</p>
6077 <p>The <tt>landingpad</tt> instruction has several restrictions:</p>
6080 <li>A landing pad block is a basic block which is the unwind destination of an
6081 '<tt>invoke</tt>' instruction.</li>
6082 <li>A landing pad block must have a '<tt>landingpad</tt>' instruction as its
6083 first non-PHI instruction.</li>
6084 <li>There can be only one '<tt>landingpad</tt>' instruction within the landing
6086 <li>A basic block that is not a landing pad block may not include a
6087 '<tt>landingpad</tt>' instruction.</li>
6088 <li>All '<tt>landingpad</tt>' instructions in a function must have the same
6089 personality function.</li>
6094 ;; A landing pad which can catch an integer.
6095 %res = landingpad { i8*, i32 } personality i32 (...)* @__gxx_personality_v0
6097 ;; A landing pad that is a cleanup.
6098 %res = landingpad { i8*, i32 } personality i32 (...)* @__gxx_personality_v0
6100 ;; A landing pad which can catch an integer and can only throw a double.
6101 %res = landingpad { i8*, i32 } personality i32 (...)* @__gxx_personality_v0
6103 filter [1 x i8**] [@_ZTId]
6112 <!-- *********************************************************************** -->
6113 <h2><a name="intrinsics">Intrinsic Functions</a></h2>
6114 <!-- *********************************************************************** -->
6118 <p>LLVM supports the notion of an "intrinsic function". These functions have
6119 well known names and semantics and are required to follow certain
6120 restrictions. Overall, these intrinsics represent an extension mechanism for
6121 the LLVM language that does not require changing all of the transformations
6122 in LLVM when adding to the language (or the bitcode reader/writer, the
6123 parser, etc...).</p>
6125 <p>Intrinsic function names must all start with an "<tt>llvm.</tt>" prefix. This
6126 prefix is reserved in LLVM for intrinsic names; thus, function names may not
6127 begin with this prefix. Intrinsic functions must always be external
6128 functions: you cannot define the body of intrinsic functions. Intrinsic
6129 functions may only be used in call or invoke instructions: it is illegal to
6130 take the address of an intrinsic function. Additionally, because intrinsic
6131 functions are part of the LLVM language, it is required if any are added that
6132 they be documented here.</p>
6134 <p>Some intrinsic functions can be overloaded, i.e., the intrinsic represents a
6135 family of functions that perform the same operation but on different data
6136 types. Because LLVM can represent over 8 million different integer types,
6137 overloading is used commonly to allow an intrinsic function to operate on any
6138 integer type. One or more of the argument types or the result type can be
6139 overloaded to accept any integer type. Argument types may also be defined as
6140 exactly matching a previous argument's type or the result type. This allows
6141 an intrinsic function which accepts multiple arguments, but needs all of them
6142 to be of the same type, to only be overloaded with respect to a single
6143 argument or the result.</p>
6145 <p>Overloaded intrinsics will have the names of its overloaded argument types
6146 encoded into its function name, each preceded by a period. Only those types
6147 which are overloaded result in a name suffix. Arguments whose type is matched
6148 against another type do not. For example, the <tt>llvm.ctpop</tt> function
6149 can take an integer of any width and returns an integer of exactly the same
6150 integer width. This leads to a family of functions such as
6151 <tt>i8 @llvm.ctpop.i8(i8 %val)</tt> and <tt>i29 @llvm.ctpop.i29(i29
6152 %val)</tt>. Only one type, the return type, is overloaded, and only one type
6153 suffix is required. Because the argument's type is matched against the return
6154 type, it does not require its own name suffix.</p>
6156 <p>To learn how to add an intrinsic function, please see the
6157 <a href="ExtendingLLVM.html">Extending LLVM Guide</a>.</p>
6159 <!-- ======================================================================= -->
6161 <a name="int_varargs">Variable Argument Handling Intrinsics</a>
6166 <p>Variable argument support is defined in LLVM with
6167 the <a href="#i_va_arg"><tt>va_arg</tt></a> instruction and these three
6168 intrinsic functions. These functions are related to the similarly named
6169 macros defined in the <tt><stdarg.h></tt> header file.</p>
6171 <p>All of these functions operate on arguments that use a target-specific value
6172 type "<tt>va_list</tt>". The LLVM assembly language reference manual does
6173 not define what this type is, so all transformations should be prepared to
6174 handle these functions regardless of the type used.</p>
6176 <p>This example shows how the <a href="#i_va_arg"><tt>va_arg</tt></a>
6177 instruction and the variable argument handling intrinsic functions are
6180 <pre class="doc_code">
6181 define i32 @test(i32 %X, ...) {
6182 ; Initialize variable argument processing
6184 %ap2 = bitcast i8** %ap to i8*
6185 call void @llvm.va_start(i8* %ap2)
6187 ; Read a single integer argument
6188 %tmp = va_arg i8** %ap, i32
6190 ; Demonstrate usage of llvm.va_copy and llvm.va_end
6192 %aq2 = bitcast i8** %aq to i8*
6193 call void @llvm.va_copy(i8* %aq2, i8* %ap2)
6194 call void @llvm.va_end(i8* %aq2)
6196 ; Stop processing of arguments.
6197 call void @llvm.va_end(i8* %ap2)
6201 declare void @llvm.va_start(i8*)
6202 declare void @llvm.va_copy(i8*, i8*)
6203 declare void @llvm.va_end(i8*)
6206 <!-- _______________________________________________________________________ -->
6208 <a name="int_va_start">'<tt>llvm.va_start</tt>' Intrinsic</a>
6216 declare void %llvm.va_start(i8* <arglist>)
6220 <p>The '<tt>llvm.va_start</tt>' intrinsic initializes <tt>*<arglist></tt>
6221 for subsequent use by <tt><a href="#i_va_arg">va_arg</a></tt>.</p>
6224 <p>The argument is a pointer to a <tt>va_list</tt> element to initialize.</p>
6227 <p>The '<tt>llvm.va_start</tt>' intrinsic works just like the <tt>va_start</tt>
6228 macro available in C. In a target-dependent way, it initializes
6229 the <tt>va_list</tt> element to which the argument points, so that the next
6230 call to <tt>va_arg</tt> will produce the first variable argument passed to
6231 the function. Unlike the C <tt>va_start</tt> macro, this intrinsic does not
6232 need to know the last argument of the function as the compiler can figure
6237 <!-- _______________________________________________________________________ -->
6239 <a name="int_va_end">'<tt>llvm.va_end</tt>' Intrinsic</a>
6246 declare void @llvm.va_end(i8* <arglist>)
6250 <p>The '<tt>llvm.va_end</tt>' intrinsic destroys <tt>*<arglist></tt>,
6251 which has been initialized previously
6252 with <tt><a href="#int_va_start">llvm.va_start</a></tt>
6253 or <tt><a href="#i_va_copy">llvm.va_copy</a></tt>.</p>
6256 <p>The argument is a pointer to a <tt>va_list</tt> to destroy.</p>
6259 <p>The '<tt>llvm.va_end</tt>' intrinsic works just like the <tt>va_end</tt>
6260 macro available in C. In a target-dependent way, it destroys
6261 the <tt>va_list</tt> element to which the argument points. Calls
6262 to <a href="#int_va_start"><tt>llvm.va_start</tt></a>
6263 and <a href="#int_va_copy"> <tt>llvm.va_copy</tt></a> must be matched exactly
6264 with calls to <tt>llvm.va_end</tt>.</p>
6268 <!-- _______________________________________________________________________ -->
6270 <a name="int_va_copy">'<tt>llvm.va_copy</tt>' Intrinsic</a>
6277 declare void @llvm.va_copy(i8* <destarglist>, i8* <srcarglist>)
6281 <p>The '<tt>llvm.va_copy</tt>' intrinsic copies the current argument position
6282 from the source argument list to the destination argument list.</p>
6285 <p>The first argument is a pointer to a <tt>va_list</tt> element to initialize.
6286 The second argument is a pointer to a <tt>va_list</tt> element to copy
6290 <p>The '<tt>llvm.va_copy</tt>' intrinsic works just like the <tt>va_copy</tt>
6291 macro available in C. In a target-dependent way, it copies the
6292 source <tt>va_list</tt> element into the destination <tt>va_list</tt>
6293 element. This intrinsic is necessary because
6294 the <tt><a href="#int_va_start"> llvm.va_start</a></tt> intrinsic may be
6295 arbitrarily complex and require, for example, memory allocation.</p>
6303 <!-- ======================================================================= -->
6305 <a name="int_gc">Accurate Garbage Collection Intrinsics</a>
6310 <p>LLVM support for <a href="GarbageCollection.html">Accurate Garbage
6311 Collection</a> (GC) requires the implementation and generation of these
6312 intrinsics. These intrinsics allow identification of <a href="#int_gcroot">GC
6313 roots on the stack</a>, as well as garbage collector implementations that
6314 require <a href="#int_gcread">read</a> and <a href="#int_gcwrite">write</a>
6315 barriers. Front-ends for type-safe garbage collected languages should generate
6316 these intrinsics to make use of the LLVM garbage collectors. For more details,
6317 see <a href="GarbageCollection.html">Accurate Garbage Collection with
6320 <p>The garbage collection intrinsics only operate on objects in the generic
6321 address space (address space zero).</p>
6323 <!-- _______________________________________________________________________ -->
6325 <a name="int_gcroot">'<tt>llvm.gcroot</tt>' Intrinsic</a>
6332 declare void @llvm.gcroot(i8** %ptrloc, i8* %metadata)
6336 <p>The '<tt>llvm.gcroot</tt>' intrinsic declares the existence of a GC root to
6337 the code generator, and allows some metadata to be associated with it.</p>
6340 <p>The first argument specifies the address of a stack object that contains the
6341 root pointer. The second pointer (which must be either a constant or a
6342 global value address) contains the meta-data to be associated with the
6346 <p>At runtime, a call to this intrinsic stores a null pointer into the "ptrloc"
6347 location. At compile-time, the code generator generates information to allow
6348 the runtime to find the pointer at GC safe points. The '<tt>llvm.gcroot</tt>'
6349 intrinsic may only be used in a function which <a href="#gc">specifies a GC
6354 <!-- _______________________________________________________________________ -->
6356 <a name="int_gcread">'<tt>llvm.gcread</tt>' Intrinsic</a>
6363 declare i8* @llvm.gcread(i8* %ObjPtr, i8** %Ptr)
6367 <p>The '<tt>llvm.gcread</tt>' intrinsic identifies reads of references from heap
6368 locations, allowing garbage collector implementations that require read
6372 <p>The second argument is the address to read from, which should be an address
6373 allocated from the garbage collector. The first object is a pointer to the
6374 start of the referenced object, if needed by the language runtime (otherwise
6378 <p>The '<tt>llvm.gcread</tt>' intrinsic has the same semantics as a load
6379 instruction, but may be replaced with substantially more complex code by the
6380 garbage collector runtime, as needed. The '<tt>llvm.gcread</tt>' intrinsic
6381 may only be used in a function which <a href="#gc">specifies a GC
6386 <!-- _______________________________________________________________________ -->
6388 <a name="int_gcwrite">'<tt>llvm.gcwrite</tt>' Intrinsic</a>
6395 declare void @llvm.gcwrite(i8* %P1, i8* %Obj, i8** %P2)
6399 <p>The '<tt>llvm.gcwrite</tt>' intrinsic identifies writes of references to heap
6400 locations, allowing garbage collector implementations that require write
6401 barriers (such as generational or reference counting collectors).</p>
6404 <p>The first argument is the reference to store, the second is the start of the
6405 object to store it to, and the third is the address of the field of Obj to
6406 store to. If the runtime does not require a pointer to the object, Obj may
6410 <p>The '<tt>llvm.gcwrite</tt>' intrinsic has the same semantics as a store
6411 instruction, but may be replaced with substantially more complex code by the
6412 garbage collector runtime, as needed. The '<tt>llvm.gcwrite</tt>' intrinsic
6413 may only be used in a function which <a href="#gc">specifies a GC
6420 <!-- ======================================================================= -->
6422 <a name="int_codegen">Code Generator Intrinsics</a>
6427 <p>These intrinsics are provided by LLVM to expose special features that may
6428 only be implemented with code generator support.</p>
6430 <!-- _______________________________________________________________________ -->
6432 <a name="int_returnaddress">'<tt>llvm.returnaddress</tt>' Intrinsic</a>
6439 declare i8 *@llvm.returnaddress(i32 <level>)
6443 <p>The '<tt>llvm.returnaddress</tt>' intrinsic attempts to compute a
6444 target-specific value indicating the return address of the current function
6445 or one of its callers.</p>
6448 <p>The argument to this intrinsic indicates which function to return the address
6449 for. Zero indicates the calling function, one indicates its caller, etc.
6450 The argument is <b>required</b> to be a constant integer value.</p>
6453 <p>The '<tt>llvm.returnaddress</tt>' intrinsic either returns a pointer
6454 indicating the return address of the specified call frame, or zero if it
6455 cannot be identified. The value returned by this intrinsic is likely to be
6456 incorrect or 0 for arguments other than zero, so it should only be used for
6457 debugging purposes.</p>
6459 <p>Note that calling this intrinsic does not prevent function inlining or other
6460 aggressive transformations, so the value returned may not be that of the
6461 obvious source-language caller.</p>
6465 <!-- _______________________________________________________________________ -->
6467 <a name="int_frameaddress">'<tt>llvm.frameaddress</tt>' Intrinsic</a>
6474 declare i8* @llvm.frameaddress(i32 <level>)
6478 <p>The '<tt>llvm.frameaddress</tt>' intrinsic attempts to return the
6479 target-specific frame pointer value for the specified stack frame.</p>
6482 <p>The argument to this intrinsic indicates which function to return the frame
6483 pointer for. Zero indicates the calling function, one indicates its caller,
6484 etc. The argument is <b>required</b> to be a constant integer value.</p>
6487 <p>The '<tt>llvm.frameaddress</tt>' intrinsic either returns a pointer
6488 indicating the frame address of the specified call frame, or zero if it
6489 cannot be identified. The value returned by this intrinsic is likely to be
6490 incorrect or 0 for arguments other than zero, so it should only be used for
6491 debugging purposes.</p>
6493 <p>Note that calling this intrinsic does not prevent function inlining or other
6494 aggressive transformations, so the value returned may not be that of the
6495 obvious source-language caller.</p>
6499 <!-- _______________________________________________________________________ -->
6501 <a name="int_stacksave">'<tt>llvm.stacksave</tt>' Intrinsic</a>
6508 declare i8* @llvm.stacksave()
6512 <p>The '<tt>llvm.stacksave</tt>' intrinsic is used to remember the current state
6513 of the function stack, for use
6514 with <a href="#int_stackrestore"> <tt>llvm.stackrestore</tt></a>. This is
6515 useful for implementing language features like scoped automatic variable
6516 sized arrays in C99.</p>
6519 <p>This intrinsic returns a opaque pointer value that can be passed
6520 to <a href="#int_stackrestore"><tt>llvm.stackrestore</tt></a>. When
6521 an <tt>llvm.stackrestore</tt> intrinsic is executed with a value saved
6522 from <tt>llvm.stacksave</tt>, it effectively restores the state of the stack
6523 to the state it was in when the <tt>llvm.stacksave</tt> intrinsic executed.
6524 In practice, this pops any <a href="#i_alloca">alloca</a> blocks from the
6525 stack that were allocated after the <tt>llvm.stacksave</tt> was executed.</p>
6529 <!-- _______________________________________________________________________ -->
6531 <a name="int_stackrestore">'<tt>llvm.stackrestore</tt>' Intrinsic</a>
6538 declare void @llvm.stackrestore(i8* %ptr)
6542 <p>The '<tt>llvm.stackrestore</tt>' intrinsic is used to restore the state of
6543 the function stack to the state it was in when the
6544 corresponding <a href="#int_stacksave"><tt>llvm.stacksave</tt></a> intrinsic
6545 executed. This is useful for implementing language features like scoped
6546 automatic variable sized arrays in C99.</p>
6549 <p>See the description
6550 for <a href="#int_stacksave"><tt>llvm.stacksave</tt></a>.</p>
6554 <!-- _______________________________________________________________________ -->
6556 <a name="int_prefetch">'<tt>llvm.prefetch</tt>' Intrinsic</a>
6563 declare void @llvm.prefetch(i8* <address>, i32 <rw>, i32 <locality>, i32 <cache type>)
6567 <p>The '<tt>llvm.prefetch</tt>' intrinsic is a hint to the code generator to
6568 insert a prefetch instruction if supported; otherwise, it is a noop.
6569 Prefetches have no effect on the behavior of the program but can change its
6570 performance characteristics.</p>
6573 <p><tt>address</tt> is the address to be prefetched, <tt>rw</tt> is the
6574 specifier determining if the fetch should be for a read (0) or write (1),
6575 and <tt>locality</tt> is a temporal locality specifier ranging from (0) - no
6576 locality, to (3) - extremely local keep in cache. The <tt>cache type</tt>
6577 specifies whether the prefetch is performed on the data (1) or instruction (0)
6578 cache. The <tt>rw</tt>, <tt>locality</tt> and <tt>cache type</tt> arguments
6579 must be constant integers.</p>
6582 <p>This intrinsic does not modify the behavior of the program. In particular,
6583 prefetches cannot trap and do not produce a value. On targets that support
6584 this intrinsic, the prefetch can provide hints to the processor cache for
6585 better performance.</p>
6589 <!-- _______________________________________________________________________ -->
6591 <a name="int_pcmarker">'<tt>llvm.pcmarker</tt>' Intrinsic</a>
6598 declare void @llvm.pcmarker(i32 <id>)
6602 <p>The '<tt>llvm.pcmarker</tt>' intrinsic is a method to export a Program
6603 Counter (PC) in a region of code to simulators and other tools. The method
6604 is target specific, but it is expected that the marker will use exported
6605 symbols to transmit the PC of the marker. The marker makes no guarantees
6606 that it will remain with any specific instruction after optimizations. It is
6607 possible that the presence of a marker will inhibit optimizations. The
6608 intended use is to be inserted after optimizations to allow correlations of
6609 simulation runs.</p>
6612 <p><tt>id</tt> is a numerical id identifying the marker.</p>
6615 <p>This intrinsic does not modify the behavior of the program. Backends that do
6616 not support this intrinsic may ignore it.</p>
6620 <!-- _______________________________________________________________________ -->
6622 <a name="int_readcyclecounter">'<tt>llvm.readcyclecounter</tt>' Intrinsic</a>
6629 declare i64 @llvm.readcyclecounter()
6633 <p>The '<tt>llvm.readcyclecounter</tt>' intrinsic provides access to the cycle
6634 counter register (or similar low latency, high accuracy clocks) on those
6635 targets that support it. On X86, it should map to RDTSC. On Alpha, it
6636 should map to RPCC. As the backing counters overflow quickly (on the order
6637 of 9 seconds on alpha), this should only be used for small timings.</p>
6640 <p>When directly supported, reading the cycle counter should not modify any
6641 memory. Implementations are allowed to either return a application specific
6642 value or a system wide value. On backends without support, this is lowered
6643 to a constant 0.</p>
6649 <!-- ======================================================================= -->
6651 <a name="int_libc">Standard C Library Intrinsics</a>
6656 <p>LLVM provides intrinsics for a few important standard C library functions.
6657 These intrinsics allow source-language front-ends to pass information about
6658 the alignment of the pointer arguments to the code generator, providing
6659 opportunity for more efficient code generation.</p>
6661 <!-- _______________________________________________________________________ -->
6663 <a name="int_memcpy">'<tt>llvm.memcpy</tt>' Intrinsic</a>
6669 <p>This is an overloaded intrinsic. You can use <tt>llvm.memcpy</tt> on any
6670 integer bit width and for different address spaces. Not all targets support
6671 all bit widths however.</p>
6674 declare void @llvm.memcpy.p0i8.p0i8.i32(i8* <dest>, i8* <src>,
6675 i32 <len>, i32 <align>, i1 <isvolatile>)
6676 declare void @llvm.memcpy.p0i8.p0i8.i64(i8* <dest>, i8* <src>,
6677 i64 <len>, i32 <align>, i1 <isvolatile>)
6681 <p>The '<tt>llvm.memcpy.*</tt>' intrinsics copy a block of memory from the
6682 source location to the destination location.</p>
6684 <p>Note that, unlike the standard libc function, the <tt>llvm.memcpy.*</tt>
6685 intrinsics do not return a value, takes extra alignment/isvolatile arguments
6686 and the pointers can be in specified address spaces.</p>
6690 <p>The first argument is a pointer to the destination, the second is a pointer
6691 to the source. The third argument is an integer argument specifying the
6692 number of bytes to copy, the fourth argument is the alignment of the
6693 source and destination locations, and the fifth is a boolean indicating a
6694 volatile access.</p>
6696 <p>If the call to this intrinsic has an alignment value that is not 0 or 1,
6697 then the caller guarantees that both the source and destination pointers are
6698 aligned to that boundary.</p>
6700 <p>If the <tt>isvolatile</tt> parameter is <tt>true</tt>, the
6701 <tt>llvm.memcpy</tt> call is a <a href="#volatile">volatile operation</a>.
6702 The detailed access behavior is not very cleanly specified and it is unwise
6703 to depend on it.</p>
6707 <p>The '<tt>llvm.memcpy.*</tt>' intrinsics copy a block of memory from the
6708 source location to the destination location, which are not allowed to
6709 overlap. It copies "len" bytes of memory over. If the argument is known to
6710 be aligned to some boundary, this can be specified as the fourth argument,
6711 otherwise it should be set to 0 or 1.</p>
6715 <!-- _______________________________________________________________________ -->
6717 <a name="int_memmove">'<tt>llvm.memmove</tt>' Intrinsic</a>
6723 <p>This is an overloaded intrinsic. You can use llvm.memmove on any integer bit
6724 width and for different address space. Not all targets support all bit
6728 declare void @llvm.memmove.p0i8.p0i8.i32(i8* <dest>, i8* <src>,
6729 i32 <len>, i32 <align>, i1 <isvolatile>)
6730 declare void @llvm.memmove.p0i8.p0i8.i64(i8* <dest>, i8* <src>,
6731 i64 <len>, i32 <align>, i1 <isvolatile>)
6735 <p>The '<tt>llvm.memmove.*</tt>' intrinsics move a block of memory from the
6736 source location to the destination location. It is similar to the
6737 '<tt>llvm.memcpy</tt>' intrinsic but allows the two memory locations to
6740 <p>Note that, unlike the standard libc function, the <tt>llvm.memmove.*</tt>
6741 intrinsics do not return a value, takes extra alignment/isvolatile arguments
6742 and the pointers can be in specified address spaces.</p>
6746 <p>The first argument is a pointer to the destination, the second is a pointer
6747 to the source. The third argument is an integer argument specifying the
6748 number of bytes to copy, the fourth argument is the alignment of the
6749 source and destination locations, and the fifth is a boolean indicating a
6750 volatile access.</p>
6752 <p>If the call to this intrinsic has an alignment value that is not 0 or 1,
6753 then the caller guarantees that the source and destination pointers are
6754 aligned to that boundary.</p>
6756 <p>If the <tt>isvolatile</tt> parameter is <tt>true</tt>, the
6757 <tt>llvm.memmove</tt> call is a <a href="#volatile">volatile operation</a>.
6758 The detailed access behavior is not very cleanly specified and it is unwise
6759 to depend on it.</p>
6763 <p>The '<tt>llvm.memmove.*</tt>' intrinsics copy a block of memory from the
6764 source location to the destination location, which may overlap. It copies
6765 "len" bytes of memory over. If the argument is known to be aligned to some
6766 boundary, this can be specified as the fourth argument, otherwise it should
6767 be set to 0 or 1.</p>
6771 <!-- _______________________________________________________________________ -->
6773 <a name="int_memset">'<tt>llvm.memset.*</tt>' Intrinsics</a>
6779 <p>This is an overloaded intrinsic. You can use llvm.memset on any integer bit
6780 width and for different address spaces. However, not all targets support all
6784 declare void @llvm.memset.p0i8.i32(i8* <dest>, i8 <val>,
6785 i32 <len>, i32 <align>, i1 <isvolatile>)
6786 declare void @llvm.memset.p0i8.i64(i8* <dest>, i8 <val>,
6787 i64 <len>, i32 <align>, i1 <isvolatile>)
6791 <p>The '<tt>llvm.memset.*</tt>' intrinsics fill a block of memory with a
6792 particular byte value.</p>
6794 <p>Note that, unlike the standard libc function, the <tt>llvm.memset</tt>
6795 intrinsic does not return a value and takes extra alignment/volatile
6796 arguments. Also, the destination can be in an arbitrary address space.</p>
6799 <p>The first argument is a pointer to the destination to fill, the second is the
6800 byte value with which to fill it, the third argument is an integer argument
6801 specifying the number of bytes to fill, and the fourth argument is the known
6802 alignment of the destination location.</p>
6804 <p>If the call to this intrinsic has an alignment value that is not 0 or 1,
6805 then the caller guarantees that the destination pointer is aligned to that
6808 <p>If the <tt>isvolatile</tt> parameter is <tt>true</tt>, the
6809 <tt>llvm.memset</tt> call is a <a href="#volatile">volatile operation</a>.
6810 The detailed access behavior is not very cleanly specified and it is unwise
6811 to depend on it.</p>
6814 <p>The '<tt>llvm.memset.*</tt>' intrinsics fill "len" bytes of memory starting
6815 at the destination location. If the argument is known to be aligned to some
6816 boundary, this can be specified as the fourth argument, otherwise it should
6817 be set to 0 or 1.</p>
6821 <!-- _______________________________________________________________________ -->
6823 <a name="int_sqrt">'<tt>llvm.sqrt.*</tt>' Intrinsic</a>
6829 <p>This is an overloaded intrinsic. You can use <tt>llvm.sqrt</tt> on any
6830 floating point or vector of floating point type. Not all targets support all
6834 declare float @llvm.sqrt.f32(float %Val)
6835 declare double @llvm.sqrt.f64(double %Val)
6836 declare x86_fp80 @llvm.sqrt.f80(x86_fp80 %Val)
6837 declare fp128 @llvm.sqrt.f128(fp128 %Val)
6838 declare ppc_fp128 @llvm.sqrt.ppcf128(ppc_fp128 %Val)
6842 <p>The '<tt>llvm.sqrt</tt>' intrinsics return the sqrt of the specified operand,
6843 returning the same value as the libm '<tt>sqrt</tt>' functions would.
6844 Unlike <tt>sqrt</tt> in libm, however, <tt>llvm.sqrt</tt> has undefined
6845 behavior for negative numbers other than -0.0 (which allows for better
6846 optimization, because there is no need to worry about errno being
6847 set). <tt>llvm.sqrt(-0.0)</tt> is defined to return -0.0 like IEEE sqrt.</p>
6850 <p>The argument and return value are floating point numbers of the same
6854 <p>This function returns the sqrt of the specified operand if it is a
6855 nonnegative floating point number.</p>
6859 <!-- _______________________________________________________________________ -->
6861 <a name="int_powi">'<tt>llvm.powi.*</tt>' Intrinsic</a>
6867 <p>This is an overloaded intrinsic. You can use <tt>llvm.powi</tt> on any
6868 floating point or vector of floating point type. Not all targets support all
6872 declare float @llvm.powi.f32(float %Val, i32 %power)
6873 declare double @llvm.powi.f64(double %Val, i32 %power)
6874 declare x86_fp80 @llvm.powi.f80(x86_fp80 %Val, i32 %power)
6875 declare fp128 @llvm.powi.f128(fp128 %Val, i32 %power)
6876 declare ppc_fp128 @llvm.powi.ppcf128(ppc_fp128 %Val, i32 %power)
6880 <p>The '<tt>llvm.powi.*</tt>' intrinsics return the first operand raised to the
6881 specified (positive or negative) power. The order of evaluation of
6882 multiplications is not defined. When a vector of floating point type is
6883 used, the second argument remains a scalar integer value.</p>
6886 <p>The second argument is an integer power, and the first is a value to raise to
6890 <p>This function returns the first value raised to the second power with an
6891 unspecified sequence of rounding operations.</p>
6895 <!-- _______________________________________________________________________ -->
6897 <a name="int_sin">'<tt>llvm.sin.*</tt>' Intrinsic</a>
6903 <p>This is an overloaded intrinsic. You can use <tt>llvm.sin</tt> on any
6904 floating point or vector of floating point type. Not all targets support all
6908 declare float @llvm.sin.f32(float %Val)
6909 declare double @llvm.sin.f64(double %Val)
6910 declare x86_fp80 @llvm.sin.f80(x86_fp80 %Val)
6911 declare fp128 @llvm.sin.f128(fp128 %Val)
6912 declare ppc_fp128 @llvm.sin.ppcf128(ppc_fp128 %Val)
6916 <p>The '<tt>llvm.sin.*</tt>' intrinsics return the sine of the operand.</p>
6919 <p>The argument and return value are floating point numbers of the same
6923 <p>This function returns the sine of the specified operand, returning the same
6924 values as the libm <tt>sin</tt> functions would, and handles error conditions
6925 in the same way.</p>
6929 <!-- _______________________________________________________________________ -->
6931 <a name="int_cos">'<tt>llvm.cos.*</tt>' Intrinsic</a>
6937 <p>This is an overloaded intrinsic. You can use <tt>llvm.cos</tt> on any
6938 floating point or vector of floating point type. Not all targets support all
6942 declare float @llvm.cos.f32(float %Val)
6943 declare double @llvm.cos.f64(double %Val)
6944 declare x86_fp80 @llvm.cos.f80(x86_fp80 %Val)
6945 declare fp128 @llvm.cos.f128(fp128 %Val)
6946 declare ppc_fp128 @llvm.cos.ppcf128(ppc_fp128 %Val)
6950 <p>The '<tt>llvm.cos.*</tt>' intrinsics return the cosine of the operand.</p>
6953 <p>The argument and return value are floating point numbers of the same
6957 <p>This function returns the cosine of the specified operand, returning the same
6958 values as the libm <tt>cos</tt> functions would, and handles error conditions
6959 in the same way.</p>
6963 <!-- _______________________________________________________________________ -->
6965 <a name="int_pow">'<tt>llvm.pow.*</tt>' Intrinsic</a>
6971 <p>This is an overloaded intrinsic. You can use <tt>llvm.pow</tt> on any
6972 floating point or vector of floating point type. Not all targets support all
6976 declare float @llvm.pow.f32(float %Val, float %Power)
6977 declare double @llvm.pow.f64(double %Val, double %Power)
6978 declare x86_fp80 @llvm.pow.f80(x86_fp80 %Val, x86_fp80 %Power)
6979 declare fp128 @llvm.pow.f128(fp128 %Val, fp128 %Power)
6980 declare ppc_fp128 @llvm.pow.ppcf128(ppc_fp128 %Val, ppc_fp128 Power)
6984 <p>The '<tt>llvm.pow.*</tt>' intrinsics return the first operand raised to the
6985 specified (positive or negative) power.</p>
6988 <p>The second argument is a floating point power, and the first is a value to
6989 raise to that power.</p>
6992 <p>This function returns the first value raised to the second power, returning
6993 the same values as the libm <tt>pow</tt> functions would, and handles error
6994 conditions in the same way.</p>
7000 <!-- _______________________________________________________________________ -->
7002 <a name="int_exp">'<tt>llvm.exp.*</tt>' Intrinsic</a>
7008 <p>This is an overloaded intrinsic. You can use <tt>llvm.exp</tt> on any
7009 floating point or vector of floating point type. Not all targets support all
7013 declare float @llvm.exp.f32(float %Val)
7014 declare double @llvm.exp.f64(double %Val)
7015 declare x86_fp80 @llvm.exp.f80(x86_fp80 %Val)
7016 declare fp128 @llvm.exp.f128(fp128 %Val)
7017 declare ppc_fp128 @llvm.exp.ppcf128(ppc_fp128 %Val)
7021 <p>The '<tt>llvm.exp.*</tt>' intrinsics perform the exp function.</p>
7024 <p>The argument and return value are floating point numbers of the same
7028 <p>This function returns the same values as the libm <tt>exp</tt> functions
7029 would, and handles error conditions in the same way.</p>
7033 <!-- _______________________________________________________________________ -->
7035 <a name="int_log">'<tt>llvm.log.*</tt>' Intrinsic</a>
7041 <p>This is an overloaded intrinsic. You can use <tt>llvm.log</tt> on any
7042 floating point or vector of floating point type. Not all targets support all
7046 declare float @llvm.log.f32(float %Val)
7047 declare double @llvm.log.f64(double %Val)
7048 declare x86_fp80 @llvm.log.f80(x86_fp80 %Val)
7049 declare fp128 @llvm.log.f128(fp128 %Val)
7050 declare ppc_fp128 @llvm.log.ppcf128(ppc_fp128 %Val)
7054 <p>The '<tt>llvm.log.*</tt>' intrinsics perform the log function.</p>
7057 <p>The argument and return value are floating point numbers of the same
7061 <p>This function returns the same values as the libm <tt>log</tt> functions
7062 would, and handles error conditions in the same way.</p>
7065 <a name="int_fma">'<tt>llvm.fma.*</tt>' Intrinsic</a>
7071 <p>This is an overloaded intrinsic. You can use <tt>llvm.fma</tt> on any
7072 floating point or vector of floating point type. Not all targets support all
7076 declare float @llvm.fma.f32(float %a, float %b, float %c)
7077 declare double @llvm.fma.f64(double %a, double %b, double %c)
7078 declare x86_fp80 @llvm.fma.f80(x86_fp80 %a, x86_fp80 %b, x86_fp80 %c)
7079 declare fp128 @llvm.fma.f128(fp128 %a, fp128 %b, fp128 %c)
7080 declare ppc_fp128 @llvm.fma.ppcf128(ppc_fp128 %a, ppc_fp128 %b, ppc_fp128 %c)
7084 <p>The '<tt>llvm.fma.*</tt>' intrinsics perform the fused multiply-add
7088 <p>The argument and return value are floating point numbers of the same
7092 <p>This function returns the same values as the libm <tt>fma</tt> functions
7097 <!-- ======================================================================= -->
7099 <a name="int_manip">Bit Manipulation Intrinsics</a>
7104 <p>LLVM provides intrinsics for a few important bit manipulation operations.
7105 These allow efficient code generation for some algorithms.</p>
7107 <!-- _______________________________________________________________________ -->
7109 <a name="int_bswap">'<tt>llvm.bswap.*</tt>' Intrinsics</a>
7115 <p>This is an overloaded intrinsic function. You can use bswap on any integer
7116 type that is an even number of bytes (i.e. BitWidth % 16 == 0).</p>
7119 declare i16 @llvm.bswap.i16(i16 <id>)
7120 declare i32 @llvm.bswap.i32(i32 <id>)
7121 declare i64 @llvm.bswap.i64(i64 <id>)
7125 <p>The '<tt>llvm.bswap</tt>' family of intrinsics is used to byte swap integer
7126 values with an even number of bytes (positive multiple of 16 bits). These
7127 are useful for performing operations on data that is not in the target's
7128 native byte order.</p>
7131 <p>The <tt>llvm.bswap.i16</tt> intrinsic returns an i16 value that has the high
7132 and low byte of the input i16 swapped. Similarly,
7133 the <tt>llvm.bswap.i32</tt> intrinsic returns an i32 value that has the four
7134 bytes of the input i32 swapped, so that if the input bytes are numbered 0, 1,
7135 2, 3 then the returned i32 will have its bytes in 3, 2, 1, 0 order.
7136 The <tt>llvm.bswap.i48</tt>, <tt>llvm.bswap.i64</tt> and other intrinsics
7137 extend this concept to additional even-byte lengths (6 bytes, 8 bytes and
7138 more, respectively).</p>
7142 <!-- _______________________________________________________________________ -->
7144 <a name="int_ctpop">'<tt>llvm.ctpop.*</tt>' Intrinsic</a>
7150 <p>This is an overloaded intrinsic. You can use llvm.ctpop on any integer bit
7151 width, or on any vector with integer elements. Not all targets support all
7152 bit widths or vector types, however.</p>
7155 declare i8 @llvm.ctpop.i8(i8 <src>)
7156 declare i16 @llvm.ctpop.i16(i16 <src>)
7157 declare i32 @llvm.ctpop.i32(i32 <src>)
7158 declare i64 @llvm.ctpop.i64(i64 <src>)
7159 declare i256 @llvm.ctpop.i256(i256 <src>)
7160 declare <2 x i32> @llvm.ctpop.v2i32(<2 x i32> <src>)
7164 <p>The '<tt>llvm.ctpop</tt>' family of intrinsics counts the number of bits set
7168 <p>The only argument is the value to be counted. The argument may be of any
7169 integer type, or a vector with integer elements.
7170 The return type must match the argument type.</p>
7173 <p>The '<tt>llvm.ctpop</tt>' intrinsic counts the 1's in a variable, or within each
7174 element of a vector.</p>
7178 <!-- _______________________________________________________________________ -->
7180 <a name="int_ctlz">'<tt>llvm.ctlz.*</tt>' Intrinsic</a>
7186 <p>This is an overloaded intrinsic. You can use <tt>llvm.ctlz</tt> on any
7187 integer bit width, or any vector whose elements are integers. Not all
7188 targets support all bit widths or vector types, however.</p>
7191 declare i8 @llvm.ctlz.i8 (i8 <src>)
7192 declare i16 @llvm.ctlz.i16(i16 <src>)
7193 declare i32 @llvm.ctlz.i32(i32 <src>)
7194 declare i64 @llvm.ctlz.i64(i64 <src>)
7195 declare i256 @llvm.ctlz.i256(i256 <src>)
7196 declare <2 x i32> @llvm.ctlz.v2i32(<2 x i32> <src;gt)
7200 <p>The '<tt>llvm.ctlz</tt>' family of intrinsic functions counts the number of
7201 leading zeros in a variable.</p>
7204 <p>The only argument is the value to be counted. The argument may be of any
7205 integer type, or any vector type with integer element type.
7206 The return type must match the argument type.</p>
7209 <p>The '<tt>llvm.ctlz</tt>' intrinsic counts the leading (most significant)
7210 zeros in a variable, or within each element of the vector if the operation
7211 is of vector type. If the src == 0 then the result is the size in bits of
7212 the type of src. For example, <tt>llvm.ctlz(i32 2) = 30</tt>.</p>
7216 <!-- _______________________________________________________________________ -->
7218 <a name="int_cttz">'<tt>llvm.cttz.*</tt>' Intrinsic</a>
7224 <p>This is an overloaded intrinsic. You can use <tt>llvm.cttz</tt> on any
7225 integer bit width, or any vector of integer elements. Not all targets
7226 support all bit widths or vector types, however.</p>
7229 declare i8 @llvm.cttz.i8 (i8 <src>)
7230 declare i16 @llvm.cttz.i16(i16 <src>)
7231 declare i32 @llvm.cttz.i32(i32 <src>)
7232 declare i64 @llvm.cttz.i64(i64 <src>)
7233 declare i256 @llvm.cttz.i256(i256 <src>)
7234 declase <2 x i32> @llvm.cttz.v2i32(<2 x i32> <src>)
7238 <p>The '<tt>llvm.cttz</tt>' family of intrinsic functions counts the number of
7242 <p>The only argument is the value to be counted. The argument may be of any
7243 integer type, or a vectory with integer element type.. The return type
7244 must match the argument type.</p>
7247 <p>The '<tt>llvm.cttz</tt>' intrinsic counts the trailing (least significant)
7248 zeros in a variable, or within each element of a vector.
7249 If the src == 0 then the result is the size in bits of
7250 the type of src. For example, <tt>llvm.cttz(2) = 1</tt>.</p>
7256 <!-- ======================================================================= -->
7258 <a name="int_overflow">Arithmetic with Overflow Intrinsics</a>
7263 <p>LLVM provides intrinsics for some arithmetic with overflow operations.</p>
7265 <!-- _______________________________________________________________________ -->
7267 <a name="int_sadd_overflow">
7268 '<tt>llvm.sadd.with.overflow.*</tt>' Intrinsics
7275 <p>This is an overloaded intrinsic. You can use <tt>llvm.sadd.with.overflow</tt>
7276 on any integer bit width.</p>
7279 declare {i16, i1} @llvm.sadd.with.overflow.i16(i16 %a, i16 %b)
7280 declare {i32, i1} @llvm.sadd.with.overflow.i32(i32 %a, i32 %b)
7281 declare {i64, i1} @llvm.sadd.with.overflow.i64(i64 %a, i64 %b)
7285 <p>The '<tt>llvm.sadd.with.overflow</tt>' family of intrinsic functions perform
7286 a signed addition of the two arguments, and indicate whether an overflow
7287 occurred during the signed summation.</p>
7290 <p>The arguments (%a and %b) and the first element of the result structure may
7291 be of integer types of any bit width, but they must have the same bit
7292 width. The second element of the result structure must be of
7293 type <tt>i1</tt>. <tt>%a</tt> and <tt>%b</tt> are the two values that will
7294 undergo signed addition.</p>
7297 <p>The '<tt>llvm.sadd.with.overflow</tt>' family of intrinsic functions perform
7298 a signed addition of the two variables. They return a structure — the
7299 first element of which is the signed summation, and the second element of
7300 which is a bit specifying if the signed summation resulted in an
7305 %res = call {i32, i1} @llvm.sadd.with.overflow.i32(i32 %a, i32 %b)
7306 %sum = extractvalue {i32, i1} %res, 0
7307 %obit = extractvalue {i32, i1} %res, 1
7308 br i1 %obit, label %overflow, label %normal
7313 <!-- _______________________________________________________________________ -->
7315 <a name="int_uadd_overflow">
7316 '<tt>llvm.uadd.with.overflow.*</tt>' Intrinsics
7323 <p>This is an overloaded intrinsic. You can use <tt>llvm.uadd.with.overflow</tt>
7324 on any integer bit width.</p>
7327 declare {i16, i1} @llvm.uadd.with.overflow.i16(i16 %a, i16 %b)
7328 declare {i32, i1} @llvm.uadd.with.overflow.i32(i32 %a, i32 %b)
7329 declare {i64, i1} @llvm.uadd.with.overflow.i64(i64 %a, i64 %b)
7333 <p>The '<tt>llvm.uadd.with.overflow</tt>' family of intrinsic functions perform
7334 an unsigned addition of the two arguments, and indicate whether a carry
7335 occurred during the unsigned summation.</p>
7338 <p>The arguments (%a and %b) and the first element of the result structure may
7339 be of integer types of any bit width, but they must have the same bit
7340 width. The second element of the result structure must be of
7341 type <tt>i1</tt>. <tt>%a</tt> and <tt>%b</tt> are the two values that will
7342 undergo unsigned addition.</p>
7345 <p>The '<tt>llvm.uadd.with.overflow</tt>' family of intrinsic functions perform
7346 an unsigned addition of the two arguments. They return a structure —
7347 the first element of which is the sum, and the second element of which is a
7348 bit specifying if the unsigned summation resulted in a carry.</p>
7352 %res = call {i32, i1} @llvm.uadd.with.overflow.i32(i32 %a, i32 %b)
7353 %sum = extractvalue {i32, i1} %res, 0
7354 %obit = extractvalue {i32, i1} %res, 1
7355 br i1 %obit, label %carry, label %normal
7360 <!-- _______________________________________________________________________ -->
7362 <a name="int_ssub_overflow">
7363 '<tt>llvm.ssub.with.overflow.*</tt>' Intrinsics
7370 <p>This is an overloaded intrinsic. You can use <tt>llvm.ssub.with.overflow</tt>
7371 on any integer bit width.</p>
7374 declare {i16, i1} @llvm.ssub.with.overflow.i16(i16 %a, i16 %b)
7375 declare {i32, i1} @llvm.ssub.with.overflow.i32(i32 %a, i32 %b)
7376 declare {i64, i1} @llvm.ssub.with.overflow.i64(i64 %a, i64 %b)
7380 <p>The '<tt>llvm.ssub.with.overflow</tt>' family of intrinsic functions perform
7381 a signed subtraction of the two arguments, and indicate whether an overflow
7382 occurred during the signed subtraction.</p>
7385 <p>The arguments (%a and %b) and the first element of the result structure may
7386 be of integer types of any bit width, but they must have the same bit
7387 width. The second element of the result structure must be of
7388 type <tt>i1</tt>. <tt>%a</tt> and <tt>%b</tt> are the two values that will
7389 undergo signed subtraction.</p>
7392 <p>The '<tt>llvm.ssub.with.overflow</tt>' family of intrinsic functions perform
7393 a signed subtraction of the two arguments. They return a structure —
7394 the first element of which is the subtraction, and the second element of
7395 which is a bit specifying if the signed subtraction resulted in an
7400 %res = call {i32, i1} @llvm.ssub.with.overflow.i32(i32 %a, i32 %b)
7401 %sum = extractvalue {i32, i1} %res, 0
7402 %obit = extractvalue {i32, i1} %res, 1
7403 br i1 %obit, label %overflow, label %normal
7408 <!-- _______________________________________________________________________ -->
7410 <a name="int_usub_overflow">
7411 '<tt>llvm.usub.with.overflow.*</tt>' Intrinsics
7418 <p>This is an overloaded intrinsic. You can use <tt>llvm.usub.with.overflow</tt>
7419 on any integer bit width.</p>
7422 declare {i16, i1} @llvm.usub.with.overflow.i16(i16 %a, i16 %b)
7423 declare {i32, i1} @llvm.usub.with.overflow.i32(i32 %a, i32 %b)
7424 declare {i64, i1} @llvm.usub.with.overflow.i64(i64 %a, i64 %b)
7428 <p>The '<tt>llvm.usub.with.overflow</tt>' family of intrinsic functions perform
7429 an unsigned subtraction of the two arguments, and indicate whether an
7430 overflow occurred during the unsigned subtraction.</p>
7433 <p>The arguments (%a and %b) and the first element of the result structure may
7434 be of integer types of any bit width, but they must have the same bit
7435 width. The second element of the result structure must be of
7436 type <tt>i1</tt>. <tt>%a</tt> and <tt>%b</tt> are the two values that will
7437 undergo unsigned subtraction.</p>
7440 <p>The '<tt>llvm.usub.with.overflow</tt>' family of intrinsic functions perform
7441 an unsigned subtraction of the two arguments. They return a structure —
7442 the first element of which is the subtraction, and the second element of
7443 which is a bit specifying if the unsigned subtraction resulted in an
7448 %res = call {i32, i1} @llvm.usub.with.overflow.i32(i32 %a, i32 %b)
7449 %sum = extractvalue {i32, i1} %res, 0
7450 %obit = extractvalue {i32, i1} %res, 1
7451 br i1 %obit, label %overflow, label %normal
7456 <!-- _______________________________________________________________________ -->
7458 <a name="int_smul_overflow">
7459 '<tt>llvm.smul.with.overflow.*</tt>' Intrinsics
7466 <p>This is an overloaded intrinsic. You can use <tt>llvm.smul.with.overflow</tt>
7467 on any integer bit width.</p>
7470 declare {i16, i1} @llvm.smul.with.overflow.i16(i16 %a, i16 %b)
7471 declare {i32, i1} @llvm.smul.with.overflow.i32(i32 %a, i32 %b)
7472 declare {i64, i1} @llvm.smul.with.overflow.i64(i64 %a, i64 %b)
7477 <p>The '<tt>llvm.smul.with.overflow</tt>' family of intrinsic functions perform
7478 a signed multiplication of the two arguments, and indicate whether an
7479 overflow occurred during the signed multiplication.</p>
7482 <p>The arguments (%a and %b) and the first element of the result structure may
7483 be of integer types of any bit width, but they must have the same bit
7484 width. The second element of the result structure must be of
7485 type <tt>i1</tt>. <tt>%a</tt> and <tt>%b</tt> are the two values that will
7486 undergo signed multiplication.</p>
7489 <p>The '<tt>llvm.smul.with.overflow</tt>' family of intrinsic functions perform
7490 a signed multiplication of the two arguments. They return a structure —
7491 the first element of which is the multiplication, and the second element of
7492 which is a bit specifying if the signed multiplication resulted in an
7497 %res = call {i32, i1} @llvm.smul.with.overflow.i32(i32 %a, i32 %b)
7498 %sum = extractvalue {i32, i1} %res, 0
7499 %obit = extractvalue {i32, i1} %res, 1
7500 br i1 %obit, label %overflow, label %normal
7505 <!-- _______________________________________________________________________ -->
7507 <a name="int_umul_overflow">
7508 '<tt>llvm.umul.with.overflow.*</tt>' Intrinsics
7515 <p>This is an overloaded intrinsic. You can use <tt>llvm.umul.with.overflow</tt>
7516 on any integer bit width.</p>
7519 declare {i16, i1} @llvm.umul.with.overflow.i16(i16 %a, i16 %b)
7520 declare {i32, i1} @llvm.umul.with.overflow.i32(i32 %a, i32 %b)
7521 declare {i64, i1} @llvm.umul.with.overflow.i64(i64 %a, i64 %b)
7525 <p>The '<tt>llvm.umul.with.overflow</tt>' family of intrinsic functions perform
7526 a unsigned multiplication of the two arguments, and indicate whether an
7527 overflow occurred during the unsigned multiplication.</p>
7530 <p>The arguments (%a and %b) and the first element of the result structure may
7531 be of integer types of any bit width, but they must have the same bit
7532 width. The second element of the result structure must be of
7533 type <tt>i1</tt>. <tt>%a</tt> and <tt>%b</tt> are the two values that will
7534 undergo unsigned multiplication.</p>
7537 <p>The '<tt>llvm.umul.with.overflow</tt>' family of intrinsic functions perform
7538 an unsigned multiplication of the two arguments. They return a structure
7539 — the first element of which is the multiplication, and the second
7540 element of which is a bit specifying if the unsigned multiplication resulted
7545 %res = call {i32, i1} @llvm.umul.with.overflow.i32(i32 %a, i32 %b)
7546 %sum = extractvalue {i32, i1} %res, 0
7547 %obit = extractvalue {i32, i1} %res, 1
7548 br i1 %obit, label %overflow, label %normal
7555 <!-- ======================================================================= -->
7557 <a name="int_fp16">Half Precision Floating Point Intrinsics</a>
7562 <p>Half precision floating point is a storage-only format. This means that it is
7563 a dense encoding (in memory) but does not support computation in the
7566 <p>This means that code must first load the half-precision floating point
7567 value as an i16, then convert it to float with <a
7568 href="#int_convert_from_fp16"><tt>llvm.convert.from.fp16</tt></a>.
7569 Computation can then be performed on the float value (including extending to
7570 double etc). To store the value back to memory, it is first converted to
7571 float if needed, then converted to i16 with
7572 <a href="#int_convert_to_fp16"><tt>llvm.convert.to.fp16</tt></a>, then
7573 storing as an i16 value.</p>
7575 <!-- _______________________________________________________________________ -->
7577 <a name="int_convert_to_fp16">
7578 '<tt>llvm.convert.to.fp16</tt>' Intrinsic
7586 declare i16 @llvm.convert.to.fp16(f32 %a)
7590 <p>The '<tt>llvm.convert.to.fp16</tt>' intrinsic function performs
7591 a conversion from single precision floating point format to half precision
7592 floating point format.</p>
7595 <p>The intrinsic function contains single argument - the value to be
7599 <p>The '<tt>llvm.convert.to.fp16</tt>' intrinsic function performs
7600 a conversion from single precision floating point format to half precision
7601 floating point format. The return value is an <tt>i16</tt> which
7602 contains the converted number.</p>
7606 %res = call i16 @llvm.convert.to.fp16(f32 %a)
7607 store i16 %res, i16* @x, align 2
7612 <!-- _______________________________________________________________________ -->
7614 <a name="int_convert_from_fp16">
7615 '<tt>llvm.convert.from.fp16</tt>' Intrinsic
7623 declare f32 @llvm.convert.from.fp16(i16 %a)
7627 <p>The '<tt>llvm.convert.from.fp16</tt>' intrinsic function performs
7628 a conversion from half precision floating point format to single precision
7629 floating point format.</p>
7632 <p>The intrinsic function contains single argument - the value to be
7636 <p>The '<tt>llvm.convert.from.fp16</tt>' intrinsic function performs a
7637 conversion from half single precision floating point format to single
7638 precision floating point format. The input half-float value is represented by
7639 an <tt>i16</tt> value.</p>
7643 %a = load i16* @x, align 2
7644 %res = call f32 @llvm.convert.from.fp16(i16 %a)
7651 <!-- ======================================================================= -->
7653 <a name="int_debugger">Debugger Intrinsics</a>
7658 <p>The LLVM debugger intrinsics (which all start with <tt>llvm.dbg.</tt>
7659 prefix), are described in
7660 the <a href="SourceLevelDebugging.html#format_common_intrinsics">LLVM Source
7661 Level Debugging</a> document.</p>
7665 <!-- ======================================================================= -->
7667 <a name="int_eh">Exception Handling Intrinsics</a>
7672 <p>The LLVM exception handling intrinsics (which all start with
7673 <tt>llvm.eh.</tt> prefix), are described in
7674 the <a href="ExceptionHandling.html#format_common_intrinsics">LLVM Exception
7675 Handling</a> document.</p>
7679 <!-- ======================================================================= -->
7681 <a name="int_trampoline">Trampoline Intrinsics</a>
7686 <p>These intrinsics make it possible to excise one parameter, marked with
7687 the <a href="#nest"><tt>nest</tt></a> attribute, from a function.
7688 The result is a callable
7689 function pointer lacking the nest parameter - the caller does not need to
7690 provide a value for it. Instead, the value to use is stored in advance in a
7691 "trampoline", a block of memory usually allocated on the stack, which also
7692 contains code to splice the nest value into the argument list. This is used
7693 to implement the GCC nested function address extension.</p>
7695 <p>For example, if the function is
7696 <tt>i32 f(i8* nest %c, i32 %x, i32 %y)</tt> then the resulting function
7697 pointer has signature <tt>i32 (i32, i32)*</tt>. It can be created as
7700 <pre class="doc_code">
7701 %tramp = alloca [10 x i8], align 4 ; size and alignment only correct for X86
7702 %tramp1 = getelementptr [10 x i8]* %tramp, i32 0, i32 0
7703 call i8* @llvm.init.trampoline(i8* %tramp1, i8* bitcast (i32 (i8*, i32, i32)* @f to i8*), i8* %nval)
7704 %p = call i8* @llvm.adjust.trampoline(i8* %tramp1)
7705 %fp = bitcast i8* %p to i32 (i32, i32)*
7708 <p>The call <tt>%val = call i32 %fp(i32 %x, i32 %y)</tt> is then equivalent
7709 to <tt>%val = call i32 %f(i8* %nval, i32 %x, i32 %y)</tt>.</p>
7711 <!-- _______________________________________________________________________ -->
7714 '<tt>llvm.init.trampoline</tt>' Intrinsic
7722 declare void @llvm.init.trampoline(i8* <tramp>, i8* <func>, i8* <nval>)
7726 <p>This fills the memory pointed to by <tt>tramp</tt> with executable code,
7727 turning it into a trampoline.</p>
7730 <p>The <tt>llvm.init.trampoline</tt> intrinsic takes three arguments, all
7731 pointers. The <tt>tramp</tt> argument must point to a sufficiently large and
7732 sufficiently aligned block of memory; this memory is written to by the
7733 intrinsic. Note that the size and the alignment are target-specific - LLVM
7734 currently provides no portable way of determining them, so a front-end that
7735 generates this intrinsic needs to have some target-specific knowledge.
7736 The <tt>func</tt> argument must hold a function bitcast to
7737 an <tt>i8*</tt>.</p>
7740 <p>The block of memory pointed to by <tt>tramp</tt> is filled with target
7741 dependent code, turning it into a function. Then <tt>tramp</tt> needs to be
7742 passed to <a href="#int_at">llvm.adjust.trampoline</a> to get a pointer
7743 which can be <a href="#int_trampoline">bitcast (to a new function) and
7744 called</a>. The new function's signature is the same as that of
7745 <tt>func</tt> with any arguments marked with the <tt>nest</tt> attribute
7746 removed. At most one such <tt>nest</tt> argument is allowed, and it must be of
7747 pointer type. Calling the new function is equivalent to calling <tt>func</tt>
7748 with the same argument list, but with <tt>nval</tt> used for the missing
7749 <tt>nest</tt> argument. If, after calling <tt>llvm.init.trampoline</tt>, the
7750 memory pointed to by <tt>tramp</tt> is modified, then the effect of any later call
7751 to the returned function pointer is undefined.</p>
7754 <!-- _______________________________________________________________________ -->
7757 '<tt>llvm.adjust.trampoline</tt>' Intrinsic
7765 declare i8* @llvm.adjust.trampoline(i8* <tramp>)
7769 <p>This performs any required machine-specific adjustment to the address of a
7770 trampoline (passed as <tt>tramp</tt>).</p>
7773 <p><tt>tramp</tt> must point to a block of memory which already has trampoline code
7774 filled in by a previous call to <a href="#int_it"><tt>llvm.init.trampoline</tt>
7778 <p>On some architectures the address of the code to be executed needs to be
7779 different to the address where the trampoline is actually stored. This
7780 intrinsic returns the executable address corresponding to <tt>tramp</tt>
7781 after performing the required machine specific adjustments.
7782 The pointer returned can then be <a href="#int_trampoline"> bitcast and
7790 <!-- ======================================================================= -->
7792 <a name="int_atomics">Atomic Operations and Synchronization Intrinsics</a>
7797 <p>These intrinsic functions expand the "universal IR" of LLVM to represent
7798 hardware constructs for atomic operations and memory synchronization. This
7799 provides an interface to the hardware, not an interface to the programmer. It
7800 is aimed at a low enough level to allow any programming models or APIs
7801 (Application Programming Interfaces) which need atomic behaviors to map
7802 cleanly onto it. It is also modeled primarily on hardware behavior. Just as
7803 hardware provides a "universal IR" for source languages, it also provides a
7804 starting point for developing a "universal" atomic operation and
7805 synchronization IR.</p>
7807 <p>These do <em>not</em> form an API such as high-level threading libraries,
7808 software transaction memory systems, atomic primitives, and intrinsic
7809 functions as found in BSD, GNU libc, atomic_ops, APR, and other system and
7810 application libraries. The hardware interface provided by LLVM should allow
7811 a clean implementation of all of these APIs and parallel programming models.
7812 No one model or paradigm should be selected above others unless the hardware
7813 itself ubiquitously does so.</p>
7815 <!-- _______________________________________________________________________ -->
7817 <a name="int_memory_barrier">'<tt>llvm.memory.barrier</tt>' Intrinsic</a>
7823 declare void @llvm.memory.barrier(i1 <ll>, i1 <ls>, i1 <sl>, i1 <ss>, i1 <device>)
7827 <p>The <tt>llvm.memory.barrier</tt> intrinsic guarantees ordering between
7828 specific pairs of memory access types.</p>
7831 <p>The <tt>llvm.memory.barrier</tt> intrinsic requires five boolean arguments.
7832 The first four arguments enables a specific barrier as listed below. The
7833 fifth argument specifies that the barrier applies to io or device or uncached
7837 <li><tt>ll</tt>: load-load barrier</li>
7838 <li><tt>ls</tt>: load-store barrier</li>
7839 <li><tt>sl</tt>: store-load barrier</li>
7840 <li><tt>ss</tt>: store-store barrier</li>
7841 <li><tt>device</tt>: barrier applies to device and uncached memory also.</li>
7845 <p>This intrinsic causes the system to enforce some ordering constraints upon
7846 the loads and stores of the program. This barrier does not
7847 indicate <em>when</em> any events will occur, it only enforces
7848 an <em>order</em> in which they occur. For any of the specified pairs of load
7849 and store operations (f.ex. load-load, or store-load), all of the first
7850 operations preceding the barrier will complete before any of the second
7851 operations succeeding the barrier begin. Specifically the semantics for each
7852 pairing is as follows:</p>
7855 <li><tt>ll</tt>: All loads before the barrier must complete before any load
7856 after the barrier begins.</li>
7857 <li><tt>ls</tt>: All loads before the barrier must complete before any
7858 store after the barrier begins.</li>
7859 <li><tt>ss</tt>: All stores before the barrier must complete before any
7860 store after the barrier begins.</li>
7861 <li><tt>sl</tt>: All stores before the barrier must complete before any
7862 load after the barrier begins.</li>
7865 <p>These semantics are applied with a logical "and" behavior when more than one
7866 is enabled in a single memory barrier intrinsic.</p>
7868 <p>Backends may implement stronger barriers than those requested when they do
7869 not support as fine grained a barrier as requested. Some architectures do
7870 not need all types of barriers and on such architectures, these become
7875 %mallocP = tail call i8* @malloc(i32 ptrtoint (i32* getelementptr (i32* null, i32 1) to i32))
7876 %ptr = bitcast i8* %mallocP to i32*
7879 %result1 = load i32* %ptr <i>; yields {i32}:result1 = 4</i>
7880 call void @llvm.memory.barrier(i1 false, i1 true, i1 false, i1 false, i1 true)
7881 <i>; guarantee the above finishes</i>
7882 store i32 8, %ptr <i>; before this begins</i>
7887 <!-- _______________________________________________________________________ -->
7889 <a name="int_atomic_cmp_swap">'<tt>llvm.atomic.cmp.swap.*</tt>' Intrinsic</a>
7895 <p>This is an overloaded intrinsic. You can use <tt>llvm.atomic.cmp.swap</tt> on
7896 any integer bit width and for different address spaces. Not all targets
7897 support all bit widths however.</p>
7900 declare i8 @llvm.atomic.cmp.swap.i8.p0i8(i8* <ptr>, i8 <cmp>, i8 <val>)
7901 declare i16 @llvm.atomic.cmp.swap.i16.p0i16(i16* <ptr>, i16 <cmp>, i16 <val>)
7902 declare i32 @llvm.atomic.cmp.swap.i32.p0i32(i32* <ptr>, i32 <cmp>, i32 <val>)
7903 declare i64 @llvm.atomic.cmp.swap.i64.p0i64(i64* <ptr>, i64 <cmp>, i64 <val>)
7907 <p>This loads a value in memory and compares it to a given value. If they are
7908 equal, it stores a new value into the memory.</p>
7911 <p>The <tt>llvm.atomic.cmp.swap</tt> intrinsic takes three arguments. The result
7912 as well as both <tt>cmp</tt> and <tt>val</tt> must be integer values with the
7913 same bit width. The <tt>ptr</tt> argument must be a pointer to a value of
7914 this integer type. While any bit width integer may be used, targets may only
7915 lower representations they support in hardware.</p>
7918 <p>This entire intrinsic must be executed atomically. It first loads the value
7919 in memory pointed to by <tt>ptr</tt> and compares it with the
7920 value <tt>cmp</tt>. If they are equal, <tt>val</tt> is stored into the
7921 memory. The loaded value is yielded in all cases. This provides the
7922 equivalent of an atomic compare-and-swap operation within the SSA
7927 %mallocP = tail call i8* @malloc(i32 ptrtoint (i32* getelementptr (i32* null, i32 1) to i32))
7928 %ptr = bitcast i8* %mallocP to i32*
7931 %val1 = add i32 4, 4
7932 %result1 = call i32 @llvm.atomic.cmp.swap.i32.p0i32(i32* %ptr, i32 4, %val1)
7933 <i>; yields {i32}:result1 = 4</i>
7934 %stored1 = icmp eq i32 %result1, 4 <i>; yields {i1}:stored1 = true</i>
7935 %memval1 = load i32* %ptr <i>; yields {i32}:memval1 = 8</i>
7937 %val2 = add i32 1, 1
7938 %result2 = call i32 @llvm.atomic.cmp.swap.i32.p0i32(i32* %ptr, i32 5, %val2)
7939 <i>; yields {i32}:result2 = 8</i>
7940 %stored2 = icmp eq i32 %result2, 5 <i>; yields {i1}:stored2 = false</i>
7942 %memval2 = load i32* %ptr <i>; yields {i32}:memval2 = 8</i>
7947 <!-- _______________________________________________________________________ -->
7949 <a name="int_atomic_swap">'<tt>llvm.atomic.swap.*</tt>' Intrinsic</a>
7955 <p>This is an overloaded intrinsic. You can use <tt>llvm.atomic.swap</tt> on any
7956 integer bit width. Not all targets support all bit widths however.</p>
7959 declare i8 @llvm.atomic.swap.i8.p0i8(i8* <ptr>, i8 <val>)
7960 declare i16 @llvm.atomic.swap.i16.p0i16(i16* <ptr>, i16 <val>)
7961 declare i32 @llvm.atomic.swap.i32.p0i32(i32* <ptr>, i32 <val>)
7962 declare i64 @llvm.atomic.swap.i64.p0i64(i64* <ptr>, i64 <val>)
7966 <p>This intrinsic loads the value stored in memory at <tt>ptr</tt> and yields
7967 the value from memory. It then stores the value in <tt>val</tt> in the memory
7968 at <tt>ptr</tt>.</p>
7971 <p>The <tt>llvm.atomic.swap</tt> intrinsic takes two arguments. Both
7972 the <tt>val</tt> argument and the result must be integers of the same bit
7973 width. The first argument, <tt>ptr</tt>, must be a pointer to a value of this
7974 integer type. The targets may only lower integer representations they
7978 <p>This intrinsic loads the value pointed to by <tt>ptr</tt>, yields it, and
7979 stores <tt>val</tt> back into <tt>ptr</tt> atomically. This provides the
7980 equivalent of an atomic swap operation within the SSA framework.</p>
7984 %mallocP = tail call i8* @malloc(i32 ptrtoint (i32* getelementptr (i32* null, i32 1) to i32))
7985 %ptr = bitcast i8* %mallocP to i32*
7988 %val1 = add i32 4, 4
7989 %result1 = call i32 @llvm.atomic.swap.i32.p0i32(i32* %ptr, i32 %val1)
7990 <i>; yields {i32}:result1 = 4</i>
7991 %stored1 = icmp eq i32 %result1, 4 <i>; yields {i1}:stored1 = true</i>
7992 %memval1 = load i32* %ptr <i>; yields {i32}:memval1 = 8</i>
7994 %val2 = add i32 1, 1
7995 %result2 = call i32 @llvm.atomic.swap.i32.p0i32(i32* %ptr, i32 %val2)
7996 <i>; yields {i32}:result2 = 8</i>
7998 %stored2 = icmp eq i32 %result2, 8 <i>; yields {i1}:stored2 = true</i>
7999 %memval2 = load i32* %ptr <i>; yields {i32}:memval2 = 2</i>
8004 <!-- _______________________________________________________________________ -->
8006 <a name="int_atomic_load_add">'<tt>llvm.atomic.load.add.*</tt>' Intrinsic</a>
8012 <p>This is an overloaded intrinsic. You can use <tt>llvm.atomic.load.add</tt> on
8013 any integer bit width. Not all targets support all bit widths however.</p>
8016 declare i8 @llvm.atomic.load.add.i8.p0i8(i8* <ptr>, i8 <delta>)
8017 declare i16 @llvm.atomic.load.add.i16.p0i16(i16* <ptr>, i16 <delta>)
8018 declare i32 @llvm.atomic.load.add.i32.p0i32(i32* <ptr>, i32 <delta>)
8019 declare i64 @llvm.atomic.load.add.i64.p0i64(i64* <ptr>, i64 <delta>)
8023 <p>This intrinsic adds <tt>delta</tt> to the value stored in memory
8024 at <tt>ptr</tt>. It yields the original value at <tt>ptr</tt>.</p>
8027 <p>The intrinsic takes two arguments, the first a pointer to an integer value
8028 and the second an integer value. The result is also an integer value. These
8029 integer types can have any bit width, but they must all have the same bit
8030 width. The targets may only lower integer representations they support.</p>
8033 <p>This intrinsic does a series of operations atomically. It first loads the
8034 value stored at <tt>ptr</tt>. It then adds <tt>delta</tt>, stores the result
8035 to <tt>ptr</tt>. It yields the original value stored at <tt>ptr</tt>.</p>
8039 %mallocP = tail call i8* @malloc(i32 ptrtoint (i32* getelementptr (i32* null, i32 1) to i32))
8040 %ptr = bitcast i8* %mallocP to i32*
8042 %result1 = call i32 @llvm.atomic.load.add.i32.p0i32(i32* %ptr, i32 4)
8043 <i>; yields {i32}:result1 = 4</i>
8044 %result2 = call i32 @llvm.atomic.load.add.i32.p0i32(i32* %ptr, i32 2)
8045 <i>; yields {i32}:result2 = 8</i>
8046 %result3 = call i32 @llvm.atomic.load.add.i32.p0i32(i32* %ptr, i32 5)
8047 <i>; yields {i32}:result3 = 10</i>
8048 %memval1 = load i32* %ptr <i>; yields {i32}:memval1 = 15</i>
8053 <!-- _______________________________________________________________________ -->
8055 <a name="int_atomic_load_sub">'<tt>llvm.atomic.load.sub.*</tt>' Intrinsic</a>
8061 <p>This is an overloaded intrinsic. You can use <tt>llvm.atomic.load.sub</tt> on
8062 any integer bit width and for different address spaces. Not all targets
8063 support all bit widths however.</p>
8066 declare i8 @llvm.atomic.load.sub.i8.p0i32(i8* <ptr>, i8 <delta>)
8067 declare i16 @llvm.atomic.load.sub.i16.p0i32(i16* <ptr>, i16 <delta>)
8068 declare i32 @llvm.atomic.load.sub.i32.p0i32(i32* <ptr>, i32 <delta>)
8069 declare i64 @llvm.atomic.load.sub.i64.p0i32(i64* <ptr>, i64 <delta>)
8073 <p>This intrinsic subtracts <tt>delta</tt> to the value stored in memory at
8074 <tt>ptr</tt>. It yields the original value at <tt>ptr</tt>.</p>
8077 <p>The intrinsic takes two arguments, the first a pointer to an integer value
8078 and the second an integer value. The result is also an integer value. These
8079 integer types can have any bit width, but they must all have the same bit
8080 width. The targets may only lower integer representations they support.</p>
8083 <p>This intrinsic does a series of operations atomically. It first loads the
8084 value stored at <tt>ptr</tt>. It then subtracts <tt>delta</tt>, stores the
8085 result to <tt>ptr</tt>. It yields the original value stored
8086 at <tt>ptr</tt>.</p>
8090 %mallocP = tail call i8* @malloc(i32 ptrtoint (i32* getelementptr (i32* null, i32 1) to i32))
8091 %ptr = bitcast i8* %mallocP to i32*
8093 %result1 = call i32 @llvm.atomic.load.sub.i32.p0i32(i32* %ptr, i32 4)
8094 <i>; yields {i32}:result1 = 8</i>
8095 %result2 = call i32 @llvm.atomic.load.sub.i32.p0i32(i32* %ptr, i32 2)
8096 <i>; yields {i32}:result2 = 4</i>
8097 %result3 = call i32 @llvm.atomic.load.sub.i32.p0i32(i32* %ptr, i32 5)
8098 <i>; yields {i32}:result3 = 2</i>
8099 %memval1 = load i32* %ptr <i>; yields {i32}:memval1 = -3</i>
8104 <!-- _______________________________________________________________________ -->
8106 <a name="int_atomic_load_and">
8107 '<tt>llvm.atomic.load.and.*</tt>' Intrinsic
8110 <a name="int_atomic_load_nand">
8111 '<tt>llvm.atomic.load.nand.*</tt>' Intrinsic
8114 <a name="int_atomic_load_or">
8115 '<tt>llvm.atomic.load.or.*</tt>' Intrinsic
8118 <a name="int_atomic_load_xor">
8119 '<tt>llvm.atomic.load.xor.*</tt>' Intrinsic
8126 <p>These are overloaded intrinsics. You can
8127 use <tt>llvm.atomic.load_and</tt>, <tt>llvm.atomic.load_nand</tt>,
8128 <tt>llvm.atomic.load_or</tt>, and <tt>llvm.atomic.load_xor</tt> on any integer
8129 bit width and for different address spaces. Not all targets support all bit
8133 declare i8 @llvm.atomic.load.and.i8.p0i8(i8* <ptr>, i8 <delta>)
8134 declare i16 @llvm.atomic.load.and.i16.p0i16(i16* <ptr>, i16 <delta>)
8135 declare i32 @llvm.atomic.load.and.i32.p0i32(i32* <ptr>, i32 <delta>)
8136 declare i64 @llvm.atomic.load.and.i64.p0i64(i64* <ptr>, i64 <delta>)
8140 declare i8 @llvm.atomic.load.or.i8.p0i8(i8* <ptr>, i8 <delta>)
8141 declare i16 @llvm.atomic.load.or.i16.p0i16(i16* <ptr>, i16 <delta>)
8142 declare i32 @llvm.atomic.load.or.i32.p0i32(i32* <ptr>, i32 <delta>)
8143 declare i64 @llvm.atomic.load.or.i64.p0i64(i64* <ptr>, i64 <delta>)
8147 declare i8 @llvm.atomic.load.nand.i8.p0i32(i8* <ptr>, i8 <delta>)
8148 declare i16 @llvm.atomic.load.nand.i16.p0i32(i16* <ptr>, i16 <delta>)
8149 declare i32 @llvm.atomic.load.nand.i32.p0i32(i32* <ptr>, i32 <delta>)
8150 declare i64 @llvm.atomic.load.nand.i64.p0i32(i64* <ptr>, i64 <delta>)
8154 declare i8 @llvm.atomic.load.xor.i8.p0i32(i8* <ptr>, i8 <delta>)
8155 declare i16 @llvm.atomic.load.xor.i16.p0i32(i16* <ptr>, i16 <delta>)
8156 declare i32 @llvm.atomic.load.xor.i32.p0i32(i32* <ptr>, i32 <delta>)
8157 declare i64 @llvm.atomic.load.xor.i64.p0i32(i64* <ptr>, i64 <delta>)
8161 <p>These intrinsics bitwise the operation (and, nand, or, xor) <tt>delta</tt> to
8162 the value stored in memory at <tt>ptr</tt>. It yields the original value
8163 at <tt>ptr</tt>.</p>
8166 <p>These intrinsics take two arguments, the first a pointer to an integer value
8167 and the second an integer value. The result is also an integer value. These
8168 integer types can have any bit width, but they must all have the same bit
8169 width. The targets may only lower integer representations they support.</p>
8172 <p>These intrinsics does a series of operations atomically. They first load the
8173 value stored at <tt>ptr</tt>. They then do the bitwise
8174 operation <tt>delta</tt>, store the result to <tt>ptr</tt>. They yield the
8175 original value stored at <tt>ptr</tt>.</p>
8179 %mallocP = tail call i8* @malloc(i32 ptrtoint (i32* getelementptr (i32* null, i32 1) to i32))
8180 %ptr = bitcast i8* %mallocP to i32*
8181 store i32 0x0F0F, %ptr
8182 %result0 = call i32 @llvm.atomic.load.nand.i32.p0i32(i32* %ptr, i32 0xFF)
8183 <i>; yields {i32}:result0 = 0x0F0F</i>
8184 %result1 = call i32 @llvm.atomic.load.and.i32.p0i32(i32* %ptr, i32 0xFF)
8185 <i>; yields {i32}:result1 = 0xFFFFFFF0</i>
8186 %result2 = call i32 @llvm.atomic.load.or.i32.p0i32(i32* %ptr, i32 0F)
8187 <i>; yields {i32}:result2 = 0xF0</i>
8188 %result3 = call i32 @llvm.atomic.load.xor.i32.p0i32(i32* %ptr, i32 0F)
8189 <i>; yields {i32}:result3 = FF</i>
8190 %memval1 = load i32* %ptr <i>; yields {i32}:memval1 = F0</i>
8195 <!-- _______________________________________________________________________ -->
8197 <a name="int_atomic_load_max">
8198 '<tt>llvm.atomic.load.max.*</tt>' Intrinsic
8201 <a name="int_atomic_load_min">
8202 '<tt>llvm.atomic.load.min.*</tt>' Intrinsic
8205 <a name="int_atomic_load_umax">
8206 '<tt>llvm.atomic.load.umax.*</tt>' Intrinsic
8209 <a name="int_atomic_load_umin">
8210 '<tt>llvm.atomic.load.umin.*</tt>' Intrinsic
8217 <p>These are overloaded intrinsics. You can use <tt>llvm.atomic.load_max</tt>,
8218 <tt>llvm.atomic.load_min</tt>, <tt>llvm.atomic.load_umax</tt>, and
8219 <tt>llvm.atomic.load_umin</tt> on any integer bit width and for different
8220 address spaces. Not all targets support all bit widths however.</p>
8223 declare i8 @llvm.atomic.load.max.i8.p0i8(i8* <ptr>, i8 <delta>)
8224 declare i16 @llvm.atomic.load.max.i16.p0i16(i16* <ptr>, i16 <delta>)
8225 declare i32 @llvm.atomic.load.max.i32.p0i32(i32* <ptr>, i32 <delta>)
8226 declare i64 @llvm.atomic.load.max.i64.p0i64(i64* <ptr>, i64 <delta>)
8230 declare i8 @llvm.atomic.load.min.i8.p0i8(i8* <ptr>, i8 <delta>)
8231 declare i16 @llvm.atomic.load.min.i16.p0i16(i16* <ptr>, i16 <delta>)
8232 declare i32 @llvm.atomic.load.min.i32.p0i32(i32* <ptr>, i32 <delta>)
8233 declare i64 @llvm.atomic.load.min.i64.p0i64(i64* <ptr>, i64 <delta>)
8237 declare i8 @llvm.atomic.load.umax.i8.p0i8(i8* <ptr>, i8 <delta>)
8238 declare i16 @llvm.atomic.load.umax.i16.p0i16(i16* <ptr>, i16 <delta>)
8239 declare i32 @llvm.atomic.load.umax.i32.p0i32(i32* <ptr>, i32 <delta>)
8240 declare i64 @llvm.atomic.load.umax.i64.p0i64(i64* <ptr>, i64 <delta>)
8244 declare i8 @llvm.atomic.load.umin.i8.p0i8(i8* <ptr>, i8 <delta>)
8245 declare i16 @llvm.atomic.load.umin.i16.p0i16(i16* <ptr>, i16 <delta>)
8246 declare i32 @llvm.atomic.load.umin.i32.p0i32(i32* <ptr>, i32 <delta>)
8247 declare i64 @llvm.atomic.load.umin.i64.p0i64(i64* <ptr>, i64 <delta>)
8251 <p>These intrinsics takes the signed or unsigned minimum or maximum of
8252 <tt>delta</tt> and the value stored in memory at <tt>ptr</tt>. It yields the
8253 original value at <tt>ptr</tt>.</p>
8256 <p>These intrinsics take two arguments, the first a pointer to an integer value
8257 and the second an integer value. The result is also an integer value. These
8258 integer types can have any bit width, but they must all have the same bit
8259 width. The targets may only lower integer representations they support.</p>
8262 <p>These intrinsics does a series of operations atomically. They first load the
8263 value stored at <tt>ptr</tt>. They then do the signed or unsigned min or
8264 max <tt>delta</tt> and the value, store the result to <tt>ptr</tt>. They
8265 yield the original value stored at <tt>ptr</tt>.</p>
8269 %mallocP = tail call i8* @malloc(i32 ptrtoint (i32* getelementptr (i32* null, i32 1) to i32))
8270 %ptr = bitcast i8* %mallocP to i32*
8272 %result0 = call i32 @llvm.atomic.load.min.i32.p0i32(i32* %ptr, i32 -2)
8273 <i>; yields {i32}:result0 = 7</i>
8274 %result1 = call i32 @llvm.atomic.load.max.i32.p0i32(i32* %ptr, i32 8)
8275 <i>; yields {i32}:result1 = -2</i>
8276 %result2 = call i32 @llvm.atomic.load.umin.i32.p0i32(i32* %ptr, i32 10)
8277 <i>; yields {i32}:result2 = 8</i>
8278 %result3 = call i32 @llvm.atomic.load.umax.i32.p0i32(i32* %ptr, i32 30)
8279 <i>; yields {i32}:result3 = 8</i>
8280 %memval1 = load i32* %ptr <i>; yields {i32}:memval1 = 30</i>
8287 <!-- ======================================================================= -->
8289 <a name="int_memorymarkers">Memory Use Markers</a>
8294 <p>This class of intrinsics exists to information about the lifetime of memory
8295 objects and ranges where variables are immutable.</p>
8297 <!-- _______________________________________________________________________ -->
8299 <a name="int_lifetime_start">'<tt>llvm.lifetime.start</tt>' Intrinsic</a>
8306 declare void @llvm.lifetime.start(i64 <size>, i8* nocapture <ptr>)
8310 <p>The '<tt>llvm.lifetime.start</tt>' intrinsic specifies the start of a memory
8311 object's lifetime.</p>
8314 <p>The first argument is a constant integer representing the size of the
8315 object, or -1 if it is variable sized. The second argument is a pointer to
8319 <p>This intrinsic indicates that before this point in the code, the value of the
8320 memory pointed to by <tt>ptr</tt> is dead. This means that it is known to
8321 never be used and has an undefined value. A load from the pointer that
8322 precedes this intrinsic can be replaced with
8323 <tt>'<a href="#undefvalues">undef</a>'</tt>.</p>
8327 <!-- _______________________________________________________________________ -->
8329 <a name="int_lifetime_end">'<tt>llvm.lifetime.end</tt>' Intrinsic</a>
8336 declare void @llvm.lifetime.end(i64 <size>, i8* nocapture <ptr>)
8340 <p>The '<tt>llvm.lifetime.end</tt>' intrinsic specifies the end of a memory
8341 object's lifetime.</p>
8344 <p>The first argument is a constant integer representing the size of the
8345 object, or -1 if it is variable sized. The second argument is a pointer to
8349 <p>This intrinsic indicates that after this point in the code, the value of the
8350 memory pointed to by <tt>ptr</tt> is dead. This means that it is known to
8351 never be used and has an undefined value. Any stores into the memory object
8352 following this intrinsic may be removed as dead.
8356 <!-- _______________________________________________________________________ -->
8358 <a name="int_invariant_start">'<tt>llvm.invariant.start</tt>' Intrinsic</a>
8365 declare {}* @llvm.invariant.start(i64 <size>, i8* nocapture <ptr>)
8369 <p>The '<tt>llvm.invariant.start</tt>' intrinsic specifies that the contents of
8370 a memory object will not change.</p>
8373 <p>The first argument is a constant integer representing the size of the
8374 object, or -1 if it is variable sized. The second argument is a pointer to
8378 <p>This intrinsic indicates that until an <tt>llvm.invariant.end</tt> that uses
8379 the return value, the referenced memory location is constant and
8384 <!-- _______________________________________________________________________ -->
8386 <a name="int_invariant_end">'<tt>llvm.invariant.end</tt>' Intrinsic</a>
8393 declare void @llvm.invariant.end({}* <start>, i64 <size>, i8* nocapture <ptr>)
8397 <p>The '<tt>llvm.invariant.end</tt>' intrinsic specifies that the contents of
8398 a memory object are mutable.</p>
8401 <p>The first argument is the matching <tt>llvm.invariant.start</tt> intrinsic.
8402 The second argument is a constant integer representing the size of the
8403 object, or -1 if it is variable sized and the third argument is a pointer
8407 <p>This intrinsic indicates that the memory is mutable again.</p>
8413 <!-- ======================================================================= -->
8415 <a name="int_general">General Intrinsics</a>
8420 <p>This class of intrinsics is designed to be generic and has no specific
8423 <!-- _______________________________________________________________________ -->
8425 <a name="int_var_annotation">'<tt>llvm.var.annotation</tt>' Intrinsic</a>
8432 declare void @llvm.var.annotation(i8* <val>, i8* <str>, i8* <str>, i32 <int>)
8436 <p>The '<tt>llvm.var.annotation</tt>' intrinsic.</p>
8439 <p>The first argument is a pointer to a value, the second is a pointer to a
8440 global string, the third is a pointer to a global string which is the source
8441 file name, and the last argument is the line number.</p>
8444 <p>This intrinsic allows annotation of local variables with arbitrary strings.
8445 This can be useful for special purpose optimizations that want to look for
8446 these annotations. These have no other defined use; they are ignored by code
8447 generation and optimization.</p>
8451 <!-- _______________________________________________________________________ -->
8453 <a name="int_annotation">'<tt>llvm.annotation.*</tt>' Intrinsic</a>
8459 <p>This is an overloaded intrinsic. You can use '<tt>llvm.annotation</tt>' on
8460 any integer bit width.</p>
8463 declare i8 @llvm.annotation.i8(i8 <val>, i8* <str>, i8* <str>, i32 <int>)
8464 declare i16 @llvm.annotation.i16(i16 <val>, i8* <str>, i8* <str>, i32 <int>)
8465 declare i32 @llvm.annotation.i32(i32 <val>, i8* <str>, i8* <str>, i32 <int>)
8466 declare i64 @llvm.annotation.i64(i64 <val>, i8* <str>, i8* <str>, i32 <int>)
8467 declare i256 @llvm.annotation.i256(i256 <val>, i8* <str>, i8* <str>, i32 <int>)
8471 <p>The '<tt>llvm.annotation</tt>' intrinsic.</p>
8474 <p>The first argument is an integer value (result of some expression), the
8475 second is a pointer to a global string, the third is a pointer to a global
8476 string which is the source file name, and the last argument is the line
8477 number. It returns the value of the first argument.</p>
8480 <p>This intrinsic allows annotations to be put on arbitrary expressions with
8481 arbitrary strings. This can be useful for special purpose optimizations that
8482 want to look for these annotations. These have no other defined use; they
8483 are ignored by code generation and optimization.</p>
8487 <!-- _______________________________________________________________________ -->
8489 <a name="int_trap">'<tt>llvm.trap</tt>' Intrinsic</a>
8496 declare void @llvm.trap()
8500 <p>The '<tt>llvm.trap</tt>' intrinsic.</p>
8506 <p>This intrinsics is lowered to the target dependent trap instruction. If the
8507 target does not have a trap instruction, this intrinsic will be lowered to
8508 the call of the <tt>abort()</tt> function.</p>
8512 <!-- _______________________________________________________________________ -->
8514 <a name="int_stackprotector">'<tt>llvm.stackprotector</tt>' Intrinsic</a>
8521 declare void @llvm.stackprotector(i8* <guard>, i8** <slot>)
8525 <p>The <tt>llvm.stackprotector</tt> intrinsic takes the <tt>guard</tt> and
8526 stores it onto the stack at <tt>slot</tt>. The stack slot is adjusted to
8527 ensure that it is placed on the stack before local variables.</p>
8530 <p>The <tt>llvm.stackprotector</tt> intrinsic requires two pointer
8531 arguments. The first argument is the value loaded from the stack
8532 guard <tt>@__stack_chk_guard</tt>. The second variable is an <tt>alloca</tt>
8533 that has enough space to hold the value of the guard.</p>
8536 <p>This intrinsic causes the prologue/epilogue inserter to force the position of
8537 the <tt>AllocaInst</tt> stack slot to be before local variables on the
8538 stack. This is to ensure that if a local variable on the stack is
8539 overwritten, it will destroy the value of the guard. When the function exits,
8540 the guard on the stack is checked against the original guard. If they are
8541 different, then the program aborts by calling the <tt>__stack_chk_fail()</tt>
8546 <!-- _______________________________________________________________________ -->
8548 <a name="int_objectsize">'<tt>llvm.objectsize</tt>' Intrinsic</a>
8555 declare i32 @llvm.objectsize.i32(i8* <object>, i1 <type>)
8556 declare i64 @llvm.objectsize.i64(i8* <object>, i1 <type>)
8560 <p>The <tt>llvm.objectsize</tt> intrinsic is designed to provide information to
8561 the optimizers to determine at compile time whether a) an operation (like
8562 memcpy) will overflow a buffer that corresponds to an object, or b) that a
8563 runtime check for overflow isn't necessary. An object in this context means
8564 an allocation of a specific class, structure, array, or other object.</p>
8567 <p>The <tt>llvm.objectsize</tt> intrinsic takes two arguments. The first
8568 argument is a pointer to or into the <tt>object</tt>. The second argument
8569 is a boolean 0 or 1. This argument determines whether you want the
8570 maximum (0) or minimum (1) bytes remaining. This needs to be a literal 0 or
8571 1, variables are not allowed.</p>
8574 <p>The <tt>llvm.objectsize</tt> intrinsic is lowered to either a constant
8575 representing the size of the object concerned, or <tt>i32/i64 -1 or 0</tt>,
8576 depending on the <tt>type</tt> argument, if the size cannot be determined at
8585 <!-- *********************************************************************** -->
8588 <a href="http://jigsaw.w3.org/css-validator/check/referer"><img
8589 src="http://jigsaw.w3.org/css-validator/images/vcss-blue" alt="Valid CSS"></a>
8590 <a href="http://validator.w3.org/check/referer"><img
8591 src="http://www.w3.org/Icons/valid-html401-blue" alt="Valid HTML 4.01"></a>
8593 <a href="mailto:sabre@nondot.org">Chris Lattner</a><br>
8594 <a href="http://llvm.org/">The LLVM Compiler Infrastructure</a><br>
8595 Last modified: $Date$