1 <!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01//EN"
2 "http://www.w3.org/TR/html4/strict.dtd">
5 <title>LLVM Assembly Language Reference Manual</title>
6 <meta http-equiv="Content-Type" content="text/html; charset=utf-8">
7 <meta name="author" content="Chris Lattner">
8 <meta name="description"
9 content="LLVM Assembly Language Reference Manual.">
10 <link rel="stylesheet" href="llvm.css" type="text/css">
15 <div class="doc_title"> LLVM Language Reference Manual </div>
17 <li><a href="#abstract">Abstract</a></li>
18 <li><a href="#introduction">Introduction</a></li>
19 <li><a href="#identifiers">Identifiers</a></li>
20 <li><a href="#highlevel">High Level Structure</a>
22 <li><a href="#modulestructure">Module Structure</a></li>
23 <li><a href="#linkage">Linkage Types</a>
25 <li><a href="#linkage_private">'<tt>private</tt>' Linkage</a></li>
26 <li><a href="#linkage_linker_private">'<tt>linker_private</tt>' Linkage</a></li>
27 <li><a href="#linkage_internal">'<tt>internal</tt>' Linkage</a></li>
28 <li><a href="#linkage_available_externally">'<tt>available_externally</tt>' Linkage</a></li>
29 <li><a href="#linkage_linkonce">'<tt>linkonce</tt>' Linkage</a></li>
30 <li><a href="#linkage_common">'<tt>common</tt>' Linkage</a></li>
31 <li><a href="#linkage_weak">'<tt>weak</tt>' Linkage</a></li>
32 <li><a href="#linkage_appending">'<tt>appending</tt>' Linkage</a></li>
33 <li><a href="#linkage_externweak">'<tt>extern_weak</tt>' Linkage</a></li>
34 <li><a href="#linkage_linkonce_odr">'<tt>linkonce_odr</tt>' Linkage</a></li>
35 <li><a href="#linkage_weak">'<tt>weak_odr</tt>' Linkage</a></li>
36 <li><a href="#linkage_external">'<tt>externally visible</tt>' Linkage</a></li>
37 <li><a href="#linkage_dllimport">'<tt>dllimport</tt>' Linkage</a></li>
38 <li><a href="#linkage_dllexport">'<tt>dllexport</tt>' Linkage</a></li>
41 <li><a href="#callingconv">Calling Conventions</a></li>
42 <li><a href="#namedtypes">Named Types</a></li>
43 <li><a href="#globalvars">Global Variables</a></li>
44 <li><a href="#functionstructure">Functions</a></li>
45 <li><a href="#aliasstructure">Aliases</a></li>
46 <li><a href="#namedmetadatastructure">Named Metadata</a></li>
47 <li><a href="#paramattrs">Parameter Attributes</a></li>
48 <li><a href="#fnattrs">Function Attributes</a></li>
49 <li><a href="#gc">Garbage Collector Names</a></li>
50 <li><a href="#moduleasm">Module-Level Inline Assembly</a></li>
51 <li><a href="#datalayout">Data Layout</a></li>
52 <li><a href="#pointeraliasing">Pointer Aliasing Rules</a></li>
55 <li><a href="#typesystem">Type System</a>
57 <li><a href="#t_classifications">Type Classifications</a></li>
58 <li><a href="#t_primitive">Primitive Types</a>
60 <li><a href="#t_integer">Integer Type</a></li>
61 <li><a href="#t_floating">Floating Point Types</a></li>
62 <li><a href="#t_void">Void Type</a></li>
63 <li><a href="#t_label">Label Type</a></li>
64 <li><a href="#t_metadata">Metadata Type</a></li>
67 <li><a href="#t_derived">Derived Types</a>
69 <li><a href="#t_aggregate">Aggregate Types</a>
71 <li><a href="#t_array">Array Type</a></li>
72 <li><a href="#t_struct">Structure Type</a></li>
73 <li><a href="#t_pstruct">Packed Structure Type</a></li>
74 <li><a href="#t_union">Union Type</a></li>
75 <li><a href="#t_vector">Vector Type</a></li>
78 <li><a href="#t_function">Function Type</a></li>
79 <li><a href="#t_pointer">Pointer Type</a></li>
80 <li><a href="#t_opaque">Opaque Type</a></li>
83 <li><a href="#t_uprefs">Type Up-references</a></li>
86 <li><a href="#constants">Constants</a>
88 <li><a href="#simpleconstants">Simple Constants</a></li>
89 <li><a href="#complexconstants">Complex Constants</a></li>
90 <li><a href="#globalconstants">Global Variable and Function Addresses</a></li>
91 <li><a href="#undefvalues">Undefined Values</a></li>
92 <li><a href="#blockaddress">Addresses of Basic Blocks</a></li>
93 <li><a href="#constantexprs">Constant Expressions</a></li>
96 <li><a href="#othervalues">Other Values</a>
98 <li><a href="#inlineasm">Inline Assembler Expressions</a></li>
99 <li><a href="#metadata">Metadata Nodes and Metadata Strings</a></li>
102 <li><a href="#intrinsic_globals">Intrinsic Global Variables</a>
104 <li><a href="#intg_used">The '<tt>llvm.used</tt>' Global Variable</a></li>
105 <li><a href="#intg_compiler_used">The '<tt>llvm.compiler.used</tt>'
106 Global Variable</a></li>
107 <li><a href="#intg_global_ctors">The '<tt>llvm.global_ctors</tt>'
108 Global Variable</a></li>
109 <li><a href="#intg_global_dtors">The '<tt>llvm.global_dtors</tt>'
110 Global Variable</a></li>
113 <li><a href="#instref">Instruction Reference</a>
115 <li><a href="#terminators">Terminator Instructions</a>
117 <li><a href="#i_ret">'<tt>ret</tt>' Instruction</a></li>
118 <li><a href="#i_br">'<tt>br</tt>' Instruction</a></li>
119 <li><a href="#i_switch">'<tt>switch</tt>' Instruction</a></li>
120 <li><a href="#i_indirectbr">'<tt>indirectbr</tt>' Instruction</a></li>
121 <li><a href="#i_invoke">'<tt>invoke</tt>' Instruction</a></li>
122 <li><a href="#i_unwind">'<tt>unwind</tt>' Instruction</a></li>
123 <li><a href="#i_unreachable">'<tt>unreachable</tt>' Instruction</a></li>
126 <li><a href="#binaryops">Binary Operations</a>
128 <li><a href="#i_add">'<tt>add</tt>' Instruction</a></li>
129 <li><a href="#i_fadd">'<tt>fadd</tt>' Instruction</a></li>
130 <li><a href="#i_sub">'<tt>sub</tt>' Instruction</a></li>
131 <li><a href="#i_fsub">'<tt>fsub</tt>' Instruction</a></li>
132 <li><a href="#i_mul">'<tt>mul</tt>' Instruction</a></li>
133 <li><a href="#i_fmul">'<tt>fmul</tt>' Instruction</a></li>
134 <li><a href="#i_udiv">'<tt>udiv</tt>' Instruction</a></li>
135 <li><a href="#i_sdiv">'<tt>sdiv</tt>' Instruction</a></li>
136 <li><a href="#i_fdiv">'<tt>fdiv</tt>' Instruction</a></li>
137 <li><a href="#i_urem">'<tt>urem</tt>' Instruction</a></li>
138 <li><a href="#i_srem">'<tt>srem</tt>' Instruction</a></li>
139 <li><a href="#i_frem">'<tt>frem</tt>' Instruction</a></li>
142 <li><a href="#bitwiseops">Bitwise Binary Operations</a>
144 <li><a href="#i_shl">'<tt>shl</tt>' Instruction</a></li>
145 <li><a href="#i_lshr">'<tt>lshr</tt>' Instruction</a></li>
146 <li><a href="#i_ashr">'<tt>ashr</tt>' Instruction</a></li>
147 <li><a href="#i_and">'<tt>and</tt>' Instruction</a></li>
148 <li><a href="#i_or">'<tt>or</tt>' Instruction</a></li>
149 <li><a href="#i_xor">'<tt>xor</tt>' Instruction</a></li>
152 <li><a href="#vectorops">Vector Operations</a>
154 <li><a href="#i_extractelement">'<tt>extractelement</tt>' Instruction</a></li>
155 <li><a href="#i_insertelement">'<tt>insertelement</tt>' Instruction</a></li>
156 <li><a href="#i_shufflevector">'<tt>shufflevector</tt>' Instruction</a></li>
159 <li><a href="#aggregateops">Aggregate Operations</a>
161 <li><a href="#i_extractvalue">'<tt>extractvalue</tt>' Instruction</a></li>
162 <li><a href="#i_insertvalue">'<tt>insertvalue</tt>' Instruction</a></li>
165 <li><a href="#memoryops">Memory Access and Addressing Operations</a>
167 <li><a href="#i_alloca">'<tt>alloca</tt>' Instruction</a></li>
168 <li><a href="#i_load">'<tt>load</tt>' Instruction</a></li>
169 <li><a href="#i_store">'<tt>store</tt>' Instruction</a></li>
170 <li><a href="#i_getelementptr">'<tt>getelementptr</tt>' Instruction</a></li>
173 <li><a href="#convertops">Conversion Operations</a>
175 <li><a href="#i_trunc">'<tt>trunc .. to</tt>' Instruction</a></li>
176 <li><a href="#i_zext">'<tt>zext .. to</tt>' Instruction</a></li>
177 <li><a href="#i_sext">'<tt>sext .. to</tt>' Instruction</a></li>
178 <li><a href="#i_fptrunc">'<tt>fptrunc .. to</tt>' Instruction</a></li>
179 <li><a href="#i_fpext">'<tt>fpext .. to</tt>' Instruction</a></li>
180 <li><a href="#i_fptoui">'<tt>fptoui .. to</tt>' Instruction</a></li>
181 <li><a href="#i_fptosi">'<tt>fptosi .. to</tt>' Instruction</a></li>
182 <li><a href="#i_uitofp">'<tt>uitofp .. to</tt>' Instruction</a></li>
183 <li><a href="#i_sitofp">'<tt>sitofp .. to</tt>' Instruction</a></li>
184 <li><a href="#i_ptrtoint">'<tt>ptrtoint .. to</tt>' Instruction</a></li>
185 <li><a href="#i_inttoptr">'<tt>inttoptr .. to</tt>' Instruction</a></li>
186 <li><a href="#i_bitcast">'<tt>bitcast .. to</tt>' Instruction</a></li>
189 <li><a href="#otherops">Other Operations</a>
191 <li><a href="#i_icmp">'<tt>icmp</tt>' Instruction</a></li>
192 <li><a href="#i_fcmp">'<tt>fcmp</tt>' Instruction</a></li>
193 <li><a href="#i_phi">'<tt>phi</tt>' Instruction</a></li>
194 <li><a href="#i_select">'<tt>select</tt>' Instruction</a></li>
195 <li><a href="#i_call">'<tt>call</tt>' Instruction</a></li>
196 <li><a href="#i_va_arg">'<tt>va_arg</tt>' Instruction</a></li>
201 <li><a href="#intrinsics">Intrinsic Functions</a>
203 <li><a href="#int_varargs">Variable Argument Handling Intrinsics</a>
205 <li><a href="#int_va_start">'<tt>llvm.va_start</tt>' Intrinsic</a></li>
206 <li><a href="#int_va_end">'<tt>llvm.va_end</tt>' Intrinsic</a></li>
207 <li><a href="#int_va_copy">'<tt>llvm.va_copy</tt>' Intrinsic</a></li>
210 <li><a href="#int_gc">Accurate Garbage Collection Intrinsics</a>
212 <li><a href="#int_gcroot">'<tt>llvm.gcroot</tt>' Intrinsic</a></li>
213 <li><a href="#int_gcread">'<tt>llvm.gcread</tt>' Intrinsic</a></li>
214 <li><a href="#int_gcwrite">'<tt>llvm.gcwrite</tt>' Intrinsic</a></li>
217 <li><a href="#int_codegen">Code Generator Intrinsics</a>
219 <li><a href="#int_returnaddress">'<tt>llvm.returnaddress</tt>' Intrinsic</a></li>
220 <li><a href="#int_frameaddress">'<tt>llvm.frameaddress</tt>' Intrinsic</a></li>
221 <li><a href="#int_stacksave">'<tt>llvm.stacksave</tt>' Intrinsic</a></li>
222 <li><a href="#int_stackrestore">'<tt>llvm.stackrestore</tt>' Intrinsic</a></li>
223 <li><a href="#int_prefetch">'<tt>llvm.prefetch</tt>' Intrinsic</a></li>
224 <li><a href="#int_pcmarker">'<tt>llvm.pcmarker</tt>' Intrinsic</a></li>
225 <li><a href="#int_readcyclecounter"><tt>llvm.readcyclecounter</tt>' Intrinsic</a></li>
228 <li><a href="#int_libc">Standard C Library Intrinsics</a>
230 <li><a href="#int_memcpy">'<tt>llvm.memcpy.*</tt>' Intrinsic</a></li>
231 <li><a href="#int_memmove">'<tt>llvm.memmove.*</tt>' Intrinsic</a></li>
232 <li><a href="#int_memset">'<tt>llvm.memset.*</tt>' Intrinsic</a></li>
233 <li><a href="#int_sqrt">'<tt>llvm.sqrt.*</tt>' Intrinsic</a></li>
234 <li><a href="#int_powi">'<tt>llvm.powi.*</tt>' Intrinsic</a></li>
235 <li><a href="#int_sin">'<tt>llvm.sin.*</tt>' Intrinsic</a></li>
236 <li><a href="#int_cos">'<tt>llvm.cos.*</tt>' Intrinsic</a></li>
237 <li><a href="#int_pow">'<tt>llvm.pow.*</tt>' Intrinsic</a></li>
240 <li><a href="#int_manip">Bit Manipulation Intrinsics</a>
242 <li><a href="#int_bswap">'<tt>llvm.bswap.*</tt>' Intrinsics</a></li>
243 <li><a href="#int_ctpop">'<tt>llvm.ctpop.*</tt>' Intrinsic </a></li>
244 <li><a href="#int_ctlz">'<tt>llvm.ctlz.*</tt>' Intrinsic </a></li>
245 <li><a href="#int_cttz">'<tt>llvm.cttz.*</tt>' Intrinsic </a></li>
248 <li><a href="#int_overflow">Arithmetic with Overflow Intrinsics</a>
250 <li><a href="#int_sadd_overflow">'<tt>llvm.sadd.with.overflow.*</tt> Intrinsics</a></li>
251 <li><a href="#int_uadd_overflow">'<tt>llvm.uadd.with.overflow.*</tt> Intrinsics</a></li>
252 <li><a href="#int_ssub_overflow">'<tt>llvm.ssub.with.overflow.*</tt> Intrinsics</a></li>
253 <li><a href="#int_usub_overflow">'<tt>llvm.usub.with.overflow.*</tt> Intrinsics</a></li>
254 <li><a href="#int_smul_overflow">'<tt>llvm.smul.with.overflow.*</tt> Intrinsics</a></li>
255 <li><a href="#int_umul_overflow">'<tt>llvm.umul.with.overflow.*</tt> Intrinsics</a></li>
258 <li><a href="#int_debugger">Debugger intrinsics</a></li>
259 <li><a href="#int_eh">Exception Handling intrinsics</a></li>
260 <li><a href="#int_trampoline">Trampoline Intrinsic</a>
262 <li><a href="#int_it">'<tt>llvm.init.trampoline</tt>' Intrinsic</a></li>
265 <li><a href="#int_atomics">Atomic intrinsics</a>
267 <li><a href="#int_memory_barrier"><tt>llvm.memory_barrier</tt></a></li>
268 <li><a href="#int_atomic_cmp_swap"><tt>llvm.atomic.cmp.swap</tt></a></li>
269 <li><a href="#int_atomic_swap"><tt>llvm.atomic.swap</tt></a></li>
270 <li><a href="#int_atomic_load_add"><tt>llvm.atomic.load.add</tt></a></li>
271 <li><a href="#int_atomic_load_sub"><tt>llvm.atomic.load.sub</tt></a></li>
272 <li><a href="#int_atomic_load_and"><tt>llvm.atomic.load.and</tt></a></li>
273 <li><a href="#int_atomic_load_nand"><tt>llvm.atomic.load.nand</tt></a></li>
274 <li><a href="#int_atomic_load_or"><tt>llvm.atomic.load.or</tt></a></li>
275 <li><a href="#int_atomic_load_xor"><tt>llvm.atomic.load.xor</tt></a></li>
276 <li><a href="#int_atomic_load_max"><tt>llvm.atomic.load.max</tt></a></li>
277 <li><a href="#int_atomic_load_min"><tt>llvm.atomic.load.min</tt></a></li>
278 <li><a href="#int_atomic_load_umax"><tt>llvm.atomic.load.umax</tt></a></li>
279 <li><a href="#int_atomic_load_umin"><tt>llvm.atomic.load.umin</tt></a></li>
282 <li><a href="#int_memorymarkers">Memory Use Markers</a>
284 <li><a href="#int_lifetime_start"><tt>llvm.lifetime.start</tt></a></li>
285 <li><a href="#int_lifetime_end"><tt>llvm.lifetime.end</tt></a></li>
286 <li><a href="#int_invariant_start"><tt>llvm.invariant.start</tt></a></li>
287 <li><a href="#int_invariant_end"><tt>llvm.invariant.end</tt></a></li>
290 <li><a href="#int_general">General intrinsics</a>
292 <li><a href="#int_var_annotation">
293 '<tt>llvm.var.annotation</tt>' Intrinsic</a></li>
294 <li><a href="#int_annotation">
295 '<tt>llvm.annotation.*</tt>' Intrinsic</a></li>
296 <li><a href="#int_trap">
297 '<tt>llvm.trap</tt>' Intrinsic</a></li>
298 <li><a href="#int_stackprotector">
299 '<tt>llvm.stackprotector</tt>' Intrinsic</a></li>
300 <li><a href="#int_objectsize">
301 '<tt>llvm.objectsize</tt>' Intrinsic</a></li>
308 <div class="doc_author">
309 <p>Written by <a href="mailto:sabre@nondot.org">Chris Lattner</a>
310 and <a href="mailto:vadve@cs.uiuc.edu">Vikram Adve</a></p>
313 <!-- *********************************************************************** -->
314 <div class="doc_section"> <a name="abstract">Abstract </a></div>
315 <!-- *********************************************************************** -->
317 <div class="doc_text">
319 <p>This document is a reference manual for the LLVM assembly language. LLVM is
320 a Static Single Assignment (SSA) based representation that provides type
321 safety, low-level operations, flexibility, and the capability of representing
322 'all' high-level languages cleanly. It is the common code representation
323 used throughout all phases of the LLVM compilation strategy.</p>
327 <!-- *********************************************************************** -->
328 <div class="doc_section"> <a name="introduction">Introduction</a> </div>
329 <!-- *********************************************************************** -->
331 <div class="doc_text">
333 <p>The LLVM code representation is designed to be used in three different forms:
334 as an in-memory compiler IR, as an on-disk bitcode representation (suitable
335 for fast loading by a Just-In-Time compiler), and as a human readable
336 assembly language representation. This allows LLVM to provide a powerful
337 intermediate representation for efficient compiler transformations and
338 analysis, while providing a natural means to debug and visualize the
339 transformations. The three different forms of LLVM are all equivalent. This
340 document describes the human readable representation and notation.</p>
342 <p>The LLVM representation aims to be light-weight and low-level while being
343 expressive, typed, and extensible at the same time. It aims to be a
344 "universal IR" of sorts, by being at a low enough level that high-level ideas
345 may be cleanly mapped to it (similar to how microprocessors are "universal
346 IR's", allowing many source languages to be mapped to them). By providing
347 type information, LLVM can be used as the target of optimizations: for
348 example, through pointer analysis, it can be proven that a C automatic
349 variable is never accessed outside of the current function, allowing it to
350 be promoted to a simple SSA value instead of a memory location.</p>
354 <!-- _______________________________________________________________________ -->
355 <div class="doc_subsubsection"> <a name="wellformed">Well-Formedness</a> </div>
357 <div class="doc_text">
359 <p>It is important to note that this document describes 'well formed' LLVM
360 assembly language. There is a difference between what the parser accepts and
361 what is considered 'well formed'. For example, the following instruction is
362 syntactically okay, but not well formed:</p>
364 <div class="doc_code">
366 %x = <a href="#i_add">add</a> i32 1, %x
370 <p>because the definition of <tt>%x</tt> does not dominate all of its uses. The
371 LLVM infrastructure provides a verification pass that may be used to verify
372 that an LLVM module is well formed. This pass is automatically run by the
373 parser after parsing input assembly and by the optimizer before it outputs
374 bitcode. The violations pointed out by the verifier pass indicate bugs in
375 transformation passes or input to the parser.</p>
379 <!-- Describe the typesetting conventions here. -->
381 <!-- *********************************************************************** -->
382 <div class="doc_section"> <a name="identifiers">Identifiers</a> </div>
383 <!-- *********************************************************************** -->
385 <div class="doc_text">
387 <p>LLVM identifiers come in two basic types: global and local. Global
388 identifiers (functions, global variables) begin with the <tt>'@'</tt>
389 character. Local identifiers (register names, types) begin with
390 the <tt>'%'</tt> character. Additionally, there are three different formats
391 for identifiers, for different purposes:</p>
394 <li>Named values are represented as a string of characters with their prefix.
395 For example, <tt>%foo</tt>, <tt>@DivisionByZero</tt>,
396 <tt>%a.really.long.identifier</tt>. The actual regular expression used is
397 '<tt>[%@][a-zA-Z$._][a-zA-Z$._0-9]*</tt>'. Identifiers which require
398 other characters in their names can be surrounded with quotes. Special
399 characters may be escaped using <tt>"\xx"</tt> where <tt>xx</tt> is the
400 ASCII code for the character in hexadecimal. In this way, any character
401 can be used in a name value, even quotes themselves.</li>
403 <li>Unnamed values are represented as an unsigned numeric value with their
404 prefix. For example, <tt>%12</tt>, <tt>@2</tt>, <tt>%44</tt>.</li>
406 <li>Constants, which are described in a <a href="#constants">section about
407 constants</a>, below.</li>
410 <p>LLVM requires that values start with a prefix for two reasons: Compilers
411 don't need to worry about name clashes with reserved words, and the set of
412 reserved words may be expanded in the future without penalty. Additionally,
413 unnamed identifiers allow a compiler to quickly come up with a temporary
414 variable without having to avoid symbol table conflicts.</p>
416 <p>Reserved words in LLVM are very similar to reserved words in other
417 languages. There are keywords for different opcodes
418 ('<tt><a href="#i_add">add</a></tt>',
419 '<tt><a href="#i_bitcast">bitcast</a></tt>',
420 '<tt><a href="#i_ret">ret</a></tt>', etc...), for primitive type names
421 ('<tt><a href="#t_void">void</a></tt>',
422 '<tt><a href="#t_primitive">i32</a></tt>', etc...), and others. These
423 reserved words cannot conflict with variable names, because none of them
424 start with a prefix character (<tt>'%'</tt> or <tt>'@'</tt>).</p>
426 <p>Here is an example of LLVM code to multiply the integer variable
427 '<tt>%X</tt>' by 8:</p>
431 <div class="doc_code">
433 %result = <a href="#i_mul">mul</a> i32 %X, 8
437 <p>After strength reduction:</p>
439 <div class="doc_code">
441 %result = <a href="#i_shl">shl</a> i32 %X, i8 3
445 <p>And the hard way:</p>
447 <div class="doc_code">
449 %0 = <a href="#i_add">add</a> i32 %X, %X <i>; yields {i32}:%0</i>
450 %1 = <a href="#i_add">add</a> i32 %0, %0 <i>; yields {i32}:%1</i>
451 %result = <a href="#i_add">add</a> i32 %1, %1
455 <p>This last way of multiplying <tt>%X</tt> by 8 illustrates several important
456 lexical features of LLVM:</p>
459 <li>Comments are delimited with a '<tt>;</tt>' and go until the end of
462 <li>Unnamed temporaries are created when the result of a computation is not
463 assigned to a named value.</li>
465 <li>Unnamed temporaries are numbered sequentially</li>
468 <p>It also shows a convention that we follow in this document. When
469 demonstrating instructions, we will follow an instruction with a comment that
470 defines the type and name of value produced. Comments are shown in italic
475 <!-- *********************************************************************** -->
476 <div class="doc_section"> <a name="highlevel">High Level Structure</a> </div>
477 <!-- *********************************************************************** -->
479 <!-- ======================================================================= -->
480 <div class="doc_subsection"> <a name="modulestructure">Module Structure</a>
483 <div class="doc_text">
485 <p>LLVM programs are composed of "Module"s, each of which is a translation unit
486 of the input programs. Each module consists of functions, global variables,
487 and symbol table entries. Modules may be combined together with the LLVM
488 linker, which merges function (and global variable) definitions, resolves
489 forward declarations, and merges symbol table entries. Here is an example of
490 the "hello world" module:</p>
492 <div class="doc_code">
494 <i>; Declare the string constant as a global constant.</i>
495 <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>
497 <i>; External declaration of the puts function</i>
498 <a href="#functionstructure">declare</a> i32 @puts(i8 *) <i>; i32(i8 *)* </i>
500 <i>; Definition of main function</i>
501 define i32 @main() { <i>; i32()* </i>
502 <i>; Convert [13 x i8]* to i8 *...</i>
503 %cast210 = <a href="#i_getelementptr">getelementptr</a> [13 x i8]* @.LC0, i64 0, i64 0 <i>; i8 *</i>
505 <i>; Call puts function to write out the string to stdout.</i>
506 <a href="#i_call">call</a> i32 @puts(i8 * %cast210) <i>; i32</i>
507 <a href="#i_ret">ret</a> i32 0<br>}
509 <i>; Named metadata</i>
510 !1 = metadata !{i32 41}
515 <p>This example is made up of a <a href="#globalvars">global variable</a> named
516 "<tt>.LC0</tt>", an external declaration of the "<tt>puts</tt>" function,
517 a <a href="#functionstructure">function definition</a> for
518 "<tt>main</tt>" and <a href="#namedmetadatastructure">named metadata</a>
521 <p>In general, a module is made up of a list of global values, where both
522 functions and global variables are global values. Global values are
523 represented by a pointer to a memory location (in this case, a pointer to an
524 array of char, and a pointer to a function), and have one of the
525 following <a href="#linkage">linkage types</a>.</p>
529 <!-- ======================================================================= -->
530 <div class="doc_subsection">
531 <a name="linkage">Linkage Types</a>
534 <div class="doc_text">
536 <p>All Global Variables and Functions have one of the following types of
540 <dt><tt><b><a name="linkage_private">private</a></b></tt></dt>
541 <dd>Global values with private linkage are only directly accessible by objects
542 in the current module. In particular, linking code into a module with an
543 private global value may cause the private to be renamed as necessary to
544 avoid collisions. Because the symbol is private to the module, all
545 references can be updated. This doesn't show up in any symbol table in the
548 <dt><tt><b><a name="linkage_linker_private">linker_private</a></b></tt></dt>
549 <dd>Similar to private, but the symbol is passed through the assembler and
550 removed by the linker after evaluation. Note that (unlike private
551 symbols) linker_private symbols are subject to coalescing by the linker:
552 weak symbols get merged and redefinitions are rejected. However, unlike
553 normal strong symbols, they are removed by the linker from the final
554 linked image (executable or dynamic library).</dd>
556 <dt><tt><b><a name="linkage_internal">internal</a></b></tt></dt>
557 <dd>Similar to private, but the value shows as a local symbol
558 (<tt>STB_LOCAL</tt> in the case of ELF) in the object file. This
559 corresponds to the notion of the '<tt>static</tt>' keyword in C.</dd>
561 <dt><tt><b><a name="linkage_available_externally">available_externally</a></b></tt></dt>
562 <dd>Globals with "<tt>available_externally</tt>" linkage are never emitted
563 into the object file corresponding to the LLVM module. They exist to
564 allow inlining and other optimizations to take place given knowledge of
565 the definition of the global, which is known to be somewhere outside the
566 module. Globals with <tt>available_externally</tt> linkage are allowed to
567 be discarded at will, and are otherwise the same as <tt>linkonce_odr</tt>.
568 This linkage type is only allowed on definitions, not declarations.</dd>
570 <dt><tt><b><a name="linkage_linkonce">linkonce</a></b></tt></dt>
571 <dd>Globals with "<tt>linkonce</tt>" linkage are merged with other globals of
572 the same name when linkage occurs. This can be used to implement
573 some forms of inline functions, templates, or other code which must be
574 generated in each translation unit that uses it, but where the body may
575 be overridden with a more definitive definition later. Unreferenced
576 <tt>linkonce</tt> globals are allowed to be discarded. Note that
577 <tt>linkonce</tt> linkage does not actually allow the optimizer to
578 inline the body of this function into callers because it doesn't know if
579 this definition of the function is the definitive definition within the
580 program or whether it will be overridden by a stronger definition.
581 To enable inlining and other optimizations, use "<tt>linkonce_odr</tt>"
584 <dt><tt><b><a name="linkage_weak">weak</a></b></tt></dt>
585 <dd>"<tt>weak</tt>" linkage has the same merging semantics as
586 <tt>linkonce</tt> linkage, except that unreferenced globals with
587 <tt>weak</tt> linkage may not be discarded. This is used for globals that
588 are declared "weak" in C source code.</dd>
590 <dt><tt><b><a name="linkage_common">common</a></b></tt></dt>
591 <dd>"<tt>common</tt>" linkage is most similar to "<tt>weak</tt>" linkage, but
592 they are used for tentative definitions in C, such as "<tt>int X;</tt>" at
594 Symbols with "<tt>common</tt>" linkage are merged in the same way as
595 <tt>weak symbols</tt>, and they may not be deleted if unreferenced.
596 <tt>common</tt> symbols may not have an explicit section,
597 must have a zero initializer, and may not be marked '<a
598 href="#globalvars"><tt>constant</tt></a>'. Functions and aliases may not
599 have common linkage.</dd>
602 <dt><tt><b><a name="linkage_appending">appending</a></b></tt></dt>
603 <dd>"<tt>appending</tt>" linkage may only be applied to global variables of
604 pointer to array type. When two global variables with appending linkage
605 are linked together, the two global arrays are appended together. This is
606 the LLVM, typesafe, equivalent of having the system linker append together
607 "sections" with identical names when .o files are linked.</dd>
609 <dt><tt><b><a name="linkage_externweak">extern_weak</a></b></tt></dt>
610 <dd>The semantics of this linkage follow the ELF object file model: the symbol
611 is weak until linked, if not linked, the symbol becomes null instead of
612 being an undefined reference.</dd>
614 <dt><tt><b><a name="linkage_linkonce_odr">linkonce_odr</a></b></tt></dt>
615 <dt><tt><b><a name="linkage_weak_odr">weak_odr</a></b></tt></dt>
616 <dd>Some languages allow differing globals to be merged, such as two functions
617 with different semantics. Other languages, such as <tt>C++</tt>, ensure
618 that only equivalent globals are ever merged (the "one definition rule" -
619 "ODR"). Such languages can use the <tt>linkonce_odr</tt>
620 and <tt>weak_odr</tt> linkage types to indicate that the global will only
621 be merged with equivalent globals. These linkage types are otherwise the
622 same as their non-<tt>odr</tt> versions.</dd>
624 <dt><tt><b><a name="linkage_external">externally visible</a></b></tt>:</dt>
625 <dd>If none of the above identifiers are used, the global is externally
626 visible, meaning that it participates in linkage and can be used to
627 resolve external symbol references.</dd>
630 <p>The next two types of linkage are targeted for Microsoft Windows platform
631 only. They are designed to support importing (exporting) symbols from (to)
632 DLLs (Dynamic Link Libraries).</p>
635 <dt><tt><b><a name="linkage_dllimport">dllimport</a></b></tt></dt>
636 <dd>"<tt>dllimport</tt>" linkage causes the compiler to reference a function
637 or variable via a global pointer to a pointer that is set up by the DLL
638 exporting the symbol. On Microsoft Windows targets, the pointer name is
639 formed by combining <code>__imp_</code> and the function or variable
642 <dt><tt><b><a name="linkage_dllexport">dllexport</a></b></tt></dt>
643 <dd>"<tt>dllexport</tt>" linkage causes the compiler to provide a global
644 pointer to a pointer in a DLL, so that it can be referenced with the
645 <tt>dllimport</tt> attribute. On Microsoft Windows targets, the pointer
646 name is formed by combining <code>__imp_</code> and the function or
650 <p>For example, since the "<tt>.LC0</tt>" variable is defined to be internal, if
651 another module defined a "<tt>.LC0</tt>" variable and was linked with this
652 one, one of the two would be renamed, preventing a collision. Since
653 "<tt>main</tt>" and "<tt>puts</tt>" are external (i.e., lacking any linkage
654 declarations), they are accessible outside of the current module.</p>
656 <p>It is illegal for a function <i>declaration</i> to have any linkage type
657 other than "externally visible", <tt>dllimport</tt>
658 or <tt>extern_weak</tt>.</p>
660 <p>Aliases can have only <tt>external</tt>, <tt>internal</tt>, <tt>weak</tt>
661 or <tt>weak_odr</tt> linkages.</p>
665 <!-- ======================================================================= -->
666 <div class="doc_subsection">
667 <a name="callingconv">Calling Conventions</a>
670 <div class="doc_text">
672 <p>LLVM <a href="#functionstructure">functions</a>, <a href="#i_call">calls</a>
673 and <a href="#i_invoke">invokes</a> can all have an optional calling
674 convention specified for the call. The calling convention of any pair of
675 dynamic caller/callee must match, or the behavior of the program is
676 undefined. The following calling conventions are supported by LLVM, and more
677 may be added in the future:</p>
680 <dt><b>"<tt>ccc</tt>" - The C calling convention</b>:</dt>
681 <dd>This calling convention (the default if no other calling convention is
682 specified) matches the target C calling conventions. This calling
683 convention supports varargs function calls and tolerates some mismatch in
684 the declared prototype and implemented declaration of the function (as
687 <dt><b>"<tt>fastcc</tt>" - The fast calling convention</b>:</dt>
688 <dd>This calling convention attempts to make calls as fast as possible
689 (e.g. by passing things in registers). This calling convention allows the
690 target to use whatever tricks it wants to produce fast code for the
691 target, without having to conform to an externally specified ABI
692 (Application Binary Interface).
693 <a href="CodeGenerator.html#tailcallopt">Tail calls can only be optimized
694 when this or the GHC convention is used.</a> This calling convention
695 does not support varargs and requires the prototype of all callees to
696 exactly match the prototype of the function definition.</dd>
698 <dt><b>"<tt>coldcc</tt>" - The cold calling convention</b>:</dt>
699 <dd>This calling convention attempts to make code in the caller as efficient
700 as possible under the assumption that the call is not commonly executed.
701 As such, these calls often preserve all registers so that the call does
702 not break any live ranges in the caller side. This calling convention
703 does not support varargs and requires the prototype of all callees to
704 exactly match the prototype of the function definition.</dd>
706 <dt><b>"<tt>cc <em>10</em></tt>" - GHC convention</b>:</dt>
707 <dd>This calling convention has been implemented specifically for use by the
708 <a href="http://www.haskell.org/ghc">Glasgow Haskell Compiler (GHC)</a>.
709 It passes everything in registers, going to extremes to achieve this by
710 disabling callee save registers. This calling convention should not be
711 used lightly but only for specific situations such as an alternative to
712 the <em>register pinning</em> performance technique often used when
713 implementing functional programming languages.At the moment only X86
714 supports this convention and it has the following limitations:
716 <li>On <em>X86-32</em> only supports up to 4 bit type parameters. No
717 floating point types are supported.</li>
718 <li>On <em>X86-64</em> only supports up to 10 bit type parameters and
719 6 floating point parameters.</li>
721 This calling convention supports
722 <a href="CodeGenerator.html#tailcallopt">tail call optimization</a> but
723 requires both the caller and callee are using it.
726 <dt><b>"<tt>cc <<em>n</em>></tt>" - Numbered convention</b>:</dt>
727 <dd>Any calling convention may be specified by number, allowing
728 target-specific calling conventions to be used. Target specific calling
729 conventions start at 64.</dd>
732 <p>More calling conventions can be added/defined on an as-needed basis, to
733 support Pascal conventions or any other well-known target-independent
738 <!-- ======================================================================= -->
739 <div class="doc_subsection">
740 <a name="visibility">Visibility Styles</a>
743 <div class="doc_text">
745 <p>All Global Variables and Functions have one of the following visibility
749 <dt><b>"<tt>default</tt>" - Default style</b>:</dt>
750 <dd>On targets that use the ELF object file format, default visibility means
751 that the declaration is visible to other modules and, in shared libraries,
752 means that the declared entity may be overridden. On Darwin, default
753 visibility means that the declaration is visible to other modules. Default
754 visibility corresponds to "external linkage" in the language.</dd>
756 <dt><b>"<tt>hidden</tt>" - Hidden style</b>:</dt>
757 <dd>Two declarations of an object with hidden visibility refer to the same
758 object if they are in the same shared object. Usually, hidden visibility
759 indicates that the symbol will not be placed into the dynamic symbol
760 table, so no other module (executable or shared library) can reference it
763 <dt><b>"<tt>protected</tt>" - Protected style</b>:</dt>
764 <dd>On ELF, protected visibility indicates that the symbol will be placed in
765 the dynamic symbol table, but that references within the defining module
766 will bind to the local symbol. That is, the symbol cannot be overridden by
772 <!-- ======================================================================= -->
773 <div class="doc_subsection">
774 <a name="namedtypes">Named Types</a>
777 <div class="doc_text">
779 <p>LLVM IR allows you to specify name aliases for certain types. This can make
780 it easier to read the IR and make the IR more condensed (particularly when
781 recursive types are involved). An example of a name specification is:</p>
783 <div class="doc_code">
785 %mytype = type { %mytype*, i32 }
789 <p>You may give a name to any <a href="#typesystem">type</a> except
790 "<a href="t_void">void</a>". Type name aliases may be used anywhere a type
791 is expected with the syntax "%mytype".</p>
793 <p>Note that type names are aliases for the structural type that they indicate,
794 and that you can therefore specify multiple names for the same type. This
795 often leads to confusing behavior when dumping out a .ll file. Since LLVM IR
796 uses structural typing, the name is not part of the type. When printing out
797 LLVM IR, the printer will pick <em>one name</em> to render all types of a
798 particular shape. This means that if you have code where two different
799 source types end up having the same LLVM type, that the dumper will sometimes
800 print the "wrong" or unexpected type. This is an important design point and
801 isn't going to change.</p>
805 <!-- ======================================================================= -->
806 <div class="doc_subsection">
807 <a name="globalvars">Global Variables</a>
810 <div class="doc_text">
812 <p>Global variables define regions of memory allocated at compilation time
813 instead of run-time. Global variables may optionally be initialized, may
814 have an explicit section to be placed in, and may have an optional explicit
815 alignment specified. A variable may be defined as "thread_local", which
816 means that it will not be shared by threads (each thread will have a
817 separated copy of the variable). A variable may be defined as a global
818 "constant," which indicates that the contents of the variable
819 will <b>never</b> be modified (enabling better optimization, allowing the
820 global data to be placed in the read-only section of an executable, etc).
821 Note that variables that need runtime initialization cannot be marked
822 "constant" as there is a store to the variable.</p>
824 <p>LLVM explicitly allows <em>declarations</em> of global variables to be marked
825 constant, even if the final definition of the global is not. This capability
826 can be used to enable slightly better optimization of the program, but
827 requires the language definition to guarantee that optimizations based on the
828 'constantness' are valid for the translation units that do not include the
831 <p>As SSA values, global variables define pointer values that are in scope
832 (i.e. they dominate) all basic blocks in the program. Global variables
833 always define a pointer to their "content" type because they describe a
834 region of memory, and all memory objects in LLVM are accessed through
837 <p>A global variable may be declared to reside in a target-specific numbered
838 address space. For targets that support them, address spaces may affect how
839 optimizations are performed and/or what target instructions are used to
840 access the variable. The default address space is zero. The address space
841 qualifier must precede any other attributes.</p>
843 <p>LLVM allows an explicit section to be specified for globals. If the target
844 supports it, it will emit globals to the section specified.</p>
846 <p>An explicit alignment may be specified for a global. If not present, or if
847 the alignment is set to zero, the alignment of the global is set by the
848 target to whatever it feels convenient. If an explicit alignment is
849 specified, the global is forced to have at least that much alignment. All
850 alignments must be a power of 2.</p>
852 <p>For example, the following defines a global in a numbered address space with
853 an initializer, section, and alignment:</p>
855 <div class="doc_code">
857 @G = addrspace(5) constant float 1.0, section "foo", align 4
864 <!-- ======================================================================= -->
865 <div class="doc_subsection">
866 <a name="functionstructure">Functions</a>
869 <div class="doc_text">
871 <p>LLVM function definitions consist of the "<tt>define</tt>" keyword, an
872 optional <a href="#linkage">linkage type</a>, an optional
873 <a href="#visibility">visibility style</a>, an optional
874 <a href="#callingconv">calling convention</a>, a return type, an optional
875 <a href="#paramattrs">parameter attribute</a> for the return type, a function
876 name, a (possibly empty) argument list (each with optional
877 <a href="#paramattrs">parameter attributes</a>), optional
878 <a href="#fnattrs">function attributes</a>, an optional section, an optional
879 alignment, an optional <a href="#gc">garbage collector name</a>, an opening
880 curly brace, a list of basic blocks, and a closing curly brace.</p>
882 <p>LLVM function declarations consist of the "<tt>declare</tt>" keyword, an
883 optional <a href="#linkage">linkage type</a>, an optional
884 <a href="#visibility">visibility style</a>, an optional
885 <a href="#callingconv">calling convention</a>, a return type, an optional
886 <a href="#paramattrs">parameter attribute</a> for the return type, a function
887 name, a possibly empty list of arguments, an optional alignment, and an
888 optional <a href="#gc">garbage collector name</a>.</p>
890 <p>A function definition contains a list of basic blocks, forming the CFG
891 (Control Flow Graph) for the function. Each basic block may optionally start
892 with a label (giving the basic block a symbol table entry), contains a list
893 of instructions, and ends with a <a href="#terminators">terminator</a>
894 instruction (such as a branch or function return).</p>
896 <p>The first basic block in a function is special in two ways: it is immediately
897 executed on entrance to the function, and it is not allowed to have
898 predecessor basic blocks (i.e. there can not be any branches to the entry
899 block of a function). Because the block can have no predecessors, it also
900 cannot have any <a href="#i_phi">PHI nodes</a>.</p>
902 <p>LLVM allows an explicit section to be specified for functions. If the target
903 supports it, it will emit functions to the section specified.</p>
905 <p>An explicit alignment may be specified for a function. If not present, or if
906 the alignment is set to zero, the alignment of the function is set by the
907 target to whatever it feels convenient. If an explicit alignment is
908 specified, the function is forced to have at least that much alignment. All
909 alignments must be a power of 2.</p>
912 <div class="doc_code">
914 define [<a href="#linkage">linkage</a>] [<a href="#visibility">visibility</a>]
915 [<a href="#callingconv">cconv</a>] [<a href="#paramattrs">ret attrs</a>]
916 <ResultType> @<FunctionName> ([argument list])
917 [<a href="#fnattrs">fn Attrs</a>] [section "name"] [align N]
918 [<a href="#gc">gc</a>] { ... }
924 <!-- ======================================================================= -->
925 <div class="doc_subsection">
926 <a name="aliasstructure">Aliases</a>
929 <div class="doc_text">
931 <p>Aliases act as "second name" for the aliasee value (which can be either
932 function, global variable, another alias or bitcast of global value). Aliases
933 may have an optional <a href="#linkage">linkage type</a>, and an
934 optional <a href="#visibility">visibility style</a>.</p>
937 <div class="doc_code">
939 @<Name> = alias [Linkage] [Visibility] <AliaseeTy> @<Aliasee>
945 <!-- ======================================================================= -->
946 <div class="doc_subsection">
947 <a name="namedmetadatastructure">Named Metadata</a>
950 <div class="doc_text">
952 <p>Named metadata is a collection of metadata. <a href="#metadata">Metadata
953 nodes</a> (but not metadata strings) and null are the only valid operands for
954 a named metadata.</p>
957 <div class="doc_code">
959 !1 = metadata !{metadata !"one"}
966 <!-- ======================================================================= -->
967 <div class="doc_subsection"><a name="paramattrs">Parameter Attributes</a></div>
969 <div class="doc_text">
971 <p>The return type and each parameter of a function type may have a set of
972 <i>parameter attributes</i> associated with them. Parameter attributes are
973 used to communicate additional information about the result or parameters of
974 a function. Parameter attributes are considered to be part of the function,
975 not of the function type, so functions with different parameter attributes
976 can have the same function type.</p>
978 <p>Parameter attributes are simple keywords that follow the type specified. If
979 multiple parameter attributes are needed, they are space separated. For
982 <div class="doc_code">
984 declare i32 @printf(i8* noalias nocapture, ...)
985 declare i32 @atoi(i8 zeroext)
986 declare signext i8 @returns_signed_char()
990 <p>Note that any attributes for the function result (<tt>nounwind</tt>,
991 <tt>readonly</tt>) come immediately after the argument list.</p>
993 <p>Currently, only the following parameter attributes are defined:</p>
996 <dt><tt><b>zeroext</b></tt></dt>
997 <dd>This indicates to the code generator that the parameter or return value
998 should be zero-extended to a 32-bit value by the caller (for a parameter)
999 or the callee (for a return value).</dd>
1001 <dt><tt><b>signext</b></tt></dt>
1002 <dd>This indicates to the code generator that the parameter or return value
1003 should be sign-extended to a 32-bit value by the caller (for a parameter)
1004 or the callee (for a return value).</dd>
1006 <dt><tt><b>inreg</b></tt></dt>
1007 <dd>This indicates that this parameter or return value should be treated in a
1008 special target-dependent fashion during while emitting code for a function
1009 call or return (usually, by putting it in a register as opposed to memory,
1010 though some targets use it to distinguish between two different kinds of
1011 registers). Use of this attribute is target-specific.</dd>
1013 <dt><tt><b><a name="byval">byval</a></b></tt></dt>
1014 <dd>This indicates that the pointer parameter should really be passed by value
1015 to the function. The attribute implies that a hidden copy of the pointee
1016 is made between the caller and the callee, so the callee is unable to
1017 modify the value in the callee. This attribute is only valid on LLVM
1018 pointer arguments. It is generally used to pass structs and arrays by
1019 value, but is also valid on pointers to scalars. The copy is considered
1020 to belong to the caller not the callee (for example,
1021 <tt><a href="#readonly">readonly</a></tt> functions should not write to
1022 <tt>byval</tt> parameters). This is not a valid attribute for return
1023 values. The byval attribute also supports specifying an alignment with
1024 the align attribute. This has a target-specific effect on the code
1025 generator that usually indicates a desired alignment for the synthesized
1028 <dt><tt><b>sret</b></tt></dt>
1029 <dd>This indicates that the pointer parameter specifies the address of a
1030 structure that is the return value of the function in the source program.
1031 This pointer must be guaranteed by the caller to be valid: loads and
1032 stores to the structure may be assumed by the callee to not to trap. This
1033 may only be applied to the first parameter. This is not a valid attribute
1034 for return values. </dd>
1036 <dt><tt><b>noalias</b></tt></dt>
1037 <dd>This indicates that the pointer does not alias any global or any other
1038 parameter. The caller is responsible for ensuring that this is the
1039 case. On a function return value, <tt>noalias</tt> additionally indicates
1040 that the pointer does not alias any other pointers visible to the
1041 caller. For further details, please see the discussion of the NoAlias
1043 <a href="http://llvm.org/docs/AliasAnalysis.html#MustMayNo">alias
1046 <dt><tt><b>nocapture</b></tt></dt>
1047 <dd>This indicates that the callee does not make any copies of the pointer
1048 that outlive the callee itself. This is not a valid attribute for return
1051 <dt><tt><b>nest</b></tt></dt>
1052 <dd>This indicates that the pointer parameter can be excised using the
1053 <a href="#int_trampoline">trampoline intrinsics</a>. This is not a valid
1054 attribute for return values.</dd>
1059 <!-- ======================================================================= -->
1060 <div class="doc_subsection">
1061 <a name="gc">Garbage Collector Names</a>
1064 <div class="doc_text">
1066 <p>Each function may specify a garbage collector name, which is simply a
1069 <div class="doc_code">
1071 define void @f() gc "name" { ... }
1075 <p>The compiler declares the supported values of <i>name</i>. Specifying a
1076 collector which will cause the compiler to alter its output in order to
1077 support the named garbage collection algorithm.</p>
1081 <!-- ======================================================================= -->
1082 <div class="doc_subsection">
1083 <a name="fnattrs">Function Attributes</a>
1086 <div class="doc_text">
1088 <p>Function attributes are set to communicate additional information about a
1089 function. Function attributes are considered to be part of the function, not
1090 of the function type, so functions with different parameter attributes can
1091 have the same function type.</p>
1093 <p>Function attributes are simple keywords that follow the type specified. If
1094 multiple attributes are needed, they are space separated. For example:</p>
1096 <div class="doc_code">
1098 define void @f() noinline { ... }
1099 define void @f() alwaysinline { ... }
1100 define void @f() alwaysinline optsize { ... }
1101 define void @f() optsize { ... }
1106 <dt><tt><b>alignstack(<<em>n</em>>)</b></tt></dt>
1107 <dd>This attribute indicates that, when emitting the prologue and epilogue,
1108 the backend should forcibly align the stack pointer. Specify the
1109 desired alignment, which must be a power of two, in parentheses.
1111 <dt><tt><b>alwaysinline</b></tt></dt>
1112 <dd>This attribute indicates that the inliner should attempt to inline this
1113 function into callers whenever possible, ignoring any active inlining size
1114 threshold for this caller.</dd>
1116 <dt><tt><b>inlinehint</b></tt></dt>
1117 <dd>This attribute indicates that the source code contained a hint that inlining
1118 this function is desirable (such as the "inline" keyword in C/C++). It
1119 is just a hint; it imposes no requirements on the inliner.</dd>
1121 <dt><tt><b>noinline</b></tt></dt>
1122 <dd>This attribute indicates that the inliner should never inline this
1123 function in any situation. This attribute may not be used together with
1124 the <tt>alwaysinline</tt> attribute.</dd>
1126 <dt><tt><b>optsize</b></tt></dt>
1127 <dd>This attribute suggests that optimization passes and code generator passes
1128 make choices that keep the code size of this function low, and otherwise
1129 do optimizations specifically to reduce code size.</dd>
1131 <dt><tt><b>noreturn</b></tt></dt>
1132 <dd>This function attribute indicates that the function never returns
1133 normally. This produces undefined behavior at runtime if the function
1134 ever does dynamically return.</dd>
1136 <dt><tt><b>nounwind</b></tt></dt>
1137 <dd>This function attribute indicates that the function never returns with an
1138 unwind or exceptional control flow. If the function does unwind, its
1139 runtime behavior is undefined.</dd>
1141 <dt><tt><b>readnone</b></tt></dt>
1142 <dd>This attribute indicates that the function computes its result (or decides
1143 to unwind an exception) based strictly on its arguments, without
1144 dereferencing any pointer arguments or otherwise accessing any mutable
1145 state (e.g. memory, control registers, etc) visible to caller functions.
1146 It does not write through any pointer arguments
1147 (including <tt><a href="#byval">byval</a></tt> arguments) and never
1148 changes any state visible to callers. This means that it cannot unwind
1149 exceptions by calling the <tt>C++</tt> exception throwing methods, but
1150 could use the <tt>unwind</tt> instruction.</dd>
1152 <dt><tt><b><a name="readonly">readonly</a></b></tt></dt>
1153 <dd>This attribute indicates that the function does not write through any
1154 pointer arguments (including <tt><a href="#byval">byval</a></tt>
1155 arguments) or otherwise modify any state (e.g. memory, control registers,
1156 etc) visible to caller functions. It may dereference pointer arguments
1157 and read state that may be set in the caller. A readonly function always
1158 returns the same value (or unwinds an exception identically) when called
1159 with the same set of arguments and global state. It cannot unwind an
1160 exception by calling the <tt>C++</tt> exception throwing methods, but may
1161 use the <tt>unwind</tt> instruction.</dd>
1163 <dt><tt><b><a name="ssp">ssp</a></b></tt></dt>
1164 <dd>This attribute indicates that the function should emit a stack smashing
1165 protector. It is in the form of a "canary"—a random value placed on
1166 the stack before the local variables that's checked upon return from the
1167 function to see if it has been overwritten. A heuristic is used to
1168 determine if a function needs stack protectors or not.<br>
1170 If a function that has an <tt>ssp</tt> attribute is inlined into a
1171 function that doesn't have an <tt>ssp</tt> attribute, then the resulting
1172 function will have an <tt>ssp</tt> attribute.</dd>
1174 <dt><tt><b>sspreq</b></tt></dt>
1175 <dd>This attribute indicates that the function should <em>always</em> emit a
1176 stack smashing protector. This overrides
1177 the <tt><a href="#ssp">ssp</a></tt> function attribute.<br>
1179 If a function that has an <tt>sspreq</tt> attribute is inlined into a
1180 function that doesn't have an <tt>sspreq</tt> attribute or which has
1181 an <tt>ssp</tt> attribute, then the resulting function will have
1182 an <tt>sspreq</tt> attribute.</dd>
1184 <dt><tt><b>noredzone</b></tt></dt>
1185 <dd>This attribute indicates that the code generator should not use a red
1186 zone, even if the target-specific ABI normally permits it.</dd>
1188 <dt><tt><b>noimplicitfloat</b></tt></dt>
1189 <dd>This attributes disables implicit floating point instructions.</dd>
1191 <dt><tt><b>naked</b></tt></dt>
1192 <dd>This attribute disables prologue / epilogue emission for the function.
1193 This can have very system-specific consequences.</dd>
1198 <!-- ======================================================================= -->
1199 <div class="doc_subsection">
1200 <a name="moduleasm">Module-Level Inline Assembly</a>
1203 <div class="doc_text">
1205 <p>Modules may contain "module-level inline asm" blocks, which corresponds to
1206 the GCC "file scope inline asm" blocks. These blocks are internally
1207 concatenated by LLVM and treated as a single unit, but may be separated in
1208 the <tt>.ll</tt> file if desired. The syntax is very simple:</p>
1210 <div class="doc_code">
1212 module asm "inline asm code goes here"
1213 module asm "more can go here"
1217 <p>The strings can contain any character by escaping non-printable characters.
1218 The escape sequence used is simply "\xx" where "xx" is the two digit hex code
1221 <p>The inline asm code is simply printed to the machine code .s file when
1222 assembly code is generated.</p>
1226 <!-- ======================================================================= -->
1227 <div class="doc_subsection">
1228 <a name="datalayout">Data Layout</a>
1231 <div class="doc_text">
1233 <p>A module may specify a target specific data layout string that specifies how
1234 data is to be laid out in memory. The syntax for the data layout is
1237 <div class="doc_code">
1239 target datalayout = "<i>layout specification</i>"
1243 <p>The <i>layout specification</i> consists of a list of specifications
1244 separated by the minus sign character ('-'). Each specification starts with
1245 a letter and may include other information after the letter to define some
1246 aspect of the data layout. The specifications accepted are as follows:</p>
1250 <dd>Specifies that the target lays out data in big-endian form. That is, the
1251 bits with the most significance have the lowest address location.</dd>
1254 <dd>Specifies that the target lays out data in little-endian form. That is,
1255 the bits with the least significance have the lowest address
1258 <dt><tt>p:<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1259 <dd>This specifies the <i>size</i> of a pointer and its <i>abi</i> and
1260 <i>preferred</i> alignments. All sizes are in bits. Specifying
1261 the <i>pref</i> alignment is optional. If omitted, the
1262 preceding <tt>:</tt> should be omitted too.</dd>
1264 <dt><tt>i<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1265 <dd>This specifies the alignment for an integer type of a given bit
1266 <i>size</i>. The value of <i>size</i> must be in the range [1,2^23).</dd>
1268 <dt><tt>v<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1269 <dd>This specifies the alignment for a vector type of a given bit
1272 <dt><tt>f<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1273 <dd>This specifies the alignment for a floating point type of a given bit
1274 <i>size</i>. The value of <i>size</i> must be either 32 (float) or 64
1277 <dt><tt>a<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1278 <dd>This specifies the alignment for an aggregate type of a given bit
1281 <dt><tt>s<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1282 <dd>This specifies the alignment for a stack object of a given bit
1285 <dt><tt>n<i>size1</i>:<i>size2</i>:<i>size3</i>...</tt></dt>
1286 <dd>This specifies a set of native integer widths for the target CPU
1287 in bits. For example, it might contain "n32" for 32-bit PowerPC,
1288 "n32:64" for PowerPC 64, or "n8:16:32:64" for X86-64. Elements of
1289 this set are considered to support most general arithmetic
1290 operations efficiently.</dd>
1293 <p>When constructing the data layout for a given target, LLVM starts with a
1294 default set of specifications which are then (possibly) overriden by the
1295 specifications in the <tt>datalayout</tt> keyword. The default specifications
1296 are given in this list:</p>
1299 <li><tt>E</tt> - big endian</li>
1300 <li><tt>p:64:64:64</tt> - 64-bit pointers with 64-bit alignment</li>
1301 <li><tt>i1:8:8</tt> - i1 is 8-bit (byte) aligned</li>
1302 <li><tt>i8:8:8</tt> - i8 is 8-bit (byte) aligned</li>
1303 <li><tt>i16:16:16</tt> - i16 is 16-bit aligned</li>
1304 <li><tt>i32:32:32</tt> - i32 is 32-bit aligned</li>
1305 <li><tt>i64:32:64</tt> - i64 has ABI alignment of 32-bits but preferred
1306 alignment of 64-bits</li>
1307 <li><tt>f32:32:32</tt> - float is 32-bit aligned</li>
1308 <li><tt>f64:64:64</tt> - double is 64-bit aligned</li>
1309 <li><tt>v64:64:64</tt> - 64-bit vector is 64-bit aligned</li>
1310 <li><tt>v128:128:128</tt> - 128-bit vector is 128-bit aligned</li>
1311 <li><tt>a0:0:1</tt> - aggregates are 8-bit aligned</li>
1312 <li><tt>s0:64:64</tt> - stack objects are 64-bit aligned</li>
1315 <p>When LLVM is determining the alignment for a given type, it uses the
1316 following rules:</p>
1319 <li>If the type sought is an exact match for one of the specifications, that
1320 specification is used.</li>
1322 <li>If no match is found, and the type sought is an integer type, then the
1323 smallest integer type that is larger than the bitwidth of the sought type
1324 is used. If none of the specifications are larger than the bitwidth then
1325 the the largest integer type is used. For example, given the default
1326 specifications above, the i7 type will use the alignment of i8 (next
1327 largest) while both i65 and i256 will use the alignment of i64 (largest
1330 <li>If no match is found, and the type sought is a vector type, then the
1331 largest vector type that is smaller than the sought vector type will be
1332 used as a fall back. This happens because <128 x double> can be
1333 implemented in terms of 64 <2 x double>, for example.</li>
1338 <!-- ======================================================================= -->
1339 <div class="doc_subsection">
1340 <a name="pointeraliasing">Pointer Aliasing Rules</a>
1343 <div class="doc_text">
1345 <p>Any memory access must be done through a pointer value associated
1346 with an address range of the memory access, otherwise the behavior
1347 is undefined. Pointer values are associated with address ranges
1348 according to the following rules:</p>
1351 <li>A pointer value formed from a
1352 <tt><a href="#i_getelementptr">getelementptr</a></tt> instruction
1353 is associated with the addresses associated with the first operand
1354 of the <tt>getelementptr</tt>.</li>
1355 <li>An address of a global variable is associated with the address
1356 range of the variable's storage.</li>
1357 <li>The result value of an allocation instruction is associated with
1358 the address range of the allocated storage.</li>
1359 <li>A null pointer in the default address-space is associated with
1361 <li>A pointer value formed by an
1362 <tt><a href="#i_inttoptr">inttoptr</a></tt> is associated with all
1363 address ranges of all pointer values that contribute (directly or
1364 indirectly) to the computation of the pointer's value.</li>
1365 <li>The result value of a
1366 <tt><a href="#i_bitcast">bitcast</a></tt> is associated with all
1367 addresses associated with the operand of the <tt>bitcast</tt>.</li>
1368 <li>An integer constant other than zero or a pointer value returned
1369 from a function not defined within LLVM may be associated with address
1370 ranges allocated through mechanisms other than those provided by
1371 LLVM. Such ranges shall not overlap with any ranges of addresses
1372 allocated by mechanisms provided by LLVM.</li>
1375 <p>LLVM IR does not associate types with memory. The result type of a
1376 <tt><a href="#i_load">load</a></tt> merely indicates the size and
1377 alignment of the memory from which to load, as well as the
1378 interpretation of the value. The first operand of a
1379 <tt><a href="#i_store">store</a></tt> similarly only indicates the size
1380 and alignment of the store.</p>
1382 <p>Consequently, type-based alias analysis, aka TBAA, aka
1383 <tt>-fstrict-aliasing</tt>, is not applicable to general unadorned
1384 LLVM IR. <a href="#metadata">Metadata</a> may be used to encode
1385 additional information which specialized optimization passes may use
1386 to implement type-based alias analysis.</p>
1390 <!-- *********************************************************************** -->
1391 <div class="doc_section"> <a name="typesystem">Type System</a> </div>
1392 <!-- *********************************************************************** -->
1394 <div class="doc_text">
1396 <p>The LLVM type system is one of the most important features of the
1397 intermediate representation. Being typed enables a number of optimizations
1398 to be performed on the intermediate representation directly, without having
1399 to do extra analyses on the side before the transformation. A strong type
1400 system makes it easier to read the generated code and enables novel analyses
1401 and transformations that are not feasible to perform on normal three address
1402 code representations.</p>
1406 <!-- ======================================================================= -->
1407 <div class="doc_subsection"> <a name="t_classifications">Type
1408 Classifications</a> </div>
1410 <div class="doc_text">
1412 <p>The types fall into a few useful classifications:</p>
1414 <table border="1" cellspacing="0" cellpadding="4">
1416 <tr><th>Classification</th><th>Types</th></tr>
1418 <td><a href="#t_integer">integer</a></td>
1419 <td><tt>i1, i2, i3, ... i8, ... i16, ... i32, ... i64, ... </tt></td>
1422 <td><a href="#t_floating">floating point</a></td>
1423 <td><tt>float, double, x86_fp80, fp128, ppc_fp128</tt></td>
1426 <td><a name="t_firstclass">first class</a></td>
1427 <td><a href="#t_integer">integer</a>,
1428 <a href="#t_floating">floating point</a>,
1429 <a href="#t_pointer">pointer</a>,
1430 <a href="#t_vector">vector</a>,
1431 <a href="#t_struct">structure</a>,
1432 <a href="#t_union">union</a>,
1433 <a href="#t_array">array</a>,
1434 <a href="#t_label">label</a>,
1435 <a href="#t_metadata">metadata</a>.
1439 <td><a href="#t_primitive">primitive</a></td>
1440 <td><a href="#t_label">label</a>,
1441 <a href="#t_void">void</a>,
1442 <a href="#t_floating">floating point</a>,
1443 <a href="#t_metadata">metadata</a>.</td>
1446 <td><a href="#t_derived">derived</a></td>
1447 <td><a href="#t_array">array</a>,
1448 <a href="#t_function">function</a>,
1449 <a href="#t_pointer">pointer</a>,
1450 <a href="#t_struct">structure</a>,
1451 <a href="#t_pstruct">packed structure</a>,
1452 <a href="#t_union">union</a>,
1453 <a href="#t_vector">vector</a>,
1454 <a href="#t_opaque">opaque</a>.
1460 <p>The <a href="#t_firstclass">first class</a> types are perhaps the most
1461 important. Values of these types are the only ones which can be produced by
1466 <!-- ======================================================================= -->
1467 <div class="doc_subsection"> <a name="t_primitive">Primitive Types</a> </div>
1469 <div class="doc_text">
1471 <p>The primitive types are the fundamental building blocks of the LLVM
1476 <!-- _______________________________________________________________________ -->
1477 <div class="doc_subsubsection"> <a name="t_integer">Integer Type</a> </div>
1479 <div class="doc_text">
1482 <p>The integer type is a very simple type that simply specifies an arbitrary
1483 bit width for the integer type desired. Any bit width from 1 bit to
1484 2<sup>23</sup>-1 (about 8 million) can be specified.</p>
1491 <p>The number of bits the integer will occupy is specified by the <tt>N</tt>
1495 <table class="layout">
1497 <td class="left"><tt>i1</tt></td>
1498 <td class="left">a single-bit integer.</td>
1501 <td class="left"><tt>i32</tt></td>
1502 <td class="left">a 32-bit integer.</td>
1505 <td class="left"><tt>i1942652</tt></td>
1506 <td class="left">a really big integer of over 1 million bits.</td>
1512 <!-- _______________________________________________________________________ -->
1513 <div class="doc_subsubsection"> <a name="t_floating">Floating Point Types</a> </div>
1515 <div class="doc_text">
1519 <tr><th>Type</th><th>Description</th></tr>
1520 <tr><td><tt>float</tt></td><td>32-bit floating point value</td></tr>
1521 <tr><td><tt>double</tt></td><td>64-bit floating point value</td></tr>
1522 <tr><td><tt>fp128</tt></td><td>128-bit floating point value (112-bit mantissa)</td></tr>
1523 <tr><td><tt>x86_fp80</tt></td><td>80-bit floating point value (X87)</td></tr>
1524 <tr><td><tt>ppc_fp128</tt></td><td>128-bit floating point value (two 64-bits)</td></tr>
1530 <!-- _______________________________________________________________________ -->
1531 <div class="doc_subsubsection"> <a name="t_void">Void Type</a> </div>
1533 <div class="doc_text">
1536 <p>The void type does not represent any value and has no size.</p>
1545 <!-- _______________________________________________________________________ -->
1546 <div class="doc_subsubsection"> <a name="t_label">Label Type</a> </div>
1548 <div class="doc_text">
1551 <p>The label type represents code labels.</p>
1560 <!-- _______________________________________________________________________ -->
1561 <div class="doc_subsubsection"> <a name="t_metadata">Metadata Type</a> </div>
1563 <div class="doc_text">
1566 <p>The metadata type represents embedded metadata. No derived types may be
1567 created from metadata except for <a href="#t_function">function</a>
1578 <!-- ======================================================================= -->
1579 <div class="doc_subsection"> <a name="t_derived">Derived Types</a> </div>
1581 <div class="doc_text">
1583 <p>The real power in LLVM comes from the derived types in the system. This is
1584 what allows a programmer to represent arrays, functions, pointers, and other
1585 useful types. Each of these types contain one or more element types which
1586 may be a primitive type, or another derived type. For example, it is
1587 possible to have a two dimensional array, using an array as the element type
1588 of another array.</p>
1593 <!-- _______________________________________________________________________ -->
1594 <div class="doc_subsubsection"> <a name="t_aggregate">Aggregate Types</a> </div>
1596 <div class="doc_text">
1598 <p>Aggregate Types are a subset of derived types that can contain multiple
1599 member types. <a href="#t_array">Arrays</a>,
1600 <a href="#t_struct">structs</a>, <a href="#t_vector">vectors</a> and
1601 <a href="#t_union">unions</a> are aggregate types.</p>
1607 <!-- _______________________________________________________________________ -->
1608 <div class="doc_subsubsection"> <a name="t_array">Array Type</a> </div>
1610 <div class="doc_text">
1613 <p>The array type is a very simple derived type that arranges elements
1614 sequentially in memory. The array type requires a size (number of elements)
1615 and an underlying data type.</p>
1619 [<# elements> x <elementtype>]
1622 <p>The number of elements is a constant integer value; <tt>elementtype</tt> may
1623 be any type with a size.</p>
1626 <table class="layout">
1628 <td class="left"><tt>[40 x i32]</tt></td>
1629 <td class="left">Array of 40 32-bit integer values.</td>
1632 <td class="left"><tt>[41 x i32]</tt></td>
1633 <td class="left">Array of 41 32-bit integer values.</td>
1636 <td class="left"><tt>[4 x i8]</tt></td>
1637 <td class="left">Array of 4 8-bit integer values.</td>
1640 <p>Here are some examples of multidimensional arrays:</p>
1641 <table class="layout">
1643 <td class="left"><tt>[3 x [4 x i32]]</tt></td>
1644 <td class="left">3x4 array of 32-bit integer values.</td>
1647 <td class="left"><tt>[12 x [10 x float]]</tt></td>
1648 <td class="left">12x10 array of single precision floating point values.</td>
1651 <td class="left"><tt>[2 x [3 x [4 x i16]]]</tt></td>
1652 <td class="left">2x3x4 array of 16-bit integer values.</td>
1656 <p>There is no restriction on indexing beyond the end of the array implied by
1657 a static type (though there are restrictions on indexing beyond the bounds
1658 of an allocated object in some cases). This means that single-dimension
1659 'variable sized array' addressing can be implemented in LLVM with a zero
1660 length array type. An implementation of 'pascal style arrays' in LLVM could
1661 use the type "<tt>{ i32, [0 x float]}</tt>", for example.</p>
1665 <!-- _______________________________________________________________________ -->
1666 <div class="doc_subsubsection"> <a name="t_function">Function Type</a> </div>
1668 <div class="doc_text">
1671 <p>The function type can be thought of as a function signature. It consists of
1672 a return type and a list of formal parameter types. The return type of a
1673 function type is a scalar type, a void type, a struct type, or a union
1674 type. If the return type is a struct type then all struct elements must be
1675 of first class types, and the struct must have at least one element.</p>
1679 <returntype> (<parameter list>)
1682 <p>...where '<tt><parameter list></tt>' is a comma-separated list of type
1683 specifiers. Optionally, the parameter list may include a type <tt>...</tt>,
1684 which indicates that the function takes a variable number of arguments.
1685 Variable argument functions can access their arguments with
1686 the <a href="#int_varargs">variable argument handling intrinsic</a>
1687 functions. '<tt><returntype></tt>' is any type except
1688 <a href="#t_label">label</a>.</p>
1691 <table class="layout">
1693 <td class="left"><tt>i32 (i32)</tt></td>
1694 <td class="left">function taking an <tt>i32</tt>, returning an <tt>i32</tt>
1696 </tr><tr class="layout">
1697 <td class="left"><tt>float (i16, i32 *) *
1699 <td class="left"><a href="#t_pointer">Pointer</a> to a function that takes
1700 an <tt>i16</tt> and a <a href="#t_pointer">pointer</a> to <tt>i32</tt>,
1701 returning <tt>float</tt>.
1703 </tr><tr class="layout">
1704 <td class="left"><tt>i32 (i8*, ...)</tt></td>
1705 <td class="left">A vararg function that takes at least one
1706 <a href="#t_pointer">pointer</a> to <tt>i8 </tt> (char in C),
1707 which returns an integer. This is the signature for <tt>printf</tt> in
1710 </tr><tr class="layout">
1711 <td class="left"><tt>{i32, i32} (i32)</tt></td>
1712 <td class="left">A function taking an <tt>i32</tt>, returning a
1713 <a href="#t_struct">structure</a> containing two <tt>i32</tt> values
1720 <!-- _______________________________________________________________________ -->
1721 <div class="doc_subsubsection"> <a name="t_struct">Structure Type</a> </div>
1723 <div class="doc_text">
1726 <p>The structure type is used to represent a collection of data members together
1727 in memory. The packing of the field types is defined to match the ABI of the
1728 underlying processor. The elements of a structure may be any type that has a
1731 <p>Structures in memory are accessed using '<tt><a href="#i_load">load</a></tt>'
1732 and '<tt><a href="#i_store">store</a></tt>' by getting a pointer to a field
1733 with the '<tt><a href="#i_getelementptr">getelementptr</a></tt>' instruction.
1734 Structures in registers are accessed using the
1735 '<tt><a href="#i_extractvalue">extractvalue</a></tt>' and
1736 '<tt><a href="#i_insertvalue">insertvalue</a></tt>' instructions.</p>
1739 { <type list> }
1743 <table class="layout">
1745 <td class="left"><tt>{ i32, i32, i32 }</tt></td>
1746 <td class="left">A triple of three <tt>i32</tt> values</td>
1747 </tr><tr class="layout">
1748 <td class="left"><tt>{ float, i32 (i32) * }</tt></td>
1749 <td class="left">A pair, where the first element is a <tt>float</tt> and the
1750 second element is a <a href="#t_pointer">pointer</a> to a
1751 <a href="#t_function">function</a> that takes an <tt>i32</tt>, returning
1752 an <tt>i32</tt>.</td>
1758 <!-- _______________________________________________________________________ -->
1759 <div class="doc_subsubsection"> <a name="t_pstruct">Packed Structure Type</a>
1762 <div class="doc_text">
1765 <p>The packed structure type is used to represent a collection of data members
1766 together in memory. There is no padding between fields. Further, the
1767 alignment of a packed structure is 1 byte. The elements of a packed
1768 structure may be any type that has a size.</p>
1770 <p>Structures are accessed using '<tt><a href="#i_load">load</a></tt> and
1771 '<tt><a href="#i_store">store</a></tt>' by getting a pointer to a field with
1772 the '<tt><a href="#i_getelementptr">getelementptr</a></tt>' instruction.</p>
1776 < { <type list> } >
1780 <table class="layout">
1782 <td class="left"><tt>< { i32, i32, i32 } ></tt></td>
1783 <td class="left">A triple of three <tt>i32</tt> values</td>
1784 </tr><tr class="layout">
1786 <tt>< { float, i32 (i32)* } ></tt></td>
1787 <td class="left">A pair, where the first element is a <tt>float</tt> and the
1788 second element is a <a href="#t_pointer">pointer</a> to a
1789 <a href="#t_function">function</a> that takes an <tt>i32</tt>, returning
1790 an <tt>i32</tt>.</td>
1796 <!-- _______________________________________________________________________ -->
1797 <div class="doc_subsubsection"> <a name="t_union">Union Type</a> </div>
1799 <div class="doc_text">
1802 <p>A union type describes an object with size and alignment suitable for
1803 an object of any one of a given set of types (also known as an "untagged"
1804 union). It is similar in concept and usage to a
1805 <a href="#t_struct">struct</a>, except that all members of the union
1806 have an offset of zero. The elements of a union may be any type that has a
1807 size. Unions must have at least one member - empty unions are not allowed.
1810 <p>The size of the union as a whole will be the size of its largest member,
1811 and the alignment requirements of the union as a whole will be the largest
1812 alignment requirement of any member.</p>
1814 <p>Union members are accessed using '<tt><a href="#i_load">load</a></tt> and
1815 '<tt><a href="#i_store">store</a></tt>' by getting a pointer to a field with
1816 the '<tt><a href="#i_getelementptr">getelementptr</a></tt>' instruction.
1817 Since all members are at offset zero, the getelementptr instruction does
1818 not affect the address, only the type of the resulting pointer.</p>
1822 union { <type list> }
1826 <table class="layout">
1828 <td class="left"><tt>union { i32, i32*, float }</tt></td>
1829 <td class="left">A union of three types: an <tt>i32</tt>, a pointer to
1830 an <tt>i32</tt>, and a <tt>float</tt>.</td>
1831 </tr><tr class="layout">
1833 <tt>union { float, i32 (i32) * }</tt></td>
1834 <td class="left">A union, where the first element is a <tt>float</tt> and the
1835 second element is a <a href="#t_pointer">pointer</a> to a
1836 <a href="#t_function">function</a> that takes an <tt>i32</tt>, returning
1837 an <tt>i32</tt>.</td>
1843 <!-- _______________________________________________________________________ -->
1844 <div class="doc_subsubsection"> <a name="t_pointer">Pointer Type</a> </div>
1846 <div class="doc_text">
1849 <p>The pointer type is used to specify memory locations.
1850 Pointers are commonly used to reference objects in memory.</p>
1852 <p>Pointer types may have an optional address space attribute defining the
1853 numbered address space where the pointed-to object resides. The default
1854 address space is number zero. The semantics of non-zero address
1855 spaces are target-specific.</p>
1857 <p>Note that LLVM does not permit pointers to void (<tt>void*</tt>) nor does it
1858 permit pointers to labels (<tt>label*</tt>). Use <tt>i8*</tt> instead.</p>
1866 <table class="layout">
1868 <td class="left"><tt>[4 x i32]*</tt></td>
1869 <td class="left">A <a href="#t_pointer">pointer</a> to <a
1870 href="#t_array">array</a> of four <tt>i32</tt> values.</td>
1873 <td class="left"><tt>i32 (i32 *) *</tt></td>
1874 <td class="left"> A <a href="#t_pointer">pointer</a> to a <a
1875 href="#t_function">function</a> that takes an <tt>i32*</tt>, returning an
1879 <td class="left"><tt>i32 addrspace(5)*</tt></td>
1880 <td class="left">A <a href="#t_pointer">pointer</a> to an <tt>i32</tt> value
1881 that resides in address space #5.</td>
1887 <!-- _______________________________________________________________________ -->
1888 <div class="doc_subsubsection"> <a name="t_vector">Vector Type</a> </div>
1890 <div class="doc_text">
1893 <p>A vector type is a simple derived type that represents a vector of elements.
1894 Vector types are used when multiple primitive data are operated in parallel
1895 using a single instruction (SIMD). A vector type requires a size (number of
1896 elements) and an underlying primitive data type. Vector types are considered
1897 <a href="#t_firstclass">first class</a>.</p>
1901 < <# elements> x <elementtype> >
1904 <p>The number of elements is a constant integer value; elementtype may be any
1905 integer or floating point type.</p>
1908 <table class="layout">
1910 <td class="left"><tt><4 x i32></tt></td>
1911 <td class="left">Vector of 4 32-bit integer values.</td>
1914 <td class="left"><tt><8 x float></tt></td>
1915 <td class="left">Vector of 8 32-bit floating-point values.</td>
1918 <td class="left"><tt><2 x i64></tt></td>
1919 <td class="left">Vector of 2 64-bit integer values.</td>
1925 <!-- _______________________________________________________________________ -->
1926 <div class="doc_subsubsection"> <a name="t_opaque">Opaque Type</a> </div>
1927 <div class="doc_text">
1930 <p>Opaque types are used to represent unknown types in the system. This
1931 corresponds (for example) to the C notion of a forward declared structure
1932 type. In LLVM, opaque types can eventually be resolved to any type (not just
1933 a structure type).</p>
1941 <table class="layout">
1943 <td class="left"><tt>opaque</tt></td>
1944 <td class="left">An opaque type.</td>
1950 <!-- ======================================================================= -->
1951 <div class="doc_subsection">
1952 <a name="t_uprefs">Type Up-references</a>
1955 <div class="doc_text">
1958 <p>An "up reference" allows you to refer to a lexically enclosing type without
1959 requiring it to have a name. For instance, a structure declaration may
1960 contain a pointer to any of the types it is lexically a member of. Example
1961 of up references (with their equivalent as named type declarations)
1965 { \2 * } %x = type { %x* }
1966 { \2 }* %y = type { %y }*
1970 <p>An up reference is needed by the asmprinter for printing out cyclic types
1971 when there is no declared name for a type in the cycle. Because the
1972 asmprinter does not want to print out an infinite type string, it needs a
1973 syntax to handle recursive types that have no names (all names are optional
1981 <p>The level is the count of the lexical type that is being referred to.</p>
1984 <table class="layout">
1986 <td class="left"><tt>\1*</tt></td>
1987 <td class="left">Self-referential pointer.</td>
1990 <td class="left"><tt>{ { \3*, i8 }, i32 }</tt></td>
1991 <td class="left">Recursive structure where the upref refers to the out-most
1998 <!-- *********************************************************************** -->
1999 <div class="doc_section"> <a name="constants">Constants</a> </div>
2000 <!-- *********************************************************************** -->
2002 <div class="doc_text">
2004 <p>LLVM has several different basic types of constants. This section describes
2005 them all and their syntax.</p>
2009 <!-- ======================================================================= -->
2010 <div class="doc_subsection"><a name="simpleconstants">Simple Constants</a></div>
2012 <div class="doc_text">
2015 <dt><b>Boolean constants</b></dt>
2016 <dd>The two strings '<tt>true</tt>' and '<tt>false</tt>' are both valid
2017 constants of the <tt><a href="#t_integer">i1</a></tt> type.</dd>
2019 <dt><b>Integer constants</b></dt>
2020 <dd>Standard integers (such as '4') are constants of
2021 the <a href="#t_integer">integer</a> type. Negative numbers may be used
2022 with integer types.</dd>
2024 <dt><b>Floating point constants</b></dt>
2025 <dd>Floating point constants use standard decimal notation (e.g. 123.421),
2026 exponential notation (e.g. 1.23421e+2), or a more precise hexadecimal
2027 notation (see below). The assembler requires the exact decimal value of a
2028 floating-point constant. For example, the assembler accepts 1.25 but
2029 rejects 1.3 because 1.3 is a repeating decimal in binary. Floating point
2030 constants must have a <a href="#t_floating">floating point</a> type. </dd>
2032 <dt><b>Null pointer constants</b></dt>
2033 <dd>The identifier '<tt>null</tt>' is recognized as a null pointer constant
2034 and must be of <a href="#t_pointer">pointer type</a>.</dd>
2037 <p>The one non-intuitive notation for constants is the hexadecimal form of
2038 floating point constants. For example, the form '<tt>double
2039 0x432ff973cafa8000</tt>' is equivalent to (but harder to read than)
2040 '<tt>double 4.5e+15</tt>'. The only time hexadecimal floating point
2041 constants are required (and the only time that they are generated by the
2042 disassembler) is when a floating point constant must be emitted but it cannot
2043 be represented as a decimal floating point number in a reasonable number of
2044 digits. For example, NaN's, infinities, and other special values are
2045 represented in their IEEE hexadecimal format so that assembly and disassembly
2046 do not cause any bits to change in the constants.</p>
2048 <p>When using the hexadecimal form, constants of types float and double are
2049 represented using the 16-digit form shown above (which matches the IEEE754
2050 representation for double); float values must, however, be exactly
2051 representable as IEE754 single precision. Hexadecimal format is always used
2052 for long double, and there are three forms of long double. The 80-bit format
2053 used by x86 is represented as <tt>0xK</tt> followed by 20 hexadecimal digits.
2054 The 128-bit format used by PowerPC (two adjacent doubles) is represented
2055 by <tt>0xM</tt> followed by 32 hexadecimal digits. The IEEE 128-bit format
2056 is represented by <tt>0xL</tt> followed by 32 hexadecimal digits; no
2057 currently supported target uses this format. Long doubles will only work if
2058 they match the long double format on your target. All hexadecimal formats
2059 are big-endian (sign bit at the left).</p>
2063 <!-- ======================================================================= -->
2064 <div class="doc_subsection">
2065 <a name="aggregateconstants"></a> <!-- old anchor -->
2066 <a name="complexconstants">Complex Constants</a>
2069 <div class="doc_text">
2071 <p>Complex constants are a (potentially recursive) combination of simple
2072 constants and smaller complex constants.</p>
2075 <dt><b>Structure constants</b></dt>
2076 <dd>Structure constants are represented with notation similar to structure
2077 type definitions (a comma separated list of elements, surrounded by braces
2078 (<tt>{}</tt>)). For example: "<tt>{ i32 4, float 17.0, i32* @G }</tt>",
2079 where "<tt>@G</tt>" is declared as "<tt>@G = external global i32</tt>".
2080 Structure constants must have <a href="#t_struct">structure type</a>, and
2081 the number and types of elements must match those specified by the
2084 <dt><b>Union constants</b></dt>
2085 <dd>Union constants are represented with notation similar to a structure with
2086 a single element - that is, a single typed element surrounded
2087 by braces (<tt>{}</tt>)). For example: "<tt>{ i32 4 }</tt>". The
2088 <a href="#t_union">union type</a> can be initialized with a single-element
2089 struct as long as the type of the struct element matches the type of
2090 one of the union members.</dd>
2092 <dt><b>Array constants</b></dt>
2093 <dd>Array constants are represented with notation similar to array type
2094 definitions (a comma separated list of elements, surrounded by square
2095 brackets (<tt>[]</tt>)). For example: "<tt>[ i32 42, i32 11, i32 74
2096 ]</tt>". Array constants must have <a href="#t_array">array type</a>, and
2097 the number and types of elements must match those specified by the
2100 <dt><b>Vector constants</b></dt>
2101 <dd>Vector constants are represented with notation similar to vector type
2102 definitions (a comma separated list of elements, surrounded by
2103 less-than/greater-than's (<tt><></tt>)). For example: "<tt>< i32
2104 42, i32 11, i32 74, i32 100 ></tt>". Vector constants must
2105 have <a href="#t_vector">vector type</a>, and the number and types of
2106 elements must match those specified by the type.</dd>
2108 <dt><b>Zero initialization</b></dt>
2109 <dd>The string '<tt>zeroinitializer</tt>' can be used to zero initialize a
2110 value to zero of <em>any</em> type, including scalar and
2111 <a href="#t_aggregate">aggregate</a> types.
2112 This is often used to avoid having to print large zero initializers
2113 (e.g. for large arrays) and is always exactly equivalent to using explicit
2114 zero initializers.</dd>
2116 <dt><b>Metadata node</b></dt>
2117 <dd>A metadata node is a structure-like constant with
2118 <a href="#t_metadata">metadata type</a>. For example: "<tt>metadata !{
2119 i32 0, metadata !"test" }</tt>". Unlike other constants that are meant to
2120 be interpreted as part of the instruction stream, metadata is a place to
2121 attach additional information such as debug info.</dd>
2126 <!-- ======================================================================= -->
2127 <div class="doc_subsection">
2128 <a name="globalconstants">Global Variable and Function Addresses</a>
2131 <div class="doc_text">
2133 <p>The addresses of <a href="#globalvars">global variables</a>
2134 and <a href="#functionstructure">functions</a> are always implicitly valid
2135 (link-time) constants. These constants are explicitly referenced when
2136 the <a href="#identifiers">identifier for the global</a> is used and always
2137 have <a href="#t_pointer">pointer</a> type. For example, the following is a
2138 legal LLVM file:</p>
2140 <div class="doc_code">
2144 @Z = global [2 x i32*] [ i32* @X, i32* @Y ]
2150 <!-- ======================================================================= -->
2151 <div class="doc_subsection"><a name="undefvalues">Undefined Values</a></div>
2152 <div class="doc_text">
2154 <p>The string '<tt>undef</tt>' can be used anywhere a constant is expected, and
2155 indicates that the user of the value may receive an unspecified bit-pattern.
2156 Undefined values may be of any type (other than label or void) and be used
2157 anywhere a constant is permitted.</p>
2159 <p>Undefined values are useful because they indicate to the compiler that the
2160 program is well defined no matter what value is used. This gives the
2161 compiler more freedom to optimize. Here are some examples of (potentially
2162 surprising) transformations that are valid (in pseudo IR):</p>
2165 <div class="doc_code">
2177 <p>This is safe because all of the output bits are affected by the undef bits.
2178 Any output bit can have a zero or one depending on the input bits.</p>
2180 <div class="doc_code">
2193 <p>These logical operations have bits that are not always affected by the input.
2194 For example, if "%X" has a zero bit, then the output of the 'and' operation will
2195 always be a zero, no matter what the corresponding bit from the undef is. As
2196 such, it is unsafe to optimize or assume that the result of the and is undef.
2197 However, it is safe to assume that all bits of the undef could be 0, and
2198 optimize the and to 0. Likewise, it is safe to assume that all the bits of
2199 the undef operand to the or could be set, allowing the or to be folded to
2202 <div class="doc_code">
2204 %A = select undef, %X, %Y
2205 %B = select undef, 42, %Y
2206 %C = select %X, %Y, undef
2218 <p>This set of examples show that undefined select (and conditional branch)
2219 conditions can go "either way" but they have to come from one of the two
2220 operands. In the %A example, if %X and %Y were both known to have a clear low
2221 bit, then %A would have to have a cleared low bit. However, in the %C example,
2222 the optimizer is allowed to assume that the undef operand could be the same as
2223 %Y, allowing the whole select to be eliminated.</p>
2226 <div class="doc_code">
2228 %A = xor undef, undef
2247 <p>This example points out that two undef operands are not necessarily the same.
2248 This can be surprising to people (and also matches C semantics) where they
2249 assume that "X^X" is always zero, even if X is undef. This isn't true for a
2250 number of reasons, but the short answer is that an undef "variable" can
2251 arbitrarily change its value over its "live range". This is true because the
2252 "variable" doesn't actually <em>have a live range</em>. Instead, the value is
2253 logically read from arbitrary registers that happen to be around when needed,
2254 so the value is not necessarily consistent over time. In fact, %A and %C need
2255 to have the same semantics or the core LLVM "replace all uses with" concept
2258 <div class="doc_code">
2268 <p>These examples show the crucial difference between an <em>undefined
2269 value</em> and <em>undefined behavior</em>. An undefined value (like undef) is
2270 allowed to have an arbitrary bit-pattern. This means that the %A operation
2271 can be constant folded to undef because the undef could be an SNaN, and fdiv is
2272 not (currently) defined on SNaN's. However, in the second example, we can make
2273 a more aggressive assumption: because the undef is allowed to be an arbitrary
2274 value, we are allowed to assume that it could be zero. Since a divide by zero
2275 has <em>undefined behavior</em>, we are allowed to assume that the operation
2276 does not execute at all. This allows us to delete the divide and all code after
2277 it: since the undefined operation "can't happen", the optimizer can assume that
2278 it occurs in dead code.
2281 <div class="doc_code">
2283 a: store undef -> %X
2284 b: store %X -> undef
2291 <p>These examples reiterate the fdiv example: a store "of" an undefined value
2292 can be assumed to not have any effect: we can assume that the value is
2293 overwritten with bits that happen to match what was already there. However, a
2294 store "to" an undefined location could clobber arbitrary memory, therefore, it
2295 has undefined behavior.</p>
2299 <!-- ======================================================================= -->
2300 <div class="doc_subsection"><a name="blockaddress">Addresses of Basic
2302 <div class="doc_text">
2304 <p><b><tt>blockaddress(@function, %block)</tt></b></p>
2306 <p>The '<tt>blockaddress</tt>' constant computes the address of the specified
2307 basic block in the specified function, and always has an i8* type. Taking
2308 the address of the entry block is illegal.</p>
2310 <p>This value only has defined behavior when used as an operand to the
2311 '<a href="#i_indirectbr"><tt>indirectbr</tt></a>' instruction or for comparisons
2312 against null. Pointer equality tests between labels addresses is undefined
2313 behavior - though, again, comparison against null is ok, and no label is
2314 equal to the null pointer. This may also be passed around as an opaque
2315 pointer sized value as long as the bits are not inspected. This allows
2316 <tt>ptrtoint</tt> and arithmetic to be performed on these values so long as
2317 the original value is reconstituted before the <tt>indirectbr</tt>.</p>
2319 <p>Finally, some targets may provide defined semantics when
2320 using the value as the operand to an inline assembly, but that is target
2327 <!-- ======================================================================= -->
2328 <div class="doc_subsection"><a name="constantexprs">Constant Expressions</a>
2331 <div class="doc_text">
2333 <p>Constant expressions are used to allow expressions involving other constants
2334 to be used as constants. Constant expressions may be of
2335 any <a href="#t_firstclass">first class</a> type and may involve any LLVM
2336 operation that does not have side effects (e.g. load and call are not
2337 supported). The following is the syntax for constant expressions:</p>
2340 <dt><b><tt>trunc ( CST to TYPE )</tt></b></dt>
2341 <dd>Truncate a constant to another type. The bit size of CST must be larger
2342 than the bit size of TYPE. Both types must be integers.</dd>
2344 <dt><b><tt>zext ( CST to TYPE )</tt></b></dt>
2345 <dd>Zero extend a constant to another type. The bit size of CST must be
2346 smaller or equal to the bit size of TYPE. Both types must be
2349 <dt><b><tt>sext ( CST to TYPE )</tt></b></dt>
2350 <dd>Sign extend a constant to another type. The bit size of CST must be
2351 smaller or equal to the bit size of TYPE. Both types must be
2354 <dt><b><tt>fptrunc ( CST to TYPE )</tt></b></dt>
2355 <dd>Truncate a floating point constant to another floating point type. The
2356 size of CST must be larger than the size of TYPE. Both types must be
2357 floating point.</dd>
2359 <dt><b><tt>fpext ( CST to TYPE )</tt></b></dt>
2360 <dd>Floating point extend a constant to another type. The size of CST must be
2361 smaller or equal to the size of TYPE. Both types must be floating
2364 <dt><b><tt>fptoui ( CST to TYPE )</tt></b></dt>
2365 <dd>Convert a floating point constant to the corresponding unsigned integer
2366 constant. TYPE must be a scalar or vector integer type. CST must be of
2367 scalar or vector floating point type. Both CST and TYPE must be scalars,
2368 or vectors of the same number of elements. If the value won't fit in the
2369 integer type, the results are undefined.</dd>
2371 <dt><b><tt>fptosi ( CST to TYPE )</tt></b></dt>
2372 <dd>Convert a floating point constant to the corresponding signed integer
2373 constant. TYPE must be a scalar or vector integer type. CST must be of
2374 scalar or vector floating point type. Both CST and TYPE must be scalars,
2375 or vectors of the same number of elements. If the value won't fit in the
2376 integer type, the results are undefined.</dd>
2378 <dt><b><tt>uitofp ( CST to TYPE )</tt></b></dt>
2379 <dd>Convert an unsigned integer constant to the corresponding floating point
2380 constant. TYPE must be a scalar or vector floating point type. CST must be
2381 of scalar or vector integer type. Both CST and TYPE must be scalars, or
2382 vectors of the same number of elements. If the value won't fit in the
2383 floating point type, the results are undefined.</dd>
2385 <dt><b><tt>sitofp ( CST to TYPE )</tt></b></dt>
2386 <dd>Convert a signed integer constant to the corresponding floating point
2387 constant. TYPE must be a scalar or vector floating point type. CST must be
2388 of scalar or vector integer type. Both CST and TYPE must be scalars, or
2389 vectors of the same number of elements. If the value won't fit in the
2390 floating point type, the results are undefined.</dd>
2392 <dt><b><tt>ptrtoint ( CST to TYPE )</tt></b></dt>
2393 <dd>Convert a pointer typed constant to the corresponding integer constant
2394 <tt>TYPE</tt> must be an integer type. <tt>CST</tt> must be of pointer
2395 type. The <tt>CST</tt> value is zero extended, truncated, or unchanged to
2396 make it fit in <tt>TYPE</tt>.</dd>
2398 <dt><b><tt>inttoptr ( CST to TYPE )</tt></b></dt>
2399 <dd>Convert a integer constant to a pointer constant. TYPE must be a pointer
2400 type. CST must be of integer type. The CST value is zero extended,
2401 truncated, or unchanged to make it fit in a pointer size. This one is
2402 <i>really</i> dangerous!</dd>
2404 <dt><b><tt>bitcast ( CST to TYPE )</tt></b></dt>
2405 <dd>Convert a constant, CST, to another TYPE. The constraints of the operands
2406 are the same as those for the <a href="#i_bitcast">bitcast
2407 instruction</a>.</dd>
2409 <dt><b><tt>getelementptr ( CSTPTR, IDX0, IDX1, ... )</tt></b></dt>
2410 <dt><b><tt>getelementptr inbounds ( CSTPTR, IDX0, IDX1, ... )</tt></b></dt>
2411 <dd>Perform the <a href="#i_getelementptr">getelementptr operation</a> on
2412 constants. As with the <a href="#i_getelementptr">getelementptr</a>
2413 instruction, the index list may have zero or more indexes, which are
2414 required to make sense for the type of "CSTPTR".</dd>
2416 <dt><b><tt>select ( COND, VAL1, VAL2 )</tt></b></dt>
2417 <dd>Perform the <a href="#i_select">select operation</a> on constants.</dd>
2419 <dt><b><tt>icmp COND ( VAL1, VAL2 )</tt></b></dt>
2420 <dd>Performs the <a href="#i_icmp">icmp operation</a> on constants.</dd>
2422 <dt><b><tt>fcmp COND ( VAL1, VAL2 )</tt></b></dt>
2423 <dd>Performs the <a href="#i_fcmp">fcmp operation</a> on constants.</dd>
2425 <dt><b><tt>extractelement ( VAL, IDX )</tt></b></dt>
2426 <dd>Perform the <a href="#i_extractelement">extractelement operation</a> on
2429 <dt><b><tt>insertelement ( VAL, ELT, IDX )</tt></b></dt>
2430 <dd>Perform the <a href="#i_insertelement">insertelement operation</a> on
2433 <dt><b><tt>shufflevector ( VEC1, VEC2, IDXMASK )</tt></b></dt>
2434 <dd>Perform the <a href="#i_shufflevector">shufflevector operation</a> on
2437 <dt><b><tt>OPCODE ( LHS, RHS )</tt></b></dt>
2438 <dd>Perform the specified operation of the LHS and RHS constants. OPCODE may
2439 be any of the <a href="#binaryops">binary</a>
2440 or <a href="#bitwiseops">bitwise binary</a> operations. The constraints
2441 on operands are the same as those for the corresponding instruction
2442 (e.g. no bitwise operations on floating point values are allowed).</dd>
2447 <!-- *********************************************************************** -->
2448 <div class="doc_section"> <a name="othervalues">Other Values</a> </div>
2449 <!-- *********************************************************************** -->
2451 <!-- ======================================================================= -->
2452 <div class="doc_subsection">
2453 <a name="inlineasm">Inline Assembler Expressions</a>
2456 <div class="doc_text">
2458 <p>LLVM supports inline assembler expressions (as opposed
2459 to <a href="#moduleasm"> Module-Level Inline Assembly</a>) through the use of
2460 a special value. This value represents the inline assembler as a string
2461 (containing the instructions to emit), a list of operand constraints (stored
2462 as a string), a flag that indicates whether or not the inline asm
2463 expression has side effects, and a flag indicating whether the function
2464 containing the asm needs to align its stack conservatively. An example
2465 inline assembler expression is:</p>
2467 <div class="doc_code">
2469 i32 (i32) asm "bswap $0", "=r,r"
2473 <p>Inline assembler expressions may <b>only</b> be used as the callee operand of
2474 a <a href="#i_call"><tt>call</tt> instruction</a>. Thus, typically we
2477 <div class="doc_code">
2479 %X = call i32 asm "<a href="#int_bswap">bswap</a> $0", "=r,r"(i32 %Y)
2483 <p>Inline asms with side effects not visible in the constraint list must be
2484 marked as having side effects. This is done through the use of the
2485 '<tt>sideeffect</tt>' keyword, like so:</p>
2487 <div class="doc_code">
2489 call void asm sideeffect "eieio", ""()
2493 <p>In some cases inline asms will contain code that will not work unless the
2494 stack is aligned in some way, such as calls or SSE instructions on x86,
2495 yet will not contain code that does that alignment within the asm.
2496 The compiler should make conservative assumptions about what the asm might
2497 contain and should generate its usual stack alignment code in the prologue
2498 if the '<tt>alignstack</tt>' keyword is present:</p>
2500 <div class="doc_code">
2502 call void asm alignstack "eieio", ""()
2506 <p>If both keywords appear the '<tt>sideeffect</tt>' keyword must come
2509 <p>TODO: The format of the asm and constraints string still need to be
2510 documented here. Constraints on what can be done (e.g. duplication, moving,
2511 etc need to be documented). This is probably best done by reference to
2512 another document that covers inline asm from a holistic perspective.</p>
2516 <!-- ======================================================================= -->
2517 <div class="doc_subsection"><a name="metadata">Metadata Nodes and Metadata
2521 <div class="doc_text">
2523 <p>LLVM IR allows metadata to be attached to instructions in the program that
2524 can convey extra information about the code to the optimizers and code
2525 generator. One example application of metadata is source-level debug
2526 information. There are two metadata primitives: strings and nodes. All
2527 metadata has the <tt>metadata</tt> type and is identified in syntax by a
2528 preceding exclamation point ('<tt>!</tt>').</p>
2530 <p>A metadata string is a string surrounded by double quotes. It can contain
2531 any character by escaping non-printable characters with "\xx" where "xx" is
2532 the two digit hex code. For example: "<tt>!"test\00"</tt>".</p>
2534 <p>Metadata nodes are represented with notation similar to structure constants
2535 (a comma separated list of elements, surrounded by braces and preceded by an
2536 exclamation point). For example: "<tt>!{ metadata !"test\00", i32
2537 10}</tt>". Metadata nodes can have any values as their operand.</p>
2539 <p>A <a href="#namedmetadatastructure">named metadata</a> is a collection of
2540 metadata nodes, which can be looked up in the module symbol table. For
2541 example: "<tt>!foo = metadata !{!4, !3}</tt>".
2543 <p>Metadata can be used as function arguments. Here <tt>llvm.dbg.value</tt>
2544 function is using two metadata arguments.
2546 <div class="doc_code">
2548 call void @llvm.dbg.value(metadata !24, i64 0, metadata !25)
2552 <p>Metadata can be attached with an instruction. Here metadata <tt>!21</tt> is
2553 attached with <tt>add</tt> instruction using <tt>!dbg</tt> identifier.
2555 <div class="doc_code">
2557 %indvar.next = add i64 %indvar, 1, !dbg !21
2563 <!-- *********************************************************************** -->
2564 <div class="doc_section">
2565 <a name="intrinsic_globals">Intrinsic Global Variables</a>
2567 <!-- *********************************************************************** -->
2569 <p>LLVM has a number of "magic" global variables that contain data that affect
2570 code generation or other IR semantics. These are documented here. All globals
2571 of this sort should have a section specified as "<tt>llvm.metadata</tt>". This
2572 section and all globals that start with "<tt>llvm.</tt>" are reserved for use
2575 <!-- ======================================================================= -->
2576 <div class="doc_subsection">
2577 <a name="intg_used">The '<tt>llvm.used</tt>' Global Variable</a>
2580 <div class="doc_text">
2582 <p>The <tt>@llvm.used</tt> global is an array with i8* element type which has <a
2583 href="#linkage_appending">appending linkage</a>. This array contains a list of
2584 pointers to global variables and functions which may optionally have a pointer
2585 cast formed of bitcast or getelementptr. For example, a legal use of it is:</p>
2591 @llvm.used = appending global [2 x i8*] [
2593 i8* bitcast (i32* @Y to i8*)
2594 ], section "llvm.metadata"
2597 <p>If a global variable appears in the <tt>@llvm.used</tt> list, then the
2598 compiler, assembler, and linker are required to treat the symbol as if there is
2599 a reference to the global that it cannot see. For example, if a variable has
2600 internal linkage and no references other than that from the <tt>@llvm.used</tt>
2601 list, it cannot be deleted. This is commonly used to represent references from
2602 inline asms and other things the compiler cannot "see", and corresponds to
2603 "attribute((used))" in GNU C.</p>
2605 <p>On some targets, the code generator must emit a directive to the assembler or
2606 object file to prevent the assembler and linker from molesting the symbol.</p>
2610 <!-- ======================================================================= -->
2611 <div class="doc_subsection">
2612 <a name="intg_compiler_used">The '<tt>llvm.compiler.used</tt>' Global Variable</a>
2615 <div class="doc_text">
2617 <p>The <tt>@llvm.compiler.used</tt> directive is the same as the
2618 <tt>@llvm.used</tt> directive, except that it only prevents the compiler from
2619 touching the symbol. On targets that support it, this allows an intelligent
2620 linker to optimize references to the symbol without being impeded as it would be
2621 by <tt>@llvm.used</tt>.</p>
2623 <p>This is a rare construct that should only be used in rare circumstances, and
2624 should not be exposed to source languages.</p>
2628 <!-- ======================================================================= -->
2629 <div class="doc_subsection">
2630 <a name="intg_global_ctors">The '<tt>llvm.global_ctors</tt>' Global Variable</a>
2633 <div class="doc_text">
2635 <p>TODO: Describe this.</p>
2639 <!-- ======================================================================= -->
2640 <div class="doc_subsection">
2641 <a name="intg_global_dtors">The '<tt>llvm.global_dtors</tt>' Global Variable</a>
2644 <div class="doc_text">
2646 <p>TODO: Describe this.</p>
2651 <!-- *********************************************************************** -->
2652 <div class="doc_section"> <a name="instref">Instruction Reference</a> </div>
2653 <!-- *********************************************************************** -->
2655 <div class="doc_text">
2657 <p>The LLVM instruction set consists of several different classifications of
2658 instructions: <a href="#terminators">terminator
2659 instructions</a>, <a href="#binaryops">binary instructions</a>,
2660 <a href="#bitwiseops">bitwise binary instructions</a>,
2661 <a href="#memoryops">memory instructions</a>, and
2662 <a href="#otherops">other instructions</a>.</p>
2666 <!-- ======================================================================= -->
2667 <div class="doc_subsection"> <a name="terminators">Terminator
2668 Instructions</a> </div>
2670 <div class="doc_text">
2672 <p>As mentioned <a href="#functionstructure">previously</a>, every basic block
2673 in a program ends with a "Terminator" instruction, which indicates which
2674 block should be executed after the current block is finished. These
2675 terminator instructions typically yield a '<tt>void</tt>' value: they produce
2676 control flow, not values (the one exception being the
2677 '<a href="#i_invoke"><tt>invoke</tt></a>' instruction).</p>
2679 <p>There are six different terminator instructions: the
2680 '<a href="#i_ret"><tt>ret</tt></a>' instruction, the
2681 '<a href="#i_br"><tt>br</tt></a>' instruction, the
2682 '<a href="#i_switch"><tt>switch</tt></a>' instruction, the
2683 '<a href="#i_indirectbr">'<tt>indirectbr</tt></a>' Instruction, the
2684 '<a href="#i_invoke"><tt>invoke</tt></a>' instruction, the
2685 '<a href="#i_unwind"><tt>unwind</tt></a>' instruction, and the
2686 '<a href="#i_unreachable"><tt>unreachable</tt></a>' instruction.</p>
2690 <!-- _______________________________________________________________________ -->
2691 <div class="doc_subsubsection"> <a name="i_ret">'<tt>ret</tt>'
2692 Instruction</a> </div>
2694 <div class="doc_text">
2698 ret <type> <value> <i>; Return a value from a non-void function</i>
2699 ret void <i>; Return from void function</i>
2703 <p>The '<tt>ret</tt>' instruction is used to return control flow (and optionally
2704 a value) from a function back to the caller.</p>
2706 <p>There are two forms of the '<tt>ret</tt>' instruction: one that returns a
2707 value and then causes control flow, and one that just causes control flow to
2711 <p>The '<tt>ret</tt>' instruction optionally accepts a single argument, the
2712 return value. The type of the return value must be a
2713 '<a href="#t_firstclass">first class</a>' type.</p>
2715 <p>A function is not <a href="#wellformed">well formed</a> if it it has a
2716 non-void return type and contains a '<tt>ret</tt>' instruction with no return
2717 value or a return value with a type that does not match its type, or if it
2718 has a void return type and contains a '<tt>ret</tt>' instruction with a
2722 <p>When the '<tt>ret</tt>' instruction is executed, control flow returns back to
2723 the calling function's context. If the caller is a
2724 "<a href="#i_call"><tt>call</tt></a>" instruction, execution continues at the
2725 instruction after the call. If the caller was an
2726 "<a href="#i_invoke"><tt>invoke</tt></a>" instruction, execution continues at
2727 the beginning of the "normal" destination block. If the instruction returns
2728 a value, that value shall set the call or invoke instruction's return
2733 ret i32 5 <i>; Return an integer value of 5</i>
2734 ret void <i>; Return from a void function</i>
2735 ret { i32, i8 } { i32 4, i8 2 } <i>; Return a struct of values 4 and 2</i>
2739 <!-- _______________________________________________________________________ -->
2740 <div class="doc_subsubsection"> <a name="i_br">'<tt>br</tt>' Instruction</a> </div>
2742 <div class="doc_text">
2746 br i1 <cond>, label <iftrue>, label <iffalse><br> br label <dest> <i>; Unconditional branch</i>
2750 <p>The '<tt>br</tt>' instruction is used to cause control flow to transfer to a
2751 different basic block in the current function. There are two forms of this
2752 instruction, corresponding to a conditional branch and an unconditional
2756 <p>The conditional branch form of the '<tt>br</tt>' instruction takes a single
2757 '<tt>i1</tt>' value and two '<tt>label</tt>' values. The unconditional form
2758 of the '<tt>br</tt>' instruction takes a single '<tt>label</tt>' value as a
2762 <p>Upon execution of a conditional '<tt>br</tt>' instruction, the '<tt>i1</tt>'
2763 argument is evaluated. If the value is <tt>true</tt>, control flows to the
2764 '<tt>iftrue</tt>' <tt>label</tt> argument. If "cond" is <tt>false</tt>,
2765 control flows to the '<tt>iffalse</tt>' <tt>label</tt> argument.</p>
2770 %cond = <a href="#i_icmp">icmp</a> eq i32 %a, %b
2771 br i1 %cond, label %IfEqual, label %IfUnequal
2773 <a href="#i_ret">ret</a> i32 1
2775 <a href="#i_ret">ret</a> i32 0
2780 <!-- _______________________________________________________________________ -->
2781 <div class="doc_subsubsection">
2782 <a name="i_switch">'<tt>switch</tt>' Instruction</a>
2785 <div class="doc_text">
2789 switch <intty> <value>, label <defaultdest> [ <intty> <val>, label <dest> ... ]
2793 <p>The '<tt>switch</tt>' instruction is used to transfer control flow to one of
2794 several different places. It is a generalization of the '<tt>br</tt>'
2795 instruction, allowing a branch to occur to one of many possible
2799 <p>The '<tt>switch</tt>' instruction uses three parameters: an integer
2800 comparison value '<tt>value</tt>', a default '<tt>label</tt>' destination,
2801 and an array of pairs of comparison value constants and '<tt>label</tt>'s.
2802 The table is not allowed to contain duplicate constant entries.</p>
2805 <p>The <tt>switch</tt> instruction specifies a table of values and
2806 destinations. When the '<tt>switch</tt>' instruction is executed, this table
2807 is searched for the given value. If the value is found, control flow is
2808 transferred to the corresponding destination; otherwise, control flow is
2809 transferred to the default destination.</p>
2811 <h5>Implementation:</h5>
2812 <p>Depending on properties of the target machine and the particular
2813 <tt>switch</tt> instruction, this instruction may be code generated in
2814 different ways. For example, it could be generated as a series of chained
2815 conditional branches or with a lookup table.</p>
2819 <i>; Emulate a conditional br instruction</i>
2820 %Val = <a href="#i_zext">zext</a> i1 %value to i32
2821 switch i32 %Val, label %truedest [ i32 0, label %falsedest ]
2823 <i>; Emulate an unconditional br instruction</i>
2824 switch i32 0, label %dest [ ]
2826 <i>; Implement a jump table:</i>
2827 switch i32 %val, label %otherwise [ i32 0, label %onzero
2829 i32 2, label %ontwo ]
2835 <!-- _______________________________________________________________________ -->
2836 <div class="doc_subsubsection">
2837 <a name="i_indirectbr">'<tt>indirectbr</tt>' Instruction</a>
2840 <div class="doc_text">
2844 indirectbr <somety>* <address>, [ label <dest1>, label <dest2>, ... ]
2849 <p>The '<tt>indirectbr</tt>' instruction implements an indirect branch to a label
2850 within the current function, whose address is specified by
2851 "<tt>address</tt>". Address must be derived from a <a
2852 href="#blockaddress">blockaddress</a> constant.</p>
2856 <p>The '<tt>address</tt>' argument is the address of the label to jump to. The
2857 rest of the arguments indicate the full set of possible destinations that the
2858 address may point to. Blocks are allowed to occur multiple times in the
2859 destination list, though this isn't particularly useful.</p>
2861 <p>This destination list is required so that dataflow analysis has an accurate
2862 understanding of the CFG.</p>
2866 <p>Control transfers to the block specified in the address argument. All
2867 possible destination blocks must be listed in the label list, otherwise this
2868 instruction has undefined behavior. This implies that jumps to labels
2869 defined in other functions have undefined behavior as well.</p>
2871 <h5>Implementation:</h5>
2873 <p>This is typically implemented with a jump through a register.</p>
2877 indirectbr i8* %Addr, [ label %bb1, label %bb2, label %bb3 ]
2883 <!-- _______________________________________________________________________ -->
2884 <div class="doc_subsubsection">
2885 <a name="i_invoke">'<tt>invoke</tt>' Instruction</a>
2888 <div class="doc_text">
2892 <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>]
2893 to label <normal label> unwind label <exception label>
2897 <p>The '<tt>invoke</tt>' instruction causes control to transfer to a specified
2898 function, with the possibility of control flow transfer to either the
2899 '<tt>normal</tt>' label or the '<tt>exception</tt>' label. If the callee
2900 function returns with the "<tt><a href="#i_ret">ret</a></tt>" instruction,
2901 control flow will return to the "normal" label. If the callee (or any
2902 indirect callees) returns with the "<a href="#i_unwind"><tt>unwind</tt></a>"
2903 instruction, control is interrupted and continued at the dynamically nearest
2904 "exception" label.</p>
2907 <p>This instruction requires several arguments:</p>
2910 <li>The optional "cconv" marker indicates which <a href="#callingconv">calling
2911 convention</a> the call should use. If none is specified, the call
2912 defaults to using C calling conventions.</li>
2914 <li>The optional <a href="#paramattrs">Parameter Attributes</a> list for
2915 return values. Only '<tt>zeroext</tt>', '<tt>signext</tt>', and
2916 '<tt>inreg</tt>' attributes are valid here.</li>
2918 <li>'<tt>ptr to function ty</tt>': shall be the signature of the pointer to
2919 function value being invoked. In most cases, this is a direct function
2920 invocation, but indirect <tt>invoke</tt>s are just as possible, branching
2921 off an arbitrary pointer to function value.</li>
2923 <li>'<tt>function ptr val</tt>': An LLVM value containing a pointer to a
2924 function to be invoked. </li>
2926 <li>'<tt>function args</tt>': argument list whose types match the function
2927 signature argument types and parameter attributes. All arguments must be
2928 of <a href="#t_firstclass">first class</a> type. If the function
2929 signature indicates the function accepts a variable number of arguments,
2930 the extra arguments can be specified.</li>
2932 <li>'<tt>normal label</tt>': the label reached when the called function
2933 executes a '<tt><a href="#i_ret">ret</a></tt>' instruction. </li>
2935 <li>'<tt>exception label</tt>': the label reached when a callee returns with
2936 the <a href="#i_unwind"><tt>unwind</tt></a> instruction. </li>
2938 <li>The optional <a href="#fnattrs">function attributes</a> list. Only
2939 '<tt>noreturn</tt>', '<tt>nounwind</tt>', '<tt>readonly</tt>' and
2940 '<tt>readnone</tt>' attributes are valid here.</li>
2944 <p>This instruction is designed to operate as a standard
2945 '<tt><a href="#i_call">call</a></tt>' instruction in most regards. The
2946 primary difference is that it establishes an association with a label, which
2947 is used by the runtime library to unwind the stack.</p>
2949 <p>This instruction is used in languages with destructors to ensure that proper
2950 cleanup is performed in the case of either a <tt>longjmp</tt> or a thrown
2951 exception. Additionally, this is important for implementation of
2952 '<tt>catch</tt>' clauses in high-level languages that support them.</p>
2954 <p>For the purposes of the SSA form, the definition of the value returned by the
2955 '<tt>invoke</tt>' instruction is deemed to occur on the edge from the current
2956 block to the "normal" label. If the callee unwinds then no return value is
2959 <p>Note that the code generator does not yet completely support unwind, and
2960 that the invoke/unwind semantics are likely to change in future versions.</p>
2964 %retval = invoke i32 @Test(i32 15) to label %Continue
2965 unwind label %TestCleanup <i>; {i32}:retval set</i>
2966 %retval = invoke <a href="#callingconv">coldcc</a> i32 %Testfnptr(i32 15) to label %Continue
2967 unwind label %TestCleanup <i>; {i32}:retval set</i>
2972 <!-- _______________________________________________________________________ -->
2974 <div class="doc_subsubsection"> <a name="i_unwind">'<tt>unwind</tt>'
2975 Instruction</a> </div>
2977 <div class="doc_text">
2985 <p>The '<tt>unwind</tt>' instruction unwinds the stack, continuing control flow
2986 at the first callee in the dynamic call stack which used
2987 an <a href="#i_invoke"><tt>invoke</tt></a> instruction to perform the call.
2988 This is primarily used to implement exception handling.</p>
2991 <p>The '<tt>unwind</tt>' instruction causes execution of the current function to
2992 immediately halt. The dynamic call stack is then searched for the
2993 first <a href="#i_invoke"><tt>invoke</tt></a> instruction on the call stack.
2994 Once found, execution continues at the "exceptional" destination block
2995 specified by the <tt>invoke</tt> instruction. If there is no <tt>invoke</tt>
2996 instruction in the dynamic call chain, undefined behavior results.</p>
2998 <p>Note that the code generator does not yet completely support unwind, and
2999 that the invoke/unwind semantics are likely to change in future versions.</p>
3003 <!-- _______________________________________________________________________ -->
3005 <div class="doc_subsubsection"> <a name="i_unreachable">'<tt>unreachable</tt>'
3006 Instruction</a> </div>
3008 <div class="doc_text">
3016 <p>The '<tt>unreachable</tt>' instruction has no defined semantics. This
3017 instruction is used to inform the optimizer that a particular portion of the
3018 code is not reachable. This can be used to indicate that the code after a
3019 no-return function cannot be reached, and other facts.</p>
3022 <p>The '<tt>unreachable</tt>' instruction has no defined semantics.</p>
3026 <!-- ======================================================================= -->
3027 <div class="doc_subsection"> <a name="binaryops">Binary Operations</a> </div>
3029 <div class="doc_text">
3031 <p>Binary operators are used to do most of the computation in a program. They
3032 require two operands of the same type, execute an operation on them, and
3033 produce a single value. The operands might represent multiple data, as is
3034 the case with the <a href="#t_vector">vector</a> data type. The result value
3035 has the same type as its operands.</p>
3037 <p>There are several different binary operators:</p>
3041 <!-- _______________________________________________________________________ -->
3042 <div class="doc_subsubsection">
3043 <a name="i_add">'<tt>add</tt>' Instruction</a>
3046 <div class="doc_text">
3050 <result> = add <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3051 <result> = add nuw <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3052 <result> = add nsw <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3053 <result> = add nuw nsw <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3057 <p>The '<tt>add</tt>' instruction returns the sum of its two operands.</p>
3060 <p>The two arguments to the '<tt>add</tt>' instruction must
3061 be <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of
3062 integer values. Both arguments must have identical types.</p>
3065 <p>The value produced is the integer sum of the two operands.</p>
3067 <p>If the sum has unsigned overflow, the result returned is the mathematical
3068 result modulo 2<sup>n</sup>, where n is the bit width of the result.</p>
3070 <p>Because LLVM integers use a two's complement representation, this instruction
3071 is appropriate for both signed and unsigned integers.</p>
3073 <p><tt>nuw</tt> and <tt>nsw</tt> stand for "No Unsigned Wrap"
3074 and "No Signed Wrap", respectively. If the <tt>nuw</tt> and/or
3075 <tt>nsw</tt> keywords are present, the result value of the <tt>add</tt>
3076 is undefined if unsigned and/or signed overflow, respectively, occurs.</p>
3080 <result> = add i32 4, %var <i>; yields {i32}:result = 4 + %var</i>
3085 <!-- _______________________________________________________________________ -->
3086 <div class="doc_subsubsection">
3087 <a name="i_fadd">'<tt>fadd</tt>' Instruction</a>
3090 <div class="doc_text">
3094 <result> = fadd <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3098 <p>The '<tt>fadd</tt>' instruction returns the sum of its two operands.</p>
3101 <p>The two arguments to the '<tt>fadd</tt>' instruction must be
3102 <a href="#t_floating">floating point</a> or <a href="#t_vector">vector</a> of
3103 floating point values. Both arguments must have identical types.</p>
3106 <p>The value produced is the floating point sum of the two operands.</p>
3110 <result> = fadd float 4.0, %var <i>; yields {float}:result = 4.0 + %var</i>
3115 <!-- _______________________________________________________________________ -->
3116 <div class="doc_subsubsection">
3117 <a name="i_sub">'<tt>sub</tt>' Instruction</a>
3120 <div class="doc_text">
3124 <result> = sub <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3125 <result> = sub nuw <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3126 <result> = sub nsw <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3127 <result> = sub nuw nsw <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3131 <p>The '<tt>sub</tt>' instruction returns the difference of its two
3134 <p>Note that the '<tt>sub</tt>' instruction is used to represent the
3135 '<tt>neg</tt>' instruction present in most other intermediate
3136 representations.</p>
3139 <p>The two arguments to the '<tt>sub</tt>' instruction must
3140 be <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of
3141 integer values. Both arguments must have identical types.</p>
3144 <p>The value produced is the integer difference of the two operands.</p>
3146 <p>If the difference has unsigned overflow, the result returned is the
3147 mathematical result modulo 2<sup>n</sup>, where n is the bit width of the
3150 <p>Because LLVM integers use a two's complement representation, this instruction
3151 is appropriate for both signed and unsigned integers.</p>
3153 <p><tt>nuw</tt> and <tt>nsw</tt> stand for "No Unsigned Wrap"
3154 and "No Signed Wrap", respectively. If the <tt>nuw</tt> and/or
3155 <tt>nsw</tt> keywords are present, the result value of the <tt>sub</tt>
3156 is undefined if unsigned and/or signed overflow, respectively, occurs.</p>
3160 <result> = sub i32 4, %var <i>; yields {i32}:result = 4 - %var</i>
3161 <result> = sub i32 0, %val <i>; yields {i32}:result = -%var</i>
3166 <!-- _______________________________________________________________________ -->
3167 <div class="doc_subsubsection">
3168 <a name="i_fsub">'<tt>fsub</tt>' Instruction</a>
3171 <div class="doc_text">
3175 <result> = fsub <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3179 <p>The '<tt>fsub</tt>' instruction returns the difference of its two
3182 <p>Note that the '<tt>fsub</tt>' instruction is used to represent the
3183 '<tt>fneg</tt>' instruction present in most other intermediate
3184 representations.</p>
3187 <p>The two arguments to the '<tt>fsub</tt>' instruction must be
3188 <a href="#t_floating">floating point</a> or <a href="#t_vector">vector</a> of
3189 floating point values. Both arguments must have identical types.</p>
3192 <p>The value produced is the floating point difference of the two operands.</p>
3196 <result> = fsub float 4.0, %var <i>; yields {float}:result = 4.0 - %var</i>
3197 <result> = fsub float -0.0, %val <i>; yields {float}:result = -%var</i>
3202 <!-- _______________________________________________________________________ -->
3203 <div class="doc_subsubsection">
3204 <a name="i_mul">'<tt>mul</tt>' Instruction</a>
3207 <div class="doc_text">
3211 <result> = mul <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3212 <result> = mul nuw <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3213 <result> = mul nsw <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3214 <result> = mul nuw nsw <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3218 <p>The '<tt>mul</tt>' instruction returns the product of its two operands.</p>
3221 <p>The two arguments to the '<tt>mul</tt>' instruction must
3222 be <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of
3223 integer values. Both arguments must have identical types.</p>
3226 <p>The value produced is the integer product of the two operands.</p>
3228 <p>If the result of the multiplication has unsigned overflow, the result
3229 returned is the mathematical result modulo 2<sup>n</sup>, where n is the bit
3230 width of the result.</p>
3232 <p>Because LLVM integers use a two's complement representation, and the result
3233 is the same width as the operands, this instruction returns the correct
3234 result for both signed and unsigned integers. If a full product
3235 (e.g. <tt>i32</tt>x<tt>i32</tt>-><tt>i64</tt>) is needed, the operands should
3236 be sign-extended or zero-extended as appropriate to the width of the full
3239 <p><tt>nuw</tt> and <tt>nsw</tt> stand for "No Unsigned Wrap"
3240 and "No Signed Wrap", respectively. If the <tt>nuw</tt> and/or
3241 <tt>nsw</tt> keywords are present, the result value of the <tt>mul</tt>
3242 is undefined if unsigned and/or signed overflow, respectively, occurs.</p>
3246 <result> = mul i32 4, %var <i>; yields {i32}:result = 4 * %var</i>
3251 <!-- _______________________________________________________________________ -->
3252 <div class="doc_subsubsection">
3253 <a name="i_fmul">'<tt>fmul</tt>' Instruction</a>
3256 <div class="doc_text">
3260 <result> = fmul <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3264 <p>The '<tt>fmul</tt>' instruction returns the product of its two operands.</p>
3267 <p>The two arguments to the '<tt>fmul</tt>' instruction must be
3268 <a href="#t_floating">floating point</a> or <a href="#t_vector">vector</a> of
3269 floating point values. Both arguments must have identical types.</p>
3272 <p>The value produced is the floating point product of the two operands.</p>
3276 <result> = fmul float 4.0, %var <i>; yields {float}:result = 4.0 * %var</i>
3281 <!-- _______________________________________________________________________ -->
3282 <div class="doc_subsubsection"> <a name="i_udiv">'<tt>udiv</tt>' Instruction
3285 <div class="doc_text">
3289 <result> = udiv <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3293 <p>The '<tt>udiv</tt>' instruction returns the quotient of its two operands.</p>
3296 <p>The two arguments to the '<tt>udiv</tt>' instruction must be
3297 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
3298 values. Both arguments must have identical types.</p>
3301 <p>The value produced is the unsigned integer quotient of the two operands.</p>
3303 <p>Note that unsigned integer division and signed integer division are distinct
3304 operations; for signed integer division, use '<tt>sdiv</tt>'.</p>
3306 <p>Division by zero leads to undefined behavior.</p>
3310 <result> = udiv i32 4, %var <i>; yields {i32}:result = 4 / %var</i>
3315 <!-- _______________________________________________________________________ -->
3316 <div class="doc_subsubsection"> <a name="i_sdiv">'<tt>sdiv</tt>' Instruction
3319 <div class="doc_text">
3323 <result> = sdiv <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3324 <result> = sdiv exact <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3328 <p>The '<tt>sdiv</tt>' instruction returns the quotient of its two operands.</p>
3331 <p>The two arguments to the '<tt>sdiv</tt>' instruction must be
3332 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
3333 values. Both arguments must have identical types.</p>
3336 <p>The value produced is the signed integer quotient of the two operands rounded
3339 <p>Note that signed integer division and unsigned integer division are distinct
3340 operations; for unsigned integer division, use '<tt>udiv</tt>'.</p>
3342 <p>Division by zero leads to undefined behavior. Overflow also leads to
3343 undefined behavior; this is a rare case, but can occur, for example, by doing
3344 a 32-bit division of -2147483648 by -1.</p>
3346 <p>If the <tt>exact</tt> keyword is present, the result value of the
3347 <tt>sdiv</tt> is undefined if the result would be rounded or if overflow
3352 <result> = sdiv i32 4, %var <i>; yields {i32}:result = 4 / %var</i>
3357 <!-- _______________________________________________________________________ -->
3358 <div class="doc_subsubsection"> <a name="i_fdiv">'<tt>fdiv</tt>'
3359 Instruction</a> </div>
3361 <div class="doc_text">
3365 <result> = fdiv <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3369 <p>The '<tt>fdiv</tt>' instruction returns the quotient of its two operands.</p>
3372 <p>The two arguments to the '<tt>fdiv</tt>' instruction must be
3373 <a href="#t_floating">floating point</a> or <a href="#t_vector">vector</a> of
3374 floating point values. Both arguments must have identical types.</p>
3377 <p>The value produced is the floating point quotient of the two operands.</p>
3381 <result> = fdiv float 4.0, %var <i>; yields {float}:result = 4.0 / %var</i>
3386 <!-- _______________________________________________________________________ -->
3387 <div class="doc_subsubsection"> <a name="i_urem">'<tt>urem</tt>' Instruction</a>
3390 <div class="doc_text">
3394 <result> = urem <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3398 <p>The '<tt>urem</tt>' instruction returns the remainder from the unsigned
3399 division of its two arguments.</p>
3402 <p>The two arguments to the '<tt>urem</tt>' instruction must be
3403 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
3404 values. Both arguments must have identical types.</p>
3407 <p>This instruction returns the unsigned integer <i>remainder</i> of a division.
3408 This instruction always performs an unsigned division to get the
3411 <p>Note that unsigned integer remainder and signed integer remainder are
3412 distinct operations; for signed integer remainder, use '<tt>srem</tt>'.</p>
3414 <p>Taking the remainder of a division by zero leads to undefined behavior.</p>
3418 <result> = urem i32 4, %var <i>; yields {i32}:result = 4 % %var</i>
3423 <!-- _______________________________________________________________________ -->
3424 <div class="doc_subsubsection">
3425 <a name="i_srem">'<tt>srem</tt>' Instruction</a>
3428 <div class="doc_text">
3432 <result> = srem <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3436 <p>The '<tt>srem</tt>' instruction returns the remainder from the signed
3437 division of its two operands. This instruction can also take
3438 <a href="#t_vector">vector</a> versions of the values in which case the
3439 elements must be integers.</p>
3442 <p>The two arguments to the '<tt>srem</tt>' instruction must be
3443 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
3444 values. Both arguments must have identical types.</p>
3447 <p>This instruction returns the <i>remainder</i> of a division (where the result
3448 has the same sign as the dividend, <tt>op1</tt>), not the <i>modulo</i>
3449 operator (where the result has the same sign as the divisor, <tt>op2</tt>) of
3450 a value. For more information about the difference,
3451 see <a href="http://mathforum.org/dr.math/problems/anne.4.28.99.html">The
3452 Math Forum</a>. For a table of how this is implemented in various languages,
3453 please see <a href="http://en.wikipedia.org/wiki/Modulo_operation">
3454 Wikipedia: modulo operation</a>.</p>
3456 <p>Note that signed integer remainder and unsigned integer remainder are
3457 distinct operations; for unsigned integer remainder, use '<tt>urem</tt>'.</p>
3459 <p>Taking the remainder of a division by zero leads to undefined behavior.
3460 Overflow also leads to undefined behavior; this is a rare case, but can
3461 occur, for example, by taking the remainder of a 32-bit division of
3462 -2147483648 by -1. (The remainder doesn't actually overflow, but this rule
3463 lets srem be implemented using instructions that return both the result of
3464 the division and the remainder.)</p>
3468 <result> = srem i32 4, %var <i>; yields {i32}:result = 4 % %var</i>
3473 <!-- _______________________________________________________________________ -->
3474 <div class="doc_subsubsection">
3475 <a name="i_frem">'<tt>frem</tt>' Instruction</a> </div>
3477 <div class="doc_text">
3481 <result> = frem <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3485 <p>The '<tt>frem</tt>' instruction returns the remainder from the division of
3486 its two operands.</p>
3489 <p>The two arguments to the '<tt>frem</tt>' instruction must be
3490 <a href="#t_floating">floating point</a> or <a href="#t_vector">vector</a> of
3491 floating point values. Both arguments must have identical types.</p>
3494 <p>This instruction returns the <i>remainder</i> of a division. The remainder
3495 has the same sign as the dividend.</p>
3499 <result> = frem float 4.0, %var <i>; yields {float}:result = 4.0 % %var</i>
3504 <!-- ======================================================================= -->
3505 <div class="doc_subsection"> <a name="bitwiseops">Bitwise Binary
3506 Operations</a> </div>
3508 <div class="doc_text">
3510 <p>Bitwise binary operators are used to do various forms of bit-twiddling in a
3511 program. They are generally very efficient instructions and can commonly be
3512 strength reduced from other instructions. They require two operands of the
3513 same type, execute an operation on them, and produce a single value. The
3514 resulting value is the same type as its operands.</p>
3518 <!-- _______________________________________________________________________ -->
3519 <div class="doc_subsubsection"> <a name="i_shl">'<tt>shl</tt>'
3520 Instruction</a> </div>
3522 <div class="doc_text">
3526 <result> = shl <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3530 <p>The '<tt>shl</tt>' instruction returns the first operand shifted to the left
3531 a specified number of bits.</p>
3534 <p>Both arguments to the '<tt>shl</tt>' instruction must be the
3535 same <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of
3536 integer type. '<tt>op2</tt>' is treated as an unsigned value.</p>
3539 <p>The value produced is <tt>op1</tt> * 2<sup><tt>op2</tt></sup> mod
3540 2<sup>n</sup>, where <tt>n</tt> is the width of the result. If <tt>op2</tt>
3541 is (statically or dynamically) negative or equal to or larger than the number
3542 of bits in <tt>op1</tt>, the result is undefined. If the arguments are
3543 vectors, each vector element of <tt>op1</tt> is shifted by the corresponding
3544 shift amount in <tt>op2</tt>.</p>
3548 <result> = shl i32 4, %var <i>; yields {i32}: 4 << %var</i>
3549 <result> = shl i32 4, 2 <i>; yields {i32}: 16</i>
3550 <result> = shl i32 1, 10 <i>; yields {i32}: 1024</i>
3551 <result> = shl i32 1, 32 <i>; undefined</i>
3552 <result> = shl <2 x i32> < i32 1, i32 1>, < i32 1, i32 2> <i>; yields: result=<2 x i32> < i32 2, i32 4></i>
3557 <!-- _______________________________________________________________________ -->
3558 <div class="doc_subsubsection"> <a name="i_lshr">'<tt>lshr</tt>'
3559 Instruction</a> </div>
3561 <div class="doc_text">
3565 <result> = lshr <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3569 <p>The '<tt>lshr</tt>' instruction (logical shift right) returns the first
3570 operand shifted to the right a specified number of bits with zero fill.</p>
3573 <p>Both arguments to the '<tt>lshr</tt>' instruction must be the same
3574 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
3575 type. '<tt>op2</tt>' is treated as an unsigned value.</p>
3578 <p>This instruction always performs a logical shift right operation. The most
3579 significant bits of the result will be filled with zero bits after the shift.
3580 If <tt>op2</tt> is (statically or dynamically) equal to or larger than the
3581 number of bits in <tt>op1</tt>, the result is undefined. If the arguments are
3582 vectors, each vector element of <tt>op1</tt> is shifted by the corresponding
3583 shift amount in <tt>op2</tt>.</p>
3587 <result> = lshr i32 4, 1 <i>; yields {i32}:result = 2</i>
3588 <result> = lshr i32 4, 2 <i>; yields {i32}:result = 1</i>
3589 <result> = lshr i8 4, 3 <i>; yields {i8}:result = 0</i>
3590 <result> = lshr i8 -2, 1 <i>; yields {i8}:result = 0x7FFFFFFF </i>
3591 <result> = lshr i32 1, 32 <i>; undefined</i>
3592 <result> = lshr <2 x i32> < i32 -2, i32 4>, < i32 1, i32 2> <i>; yields: result=<2 x i32> < i32 0x7FFFFFFF, i32 1></i>
3597 <!-- _______________________________________________________________________ -->
3598 <div class="doc_subsubsection"> <a name="i_ashr">'<tt>ashr</tt>'
3599 Instruction</a> </div>
3600 <div class="doc_text">
3604 <result> = ashr <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3608 <p>The '<tt>ashr</tt>' instruction (arithmetic shift right) returns the first
3609 operand shifted to the right a specified number of bits with sign
3613 <p>Both arguments to the '<tt>ashr</tt>' instruction must be the same
3614 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
3615 type. '<tt>op2</tt>' is treated as an unsigned value.</p>
3618 <p>This instruction always performs an arithmetic shift right operation, The
3619 most significant bits of the result will be filled with the sign bit
3620 of <tt>op1</tt>. If <tt>op2</tt> is (statically or dynamically) equal to or
3621 larger than the number of bits in <tt>op1</tt>, the result is undefined. If
3622 the arguments are vectors, each vector element of <tt>op1</tt> is shifted by
3623 the corresponding shift amount in <tt>op2</tt>.</p>
3627 <result> = ashr i32 4, 1 <i>; yields {i32}:result = 2</i>
3628 <result> = ashr i32 4, 2 <i>; yields {i32}:result = 1</i>
3629 <result> = ashr i8 4, 3 <i>; yields {i8}:result = 0</i>
3630 <result> = ashr i8 -2, 1 <i>; yields {i8}:result = -1</i>
3631 <result> = ashr i32 1, 32 <i>; undefined</i>
3632 <result> = ashr <2 x i32> < i32 -2, i32 4>, < i32 1, i32 3> <i>; yields: result=<2 x i32> < i32 -1, i32 0></i>
3637 <!-- _______________________________________________________________________ -->
3638 <div class="doc_subsubsection"> <a name="i_and">'<tt>and</tt>'
3639 Instruction</a> </div>
3641 <div class="doc_text">
3645 <result> = and <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3649 <p>The '<tt>and</tt>' instruction returns the bitwise logical and of its two
3653 <p>The two arguments to the '<tt>and</tt>' instruction must be
3654 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
3655 values. Both arguments must have identical types.</p>
3658 <p>The truth table used for the '<tt>and</tt>' instruction is:</p>
3660 <table border="1" cellspacing="0" cellpadding="4">
3692 <result> = and i32 4, %var <i>; yields {i32}:result = 4 & %var</i>
3693 <result> = and i32 15, 40 <i>; yields {i32}:result = 8</i>
3694 <result> = and i32 4, 8 <i>; yields {i32}:result = 0</i>
3697 <!-- _______________________________________________________________________ -->
3698 <div class="doc_subsubsection"> <a name="i_or">'<tt>or</tt>' Instruction</a> </div>
3700 <div class="doc_text">
3704 <result> = or <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3708 <p>The '<tt>or</tt>' instruction returns the bitwise logical inclusive or of its
3712 <p>The two arguments to the '<tt>or</tt>' instruction must be
3713 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
3714 values. Both arguments must have identical types.</p>
3717 <p>The truth table used for the '<tt>or</tt>' instruction is:</p>
3719 <table border="1" cellspacing="0" cellpadding="4">
3751 <result> = or i32 4, %var <i>; yields {i32}:result = 4 | %var</i>
3752 <result> = or i32 15, 40 <i>; yields {i32}:result = 47</i>
3753 <result> = or i32 4, 8 <i>; yields {i32}:result = 12</i>
3758 <!-- _______________________________________________________________________ -->
3759 <div class="doc_subsubsection"> <a name="i_xor">'<tt>xor</tt>'
3760 Instruction</a> </div>
3762 <div class="doc_text">
3766 <result> = xor <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3770 <p>The '<tt>xor</tt>' instruction returns the bitwise logical exclusive or of
3771 its two operands. The <tt>xor</tt> is used to implement the "one's
3772 complement" operation, which is the "~" operator in C.</p>
3775 <p>The two arguments to the '<tt>xor</tt>' instruction must be
3776 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
3777 values. Both arguments must have identical types.</p>
3780 <p>The truth table used for the '<tt>xor</tt>' instruction is:</p>
3782 <table border="1" cellspacing="0" cellpadding="4">
3814 <result> = xor i32 4, %var <i>; yields {i32}:result = 4 ^ %var</i>
3815 <result> = xor i32 15, 40 <i>; yields {i32}:result = 39</i>
3816 <result> = xor i32 4, 8 <i>; yields {i32}:result = 12</i>
3817 <result> = xor i32 %V, -1 <i>; yields {i32}:result = ~%V</i>
3822 <!-- ======================================================================= -->
3823 <div class="doc_subsection">
3824 <a name="vectorops">Vector Operations</a>
3827 <div class="doc_text">
3829 <p>LLVM supports several instructions to represent vector operations in a
3830 target-independent manner. These instructions cover the element-access and
3831 vector-specific operations needed to process vectors effectively. While LLVM
3832 does directly support these vector operations, many sophisticated algorithms
3833 will want to use target-specific intrinsics to take full advantage of a
3834 specific target.</p>
3838 <!-- _______________________________________________________________________ -->
3839 <div class="doc_subsubsection">
3840 <a name="i_extractelement">'<tt>extractelement</tt>' Instruction</a>
3843 <div class="doc_text">
3847 <result> = extractelement <n x <ty>> <val>, i32 <idx> <i>; yields <ty></i>
3851 <p>The '<tt>extractelement</tt>' instruction extracts a single scalar element
3852 from a vector at a specified index.</p>
3856 <p>The first operand of an '<tt>extractelement</tt>' instruction is a value
3857 of <a href="#t_vector">vector</a> type. The second operand is an index
3858 indicating the position from which to extract the element. The index may be
3862 <p>The result is a scalar of the same type as the element type of
3863 <tt>val</tt>. Its value is the value at position <tt>idx</tt> of
3864 <tt>val</tt>. If <tt>idx</tt> exceeds the length of <tt>val</tt>, the
3865 results are undefined.</p>
3869 <result> = extractelement <4 x i32> %vec, i32 0 <i>; yields i32</i>
3874 <!-- _______________________________________________________________________ -->
3875 <div class="doc_subsubsection">
3876 <a name="i_insertelement">'<tt>insertelement</tt>' Instruction</a>
3879 <div class="doc_text">
3883 <result> = insertelement <n x <ty>> <val>, <ty> <elt>, i32 <idx> <i>; yields <n x <ty>></i>
3887 <p>The '<tt>insertelement</tt>' instruction inserts a scalar element into a
3888 vector at a specified index.</p>
3891 <p>The first operand of an '<tt>insertelement</tt>' instruction is a value
3892 of <a href="#t_vector">vector</a> type. The second operand is a scalar value
3893 whose type must equal the element type of the first operand. The third
3894 operand is an index indicating the position at which to insert the value.
3895 The index may be a variable.</p>
3898 <p>The result is a vector of the same type as <tt>val</tt>. Its element values
3899 are those of <tt>val</tt> except at position <tt>idx</tt>, where it gets the
3900 value <tt>elt</tt>. If <tt>idx</tt> exceeds the length of <tt>val</tt>, the
3901 results are undefined.</p>
3905 <result> = insertelement <4 x i32> %vec, i32 1, i32 0 <i>; yields <4 x i32></i>
3910 <!-- _______________________________________________________________________ -->
3911 <div class="doc_subsubsection">
3912 <a name="i_shufflevector">'<tt>shufflevector</tt>' Instruction</a>
3915 <div class="doc_text">
3919 <result> = shufflevector <n x <ty>> <v1>, <n x <ty>> <v2>, <m x i32> <mask> <i>; yields <m x <ty>></i>
3923 <p>The '<tt>shufflevector</tt>' instruction constructs a permutation of elements
3924 from two input vectors, returning a vector with the same element type as the
3925 input and length that is the same as the shuffle mask.</p>
3928 <p>The first two operands of a '<tt>shufflevector</tt>' instruction are vectors
3929 with types that match each other. The third argument is a shuffle mask whose
3930 element type is always 'i32'. The result of the instruction is a vector
3931 whose length is the same as the shuffle mask and whose element type is the
3932 same as the element type of the first two operands.</p>
3934 <p>The shuffle mask operand is required to be a constant vector with either
3935 constant integer or undef values.</p>
3938 <p>The elements of the two input vectors are numbered from left to right across
3939 both of the vectors. The shuffle mask operand specifies, for each element of
3940 the result vector, which element of the two input vectors the result element
3941 gets. The element selector may be undef (meaning "don't care") and the
3942 second operand may be undef if performing a shuffle from only one vector.</p>
3946 <result> = shufflevector <4 x i32> %v1, <4 x i32> %v2,
3947 <4 x i32> <i32 0, i32 4, i32 1, i32 5> <i>; yields <4 x i32></i>
3948 <result> = shufflevector <4 x i32> %v1, <4 x i32> undef,
3949 <4 x i32> <i32 0, i32 1, i32 2, i32 3> <i>; yields <4 x i32></i> - Identity shuffle.
3950 <result> = shufflevector <8 x i32> %v1, <8 x i32> undef,
3951 <4 x i32> <i32 0, i32 1, i32 2, i32 3> <i>; yields <4 x i32></i>
3952 <result> = shufflevector <4 x i32> %v1, <4 x i32> %v2,
3953 <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>
3958 <!-- ======================================================================= -->
3959 <div class="doc_subsection">
3960 <a name="aggregateops">Aggregate Operations</a>
3963 <div class="doc_text">
3965 <p>LLVM supports several instructions for working with
3966 <a href="#t_aggregate">aggregate</a> values.</p>
3970 <!-- _______________________________________________________________________ -->
3971 <div class="doc_subsubsection">
3972 <a name="i_extractvalue">'<tt>extractvalue</tt>' Instruction</a>
3975 <div class="doc_text">
3979 <result> = extractvalue <aggregate type> <val>, <idx>{, <idx>}*
3983 <p>The '<tt>extractvalue</tt>' instruction extracts the value of a member field
3984 from an <a href="#t_aggregate">aggregate</a> value.</p>
3987 <p>The first operand of an '<tt>extractvalue</tt>' instruction is a value
3988 of <a href="#t_struct">struct</a>, <a href="#t_union">union</a> or
3989 <a href="#t_array">array</a> type. The operands are constant indices to
3990 specify which value to extract in a similar manner as indices in a
3991 '<tt><a href="#i_getelementptr">getelementptr</a></tt>' instruction.</p>
3994 <p>The result is the value at the position in the aggregate specified by the
3999 <result> = extractvalue {i32, float} %agg, 0 <i>; yields i32</i>
4004 <!-- _______________________________________________________________________ -->
4005 <div class="doc_subsubsection">
4006 <a name="i_insertvalue">'<tt>insertvalue</tt>' Instruction</a>
4009 <div class="doc_text">
4013 <result> = insertvalue <aggregate type> <val>, <ty> <elt>, <idx> <i>; yields <aggregate type></i>
4017 <p>The '<tt>insertvalue</tt>' instruction inserts a value into a member field
4018 in an <a href="#t_aggregate">aggregate</a> value.</p>
4021 <p>The first operand of an '<tt>insertvalue</tt>' instruction is a value
4022 of <a href="#t_struct">struct</a>, <a href="#t_union">union</a> or
4023 <a href="#t_array">array</a> type. The second operand is a first-class
4024 value to insert. The following operands are constant indices indicating
4025 the position at which to insert the value in a similar manner as indices in a
4026 '<tt><a href="#i_getelementptr">getelementptr</a></tt>' instruction. The
4027 value to insert must have the same type as the value identified by the
4031 <p>The result is an aggregate of the same type as <tt>val</tt>. Its value is
4032 that of <tt>val</tt> except that the value at the position specified by the
4033 indices is that of <tt>elt</tt>.</p>
4037 %agg1 = insertvalue {i32, float} undef, i32 1, 0 <i>; yields {i32 1, float undef}</i>
4038 %agg2 = insertvalue {i32, float} %agg1, float %val, 1 <i>; yields {i32 1, float %val}</i>
4044 <!-- ======================================================================= -->
4045 <div class="doc_subsection">
4046 <a name="memoryops">Memory Access and Addressing Operations</a>
4049 <div class="doc_text">
4051 <p>A key design point of an SSA-based representation is how it represents
4052 memory. In LLVM, no memory locations are in SSA form, which makes things
4053 very simple. This section describes how to read, write, and allocate
4058 <!-- _______________________________________________________________________ -->
4059 <div class="doc_subsubsection">
4060 <a name="i_alloca">'<tt>alloca</tt>' Instruction</a>
4063 <div class="doc_text">
4067 <result> = alloca <type>[, i32 <NumElements>][, align <alignment>] <i>; yields {type*}:result</i>
4071 <p>The '<tt>alloca</tt>' instruction allocates memory on the stack frame of the
4072 currently executing function, to be automatically released when this function
4073 returns to its caller. The object is always allocated in the generic address
4074 space (address space zero).</p>
4077 <p>The '<tt>alloca</tt>' instruction
4078 allocates <tt>sizeof(<type>)*NumElements</tt> bytes of memory on the
4079 runtime stack, returning a pointer of the appropriate type to the program.
4080 If "NumElements" is specified, it is the number of elements allocated,
4081 otherwise "NumElements" is defaulted to be one. If a constant alignment is
4082 specified, the value result of the allocation is guaranteed to be aligned to
4083 at least that boundary. If not specified, or if zero, the target can choose
4084 to align the allocation on any convenient boundary compatible with the
4087 <p>'<tt>type</tt>' may be any sized type.</p>
4090 <p>Memory is allocated; a pointer is returned. The operation is undefined if
4091 there is insufficient stack space for the allocation. '<tt>alloca</tt>'d
4092 memory is automatically released when the function returns. The
4093 '<tt>alloca</tt>' instruction is commonly used to represent automatic
4094 variables that must have an address available. When the function returns
4095 (either with the <tt><a href="#i_ret">ret</a></tt>
4096 or <tt><a href="#i_unwind">unwind</a></tt> instructions), the memory is
4097 reclaimed. Allocating zero bytes is legal, but the result is undefined.</p>
4101 %ptr = alloca i32 <i>; yields {i32*}:ptr</i>
4102 %ptr = alloca i32, i32 4 <i>; yields {i32*}:ptr</i>
4103 %ptr = alloca i32, i32 4, align 1024 <i>; yields {i32*}:ptr</i>
4104 %ptr = alloca i32, align 1024 <i>; yields {i32*}:ptr</i>
4109 <!-- _______________________________________________________________________ -->
4110 <div class="doc_subsubsection"> <a name="i_load">'<tt>load</tt>'
4111 Instruction</a> </div>
4113 <div class="doc_text">
4117 <result> = load <ty>* <pointer>[, align <alignment>][, !nontemporal !<index>]
4118 <result> = volatile load <ty>* <pointer>[, align <alignment>][, !nontemporal !<index>]
4119 !<index> = !{ i32 1 }
4123 <p>The '<tt>load</tt>' instruction is used to read from memory.</p>
4126 <p>The argument to the '<tt>load</tt>' instruction specifies the memory address
4127 from which to load. The pointer must point to
4128 a <a href="#t_firstclass">first class</a> type. If the <tt>load</tt> is
4129 marked as <tt>volatile</tt>, then the optimizer is not allowed to modify the
4130 number or order of execution of this <tt>load</tt> with other
4131 volatile <tt>load</tt> and <tt><a href="#i_store">store</a></tt>
4134 <p>The optional constant <tt>align</tt> argument specifies the alignment of the
4135 operation (that is, the alignment of the memory address). A value of 0 or an
4136 omitted <tt>align</tt> argument means that the operation has the preferential
4137 alignment for the target. It is the responsibility of the code emitter to
4138 ensure that the alignment information is correct. Overestimating the
4139 alignment results in undefined behavior. Underestimating the alignment may
4140 produce less efficient code. An alignment of 1 is always safe.</p>
4142 <p>The optional <tt>!nontemporal</tt> metadata must reference a single
4143 metatadata name <index> corresponding to a metadata node with
4144 one <tt>i32</tt> entry of value 1. The existence of
4145 the <tt>!nontemporal</tt> metatadata on the instruction tells the optimizer
4146 and code generator that this load is not expected to be reused in the cache.
4147 The code generator may select special instructions to save cache bandwidth,
4148 such as the <tt>MOVNT</tt> instruction on x86.</p>
4151 <p>The location of memory pointed to is loaded. If the value being loaded is of
4152 scalar type then the number of bytes read does not exceed the minimum number
4153 of bytes needed to hold all bits of the type. For example, loading an
4154 <tt>i24</tt> reads at most three bytes. When loading a value of a type like
4155 <tt>i20</tt> with a size that is not an integral number of bytes, the result
4156 is undefined if the value was not originally written using a store of the
4161 %ptr = <a href="#i_alloca">alloca</a> i32 <i>; yields {i32*}:ptr</i>
4162 <a href="#i_store">store</a> i32 3, i32* %ptr <i>; yields {void}</i>
4163 %val = load i32* %ptr <i>; yields {i32}:val = i32 3</i>
4168 <!-- _______________________________________________________________________ -->
4169 <div class="doc_subsubsection"> <a name="i_store">'<tt>store</tt>'
4170 Instruction</a> </div>
4172 <div class="doc_text">
4176 store <ty> <value>, <ty>* <pointer>[, align <alignment>][, !nontemporal !<index>] <i>; yields {void}</i>
4177 volatile store <ty> <value>, <ty>* <pointer>[, align <alignment>][, !nontemporal !<index>] <i>; yields {void}</i>
4181 <p>The '<tt>store</tt>' instruction is used to write to memory.</p>
4184 <p>There are two arguments to the '<tt>store</tt>' instruction: a value to store
4185 and an address at which to store it. The type of the
4186 '<tt><pointer></tt>' operand must be a pointer to
4187 the <a href="#t_firstclass">first class</a> type of the
4188 '<tt><value></tt>' operand. If the <tt>store</tt> is marked
4189 as <tt>volatile</tt>, then the optimizer is not allowed to modify the number
4190 or order of execution of this <tt>store</tt> with other
4191 volatile <tt>load</tt> and <tt><a href="#i_store">store</a></tt>
4194 <p>The optional constant "align" argument specifies the alignment of the
4195 operation (that is, the alignment of the memory address). A value of 0 or an
4196 omitted "align" argument means that the operation has the preferential
4197 alignment for the target. It is the responsibility of the code emitter to
4198 ensure that the alignment information is correct. Overestimating the
4199 alignment results in an undefined behavior. Underestimating the alignment may
4200 produce less efficient code. An alignment of 1 is always safe.</p>
4202 <p>The optional !nontemporal metadata must reference a single metatadata
4203 name <index> corresponding to a metadata node with one i32 entry of
4204 value 1. The existence of the !nontemporal metatadata on the
4205 instruction tells the optimizer and code generator that this load is
4206 not expected to be reused in the cache. The code generator may
4207 select special instructions to save cache bandwidth, such as the
4208 MOVNT instruction on x86.</p>
4212 <p>The contents of memory are updated to contain '<tt><value></tt>' at the
4213 location specified by the '<tt><pointer></tt>' operand. If
4214 '<tt><value></tt>' is of scalar type then the number of bytes written
4215 does not exceed the minimum number of bytes needed to hold all bits of the
4216 type. For example, storing an <tt>i24</tt> writes at most three bytes. When
4217 writing a value of a type like <tt>i20</tt> with a size that is not an
4218 integral number of bytes, it is unspecified what happens to the extra bits
4219 that do not belong to the type, but they will typically be overwritten.</p>
4223 %ptr = <a href="#i_alloca">alloca</a> i32 <i>; yields {i32*}:ptr</i>
4224 store i32 3, i32* %ptr <i>; yields {void}</i>
4225 %val = <a href="#i_load">load</a> i32* %ptr <i>; yields {i32}:val = i32 3</i>
4230 <!-- _______________________________________________________________________ -->
4231 <div class="doc_subsubsection">
4232 <a name="i_getelementptr">'<tt>getelementptr</tt>' Instruction</a>
4235 <div class="doc_text">
4239 <result> = getelementptr <pty>* <ptrval>{, <ty> <idx>}*
4240 <result> = getelementptr inbounds <pty>* <ptrval>{, <ty> <idx>}*
4244 <p>The '<tt>getelementptr</tt>' instruction is used to get the address of a
4245 subelement of an <a href="#t_aggregate">aggregate</a> data structure.
4246 It performs address calculation only and does not access memory.</p>
4249 <p>The first argument is always a pointer, and forms the basis of the
4250 calculation. The remaining arguments are indices that indicate which of the
4251 elements of the aggregate object are indexed. The interpretation of each
4252 index is dependent on the type being indexed into. The first index always
4253 indexes the pointer value given as the first argument, the second index
4254 indexes a value of the type pointed to (not necessarily the value directly
4255 pointed to, since the first index can be non-zero), etc. The first type
4256 indexed into must be a pointer value, subsequent types can be arrays,
4257 vectors, structs and unions. Note that subsequent types being indexed into
4258 can never be pointers, since that would require loading the pointer before
4259 continuing calculation.</p>
4261 <p>The type of each index argument depends on the type it is indexing into.
4262 When indexing into a (optionally packed) structure or union, only <tt>i32</tt>
4263 integer <b>constants</b> are allowed. When indexing into an array, pointer
4264 or vector, integers of any width are allowed, and they are not required to be
4267 <p>For example, let's consider a C code fragment and how it gets compiled to
4270 <div class="doc_code">
4283 int *foo(struct ST *s) {
4284 return &s[1].Z.B[5][13];
4289 <p>The LLVM code generated by the GCC frontend is:</p>
4291 <div class="doc_code">
4293 %RT = <a href="#namedtypes">type</a> { i8 , [10 x [20 x i32]], i8 }
4294 %ST = <a href="#namedtypes">type</a> { i32, double, %RT }
4296 define i32* @foo(%ST* %s) {
4298 %reg = getelementptr %ST* %s, i32 1, i32 2, i32 1, i32 5, i32 13
4305 <p>In the example above, the first index is indexing into the '<tt>%ST*</tt>'
4306 type, which is a pointer, yielding a '<tt>%ST</tt>' = '<tt>{ i32, double, %RT
4307 }</tt>' type, a structure. The second index indexes into the third element
4308 of the structure, yielding a '<tt>%RT</tt>' = '<tt>{ i8 , [10 x [20 x i32]],
4309 i8 }</tt>' type, another structure. The third index indexes into the second
4310 element of the structure, yielding a '<tt>[10 x [20 x i32]]</tt>' type, an
4311 array. The two dimensions of the array are subscripted into, yielding an
4312 '<tt>i32</tt>' type. The '<tt>getelementptr</tt>' instruction returns a
4313 pointer to this element, thus computing a value of '<tt>i32*</tt>' type.</p>
4315 <p>Note that it is perfectly legal to index partially through a structure,
4316 returning a pointer to an inner element. Because of this, the LLVM code for
4317 the given testcase is equivalent to:</p>
4320 define i32* @foo(%ST* %s) {
4321 %t1 = getelementptr %ST* %s, i32 1 <i>; yields %ST*:%t1</i>
4322 %t2 = getelementptr %ST* %t1, i32 0, i32 2 <i>; yields %RT*:%t2</i>
4323 %t3 = getelementptr %RT* %t2, i32 0, i32 1 <i>; yields [10 x [20 x i32]]*:%t3</i>
4324 %t4 = getelementptr [10 x [20 x i32]]* %t3, i32 0, i32 5 <i>; yields [20 x i32]*:%t4</i>
4325 %t5 = getelementptr [20 x i32]* %t4, i32 0, i32 13 <i>; yields i32*:%t5</i>
4330 <p>If the <tt>inbounds</tt> keyword is present, the result value of the
4331 <tt>getelementptr</tt> is undefined if the base pointer is not an
4332 <i>in bounds</i> address of an allocated object, or if any of the addresses
4333 that would be formed by successive addition of the offsets implied by the
4334 indices to the base address with infinitely precise arithmetic are not an
4335 <i>in bounds</i> address of that allocated object.
4336 The <i>in bounds</i> addresses for an allocated object are all the addresses
4337 that point into the object, plus the address one byte past the end.</p>
4339 <p>If the <tt>inbounds</tt> keyword is not present, the offsets are added to
4340 the base address with silently-wrapping two's complement arithmetic, and
4341 the result value of the <tt>getelementptr</tt> may be outside the object
4342 pointed to by the base pointer. The result value may not necessarily be
4343 used to access memory though, even if it happens to point into allocated
4344 storage. See the <a href="#pointeraliasing">Pointer Aliasing Rules</a>
4345 section for more information.</p>
4347 <p>The getelementptr instruction is often confusing. For some more insight into
4348 how it works, see <a href="GetElementPtr.html">the getelementptr FAQ</a>.</p>
4352 <i>; yields [12 x i8]*:aptr</i>
4353 %aptr = getelementptr {i32, [12 x i8]}* %saptr, i64 0, i32 1
4354 <i>; yields i8*:vptr</i>
4355 %vptr = getelementptr {i32, <2 x i8>}* %svptr, i64 0, i32 1, i32 1
4356 <i>; yields i8*:eptr</i>
4357 %eptr = getelementptr [12 x i8]* %aptr, i64 0, i32 1
4358 <i>; yields i32*:iptr</i>
4359 %iptr = getelementptr [10 x i32]* @arr, i16 0, i16 0
4364 <!-- ======================================================================= -->
4365 <div class="doc_subsection"> <a name="convertops">Conversion Operations</a>
4368 <div class="doc_text">
4370 <p>The instructions in this category are the conversion instructions (casting)
4371 which all take a single operand and a type. They perform various bit
4372 conversions on the operand.</p>
4376 <!-- _______________________________________________________________________ -->
4377 <div class="doc_subsubsection">
4378 <a name="i_trunc">'<tt>trunc .. to</tt>' Instruction</a>
4380 <div class="doc_text">
4384 <result> = trunc <ty> <value> to <ty2> <i>; yields ty2</i>
4388 <p>The '<tt>trunc</tt>' instruction truncates its operand to the
4389 type <tt>ty2</tt>.</p>
4392 <p>The '<tt>trunc</tt>' instruction takes a <tt>value</tt> to trunc, which must
4393 be an <a href="#t_integer">integer</a> type, and a type that specifies the
4394 size and type of the result, which must be
4395 an <a href="#t_integer">integer</a> type. The bit size of <tt>value</tt> must
4396 be larger than the bit size of <tt>ty2</tt>. Equal sized types are not
4400 <p>The '<tt>trunc</tt>' instruction truncates the high order bits
4401 in <tt>value</tt> and converts the remaining bits to <tt>ty2</tt>. Since the
4402 source size must be larger than the destination size, <tt>trunc</tt> cannot
4403 be a <i>no-op cast</i>. It will always truncate bits.</p>
4407 %X = trunc i32 257 to i8 <i>; yields i8:1</i>
4408 %Y = trunc i32 123 to i1 <i>; yields i1:true</i>
4409 %Z = trunc i32 122 to i1 <i>; yields i1:false</i>
4414 <!-- _______________________________________________________________________ -->
4415 <div class="doc_subsubsection">
4416 <a name="i_zext">'<tt>zext .. to</tt>' Instruction</a>
4418 <div class="doc_text">
4422 <result> = zext <ty> <value> to <ty2> <i>; yields ty2</i>
4426 <p>The '<tt>zext</tt>' instruction zero extends its operand to type
4431 <p>The '<tt>zext</tt>' instruction takes a value to cast, which must be of
4432 <a href="#t_integer">integer</a> type, and a type to cast it to, which must
4433 also be of <a href="#t_integer">integer</a> type. The bit size of the
4434 <tt>value</tt> must be smaller than the bit size of the destination type,
4438 <p>The <tt>zext</tt> fills the high order bits of the <tt>value</tt> with zero
4439 bits until it reaches the size of the destination type, <tt>ty2</tt>.</p>
4441 <p>When zero extending from i1, the result will always be either 0 or 1.</p>
4445 %X = zext i32 257 to i64 <i>; yields i64:257</i>
4446 %Y = zext i1 true to i32 <i>; yields i32:1</i>
4451 <!-- _______________________________________________________________________ -->
4452 <div class="doc_subsubsection">
4453 <a name="i_sext">'<tt>sext .. to</tt>' Instruction</a>
4455 <div class="doc_text">
4459 <result> = sext <ty> <value> to <ty2> <i>; yields ty2</i>
4463 <p>The '<tt>sext</tt>' sign extends <tt>value</tt> to the type <tt>ty2</tt>.</p>
4466 <p>The '<tt>sext</tt>' instruction takes a value to cast, which must be of
4467 <a href="#t_integer">integer</a> type, and a type to cast it to, which must
4468 also be of <a href="#t_integer">integer</a> type. The bit size of the
4469 <tt>value</tt> must be smaller than the bit size of the destination type,
4473 <p>The '<tt>sext</tt>' instruction performs a sign extension by copying the sign
4474 bit (highest order bit) of the <tt>value</tt> until it reaches the bit size
4475 of the type <tt>ty2</tt>.</p>
4477 <p>When sign extending from i1, the extension always results in -1 or 0.</p>
4481 %X = sext i8 -1 to i16 <i>; yields i16 :65535</i>
4482 %Y = sext i1 true to i32 <i>; yields i32:-1</i>
4487 <!-- _______________________________________________________________________ -->
4488 <div class="doc_subsubsection">
4489 <a name="i_fptrunc">'<tt>fptrunc .. to</tt>' Instruction</a>
4492 <div class="doc_text">
4496 <result> = fptrunc <ty> <value> to <ty2> <i>; yields ty2</i>
4500 <p>The '<tt>fptrunc</tt>' instruction truncates <tt>value</tt> to type
4504 <p>The '<tt>fptrunc</tt>' instruction takes a <a href="#t_floating">floating
4505 point</a> value to cast and a <a href="#t_floating">floating point</a> type
4506 to cast it to. The size of <tt>value</tt> must be larger than the size of
4507 <tt>ty2</tt>. This implies that <tt>fptrunc</tt> cannot be used to make a
4508 <i>no-op cast</i>.</p>
4511 <p>The '<tt>fptrunc</tt>' instruction truncates a <tt>value</tt> from a larger
4512 <a href="#t_floating">floating point</a> type to a smaller
4513 <a href="#t_floating">floating point</a> type. If the value cannot fit
4514 within the destination type, <tt>ty2</tt>, then the results are
4519 %X = fptrunc double 123.0 to float <i>; yields float:123.0</i>
4520 %Y = fptrunc double 1.0E+300 to float <i>; yields undefined</i>
4525 <!-- _______________________________________________________________________ -->
4526 <div class="doc_subsubsection">
4527 <a name="i_fpext">'<tt>fpext .. to</tt>' Instruction</a>
4529 <div class="doc_text">
4533 <result> = fpext <ty> <value> to <ty2> <i>; yields ty2</i>
4537 <p>The '<tt>fpext</tt>' extends a floating point <tt>value</tt> to a larger
4538 floating point value.</p>
4541 <p>The '<tt>fpext</tt>' instruction takes a
4542 <a href="#t_floating">floating point</a> <tt>value</tt> to cast, and
4543 a <a href="#t_floating">floating point</a> type to cast it to. The source
4544 type must be smaller than the destination type.</p>
4547 <p>The '<tt>fpext</tt>' instruction extends the <tt>value</tt> from a smaller
4548 <a href="#t_floating">floating point</a> type to a larger
4549 <a href="#t_floating">floating point</a> type. The <tt>fpext</tt> cannot be
4550 used to make a <i>no-op cast</i> because it always changes bits. Use
4551 <tt>bitcast</tt> to make a <i>no-op cast</i> for a floating point cast.</p>
4555 %X = fpext float 3.1415 to double <i>; yields double:3.1415</i>
4556 %Y = fpext float 1.0 to float <i>; yields float:1.0 (no-op)</i>
4561 <!-- _______________________________________________________________________ -->
4562 <div class="doc_subsubsection">
4563 <a name="i_fptoui">'<tt>fptoui .. to</tt>' Instruction</a>
4565 <div class="doc_text">
4569 <result> = fptoui <ty> <value> to <ty2> <i>; yields ty2</i>
4573 <p>The '<tt>fptoui</tt>' converts a floating point <tt>value</tt> to its
4574 unsigned integer equivalent of type <tt>ty2</tt>.</p>
4577 <p>The '<tt>fptoui</tt>' instruction takes a value to cast, which must be a
4578 scalar or vector <a href="#t_floating">floating point</a> value, and a type
4579 to cast it to <tt>ty2</tt>, which must be an <a href="#t_integer">integer</a>
4580 type. If <tt>ty</tt> is a vector floating point type, <tt>ty2</tt> must be a
4581 vector integer type with the same number of elements as <tt>ty</tt></p>
4584 <p>The '<tt>fptoui</tt>' instruction converts its
4585 <a href="#t_floating">floating point</a> operand into the nearest (rounding
4586 towards zero) unsigned integer value. If the value cannot fit
4587 in <tt>ty2</tt>, the results are undefined.</p>
4591 %X = fptoui double 123.0 to i32 <i>; yields i32:123</i>
4592 %Y = fptoui float 1.0E+300 to i1 <i>; yields undefined:1</i>
4593 %Z = fptoui float 1.04E+17 to i8 <i>; yields undefined:1</i>
4598 <!-- _______________________________________________________________________ -->
4599 <div class="doc_subsubsection">
4600 <a name="i_fptosi">'<tt>fptosi .. to</tt>' Instruction</a>
4602 <div class="doc_text">
4606 <result> = fptosi <ty> <value> to <ty2> <i>; yields ty2</i>
4610 <p>The '<tt>fptosi</tt>' instruction converts
4611 <a href="#t_floating">floating point</a> <tt>value</tt> to
4612 type <tt>ty2</tt>.</p>
4615 <p>The '<tt>fptosi</tt>' instruction takes a value to cast, which must be a
4616 scalar or vector <a href="#t_floating">floating point</a> value, and a type
4617 to cast it to <tt>ty2</tt>, which must be an <a href="#t_integer">integer</a>
4618 type. If <tt>ty</tt> is a vector floating point type, <tt>ty2</tt> must be a
4619 vector integer type with the same number of elements as <tt>ty</tt></p>
4622 <p>The '<tt>fptosi</tt>' instruction converts its
4623 <a href="#t_floating">floating point</a> operand into the nearest (rounding
4624 towards zero) signed integer value. If the value cannot fit in <tt>ty2</tt>,
4625 the results are undefined.</p>
4629 %X = fptosi double -123.0 to i32 <i>; yields i32:-123</i>
4630 %Y = fptosi float 1.0E-247 to i1 <i>; yields undefined:1</i>
4631 %Z = fptosi float 1.04E+17 to i8 <i>; yields undefined:1</i>
4636 <!-- _______________________________________________________________________ -->
4637 <div class="doc_subsubsection">
4638 <a name="i_uitofp">'<tt>uitofp .. to</tt>' Instruction</a>
4640 <div class="doc_text">
4644 <result> = uitofp <ty> <value> to <ty2> <i>; yields ty2</i>
4648 <p>The '<tt>uitofp</tt>' instruction regards <tt>value</tt> as an unsigned
4649 integer and converts that value to the <tt>ty2</tt> type.</p>
4652 <p>The '<tt>uitofp</tt>' instruction takes a value to cast, which must be a
4653 scalar or vector <a href="#t_integer">integer</a> value, and a type to cast
4654 it to <tt>ty2</tt>, which must be an <a href="#t_floating">floating point</a>
4655 type. If <tt>ty</tt> is a vector integer type, <tt>ty2</tt> must be a vector
4656 floating point type with the same number of elements as <tt>ty</tt></p>
4659 <p>The '<tt>uitofp</tt>' instruction interprets its operand as an unsigned
4660 integer quantity and converts it to the corresponding floating point
4661 value. If the value cannot fit in the floating point value, the results are
4666 %X = uitofp i32 257 to float <i>; yields float:257.0</i>
4667 %Y = uitofp i8 -1 to double <i>; yields double:255.0</i>
4672 <!-- _______________________________________________________________________ -->
4673 <div class="doc_subsubsection">
4674 <a name="i_sitofp">'<tt>sitofp .. to</tt>' Instruction</a>
4676 <div class="doc_text">
4680 <result> = sitofp <ty> <value> to <ty2> <i>; yields ty2</i>
4684 <p>The '<tt>sitofp</tt>' instruction regards <tt>value</tt> as a signed integer
4685 and converts that value to the <tt>ty2</tt> type.</p>
4688 <p>The '<tt>sitofp</tt>' instruction takes a value to cast, which must be a
4689 scalar or vector <a href="#t_integer">integer</a> value, and a type to cast
4690 it to <tt>ty2</tt>, which must be an <a href="#t_floating">floating point</a>
4691 type. If <tt>ty</tt> is a vector integer type, <tt>ty2</tt> must be a vector
4692 floating point type with the same number of elements as <tt>ty</tt></p>
4695 <p>The '<tt>sitofp</tt>' instruction interprets its operand as a signed integer
4696 quantity and converts it to the corresponding floating point value. If the
4697 value cannot fit in the floating point value, the results are undefined.</p>
4701 %X = sitofp i32 257 to float <i>; yields float:257.0</i>
4702 %Y = sitofp i8 -1 to double <i>; yields double:-1.0</i>
4707 <!-- _______________________________________________________________________ -->
4708 <div class="doc_subsubsection">
4709 <a name="i_ptrtoint">'<tt>ptrtoint .. to</tt>' Instruction</a>
4711 <div class="doc_text">
4715 <result> = ptrtoint <ty> <value> to <ty2> <i>; yields ty2</i>
4719 <p>The '<tt>ptrtoint</tt>' instruction converts the pointer <tt>value</tt> to
4720 the integer type <tt>ty2</tt>.</p>
4723 <p>The '<tt>ptrtoint</tt>' instruction takes a <tt>value</tt> to cast, which
4724 must be a <a href="#t_pointer">pointer</a> value, and a type to cast it to
4725 <tt>ty2</tt>, which must be an <a href="#t_integer">integer</a> type.</p>
4728 <p>The '<tt>ptrtoint</tt>' instruction converts <tt>value</tt> to integer type
4729 <tt>ty2</tt> by interpreting the pointer value as an integer and either
4730 truncating or zero extending that value to the size of the integer type. If
4731 <tt>value</tt> is smaller than <tt>ty2</tt> then a zero extension is done. If
4732 <tt>value</tt> is larger than <tt>ty2</tt> then a truncation is done. If they
4733 are the same size, then nothing is done (<i>no-op cast</i>) other than a type
4738 %X = ptrtoint i32* %X to i8 <i>; yields truncation on 32-bit architecture</i>
4739 %Y = ptrtoint i32* %x to i64 <i>; yields zero extension on 32-bit architecture</i>
4744 <!-- _______________________________________________________________________ -->
4745 <div class="doc_subsubsection">
4746 <a name="i_inttoptr">'<tt>inttoptr .. to</tt>' Instruction</a>
4748 <div class="doc_text">
4752 <result> = inttoptr <ty> <value> to <ty2> <i>; yields ty2</i>
4756 <p>The '<tt>inttoptr</tt>' instruction converts an integer <tt>value</tt> to a
4757 pointer type, <tt>ty2</tt>.</p>
4760 <p>The '<tt>inttoptr</tt>' instruction takes an <a href="#t_integer">integer</a>
4761 value to cast, and a type to cast it to, which must be a
4762 <a href="#t_pointer">pointer</a> type.</p>
4765 <p>The '<tt>inttoptr</tt>' instruction converts <tt>value</tt> to type
4766 <tt>ty2</tt> by applying either a zero extension or a truncation depending on
4767 the size of the integer <tt>value</tt>. If <tt>value</tt> is larger than the
4768 size of a pointer then a truncation is done. If <tt>value</tt> is smaller
4769 than the size of a pointer then a zero extension is done. If they are the
4770 same size, nothing is done (<i>no-op cast</i>).</p>
4774 %X = inttoptr i32 255 to i32* <i>; yields zero extension on 64-bit architecture</i>
4775 %Y = inttoptr i32 255 to i32* <i>; yields no-op on 32-bit architecture</i>
4776 %Z = inttoptr i64 0 to i32* <i>; yields truncation on 32-bit architecture</i>
4781 <!-- _______________________________________________________________________ -->
4782 <div class="doc_subsubsection">
4783 <a name="i_bitcast">'<tt>bitcast .. to</tt>' Instruction</a>
4785 <div class="doc_text">
4789 <result> = bitcast <ty> <value> to <ty2> <i>; yields ty2</i>
4793 <p>The '<tt>bitcast</tt>' instruction converts <tt>value</tt> to type
4794 <tt>ty2</tt> without changing any bits.</p>
4797 <p>The '<tt>bitcast</tt>' instruction takes a value to cast, which must be a
4798 non-aggregate first class value, and a type to cast it to, which must also be
4799 a non-aggregate <a href="#t_firstclass">first class</a> type. The bit sizes
4800 of <tt>value</tt> and the destination type, <tt>ty2</tt>, must be
4801 identical. If the source type is a pointer, the destination type must also be
4802 a pointer. This instruction supports bitwise conversion of vectors to
4803 integers and to vectors of other types (as long as they have the same
4807 <p>The '<tt>bitcast</tt>' instruction converts <tt>value</tt> to type
4808 <tt>ty2</tt>. It is always a <i>no-op cast</i> because no bits change with
4809 this conversion. The conversion is done as if the <tt>value</tt> had been
4810 stored to memory and read back as type <tt>ty2</tt>. Pointer types may only
4811 be converted to other pointer types with this instruction. To convert
4812 pointers to other types, use the <a href="#i_inttoptr">inttoptr</a> or
4813 <a href="#i_ptrtoint">ptrtoint</a> instructions first.</p>
4817 %X = bitcast i8 255 to i8 <i>; yields i8 :-1</i>
4818 %Y = bitcast i32* %x to sint* <i>; yields sint*:%x</i>
4819 %Z = bitcast <2 x int> %V to i64; <i>; yields i64: %V</i>
4824 <!-- ======================================================================= -->
4825 <div class="doc_subsection"> <a name="otherops">Other Operations</a> </div>
4827 <div class="doc_text">
4829 <p>The instructions in this category are the "miscellaneous" instructions, which
4830 defy better classification.</p>
4834 <!-- _______________________________________________________________________ -->
4835 <div class="doc_subsubsection"><a name="i_icmp">'<tt>icmp</tt>' Instruction</a>
4838 <div class="doc_text">
4842 <result> = icmp <cond> <ty> <op1>, <op2> <i>; yields {i1} or {<N x i1>}:result</i>
4846 <p>The '<tt>icmp</tt>' instruction returns a boolean value or a vector of
4847 boolean values based on comparison of its two integer, integer vector, or
4848 pointer operands.</p>
4851 <p>The '<tt>icmp</tt>' instruction takes three operands. The first operand is
4852 the condition code indicating the kind of comparison to perform. It is not a
4853 value, just a keyword. The possible condition code are:</p>
4856 <li><tt>eq</tt>: equal</li>
4857 <li><tt>ne</tt>: not equal </li>
4858 <li><tt>ugt</tt>: unsigned greater than</li>
4859 <li><tt>uge</tt>: unsigned greater or equal</li>
4860 <li><tt>ult</tt>: unsigned less than</li>
4861 <li><tt>ule</tt>: unsigned less or equal</li>
4862 <li><tt>sgt</tt>: signed greater than</li>
4863 <li><tt>sge</tt>: signed greater or equal</li>
4864 <li><tt>slt</tt>: signed less than</li>
4865 <li><tt>sle</tt>: signed less or equal</li>
4868 <p>The remaining two arguments must be <a href="#t_integer">integer</a> or
4869 <a href="#t_pointer">pointer</a> or integer <a href="#t_vector">vector</a>
4870 typed. They must also be identical types.</p>
4873 <p>The '<tt>icmp</tt>' compares <tt>op1</tt> and <tt>op2</tt> according to the
4874 condition code given as <tt>cond</tt>. The comparison performed always yields
4875 either an <a href="#t_integer"><tt>i1</tt></a> or vector of <tt>i1</tt>
4876 result, as follows:</p>
4879 <li><tt>eq</tt>: yields <tt>true</tt> if the operands are equal,
4880 <tt>false</tt> otherwise. No sign interpretation is necessary or
4883 <li><tt>ne</tt>: yields <tt>true</tt> if the operands are unequal,
4884 <tt>false</tt> otherwise. No sign interpretation is necessary or
4887 <li><tt>ugt</tt>: interprets the operands as unsigned values and yields
4888 <tt>true</tt> if <tt>op1</tt> is greater than <tt>op2</tt>.</li>
4890 <li><tt>uge</tt>: interprets the operands as unsigned values and yields
4891 <tt>true</tt> if <tt>op1</tt> is greater than or equal
4892 to <tt>op2</tt>.</li>
4894 <li><tt>ult</tt>: interprets the operands as unsigned values and yields
4895 <tt>true</tt> if <tt>op1</tt> is less than <tt>op2</tt>.</li>
4897 <li><tt>ule</tt>: interprets the operands as unsigned values and yields
4898 <tt>true</tt> if <tt>op1</tt> is less than or equal to <tt>op2</tt>.</li>
4900 <li><tt>sgt</tt>: interprets the operands as signed values and yields
4901 <tt>true</tt> if <tt>op1</tt> is greater than <tt>op2</tt>.</li>
4903 <li><tt>sge</tt>: interprets the operands as signed values and yields
4904 <tt>true</tt> if <tt>op1</tt> is greater than or equal
4905 to <tt>op2</tt>.</li>
4907 <li><tt>slt</tt>: interprets the operands as signed values and yields
4908 <tt>true</tt> if <tt>op1</tt> is less than <tt>op2</tt>.</li>
4910 <li><tt>sle</tt>: interprets the operands as signed values and yields
4911 <tt>true</tt> if <tt>op1</tt> is less than or equal to <tt>op2</tt>.</li>
4914 <p>If the operands are <a href="#t_pointer">pointer</a> typed, the pointer
4915 values are compared as if they were integers.</p>
4917 <p>If the operands are integer vectors, then they are compared element by
4918 element. The result is an <tt>i1</tt> vector with the same number of elements
4919 as the values being compared. Otherwise, the result is an <tt>i1</tt>.</p>
4923 <result> = icmp eq i32 4, 5 <i>; yields: result=false</i>
4924 <result> = icmp ne float* %X, %X <i>; yields: result=false</i>
4925 <result> = icmp ult i16 4, 5 <i>; yields: result=true</i>
4926 <result> = icmp sgt i16 4, 5 <i>; yields: result=false</i>
4927 <result> = icmp ule i16 -4, 5 <i>; yields: result=false</i>
4928 <result> = icmp sge i16 4, 5 <i>; yields: result=false</i>
4931 <p>Note that the code generator does not yet support vector types with
4932 the <tt>icmp</tt> instruction.</p>
4936 <!-- _______________________________________________________________________ -->
4937 <div class="doc_subsubsection"><a name="i_fcmp">'<tt>fcmp</tt>' Instruction</a>
4940 <div class="doc_text">
4944 <result> = fcmp <cond> <ty> <op1>, <op2> <i>; yields {i1} or {<N x i1>}:result</i>
4948 <p>The '<tt>fcmp</tt>' instruction returns a boolean value or vector of boolean
4949 values based on comparison of its operands.</p>
4951 <p>If the operands are floating point scalars, then the result type is a boolean
4952 (<a href="#t_integer"><tt>i1</tt></a>).</p>
4954 <p>If the operands are floating point vectors, then the result type is a vector
4955 of boolean with the same number of elements as the operands being
4959 <p>The '<tt>fcmp</tt>' instruction takes three operands. The first operand is
4960 the condition code indicating the kind of comparison to perform. It is not a
4961 value, just a keyword. The possible condition code are:</p>
4964 <li><tt>false</tt>: no comparison, always returns false</li>
4965 <li><tt>oeq</tt>: ordered and equal</li>
4966 <li><tt>ogt</tt>: ordered and greater than </li>
4967 <li><tt>oge</tt>: ordered and greater than or equal</li>
4968 <li><tt>olt</tt>: ordered and less than </li>
4969 <li><tt>ole</tt>: ordered and less than or equal</li>
4970 <li><tt>one</tt>: ordered and not equal</li>
4971 <li><tt>ord</tt>: ordered (no nans)</li>
4972 <li><tt>ueq</tt>: unordered or equal</li>
4973 <li><tt>ugt</tt>: unordered or greater than </li>
4974 <li><tt>uge</tt>: unordered or greater than or equal</li>
4975 <li><tt>ult</tt>: unordered or less than </li>
4976 <li><tt>ule</tt>: unordered or less than or equal</li>
4977 <li><tt>une</tt>: unordered or not equal</li>
4978 <li><tt>uno</tt>: unordered (either nans)</li>
4979 <li><tt>true</tt>: no comparison, always returns true</li>
4982 <p><i>Ordered</i> means that neither operand is a QNAN while
4983 <i>unordered</i> means that either operand may be a QNAN.</p>
4985 <p>Each of <tt>val1</tt> and <tt>val2</tt> arguments must be either
4986 a <a href="#t_floating">floating point</a> type or
4987 a <a href="#t_vector">vector</a> of floating point type. They must have
4988 identical types.</p>
4991 <p>The '<tt>fcmp</tt>' instruction compares <tt>op1</tt> and <tt>op2</tt>
4992 according to the condition code given as <tt>cond</tt>. If the operands are
4993 vectors, then the vectors are compared element by element. Each comparison
4994 performed always yields an <a href="#t_integer">i1</a> result, as
4998 <li><tt>false</tt>: always yields <tt>false</tt>, regardless of operands.</li>
5000 <li><tt>oeq</tt>: yields <tt>true</tt> if both operands are not a QNAN and
5001 <tt>op1</tt> is equal to <tt>op2</tt>.</li>
5003 <li><tt>ogt</tt>: yields <tt>true</tt> if both operands are not a QNAN and
5004 <tt>op1</tt> is greater than <tt>op2</tt>.</li>
5006 <li><tt>oge</tt>: yields <tt>true</tt> if both operands are not a QNAN and
5007 <tt>op1</tt> is greater than or equal to <tt>op2</tt>.</li>
5009 <li><tt>olt</tt>: yields <tt>true</tt> if both operands are not a QNAN and
5010 <tt>op1</tt> is less than <tt>op2</tt>.</li>
5012 <li><tt>ole</tt>: yields <tt>true</tt> if both operands are not a QNAN and
5013 <tt>op1</tt> is less than or equal to <tt>op2</tt>.</li>
5015 <li><tt>one</tt>: yields <tt>true</tt> if both operands are not a QNAN and
5016 <tt>op1</tt> is not equal to <tt>op2</tt>.</li>
5018 <li><tt>ord</tt>: yields <tt>true</tt> if both operands are not a QNAN.</li>
5020 <li><tt>ueq</tt>: yields <tt>true</tt> if either operand is a QNAN or
5021 <tt>op1</tt> is equal to <tt>op2</tt>.</li>
5023 <li><tt>ugt</tt>: yields <tt>true</tt> if either operand is a QNAN or
5024 <tt>op1</tt> is greater than <tt>op2</tt>.</li>
5026 <li><tt>uge</tt>: yields <tt>true</tt> if either operand is a QNAN or
5027 <tt>op1</tt> is greater than or equal to <tt>op2</tt>.</li>
5029 <li><tt>ult</tt>: yields <tt>true</tt> if either operand is a QNAN or
5030 <tt>op1</tt> is less than <tt>op2</tt>.</li>
5032 <li><tt>ule</tt>: yields <tt>true</tt> if either operand is a QNAN or
5033 <tt>op1</tt> is less than or equal to <tt>op2</tt>.</li>
5035 <li><tt>une</tt>: yields <tt>true</tt> if either operand is a QNAN or
5036 <tt>op1</tt> is not equal to <tt>op2</tt>.</li>
5038 <li><tt>uno</tt>: yields <tt>true</tt> if either operand is a QNAN.</li>
5040 <li><tt>true</tt>: always yields <tt>true</tt>, regardless of operands.</li>
5045 <result> = fcmp oeq float 4.0, 5.0 <i>; yields: result=false</i>
5046 <result> = fcmp one float 4.0, 5.0 <i>; yields: result=true</i>
5047 <result> = fcmp olt float 4.0, 5.0 <i>; yields: result=true</i>
5048 <result> = fcmp ueq double 1.0, 2.0 <i>; yields: result=false</i>
5051 <p>Note that the code generator does not yet support vector types with
5052 the <tt>fcmp</tt> instruction.</p>
5056 <!-- _______________________________________________________________________ -->
5057 <div class="doc_subsubsection">
5058 <a name="i_phi">'<tt>phi</tt>' Instruction</a>
5061 <div class="doc_text">
5065 <result> = phi <ty> [ <val0>, <label0>], ...
5069 <p>The '<tt>phi</tt>' instruction is used to implement the φ node in the
5070 SSA graph representing the function.</p>
5073 <p>The type of the incoming values is specified with the first type field. After
5074 this, the '<tt>phi</tt>' instruction takes a list of pairs as arguments, with
5075 one pair for each predecessor basic block of the current block. Only values
5076 of <a href="#t_firstclass">first class</a> type may be used as the value
5077 arguments to the PHI node. Only labels may be used as the label
5080 <p>There must be no non-phi instructions between the start of a basic block and
5081 the PHI instructions: i.e. PHI instructions must be first in a basic
5084 <p>For the purposes of the SSA form, the use of each incoming value is deemed to
5085 occur on the edge from the corresponding predecessor block to the current
5086 block (but after any definition of an '<tt>invoke</tt>' instruction's return
5087 value on the same edge).</p>
5090 <p>At runtime, the '<tt>phi</tt>' instruction logically takes on the value
5091 specified by the pair corresponding to the predecessor basic block that
5092 executed just prior to the current block.</p>
5096 Loop: ; Infinite loop that counts from 0 on up...
5097 %indvar = phi i32 [ 0, %LoopHeader ], [ %nextindvar, %Loop ]
5098 %nextindvar = add i32 %indvar, 1
5104 <!-- _______________________________________________________________________ -->
5105 <div class="doc_subsubsection">
5106 <a name="i_select">'<tt>select</tt>' Instruction</a>
5109 <div class="doc_text">
5113 <result> = select <i>selty</i> <cond>, <ty> <val1>, <ty> <val2> <i>; yields ty</i>
5115 <i>selty</i> is either i1 or {<N x i1>}
5119 <p>The '<tt>select</tt>' instruction is used to choose one value based on a
5120 condition, without branching.</p>
5124 <p>The '<tt>select</tt>' instruction requires an 'i1' value or a vector of 'i1'
5125 values indicating the condition, and two values of the
5126 same <a href="#t_firstclass">first class</a> type. If the val1/val2 are
5127 vectors and the condition is a scalar, then entire vectors are selected, not
5128 individual elements.</p>
5131 <p>If the condition is an i1 and it evaluates to 1, the instruction returns the
5132 first value argument; otherwise, it returns the second value argument.</p>
5134 <p>If the condition is a vector of i1, then the value arguments must be vectors
5135 of the same size, and the selection is done element by element.</p>
5139 %X = select i1 true, i8 17, i8 42 <i>; yields i8:17</i>
5142 <p>Note that the code generator does not yet support conditions
5143 with vector type.</p>
5147 <!-- _______________________________________________________________________ -->
5148 <div class="doc_subsubsection">
5149 <a name="i_call">'<tt>call</tt>' Instruction</a>
5152 <div class="doc_text">
5156 <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>]
5160 <p>The '<tt>call</tt>' instruction represents a simple function call.</p>
5163 <p>This instruction requires several arguments:</p>
5166 <li>The optional "tail" marker indicates that the callee function does not
5167 access any allocas or varargs in the caller. Note that calls may be
5168 marked "tail" even if they do not occur before
5169 a <a href="#i_ret"><tt>ret</tt></a> instruction. If the "tail" marker is
5170 present, the function call is eligible for tail call optimization,
5171 but <a href="CodeGenerator.html#tailcallopt">might not in fact be
5172 optimized into a jump</a>. The code generator may optimize calls marked
5173 "tail" with either 1) automatic <a href="CodeGenerator.html#sibcallopt">
5174 sibling call optimization</a> when the caller and callee have
5175 matching signatures, or 2) forced tail call optimization when the
5176 following extra requirements are met:
5178 <li>Caller and callee both have the calling
5179 convention <tt>fastcc</tt>.</li>
5180 <li>The call is in tail position (ret immediately follows call and ret
5181 uses value of call or is void).</li>
5182 <li>Option <tt>-tailcallopt</tt> is enabled,
5183 or <code>llvm::GuaranteedTailCallOpt</code> is <code>true</code>.</li>
5184 <li><a href="CodeGenerator.html#tailcallopt">Platform specific
5185 constraints are met.</a></li>
5189 <li>The optional "cconv" marker indicates which <a href="#callingconv">calling
5190 convention</a> the call should use. If none is specified, the call
5191 defaults to using C calling conventions. The calling convention of the
5192 call must match the calling convention of the target function, or else the
5193 behavior is undefined.</li>
5195 <li>The optional <a href="#paramattrs">Parameter Attributes</a> list for
5196 return values. Only '<tt>zeroext</tt>', '<tt>signext</tt>', and
5197 '<tt>inreg</tt>' attributes are valid here.</li>
5199 <li>'<tt>ty</tt>': the type of the call instruction itself which is also the
5200 type of the return value. Functions that return no value are marked
5201 <tt><a href="#t_void">void</a></tt>.</li>
5203 <li>'<tt>fnty</tt>': shall be the signature of the pointer to function value
5204 being invoked. The argument types must match the types implied by this
5205 signature. This type can be omitted if the function is not varargs and if
5206 the function type does not return a pointer to a function.</li>
5208 <li>'<tt>fnptrval</tt>': An LLVM value containing a pointer to a function to
5209 be invoked. In most cases, this is a direct function invocation, but
5210 indirect <tt>call</tt>s are just as possible, calling an arbitrary pointer
5211 to function value.</li>
5213 <li>'<tt>function args</tt>': argument list whose types match the function
5214 signature argument types and parameter attributes. All arguments must be
5215 of <a href="#t_firstclass">first class</a> type. If the function
5216 signature indicates the function accepts a variable number of arguments,
5217 the extra arguments can be specified.</li>
5219 <li>The optional <a href="#fnattrs">function attributes</a> list. Only
5220 '<tt>noreturn</tt>', '<tt>nounwind</tt>', '<tt>readonly</tt>' and
5221 '<tt>readnone</tt>' attributes are valid here.</li>
5225 <p>The '<tt>call</tt>' instruction is used to cause control flow to transfer to
5226 a specified function, with its incoming arguments bound to the specified
5227 values. Upon a '<tt><a href="#i_ret">ret</a></tt>' instruction in the called
5228 function, control flow continues with the instruction after the function
5229 call, and the return value of the function is bound to the result
5234 %retval = call i32 @test(i32 %argc)
5235 call i32 (i8 *, ...)* @printf(i8 * %msg, i32 12, i8 42) <i>; yields i32</i>
5236 %X = tail call i32 @foo() <i>; yields i32</i>
5237 %Y = tail call <a href="#callingconv">fastcc</a> i32 @foo() <i>; yields i32</i>
5238 call void %foo(i8 97 signext)
5240 %struct.A = type { i32, i8 }
5241 %r = call %struct.A @foo() <i>; yields { 32, i8 }</i>
5242 %gr = extractvalue %struct.A %r, 0 <i>; yields i32</i>
5243 %gr1 = extractvalue %struct.A %r, 1 <i>; yields i8</i>
5244 %Z = call void @foo() noreturn <i>; indicates that %foo never returns normally</i>
5245 %ZZ = call zeroext i32 @bar() <i>; Return value is %zero extended</i>
5248 <p>llvm treats calls to some functions with names and arguments that match the
5249 standard C99 library as being the C99 library functions, and may perform
5250 optimizations or generate code for them under that assumption. This is
5251 something we'd like to change in the future to provide better support for
5252 freestanding environments and non-C-based languages.</p>
5256 <!-- _______________________________________________________________________ -->
5257 <div class="doc_subsubsection">
5258 <a name="i_va_arg">'<tt>va_arg</tt>' Instruction</a>
5261 <div class="doc_text">
5265 <resultval> = va_arg <va_list*> <arglist>, <argty>
5269 <p>The '<tt>va_arg</tt>' instruction is used to access arguments passed through
5270 the "variable argument" area of a function call. It is used to implement the
5271 <tt>va_arg</tt> macro in C.</p>
5274 <p>This instruction takes a <tt>va_list*</tt> value and the type of the
5275 argument. It returns a value of the specified argument type and increments
5276 the <tt>va_list</tt> to point to the next argument. The actual type
5277 of <tt>va_list</tt> is target specific.</p>
5280 <p>The '<tt>va_arg</tt>' instruction loads an argument of the specified type
5281 from the specified <tt>va_list</tt> and causes the <tt>va_list</tt> to point
5282 to the next argument. For more information, see the variable argument
5283 handling <a href="#int_varargs">Intrinsic Functions</a>.</p>
5285 <p>It is legal for this instruction to be called in a function which does not
5286 take a variable number of arguments, for example, the <tt>vfprintf</tt>
5289 <p><tt>va_arg</tt> is an LLVM instruction instead of
5290 an <a href="#intrinsics">intrinsic function</a> because it takes a type as an
5294 <p>See the <a href="#int_varargs">variable argument processing</a> section.</p>
5296 <p>Note that the code generator does not yet fully support va_arg on many
5297 targets. Also, it does not currently support va_arg with aggregate types on
5302 <!-- *********************************************************************** -->
5303 <div class="doc_section"> <a name="intrinsics">Intrinsic Functions</a> </div>
5304 <!-- *********************************************************************** -->
5306 <div class="doc_text">
5308 <p>LLVM supports the notion of an "intrinsic function". These functions have
5309 well known names and semantics and are required to follow certain
5310 restrictions. Overall, these intrinsics represent an extension mechanism for
5311 the LLVM language that does not require changing all of the transformations
5312 in LLVM when adding to the language (or the bitcode reader/writer, the
5313 parser, etc...).</p>
5315 <p>Intrinsic function names must all start with an "<tt>llvm.</tt>" prefix. This
5316 prefix is reserved in LLVM for intrinsic names; thus, function names may not
5317 begin with this prefix. Intrinsic functions must always be external
5318 functions: you cannot define the body of intrinsic functions. Intrinsic
5319 functions may only be used in call or invoke instructions: it is illegal to
5320 take the address of an intrinsic function. Additionally, because intrinsic
5321 functions are part of the LLVM language, it is required if any are added that
5322 they be documented here.</p>
5324 <p>Some intrinsic functions can be overloaded, i.e., the intrinsic represents a
5325 family of functions that perform the same operation but on different data
5326 types. Because LLVM can represent over 8 million different integer types,
5327 overloading is used commonly to allow an intrinsic function to operate on any
5328 integer type. One or more of the argument types or the result type can be
5329 overloaded to accept any integer type. Argument types may also be defined as
5330 exactly matching a previous argument's type or the result type. This allows
5331 an intrinsic function which accepts multiple arguments, but needs all of them
5332 to be of the same type, to only be overloaded with respect to a single
5333 argument or the result.</p>
5335 <p>Overloaded intrinsics will have the names of its overloaded argument types
5336 encoded into its function name, each preceded by a period. Only those types
5337 which are overloaded result in a name suffix. Arguments whose type is matched
5338 against another type do not. For example, the <tt>llvm.ctpop</tt> function
5339 can take an integer of any width and returns an integer of exactly the same
5340 integer width. This leads to a family of functions such as
5341 <tt>i8 @llvm.ctpop.i8(i8 %val)</tt> and <tt>i29 @llvm.ctpop.i29(i29
5342 %val)</tt>. Only one type, the return type, is overloaded, and only one type
5343 suffix is required. Because the argument's type is matched against the return
5344 type, it does not require its own name suffix.</p>
5346 <p>To learn how to add an intrinsic function, please see the
5347 <a href="ExtendingLLVM.html">Extending LLVM Guide</a>.</p>
5351 <!-- ======================================================================= -->
5352 <div class="doc_subsection">
5353 <a name="int_varargs">Variable Argument Handling Intrinsics</a>
5356 <div class="doc_text">
5358 <p>Variable argument support is defined in LLVM with
5359 the <a href="#i_va_arg"><tt>va_arg</tt></a> instruction and these three
5360 intrinsic functions. These functions are related to the similarly named
5361 macros defined in the <tt><stdarg.h></tt> header file.</p>
5363 <p>All of these functions operate on arguments that use a target-specific value
5364 type "<tt>va_list</tt>". The LLVM assembly language reference manual does
5365 not define what this type is, so all transformations should be prepared to
5366 handle these functions regardless of the type used.</p>
5368 <p>This example shows how the <a href="#i_va_arg"><tt>va_arg</tt></a>
5369 instruction and the variable argument handling intrinsic functions are
5372 <div class="doc_code">
5374 define i32 @test(i32 %X, ...) {
5375 ; Initialize variable argument processing
5377 %ap2 = bitcast i8** %ap to i8*
5378 call void @llvm.va_start(i8* %ap2)
5380 ; Read a single integer argument
5381 %tmp = va_arg i8** %ap, i32
5383 ; Demonstrate usage of llvm.va_copy and llvm.va_end
5385 %aq2 = bitcast i8** %aq to i8*
5386 call void @llvm.va_copy(i8* %aq2, i8* %ap2)
5387 call void @llvm.va_end(i8* %aq2)
5389 ; Stop processing of arguments.
5390 call void @llvm.va_end(i8* %ap2)
5394 declare void @llvm.va_start(i8*)
5395 declare void @llvm.va_copy(i8*, i8*)
5396 declare void @llvm.va_end(i8*)
5402 <!-- _______________________________________________________________________ -->
5403 <div class="doc_subsubsection">
5404 <a name="int_va_start">'<tt>llvm.va_start</tt>' Intrinsic</a>
5408 <div class="doc_text">
5412 declare void %llvm.va_start(i8* <arglist>)
5416 <p>The '<tt>llvm.va_start</tt>' intrinsic initializes <tt>*<arglist></tt>
5417 for subsequent use by <tt><a href="#i_va_arg">va_arg</a></tt>.</p>
5420 <p>The argument is a pointer to a <tt>va_list</tt> element to initialize.</p>
5423 <p>The '<tt>llvm.va_start</tt>' intrinsic works just like the <tt>va_start</tt>
5424 macro available in C. In a target-dependent way, it initializes
5425 the <tt>va_list</tt> element to which the argument points, so that the next
5426 call to <tt>va_arg</tt> will produce the first variable argument passed to
5427 the function. Unlike the C <tt>va_start</tt> macro, this intrinsic does not
5428 need to know the last argument of the function as the compiler can figure
5433 <!-- _______________________________________________________________________ -->
5434 <div class="doc_subsubsection">
5435 <a name="int_va_end">'<tt>llvm.va_end</tt>' Intrinsic</a>
5438 <div class="doc_text">
5442 declare void @llvm.va_end(i8* <arglist>)
5446 <p>The '<tt>llvm.va_end</tt>' intrinsic destroys <tt>*<arglist></tt>,
5447 which has been initialized previously
5448 with <tt><a href="#int_va_start">llvm.va_start</a></tt>
5449 or <tt><a href="#i_va_copy">llvm.va_copy</a></tt>.</p>
5452 <p>The argument is a pointer to a <tt>va_list</tt> to destroy.</p>
5455 <p>The '<tt>llvm.va_end</tt>' intrinsic works just like the <tt>va_end</tt>
5456 macro available in C. In a target-dependent way, it destroys
5457 the <tt>va_list</tt> element to which the argument points. Calls
5458 to <a href="#int_va_start"><tt>llvm.va_start</tt></a>
5459 and <a href="#int_va_copy"> <tt>llvm.va_copy</tt></a> must be matched exactly
5460 with calls to <tt>llvm.va_end</tt>.</p>
5464 <!-- _______________________________________________________________________ -->
5465 <div class="doc_subsubsection">
5466 <a name="int_va_copy">'<tt>llvm.va_copy</tt>' Intrinsic</a>
5469 <div class="doc_text">
5473 declare void @llvm.va_copy(i8* <destarglist>, i8* <srcarglist>)
5477 <p>The '<tt>llvm.va_copy</tt>' intrinsic copies the current argument position
5478 from the source argument list to the destination argument list.</p>
5481 <p>The first argument is a pointer to a <tt>va_list</tt> element to initialize.
5482 The second argument is a pointer to a <tt>va_list</tt> element to copy
5486 <p>The '<tt>llvm.va_copy</tt>' intrinsic works just like the <tt>va_copy</tt>
5487 macro available in C. In a target-dependent way, it copies the
5488 source <tt>va_list</tt> element into the destination <tt>va_list</tt>
5489 element. This intrinsic is necessary because
5490 the <tt><a href="#int_va_start"> llvm.va_start</a></tt> intrinsic may be
5491 arbitrarily complex and require, for example, memory allocation.</p>
5495 <!-- ======================================================================= -->
5496 <div class="doc_subsection">
5497 <a name="int_gc">Accurate Garbage Collection Intrinsics</a>
5500 <div class="doc_text">
5502 <p>LLVM support for <a href="GarbageCollection.html">Accurate Garbage
5503 Collection</a> (GC) requires the implementation and generation of these
5504 intrinsics. These intrinsics allow identification of <a href="#int_gcroot">GC
5505 roots on the stack</a>, as well as garbage collector implementations that
5506 require <a href="#int_gcread">read</a> and <a href="#int_gcwrite">write</a>
5507 barriers. Front-ends for type-safe garbage collected languages should generate
5508 these intrinsics to make use of the LLVM garbage collectors. For more details,
5509 see <a href="GarbageCollection.html">Accurate Garbage Collection with
5512 <p>The garbage collection intrinsics only operate on objects in the generic
5513 address space (address space zero).</p>
5517 <!-- _______________________________________________________________________ -->
5518 <div class="doc_subsubsection">
5519 <a name="int_gcroot">'<tt>llvm.gcroot</tt>' Intrinsic</a>
5522 <div class="doc_text">
5526 declare void @llvm.gcroot(i8** %ptrloc, i8* %metadata)
5530 <p>The '<tt>llvm.gcroot</tt>' intrinsic declares the existence of a GC root to
5531 the code generator, and allows some metadata to be associated with it.</p>
5534 <p>The first argument specifies the address of a stack object that contains the
5535 root pointer. The second pointer (which must be either a constant or a
5536 global value address) contains the meta-data to be associated with the
5540 <p>At runtime, a call to this intrinsic stores a null pointer into the "ptrloc"
5541 location. At compile-time, the code generator generates information to allow
5542 the runtime to find the pointer at GC safe points. The '<tt>llvm.gcroot</tt>'
5543 intrinsic may only be used in a function which <a href="#gc">specifies a GC
5548 <!-- _______________________________________________________________________ -->
5549 <div class="doc_subsubsection">
5550 <a name="int_gcread">'<tt>llvm.gcread</tt>' Intrinsic</a>
5553 <div class="doc_text">
5557 declare i8* @llvm.gcread(i8* %ObjPtr, i8** %Ptr)
5561 <p>The '<tt>llvm.gcread</tt>' intrinsic identifies reads of references from heap
5562 locations, allowing garbage collector implementations that require read
5566 <p>The second argument is the address to read from, which should be an address
5567 allocated from the garbage collector. The first object is a pointer to the
5568 start of the referenced object, if needed by the language runtime (otherwise
5572 <p>The '<tt>llvm.gcread</tt>' intrinsic has the same semantics as a load
5573 instruction, but may be replaced with substantially more complex code by the
5574 garbage collector runtime, as needed. The '<tt>llvm.gcread</tt>' intrinsic
5575 may only be used in a function which <a href="#gc">specifies a GC
5580 <!-- _______________________________________________________________________ -->
5581 <div class="doc_subsubsection">
5582 <a name="int_gcwrite">'<tt>llvm.gcwrite</tt>' Intrinsic</a>
5585 <div class="doc_text">
5589 declare void @llvm.gcwrite(i8* %P1, i8* %Obj, i8** %P2)
5593 <p>The '<tt>llvm.gcwrite</tt>' intrinsic identifies writes of references to heap
5594 locations, allowing garbage collector implementations that require write
5595 barriers (such as generational or reference counting collectors).</p>
5598 <p>The first argument is the reference to store, the second is the start of the
5599 object to store it to, and the third is the address of the field of Obj to
5600 store to. If the runtime does not require a pointer to the object, Obj may
5604 <p>The '<tt>llvm.gcwrite</tt>' intrinsic has the same semantics as a store
5605 instruction, but may be replaced with substantially more complex code by the
5606 garbage collector runtime, as needed. The '<tt>llvm.gcwrite</tt>' intrinsic
5607 may only be used in a function which <a href="#gc">specifies a GC
5612 <!-- ======================================================================= -->
5613 <div class="doc_subsection">
5614 <a name="int_codegen">Code Generator Intrinsics</a>
5617 <div class="doc_text">
5619 <p>These intrinsics are provided by LLVM to expose special features that may
5620 only be implemented with code generator support.</p>
5624 <!-- _______________________________________________________________________ -->
5625 <div class="doc_subsubsection">
5626 <a name="int_returnaddress">'<tt>llvm.returnaddress</tt>' Intrinsic</a>
5629 <div class="doc_text">
5633 declare i8 *@llvm.returnaddress(i32 <level>)
5637 <p>The '<tt>llvm.returnaddress</tt>' intrinsic attempts to compute a
5638 target-specific value indicating the return address of the current function
5639 or one of its callers.</p>
5642 <p>The argument to this intrinsic indicates which function to return the address
5643 for. Zero indicates the calling function, one indicates its caller, etc.
5644 The argument is <b>required</b> to be a constant integer value.</p>
5647 <p>The '<tt>llvm.returnaddress</tt>' intrinsic either returns a pointer
5648 indicating the return address of the specified call frame, or zero if it
5649 cannot be identified. The value returned by this intrinsic is likely to be
5650 incorrect or 0 for arguments other than zero, so it should only be used for
5651 debugging purposes.</p>
5653 <p>Note that calling this intrinsic does not prevent function inlining or other
5654 aggressive transformations, so the value returned may not be that of the
5655 obvious source-language caller.</p>
5659 <!-- _______________________________________________________________________ -->
5660 <div class="doc_subsubsection">
5661 <a name="int_frameaddress">'<tt>llvm.frameaddress</tt>' Intrinsic</a>
5664 <div class="doc_text">
5668 declare i8 *@llvm.frameaddress(i32 <level>)
5672 <p>The '<tt>llvm.frameaddress</tt>' intrinsic attempts to return the
5673 target-specific frame pointer value for the specified stack frame.</p>
5676 <p>The argument to this intrinsic indicates which function to return the frame
5677 pointer for. Zero indicates the calling function, one indicates its caller,
5678 etc. The argument is <b>required</b> to be a constant integer value.</p>
5681 <p>The '<tt>llvm.frameaddress</tt>' intrinsic either returns a pointer
5682 indicating the frame address of the specified call frame, or zero if it
5683 cannot be identified. The value returned by this intrinsic is likely to be
5684 incorrect or 0 for arguments other than zero, so it should only be used for
5685 debugging purposes.</p>
5687 <p>Note that calling this intrinsic does not prevent function inlining or other
5688 aggressive transformations, so the value returned may not be that of the
5689 obvious source-language caller.</p>
5693 <!-- _______________________________________________________________________ -->
5694 <div class="doc_subsubsection">
5695 <a name="int_stacksave">'<tt>llvm.stacksave</tt>' Intrinsic</a>
5698 <div class="doc_text">
5702 declare i8 *@llvm.stacksave()
5706 <p>The '<tt>llvm.stacksave</tt>' intrinsic is used to remember the current state
5707 of the function stack, for use
5708 with <a href="#int_stackrestore"> <tt>llvm.stackrestore</tt></a>. This is
5709 useful for implementing language features like scoped automatic variable
5710 sized arrays in C99.</p>
5713 <p>This intrinsic returns a opaque pointer value that can be passed
5714 to <a href="#int_stackrestore"><tt>llvm.stackrestore</tt></a>. When
5715 an <tt>llvm.stackrestore</tt> intrinsic is executed with a value saved
5716 from <tt>llvm.stacksave</tt>, it effectively restores the state of the stack
5717 to the state it was in when the <tt>llvm.stacksave</tt> intrinsic executed.
5718 In practice, this pops any <a href="#i_alloca">alloca</a> blocks from the
5719 stack that were allocated after the <tt>llvm.stacksave</tt> was executed.</p>
5723 <!-- _______________________________________________________________________ -->
5724 <div class="doc_subsubsection">
5725 <a name="int_stackrestore">'<tt>llvm.stackrestore</tt>' Intrinsic</a>
5728 <div class="doc_text">
5732 declare void @llvm.stackrestore(i8 * %ptr)
5736 <p>The '<tt>llvm.stackrestore</tt>' intrinsic is used to restore the state of
5737 the function stack to the state it was in when the
5738 corresponding <a href="#int_stacksave"><tt>llvm.stacksave</tt></a> intrinsic
5739 executed. This is useful for implementing language features like scoped
5740 automatic variable sized arrays in C99.</p>
5743 <p>See the description
5744 for <a href="#int_stacksave"><tt>llvm.stacksave</tt></a>.</p>
5748 <!-- _______________________________________________________________________ -->
5749 <div class="doc_subsubsection">
5750 <a name="int_prefetch">'<tt>llvm.prefetch</tt>' Intrinsic</a>
5753 <div class="doc_text">
5757 declare void @llvm.prefetch(i8* <address>, i32 <rw>, i32 <locality>)
5761 <p>The '<tt>llvm.prefetch</tt>' intrinsic is a hint to the code generator to
5762 insert a prefetch instruction if supported; otherwise, it is a noop.
5763 Prefetches have no effect on the behavior of the program but can change its
5764 performance characteristics.</p>
5767 <p><tt>address</tt> is the address to be prefetched, <tt>rw</tt> is the
5768 specifier determining if the fetch should be for a read (0) or write (1),
5769 and <tt>locality</tt> is a temporal locality specifier ranging from (0) - no
5770 locality, to (3) - extremely local keep in cache. The <tt>rw</tt>
5771 and <tt>locality</tt> arguments must be constant integers.</p>
5774 <p>This intrinsic does not modify the behavior of the program. In particular,
5775 prefetches cannot trap and do not produce a value. On targets that support
5776 this intrinsic, the prefetch can provide hints to the processor cache for
5777 better performance.</p>
5781 <!-- _______________________________________________________________________ -->
5782 <div class="doc_subsubsection">
5783 <a name="int_pcmarker">'<tt>llvm.pcmarker</tt>' Intrinsic</a>
5786 <div class="doc_text">
5790 declare void @llvm.pcmarker(i32 <id>)
5794 <p>The '<tt>llvm.pcmarker</tt>' intrinsic is a method to export a Program
5795 Counter (PC) in a region of code to simulators and other tools. The method
5796 is target specific, but it is expected that the marker will use exported
5797 symbols to transmit the PC of the marker. The marker makes no guarantees
5798 that it will remain with any specific instruction after optimizations. It is
5799 possible that the presence of a marker will inhibit optimizations. The
5800 intended use is to be inserted after optimizations to allow correlations of
5801 simulation runs.</p>
5804 <p><tt>id</tt> is a numerical id identifying the marker.</p>
5807 <p>This intrinsic does not modify the behavior of the program. Backends that do
5808 not support this intrinsic may ignore it.</p>
5812 <!-- _______________________________________________________________________ -->
5813 <div class="doc_subsubsection">
5814 <a name="int_readcyclecounter">'<tt>llvm.readcyclecounter</tt>' Intrinsic</a>
5817 <div class="doc_text">
5821 declare i64 @llvm.readcyclecounter( )
5825 <p>The '<tt>llvm.readcyclecounter</tt>' intrinsic provides access to the cycle
5826 counter register (or similar low latency, high accuracy clocks) on those
5827 targets that support it. On X86, it should map to RDTSC. On Alpha, it
5828 should map to RPCC. As the backing counters overflow quickly (on the order
5829 of 9 seconds on alpha), this should only be used for small timings.</p>
5832 <p>When directly supported, reading the cycle counter should not modify any
5833 memory. Implementations are allowed to either return a application specific
5834 value or a system wide value. On backends without support, this is lowered
5835 to a constant 0.</p>
5839 <!-- ======================================================================= -->
5840 <div class="doc_subsection">
5841 <a name="int_libc">Standard C Library Intrinsics</a>
5844 <div class="doc_text">
5846 <p>LLVM provides intrinsics for a few important standard C library functions.
5847 These intrinsics allow source-language front-ends to pass information about
5848 the alignment of the pointer arguments to the code generator, providing
5849 opportunity for more efficient code generation.</p>
5853 <!-- _______________________________________________________________________ -->
5854 <div class="doc_subsubsection">
5855 <a name="int_memcpy">'<tt>llvm.memcpy</tt>' Intrinsic</a>
5858 <div class="doc_text">
5861 <p>This is an overloaded intrinsic. You can use <tt>llvm.memcpy</tt> on any
5862 integer bit width. Not all targets support all bit widths however.</p>
5865 declare void @llvm.memcpy.i8(i8 * <dest>, i8 * <src>,
5866 i8 <len>, i32 <align>)
5867 declare void @llvm.memcpy.i16(i8 * <dest>, i8 * <src>,
5868 i16 <len>, i32 <align>)
5869 declare void @llvm.memcpy.i32(i8 * <dest>, i8 * <src>,
5870 i32 <len>, i32 <align>)
5871 declare void @llvm.memcpy.i64(i8 * <dest>, i8 * <src>,
5872 i64 <len>, i32 <align>)
5876 <p>The '<tt>llvm.memcpy.*</tt>' intrinsics copy a block of memory from the
5877 source location to the destination location.</p>
5879 <p>Note that, unlike the standard libc function, the <tt>llvm.memcpy.*</tt>
5880 intrinsics do not return a value, and takes an extra alignment argument.</p>
5883 <p>The first argument is a pointer to the destination, the second is a pointer
5884 to the source. The third argument is an integer argument specifying the
5885 number of bytes to copy, and the fourth argument is the alignment of the
5886 source and destination locations.</p>
5888 <p>If the call to this intrinsic has an alignment value that is not 0 or 1,
5889 then the caller guarantees that both the source and destination pointers are
5890 aligned to that boundary.</p>
5893 <p>The '<tt>llvm.memcpy.*</tt>' intrinsics copy a block of memory from the
5894 source location to the destination location, which are not allowed to
5895 overlap. It copies "len" bytes of memory over. If the argument is known to
5896 be aligned to some boundary, this can be specified as the fourth argument,
5897 otherwise it should be set to 0 or 1.</p>
5901 <!-- _______________________________________________________________________ -->
5902 <div class="doc_subsubsection">
5903 <a name="int_memmove">'<tt>llvm.memmove</tt>' Intrinsic</a>
5906 <div class="doc_text">
5909 <p>This is an overloaded intrinsic. You can use llvm.memmove on any integer bit
5910 width. Not all targets support all bit widths however.</p>
5913 declare void @llvm.memmove.i8(i8 * <dest>, i8 * <src>,
5914 i8 <len>, i32 <align>)
5915 declare void @llvm.memmove.i16(i8 * <dest>, i8 * <src>,
5916 i16 <len>, i32 <align>)
5917 declare void @llvm.memmove.i32(i8 * <dest>, i8 * <src>,
5918 i32 <len>, i32 <align>)
5919 declare void @llvm.memmove.i64(i8 * <dest>, i8 * <src>,
5920 i64 <len>, i32 <align>)
5924 <p>The '<tt>llvm.memmove.*</tt>' intrinsics move a block of memory from the
5925 source location to the destination location. It is similar to the
5926 '<tt>llvm.memcpy</tt>' intrinsic but allows the two memory locations to
5929 <p>Note that, unlike the standard libc function, the <tt>llvm.memmove.*</tt>
5930 intrinsics do not return a value, and takes an extra alignment argument.</p>
5933 <p>The first argument is a pointer to the destination, the second is a pointer
5934 to the source. The third argument is an integer argument specifying the
5935 number of bytes to copy, and the fourth argument is the alignment of the
5936 source and destination locations.</p>
5938 <p>If the call to this intrinsic has an alignment value that is not 0 or 1,
5939 then the caller guarantees that the source and destination pointers are
5940 aligned to that boundary.</p>
5943 <p>The '<tt>llvm.memmove.*</tt>' intrinsics copy a block of memory from the
5944 source location to the destination location, which may overlap. It copies
5945 "len" bytes of memory over. If the argument is known to be aligned to some
5946 boundary, this can be specified as the fourth argument, otherwise it should
5947 be set to 0 or 1.</p>
5951 <!-- _______________________________________________________________________ -->
5952 <div class="doc_subsubsection">
5953 <a name="int_memset">'<tt>llvm.memset.*</tt>' Intrinsics</a>
5956 <div class="doc_text">
5959 <p>This is an overloaded intrinsic. You can use llvm.memset on any integer bit
5960 width. Not all targets support all bit widths however.</p>
5963 declare void @llvm.memset.i8(i8 * <dest>, i8 <val>,
5964 i8 <len>, i32 <align>)
5965 declare void @llvm.memset.i16(i8 * <dest>, i8 <val>,
5966 i16 <len>, i32 <align>)
5967 declare void @llvm.memset.i32(i8 * <dest>, i8 <val>,
5968 i32 <len>, i32 <align>)
5969 declare void @llvm.memset.i64(i8 * <dest>, i8 <val>,
5970 i64 <len>, i32 <align>)
5974 <p>The '<tt>llvm.memset.*</tt>' intrinsics fill a block of memory with a
5975 particular byte value.</p>
5977 <p>Note that, unlike the standard libc function, the <tt>llvm.memset</tt>
5978 intrinsic does not return a value, and takes an extra alignment argument.</p>
5981 <p>The first argument is a pointer to the destination to fill, the second is the
5982 byte value to fill it with, the third argument is an integer argument
5983 specifying the number of bytes to fill, and the fourth argument is the known
5984 alignment of destination location.</p>
5986 <p>If the call to this intrinsic has an alignment value that is not 0 or 1,
5987 then the caller guarantees that the destination pointer is aligned to that
5991 <p>The '<tt>llvm.memset.*</tt>' intrinsics fill "len" bytes of memory starting
5992 at the destination location. If the argument is known to be aligned to some
5993 boundary, this can be specified as the fourth argument, otherwise it should
5994 be set to 0 or 1.</p>
5998 <!-- _______________________________________________________________________ -->
5999 <div class="doc_subsubsection">
6000 <a name="int_sqrt">'<tt>llvm.sqrt.*</tt>' Intrinsic</a>
6003 <div class="doc_text">
6006 <p>This is an overloaded intrinsic. You can use <tt>llvm.sqrt</tt> on any
6007 floating point or vector of floating point type. Not all targets support all
6011 declare float @llvm.sqrt.f32(float %Val)
6012 declare double @llvm.sqrt.f64(double %Val)
6013 declare x86_fp80 @llvm.sqrt.f80(x86_fp80 %Val)
6014 declare fp128 @llvm.sqrt.f128(fp128 %Val)
6015 declare ppc_fp128 @llvm.sqrt.ppcf128(ppc_fp128 %Val)
6019 <p>The '<tt>llvm.sqrt</tt>' intrinsics return the sqrt of the specified operand,
6020 returning the same value as the libm '<tt>sqrt</tt>' functions would.
6021 Unlike <tt>sqrt</tt> in libm, however, <tt>llvm.sqrt</tt> has undefined
6022 behavior for negative numbers other than -0.0 (which allows for better
6023 optimization, because there is no need to worry about errno being
6024 set). <tt>llvm.sqrt(-0.0)</tt> is defined to return -0.0 like IEEE sqrt.</p>
6027 <p>The argument and return value are floating point numbers of the same
6031 <p>This function returns the sqrt of the specified operand if it is a
6032 nonnegative floating point number.</p>
6036 <!-- _______________________________________________________________________ -->
6037 <div class="doc_subsubsection">
6038 <a name="int_powi">'<tt>llvm.powi.*</tt>' Intrinsic</a>
6041 <div class="doc_text">
6044 <p>This is an overloaded intrinsic. You can use <tt>llvm.powi</tt> on any
6045 floating point or vector of floating point type. Not all targets support all
6049 declare float @llvm.powi.f32(float %Val, i32 %power)
6050 declare double @llvm.powi.f64(double %Val, i32 %power)
6051 declare x86_fp80 @llvm.powi.f80(x86_fp80 %Val, i32 %power)
6052 declare fp128 @llvm.powi.f128(fp128 %Val, i32 %power)
6053 declare ppc_fp128 @llvm.powi.ppcf128(ppc_fp128 %Val, i32 %power)
6057 <p>The '<tt>llvm.powi.*</tt>' intrinsics return the first operand raised to the
6058 specified (positive or negative) power. The order of evaluation of
6059 multiplications is not defined. When a vector of floating point type is
6060 used, the second argument remains a scalar integer value.</p>
6063 <p>The second argument is an integer power, and the first is a value to raise to
6067 <p>This function returns the first value raised to the second power with an
6068 unspecified sequence of rounding operations.</p>
6072 <!-- _______________________________________________________________________ -->
6073 <div class="doc_subsubsection">
6074 <a name="int_sin">'<tt>llvm.sin.*</tt>' Intrinsic</a>
6077 <div class="doc_text">
6080 <p>This is an overloaded intrinsic. You can use <tt>llvm.sin</tt> on any
6081 floating point or vector of floating point type. Not all targets support all
6085 declare float @llvm.sin.f32(float %Val)
6086 declare double @llvm.sin.f64(double %Val)
6087 declare x86_fp80 @llvm.sin.f80(x86_fp80 %Val)
6088 declare fp128 @llvm.sin.f128(fp128 %Val)
6089 declare ppc_fp128 @llvm.sin.ppcf128(ppc_fp128 %Val)
6093 <p>The '<tt>llvm.sin.*</tt>' intrinsics return the sine of the operand.</p>
6096 <p>The argument and return value are floating point numbers of the same
6100 <p>This function returns the sine of the specified operand, returning the same
6101 values as the libm <tt>sin</tt> functions would, and handles error conditions
6102 in the same way.</p>
6106 <!-- _______________________________________________________________________ -->
6107 <div class="doc_subsubsection">
6108 <a name="int_cos">'<tt>llvm.cos.*</tt>' Intrinsic</a>
6111 <div class="doc_text">
6114 <p>This is an overloaded intrinsic. You can use <tt>llvm.cos</tt> on any
6115 floating point or vector of floating point type. Not all targets support all
6119 declare float @llvm.cos.f32(float %Val)
6120 declare double @llvm.cos.f64(double %Val)
6121 declare x86_fp80 @llvm.cos.f80(x86_fp80 %Val)
6122 declare fp128 @llvm.cos.f128(fp128 %Val)
6123 declare ppc_fp128 @llvm.cos.ppcf128(ppc_fp128 %Val)
6127 <p>The '<tt>llvm.cos.*</tt>' intrinsics return the cosine of the operand.</p>
6130 <p>The argument and return value are floating point numbers of the same
6134 <p>This function returns the cosine of the specified operand, returning the same
6135 values as the libm <tt>cos</tt> functions would, and handles error conditions
6136 in the same way.</p>
6140 <!-- _______________________________________________________________________ -->
6141 <div class="doc_subsubsection">
6142 <a name="int_pow">'<tt>llvm.pow.*</tt>' Intrinsic</a>
6145 <div class="doc_text">
6148 <p>This is an overloaded intrinsic. You can use <tt>llvm.pow</tt> on any
6149 floating point or vector of floating point type. Not all targets support all
6153 declare float @llvm.pow.f32(float %Val, float %Power)
6154 declare double @llvm.pow.f64(double %Val, double %Power)
6155 declare x86_fp80 @llvm.pow.f80(x86_fp80 %Val, x86_fp80 %Power)
6156 declare fp128 @llvm.pow.f128(fp128 %Val, fp128 %Power)
6157 declare ppc_fp128 @llvm.pow.ppcf128(ppc_fp128 %Val, ppc_fp128 Power)
6161 <p>The '<tt>llvm.pow.*</tt>' intrinsics return the first operand raised to the
6162 specified (positive or negative) power.</p>
6165 <p>The second argument is a floating point power, and the first is a value to
6166 raise to that power.</p>
6169 <p>This function returns the first value raised to the second power, returning
6170 the same values as the libm <tt>pow</tt> functions would, and handles error
6171 conditions in the same way.</p>
6175 <!-- ======================================================================= -->
6176 <div class="doc_subsection">
6177 <a name="int_manip">Bit Manipulation Intrinsics</a>
6180 <div class="doc_text">
6182 <p>LLVM provides intrinsics for a few important bit manipulation operations.
6183 These allow efficient code generation for some algorithms.</p>
6187 <!-- _______________________________________________________________________ -->
6188 <div class="doc_subsubsection">
6189 <a name="int_bswap">'<tt>llvm.bswap.*</tt>' Intrinsics</a>
6192 <div class="doc_text">
6195 <p>This is an overloaded intrinsic function. You can use bswap on any integer
6196 type that is an even number of bytes (i.e. BitWidth % 16 == 0).</p>
6199 declare i16 @llvm.bswap.i16(i16 <id>)
6200 declare i32 @llvm.bswap.i32(i32 <id>)
6201 declare i64 @llvm.bswap.i64(i64 <id>)
6205 <p>The '<tt>llvm.bswap</tt>' family of intrinsics is used to byte swap integer
6206 values with an even number of bytes (positive multiple of 16 bits). These
6207 are useful for performing operations on data that is not in the target's
6208 native byte order.</p>
6211 <p>The <tt>llvm.bswap.i16</tt> intrinsic returns an i16 value that has the high
6212 and low byte of the input i16 swapped. Similarly,
6213 the <tt>llvm.bswap.i32</tt> intrinsic returns an i32 value that has the four
6214 bytes of the input i32 swapped, so that if the input bytes are numbered 0, 1,
6215 2, 3 then the returned i32 will have its bytes in 3, 2, 1, 0 order.
6216 The <tt>llvm.bswap.i48</tt>, <tt>llvm.bswap.i64</tt> and other intrinsics
6217 extend this concept to additional even-byte lengths (6 bytes, 8 bytes and
6218 more, respectively).</p>
6222 <!-- _______________________________________________________________________ -->
6223 <div class="doc_subsubsection">
6224 <a name="int_ctpop">'<tt>llvm.ctpop.*</tt>' Intrinsic</a>
6227 <div class="doc_text">
6230 <p>This is an overloaded intrinsic. You can use llvm.ctpop on any integer bit
6231 width. Not all targets support all bit widths however.</p>
6234 declare i8 @llvm.ctpop.i8(i8 <src>)
6235 declare i16 @llvm.ctpop.i16(i16 <src>)
6236 declare i32 @llvm.ctpop.i32(i32 <src>)
6237 declare i64 @llvm.ctpop.i64(i64 <src>)
6238 declare i256 @llvm.ctpop.i256(i256 <src>)
6242 <p>The '<tt>llvm.ctpop</tt>' family of intrinsics counts the number of bits set
6246 <p>The only argument is the value to be counted. The argument may be of any
6247 integer type. The return type must match the argument type.</p>
6250 <p>The '<tt>llvm.ctpop</tt>' intrinsic counts the 1's in a variable.</p>
6254 <!-- _______________________________________________________________________ -->
6255 <div class="doc_subsubsection">
6256 <a name="int_ctlz">'<tt>llvm.ctlz.*</tt>' Intrinsic</a>
6259 <div class="doc_text">
6262 <p>This is an overloaded intrinsic. You can use <tt>llvm.ctlz</tt> on any
6263 integer bit width. Not all targets support all bit widths however.</p>
6266 declare i8 @llvm.ctlz.i8 (i8 <src>)
6267 declare i16 @llvm.ctlz.i16(i16 <src>)
6268 declare i32 @llvm.ctlz.i32(i32 <src>)
6269 declare i64 @llvm.ctlz.i64(i64 <src>)
6270 declare i256 @llvm.ctlz.i256(i256 <src>)
6274 <p>The '<tt>llvm.ctlz</tt>' family of intrinsic functions counts the number of
6275 leading zeros in a variable.</p>
6278 <p>The only argument is the value to be counted. The argument may be of any
6279 integer type. The return type must match the argument type.</p>
6282 <p>The '<tt>llvm.ctlz</tt>' intrinsic counts the leading (most significant)
6283 zeros in a variable. If the src == 0 then the result is the size in bits of
6284 the type of src. For example, <tt>llvm.ctlz(i32 2) = 30</tt>.</p>
6288 <!-- _______________________________________________________________________ -->
6289 <div class="doc_subsubsection">
6290 <a name="int_cttz">'<tt>llvm.cttz.*</tt>' Intrinsic</a>
6293 <div class="doc_text">
6296 <p>This is an overloaded intrinsic. You can use <tt>llvm.cttz</tt> on any
6297 integer bit width. Not all targets support all bit widths however.</p>
6300 declare i8 @llvm.cttz.i8 (i8 <src>)
6301 declare i16 @llvm.cttz.i16(i16 <src>)
6302 declare i32 @llvm.cttz.i32(i32 <src>)
6303 declare i64 @llvm.cttz.i64(i64 <src>)
6304 declare i256 @llvm.cttz.i256(i256 <src>)
6308 <p>The '<tt>llvm.cttz</tt>' family of intrinsic functions counts the number of
6312 <p>The only argument is the value to be counted. The argument may be of any
6313 integer type. The return type must match the argument type.</p>
6316 <p>The '<tt>llvm.cttz</tt>' intrinsic counts the trailing (least significant)
6317 zeros in a variable. If the src == 0 then the result is the size in bits of
6318 the type of src. For example, <tt>llvm.cttz(2) = 1</tt>.</p>
6322 <!-- ======================================================================= -->
6323 <div class="doc_subsection">
6324 <a name="int_overflow">Arithmetic with Overflow Intrinsics</a>
6327 <div class="doc_text">
6329 <p>LLVM provides intrinsics for some arithmetic with overflow operations.</p>
6333 <!-- _______________________________________________________________________ -->
6334 <div class="doc_subsubsection">
6335 <a name="int_sadd_overflow">'<tt>llvm.sadd.with.overflow.*</tt>' Intrinsics</a>
6338 <div class="doc_text">
6341 <p>This is an overloaded intrinsic. You can use <tt>llvm.sadd.with.overflow</tt>
6342 on any integer bit width.</p>
6345 declare {i16, i1} @llvm.sadd.with.overflow.i16(i16 %a, i16 %b)
6346 declare {i32, i1} @llvm.sadd.with.overflow.i32(i32 %a, i32 %b)
6347 declare {i64, i1} @llvm.sadd.with.overflow.i64(i64 %a, i64 %b)
6351 <p>The '<tt>llvm.sadd.with.overflow</tt>' family of intrinsic functions perform
6352 a signed addition of the two arguments, and indicate whether an overflow
6353 occurred during the signed summation.</p>
6356 <p>The arguments (%a and %b) and the first element of the result structure may
6357 be of integer types of any bit width, but they must have the same bit
6358 width. The second element of the result structure must be of
6359 type <tt>i1</tt>. <tt>%a</tt> and <tt>%b</tt> are the two values that will
6360 undergo signed addition.</p>
6363 <p>The '<tt>llvm.sadd.with.overflow</tt>' family of intrinsic functions perform
6364 a signed addition of the two variables. They return a structure — the
6365 first element of which is the signed summation, and the second element of
6366 which is a bit specifying if the signed summation resulted in an
6371 %res = call {i32, i1} @llvm.sadd.with.overflow.i32(i32 %a, i32 %b)
6372 %sum = extractvalue {i32, i1} %res, 0
6373 %obit = extractvalue {i32, i1} %res, 1
6374 br i1 %obit, label %overflow, label %normal
6379 <!-- _______________________________________________________________________ -->
6380 <div class="doc_subsubsection">
6381 <a name="int_uadd_overflow">'<tt>llvm.uadd.with.overflow.*</tt>' Intrinsics</a>
6384 <div class="doc_text">
6387 <p>This is an overloaded intrinsic. You can use <tt>llvm.uadd.with.overflow</tt>
6388 on any integer bit width.</p>
6391 declare {i16, i1} @llvm.uadd.with.overflow.i16(i16 %a, i16 %b)
6392 declare {i32, i1} @llvm.uadd.with.overflow.i32(i32 %a, i32 %b)
6393 declare {i64, i1} @llvm.uadd.with.overflow.i64(i64 %a, i64 %b)
6397 <p>The '<tt>llvm.uadd.with.overflow</tt>' family of intrinsic functions perform
6398 an unsigned addition of the two arguments, and indicate whether a carry
6399 occurred during the unsigned summation.</p>
6402 <p>The arguments (%a and %b) and the first element of the result structure may
6403 be of integer types of any bit width, but they must have the same bit
6404 width. The second element of the result structure must be of
6405 type <tt>i1</tt>. <tt>%a</tt> and <tt>%b</tt> are the two values that will
6406 undergo unsigned addition.</p>
6409 <p>The '<tt>llvm.uadd.with.overflow</tt>' family of intrinsic functions perform
6410 an unsigned addition of the two arguments. They return a structure —
6411 the first element of which is the sum, and the second element of which is a
6412 bit specifying if the unsigned summation resulted in a carry.</p>
6416 %res = call {i32, i1} @llvm.uadd.with.overflow.i32(i32 %a, i32 %b)
6417 %sum = extractvalue {i32, i1} %res, 0
6418 %obit = extractvalue {i32, i1} %res, 1
6419 br i1 %obit, label %carry, label %normal
6424 <!-- _______________________________________________________________________ -->
6425 <div class="doc_subsubsection">
6426 <a name="int_ssub_overflow">'<tt>llvm.ssub.with.overflow.*</tt>' Intrinsics</a>
6429 <div class="doc_text">
6432 <p>This is an overloaded intrinsic. You can use <tt>llvm.ssub.with.overflow</tt>
6433 on any integer bit width.</p>
6436 declare {i16, i1} @llvm.ssub.with.overflow.i16(i16 %a, i16 %b)
6437 declare {i32, i1} @llvm.ssub.with.overflow.i32(i32 %a, i32 %b)
6438 declare {i64, i1} @llvm.ssub.with.overflow.i64(i64 %a, i64 %b)
6442 <p>The '<tt>llvm.ssub.with.overflow</tt>' family of intrinsic functions perform
6443 a signed subtraction of the two arguments, and indicate whether an overflow
6444 occurred during the signed subtraction.</p>
6447 <p>The arguments (%a and %b) and the first element of the result structure may
6448 be of integer types of any bit width, but they must have the same bit
6449 width. The second element of the result structure must be of
6450 type <tt>i1</tt>. <tt>%a</tt> and <tt>%b</tt> are the two values that will
6451 undergo signed subtraction.</p>
6454 <p>The '<tt>llvm.ssub.with.overflow</tt>' family of intrinsic functions perform
6455 a signed subtraction of the two arguments. They return a structure —
6456 the first element of which is the subtraction, and the second element of
6457 which is a bit specifying if the signed subtraction resulted in an
6462 %res = call {i32, i1} @llvm.ssub.with.overflow.i32(i32 %a, i32 %b)
6463 %sum = extractvalue {i32, i1} %res, 0
6464 %obit = extractvalue {i32, i1} %res, 1
6465 br i1 %obit, label %overflow, label %normal
6470 <!-- _______________________________________________________________________ -->
6471 <div class="doc_subsubsection">
6472 <a name="int_usub_overflow">'<tt>llvm.usub.with.overflow.*</tt>' Intrinsics</a>
6475 <div class="doc_text">
6478 <p>This is an overloaded intrinsic. You can use <tt>llvm.usub.with.overflow</tt>
6479 on any integer bit width.</p>
6482 declare {i16, i1} @llvm.usub.with.overflow.i16(i16 %a, i16 %b)
6483 declare {i32, i1} @llvm.usub.with.overflow.i32(i32 %a, i32 %b)
6484 declare {i64, i1} @llvm.usub.with.overflow.i64(i64 %a, i64 %b)
6488 <p>The '<tt>llvm.usub.with.overflow</tt>' family of intrinsic functions perform
6489 an unsigned subtraction of the two arguments, and indicate whether an
6490 overflow occurred during the unsigned subtraction.</p>
6493 <p>The arguments (%a and %b) and the first element of the result structure may
6494 be of integer types of any bit width, but they must have the same bit
6495 width. The second element of the result structure must be of
6496 type <tt>i1</tt>. <tt>%a</tt> and <tt>%b</tt> are the two values that will
6497 undergo unsigned subtraction.</p>
6500 <p>The '<tt>llvm.usub.with.overflow</tt>' family of intrinsic functions perform
6501 an unsigned subtraction of the two arguments. They return a structure —
6502 the first element of which is the subtraction, and the second element of
6503 which is a bit specifying if the unsigned subtraction resulted in an
6508 %res = call {i32, i1} @llvm.usub.with.overflow.i32(i32 %a, i32 %b)
6509 %sum = extractvalue {i32, i1} %res, 0
6510 %obit = extractvalue {i32, i1} %res, 1
6511 br i1 %obit, label %overflow, label %normal
6516 <!-- _______________________________________________________________________ -->
6517 <div class="doc_subsubsection">
6518 <a name="int_smul_overflow">'<tt>llvm.smul.with.overflow.*</tt>' Intrinsics</a>
6521 <div class="doc_text">
6524 <p>This is an overloaded intrinsic. You can use <tt>llvm.smul.with.overflow</tt>
6525 on any integer bit width.</p>
6528 declare {i16, i1} @llvm.smul.with.overflow.i16(i16 %a, i16 %b)
6529 declare {i32, i1} @llvm.smul.with.overflow.i32(i32 %a, i32 %b)
6530 declare {i64, i1} @llvm.smul.with.overflow.i64(i64 %a, i64 %b)
6535 <p>The '<tt>llvm.smul.with.overflow</tt>' family of intrinsic functions perform
6536 a signed multiplication of the two arguments, and indicate whether an
6537 overflow occurred during the signed multiplication.</p>
6540 <p>The arguments (%a and %b) and the first element of the result structure may
6541 be of integer types of any bit width, but they must have the same bit
6542 width. The second element of the result structure must be of
6543 type <tt>i1</tt>. <tt>%a</tt> and <tt>%b</tt> are the two values that will
6544 undergo signed multiplication.</p>
6547 <p>The '<tt>llvm.smul.with.overflow</tt>' family of intrinsic functions perform
6548 a signed multiplication of the two arguments. They return a structure —
6549 the first element of which is the multiplication, and the second element of
6550 which is a bit specifying if the signed multiplication resulted in an
6555 %res = call {i32, i1} @llvm.smul.with.overflow.i32(i32 %a, i32 %b)
6556 %sum = extractvalue {i32, i1} %res, 0
6557 %obit = extractvalue {i32, i1} %res, 1
6558 br i1 %obit, label %overflow, label %normal
6563 <!-- _______________________________________________________________________ -->
6564 <div class="doc_subsubsection">
6565 <a name="int_umul_overflow">'<tt>llvm.umul.with.overflow.*</tt>' Intrinsics</a>
6568 <div class="doc_text">
6571 <p>This is an overloaded intrinsic. You can use <tt>llvm.umul.with.overflow</tt>
6572 on any integer bit width.</p>
6575 declare {i16, i1} @llvm.umul.with.overflow.i16(i16 %a, i16 %b)
6576 declare {i32, i1} @llvm.umul.with.overflow.i32(i32 %a, i32 %b)
6577 declare {i64, i1} @llvm.umul.with.overflow.i64(i64 %a, i64 %b)
6581 <p>The '<tt>llvm.umul.with.overflow</tt>' family of intrinsic functions perform
6582 a unsigned multiplication of the two arguments, and indicate whether an
6583 overflow occurred during the unsigned multiplication.</p>
6586 <p>The arguments (%a and %b) and the first element of the result structure may
6587 be of integer types of any bit width, but they must have the same bit
6588 width. The second element of the result structure must be of
6589 type <tt>i1</tt>. <tt>%a</tt> and <tt>%b</tt> are the two values that will
6590 undergo unsigned multiplication.</p>
6593 <p>The '<tt>llvm.umul.with.overflow</tt>' family of intrinsic functions perform
6594 an unsigned multiplication of the two arguments. They return a structure
6595 — the first element of which is the multiplication, and the second
6596 element of which is a bit specifying if the unsigned multiplication resulted
6601 %res = call {i32, i1} @llvm.umul.with.overflow.i32(i32 %a, i32 %b)
6602 %sum = extractvalue {i32, i1} %res, 0
6603 %obit = extractvalue {i32, i1} %res, 1
6604 br i1 %obit, label %overflow, label %normal
6609 <!-- ======================================================================= -->
6610 <div class="doc_subsection">
6611 <a name="int_debugger">Debugger Intrinsics</a>
6614 <div class="doc_text">
6616 <p>The LLVM debugger intrinsics (which all start with <tt>llvm.dbg.</tt>
6617 prefix), are described in
6618 the <a href="SourceLevelDebugging.html#format_common_intrinsics">LLVM Source
6619 Level Debugging</a> document.</p>
6623 <!-- ======================================================================= -->
6624 <div class="doc_subsection">
6625 <a name="int_eh">Exception Handling Intrinsics</a>
6628 <div class="doc_text">
6630 <p>The LLVM exception handling intrinsics (which all start with
6631 <tt>llvm.eh.</tt> prefix), are described in
6632 the <a href="ExceptionHandling.html#format_common_intrinsics">LLVM Exception
6633 Handling</a> document.</p>
6637 <!-- ======================================================================= -->
6638 <div class="doc_subsection">
6639 <a name="int_trampoline">Trampoline Intrinsic</a>
6642 <div class="doc_text">
6644 <p>This intrinsic makes it possible to excise one parameter, marked with
6645 the <tt>nest</tt> attribute, from a function. The result is a callable
6646 function pointer lacking the nest parameter - the caller does not need to
6647 provide a value for it. Instead, the value to use is stored in advance in a
6648 "trampoline", a block of memory usually allocated on the stack, which also
6649 contains code to splice the nest value into the argument list. This is used
6650 to implement the GCC nested function address extension.</p>
6652 <p>For example, if the function is
6653 <tt>i32 f(i8* nest %c, i32 %x, i32 %y)</tt> then the resulting function
6654 pointer has signature <tt>i32 (i32, i32)*</tt>. It can be created as
6657 <div class="doc_code">
6659 %tramp = alloca [10 x i8], align 4 ; size and alignment only correct for X86
6660 %tramp1 = getelementptr [10 x i8]* %tramp, i32 0, i32 0
6661 %p = call i8* @llvm.init.trampoline( i8* %tramp1, i8* bitcast (i32 (i8* nest , i32, i32)* @f to i8*), i8* %nval )
6662 %fp = bitcast i8* %p to i32 (i32, i32)*
6666 <p>The call <tt>%val = call i32 %fp( i32 %x, i32 %y )</tt> is then equivalent
6667 to <tt>%val = call i32 %f( i8* %nval, i32 %x, i32 %y )</tt>.</p>
6671 <!-- _______________________________________________________________________ -->
6672 <div class="doc_subsubsection">
6673 <a name="int_it">'<tt>llvm.init.trampoline</tt>' Intrinsic</a>
6676 <div class="doc_text">
6680 declare i8* @llvm.init.trampoline(i8* <tramp>, i8* <func>, i8* <nval>)
6684 <p>This fills the memory pointed to by <tt>tramp</tt> with code and returns a
6685 function pointer suitable for executing it.</p>
6688 <p>The <tt>llvm.init.trampoline</tt> intrinsic takes three arguments, all
6689 pointers. The <tt>tramp</tt> argument must point to a sufficiently large and
6690 sufficiently aligned block of memory; this memory is written to by the
6691 intrinsic. Note that the size and the alignment are target-specific - LLVM
6692 currently provides no portable way of determining them, so a front-end that
6693 generates this intrinsic needs to have some target-specific knowledge.
6694 The <tt>func</tt> argument must hold a function bitcast to
6695 an <tt>i8*</tt>.</p>
6698 <p>The block of memory pointed to by <tt>tramp</tt> is filled with target
6699 dependent code, turning it into a function. A pointer to this function is
6700 returned, but needs to be bitcast to an <a href="#int_trampoline">appropriate
6701 function pointer type</a> before being called. The new function's signature
6702 is the same as that of <tt>func</tt> with any arguments marked with
6703 the <tt>nest</tt> attribute removed. At most one such <tt>nest</tt> argument
6704 is allowed, and it must be of pointer type. Calling the new function is
6705 equivalent to calling <tt>func</tt> with the same argument list, but
6706 with <tt>nval</tt> used for the missing <tt>nest</tt> argument. If, after
6707 calling <tt>llvm.init.trampoline</tt>, the memory pointed to
6708 by <tt>tramp</tt> is modified, then the effect of any later call to the
6709 returned function pointer is undefined.</p>
6713 <!-- ======================================================================= -->
6714 <div class="doc_subsection">
6715 <a name="int_atomics">Atomic Operations and Synchronization Intrinsics</a>
6718 <div class="doc_text">
6720 <p>These intrinsic functions expand the "universal IR" of LLVM to represent
6721 hardware constructs for atomic operations and memory synchronization. This
6722 provides an interface to the hardware, not an interface to the programmer. It
6723 is aimed at a low enough level to allow any programming models or APIs
6724 (Application Programming Interfaces) which need atomic behaviors to map
6725 cleanly onto it. It is also modeled primarily on hardware behavior. Just as
6726 hardware provides a "universal IR" for source languages, it also provides a
6727 starting point for developing a "universal" atomic operation and
6728 synchronization IR.</p>
6730 <p>These do <em>not</em> form an API such as high-level threading libraries,
6731 software transaction memory systems, atomic primitives, and intrinsic
6732 functions as found in BSD, GNU libc, atomic_ops, APR, and other system and
6733 application libraries. The hardware interface provided by LLVM should allow
6734 a clean implementation of all of these APIs and parallel programming models.
6735 No one model or paradigm should be selected above others unless the hardware
6736 itself ubiquitously does so.</p>
6740 <!-- _______________________________________________________________________ -->
6741 <div class="doc_subsubsection">
6742 <a name="int_memory_barrier">'<tt>llvm.memory.barrier</tt>' Intrinsic</a>
6744 <div class="doc_text">
6747 declare void @llvm.memory.barrier( i1 <ll>, i1 <ls>, i1 <sl>, i1 <ss>, i1 <device> )
6751 <p>The <tt>llvm.memory.barrier</tt> intrinsic guarantees ordering between
6752 specific pairs of memory access types.</p>
6755 <p>The <tt>llvm.memory.barrier</tt> intrinsic requires five boolean arguments.
6756 The first four arguments enables a specific barrier as listed below. The
6757 fifth argument specifies that the barrier applies to io or device or uncached
6761 <li><tt>ll</tt>: load-load barrier</li>
6762 <li><tt>ls</tt>: load-store barrier</li>
6763 <li><tt>sl</tt>: store-load barrier</li>
6764 <li><tt>ss</tt>: store-store barrier</li>
6765 <li><tt>device</tt>: barrier applies to device and uncached memory also.</li>
6769 <p>This intrinsic causes the system to enforce some ordering constraints upon
6770 the loads and stores of the program. This barrier does not
6771 indicate <em>when</em> any events will occur, it only enforces
6772 an <em>order</em> in which they occur. For any of the specified pairs of load
6773 and store operations (f.ex. load-load, or store-load), all of the first
6774 operations preceding the barrier will complete before any of the second
6775 operations succeeding the barrier begin. Specifically the semantics for each
6776 pairing is as follows:</p>
6779 <li><tt>ll</tt>: All loads before the barrier must complete before any load
6780 after the barrier begins.</li>
6781 <li><tt>ls</tt>: All loads before the barrier must complete before any
6782 store after the barrier begins.</li>
6783 <li><tt>ss</tt>: All stores before the barrier must complete before any
6784 store after the barrier begins.</li>
6785 <li><tt>sl</tt>: All stores before the barrier must complete before any
6786 load after the barrier begins.</li>
6789 <p>These semantics are applied with a logical "and" behavior when more than one
6790 is enabled in a single memory barrier intrinsic.</p>
6792 <p>Backends may implement stronger barriers than those requested when they do
6793 not support as fine grained a barrier as requested. Some architectures do
6794 not need all types of barriers and on such architectures, these become
6799 %mallocP = tail call i8* @malloc(i32 ptrtoint (i32* getelementptr (i32* null, i32 1) to i32))
6800 %ptr = bitcast i8* %mallocP to i32*
6803 %result1 = load i32* %ptr <i>; yields {i32}:result1 = 4</i>
6804 call void @llvm.memory.barrier( i1 false, i1 true, i1 false, i1 false )
6805 <i>; guarantee the above finishes</i>
6806 store i32 8, %ptr <i>; before this begins</i>
6811 <!-- _______________________________________________________________________ -->
6812 <div class="doc_subsubsection">
6813 <a name="int_atomic_cmp_swap">'<tt>llvm.atomic.cmp.swap.*</tt>' Intrinsic</a>
6816 <div class="doc_text">
6819 <p>This is an overloaded intrinsic. You can use <tt>llvm.atomic.cmp.swap</tt> on
6820 any integer bit width and for different address spaces. Not all targets
6821 support all bit widths however.</p>
6824 declare i8 @llvm.atomic.cmp.swap.i8.p0i8( i8* <ptr>, i8 <cmp>, i8 <val> )
6825 declare i16 @llvm.atomic.cmp.swap.i16.p0i16( i16* <ptr>, i16 <cmp>, i16 <val> )
6826 declare i32 @llvm.atomic.cmp.swap.i32.p0i32( i32* <ptr>, i32 <cmp>, i32 <val> )
6827 declare i64 @llvm.atomic.cmp.swap.i64.p0i64( i64* <ptr>, i64 <cmp>, i64 <val> )
6831 <p>This loads a value in memory and compares it to a given value. If they are
6832 equal, it stores a new value into the memory.</p>
6835 <p>The <tt>llvm.atomic.cmp.swap</tt> intrinsic takes three arguments. The result
6836 as well as both <tt>cmp</tt> and <tt>val</tt> must be integer values with the
6837 same bit width. The <tt>ptr</tt> argument must be a pointer to a value of
6838 this integer type. While any bit width integer may be used, targets may only
6839 lower representations they support in hardware.</p>
6842 <p>This entire intrinsic must be executed atomically. It first loads the value
6843 in memory pointed to by <tt>ptr</tt> and compares it with the
6844 value <tt>cmp</tt>. If they are equal, <tt>val</tt> is stored into the
6845 memory. The loaded value is yielded in all cases. This provides the
6846 equivalent of an atomic compare-and-swap operation within the SSA
6851 %mallocP = tail call i8* @malloc(i32 ptrtoint (i32* getelementptr (i32* null, i32 1) to i32))
6852 %ptr = bitcast i8* %mallocP to i32*
6855 %val1 = add i32 4, 4
6856 %result1 = call i32 @llvm.atomic.cmp.swap.i32.p0i32( i32* %ptr, i32 4, %val1 )
6857 <i>; yields {i32}:result1 = 4</i>
6858 %stored1 = icmp eq i32 %result1, 4 <i>; yields {i1}:stored1 = true</i>
6859 %memval1 = load i32* %ptr <i>; yields {i32}:memval1 = 8</i>
6861 %val2 = add i32 1, 1
6862 %result2 = call i32 @llvm.atomic.cmp.swap.i32.p0i32( i32* %ptr, i32 5, %val2 )
6863 <i>; yields {i32}:result2 = 8</i>
6864 %stored2 = icmp eq i32 %result2, 5 <i>; yields {i1}:stored2 = false</i>
6866 %memval2 = load i32* %ptr <i>; yields {i32}:memval2 = 8</i>
6871 <!-- _______________________________________________________________________ -->
6872 <div class="doc_subsubsection">
6873 <a name="int_atomic_swap">'<tt>llvm.atomic.swap.*</tt>' Intrinsic</a>
6875 <div class="doc_text">
6878 <p>This is an overloaded intrinsic. You can use <tt>llvm.atomic.swap</tt> on any
6879 integer bit width. Not all targets support all bit widths however.</p>
6882 declare i8 @llvm.atomic.swap.i8.p0i8( i8* <ptr>, i8 <val> )
6883 declare i16 @llvm.atomic.swap.i16.p0i16( i16* <ptr>, i16 <val> )
6884 declare i32 @llvm.atomic.swap.i32.p0i32( i32* <ptr>, i32 <val> )
6885 declare i64 @llvm.atomic.swap.i64.p0i64( i64* <ptr>, i64 <val> )
6889 <p>This intrinsic loads the value stored in memory at <tt>ptr</tt> and yields
6890 the value from memory. It then stores the value in <tt>val</tt> in the memory
6891 at <tt>ptr</tt>.</p>
6894 <p>The <tt>llvm.atomic.swap</tt> intrinsic takes two arguments. Both
6895 the <tt>val</tt> argument and the result must be integers of the same bit
6896 width. The first argument, <tt>ptr</tt>, must be a pointer to a value of this
6897 integer type. The targets may only lower integer representations they
6901 <p>This intrinsic loads the value pointed to by <tt>ptr</tt>, yields it, and
6902 stores <tt>val</tt> back into <tt>ptr</tt> atomically. This provides the
6903 equivalent of an atomic swap operation within the SSA framework.</p>
6907 %mallocP = tail call i8* @malloc(i32 ptrtoint (i32* getelementptr (i32* null, i32 1) to i32))
6908 %ptr = bitcast i8* %mallocP to i32*
6911 %val1 = add i32 4, 4
6912 %result1 = call i32 @llvm.atomic.swap.i32.p0i32( i32* %ptr, i32 %val1 )
6913 <i>; yields {i32}:result1 = 4</i>
6914 %stored1 = icmp eq i32 %result1, 4 <i>; yields {i1}:stored1 = true</i>
6915 %memval1 = load i32* %ptr <i>; yields {i32}:memval1 = 8</i>
6917 %val2 = add i32 1, 1
6918 %result2 = call i32 @llvm.atomic.swap.i32.p0i32( i32* %ptr, i32 %val2 )
6919 <i>; yields {i32}:result2 = 8</i>
6921 %stored2 = icmp eq i32 %result2, 8 <i>; yields {i1}:stored2 = true</i>
6922 %memval2 = load i32* %ptr <i>; yields {i32}:memval2 = 2</i>
6927 <!-- _______________________________________________________________________ -->
6928 <div class="doc_subsubsection">
6929 <a name="int_atomic_load_add">'<tt>llvm.atomic.load.add.*</tt>' Intrinsic</a>
6933 <div class="doc_text">
6936 <p>This is an overloaded intrinsic. You can use <tt>llvm.atomic.load.add</tt> on
6937 any integer bit width. Not all targets support all bit widths however.</p>
6940 declare i8 @llvm.atomic.load.add.i8..p0i8( i8* <ptr>, i8 <delta> )
6941 declare i16 @llvm.atomic.load.add.i16..p0i16( i16* <ptr>, i16 <delta> )
6942 declare i32 @llvm.atomic.load.add.i32..p0i32( i32* <ptr>, i32 <delta> )
6943 declare i64 @llvm.atomic.load.add.i64..p0i64( i64* <ptr>, i64 <delta> )
6947 <p>This intrinsic adds <tt>delta</tt> to the value stored in memory
6948 at <tt>ptr</tt>. It yields the original value at <tt>ptr</tt>.</p>
6951 <p>The intrinsic takes two arguments, the first a pointer to an integer value
6952 and the second an integer value. The result is also an integer value. These
6953 integer types can have any bit width, but they must all have the same bit
6954 width. The targets may only lower integer representations they support.</p>
6957 <p>This intrinsic does a series of operations atomically. It first loads the
6958 value stored at <tt>ptr</tt>. It then adds <tt>delta</tt>, stores the result
6959 to <tt>ptr</tt>. It yields the original value stored at <tt>ptr</tt>.</p>
6963 %mallocP = tail call i8* @malloc(i32 ptrtoint (i32* getelementptr (i32* null, i32 1) to i32))
6964 %ptr = bitcast i8* %mallocP to i32*
6966 %result1 = call i32 @llvm.atomic.load.add.i32.p0i32( i32* %ptr, i32 4 )
6967 <i>; yields {i32}:result1 = 4</i>
6968 %result2 = call i32 @llvm.atomic.load.add.i32.p0i32( i32* %ptr, i32 2 )
6969 <i>; yields {i32}:result2 = 8</i>
6970 %result3 = call i32 @llvm.atomic.load.add.i32.p0i32( i32* %ptr, i32 5 )
6971 <i>; yields {i32}:result3 = 10</i>
6972 %memval1 = load i32* %ptr <i>; yields {i32}:memval1 = 15</i>
6977 <!-- _______________________________________________________________________ -->
6978 <div class="doc_subsubsection">
6979 <a name="int_atomic_load_sub">'<tt>llvm.atomic.load.sub.*</tt>' Intrinsic</a>
6983 <div class="doc_text">
6986 <p>This is an overloaded intrinsic. You can use <tt>llvm.atomic.load.sub</tt> on
6987 any integer bit width and for different address spaces. Not all targets
6988 support all bit widths however.</p>
6991 declare i8 @llvm.atomic.load.sub.i8.p0i32( i8* <ptr>, i8 <delta> )
6992 declare i16 @llvm.atomic.load.sub.i16.p0i32( i16* <ptr>, i16 <delta> )
6993 declare i32 @llvm.atomic.load.sub.i32.p0i32( i32* <ptr>, i32 <delta> )
6994 declare i64 @llvm.atomic.load.sub.i64.p0i32( i64* <ptr>, i64 <delta> )
6998 <p>This intrinsic subtracts <tt>delta</tt> to the value stored in memory at
6999 <tt>ptr</tt>. It yields the original value at <tt>ptr</tt>.</p>
7002 <p>The intrinsic takes two arguments, the first a pointer to an integer value
7003 and the second an integer value. The result is also an integer value. These
7004 integer types can have any bit width, but they must all have the same bit
7005 width. The targets may only lower integer representations they support.</p>
7008 <p>This intrinsic does a series of operations atomically. It first loads the
7009 value stored at <tt>ptr</tt>. It then subtracts <tt>delta</tt>, stores the
7010 result to <tt>ptr</tt>. It yields the original value stored
7011 at <tt>ptr</tt>.</p>
7015 %mallocP = tail call i8* @malloc(i32 ptrtoint (i32* getelementptr (i32* null, i32 1) to i32))
7016 %ptr = bitcast i8* %mallocP to i32*
7018 %result1 = call i32 @llvm.atomic.load.sub.i32.p0i32( i32* %ptr, i32 4 )
7019 <i>; yields {i32}:result1 = 8</i>
7020 %result2 = call i32 @llvm.atomic.load.sub.i32.p0i32( i32* %ptr, i32 2 )
7021 <i>; yields {i32}:result2 = 4</i>
7022 %result3 = call i32 @llvm.atomic.load.sub.i32.p0i32( i32* %ptr, i32 5 )
7023 <i>; yields {i32}:result3 = 2</i>
7024 %memval1 = load i32* %ptr <i>; yields {i32}:memval1 = -3</i>
7029 <!-- _______________________________________________________________________ -->
7030 <div class="doc_subsubsection">
7031 <a name="int_atomic_load_and">'<tt>llvm.atomic.load.and.*</tt>' Intrinsic</a><br>
7032 <a name="int_atomic_load_nand">'<tt>llvm.atomic.load.nand.*</tt>' Intrinsic</a><br>
7033 <a name="int_atomic_load_or">'<tt>llvm.atomic.load.or.*</tt>' Intrinsic</a><br>
7034 <a name="int_atomic_load_xor">'<tt>llvm.atomic.load.xor.*</tt>' Intrinsic</a><br>
7037 <div class="doc_text">
7040 <p>These are overloaded intrinsics. You can
7041 use <tt>llvm.atomic.load_and</tt>, <tt>llvm.atomic.load_nand</tt>,
7042 <tt>llvm.atomic.load_or</tt>, and <tt>llvm.atomic.load_xor</tt> on any integer
7043 bit width and for different address spaces. Not all targets support all bit
7047 declare i8 @llvm.atomic.load.and.i8.p0i8( i8* <ptr>, i8 <delta> )
7048 declare i16 @llvm.atomic.load.and.i16.p0i16( i16* <ptr>, i16 <delta> )
7049 declare i32 @llvm.atomic.load.and.i32.p0i32( i32* <ptr>, i32 <delta> )
7050 declare i64 @llvm.atomic.load.and.i64.p0i64( i64* <ptr>, i64 <delta> )
7054 declare i8 @llvm.atomic.load.or.i8.p0i8( i8* <ptr>, i8 <delta> )
7055 declare i16 @llvm.atomic.load.or.i16.p0i16( i16* <ptr>, i16 <delta> )
7056 declare i32 @llvm.atomic.load.or.i32.p0i32( i32* <ptr>, i32 <delta> )
7057 declare i64 @llvm.atomic.load.or.i64.p0i64( i64* <ptr>, i64 <delta> )
7061 declare i8 @llvm.atomic.load.nand.i8.p0i32( i8* <ptr>, i8 <delta> )
7062 declare i16 @llvm.atomic.load.nand.i16.p0i32( i16* <ptr>, i16 <delta> )
7063 declare i32 @llvm.atomic.load.nand.i32.p0i32( i32* <ptr>, i32 <delta> )
7064 declare i64 @llvm.atomic.load.nand.i64.p0i32( i64* <ptr>, i64 <delta> )
7068 declare i8 @llvm.atomic.load.xor.i8.p0i32( i8* <ptr>, i8 <delta> )
7069 declare i16 @llvm.atomic.load.xor.i16.p0i32( i16* <ptr>, i16 <delta> )
7070 declare i32 @llvm.atomic.load.xor.i32.p0i32( i32* <ptr>, i32 <delta> )
7071 declare i64 @llvm.atomic.load.xor.i64.p0i32( i64* <ptr>, i64 <delta> )
7075 <p>These intrinsics bitwise the operation (and, nand, or, xor) <tt>delta</tt> to
7076 the value stored in memory at <tt>ptr</tt>. It yields the original value
7077 at <tt>ptr</tt>.</p>
7080 <p>These intrinsics take two arguments, the first a pointer to an integer value
7081 and the second an integer value. The result is also an integer value. These
7082 integer types can have any bit width, but they must all have the same bit
7083 width. The targets may only lower integer representations they support.</p>
7086 <p>These intrinsics does a series of operations atomically. They first load the
7087 value stored at <tt>ptr</tt>. They then do the bitwise
7088 operation <tt>delta</tt>, store the result to <tt>ptr</tt>. They yield the
7089 original value stored at <tt>ptr</tt>.</p>
7093 %mallocP = tail call i8* @malloc(i32 ptrtoint (i32* getelementptr (i32* null, i32 1) to i32))
7094 %ptr = bitcast i8* %mallocP to i32*
7095 store i32 0x0F0F, %ptr
7096 %result0 = call i32 @llvm.atomic.load.nand.i32.p0i32( i32* %ptr, i32 0xFF )
7097 <i>; yields {i32}:result0 = 0x0F0F</i>
7098 %result1 = call i32 @llvm.atomic.load.and.i32.p0i32( i32* %ptr, i32 0xFF )
7099 <i>; yields {i32}:result1 = 0xFFFFFFF0</i>
7100 %result2 = call i32 @llvm.atomic.load.or.i32.p0i32( i32* %ptr, i32 0F )
7101 <i>; yields {i32}:result2 = 0xF0</i>
7102 %result3 = call i32 @llvm.atomic.load.xor.i32.p0i32( i32* %ptr, i32 0F )
7103 <i>; yields {i32}:result3 = FF</i>
7104 %memval1 = load i32* %ptr <i>; yields {i32}:memval1 = F0</i>
7109 <!-- _______________________________________________________________________ -->
7110 <div class="doc_subsubsection">
7111 <a name="int_atomic_load_max">'<tt>llvm.atomic.load.max.*</tt>' Intrinsic</a><br>
7112 <a name="int_atomic_load_min">'<tt>llvm.atomic.load.min.*</tt>' Intrinsic</a><br>
7113 <a name="int_atomic_load_umax">'<tt>llvm.atomic.load.umax.*</tt>' Intrinsic</a><br>
7114 <a name="int_atomic_load_umin">'<tt>llvm.atomic.load.umin.*</tt>' Intrinsic</a><br>
7117 <div class="doc_text">
7120 <p>These are overloaded intrinsics. You can use <tt>llvm.atomic.load_max</tt>,
7121 <tt>llvm.atomic.load_min</tt>, <tt>llvm.atomic.load_umax</tt>, and
7122 <tt>llvm.atomic.load_umin</tt> on any integer bit width and for different
7123 address spaces. Not all targets support all bit widths however.</p>
7126 declare i8 @llvm.atomic.load.max.i8.p0i8( i8* <ptr>, i8 <delta> )
7127 declare i16 @llvm.atomic.load.max.i16.p0i16( i16* <ptr>, i16 <delta> )
7128 declare i32 @llvm.atomic.load.max.i32.p0i32( i32* <ptr>, i32 <delta> )
7129 declare i64 @llvm.atomic.load.max.i64.p0i64( i64* <ptr>, i64 <delta> )
7133 declare i8 @llvm.atomic.load.min.i8.p0i8( i8* <ptr>, i8 <delta> )
7134 declare i16 @llvm.atomic.load.min.i16.p0i16( i16* <ptr>, i16 <delta> )
7135 declare i32 @llvm.atomic.load.min.i32..p0i32( i32* <ptr>, i32 <delta> )
7136 declare i64 @llvm.atomic.load.min.i64..p0i64( i64* <ptr>, i64 <delta> )
7140 declare i8 @llvm.atomic.load.umax.i8.p0i8( i8* <ptr>, i8 <delta> )
7141 declare i16 @llvm.atomic.load.umax.i16.p0i16( i16* <ptr>, i16 <delta> )
7142 declare i32 @llvm.atomic.load.umax.i32.p0i32( i32* <ptr>, i32 <delta> )
7143 declare i64 @llvm.atomic.load.umax.i64.p0i64( i64* <ptr>, i64 <delta> )
7147 declare i8 @llvm.atomic.load.umin.i8..p0i8( i8* <ptr>, i8 <delta> )
7148 declare i16 @llvm.atomic.load.umin.i16.p0i16( i16* <ptr>, i16 <delta> )
7149 declare i32 @llvm.atomic.load.umin.i32..p0i32( i32* <ptr>, i32 <delta> )
7150 declare i64 @llvm.atomic.load.umin.i64..p0i64( i64* <ptr>, i64 <delta> )
7154 <p>These intrinsics takes the signed or unsigned minimum or maximum of
7155 <tt>delta</tt> and the value stored in memory at <tt>ptr</tt>. It yields the
7156 original value at <tt>ptr</tt>.</p>
7159 <p>These intrinsics take two arguments, the first a pointer to an integer value
7160 and the second an integer value. The result is also an integer value. These
7161 integer types can have any bit width, but they must all have the same bit
7162 width. The targets may only lower integer representations they support.</p>
7165 <p>These intrinsics does a series of operations atomically. They first load the
7166 value stored at <tt>ptr</tt>. They then do the signed or unsigned min or
7167 max <tt>delta</tt> and the value, store the result to <tt>ptr</tt>. They
7168 yield the original value stored at <tt>ptr</tt>.</p>
7172 %mallocP = tail call i8* @malloc(i32 ptrtoint (i32* getelementptr (i32* null, i32 1) to i32))
7173 %ptr = bitcast i8* %mallocP to i32*
7175 %result0 = call i32 @llvm.atomic.load.min.i32.p0i32( i32* %ptr, i32 -2 )
7176 <i>; yields {i32}:result0 = 7</i>
7177 %result1 = call i32 @llvm.atomic.load.max.i32.p0i32( i32* %ptr, i32 8 )
7178 <i>; yields {i32}:result1 = -2</i>
7179 %result2 = call i32 @llvm.atomic.load.umin.i32.p0i32( i32* %ptr, i32 10 )
7180 <i>; yields {i32}:result2 = 8</i>
7181 %result3 = call i32 @llvm.atomic.load.umax.i32.p0i32( i32* %ptr, i32 30 )
7182 <i>; yields {i32}:result3 = 8</i>
7183 %memval1 = load i32* %ptr <i>; yields {i32}:memval1 = 30</i>
7189 <!-- ======================================================================= -->
7190 <div class="doc_subsection">
7191 <a name="int_memorymarkers">Memory Use Markers</a>
7194 <div class="doc_text">
7196 <p>This class of intrinsics exists to information about the lifetime of memory
7197 objects and ranges where variables are immutable.</p>
7201 <!-- _______________________________________________________________________ -->
7202 <div class="doc_subsubsection">
7203 <a name="int_lifetime_start">'<tt>llvm.lifetime.start</tt>' Intrinsic</a>
7206 <div class="doc_text">
7210 declare void @llvm.lifetime.start(i64 <size>, i8* nocapture <ptr>)
7214 <p>The '<tt>llvm.lifetime.start</tt>' intrinsic specifies the start of a memory
7215 object's lifetime.</p>
7218 <p>The first argument is a constant integer representing the size of the
7219 object, or -1 if it is variable sized. The second argument is a pointer to
7223 <p>This intrinsic indicates that before this point in the code, the value of the
7224 memory pointed to by <tt>ptr</tt> is dead. This means that it is known to
7225 never be used and has an undefined value. A load from the pointer that
7226 precedes this intrinsic can be replaced with
7227 <tt>'<a href="#undefvalues">undef</a>'</tt>.</p>
7231 <!-- _______________________________________________________________________ -->
7232 <div class="doc_subsubsection">
7233 <a name="int_lifetime_end">'<tt>llvm.lifetime.end</tt>' Intrinsic</a>
7236 <div class="doc_text">
7240 declare void @llvm.lifetime.end(i64 <size>, i8* nocapture <ptr>)
7244 <p>The '<tt>llvm.lifetime.end</tt>' intrinsic specifies the end of a memory
7245 object's lifetime.</p>
7248 <p>The first argument is a constant integer representing the size of the
7249 object, or -1 if it is variable sized. The second argument is a pointer to
7253 <p>This intrinsic indicates that after this point in the code, the value of the
7254 memory pointed to by <tt>ptr</tt> is dead. This means that it is known to
7255 never be used and has an undefined value. Any stores into the memory object
7256 following this intrinsic may be removed as dead.
7260 <!-- _______________________________________________________________________ -->
7261 <div class="doc_subsubsection">
7262 <a name="int_invariant_start">'<tt>llvm.invariant.start</tt>' Intrinsic</a>
7265 <div class="doc_text">
7269 declare {}* @llvm.invariant.start(i64 <size>, i8* nocapture <ptr>) readonly
7273 <p>The '<tt>llvm.invariant.start</tt>' intrinsic specifies that the contents of
7274 a memory object will not change.</p>
7277 <p>The first argument is a constant integer representing the size of the
7278 object, or -1 if it is variable sized. The second argument is a pointer to
7282 <p>This intrinsic indicates that until an <tt>llvm.invariant.end</tt> that uses
7283 the return value, the referenced memory location is constant and
7288 <!-- _______________________________________________________________________ -->
7289 <div class="doc_subsubsection">
7290 <a name="int_invariant_end">'<tt>llvm.invariant.end</tt>' Intrinsic</a>
7293 <div class="doc_text">
7297 declare void @llvm.invariant.end({}* <start>, i64 <size>, i8* nocapture <ptr>)
7301 <p>The '<tt>llvm.invariant.end</tt>' intrinsic specifies that the contents of
7302 a memory object are mutable.</p>
7305 <p>The first argument is the matching <tt>llvm.invariant.start</tt> intrinsic.
7306 The second argument is a constant integer representing the size of the
7307 object, or -1 if it is variable sized and the third argument is a pointer
7311 <p>This intrinsic indicates that the memory is mutable again.</p>
7315 <!-- ======================================================================= -->
7316 <div class="doc_subsection">
7317 <a name="int_general">General Intrinsics</a>
7320 <div class="doc_text">
7322 <p>This class of intrinsics is designed to be generic and has no specific
7327 <!-- _______________________________________________________________________ -->
7328 <div class="doc_subsubsection">
7329 <a name="int_var_annotation">'<tt>llvm.var.annotation</tt>' Intrinsic</a>
7332 <div class="doc_text">
7336 declare void @llvm.var.annotation(i8* <val>, i8* <str>, i8* <str>, i32 <int> )
7340 <p>The '<tt>llvm.var.annotation</tt>' intrinsic.</p>
7343 <p>The first argument is a pointer to a value, the second is a pointer to a
7344 global string, the third is a pointer to a global string which is the source
7345 file name, and the last argument is the line number.</p>
7348 <p>This intrinsic allows annotation of local variables with arbitrary strings.
7349 This can be useful for special purpose optimizations that want to look for
7350 these annotations. These have no other defined use, they are ignored by code
7351 generation and optimization.</p>
7355 <!-- _______________________________________________________________________ -->
7356 <div class="doc_subsubsection">
7357 <a name="int_annotation">'<tt>llvm.annotation.*</tt>' Intrinsic</a>
7360 <div class="doc_text">
7363 <p>This is an overloaded intrinsic. You can use '<tt>llvm.annotation</tt>' on
7364 any integer bit width.</p>
7367 declare i8 @llvm.annotation.i8(i8 <val>, i8* <str>, i8* <str>, i32 <int> )
7368 declare i16 @llvm.annotation.i16(i16 <val>, i8* <str>, i8* <str>, i32 <int> )
7369 declare i32 @llvm.annotation.i32(i32 <val>, i8* <str>, i8* <str>, i32 <int> )
7370 declare i64 @llvm.annotation.i64(i64 <val>, i8* <str>, i8* <str>, i32 <int> )
7371 declare i256 @llvm.annotation.i256(i256 <val>, i8* <str>, i8* <str>, i32 <int> )
7375 <p>The '<tt>llvm.annotation</tt>' intrinsic.</p>
7378 <p>The first argument is an integer value (result of some expression), the
7379 second is a pointer to a global string, the third is a pointer to a global
7380 string which is the source file name, and the last argument is the line
7381 number. It returns the value of the first argument.</p>
7384 <p>This intrinsic allows annotations to be put on arbitrary expressions with
7385 arbitrary strings. This can be useful for special purpose optimizations that
7386 want to look for these annotations. These have no other defined use, they
7387 are ignored by code generation and optimization.</p>
7391 <!-- _______________________________________________________________________ -->
7392 <div class="doc_subsubsection">
7393 <a name="int_trap">'<tt>llvm.trap</tt>' Intrinsic</a>
7396 <div class="doc_text">
7400 declare void @llvm.trap()
7404 <p>The '<tt>llvm.trap</tt>' intrinsic.</p>
7410 <p>This intrinsics is lowered to the target dependent trap instruction. If the
7411 target does not have a trap instruction, this intrinsic will be lowered to
7412 the call of the <tt>abort()</tt> function.</p>
7416 <!-- _______________________________________________________________________ -->
7417 <div class="doc_subsubsection">
7418 <a name="int_stackprotector">'<tt>llvm.stackprotector</tt>' Intrinsic</a>
7421 <div class="doc_text">
7425 declare void @llvm.stackprotector( i8* <guard>, i8** <slot> )
7429 <p>The <tt>llvm.stackprotector</tt> intrinsic takes the <tt>guard</tt> and
7430 stores it onto the stack at <tt>slot</tt>. The stack slot is adjusted to
7431 ensure that it is placed on the stack before local variables.</p>
7434 <p>The <tt>llvm.stackprotector</tt> intrinsic requires two pointer
7435 arguments. The first argument is the value loaded from the stack
7436 guard <tt>@__stack_chk_guard</tt>. The second variable is an <tt>alloca</tt>
7437 that has enough space to hold the value of the guard.</p>
7440 <p>This intrinsic causes the prologue/epilogue inserter to force the position of
7441 the <tt>AllocaInst</tt> stack slot to be before local variables on the
7442 stack. This is to ensure that if a local variable on the stack is
7443 overwritten, it will destroy the value of the guard. When the function exits,
7444 the guard on the stack is checked against the original guard. If they're
7445 different, then the program aborts by calling the <tt>__stack_chk_fail()</tt>
7450 <!-- _______________________________________________________________________ -->
7451 <div class="doc_subsubsection">
7452 <a name="int_objectsize">'<tt>llvm.objectsize</tt>' Intrinsic</a>
7455 <div class="doc_text">
7459 declare i32 @llvm.objectsize.i32( i8* <object>, i1 <type> )
7460 declare i64 @llvm.objectsize.i64( i8* <object>, i1 <type> )
7464 <p>The <tt>llvm.objectsize</tt> intrinsic is designed to provide information
7465 to the optimizers to discover at compile time either a) when an
7466 operation like memcpy will either overflow a buffer that corresponds to
7467 an object, or b) to determine that a runtime check for overflow isn't
7468 necessary. An object in this context means an allocation of a
7469 specific class, structure, array, or other object.</p>
7472 <p>The <tt>llvm.objectsize</tt> intrinsic takes two arguments. The first
7473 argument is a pointer to or into the <tt>object</tt>. The second argument
7474 is a boolean 0 or 1. This argument determines whether you want the
7475 maximum (0) or minimum (1) bytes remaining. This needs to be a literal 0 or
7476 1, variables are not allowed.</p>
7479 <p>The <tt>llvm.objectsize</tt> intrinsic is lowered to either a constant
7480 representing the size of the object concerned or <tt>i32/i64 -1 or 0</tt>
7481 (depending on the <tt>type</tt> argument if the size cannot be determined
7482 at compile time.</p>
7486 <!-- *********************************************************************** -->
7489 <a href="http://jigsaw.w3.org/css-validator/check/referer"><img
7490 src="http://jigsaw.w3.org/css-validator/images/vcss-blue" alt="Valid CSS"></a>
7491 <a href="http://validator.w3.org/check/referer"><img
7492 src="http://www.w3.org/Icons/valid-html401-blue" alt="Valid HTML 4.01"></a>
7494 <a href="mailto:sabre@nondot.org">Chris Lattner</a><br>
7495 <a href="http://llvm.org">The LLVM Compiler Infrastructure</a><br>
7496 Last modified: $Date$