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></li>
24 <li><a href="#callingconv">Calling Conventions</a></li>
25 <li><a href="#globalvars">Global Variables</a></li>
26 <li><a href="#functionstructure">Functions</a></li>
27 <li><a href="#aliasstructure">Aliases</a>
28 <li><a href="#paramattrs">Parameter Attributes</a></li>
29 <li><a href="#moduleasm">Module-Level Inline Assembly</a></li>
30 <li><a href="#datalayout">Data Layout</a></li>
33 <li><a href="#typesystem">Type System</a>
35 <li><a href="#t_primitive">Primitive Types</a>
37 <li><a href="#t_classifications">Type Classifications</a></li>
40 <li><a href="#t_derived">Derived Types</a>
42 <li><a href="#t_array">Array Type</a></li>
43 <li><a href="#t_function">Function Type</a></li>
44 <li><a href="#t_pointer">Pointer Type</a></li>
45 <li><a href="#t_struct">Structure Type</a></li>
46 <li><a href="#t_pstruct">Packed Structure Type</a></li>
47 <li><a href="#t_vector">Vector Type</a></li>
48 <li><a href="#t_opaque">Opaque Type</a></li>
53 <li><a href="#constants">Constants</a>
55 <li><a href="#simpleconstants">Simple Constants</a>
56 <li><a href="#aggregateconstants">Aggregate Constants</a>
57 <li><a href="#globalconstants">Global Variable and Function Addresses</a>
58 <li><a href="#undefvalues">Undefined Values</a>
59 <li><a href="#constantexprs">Constant Expressions</a>
62 <li><a href="#othervalues">Other Values</a>
64 <li><a href="#inlineasm">Inline Assembler Expressions</a>
67 <li><a href="#instref">Instruction Reference</a>
69 <li><a href="#terminators">Terminator Instructions</a>
71 <li><a href="#i_ret">'<tt>ret</tt>' Instruction</a></li>
72 <li><a href="#i_br">'<tt>br</tt>' Instruction</a></li>
73 <li><a href="#i_switch">'<tt>switch</tt>' Instruction</a></li>
74 <li><a href="#i_invoke">'<tt>invoke</tt>' Instruction</a></li>
75 <li><a href="#i_unwind">'<tt>unwind</tt>' Instruction</a></li>
76 <li><a href="#i_unreachable">'<tt>unreachable</tt>' Instruction</a></li>
79 <li><a href="#binaryops">Binary Operations</a>
81 <li><a href="#i_add">'<tt>add</tt>' Instruction</a></li>
82 <li><a href="#i_sub">'<tt>sub</tt>' Instruction</a></li>
83 <li><a href="#i_mul">'<tt>mul</tt>' Instruction</a></li>
84 <li><a href="#i_udiv">'<tt>udiv</tt>' Instruction</a></li>
85 <li><a href="#i_sdiv">'<tt>sdiv</tt>' Instruction</a></li>
86 <li><a href="#i_fdiv">'<tt>fdiv</tt>' Instruction</a></li>
87 <li><a href="#i_urem">'<tt>urem</tt>' Instruction</a></li>
88 <li><a href="#i_srem">'<tt>srem</tt>' Instruction</a></li>
89 <li><a href="#i_frem">'<tt>frem</tt>' Instruction</a></li>
92 <li><a href="#bitwiseops">Bitwise Binary Operations</a>
94 <li><a href="#i_shl">'<tt>shl</tt>' Instruction</a></li>
95 <li><a href="#i_lshr">'<tt>lshr</tt>' Instruction</a></li>
96 <li><a href="#i_ashr">'<tt>ashr</tt>' Instruction</a></li>
97 <li><a href="#i_and">'<tt>and</tt>' Instruction</a></li>
98 <li><a href="#i_or">'<tt>or</tt>' Instruction</a></li>
99 <li><a href="#i_xor">'<tt>xor</tt>' Instruction</a></li>
102 <li><a href="#vectorops">Vector Operations</a>
104 <li><a href="#i_extractelement">'<tt>extractelement</tt>' Instruction</a></li>
105 <li><a href="#i_insertelement">'<tt>insertelement</tt>' Instruction</a></li>
106 <li><a href="#i_shufflevector">'<tt>shufflevector</tt>' Instruction</a></li>
109 <li><a href="#memoryops">Memory Access and Addressing Operations</a>
111 <li><a href="#i_malloc">'<tt>malloc</tt>' Instruction</a></li>
112 <li><a href="#i_free">'<tt>free</tt>' Instruction</a></li>
113 <li><a href="#i_alloca">'<tt>alloca</tt>' Instruction</a></li>
114 <li><a href="#i_load">'<tt>load</tt>' Instruction</a></li>
115 <li><a href="#i_store">'<tt>store</tt>' Instruction</a></li>
116 <li><a href="#i_getelementptr">'<tt>getelementptr</tt>' Instruction</a></li>
119 <li><a href="#convertops">Conversion Operations</a>
121 <li><a href="#i_trunc">'<tt>trunc .. to</tt>' Instruction</a></li>
122 <li><a href="#i_zext">'<tt>zext .. to</tt>' Instruction</a></li>
123 <li><a href="#i_sext">'<tt>sext .. to</tt>' Instruction</a></li>
124 <li><a href="#i_fptrunc">'<tt>fptrunc .. to</tt>' Instruction</a></li>
125 <li><a href="#i_fpext">'<tt>fpext .. to</tt>' Instruction</a></li>
126 <li><a href="#i_fptoui">'<tt>fptoui .. to</tt>' Instruction</a></li>
127 <li><a href="#i_fptosi">'<tt>fptosi .. to</tt>' Instruction</a></li>
128 <li><a href="#i_uitofp">'<tt>uitofp .. to</tt>' Instruction</a></li>
129 <li><a href="#i_sitofp">'<tt>sitofp .. to</tt>' Instruction</a></li>
130 <li><a href="#i_ptrtoint">'<tt>ptrtoint .. to</tt>' Instruction</a></li>
131 <li><a href="#i_inttoptr">'<tt>inttoptr .. to</tt>' Instruction</a></li>
132 <li><a href="#i_bitcast">'<tt>bitcast .. to</tt>' Instruction</a></li>
134 <li><a href="#otherops">Other Operations</a>
136 <li><a href="#i_icmp">'<tt>icmp</tt>' Instruction</a></li>
137 <li><a href="#i_fcmp">'<tt>fcmp</tt>' Instruction</a></li>
138 <li><a href="#i_phi">'<tt>phi</tt>' Instruction</a></li>
139 <li><a href="#i_select">'<tt>select</tt>' Instruction</a></li>
140 <li><a href="#i_call">'<tt>call</tt>' Instruction</a></li>
141 <li><a href="#i_va_arg">'<tt>va_arg</tt>' Instruction</a></li>
146 <li><a href="#intrinsics">Intrinsic Functions</a>
148 <li><a href="#int_varargs">Variable Argument Handling Intrinsics</a>
150 <li><a href="#int_va_start">'<tt>llvm.va_start</tt>' Intrinsic</a></li>
151 <li><a href="#int_va_end">'<tt>llvm.va_end</tt>' Intrinsic</a></li>
152 <li><a href="#int_va_copy">'<tt>llvm.va_copy</tt>' Intrinsic</a></li>
155 <li><a href="#int_gc">Accurate Garbage Collection Intrinsics</a>
157 <li><a href="#int_gcroot">'<tt>llvm.gcroot</tt>' Intrinsic</a></li>
158 <li><a href="#int_gcread">'<tt>llvm.gcread</tt>' Intrinsic</a></li>
159 <li><a href="#int_gcwrite">'<tt>llvm.gcwrite</tt>' Intrinsic</a></li>
162 <li><a href="#int_codegen">Code Generator Intrinsics</a>
164 <li><a href="#int_returnaddress">'<tt>llvm.returnaddress</tt>' Intrinsic</a></li>
165 <li><a href="#int_frameaddress">'<tt>llvm.frameaddress</tt>' Intrinsic</a></li>
166 <li><a href="#int_stacksave">'<tt>llvm.stacksave</tt>' Intrinsic</a></li>
167 <li><a href="#int_stackrestore">'<tt>llvm.stackrestore</tt>' Intrinsic</a></li>
168 <li><a href="#int_prefetch">'<tt>llvm.prefetch</tt>' Intrinsic</a></li>
169 <li><a href="#int_pcmarker">'<tt>llvm.pcmarker</tt>' Intrinsic</a></li>
170 <li><a href="#int_readcyclecounter"><tt>llvm.readcyclecounter</tt>' Intrinsic</a></li>
173 <li><a href="#int_libc">Standard C Library Intrinsics</a>
175 <li><a href="#int_memcpy">'<tt>llvm.memcpy.*</tt>' Intrinsic</a></li>
176 <li><a href="#int_memmove">'<tt>llvm.memmove.*</tt>' Intrinsic</a></li>
177 <li><a href="#int_memset">'<tt>llvm.memset.*</tt>' Intrinsic</a></li>
178 <li><a href="#int_sqrt">'<tt>llvm.sqrt.*</tt>' Intrinsic</a></li>
179 <li><a href="#int_powi">'<tt>llvm.powi.*</tt>' Intrinsic</a></li>
182 <li><a href="#int_manip">Bit Manipulation Intrinsics</a>
184 <li><a href="#int_bswap">'<tt>llvm.bswap.*</tt>' Intrinsics</a></li>
185 <li><a href="#int_ctpop">'<tt>llvm.ctpop.*</tt>' Intrinsic </a></li>
186 <li><a href="#int_ctlz">'<tt>llvm.ctlz.*</tt>' Intrinsic </a></li>
187 <li><a href="#int_cttz">'<tt>llvm.cttz.*</tt>' Intrinsic </a></li>
188 <li><a href="#int_part_select">'<tt>llvm.part.select.*</tt>' Intrinsic </a></li>
189 <li><a href="#int_part_set">'<tt>llvm.part.set.*</tt>' Intrinsic </a></li>
192 <li><a href="#int_debugger">Debugger intrinsics</a></li>
193 <li><a href="#int_eh">Exception Handling intrinsics</a></li>
194 <li><a href="#int_atomics">Atomic Operations and Synchronization Intrinsics</a>
196 <li><a href="#int_lcs">'<tt>llvm.atomic.lcs.*</tt>' Intrinsic</a></li>
197 <li><a href="#int_ls">'<tt>llvm.atomic.ls.*</tt>' Intrinsic</a></li>
198 <li><a href="#int_las">'<tt>llvm.atomic.las.*</tt>' Intrinsic</a></li>
199 <li><a href="#int_lss">'<tt>llvm.atomic.lss.*</tt>' Intrinsic</a></li>
200 <li><a href="#int_memory_barrier">'<tt>llvm.memory.barrier</tt>' Intrinsic</a></li>
203 <li><a href="#int_trampoline">Trampoline Intrinsics</a>
205 <li><a href="#int_it">'<tt>llvm.init.trampoline</tt>' Intrinsic</a></li>
206 <li><a href="#int_at">'<tt>llvm.adjust.trampoline</tt>' Intrinsic</a></li>
209 <li><a href="#int_general">General intrinsics</a>
211 <li><a href="#int_var_annotation">
212 <tt>llvm.var.annotation</tt>' Intrinsic</a></li>
219 <div class="doc_author">
220 <p>Written by <a href="mailto:sabre@nondot.org">Chris Lattner</a>
221 and <a href="mailto:vadve@cs.uiuc.edu">Vikram Adve</a></p>
224 <!-- *********************************************************************** -->
225 <div class="doc_section"> <a name="abstract">Abstract </a></div>
226 <!-- *********************************************************************** -->
228 <div class="doc_text">
229 <p>This document is a reference manual for the LLVM assembly language.
230 LLVM is an SSA based representation that provides type safety,
231 low-level operations, flexibility, and the capability of representing
232 'all' high-level languages cleanly. It is the common code
233 representation used throughout all phases of the LLVM compilation
237 <!-- *********************************************************************** -->
238 <div class="doc_section"> <a name="introduction">Introduction</a> </div>
239 <!-- *********************************************************************** -->
241 <div class="doc_text">
243 <p>The LLVM code representation is designed to be used in three
244 different forms: as an in-memory compiler IR, as an on-disk bitcode
245 representation (suitable for fast loading by a Just-In-Time compiler),
246 and as a human readable assembly language representation. This allows
247 LLVM to provide a powerful intermediate representation for efficient
248 compiler transformations and analysis, while providing a natural means
249 to debug and visualize the transformations. The three different forms
250 of LLVM are all equivalent. This document describes the human readable
251 representation and notation.</p>
253 <p>The LLVM representation aims to be light-weight and low-level
254 while being expressive, typed, and extensible at the same time. It
255 aims to be a "universal IR" of sorts, by being at a low enough level
256 that high-level ideas may be cleanly mapped to it (similar to how
257 microprocessors are "universal IR's", allowing many source languages to
258 be mapped to them). By providing type information, LLVM can be used as
259 the target of optimizations: for example, through pointer analysis, it
260 can be proven that a C automatic variable is never accessed outside of
261 the current function... allowing it to be promoted to a simple SSA
262 value instead of a memory location.</p>
266 <!-- _______________________________________________________________________ -->
267 <div class="doc_subsubsection"> <a name="wellformed">Well-Formedness</a> </div>
269 <div class="doc_text">
271 <p>It is important to note that this document describes 'well formed'
272 LLVM assembly language. There is a difference between what the parser
273 accepts and what is considered 'well formed'. For example, the
274 following instruction is syntactically okay, but not well formed:</p>
276 <div class="doc_code">
278 %x = <a href="#i_add">add</a> i32 1, %x
282 <p>...because the definition of <tt>%x</tt> does not dominate all of
283 its uses. The LLVM infrastructure provides a verification pass that may
284 be used to verify that an LLVM module is well formed. This pass is
285 automatically run by the parser after parsing input assembly and by
286 the optimizer before it outputs bitcode. The violations pointed out
287 by the verifier pass indicate bugs in transformation passes or input to
291 <!-- Describe the typesetting conventions here. -->
293 <!-- *********************************************************************** -->
294 <div class="doc_section"> <a name="identifiers">Identifiers</a> </div>
295 <!-- *********************************************************************** -->
297 <div class="doc_text">
299 <p>LLVM uses three different forms of identifiers, for different
303 <li>Named values are represented as a string of characters with a '%' prefix.
304 For example, %foo, %DivisionByZero, %a.really.long.identifier. The actual
305 regular expression used is '<tt>%[a-zA-Z$._][a-zA-Z$._0-9]*</tt>'.
306 Identifiers which require other characters in their names can be surrounded
307 with quotes. In this way, anything except a <tt>"</tt> character can be used
310 <li>Unnamed values are represented as an unsigned numeric value with a '%'
311 prefix. For example, %12, %2, %44.</li>
313 <li>Constants, which are described in a <a href="#constants">section about
314 constants</a>, below.</li>
317 <p>LLVM requires that values start with a '%' sign for two reasons: Compilers
318 don't need to worry about name clashes with reserved words, and the set of
319 reserved words may be expanded in the future without penalty. Additionally,
320 unnamed identifiers allow a compiler to quickly come up with a temporary
321 variable without having to avoid symbol table conflicts.</p>
323 <p>Reserved words in LLVM are very similar to reserved words in other
324 languages. There are keywords for different opcodes
325 ('<tt><a href="#i_add">add</a></tt>',
326 '<tt><a href="#i_bitcast">bitcast</a></tt>',
327 '<tt><a href="#i_ret">ret</a></tt>', etc...), for primitive type names ('<tt><a
328 href="#t_void">void</a></tt>', '<tt><a href="#t_primitive">i32</a></tt>', etc...),
329 and others. These reserved words cannot conflict with variable names, because
330 none of them start with a '%' character.</p>
332 <p>Here is an example of LLVM code to multiply the integer variable
333 '<tt>%X</tt>' by 8:</p>
337 <div class="doc_code">
339 %result = <a href="#i_mul">mul</a> i32 %X, 8
343 <p>After strength reduction:</p>
345 <div class="doc_code">
347 %result = <a href="#i_shl">shl</a> i32 %X, i8 3
351 <p>And the hard way:</p>
353 <div class="doc_code">
355 <a href="#i_add">add</a> i32 %X, %X <i>; yields {i32}:%0</i>
356 <a href="#i_add">add</a> i32 %0, %0 <i>; yields {i32}:%1</i>
357 %result = <a href="#i_add">add</a> i32 %1, %1
361 <p>This last way of multiplying <tt>%X</tt> by 8 illustrates several
362 important lexical features of LLVM:</p>
366 <li>Comments are delimited with a '<tt>;</tt>' and go until the end of
369 <li>Unnamed temporaries are created when the result of a computation is not
370 assigned to a named value.</li>
372 <li>Unnamed temporaries are numbered sequentially</li>
376 <p>...and it also shows a convention that we follow in this document. When
377 demonstrating instructions, we will follow an instruction with a comment that
378 defines the type and name of value produced. Comments are shown in italic
383 <!-- *********************************************************************** -->
384 <div class="doc_section"> <a name="highlevel">High Level Structure</a> </div>
385 <!-- *********************************************************************** -->
387 <!-- ======================================================================= -->
388 <div class="doc_subsection"> <a name="modulestructure">Module Structure</a>
391 <div class="doc_text">
393 <p>LLVM programs are composed of "Module"s, each of which is a
394 translation unit of the input programs. Each module consists of
395 functions, global variables, and symbol table entries. Modules may be
396 combined together with the LLVM linker, which merges function (and
397 global variable) definitions, resolves forward declarations, and merges
398 symbol table entries. Here is an example of the "hello world" module:</p>
400 <div class="doc_code">
401 <pre><i>; Declare the string constant as a global constant...</i>
402 <a href="#identifiers">@.LC0</a> = <a href="#linkage_internal">internal</a> <a
403 href="#globalvars">constant</a> <a href="#t_array">[13 x i8]</a> c"hello world\0A\00" <i>; [13 x i8]*</i>
405 <i>; External declaration of the puts function</i>
406 <a href="#functionstructure">declare</a> i32 @puts(i8 *) <i>; i32(i8 *)* </i>
408 <i>; Definition of main function</i>
409 define i32 @main() { <i>; i32()* </i>
410 <i>; Convert [13x i8 ]* to i8 *...</i>
412 href="#i_getelementptr">getelementptr</a> [13 x i8 ]* @.LC0, i64 0, i64 0 <i>; i8 *</i>
414 <i>; Call puts function to write out the string to stdout...</i>
416 href="#i_call">call</a> i32 @puts(i8 * %cast210) <i>; i32</i>
418 href="#i_ret">ret</a> i32 0<br>}<br>
422 <p>This example is made up of a <a href="#globalvars">global variable</a>
423 named "<tt>.LC0</tt>", an external declaration of the "<tt>puts</tt>"
424 function, and a <a href="#functionstructure">function definition</a>
425 for "<tt>main</tt>".</p>
427 <p>In general, a module is made up of a list of global values,
428 where both functions and global variables are global values. Global values are
429 represented by a pointer to a memory location (in this case, a pointer to an
430 array of char, and a pointer to a function), and have one of the following <a
431 href="#linkage">linkage types</a>.</p>
435 <!-- ======================================================================= -->
436 <div class="doc_subsection">
437 <a name="linkage">Linkage Types</a>
440 <div class="doc_text">
443 All Global Variables and Functions have one of the following types of linkage:
448 <dt><tt><b><a name="linkage_internal">internal</a></b></tt> </dt>
450 <dd>Global values with internal linkage are only directly accessible by
451 objects in the current module. In particular, linking code into a module with
452 an internal global value may cause the internal to be renamed as necessary to
453 avoid collisions. Because the symbol is internal to the module, all
454 references can be updated. This corresponds to the notion of the
455 '<tt>static</tt>' keyword in C.
458 <dt><tt><b><a name="linkage_linkonce">linkonce</a></b></tt>: </dt>
460 <dd>Globals with "<tt>linkonce</tt>" linkage are merged with other globals of
461 the same name when linkage occurs. This is typically used to implement
462 inline functions, templates, or other code which must be generated in each
463 translation unit that uses it. Unreferenced <tt>linkonce</tt> globals are
464 allowed to be discarded.
467 <dt><tt><b><a name="linkage_weak">weak</a></b></tt>: </dt>
469 <dd>"<tt>weak</tt>" linkage is exactly the same as <tt>linkonce</tt> linkage,
470 except that unreferenced <tt>weak</tt> globals may not be discarded. This is
471 used for globals that may be emitted in multiple translation units, but that
472 are not guaranteed to be emitted into every translation unit that uses them.
473 One example of this are common globals in C, such as "<tt>int X;</tt>" at
477 <dt><tt><b><a name="linkage_appending">appending</a></b></tt>: </dt>
479 <dd>"<tt>appending</tt>" linkage may only be applied to global variables of
480 pointer to array type. When two global variables with appending linkage are
481 linked together, the two global arrays are appended together. This is the
482 LLVM, typesafe, equivalent of having the system linker append together
483 "sections" with identical names when .o files are linked.
486 <dt><tt><b><a name="linkage_externweak">extern_weak</a></b></tt>: </dt>
487 <dd>The semantics of this linkage follow the ELF model: the symbol is weak
488 until linked, if not linked, the symbol becomes null instead of being an
492 <dt><tt><b><a name="linkage_external">externally visible</a></b></tt>:</dt>
494 <dd>If none of the above identifiers are used, the global is externally
495 visible, meaning that it participates in linkage and can be used to resolve
496 external symbol references.
501 The next two types of linkage are targeted for Microsoft Windows platform
502 only. They are designed to support importing (exporting) symbols from (to)
507 <dt><tt><b><a name="linkage_dllimport">dllimport</a></b></tt>: </dt>
509 <dd>"<tt>dllimport</tt>" linkage causes the compiler to reference a function
510 or variable via a global pointer to a pointer that is set up by the DLL
511 exporting the symbol. On Microsoft Windows targets, the pointer name is
512 formed by combining <code>_imp__</code> and the function or variable name.
515 <dt><tt><b><a name="linkage_dllexport">dllexport</a></b></tt>: </dt>
517 <dd>"<tt>dllexport</tt>" linkage causes the compiler to provide a global
518 pointer to a pointer in a DLL, so that it can be referenced with the
519 <tt>dllimport</tt> attribute. On Microsoft Windows targets, the pointer
520 name is formed by combining <code>_imp__</code> and the function or variable
526 <p><a name="linkage_external"></a>For example, since the "<tt>.LC0</tt>"
527 variable is defined to be internal, if another module defined a "<tt>.LC0</tt>"
528 variable and was linked with this one, one of the two would be renamed,
529 preventing a collision. Since "<tt>main</tt>" and "<tt>puts</tt>" are
530 external (i.e., lacking any linkage declarations), they are accessible
531 outside of the current module.</p>
532 <p>It is illegal for a function <i>declaration</i>
533 to have any linkage type other than "externally visible", <tt>dllimport</tt>,
534 or <tt>extern_weak</tt>.</p>
535 <p>Aliases can have only <tt>external</tt>, <tt>internal</tt> and <tt>weak</tt>
539 <!-- ======================================================================= -->
540 <div class="doc_subsection">
541 <a name="callingconv">Calling Conventions</a>
544 <div class="doc_text">
546 <p>LLVM <a href="#functionstructure">functions</a>, <a href="#i_call">calls</a>
547 and <a href="#i_invoke">invokes</a> can all have an optional calling convention
548 specified for the call. The calling convention of any pair of dynamic
549 caller/callee must match, or the behavior of the program is undefined. The
550 following calling conventions are supported by LLVM, and more may be added in
554 <dt><b>"<tt>ccc</tt>" - The C calling convention</b>:</dt>
556 <dd>This calling convention (the default if no other calling convention is
557 specified) matches the target C calling conventions. This calling convention
558 supports varargs function calls and tolerates some mismatch in the declared
559 prototype and implemented declaration of the function (as does normal C).
562 <dt><b>"<tt>fastcc</tt>" - The fast calling convention</b>:</dt>
564 <dd>This calling convention attempts to make calls as fast as possible
565 (e.g. by passing things in registers). This calling convention allows the
566 target to use whatever tricks it wants to produce fast code for the target,
567 without having to conform to an externally specified ABI. Implementations of
568 this convention should allow arbitrary tail call optimization to be supported.
569 This calling convention does not support varargs and requires the prototype of
570 all callees to exactly match the prototype of the function definition.
573 <dt><b>"<tt>coldcc</tt>" - The cold calling convention</b>:</dt>
575 <dd>This calling convention attempts to make code in the caller as efficient
576 as possible under the assumption that the call is not commonly executed. As
577 such, these calls often preserve all registers so that the call does not break
578 any live ranges in the caller side. This calling convention does not support
579 varargs and requires the prototype of all callees to exactly match the
580 prototype of the function definition.
583 <dt><b>"<tt>cc <<em>n</em>></tt>" - Numbered convention</b>:</dt>
585 <dd>Any calling convention may be specified by number, allowing
586 target-specific calling conventions to be used. Target specific calling
587 conventions start at 64.
591 <p>More calling conventions can be added/defined on an as-needed basis, to
592 support pascal conventions or any other well-known target-independent
597 <!-- ======================================================================= -->
598 <div class="doc_subsection">
599 <a name="visibility">Visibility Styles</a>
602 <div class="doc_text">
605 All Global Variables and Functions have one of the following visibility styles:
609 <dt><b>"<tt>default</tt>" - Default style</b>:</dt>
611 <dd>On ELF, default visibility means that the declaration is visible to other
612 modules and, in shared libraries, means that the declared entity may be
613 overridden. On Darwin, default visibility means that the declaration is
614 visible to other modules. Default visibility corresponds to "external
615 linkage" in the language.
618 <dt><b>"<tt>hidden</tt>" - Hidden style</b>:</dt>
620 <dd>Two declarations of an object with hidden visibility refer to the same
621 object if they are in the same shared object. Usually, hidden visibility
622 indicates that the symbol will not be placed into the dynamic symbol table,
623 so no other module (executable or shared library) can reference it
627 <dt><b>"<tt>protected</tt>" - Protected style</b>:</dt>
629 <dd>On ELF, protected visibility indicates that the symbol will be placed in
630 the dynamic symbol table, but that references within the defining module will
631 bind to the local symbol. That is, the symbol cannot be overridden by another
638 <!-- ======================================================================= -->
639 <div class="doc_subsection">
640 <a name="globalvars">Global Variables</a>
643 <div class="doc_text">
645 <p>Global variables define regions of memory allocated at compilation time
646 instead of run-time. Global variables may optionally be initialized, may have
647 an explicit section to be placed in, and may have an optional explicit alignment
648 specified. A variable may be defined as "thread_local", which means that it
649 will not be shared by threads (each thread will have a separated copy of the
650 variable). A variable may be defined as a global "constant," which indicates
651 that the contents of the variable will <b>never</b> be modified (enabling better
652 optimization, allowing the global data to be placed in the read-only section of
653 an executable, etc). Note that variables that need runtime initialization
654 cannot be marked "constant" as there is a store to the variable.</p>
657 LLVM explicitly allows <em>declarations</em> of global variables to be marked
658 constant, even if the final definition of the global is not. This capability
659 can be used to enable slightly better optimization of the program, but requires
660 the language definition to guarantee that optimizations based on the
661 'constantness' are valid for the translation units that do not include the
665 <p>As SSA values, global variables define pointer values that are in
666 scope (i.e. they dominate) all basic blocks in the program. Global
667 variables always define a pointer to their "content" type because they
668 describe a region of memory, and all memory objects in LLVM are
669 accessed through pointers.</p>
671 <p>LLVM allows an explicit section to be specified for globals. If the target
672 supports it, it will emit globals to the section specified.</p>
674 <p>An explicit alignment may be specified for a global. If not present, or if
675 the alignment is set to zero, the alignment of the global is set by the target
676 to whatever it feels convenient. If an explicit alignment is specified, the
677 global is forced to have at least that much alignment. All alignments must be
680 <p>For example, the following defines a global with an initializer, section,
683 <div class="doc_code">
685 @G = constant float 1.0, section "foo", align 4
692 <!-- ======================================================================= -->
693 <div class="doc_subsection">
694 <a name="functionstructure">Functions</a>
697 <div class="doc_text">
699 <p>LLVM function definitions consist of the "<tt>define</tt>" keyord,
700 an optional <a href="#linkage">linkage type</a>, an optional
701 <a href="#visibility">visibility style</a>, an optional
702 <a href="#callingconv">calling convention</a>, a return type, an optional
703 <a href="#paramattrs">parameter attribute</a> for the return type, a function
704 name, a (possibly empty) argument list (each with optional
705 <a href="#paramattrs">parameter attributes</a>), an optional section, an
706 optional alignment, an opening curly brace, a list of basic blocks, and a
709 LLVM function declarations consist of the "<tt>declare</tt>" keyword, an
710 optional <a href="#linkage">linkage type</a>, an optional
711 <a href="#visibility">visibility style</a>, an optional
712 <a href="#callingconv">calling convention</a>, a return type, an optional
713 <a href="#paramattrs">parameter attribute</a> for the return type, a function
714 name, a possibly empty list of arguments, and an optional alignment.</p>
716 <p>A function definition contains a list of basic blocks, forming the CFG for
717 the function. Each basic block may optionally start with a label (giving the
718 basic block a symbol table entry), contains a list of instructions, and ends
719 with a <a href="#terminators">terminator</a> instruction (such as a branch or
720 function return).</p>
722 <p>The first basic block in a function is special in two ways: it is immediately
723 executed on entrance to the function, and it is not allowed to have predecessor
724 basic blocks (i.e. there can not be any branches to the entry block of a
725 function). Because the block can have no predecessors, it also cannot have any
726 <a href="#i_phi">PHI nodes</a>.</p>
728 <p>LLVM allows an explicit section to be specified for functions. If the target
729 supports it, it will emit functions to the section specified.</p>
731 <p>An explicit alignment may be specified for a function. If not present, or if
732 the alignment is set to zero, the alignment of the function is set by the target
733 to whatever it feels convenient. If an explicit alignment is specified, the
734 function is forced to have at least that much alignment. All alignments must be
740 <!-- ======================================================================= -->
741 <div class="doc_subsection">
742 <a name="aliasstructure">Aliases</a>
744 <div class="doc_text">
745 <p>Aliases act as "second name" for the aliasee value (which can be either
746 function or global variable or bitcast of global value). Aliases may have an
747 optional <a href="#linkage">linkage type</a>, and an
748 optional <a href="#visibility">visibility style</a>.</p>
752 <div class="doc_code">
754 @<Name> = [Linkage] [Visibility] alias <AliaseeTy> @<Aliasee>
762 <!-- ======================================================================= -->
763 <div class="doc_subsection"><a name="paramattrs">Parameter Attributes</a></div>
764 <div class="doc_text">
765 <p>The return type and each parameter of a function type may have a set of
766 <i>parameter attributes</i> associated with them. Parameter attributes are
767 used to communicate additional information about the result or parameters of
768 a function. Parameter attributes are considered to be part of the function
769 type so two functions types that differ only by the parameter attributes
770 are different function types.</p>
772 <p>Parameter attributes are simple keywords that follow the type specified. If
773 multiple parameter attributes are needed, they are space separated. For
776 <div class="doc_code">
778 %someFunc = i16 (i8 signext %someParam) zeroext
779 %someFunc = i16 (i8 zeroext %someParam) zeroext
783 <p>Note that the two function types above are unique because the parameter has
784 a different attribute (<tt>signext</tt> in the first one, <tt>zeroext</tt> in
785 the second). Also note that the attribute for the function result
786 (<tt>zeroext</tt>) comes immediately after the argument list.</p>
788 <p>Currently, only the following parameter attributes are defined:</p>
790 <dt><tt>zeroext</tt></dt>
791 <dd>This indicates that the parameter should be zero extended just before
792 a call to this function.</dd>
793 <dt><tt>signext</tt></dt>
794 <dd>This indicates that the parameter should be sign extended just before
795 a call to this function.</dd>
796 <dt><tt>inreg</tt></dt>
797 <dd>This indicates that the parameter should be placed in register (if
798 possible) during assembling function call. Support for this attribute is
800 <dt><tt>sret</tt></dt>
801 <dd>This indicates that the parameter specifies the address of a structure
802 that is the return value of the function in the source program.</dd>
803 <dt><tt>noalias</tt></dt>
804 <dd>This indicates that the parameter not alias any other object or any
805 other "noalias" objects during the function call.
806 <dt><tt>noreturn</tt></dt>
807 <dd>This function attribute indicates that the function never returns. This
808 indicates to LLVM that every call to this function should be treated as if
809 an <tt>unreachable</tt> instruction immediately followed the call.</dd>
810 <dt><tt>nounwind</tt></dt>
811 <dd>This function attribute indicates that the function type does not use
812 the unwind instruction and does not allow stack unwinding to propagate
818 <!-- ======================================================================= -->
819 <div class="doc_subsection">
820 <a name="moduleasm">Module-Level Inline Assembly</a>
823 <div class="doc_text">
825 Modules may contain "module-level inline asm" blocks, which corresponds to the
826 GCC "file scope inline asm" blocks. These blocks are internally concatenated by
827 LLVM and treated as a single unit, but may be separated in the .ll file if
828 desired. The syntax is very simple:
831 <div class="doc_code">
833 module asm "inline asm code goes here"
834 module asm "more can go here"
838 <p>The strings can contain any character by escaping non-printable characters.
839 The escape sequence used is simply "\xx" where "xx" is the two digit hex code
844 The inline asm code is simply printed to the machine code .s file when
845 assembly code is generated.
849 <!-- ======================================================================= -->
850 <div class="doc_subsection">
851 <a name="datalayout">Data Layout</a>
854 <div class="doc_text">
855 <p>A module may specify a target specific data layout string that specifies how
856 data is to be laid out in memory. The syntax for the data layout is simply:</p>
857 <pre> target datalayout = "<i>layout specification</i>"</pre>
858 <p>The <i>layout specification</i> consists of a list of specifications
859 separated by the minus sign character ('-'). Each specification starts with a
860 letter and may include other information after the letter to define some
861 aspect of the data layout. The specifications accepted are as follows: </p>
864 <dd>Specifies that the target lays out data in big-endian form. That is, the
865 bits with the most significance have the lowest address location.</dd>
867 <dd>Specifies that hte target lays out data in little-endian form. That is,
868 the bits with the least significance have the lowest address location.</dd>
869 <dt><tt>p:<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
870 <dd>This specifies the <i>size</i> of a pointer and its <i>abi</i> and
871 <i>preferred</i> alignments. All sizes are in bits. Specifying the <i>pref</i>
872 alignment is optional. If omitted, the preceding <tt>:</tt> should be omitted
874 <dt><tt>i<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
875 <dd>This specifies the alignment for an integer type of a given bit
876 <i>size</i>. The value of <i>size</i> must be in the range [1,2^23).</dd>
877 <dt><tt>v<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
878 <dd>This specifies the alignment for a vector type of a given bit
880 <dt><tt>f<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
881 <dd>This specifies the alignment for a floating point type of a given bit
882 <i>size</i>. The value of <i>size</i> must be either 32 (float) or 64
884 <dt><tt>a<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
885 <dd>This specifies the alignment for an aggregate type of a given bit
888 <p>When constructing the data layout for a given target, LLVM starts with a
889 default set of specifications which are then (possibly) overriden by the
890 specifications in the <tt>datalayout</tt> keyword. The default specifications
891 are given in this list:</p>
893 <li><tt>E</tt> - big endian</li>
894 <li><tt>p:32:64:64</tt> - 32-bit pointers with 64-bit alignment</li>
895 <li><tt>i1:8:8</tt> - i1 is 8-bit (byte) aligned</li>
896 <li><tt>i8:8:8</tt> - i8 is 8-bit (byte) aligned</li>
897 <li><tt>i16:16:16</tt> - i16 is 16-bit aligned</li>
898 <li><tt>i32:32:32</tt> - i32 is 32-bit aligned</li>
899 <li><tt>i64:32:64</tt> - i64 has abi alignment of 32-bits but preferred
900 alignment of 64-bits</li>
901 <li><tt>f32:32:32</tt> - float is 32-bit aligned</li>
902 <li><tt>f64:64:64</tt> - double is 64-bit aligned</li>
903 <li><tt>v64:64:64</tt> - 64-bit vector is 64-bit aligned</li>
904 <li><tt>v128:128:128</tt> - 128-bit vector is 128-bit aligned</li>
905 <li><tt>a0:0:1</tt> - aggregates are 8-bit aligned</li>
907 <p>When llvm is determining the alignment for a given type, it uses the
910 <li>If the type sought is an exact match for one of the specifications, that
911 specification is used.</li>
912 <li>If no match is found, and the type sought is an integer type, then the
913 smallest integer type that is larger than the bitwidth of the sought type is
914 used. If none of the specifications are larger than the bitwidth then the the
915 largest integer type is used. For example, given the default specifications
916 above, the i7 type will use the alignment of i8 (next largest) while both
917 i65 and i256 will use the alignment of i64 (largest specified).</li>
918 <li>If no match is found, and the type sought is a vector type, then the
919 largest vector type that is smaller than the sought vector type will be used
920 as a fall back. This happens because <128 x double> can be implemented in
921 terms of 64 <2 x double>, for example.</li>
925 <!-- *********************************************************************** -->
926 <div class="doc_section"> <a name="typesystem">Type System</a> </div>
927 <!-- *********************************************************************** -->
929 <div class="doc_text">
931 <p>The LLVM type system is one of the most important features of the
932 intermediate representation. Being typed enables a number of
933 optimizations to be performed on the IR directly, without having to do
934 extra analyses on the side before the transformation. A strong type
935 system makes it easier to read the generated code and enables novel
936 analyses and transformations that are not feasible to perform on normal
937 three address code representations.</p>
941 <!-- ======================================================================= -->
942 <div class="doc_subsection"> <a name="t_primitive">Primitive Types</a> </div>
943 <div class="doc_text">
944 <p>The primitive types are the fundamental building blocks of the LLVM
945 system. The current set of primitive types is as follows:</p>
947 <table class="layout">
952 <tr><th>Type</th><th>Description</th></tr>
953 <tr><td><tt><a name="t_void">void</a></tt></td><td>No value</td></tr>
954 <tr><td><tt>label</tt></td><td>Branch destination</td></tr>
961 <tr><th>Type</th><th>Description</th></tr>
962 <tr><td><tt>float</tt></td><td>32-bit floating point value</td></tr>
963 <tr><td><tt>double</tt></td><td>64-bit floating point value</td></tr>
971 <!-- _______________________________________________________________________ -->
972 <div class="doc_subsubsection"> <a name="t_classifications">Type
973 Classifications</a> </div>
974 <div class="doc_text">
975 <p>These different primitive types fall into a few useful
978 <table border="1" cellspacing="0" cellpadding="4">
980 <tr><th>Classification</th><th>Types</th></tr>
982 <td><a name="t_integer">integer</a></td>
983 <td><tt>i1, i2, i3, ... i8, ... i16, ... i32, ... i64, ... </tt></td>
986 <td><a name="t_floating">floating point</a></td>
987 <td><tt>float, double</tt></td>
990 <td><a name="t_firstclass">first class</a></td>
991 <td><tt>i1, ..., float, double, <br/>
992 <a href="#t_pointer">pointer</a>,<a href="#t_vector">vector</a></tt>
998 <p>The <a href="#t_firstclass">first class</a> types are perhaps the
999 most important. Values of these types are the only ones which can be
1000 produced by instructions, passed as arguments, or used as operands to
1001 instructions. This means that all structures and arrays must be
1002 manipulated either by pointer or by component.</p>
1005 <!-- ======================================================================= -->
1006 <div class="doc_subsection"> <a name="t_derived">Derived Types</a> </div>
1008 <div class="doc_text">
1010 <p>The real power in LLVM comes from the derived types in the system.
1011 This is what allows a programmer to represent arrays, functions,
1012 pointers, and other useful types. Note that these derived types may be
1013 recursive: For example, it is possible to have a two dimensional array.</p>
1017 <!-- _______________________________________________________________________ -->
1018 <div class="doc_subsubsection"> <a name="t_integer">Integer Type</a> </div>
1020 <div class="doc_text">
1023 <p>The integer type is a very simple derived type that simply specifies an
1024 arbitrary bit width for the integer type desired. Any bit width from 1 bit to
1025 2^23-1 (about 8 million) can be specified.</p>
1033 <p>The number of bits the integer will occupy is specified by the <tt>N</tt>
1037 <table class="layout">
1047 <tt>i1942652</tt><br/>
1050 A boolean integer of 1 bit<br/>
1051 A nibble sized integer of 4 bits.<br/>
1052 A byte sized integer of 8 bits.<br/>
1053 A half word sized integer of 16 bits.<br/>
1054 A word sized integer of 32 bits.<br/>
1055 An integer whose bit width is the answer. <br/>
1056 A double word sized integer of 64 bits.<br/>
1057 A really big integer of over 1 million bits.<br/>
1063 <!-- _______________________________________________________________________ -->
1064 <div class="doc_subsubsection"> <a name="t_array">Array Type</a> </div>
1066 <div class="doc_text">
1070 <p>The array type is a very simple derived type that arranges elements
1071 sequentially in memory. The array type requires a size (number of
1072 elements) and an underlying data type.</p>
1077 [<# elements> x <elementtype>]
1080 <p>The number of elements is a constant integer value; elementtype may
1081 be any type with a size.</p>
1084 <table class="layout">
1087 <tt>[40 x i32 ]</tt><br/>
1088 <tt>[41 x i32 ]</tt><br/>
1089 <tt>[40 x i8]</tt><br/>
1092 Array of 40 32-bit integer values.<br/>
1093 Array of 41 32-bit integer values.<br/>
1094 Array of 40 8-bit integer values.<br/>
1098 <p>Here are some examples of multidimensional arrays:</p>
1099 <table class="layout">
1102 <tt>[3 x [4 x i32]]</tt><br/>
1103 <tt>[12 x [10 x float]]</tt><br/>
1104 <tt>[2 x [3 x [4 x i16]]]</tt><br/>
1107 3x4 array of 32-bit integer values.<br/>
1108 12x10 array of single precision floating point values.<br/>
1109 2x3x4 array of 16-bit integer values.<br/>
1114 <p>Note that 'variable sized arrays' can be implemented in LLVM with a zero
1115 length array. Normally, accesses past the end of an array are undefined in
1116 LLVM (e.g. it is illegal to access the 5th element of a 3 element array).
1117 As a special case, however, zero length arrays are recognized to be variable
1118 length. This allows implementation of 'pascal style arrays' with the LLVM
1119 type "{ i32, [0 x float]}", for example.</p>
1123 <!-- _______________________________________________________________________ -->
1124 <div class="doc_subsubsection"> <a name="t_function">Function Type</a> </div>
1125 <div class="doc_text">
1127 <p>The function type can be thought of as a function signature. It
1128 consists of a return type and a list of formal parameter types.
1129 Function types are usually used to build virtual function tables
1130 (which are structures of pointers to functions), for indirect function
1131 calls, and when defining a function.</p>
1133 The return type of a function type cannot be an aggregate type.
1136 <pre> <returntype> (<parameter list>)<br></pre>
1137 <p>...where '<tt><parameter list></tt>' is a comma-separated list of type
1138 specifiers. Optionally, the parameter list may include a type <tt>...</tt>,
1139 which indicates that the function takes a variable number of arguments.
1140 Variable argument functions can access their arguments with the <a
1141 href="#int_varargs">variable argument handling intrinsic</a> functions.</p>
1143 <table class="layout">
1145 <td class="left"><tt>i32 (i32)</tt></td>
1146 <td class="left">function taking an <tt>i32</tt>, returning an <tt>i32</tt>
1148 </tr><tr class="layout">
1149 <td class="left"><tt>float (i16 signext, i32 *) *
1151 <td class="left"><a href="#t_pointer">Pointer</a> to a function that takes
1152 an <tt>i16</tt> that should be sign extended and a
1153 <a href="#t_pointer">pointer</a> to <tt>i32</tt>, returning
1156 </tr><tr class="layout">
1157 <td class="left"><tt>i32 (i8*, ...)</tt></td>
1158 <td class="left">A vararg function that takes at least one
1159 <a href="#t_pointer">pointer</a> to <tt>i8 </tt> (char in C),
1160 which returns an integer. This is the signature for <tt>printf</tt> in
1167 <!-- _______________________________________________________________________ -->
1168 <div class="doc_subsubsection"> <a name="t_struct">Structure Type</a> </div>
1169 <div class="doc_text">
1171 <p>The structure type is used to represent a collection of data members
1172 together in memory. The packing of the field types is defined to match
1173 the ABI of the underlying processor. The elements of a structure may
1174 be any type that has a size.</p>
1175 <p>Structures are accessed using '<tt><a href="#i_load">load</a></tt>
1176 and '<tt><a href="#i_store">store</a></tt>' by getting a pointer to a
1177 field with the '<tt><a href="#i_getelementptr">getelementptr</a></tt>'
1180 <pre> { <type list> }<br></pre>
1182 <table class="layout">
1184 <td class="left"><tt>{ i32, i32, i32 }</tt></td>
1185 <td class="left">A triple of three <tt>i32</tt> values</td>
1186 </tr><tr class="layout">
1187 <td class="left"><tt>{ float, i32 (i32) * }</tt></td>
1188 <td class="left">A pair, where the first element is a <tt>float</tt> and the
1189 second element is a <a href="#t_pointer">pointer</a> to a
1190 <a href="#t_function">function</a> that takes an <tt>i32</tt>, returning
1191 an <tt>i32</tt>.</td>
1196 <!-- _______________________________________________________________________ -->
1197 <div class="doc_subsubsection"> <a name="t_pstruct">Packed Structure Type</a>
1199 <div class="doc_text">
1201 <p>The packed structure type is used to represent a collection of data members
1202 together in memory. There is no padding between fields. Further, the alignment
1203 of a packed structure is 1 byte. The elements of a packed structure may
1204 be any type that has a size.</p>
1205 <p>Structures are accessed using '<tt><a href="#i_load">load</a></tt>
1206 and '<tt><a href="#i_store">store</a></tt>' by getting a pointer to a
1207 field with the '<tt><a href="#i_getelementptr">getelementptr</a></tt>'
1210 <pre> < { <type list> } > <br></pre>
1212 <table class="layout">
1214 <td class="left"><tt>< { i32, i32, i32 } ></tt></td>
1215 <td class="left">A triple of three <tt>i32</tt> values</td>
1216 </tr><tr class="layout">
1217 <td class="left"><tt>< { float, i32 (i32) * } ></tt></td>
1218 <td class="left">A pair, where the first element is a <tt>float</tt> and the
1219 second element is a <a href="#t_pointer">pointer</a> to a
1220 <a href="#t_function">function</a> that takes an <tt>i32</tt>, returning
1221 an <tt>i32</tt>.</td>
1226 <!-- _______________________________________________________________________ -->
1227 <div class="doc_subsubsection"> <a name="t_pointer">Pointer Type</a> </div>
1228 <div class="doc_text">
1230 <p>As in many languages, the pointer type represents a pointer or
1231 reference to another object, which must live in memory.</p>
1233 <pre> <type> *<br></pre>
1235 <table class="layout">
1238 <tt>[4x i32]*</tt><br/>
1239 <tt>i32 (i32 *) *</tt><br/>
1242 A <a href="#t_pointer">pointer</a> to <a href="#t_array">array</a> of
1243 four <tt>i32</tt> values<br/>
1244 A <a href="#t_pointer">pointer</a> to a <a
1245 href="#t_function">function</a> that takes an <tt>i32*</tt>, returning an
1252 <!-- _______________________________________________________________________ -->
1253 <div class="doc_subsubsection"> <a name="t_vector">Vector Type</a> </div>
1254 <div class="doc_text">
1258 <p>A vector type is a simple derived type that represents a vector
1259 of elements. Vector types are used when multiple primitive data
1260 are operated in parallel using a single instruction (SIMD).
1261 A vector type requires a size (number of
1262 elements) and an underlying primitive data type. Vectors must have a power
1263 of two length (1, 2, 4, 8, 16 ...). Vector types are
1264 considered <a href="#t_firstclass">first class</a>.</p>
1269 < <# elements> x <elementtype> >
1272 <p>The number of elements is a constant integer value; elementtype may
1273 be any integer or floating point type.</p>
1277 <table class="layout">
1280 <tt><4 x i32></tt><br/>
1281 <tt><8 x float></tt><br/>
1282 <tt><2 x i64></tt><br/>
1285 Vector of 4 32-bit integer values.<br/>
1286 Vector of 8 floating-point values.<br/>
1287 Vector of 2 64-bit integer values.<br/>
1293 <!-- _______________________________________________________________________ -->
1294 <div class="doc_subsubsection"> <a name="t_opaque">Opaque Type</a> </div>
1295 <div class="doc_text">
1299 <p>Opaque types are used to represent unknown types in the system. This
1300 corresponds (for example) to the C notion of a foward declared structure type.
1301 In LLVM, opaque types can eventually be resolved to any type (not just a
1302 structure type).</p>
1312 <table class="layout">
1318 An opaque type.<br/>
1325 <!-- *********************************************************************** -->
1326 <div class="doc_section"> <a name="constants">Constants</a> </div>
1327 <!-- *********************************************************************** -->
1329 <div class="doc_text">
1331 <p>LLVM has several different basic types of constants. This section describes
1332 them all and their syntax.</p>
1336 <!-- ======================================================================= -->
1337 <div class="doc_subsection"><a name="simpleconstants">Simple Constants</a></div>
1339 <div class="doc_text">
1342 <dt><b>Boolean constants</b></dt>
1344 <dd>The two strings '<tt>true</tt>' and '<tt>false</tt>' are both valid
1345 constants of the <tt><a href="#t_primitive">i1</a></tt> type.
1348 <dt><b>Integer constants</b></dt>
1350 <dd>Standard integers (such as '4') are constants of the <a
1351 href="#t_integer">integer</a> type. Negative numbers may be used with
1355 <dt><b>Floating point constants</b></dt>
1357 <dd>Floating point constants use standard decimal notation (e.g. 123.421),
1358 exponential notation (e.g. 1.23421e+2), or a more precise hexadecimal
1359 notation (see below). Floating point constants must have a <a
1360 href="#t_floating">floating point</a> type. </dd>
1362 <dt><b>Null pointer constants</b></dt>
1364 <dd>The identifier '<tt>null</tt>' is recognized as a null pointer constant
1365 and must be of <a href="#t_pointer">pointer type</a>.</dd>
1369 <p>The one non-intuitive notation for constants is the optional hexadecimal form
1370 of floating point constants. For example, the form '<tt>double
1371 0x432ff973cafa8000</tt>' is equivalent to (but harder to read than) '<tt>double
1372 4.5e+15</tt>'. The only time hexadecimal floating point constants are required
1373 (and the only time that they are generated by the disassembler) is when a
1374 floating point constant must be emitted but it cannot be represented as a
1375 decimal floating point number. For example, NaN's, infinities, and other
1376 special values are represented in their IEEE hexadecimal format so that
1377 assembly and disassembly do not cause any bits to change in the constants.</p>
1381 <!-- ======================================================================= -->
1382 <div class="doc_subsection"><a name="aggregateconstants">Aggregate Constants</a>
1385 <div class="doc_text">
1386 <p>Aggregate constants arise from aggregation of simple constants
1387 and smaller aggregate constants.</p>
1390 <dt><b>Structure constants</b></dt>
1392 <dd>Structure constants are represented with notation similar to structure
1393 type definitions (a comma separated list of elements, surrounded by braces
1394 (<tt>{}</tt>)). For example: "<tt>{ i32 4, float 17.0, i32* %G }</tt>",
1395 where "<tt>%G</tt>" is declared as "<tt>@G = external global i32</tt>". Structure constants
1396 must have <a href="#t_struct">structure type</a>, and the number and
1397 types of elements must match those specified by the type.
1400 <dt><b>Array constants</b></dt>
1402 <dd>Array constants are represented with notation similar to array type
1403 definitions (a comma separated list of elements, surrounded by square brackets
1404 (<tt>[]</tt>)). For example: "<tt>[ i32 42, i32 11, i32 74 ]</tt>". Array
1405 constants must have <a href="#t_array">array type</a>, and the number and
1406 types of elements must match those specified by the type.
1409 <dt><b>Vector constants</b></dt>
1411 <dd>Vector constants are represented with notation similar to vector type
1412 definitions (a comma separated list of elements, surrounded by
1413 less-than/greater-than's (<tt><></tt>)). For example: "<tt>< i32 42,
1414 i32 11, i32 74, i32 100 ></tt>". Vector constants must have <a
1415 href="#t_vector">vector type</a>, and the number and types of elements must
1416 match those specified by the type.
1419 <dt><b>Zero initialization</b></dt>
1421 <dd>The string '<tt>zeroinitializer</tt>' can be used to zero initialize a
1422 value to zero of <em>any</em> type, including scalar and aggregate types.
1423 This is often used to avoid having to print large zero initializers (e.g. for
1424 large arrays) and is always exactly equivalent to using explicit zero
1431 <!-- ======================================================================= -->
1432 <div class="doc_subsection">
1433 <a name="globalconstants">Global Variable and Function Addresses</a>
1436 <div class="doc_text">
1438 <p>The addresses of <a href="#globalvars">global variables</a> and <a
1439 href="#functionstructure">functions</a> are always implicitly valid (link-time)
1440 constants. These constants are explicitly referenced when the <a
1441 href="#identifiers">identifier for the global</a> is used and always have <a
1442 href="#t_pointer">pointer</a> type. For example, the following is a legal LLVM
1445 <div class="doc_code">
1449 @Z = global [2 x i32*] [ i32* @X, i32* @Y ]
1455 <!-- ======================================================================= -->
1456 <div class="doc_subsection"><a name="undefvalues">Undefined Values</a></div>
1457 <div class="doc_text">
1458 <p>The string '<tt>undef</tt>' is recognized as a type-less constant that has
1459 no specific value. Undefined values may be of any type and be used anywhere
1460 a constant is permitted.</p>
1462 <p>Undefined values indicate to the compiler that the program is well defined
1463 no matter what value is used, giving the compiler more freedom to optimize.
1467 <!-- ======================================================================= -->
1468 <div class="doc_subsection"><a name="constantexprs">Constant Expressions</a>
1471 <div class="doc_text">
1473 <p>Constant expressions are used to allow expressions involving other constants
1474 to be used as constants. Constant expressions may be of any <a
1475 href="#t_firstclass">first class</a> type and may involve any LLVM operation
1476 that does not have side effects (e.g. load and call are not supported). The
1477 following is the syntax for constant expressions:</p>
1480 <dt><b><tt>trunc ( CST to TYPE )</tt></b></dt>
1481 <dd>Truncate a constant to another type. The bit size of CST must be larger
1482 than the bit size of TYPE. Both types must be integers.</dd>
1484 <dt><b><tt>zext ( CST to TYPE )</tt></b></dt>
1485 <dd>Zero extend a constant to another type. The bit size of CST must be
1486 smaller or equal to the bit size of TYPE. Both types must be integers.</dd>
1488 <dt><b><tt>sext ( CST to TYPE )</tt></b></dt>
1489 <dd>Sign extend a constant to another type. The bit size of CST must be
1490 smaller or equal to the bit size of TYPE. Both types must be integers.</dd>
1492 <dt><b><tt>fptrunc ( CST to TYPE )</tt></b></dt>
1493 <dd>Truncate a floating point constant to another floating point type. The
1494 size of CST must be larger than the size of TYPE. Both types must be
1495 floating point.</dd>
1497 <dt><b><tt>fpext ( CST to TYPE )</tt></b></dt>
1498 <dd>Floating point extend a constant to another type. The size of CST must be
1499 smaller or equal to the size of TYPE. Both types must be floating point.</dd>
1501 <dt><b><tt>fp2uint ( CST to TYPE )</tt></b></dt>
1502 <dd>Convert a floating point constant to the corresponding unsigned integer
1503 constant. TYPE must be an integer type. CST must be floating point. If the
1504 value won't fit in the integer type, the results are undefined.</dd>
1506 <dt><b><tt>fptosi ( CST to TYPE )</tt></b></dt>
1507 <dd>Convert a floating point constant to the corresponding signed integer
1508 constant. TYPE must be an integer type. CST must be floating point. If the
1509 value won't fit in the integer type, the results are undefined.</dd>
1511 <dt><b><tt>uitofp ( CST to TYPE )</tt></b></dt>
1512 <dd>Convert an unsigned integer constant to the corresponding floating point
1513 constant. TYPE must be floating point. CST must be of integer type. If the
1514 value won't fit in the floating point type, the results are undefined.</dd>
1516 <dt><b><tt>sitofp ( CST to TYPE )</tt></b></dt>
1517 <dd>Convert a signed integer constant to the corresponding floating point
1518 constant. TYPE must be floating point. CST must be of integer type. If the
1519 value won't fit in the floating point type, the results are undefined.</dd>
1521 <dt><b><tt>ptrtoint ( CST to TYPE )</tt></b></dt>
1522 <dd>Convert a pointer typed constant to the corresponding integer constant
1523 TYPE must be an integer type. CST must be of pointer type. The CST value is
1524 zero extended, truncated, or unchanged to make it fit in TYPE.</dd>
1526 <dt><b><tt>inttoptr ( CST to TYPE )</tt></b></dt>
1527 <dd>Convert a integer constant to a pointer constant. TYPE must be a
1528 pointer type. CST must be of integer type. The CST value is zero extended,
1529 truncated, or unchanged to make it fit in a pointer size. This one is
1530 <i>really</i> dangerous!</dd>
1532 <dt><b><tt>bitcast ( CST to TYPE )</tt></b></dt>
1533 <dd>Convert a constant, CST, to another TYPE. The size of CST and TYPE must be
1534 identical (same number of bits). The conversion is done as if the CST value
1535 was stored to memory and read back as TYPE. In other words, no bits change
1536 with this operator, just the type. This can be used for conversion of
1537 vector types to any other type, as long as they have the same bit width. For
1538 pointers it is only valid to cast to another pointer type.
1541 <dt><b><tt>getelementptr ( CSTPTR, IDX0, IDX1, ... )</tt></b></dt>
1543 <dd>Perform the <a href="#i_getelementptr">getelementptr operation</a> on
1544 constants. As with the <a href="#i_getelementptr">getelementptr</a>
1545 instruction, the index list may have zero or more indexes, which are required
1546 to make sense for the type of "CSTPTR".</dd>
1548 <dt><b><tt>select ( COND, VAL1, VAL2 )</tt></b></dt>
1550 <dd>Perform the <a href="#i_select">select operation</a> on
1553 <dt><b><tt>icmp COND ( VAL1, VAL2 )</tt></b></dt>
1554 <dd>Performs the <a href="#i_icmp">icmp operation</a> on constants.</dd>
1556 <dt><b><tt>fcmp COND ( VAL1, VAL2 )</tt></b></dt>
1557 <dd>Performs the <a href="#i_fcmp">fcmp operation</a> on constants.</dd>
1559 <dt><b><tt>extractelement ( VAL, IDX )</tt></b></dt>
1561 <dd>Perform the <a href="#i_extractelement">extractelement
1562 operation</a> on constants.
1564 <dt><b><tt>insertelement ( VAL, ELT, IDX )</tt></b></dt>
1566 <dd>Perform the <a href="#i_insertelement">insertelement
1567 operation</a> on constants.</dd>
1570 <dt><b><tt>shufflevector ( VEC1, VEC2, IDXMASK )</tt></b></dt>
1572 <dd>Perform the <a href="#i_shufflevector">shufflevector
1573 operation</a> on constants.</dd>
1575 <dt><b><tt>OPCODE ( LHS, RHS )</tt></b></dt>
1577 <dd>Perform the specified operation of the LHS and RHS constants. OPCODE may
1578 be any of the <a href="#binaryops">binary</a> or <a href="#bitwiseops">bitwise
1579 binary</a> operations. The constraints on operands are the same as those for
1580 the corresponding instruction (e.g. no bitwise operations on floating point
1581 values are allowed).</dd>
1585 <!-- *********************************************************************** -->
1586 <div class="doc_section"> <a name="othervalues">Other Values</a> </div>
1587 <!-- *********************************************************************** -->
1589 <!-- ======================================================================= -->
1590 <div class="doc_subsection">
1591 <a name="inlineasm">Inline Assembler Expressions</a>
1594 <div class="doc_text">
1597 LLVM supports inline assembler expressions (as opposed to <a href="#moduleasm">
1598 Module-Level Inline Assembly</a>) through the use of a special value. This
1599 value represents the inline assembler as a string (containing the instructions
1600 to emit), a list of operand constraints (stored as a string), and a flag that
1601 indicates whether or not the inline asm expression has side effects. An example
1602 inline assembler expression is:
1605 <div class="doc_code">
1607 i32 (i32) asm "bswap $0", "=r,r"
1612 Inline assembler expressions may <b>only</b> be used as the callee operand of
1613 a <a href="#i_call"><tt>call</tt> instruction</a>. Thus, typically we have:
1616 <div class="doc_code">
1618 %X = call i32 asm "<a href="#int_bswap">bswap</a> $0", "=r,r"(i32 %Y)
1623 Inline asms with side effects not visible in the constraint list must be marked
1624 as having side effects. This is done through the use of the
1625 '<tt>sideeffect</tt>' keyword, like so:
1628 <div class="doc_code">
1630 call void asm sideeffect "eieio", ""()
1634 <p>TODO: The format of the asm and constraints string still need to be
1635 documented here. Constraints on what can be done (e.g. duplication, moving, etc
1636 need to be documented).
1641 <!-- *********************************************************************** -->
1642 <div class="doc_section"> <a name="instref">Instruction Reference</a> </div>
1643 <!-- *********************************************************************** -->
1645 <div class="doc_text">
1647 <p>The LLVM instruction set consists of several different
1648 classifications of instructions: <a href="#terminators">terminator
1649 instructions</a>, <a href="#binaryops">binary instructions</a>,
1650 <a href="#bitwiseops">bitwise binary instructions</a>, <a
1651 href="#memoryops">memory instructions</a>, and <a href="#otherops">other
1652 instructions</a>.</p>
1656 <!-- ======================================================================= -->
1657 <div class="doc_subsection"> <a name="terminators">Terminator
1658 Instructions</a> </div>
1660 <div class="doc_text">
1662 <p>As mentioned <a href="#functionstructure">previously</a>, every
1663 basic block in a program ends with a "Terminator" instruction, which
1664 indicates which block should be executed after the current block is
1665 finished. These terminator instructions typically yield a '<tt>void</tt>'
1666 value: they produce control flow, not values (the one exception being
1667 the '<a href="#i_invoke"><tt>invoke</tt></a>' instruction).</p>
1668 <p>There are six different terminator instructions: the '<a
1669 href="#i_ret"><tt>ret</tt></a>' instruction, the '<a href="#i_br"><tt>br</tt></a>'
1670 instruction, the '<a href="#i_switch"><tt>switch</tt></a>' instruction,
1671 the '<a href="#i_invoke"><tt>invoke</tt></a>' instruction, the '<a
1672 href="#i_unwind"><tt>unwind</tt></a>' instruction, and the '<a
1673 href="#i_unreachable"><tt>unreachable</tt></a>' instruction.</p>
1677 <!-- _______________________________________________________________________ -->
1678 <div class="doc_subsubsection"> <a name="i_ret">'<tt>ret</tt>'
1679 Instruction</a> </div>
1680 <div class="doc_text">
1682 <pre> ret <type> <value> <i>; Return a value from a non-void function</i>
1683 ret void <i>; Return from void function</i>
1686 <p>The '<tt>ret</tt>' instruction is used to return control flow (and a
1687 value) from a function back to the caller.</p>
1688 <p>There are two forms of the '<tt>ret</tt>' instruction: one that
1689 returns a value and then causes control flow, and one that just causes
1690 control flow to occur.</p>
1692 <p>The '<tt>ret</tt>' instruction may return any '<a
1693 href="#t_firstclass">first class</a>' type. Notice that a function is
1694 not <a href="#wellformed">well formed</a> if there exists a '<tt>ret</tt>'
1695 instruction inside of the function that returns a value that does not
1696 match the return type of the function.</p>
1698 <p>When the '<tt>ret</tt>' instruction is executed, control flow
1699 returns back to the calling function's context. If the caller is a "<a
1700 href="#i_call"><tt>call</tt></a>" instruction, execution continues at
1701 the instruction after the call. If the caller was an "<a
1702 href="#i_invoke"><tt>invoke</tt></a>" instruction, execution continues
1703 at the beginning of the "normal" destination block. If the instruction
1704 returns a value, that value shall set the call or invoke instruction's
1707 <pre> ret i32 5 <i>; Return an integer value of 5</i>
1708 ret void <i>; Return from a void function</i>
1711 <!-- _______________________________________________________________________ -->
1712 <div class="doc_subsubsection"> <a name="i_br">'<tt>br</tt>' Instruction</a> </div>
1713 <div class="doc_text">
1715 <pre> br i1 <cond>, label <iftrue>, label <iffalse><br> br label <dest> <i>; Unconditional branch</i>
1718 <p>The '<tt>br</tt>' instruction is used to cause control flow to
1719 transfer to a different basic block in the current function. There are
1720 two forms of this instruction, corresponding to a conditional branch
1721 and an unconditional branch.</p>
1723 <p>The conditional branch form of the '<tt>br</tt>' instruction takes a
1724 single '<tt>i1</tt>' value and two '<tt>label</tt>' values. The
1725 unconditional form of the '<tt>br</tt>' instruction takes a single
1726 '<tt>label</tt>' value as a target.</p>
1728 <p>Upon execution of a conditional '<tt>br</tt>' instruction, the '<tt>i1</tt>'
1729 argument is evaluated. If the value is <tt>true</tt>, control flows
1730 to the '<tt>iftrue</tt>' <tt>label</tt> argument. If "cond" is <tt>false</tt>,
1731 control flows to the '<tt>iffalse</tt>' <tt>label</tt> argument.</p>
1733 <pre>Test:<br> %cond = <a href="#i_icmp">icmp</a> eq, i32 %a, %b<br> br i1 %cond, label %IfEqual, label %IfUnequal<br>IfEqual:<br> <a
1734 href="#i_ret">ret</a> i32 1<br>IfUnequal:<br> <a href="#i_ret">ret</a> i32 0<br></pre>
1736 <!-- _______________________________________________________________________ -->
1737 <div class="doc_subsubsection">
1738 <a name="i_switch">'<tt>switch</tt>' Instruction</a>
1741 <div class="doc_text">
1745 switch <intty> <value>, label <defaultdest> [ <intty> <val>, label <dest> ... ]
1750 <p>The '<tt>switch</tt>' instruction is used to transfer control flow to one of
1751 several different places. It is a generalization of the '<tt>br</tt>'
1752 instruction, allowing a branch to occur to one of many possible
1758 <p>The '<tt>switch</tt>' instruction uses three parameters: an integer
1759 comparison value '<tt>value</tt>', a default '<tt>label</tt>' destination, and
1760 an array of pairs of comparison value constants and '<tt>label</tt>'s. The
1761 table is not allowed to contain duplicate constant entries.</p>
1765 <p>The <tt>switch</tt> instruction specifies a table of values and
1766 destinations. When the '<tt>switch</tt>' instruction is executed, this
1767 table is searched for the given value. If the value is found, control flow is
1768 transfered to the corresponding destination; otherwise, control flow is
1769 transfered to the default destination.</p>
1771 <h5>Implementation:</h5>
1773 <p>Depending on properties of the target machine and the particular
1774 <tt>switch</tt> instruction, this instruction may be code generated in different
1775 ways. For example, it could be generated as a series of chained conditional
1776 branches or with a lookup table.</p>
1781 <i>; Emulate a conditional br instruction</i>
1782 %Val = <a href="#i_zext">zext</a> i1 %value to i32
1783 switch i32 %Val, label %truedest [i32 0, label %falsedest ]
1785 <i>; Emulate an unconditional br instruction</i>
1786 switch i32 0, label %dest [ ]
1788 <i>; Implement a jump table:</i>
1789 switch i32 %val, label %otherwise [ i32 0, label %onzero
1791 i32 2, label %ontwo ]
1795 <!-- _______________________________________________________________________ -->
1796 <div class="doc_subsubsection">
1797 <a name="i_invoke">'<tt>invoke</tt>' Instruction</a>
1800 <div class="doc_text">
1805 <result> = invoke [<a href="#callingconv">cconv</a>] <ptr to function ty> %<function ptr val>(<function args>)
1806 to label <normal label> unwind label <exception label>
1811 <p>The '<tt>invoke</tt>' instruction causes control to transfer to a specified
1812 function, with the possibility of control flow transfer to either the
1813 '<tt>normal</tt>' label or the
1814 '<tt>exception</tt>' label. If the callee function returns with the
1815 "<tt><a href="#i_ret">ret</a></tt>" instruction, control flow will return to the
1816 "normal" label. If the callee (or any indirect callees) returns with the "<a
1817 href="#i_unwind"><tt>unwind</tt></a>" instruction, control is interrupted and
1818 continued at the dynamically nearest "exception" label.</p>
1822 <p>This instruction requires several arguments:</p>
1826 The optional "cconv" marker indicates which <a href="#callingconv">calling
1827 convention</a> the call should use. If none is specified, the call defaults
1828 to using C calling conventions.
1830 <li>'<tt>ptr to function ty</tt>': shall be the signature of the pointer to
1831 function value being invoked. In most cases, this is a direct function
1832 invocation, but indirect <tt>invoke</tt>s are just as possible, branching off
1833 an arbitrary pointer to function value.
1836 <li>'<tt>function ptr val</tt>': An LLVM value containing a pointer to a
1837 function to be invoked. </li>
1839 <li>'<tt>function args</tt>': argument list whose types match the function
1840 signature argument types. If the function signature indicates the function
1841 accepts a variable number of arguments, the extra arguments can be
1844 <li>'<tt>normal label</tt>': the label reached when the called function
1845 executes a '<tt><a href="#i_ret">ret</a></tt>' instruction. </li>
1847 <li>'<tt>exception label</tt>': the label reached when a callee returns with
1848 the <a href="#i_unwind"><tt>unwind</tt></a> instruction. </li>
1854 <p>This instruction is designed to operate as a standard '<tt><a
1855 href="#i_call">call</a></tt>' instruction in most regards. The primary
1856 difference is that it establishes an association with a label, which is used by
1857 the runtime library to unwind the stack.</p>
1859 <p>This instruction is used in languages with destructors to ensure that proper
1860 cleanup is performed in the case of either a <tt>longjmp</tt> or a thrown
1861 exception. Additionally, this is important for implementation of
1862 '<tt>catch</tt>' clauses in high-level languages that support them.</p>
1866 %retval = invoke i32 %Test(i32 15) to label %Continue
1867 unwind label %TestCleanup <i>; {i32}:retval set</i>
1868 %retval = invoke <a href="#callingconv">coldcc</a> i32 %Test(i32 15) to label %Continue
1869 unwind label %TestCleanup <i>; {i32}:retval set</i>
1874 <!-- _______________________________________________________________________ -->
1876 <div class="doc_subsubsection"> <a name="i_unwind">'<tt>unwind</tt>'
1877 Instruction</a> </div>
1879 <div class="doc_text">
1888 <p>The '<tt>unwind</tt>' instruction unwinds the stack, continuing control flow
1889 at the first callee in the dynamic call stack which used an <a
1890 href="#i_invoke"><tt>invoke</tt></a> instruction to perform the call. This is
1891 primarily used to implement exception handling.</p>
1895 <p>The '<tt>unwind</tt>' intrinsic causes execution of the current function to
1896 immediately halt. The dynamic call stack is then searched for the first <a
1897 href="#i_invoke"><tt>invoke</tt></a> instruction on the call stack. Once found,
1898 execution continues at the "exceptional" destination block specified by the
1899 <tt>invoke</tt> instruction. If there is no <tt>invoke</tt> instruction in the
1900 dynamic call chain, undefined behavior results.</p>
1903 <!-- _______________________________________________________________________ -->
1905 <div class="doc_subsubsection"> <a name="i_unreachable">'<tt>unreachable</tt>'
1906 Instruction</a> </div>
1908 <div class="doc_text">
1917 <p>The '<tt>unreachable</tt>' instruction has no defined semantics. This
1918 instruction is used to inform the optimizer that a particular portion of the
1919 code is not reachable. This can be used to indicate that the code after a
1920 no-return function cannot be reached, and other facts.</p>
1924 <p>The '<tt>unreachable</tt>' instruction has no defined semantics.</p>
1929 <!-- ======================================================================= -->
1930 <div class="doc_subsection"> <a name="binaryops">Binary Operations</a> </div>
1931 <div class="doc_text">
1932 <p>Binary operators are used to do most of the computation in a
1933 program. They require two operands, execute an operation on them, and
1934 produce a single value. The operands might represent
1935 multiple data, as is the case with the <a href="#t_vector">vector</a> data type.
1936 The result value of a binary operator is not
1937 necessarily the same type as its operands.</p>
1938 <p>There are several different binary operators:</p>
1940 <!-- _______________________________________________________________________ -->
1941 <div class="doc_subsubsection"> <a name="i_add">'<tt>add</tt>'
1942 Instruction</a> </div>
1943 <div class="doc_text">
1945 <pre> <result> = add <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1948 <p>The '<tt>add</tt>' instruction returns the sum of its two operands.</p>
1950 <p>The two arguments to the '<tt>add</tt>' instruction must be either <a
1951 href="#t_integer">integer</a> or <a href="#t_floating">floating point</a> values.
1952 This instruction can also take <a href="#t_vector">vector</a> versions of the values.
1953 Both arguments must have identical types.</p>
1955 <p>The value produced is the integer or floating point sum of the two
1958 <pre> <result> = add i32 4, %var <i>; yields {i32}:result = 4 + %var</i>
1961 <!-- _______________________________________________________________________ -->
1962 <div class="doc_subsubsection"> <a name="i_sub">'<tt>sub</tt>'
1963 Instruction</a> </div>
1964 <div class="doc_text">
1966 <pre> <result> = sub <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1969 <p>The '<tt>sub</tt>' instruction returns the difference of its two
1971 <p>Note that the '<tt>sub</tt>' instruction is used to represent the '<tt>neg</tt>'
1972 instruction present in most other intermediate representations.</p>
1974 <p>The two arguments to the '<tt>sub</tt>' instruction must be either <a
1975 href="#t_integer">integer</a> or <a href="#t_floating">floating point</a>
1977 This instruction can also take <a href="#t_vector">vector</a> versions of the values.
1978 Both arguments must have identical types.</p>
1980 <p>The value produced is the integer or floating point difference of
1981 the two operands.</p>
1984 <result> = sub i32 4, %var <i>; yields {i32}:result = 4 - %var</i>
1985 <result> = sub i32 0, %val <i>; yields {i32}:result = -%var</i>
1988 <!-- _______________________________________________________________________ -->
1989 <div class="doc_subsubsection"> <a name="i_mul">'<tt>mul</tt>'
1990 Instruction</a> </div>
1991 <div class="doc_text">
1993 <pre> <result> = mul <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1996 <p>The '<tt>mul</tt>' instruction returns the product of its two
1999 <p>The two arguments to the '<tt>mul</tt>' instruction must be either <a
2000 href="#t_integer">integer</a> or <a href="#t_floating">floating point</a>
2002 This instruction can also take <a href="#t_vector">vector</a> versions of the values.
2003 Both arguments must have identical types.</p>
2005 <p>The value produced is the integer or floating point product of the
2007 <p>Because the operands are the same width, the result of an integer
2008 multiplication is the same whether the operands should be deemed unsigned or
2011 <pre> <result> = mul i32 4, %var <i>; yields {i32}:result = 4 * %var</i>
2014 <!-- _______________________________________________________________________ -->
2015 <div class="doc_subsubsection"> <a name="i_udiv">'<tt>udiv</tt>' Instruction
2017 <div class="doc_text">
2019 <pre> <result> = udiv <ty> <var1>, <var2> <i>; yields {ty}:result</i>
2022 <p>The '<tt>udiv</tt>' instruction returns the quotient of its two
2025 <p>The two arguments to the '<tt>udiv</tt>' instruction must be
2026 <a href="#t_integer">integer</a> values. Both arguments must have identical
2027 types. This instruction can also take <a href="#t_vector">vector</a> versions
2028 of the values in which case the elements must be integers.</p>
2030 <p>The value produced is the unsigned integer quotient of the two operands. This
2031 instruction always performs an unsigned division operation, regardless of
2032 whether the arguments are unsigned or not.</p>
2034 <pre> <result> = udiv i32 4, %var <i>; yields {i32}:result = 4 / %var</i>
2037 <!-- _______________________________________________________________________ -->
2038 <div class="doc_subsubsection"> <a name="i_sdiv">'<tt>sdiv</tt>' Instruction
2040 <div class="doc_text">
2042 <pre> <result> = sdiv <ty> <var1>, <var2> <i>; yields {ty}:result</i>
2045 <p>The '<tt>sdiv</tt>' instruction returns the quotient of its two
2048 <p>The two arguments to the '<tt>sdiv</tt>' instruction must be
2049 <a href="#t_integer">integer</a> values. Both arguments must have identical
2050 types. This instruction can also take <a href="#t_vector">vector</a> versions
2051 of the values in which case the elements must be integers.</p>
2053 <p>The value produced is the signed integer quotient of the two operands. This
2054 instruction always performs a signed division operation, regardless of whether
2055 the arguments are signed or not.</p>
2057 <pre> <result> = sdiv i32 4, %var <i>; yields {i32}:result = 4 / %var</i>
2060 <!-- _______________________________________________________________________ -->
2061 <div class="doc_subsubsection"> <a name="i_fdiv">'<tt>fdiv</tt>'
2062 Instruction</a> </div>
2063 <div class="doc_text">
2065 <pre> <result> = fdiv <ty> <var1>, <var2> <i>; yields {ty}:result</i>
2068 <p>The '<tt>fdiv</tt>' instruction returns the quotient of its two
2071 <p>The two arguments to the '<tt>fdiv</tt>' instruction must be
2072 <a href="#t_floating">floating point</a> values. Both arguments must have
2073 identical types. This instruction can also take <a href="#t_vector">vector</a>
2074 versions of floating point values.</p>
2076 <p>The value produced is the floating point quotient of the two operands.</p>
2078 <pre> <result> = fdiv float 4.0, %var <i>; yields {float}:result = 4.0 / %var</i>
2081 <!-- _______________________________________________________________________ -->
2082 <div class="doc_subsubsection"> <a name="i_urem">'<tt>urem</tt>' Instruction</a>
2084 <div class="doc_text">
2086 <pre> <result> = urem <ty> <var1>, <var2> <i>; yields {ty}:result</i>
2089 <p>The '<tt>urem</tt>' instruction returns the remainder from the
2090 unsigned division of its two arguments.</p>
2092 <p>The two arguments to the '<tt>urem</tt>' instruction must be
2093 <a href="#t_integer">integer</a> values. Both arguments must have identical
2096 <p>This instruction returns the unsigned integer <i>remainder</i> of a division.
2097 This instruction always performs an unsigned division to get the remainder,
2098 regardless of whether the arguments are unsigned or not.</p>
2100 <pre> <result> = urem i32 4, %var <i>; yields {i32}:result = 4 % %var</i>
2104 <!-- _______________________________________________________________________ -->
2105 <div class="doc_subsubsection"> <a name="i_srem">'<tt>srem</tt>'
2106 Instruction</a> </div>
2107 <div class="doc_text">
2109 <pre> <result> = srem <ty> <var1>, <var2> <i>; yields {ty}:result</i>
2112 <p>The '<tt>srem</tt>' instruction returns the remainder from the
2113 signed division of its two operands.</p>
2115 <p>The two arguments to the '<tt>srem</tt>' instruction must be
2116 <a href="#t_integer">integer</a> values. Both arguments must have identical
2119 <p>This instruction returns the <i>remainder</i> of a division (where the result
2120 has the same sign as the dividend, <tt>var1</tt>), not the <i>modulo</i>
2121 operator (where the result has the same sign as the divisor, <tt>var2</tt>) of
2122 a value. For more information about the difference, see <a
2123 href="http://mathforum.org/dr.math/problems/anne.4.28.99.html">The
2124 Math Forum</a>. For a table of how this is implemented in various languages,
2125 please see <a href="http://en.wikipedia.org/wiki/Modulo_operation">
2126 Wikipedia: modulo operation</a>.</p>
2128 <pre> <result> = srem i32 4, %var <i>; yields {i32}:result = 4 % %var</i>
2132 <!-- _______________________________________________________________________ -->
2133 <div class="doc_subsubsection"> <a name="i_frem">'<tt>frem</tt>'
2134 Instruction</a> </div>
2135 <div class="doc_text">
2137 <pre> <result> = frem <ty> <var1>, <var2> <i>; yields {ty}:result</i>
2140 <p>The '<tt>frem</tt>' instruction returns the remainder from the
2141 division of its two operands.</p>
2143 <p>The two arguments to the '<tt>frem</tt>' instruction must be
2144 <a href="#t_floating">floating point</a> values. Both arguments must have
2145 identical types.</p>
2147 <p>This instruction returns the <i>remainder</i> of a division.</p>
2149 <pre> <result> = frem float 4.0, %var <i>; yields {float}:result = 4.0 % %var</i>
2153 <!-- ======================================================================= -->
2154 <div class="doc_subsection"> <a name="bitwiseops">Bitwise Binary
2155 Operations</a> </div>
2156 <div class="doc_text">
2157 <p>Bitwise binary operators are used to do various forms of
2158 bit-twiddling in a program. They are generally very efficient
2159 instructions and can commonly be strength reduced from other
2160 instructions. They require two operands, execute an operation on them,
2161 and produce a single value. The resulting value of the bitwise binary
2162 operators is always the same type as its first operand.</p>
2165 <!-- _______________________________________________________________________ -->
2166 <div class="doc_subsubsection"> <a name="i_shl">'<tt>shl</tt>'
2167 Instruction</a> </div>
2168 <div class="doc_text">
2170 <pre> <result> = shl <ty> <var1>, <var2> <i>; yields {ty}:result</i>
2173 <p>The '<tt>shl</tt>' instruction returns the first operand shifted to
2174 the left a specified number of bits.</p>
2176 <p>Both arguments to the '<tt>shl</tt>' instruction must be the same <a
2177 href="#t_integer">integer</a> type.</p>
2179 <p>The value produced is <tt>var1</tt> * 2<sup><tt>var2</tt></sup>.</p>
2180 <h5>Example:</h5><pre>
2181 <result> = shl i32 4, %var <i>; yields {i32}: 4 << %var</i>
2182 <result> = shl i32 4, 2 <i>; yields {i32}: 16</i>
2183 <result> = shl i32 1, 10 <i>; yields {i32}: 1024</i>
2186 <!-- _______________________________________________________________________ -->
2187 <div class="doc_subsubsection"> <a name="i_lshr">'<tt>lshr</tt>'
2188 Instruction</a> </div>
2189 <div class="doc_text">
2191 <pre> <result> = lshr <ty> <var1>, <var2> <i>; yields {ty}:result</i>
2195 <p>The '<tt>lshr</tt>' instruction (logical shift right) returns the first
2196 operand shifted to the right a specified number of bits with zero fill.</p>
2199 <p>Both arguments to the '<tt>lshr</tt>' instruction must be the same
2200 <a href="#t_integer">integer</a> type.</p>
2203 <p>This instruction always performs a logical shift right operation. The most
2204 significant bits of the result will be filled with zero bits after the
2209 <result> = lshr i32 4, 1 <i>; yields {i32}:result = 2</i>
2210 <result> = lshr i32 4, 2 <i>; yields {i32}:result = 1</i>
2211 <result> = lshr i8 4, 3 <i>; yields {i8}:result = 0</i>
2212 <result> = lshr i8 -2, 1 <i>; yields {i8}:result = 0x7FFFFFFF </i>
2216 <!-- _______________________________________________________________________ -->
2217 <div class="doc_subsubsection"> <a name="i_ashr">'<tt>ashr</tt>'
2218 Instruction</a> </div>
2219 <div class="doc_text">
2222 <pre> <result> = ashr <ty> <var1>, <var2> <i>; yields {ty}:result</i>
2226 <p>The '<tt>ashr</tt>' instruction (arithmetic shift right) returns the first
2227 operand shifted to the right a specified number of bits with sign extension.</p>
2230 <p>Both arguments to the '<tt>ashr</tt>' instruction must be the same
2231 <a href="#t_integer">integer</a> type.</p>
2234 <p>This instruction always performs an arithmetic shift right operation,
2235 The most significant bits of the result will be filled with the sign bit
2236 of <tt>var1</tt>.</p>
2240 <result> = ashr i32 4, 1 <i>; yields {i32}:result = 2</i>
2241 <result> = ashr i32 4, 2 <i>; yields {i32}:result = 1</i>
2242 <result> = ashr i8 4, 3 <i>; yields {i8}:result = 0</i>
2243 <result> = ashr i8 -2, 1 <i>; yields {i8}:result = -1</i>
2247 <!-- _______________________________________________________________________ -->
2248 <div class="doc_subsubsection"> <a name="i_and">'<tt>and</tt>'
2249 Instruction</a> </div>
2250 <div class="doc_text">
2252 <pre> <result> = and <ty> <var1>, <var2> <i>; yields {ty}:result</i>
2255 <p>The '<tt>and</tt>' instruction returns the bitwise logical and of
2256 its two operands.</p>
2258 <p>The two arguments to the '<tt>and</tt>' instruction must be <a
2259 href="#t_integer">integer</a> values. Both arguments must have
2260 identical types.</p>
2262 <p>The truth table used for the '<tt>and</tt>' instruction is:</p>
2264 <div style="align: center">
2265 <table border="1" cellspacing="0" cellpadding="4">
2296 <pre> <result> = and i32 4, %var <i>; yields {i32}:result = 4 & %var</i>
2297 <result> = and i32 15, 40 <i>; yields {i32}:result = 8</i>
2298 <result> = and i32 4, 8 <i>; yields {i32}:result = 0</i>
2301 <!-- _______________________________________________________________________ -->
2302 <div class="doc_subsubsection"> <a name="i_or">'<tt>or</tt>' Instruction</a> </div>
2303 <div class="doc_text">
2305 <pre> <result> = or <ty> <var1>, <var2> <i>; yields {ty}:result</i>
2308 <p>The '<tt>or</tt>' instruction returns the bitwise logical inclusive
2309 or of its two operands.</p>
2311 <p>The two arguments to the '<tt>or</tt>' instruction must be <a
2312 href="#t_integer">integer</a> values. Both arguments must have
2313 identical types.</p>
2315 <p>The truth table used for the '<tt>or</tt>' instruction is:</p>
2317 <div style="align: center">
2318 <table border="1" cellspacing="0" cellpadding="4">
2349 <pre> <result> = or i32 4, %var <i>; yields {i32}:result = 4 | %var</i>
2350 <result> = or i32 15, 40 <i>; yields {i32}:result = 47</i>
2351 <result> = or i32 4, 8 <i>; yields {i32}:result = 12</i>
2354 <!-- _______________________________________________________________________ -->
2355 <div class="doc_subsubsection"> <a name="i_xor">'<tt>xor</tt>'
2356 Instruction</a> </div>
2357 <div class="doc_text">
2359 <pre> <result> = xor <ty> <var1>, <var2> <i>; yields {ty}:result</i>
2362 <p>The '<tt>xor</tt>' instruction returns the bitwise logical exclusive
2363 or of its two operands. The <tt>xor</tt> is used to implement the
2364 "one's complement" operation, which is the "~" operator in C.</p>
2366 <p>The two arguments to the '<tt>xor</tt>' instruction must be <a
2367 href="#t_integer">integer</a> values. Both arguments must have
2368 identical types.</p>
2370 <p>The truth table used for the '<tt>xor</tt>' instruction is:</p>
2372 <div style="align: center">
2373 <table border="1" cellspacing="0" cellpadding="4">
2405 <pre> <result> = xor i32 4, %var <i>; yields {i32}:result = 4 ^ %var</i>
2406 <result> = xor i32 15, 40 <i>; yields {i32}:result = 39</i>
2407 <result> = xor i32 4, 8 <i>; yields {i32}:result = 12</i>
2408 <result> = xor i32 %V, -1 <i>; yields {i32}:result = ~%V</i>
2412 <!-- ======================================================================= -->
2413 <div class="doc_subsection">
2414 <a name="vectorops">Vector Operations</a>
2417 <div class="doc_text">
2419 <p>LLVM supports several instructions to represent vector operations in a
2420 target-independent manner. These instructions cover the element-access and
2421 vector-specific operations needed to process vectors effectively. While LLVM
2422 does directly support these vector operations, many sophisticated algorithms
2423 will want to use target-specific intrinsics to take full advantage of a specific
2428 <!-- _______________________________________________________________________ -->
2429 <div class="doc_subsubsection">
2430 <a name="i_extractelement">'<tt>extractelement</tt>' Instruction</a>
2433 <div class="doc_text">
2438 <result> = extractelement <n x <ty>> <val>, i32 <idx> <i>; yields <ty></i>
2444 The '<tt>extractelement</tt>' instruction extracts a single scalar
2445 element from a vector at a specified index.
2452 The first operand of an '<tt>extractelement</tt>' instruction is a
2453 value of <a href="#t_vector">vector</a> type. The second operand is
2454 an index indicating the position from which to extract the element.
2455 The index may be a variable.</p>
2460 The result is a scalar of the same type as the element type of
2461 <tt>val</tt>. Its value is the value at position <tt>idx</tt> of
2462 <tt>val</tt>. If <tt>idx</tt> exceeds the length of <tt>val</tt>, the
2463 results are undefined.
2469 %result = extractelement <4 x i32> %vec, i32 0 <i>; yields i32</i>
2474 <!-- _______________________________________________________________________ -->
2475 <div class="doc_subsubsection">
2476 <a name="i_insertelement">'<tt>insertelement</tt>' Instruction</a>
2479 <div class="doc_text">
2484 <result> = insertelement <n x <ty>> <val>, <ty> <elt>, i32 <idx> <i>; yields <n x <ty>></i>
2490 The '<tt>insertelement</tt>' instruction inserts a scalar
2491 element into a vector at a specified index.
2498 The first operand of an '<tt>insertelement</tt>' instruction is a
2499 value of <a href="#t_vector">vector</a> type. The second operand is a
2500 scalar value whose type must equal the element type of the first
2501 operand. The third operand is an index indicating the position at
2502 which to insert the value. The index may be a variable.</p>
2507 The result is a vector of the same type as <tt>val</tt>. Its
2508 element values are those of <tt>val</tt> except at position
2509 <tt>idx</tt>, where it gets the value <tt>elt</tt>. If <tt>idx</tt>
2510 exceeds the length of <tt>val</tt>, the results are undefined.
2516 %result = insertelement <4 x i32> %vec, i32 1, i32 0 <i>; yields <4 x i32></i>
2520 <!-- _______________________________________________________________________ -->
2521 <div class="doc_subsubsection">
2522 <a name="i_shufflevector">'<tt>shufflevector</tt>' Instruction</a>
2525 <div class="doc_text">
2530 <result> = shufflevector <n x <ty>> <v1>, <n x <ty>> <v2>, <n x i32> <mask> <i>; yields <n x <ty>></i>
2536 The '<tt>shufflevector</tt>' instruction constructs a permutation of elements
2537 from two input vectors, returning a vector of the same type.
2543 The first two operands of a '<tt>shufflevector</tt>' instruction are vectors
2544 with types that match each other and types that match the result of the
2545 instruction. The third argument is a shuffle mask, which has the same number
2546 of elements as the other vector type, but whose element type is always 'i32'.
2550 The shuffle mask operand is required to be a constant vector with either
2551 constant integer or undef values.
2557 The elements of the two input vectors are numbered from left to right across
2558 both of the vectors. The shuffle mask operand specifies, for each element of
2559 the result vector, which element of the two input registers the result element
2560 gets. The element selector may be undef (meaning "don't care") and the second
2561 operand may be undef if performing a shuffle from only one vector.
2567 %result = shufflevector <4 x i32> %v1, <4 x i32> %v2,
2568 <4 x i32> <i32 0, i32 4, i32 1, i32 5> <i>; yields <4 x i32></i>
2569 %result = shufflevector <4 x i32> %v1, <4 x i32> undef,
2570 <4 x i32> <i32 0, i32 1, i32 2, i32 3> <i>; yields <4 x i32></i> - Identity shuffle.
2575 <!-- ======================================================================= -->
2576 <div class="doc_subsection">
2577 <a name="memoryops">Memory Access and Addressing Operations</a>
2580 <div class="doc_text">
2582 <p>A key design point of an SSA-based representation is how it
2583 represents memory. In LLVM, no memory locations are in SSA form, which
2584 makes things very simple. This section describes how to read, write,
2585 allocate, and free memory in LLVM.</p>
2589 <!-- _______________________________________________________________________ -->
2590 <div class="doc_subsubsection">
2591 <a name="i_malloc">'<tt>malloc</tt>' Instruction</a>
2594 <div class="doc_text">
2599 <result> = malloc <type>[, i32 <NumElements>][, align <alignment>] <i>; yields {type*}:result</i>
2604 <p>The '<tt>malloc</tt>' instruction allocates memory from the system
2605 heap and returns a pointer to it.</p>
2609 <p>The '<tt>malloc</tt>' instruction allocates
2610 <tt>sizeof(<type>)*NumElements</tt>
2611 bytes of memory from the operating system and returns a pointer of the
2612 appropriate type to the program. If "NumElements" is specified, it is the
2613 number of elements allocated. If an alignment is specified, the value result
2614 of the allocation is guaranteed to be aligned to at least that boundary. If
2615 not specified, or if zero, the target can choose to align the allocation on any
2616 convenient boundary.</p>
2618 <p>'<tt>type</tt>' must be a sized type.</p>
2622 <p>Memory is allocated using the system "<tt>malloc</tt>" function, and
2623 a pointer is returned.</p>
2628 %array = malloc [4 x i8 ] <i>; yields {[%4 x i8]*}:array</i>
2630 %size = <a href="#i_add">add</a> i32 2, 2 <i>; yields {i32}:size = i32 4</i>
2631 %array1 = malloc i8, i32 4 <i>; yields {i8*}:array1</i>
2632 %array2 = malloc [12 x i8], i32 %size <i>; yields {[12 x i8]*}:array2</i>
2633 %array3 = malloc i32, i32 4, align 1024 <i>; yields {i32*}:array3</i>
2634 %array4 = malloc i32, align 1024 <i>; yields {i32*}:array4</i>
2638 <!-- _______________________________________________________________________ -->
2639 <div class="doc_subsubsection">
2640 <a name="i_free">'<tt>free</tt>' Instruction</a>
2643 <div class="doc_text">
2648 free <type> <value> <i>; yields {void}</i>
2653 <p>The '<tt>free</tt>' instruction returns memory back to the unused
2654 memory heap to be reallocated in the future.</p>
2658 <p>'<tt>value</tt>' shall be a pointer value that points to a value
2659 that was allocated with the '<tt><a href="#i_malloc">malloc</a></tt>'
2664 <p>Access to the memory pointed to by the pointer is no longer defined
2665 after this instruction executes.</p>
2670 %array = <a href="#i_malloc">malloc</a> [4 x i8] <i>; yields {[4 x i8]*}:array</i>
2671 free [4 x i8]* %array
2675 <!-- _______________________________________________________________________ -->
2676 <div class="doc_subsubsection">
2677 <a name="i_alloca">'<tt>alloca</tt>' Instruction</a>
2680 <div class="doc_text">
2685 <result> = alloca <type>[, i32 <NumElements>][, align <alignment>] <i>; yields {type*}:result</i>
2690 <p>The '<tt>alloca</tt>' instruction allocates memory on the stack frame of the
2691 currently executing function, to be automatically released when this function
2692 returns to its caller.</p>
2696 <p>The '<tt>alloca</tt>' instruction allocates <tt>sizeof(<type>)*NumElements</tt>
2697 bytes of memory on the runtime stack, returning a pointer of the
2698 appropriate type to the program. If "NumElements" is specified, it is the
2699 number of elements allocated. If an alignment is specified, the value result
2700 of the allocation is guaranteed to be aligned to at least that boundary. If
2701 not specified, or if zero, the target can choose to align the allocation on any
2702 convenient boundary.</p>
2704 <p>'<tt>type</tt>' may be any sized type.</p>
2708 <p>Memory is allocated; a pointer is returned. '<tt>alloca</tt>'d
2709 memory is automatically released when the function returns. The '<tt>alloca</tt>'
2710 instruction is commonly used to represent automatic variables that must
2711 have an address available. When the function returns (either with the <tt><a
2712 href="#i_ret">ret</a></tt> or <tt><a href="#i_unwind">unwind</a></tt>
2713 instructions), the memory is reclaimed.</p>
2718 %ptr = alloca i32 <i>; yields {i32*}:ptr</i>
2719 %ptr = alloca i32, i32 4 <i>; yields {i32*}:ptr</i>
2720 %ptr = alloca i32, i32 4, align 1024 <i>; yields {i32*}:ptr</i>
2721 %ptr = alloca i32, align 1024 <i>; yields {i32*}:ptr</i>
2725 <!-- _______________________________________________________________________ -->
2726 <div class="doc_subsubsection"> <a name="i_load">'<tt>load</tt>'
2727 Instruction</a> </div>
2728 <div class="doc_text">
2730 <pre> <result> = load <ty>* <pointer>[, align <alignment>]<br> <result> = volatile load <ty>* <pointer>[, align <alignment>]<br></pre>
2732 <p>The '<tt>load</tt>' instruction is used to read from memory.</p>
2734 <p>The argument to the '<tt>load</tt>' instruction specifies the memory
2735 address from which to load. The pointer must point to a <a
2736 href="#t_firstclass">first class</a> type. If the <tt>load</tt> is
2737 marked as <tt>volatile</tt>, then the optimizer is not allowed to modify
2738 the number or order of execution of this <tt>load</tt> with other
2739 volatile <tt>load</tt> and <tt><a href="#i_store">store</a></tt>
2742 <p>The location of memory pointed to is loaded.</p>
2744 <pre> %ptr = <a href="#i_alloca">alloca</a> i32 <i>; yields {i32*}:ptr</i>
2746 href="#i_store">store</a> i32 3, i32* %ptr <i>; yields {void}</i>
2747 %val = load i32* %ptr <i>; yields {i32}:val = i32 3</i>
2750 <!-- _______________________________________________________________________ -->
2751 <div class="doc_subsubsection"> <a name="i_store">'<tt>store</tt>'
2752 Instruction</a> </div>
2753 <div class="doc_text">
2755 <pre> store <ty> <value>, <ty>* <pointer>[, align <alignment>] <i>; yields {void}</i>
2756 volatile store <ty> <value>, <ty>* <pointer>[, align <alignment>] <i>; yields {void}</i>
2759 <p>The '<tt>store</tt>' instruction is used to write to memory.</p>
2761 <p>There are two arguments to the '<tt>store</tt>' instruction: a value
2762 to store and an address at which to store it. The type of the '<tt><pointer></tt>'
2763 operand must be a pointer to the type of the '<tt><value></tt>'
2764 operand. If the <tt>store</tt> is marked as <tt>volatile</tt>, then the
2765 optimizer is not allowed to modify the number or order of execution of
2766 this <tt>store</tt> with other volatile <tt>load</tt> and <tt><a
2767 href="#i_store">store</a></tt> instructions.</p>
2769 <p>The contents of memory are updated to contain '<tt><value></tt>'
2770 at the location specified by the '<tt><pointer></tt>' operand.</p>
2772 <pre> %ptr = <a href="#i_alloca">alloca</a> i32 <i>; yields {i32*}:ptr</i>
2774 href="#i_store">store</a> i32 3, i32* %ptr <i>; yields {void}</i>
2775 %val = load i32* %ptr <i>; yields {i32}:val = i32 3</i>
2779 <!-- _______________________________________________________________________ -->
2780 <div class="doc_subsubsection">
2781 <a name="i_getelementptr">'<tt>getelementptr</tt>' Instruction</a>
2784 <div class="doc_text">
2787 <result> = getelementptr <ty>* <ptrval>{, <ty> <idx>}*
2793 The '<tt>getelementptr</tt>' instruction is used to get the address of a
2794 subelement of an aggregate data structure.</p>
2798 <p>This instruction takes a list of integer operands that indicate what
2799 elements of the aggregate object to index to. The actual types of the arguments
2800 provided depend on the type of the first pointer argument. The
2801 '<tt>getelementptr</tt>' instruction is used to index down through the type
2802 levels of a structure or to a specific index in an array. When indexing into a
2803 structure, only <tt>i32</tt> integer constants are allowed. When indexing
2804 into an array or pointer, only integers of 32 or 64 bits are allowed, and will
2805 be sign extended to 64-bit values.</p>
2807 <p>For example, let's consider a C code fragment and how it gets
2808 compiled to LLVM:</p>
2810 <div class="doc_code">
2823 int *foo(struct ST *s) {
2824 return &s[1].Z.B[5][13];
2829 <p>The LLVM code generated by the GCC frontend is:</p>
2831 <div class="doc_code">
2833 %RT = type { i8 , [10 x [20 x i32]], i8 }
2834 %ST = type { i32, double, %RT }
2836 define i32* %foo(%ST* %s) {
2838 %reg = getelementptr %ST* %s, i32 1, i32 2, i32 1, i32 5, i32 13
2846 <p>The index types specified for the '<tt>getelementptr</tt>' instruction depend
2847 on the pointer type that is being indexed into. <a href="#t_pointer">Pointer</a>
2848 and <a href="#t_array">array</a> types can use a 32-bit or 64-bit
2849 <a href="#t_integer">integer</a> type but the value will always be sign extended
2850 to 64-bits. <a href="#t_struct">Structure</a> types require <tt>i32</tt>
2851 <b>constants</b>.</p>
2853 <p>In the example above, the first index is indexing into the '<tt>%ST*</tt>'
2854 type, which is a pointer, yielding a '<tt>%ST</tt>' = '<tt>{ i32, double, %RT
2855 }</tt>' type, a structure. The second index indexes into the third element of
2856 the structure, yielding a '<tt>%RT</tt>' = '<tt>{ i8 , [10 x [20 x i32]],
2857 i8 }</tt>' type, another structure. The third index indexes into the second
2858 element of the structure, yielding a '<tt>[10 x [20 x i32]]</tt>' type, an
2859 array. The two dimensions of the array are subscripted into, yielding an
2860 '<tt>i32</tt>' type. The '<tt>getelementptr</tt>' instruction returns a pointer
2861 to this element, thus computing a value of '<tt>i32*</tt>' type.</p>
2863 <p>Note that it is perfectly legal to index partially through a
2864 structure, returning a pointer to an inner element. Because of this,
2865 the LLVM code for the given testcase is equivalent to:</p>
2868 define i32* %foo(%ST* %s) {
2869 %t1 = getelementptr %ST* %s, i32 1 <i>; yields %ST*:%t1</i>
2870 %t2 = getelementptr %ST* %t1, i32 0, i32 2 <i>; yields %RT*:%t2</i>
2871 %t3 = getelementptr %RT* %t2, i32 0, i32 1 <i>; yields [10 x [20 x i32]]*:%t3</i>
2872 %t4 = getelementptr [10 x [20 x i32]]* %t3, i32 0, i32 5 <i>; yields [20 x i32]*:%t4</i>
2873 %t5 = getelementptr [20 x i32]* %t4, i32 0, i32 13 <i>; yields i32*:%t5</i>
2878 <p>Note that it is undefined to access an array out of bounds: array and
2879 pointer indexes must always be within the defined bounds of the array type.
2880 The one exception for this rules is zero length arrays. These arrays are
2881 defined to be accessible as variable length arrays, which requires access
2882 beyond the zero'th element.</p>
2884 <p>The getelementptr instruction is often confusing. For some more insight
2885 into how it works, see <a href="GetElementPtr.html">the getelementptr
2891 <i>; yields [12 x i8]*:aptr</i>
2892 %aptr = getelementptr {i32, [12 x i8]}* %sptr, i64 0, i32 1
2896 <!-- ======================================================================= -->
2897 <div class="doc_subsection"> <a name="convertops">Conversion Operations</a>
2899 <div class="doc_text">
2900 <p>The instructions in this category are the conversion instructions (casting)
2901 which all take a single operand and a type. They perform various bit conversions
2905 <!-- _______________________________________________________________________ -->
2906 <div class="doc_subsubsection">
2907 <a name="i_trunc">'<tt>trunc .. to</tt>' Instruction</a>
2909 <div class="doc_text">
2913 <result> = trunc <ty> <value> to <ty2> <i>; yields ty2</i>
2918 The '<tt>trunc</tt>' instruction truncates its operand to the type <tt>ty2</tt>.
2923 The '<tt>trunc</tt>' instruction takes a <tt>value</tt> to trunc, which must
2924 be an <a href="#t_integer">integer</a> type, and a type that specifies the size
2925 and type of the result, which must be an <a href="#t_integer">integer</a>
2926 type. The bit size of <tt>value</tt> must be larger than the bit size of
2927 <tt>ty2</tt>. Equal sized types are not allowed.</p>
2931 The '<tt>trunc</tt>' instruction truncates the high order bits in <tt>value</tt>
2932 and converts the remaining bits to <tt>ty2</tt>. Since the source size must be
2933 larger than the destination size, <tt>trunc</tt> cannot be a <i>no-op cast</i>.
2934 It will always truncate bits.</p>
2938 %X = trunc i32 257 to i8 <i>; yields i8:1</i>
2939 %Y = trunc i32 123 to i1 <i>; yields i1:true</i>
2940 %Y = trunc i32 122 to i1 <i>; yields i1:false</i>
2944 <!-- _______________________________________________________________________ -->
2945 <div class="doc_subsubsection">
2946 <a name="i_zext">'<tt>zext .. to</tt>' Instruction</a>
2948 <div class="doc_text">
2952 <result> = zext <ty> <value> to <ty2> <i>; yields ty2</i>
2956 <p>The '<tt>zext</tt>' instruction zero extends its operand to type
2961 <p>The '<tt>zext</tt>' instruction takes a value to cast, which must be of
2962 <a href="#t_integer">integer</a> type, and a type to cast it to, which must
2963 also be of <a href="#t_integer">integer</a> type. The bit size of the
2964 <tt>value</tt> must be smaller than the bit size of the destination type,
2968 <p>The <tt>zext</tt> fills the high order bits of the <tt>value</tt> with zero
2969 bits until it reaches the size of the destination type, <tt>ty2</tt>.</p>
2971 <p>When zero extending from i1, the result will always be either 0 or 1.</p>
2975 %X = zext i32 257 to i64 <i>; yields i64:257</i>
2976 %Y = zext i1 true to i32 <i>; yields i32:1</i>
2980 <!-- _______________________________________________________________________ -->
2981 <div class="doc_subsubsection">
2982 <a name="i_sext">'<tt>sext .. to</tt>' Instruction</a>
2984 <div class="doc_text">
2988 <result> = sext <ty> <value> to <ty2> <i>; yields ty2</i>
2992 <p>The '<tt>sext</tt>' sign extends <tt>value</tt> to the type <tt>ty2</tt>.</p>
2996 The '<tt>sext</tt>' instruction takes a value to cast, which must be of
2997 <a href="#t_integer">integer</a> type, and a type to cast it to, which must
2998 also be of <a href="#t_integer">integer</a> type. The bit size of the
2999 <tt>value</tt> must be smaller than the bit size of the destination type,
3004 The '<tt>sext</tt>' instruction performs a sign extension by copying the sign
3005 bit (highest order bit) of the <tt>value</tt> until it reaches the bit size of
3006 the type <tt>ty2</tt>.</p>
3008 <p>When sign extending from i1, the extension always results in -1 or 0.</p>
3012 %X = sext i8 -1 to i16 <i>; yields i16 :65535</i>
3013 %Y = sext i1 true to i32 <i>; yields i32:-1</i>
3017 <!-- _______________________________________________________________________ -->
3018 <div class="doc_subsubsection">
3019 <a name="i_fptrunc">'<tt>fptrunc .. to</tt>' Instruction</a>
3022 <div class="doc_text">
3027 <result> = fptrunc <ty> <value> to <ty2> <i>; yields ty2</i>
3031 <p>The '<tt>fptrunc</tt>' instruction truncates <tt>value</tt> to type
3036 <p>The '<tt>fptrunc</tt>' instruction takes a <a href="#t_floating">floating
3037 point</a> value to cast and a <a href="#t_floating">floating point</a> type to
3038 cast it to. The size of <tt>value</tt> must be larger than the size of
3039 <tt>ty2</tt>. This implies that <tt>fptrunc</tt> cannot be used to make a
3040 <i>no-op cast</i>.</p>
3043 <p> The '<tt>fptrunc</tt>' instruction truncates a <tt>value</tt> from a larger
3044 <a href="#t_floating">floating point</a> type to a smaller
3045 <a href="#t_floating">floating point</a> type. If the value cannot fit within
3046 the destination type, <tt>ty2</tt>, then the results are undefined.</p>
3050 %X = fptrunc double 123.0 to float <i>; yields float:123.0</i>
3051 %Y = fptrunc double 1.0E+300 to float <i>; yields undefined</i>
3055 <!-- _______________________________________________________________________ -->
3056 <div class="doc_subsubsection">
3057 <a name="i_fpext">'<tt>fpext .. to</tt>' Instruction</a>
3059 <div class="doc_text">
3063 <result> = fpext <ty> <value> to <ty2> <i>; yields ty2</i>
3067 <p>The '<tt>fpext</tt>' extends a floating point <tt>value</tt> to a larger
3068 floating point value.</p>
3071 <p>The '<tt>fpext</tt>' instruction takes a
3072 <a href="#t_floating">floating point</a> <tt>value</tt> to cast,
3073 and a <a href="#t_floating">floating point</a> type to cast it to. The source
3074 type must be smaller than the destination type.</p>
3077 <p>The '<tt>fpext</tt>' instruction extends the <tt>value</tt> from a smaller
3078 <a href="#t_floating">floating point</a> type to a larger
3079 <a href="#t_floating">floating point</a> type. The <tt>fpext</tt> cannot be
3080 used to make a <i>no-op cast</i> because it always changes bits. Use
3081 <tt>bitcast</tt> to make a <i>no-op cast</i> for a floating point cast.</p>
3085 %X = fpext float 3.1415 to double <i>; yields double:3.1415</i>
3086 %Y = fpext float 1.0 to float <i>; yields float:1.0 (no-op)</i>
3090 <!-- _______________________________________________________________________ -->
3091 <div class="doc_subsubsection">
3092 <a name="i_fptoui">'<tt>fptoui .. to</tt>' Instruction</a>
3094 <div class="doc_text">
3098 <result> = fp2uint <ty> <value> to <ty2> <i>; yields ty2</i>
3102 <p>The '<tt>fp2uint</tt>' converts a floating point <tt>value</tt> to its
3103 unsigned integer equivalent of type <tt>ty2</tt>.
3107 <p>The '<tt>fp2uint</tt>' instruction takes a value to cast, which must be a
3108 <a href="#t_floating">floating point</a> value, and a type to cast it to, which
3109 must be an <a href="#t_integer">integer</a> type.</p>
3112 <p> The '<tt>fp2uint</tt>' instruction converts its
3113 <a href="#t_floating">floating point</a> operand into the nearest (rounding
3114 towards zero) unsigned integer value. If the value cannot fit in <tt>ty2</tt>,
3115 the results are undefined.</p>
3117 <p>When converting to i1, the conversion is done as a comparison against
3118 zero. If the <tt>value</tt> was zero, the i1 result will be <tt>false</tt>.
3119 If the <tt>value</tt> was non-zero, the i1 result will be <tt>true</tt>.</p>
3123 %X = fp2uint double 123.0 to i32 <i>; yields i32:123</i>
3124 %Y = fp2uint float 1.0E+300 to i1 <i>; yields i1:true</i>
3125 %X = fp2uint float 1.04E+17 to i8 <i>; yields undefined:1</i>
3129 <!-- _______________________________________________________________________ -->
3130 <div class="doc_subsubsection">
3131 <a name="i_fptosi">'<tt>fptosi .. to</tt>' Instruction</a>
3133 <div class="doc_text">
3137 <result> = fptosi <ty> <value> to <ty2> <i>; yields ty2</i>
3141 <p>The '<tt>fptosi</tt>' instruction converts
3142 <a href="#t_floating">floating point</a> <tt>value</tt> to type <tt>ty2</tt>.
3147 <p> The '<tt>fptosi</tt>' instruction takes a value to cast, which must be a
3148 <a href="#t_floating">floating point</a> value, and a type to cast it to, which
3149 must also be an <a href="#t_integer">integer</a> type.</p>
3152 <p>The '<tt>fptosi</tt>' instruction converts its
3153 <a href="#t_floating">floating point</a> operand into the nearest (rounding
3154 towards zero) signed integer value. If the value cannot fit in <tt>ty2</tt>,
3155 the results are undefined.</p>
3157 <p>When converting to i1, the conversion is done as a comparison against
3158 zero. If the <tt>value</tt> was zero, the i1 result will be <tt>false</tt>.
3159 If the <tt>value</tt> was non-zero, the i1 result will be <tt>true</tt>.</p>
3163 %X = fptosi double -123.0 to i32 <i>; yields i32:-123</i>
3164 %Y = fptosi float 1.0E-247 to i1 <i>; yields i1:true</i>
3165 %X = fptosi float 1.04E+17 to i8 <i>; yields undefined:1</i>
3169 <!-- _______________________________________________________________________ -->
3170 <div class="doc_subsubsection">
3171 <a name="i_uitofp">'<tt>uitofp .. to</tt>' Instruction</a>
3173 <div class="doc_text">
3177 <result> = uitofp <ty> <value> to <ty2> <i>; yields ty2</i>
3181 <p>The '<tt>uitofp</tt>' instruction regards <tt>value</tt> as an unsigned
3182 integer and converts that value to the <tt>ty2</tt> type.</p>
3186 <p>The '<tt>uitofp</tt>' instruction takes a value to cast, which must be an
3187 <a href="#t_integer">integer</a> value, and a type to cast it to, which must
3188 be a <a href="#t_floating">floating point</a> type.</p>
3191 <p>The '<tt>uitofp</tt>' instruction interprets its operand as an unsigned
3192 integer quantity and converts it to the corresponding floating point value. If
3193 the value cannot fit in the floating point value, the results are undefined.</p>
3198 %X = uitofp i32 257 to float <i>; yields float:257.0</i>
3199 %Y = uitofp i8 -1 to double <i>; yields double:255.0</i>
3203 <!-- _______________________________________________________________________ -->
3204 <div class="doc_subsubsection">
3205 <a name="i_sitofp">'<tt>sitofp .. to</tt>' Instruction</a>
3207 <div class="doc_text">
3211 <result> = sitofp <ty> <value> to <ty2> <i>; yields ty2</i>
3215 <p>The '<tt>sitofp</tt>' instruction regards <tt>value</tt> as a signed
3216 integer and converts that value to the <tt>ty2</tt> type.</p>
3219 <p>The '<tt>sitofp</tt>' instruction takes a value to cast, which must be an
3220 <a href="#t_integer">integer</a> value, and a type to cast it to, which must be
3221 a <a href="#t_floating">floating point</a> type.</p>
3224 <p>The '<tt>sitofp</tt>' instruction interprets its operand as a signed
3225 integer quantity and converts it to the corresponding floating point value. If
3226 the value cannot fit in the floating point value, the results are undefined.</p>
3230 %X = sitofp i32 257 to float <i>; yields float:257.0</i>
3231 %Y = sitofp i8 -1 to double <i>; yields double:-1.0</i>
3235 <!-- _______________________________________________________________________ -->
3236 <div class="doc_subsubsection">
3237 <a name="i_ptrtoint">'<tt>ptrtoint .. to</tt>' Instruction</a>
3239 <div class="doc_text">
3243 <result> = ptrtoint <ty> <value> to <ty2> <i>; yields ty2</i>
3247 <p>The '<tt>ptrtoint</tt>' instruction converts the pointer <tt>value</tt> to
3248 the integer type <tt>ty2</tt>.</p>
3251 <p>The '<tt>ptrtoint</tt>' instruction takes a <tt>value</tt> to cast, which
3252 must be a <a href="#t_pointer">pointer</a> value, and a type to cast it to
3253 <tt>ty2</tt>, which must be an <a href="#t_integer">integer</a> type.
3256 <p>The '<tt>ptrtoint</tt>' instruction converts <tt>value</tt> to integer type
3257 <tt>ty2</tt> by interpreting the pointer value as an integer and either
3258 truncating or zero extending that value to the size of the integer type. If
3259 <tt>value</tt> is smaller than <tt>ty2</tt> then a zero extension is done. If
3260 <tt>value</tt> is larger than <tt>ty2</tt> then a truncation is done. If they
3261 are the same size, then nothing is done (<i>no-op cast</i>) other than a type
3266 %X = ptrtoint i32* %X to i8 <i>; yields truncation on 32-bit architecture</i>
3267 %Y = ptrtoint i32* %x to i64 <i>; yields zero extension on 32-bit architecture</i>
3271 <!-- _______________________________________________________________________ -->
3272 <div class="doc_subsubsection">
3273 <a name="i_inttoptr">'<tt>inttoptr .. to</tt>' Instruction</a>
3275 <div class="doc_text">
3279 <result> = inttoptr <ty> <value> to <ty2> <i>; yields ty2</i>
3283 <p>The '<tt>inttoptr</tt>' instruction converts an integer <tt>value</tt> to
3284 a pointer type, <tt>ty2</tt>.</p>
3287 <p>The '<tt>inttoptr</tt>' instruction takes an <a href="#t_integer">integer</a>
3288 value to cast, and a type to cast it to, which must be a
3289 <a href="#t_pointer">pointer</a> type.
3292 <p>The '<tt>inttoptr</tt>' instruction converts <tt>value</tt> to type
3293 <tt>ty2</tt> by applying either a zero extension or a truncation depending on
3294 the size of the integer <tt>value</tt>. If <tt>value</tt> is larger than the
3295 size of a pointer then a truncation is done. If <tt>value</tt> is smaller than
3296 the size of a pointer then a zero extension is done. If they are the same size,
3297 nothing is done (<i>no-op cast</i>).</p>
3301 %X = inttoptr i32 255 to i32* <i>; yields zero extension on 64-bit architecture</i>
3302 %X = inttoptr i32 255 to i32* <i>; yields no-op on 32-bit architecture</i>
3303 %Y = inttoptr i64 0 to i32* <i>; yields truncation on 32-bit architecture</i>
3307 <!-- _______________________________________________________________________ -->
3308 <div class="doc_subsubsection">
3309 <a name="i_bitcast">'<tt>bitcast .. to</tt>' Instruction</a>
3311 <div class="doc_text">
3315 <result> = bitcast <ty> <value> to <ty2> <i>; yields ty2</i>
3319 <p>The '<tt>bitcast</tt>' instruction converts <tt>value</tt> to type
3320 <tt>ty2</tt> without changing any bits.</p>
3323 <p>The '<tt>bitcast</tt>' instruction takes a value to cast, which must be
3324 a first class value, and a type to cast it to, which must also be a <a
3325 href="#t_firstclass">first class</a> type. The bit sizes of <tt>value</tt>
3326 and the destination type, <tt>ty2</tt>, must be identical. If the source
3327 type is a pointer, the destination type must also be a pointer.</p>
3330 <p>The '<tt>bitcast</tt>' instruction converts <tt>value</tt> to type
3331 <tt>ty2</tt>. It is always a <i>no-op cast</i> because no bits change with
3332 this conversion. The conversion is done as if the <tt>value</tt> had been
3333 stored to memory and read back as type <tt>ty2</tt>. Pointer types may only be
3334 converted to other pointer types with this instruction. To convert pointers to
3335 other types, use the <a href="#i_inttoptr">inttoptr</a> or
3336 <a href="#i_ptrtoint">ptrtoint</a> instructions first.</p>
3340 %X = bitcast i8 255 to i8 <i>; yields i8 :-1</i>
3341 %Y = bitcast i32* %x to sint* <i>; yields sint*:%x</i>
3342 %Z = bitcast <2xint> %V to i64; <i>; yields i64: %V</i>
3346 <!-- ======================================================================= -->
3347 <div class="doc_subsection"> <a name="otherops">Other Operations</a> </div>
3348 <div class="doc_text">
3349 <p>The instructions in this category are the "miscellaneous"
3350 instructions, which defy better classification.</p>
3353 <!-- _______________________________________________________________________ -->
3354 <div class="doc_subsubsection"><a name="i_icmp">'<tt>icmp</tt>' Instruction</a>
3356 <div class="doc_text">
3358 <pre> <result> = icmp <cond> <ty> <var1>, <var2> <i>; yields {i1}:result</i>
3361 <p>The '<tt>icmp</tt>' instruction returns a boolean value based on comparison
3362 of its two integer operands.</p>
3364 <p>The '<tt>icmp</tt>' instruction takes three operands. The first operand is
3365 the condition code indicating the kind of comparison to perform. It is not
3366 a value, just a keyword. The possible condition code are:
3368 <li><tt>eq</tt>: equal</li>
3369 <li><tt>ne</tt>: not equal </li>
3370 <li><tt>ugt</tt>: unsigned greater than</li>
3371 <li><tt>uge</tt>: unsigned greater or equal</li>
3372 <li><tt>ult</tt>: unsigned less than</li>
3373 <li><tt>ule</tt>: unsigned less or equal</li>
3374 <li><tt>sgt</tt>: signed greater than</li>
3375 <li><tt>sge</tt>: signed greater or equal</li>
3376 <li><tt>slt</tt>: signed less than</li>
3377 <li><tt>sle</tt>: signed less or equal</li>
3379 <p>The remaining two arguments must be <a href="#t_integer">integer</a> or
3380 <a href="#t_pointer">pointer</a> typed. They must also be identical types.</p>
3382 <p>The '<tt>icmp</tt>' compares <tt>var1</tt> and <tt>var2</tt> according to
3383 the condition code given as <tt>cond</tt>. The comparison performed always
3384 yields a <a href="#t_primitive">i1</a> result, as follows:
3386 <li><tt>eq</tt>: yields <tt>true</tt> if the operands are equal,
3387 <tt>false</tt> otherwise. No sign interpretation is necessary or performed.
3389 <li><tt>ne</tt>: yields <tt>true</tt> if the operands are unequal,
3390 <tt>false</tt> otherwise. No sign interpretation is necessary or performed.
3391 <li><tt>ugt</tt>: interprets the operands as unsigned values and yields
3392 <tt>true</tt> if <tt>var1</tt> is greater than <tt>var2</tt>.</li>
3393 <li><tt>uge</tt>: interprets the operands as unsigned values and yields
3394 <tt>true</tt> if <tt>var1</tt> is greater than or equal to <tt>var2</tt>.</li>
3395 <li><tt>ult</tt>: interprets the operands as unsigned values and yields
3396 <tt>true</tt> if <tt>var1</tt> is less than <tt>var2</tt>.</li>
3397 <li><tt>ule</tt>: interprets the operands as unsigned values and yields
3398 <tt>true</tt> if <tt>var1</tt> is less than or equal to <tt>var2</tt>.</li>
3399 <li><tt>sgt</tt>: interprets the operands as signed values and yields
3400 <tt>true</tt> if <tt>var1</tt> is greater than <tt>var2</tt>.</li>
3401 <li><tt>sge</tt>: interprets the operands as signed values and yields
3402 <tt>true</tt> if <tt>var1</tt> is greater than or equal to <tt>var2</tt>.</li>
3403 <li><tt>slt</tt>: interprets the operands as signed values and yields
3404 <tt>true</tt> if <tt>var1</tt> is less than <tt>var2</tt>.</li>
3405 <li><tt>sle</tt>: interprets the operands as signed values and yields
3406 <tt>true</tt> if <tt>var1</tt> is less than or equal to <tt>var2</tt>.</li>
3408 <p>If the operands are <a href="#t_pointer">pointer</a> typed, the pointer
3409 values are compared as if they were integers.</p>
3412 <pre> <result> = icmp eq i32 4, 5 <i>; yields: result=false</i>
3413 <result> = icmp ne float* %X, %X <i>; yields: result=false</i>
3414 <result> = icmp ult i16 4, 5 <i>; yields: result=true</i>
3415 <result> = icmp sgt i16 4, 5 <i>; yields: result=false</i>
3416 <result> = icmp ule i16 -4, 5 <i>; yields: result=false</i>
3417 <result> = icmp sge i16 4, 5 <i>; yields: result=false</i>
3421 <!-- _______________________________________________________________________ -->
3422 <div class="doc_subsubsection"><a name="i_fcmp">'<tt>fcmp</tt>' Instruction</a>
3424 <div class="doc_text">
3426 <pre> <result> = fcmp <cond> <ty> <var1>, <var2> <i>; yields {i1}:result</i>
3429 <p>The '<tt>fcmp</tt>' instruction returns a boolean value based on comparison
3430 of its floating point operands.</p>
3432 <p>The '<tt>fcmp</tt>' instruction takes three operands. The first operand is
3433 the condition code indicating the kind of comparison to perform. It is not
3434 a value, just a keyword. The possible condition code are:
3436 <li><tt>false</tt>: no comparison, always returns false</li>
3437 <li><tt>oeq</tt>: ordered and equal</li>
3438 <li><tt>ogt</tt>: ordered and greater than </li>
3439 <li><tt>oge</tt>: ordered and greater than or equal</li>
3440 <li><tt>olt</tt>: ordered and less than </li>
3441 <li><tt>ole</tt>: ordered and less than or equal</li>
3442 <li><tt>one</tt>: ordered and not equal</li>
3443 <li><tt>ord</tt>: ordered (no nans)</li>
3444 <li><tt>ueq</tt>: unordered or equal</li>
3445 <li><tt>ugt</tt>: unordered or greater than </li>
3446 <li><tt>uge</tt>: unordered or greater than or equal</li>
3447 <li><tt>ult</tt>: unordered or less than </li>
3448 <li><tt>ule</tt>: unordered or less than or equal</li>
3449 <li><tt>une</tt>: unordered or not equal</li>
3450 <li><tt>uno</tt>: unordered (either nans)</li>
3451 <li><tt>true</tt>: no comparison, always returns true</li>
3453 <p><i>Ordered</i> means that neither operand is a QNAN while
3454 <i>unordered</i> means that either operand may be a QNAN.</p>
3455 <p>The <tt>val1</tt> and <tt>val2</tt> arguments must be
3456 <a href="#t_floating">floating point</a> typed. They must have identical
3459 <p>The '<tt>fcmp</tt>' compares <tt>var1</tt> and <tt>var2</tt> according to
3460 the condition code given as <tt>cond</tt>. The comparison performed always
3461 yields a <a href="#t_primitive">i1</a> result, as follows:
3463 <li><tt>false</tt>: always yields <tt>false</tt>, regardless of operands.</li>
3464 <li><tt>oeq</tt>: yields <tt>true</tt> if both operands are not a QNAN and
3465 <tt>var1</tt> is equal to <tt>var2</tt>.</li>
3466 <li><tt>ogt</tt>: yields <tt>true</tt> if both operands are not a QNAN and
3467 <tt>var1</tt> is greather than <tt>var2</tt>.</li>
3468 <li><tt>oge</tt>: yields <tt>true</tt> if both operands are not a QNAN and
3469 <tt>var1</tt> is greater than or equal to <tt>var2</tt>.</li>
3470 <li><tt>olt</tt>: yields <tt>true</tt> if both operands are not a QNAN and
3471 <tt>var1</tt> is less than <tt>var2</tt>.</li>
3472 <li><tt>ole</tt>: yields <tt>true</tt> if both operands are not a QNAN and
3473 <tt>var1</tt> is less than or equal to <tt>var2</tt>.</li>
3474 <li><tt>one</tt>: yields <tt>true</tt> if both operands are not a QNAN and
3475 <tt>var1</tt> is not equal to <tt>var2</tt>.</li>
3476 <li><tt>ord</tt>: yields <tt>true</tt> if both operands are not a QNAN.</li>
3477 <li><tt>ueq</tt>: yields <tt>true</tt> if either operand is a QNAN or
3478 <tt>var1</tt> is equal to <tt>var2</tt>.</li>
3479 <li><tt>ugt</tt>: yields <tt>true</tt> if either operand is a QNAN or
3480 <tt>var1</tt> is greater than <tt>var2</tt>.</li>
3481 <li><tt>uge</tt>: yields <tt>true</tt> if either operand is a QNAN or
3482 <tt>var1</tt> is greater than or equal to <tt>var2</tt>.</li>
3483 <li><tt>ult</tt>: yields <tt>true</tt> if either operand is a QNAN or
3484 <tt>var1</tt> is less than <tt>var2</tt>.</li>
3485 <li><tt>ule</tt>: yields <tt>true</tt> if either operand is a QNAN or
3486 <tt>var1</tt> is less than or equal to <tt>var2</tt>.</li>
3487 <li><tt>une</tt>: yields <tt>true</tt> if either operand is a QNAN or
3488 <tt>var1</tt> is not equal to <tt>var2</tt>.</li>
3489 <li><tt>uno</tt>: yields <tt>true</tt> if either operand is a QNAN.</li>
3490 <li><tt>true</tt>: always yields <tt>true</tt>, regardless of operands.</li>
3494 <pre> <result> = fcmp oeq float 4.0, 5.0 <i>; yields: result=false</i>
3495 <result> = icmp one float 4.0, 5.0 <i>; yields: result=true</i>
3496 <result> = icmp olt float 4.0, 5.0 <i>; yields: result=true</i>
3497 <result> = icmp ueq double 1.0, 2.0 <i>; yields: result=false</i>
3501 <!-- _______________________________________________________________________ -->
3502 <div class="doc_subsubsection"> <a name="i_phi">'<tt>phi</tt>'
3503 Instruction</a> </div>
3504 <div class="doc_text">
3506 <pre> <result> = phi <ty> [ <val0>, <label0>], ...<br></pre>
3508 <p>The '<tt>phi</tt>' instruction is used to implement the φ node in
3509 the SSA graph representing the function.</p>
3511 <p>The type of the incoming values is specified with the first type
3512 field. After this, the '<tt>phi</tt>' instruction takes a list of pairs
3513 as arguments, with one pair for each predecessor basic block of the
3514 current block. Only values of <a href="#t_firstclass">first class</a>
3515 type may be used as the value arguments to the PHI node. Only labels
3516 may be used as the label arguments.</p>
3517 <p>There must be no non-phi instructions between the start of a basic
3518 block and the PHI instructions: i.e. PHI instructions must be first in
3521 <p>At runtime, the '<tt>phi</tt>' instruction logically takes on the value
3522 specified by the pair corresponding to the predecessor basic block that executed
3523 just prior to the current block.</p>
3525 <pre>Loop: ; Infinite loop that counts from 0 on up...<br> %indvar = phi i32 [ 0, %LoopHeader ], [ %nextindvar, %Loop ]<br> %nextindvar = add i32 %indvar, 1<br> br label %Loop<br></pre>
3528 <!-- _______________________________________________________________________ -->
3529 <div class="doc_subsubsection">
3530 <a name="i_select">'<tt>select</tt>' Instruction</a>
3533 <div class="doc_text">
3538 <result> = select i1 <cond>, <ty> <val1>, <ty> <val2> <i>; yields ty</i>
3544 The '<tt>select</tt>' instruction is used to choose one value based on a
3545 condition, without branching.
3552 The '<tt>select</tt>' instruction requires a boolean value indicating the condition, and two values of the same <a href="#t_firstclass">first class</a> type.
3558 If the boolean condition evaluates to true, the instruction returns the first
3559 value argument; otherwise, it returns the second value argument.
3565 %X = select i1 true, i8 17, i8 42 <i>; yields i8:17</i>
3570 <!-- _______________________________________________________________________ -->
3571 <div class="doc_subsubsection">
3572 <a name="i_call">'<tt>call</tt>' Instruction</a>
3575 <div class="doc_text">
3579 <result> = [tail] call [<a href="#callingconv">cconv</a>] <ty>* <fnptrval>(<param list>)
3584 <p>The '<tt>call</tt>' instruction represents a simple function call.</p>
3588 <p>This instruction requires several arguments:</p>
3592 <p>The optional "tail" marker indicates whether the callee function accesses
3593 any allocas or varargs in the caller. If the "tail" marker is present, the
3594 function call is eligible for tail call optimization. Note that calls may
3595 be marked "tail" even if they do not occur before a <a
3596 href="#i_ret"><tt>ret</tt></a> instruction.
3599 <p>The optional "cconv" marker indicates which <a href="#callingconv">calling
3600 convention</a> the call should use. If none is specified, the call defaults
3601 to using C calling conventions.
3604 <p>'<tt>ty</tt>': shall be the signature of the pointer to function value
3605 being invoked. The argument types must match the types implied by this
3606 signature. This type can be omitted if the function is not varargs and
3607 if the function type does not return a pointer to a function.</p>
3610 <p>'<tt>fnptrval</tt>': An LLVM value containing a pointer to a function to
3611 be invoked. In most cases, this is a direct function invocation, but
3612 indirect <tt>call</tt>s are just as possible, calling an arbitrary pointer
3613 to function value.</p>
3616 <p>'<tt>function args</tt>': argument list whose types match the
3617 function signature argument types. All arguments must be of
3618 <a href="#t_firstclass">first class</a> type. If the function signature
3619 indicates the function accepts a variable number of arguments, the extra
3620 arguments can be specified.</p>
3626 <p>The '<tt>call</tt>' instruction is used to cause control flow to
3627 transfer to a specified function, with its incoming arguments bound to
3628 the specified values. Upon a '<tt><a href="#i_ret">ret</a></tt>'
3629 instruction in the called function, control flow continues with the
3630 instruction after the function call, and the return value of the
3631 function is bound to the result argument. This is a simpler case of
3632 the <a href="#i_invoke">invoke</a> instruction.</p>
3637 %retval = call i32 %test(i32 %argc)
3638 call i32(i8 *, ...) *%printf(i8 * %msg, i32 12, i8 42);
3639 %X = tail call i32 %foo()
3640 %Y = tail call <a href="#callingconv">fastcc</a> i32 %foo()
3645 <!-- _______________________________________________________________________ -->
3646 <div class="doc_subsubsection">
3647 <a name="i_va_arg">'<tt>va_arg</tt>' Instruction</a>
3650 <div class="doc_text">
3655 <resultval> = va_arg <va_list*> <arglist>, <argty>
3660 <p>The '<tt>va_arg</tt>' instruction is used to access arguments passed through
3661 the "variable argument" area of a function call. It is used to implement the
3662 <tt>va_arg</tt> macro in C.</p>
3666 <p>This instruction takes a <tt>va_list*</tt> value and the type of
3667 the argument. It returns a value of the specified argument type and
3668 increments the <tt>va_list</tt> to point to the next argument. The
3669 actual type of <tt>va_list</tt> is target specific.</p>
3673 <p>The '<tt>va_arg</tt>' instruction loads an argument of the specified
3674 type from the specified <tt>va_list</tt> and causes the
3675 <tt>va_list</tt> to point to the next argument. For more information,
3676 see the variable argument handling <a href="#int_varargs">Intrinsic
3679 <p>It is legal for this instruction to be called in a function which does not
3680 take a variable number of arguments, for example, the <tt>vfprintf</tt>
3683 <p><tt>va_arg</tt> is an LLVM instruction instead of an <a
3684 href="#intrinsics">intrinsic function</a> because it takes a type as an
3689 <p>See the <a href="#int_varargs">variable argument processing</a> section.</p>
3693 <!-- *********************************************************************** -->
3694 <div class="doc_section"> <a name="intrinsics">Intrinsic Functions</a> </div>
3695 <!-- *********************************************************************** -->
3697 <div class="doc_text">
3699 <p>LLVM supports the notion of an "intrinsic function". These functions have
3700 well known names and semantics and are required to follow certain restrictions.
3701 Overall, these intrinsics represent an extension mechanism for the LLVM
3702 language that does not require changing all of the transformations in LLVM when
3703 adding to the language (or the bitcode reader/writer, the parser, etc...).</p>
3705 <p>Intrinsic function names must all start with an "<tt>llvm.</tt>" prefix. This
3706 prefix is reserved in LLVM for intrinsic names; thus, function names may not
3707 begin with this prefix. Intrinsic functions must always be external functions:
3708 you cannot define the body of intrinsic functions. Intrinsic functions may
3709 only be used in call or invoke instructions: it is illegal to take the address
3710 of an intrinsic function. Additionally, because intrinsic functions are part
3711 of the LLVM language, it is required if any are added that they be documented
3714 <p>Some intrinsic functions can be overloaded, i.e., the intrinsic represents
3715 a family of functions that perform the same operation but on different data
3716 types. This is most frequent with the integer types. Since LLVM can represent
3717 over 8 million different integer types, there is a way to declare an intrinsic
3718 that can be overloaded based on its arguments. Such an intrinsic will have the
3719 names of its argument types encoded into its function name, each
3720 preceded by a period. For example, the <tt>llvm.ctpop</tt> function can take an
3721 integer of any width. This leads to a family of functions such as
3722 <tt>i32 @llvm.ctpop.i8(i8 %val)</tt> and <tt>i32 @llvm.ctpop.i29(i29 %val)</tt>.
3726 <p>To learn how to add an intrinsic function, please see the
3727 <a href="ExtendingLLVM.html">Extending LLVM Guide</a>.
3732 <!-- ======================================================================= -->
3733 <div class="doc_subsection">
3734 <a name="int_varargs">Variable Argument Handling Intrinsics</a>
3737 <div class="doc_text">
3739 <p>Variable argument support is defined in LLVM with the <a
3740 href="#i_va_arg"><tt>va_arg</tt></a> instruction and these three
3741 intrinsic functions. These functions are related to the similarly
3742 named macros defined in the <tt><stdarg.h></tt> header file.</p>
3744 <p>All of these functions operate on arguments that use a
3745 target-specific value type "<tt>va_list</tt>". The LLVM assembly
3746 language reference manual does not define what this type is, so all
3747 transformations should be prepared to handle these functions regardless of
3750 <p>This example shows how the <a href="#i_va_arg"><tt>va_arg</tt></a>
3751 instruction and the variable argument handling intrinsic functions are
3754 <div class="doc_code">
3756 define i32 @test(i32 %X, ...) {
3757 ; Initialize variable argument processing
3759 %ap2 = bitcast i8** %ap to i8*
3760 call void @llvm.va_start(i8* %ap2)
3762 ; Read a single integer argument
3763 %tmp = va_arg i8** %ap, i32
3765 ; Demonstrate usage of llvm.va_copy and llvm.va_end
3767 %aq2 = bitcast i8** %aq to i8*
3768 call void @llvm.va_copy(i8* %aq2, i8* %ap2)
3769 call void @llvm.va_end(i8* %aq2)
3771 ; Stop processing of arguments.
3772 call void @llvm.va_end(i8* %ap2)
3776 declare void @llvm.va_start(i8*)
3777 declare void @llvm.va_copy(i8*, i8*)
3778 declare void @llvm.va_end(i8*)
3784 <!-- _______________________________________________________________________ -->
3785 <div class="doc_subsubsection">
3786 <a name="int_va_start">'<tt>llvm.va_start</tt>' Intrinsic</a>
3790 <div class="doc_text">
3792 <pre> declare void %llvm.va_start(i8* <arglist>)<br></pre>
3794 <P>The '<tt>llvm.va_start</tt>' intrinsic initializes
3795 <tt>*<arglist></tt> for subsequent use by <tt><a
3796 href="#i_va_arg">va_arg</a></tt>.</p>
3800 <P>The argument is a pointer to a <tt>va_list</tt> element to initialize.</p>
3804 <P>The '<tt>llvm.va_start</tt>' intrinsic works just like the <tt>va_start</tt>
3805 macro available in C. In a target-dependent way, it initializes the
3806 <tt>va_list</tt> element to which the argument points, so that the next call to
3807 <tt>va_arg</tt> will produce the first variable argument passed to the function.
3808 Unlike the C <tt>va_start</tt> macro, this intrinsic does not need to know the
3809 last argument of the function as the compiler can figure that out.</p>
3813 <!-- _______________________________________________________________________ -->
3814 <div class="doc_subsubsection">
3815 <a name="int_va_end">'<tt>llvm.va_end</tt>' Intrinsic</a>
3818 <div class="doc_text">
3820 <pre> declare void @llvm.va_end(i8* <arglist>)<br></pre>
3823 <p>The '<tt>llvm.va_end</tt>' intrinsic destroys <tt>*<arglist></tt>,
3824 which has been initialized previously with <tt><a href="#int_va_start">llvm.va_start</a></tt>
3825 or <tt><a href="#i_va_copy">llvm.va_copy</a></tt>.</p>
3829 <p>The argument is a pointer to a <tt>va_list</tt> to destroy.</p>
3833 <p>The '<tt>llvm.va_end</tt>' intrinsic works just like the <tt>va_end</tt>
3834 macro available in C. In a target-dependent way, it destroys the
3835 <tt>va_list</tt> element to which the argument points. Calls to <a
3836 href="#int_va_start"><tt>llvm.va_start</tt></a> and <a href="#int_va_copy">
3837 <tt>llvm.va_copy</tt></a> must be matched exactly with calls to
3838 <tt>llvm.va_end</tt>.</p>
3842 <!-- _______________________________________________________________________ -->
3843 <div class="doc_subsubsection">
3844 <a name="int_va_copy">'<tt>llvm.va_copy</tt>' Intrinsic</a>
3847 <div class="doc_text">
3852 declare void @llvm.va_copy(i8* <destarglist>, i8* <srcarglist>)
3857 <p>The '<tt>llvm.va_copy</tt>' intrinsic copies the current argument position
3858 from the source argument list to the destination argument list.</p>
3862 <p>The first argument is a pointer to a <tt>va_list</tt> element to initialize.
3863 The second argument is a pointer to a <tt>va_list</tt> element to copy from.</p>
3868 <p>The '<tt>llvm.va_copy</tt>' intrinsic works just like the <tt>va_copy</tt>
3869 macro available in C. In a target-dependent way, it copies the source
3870 <tt>va_list</tt> element into the destination <tt>va_list</tt> element. This
3871 intrinsic is necessary because the <tt><a href="#int_va_start">
3872 llvm.va_start</a></tt> intrinsic may be arbitrarily complex and require, for
3873 example, memory allocation.</p>
3877 <!-- ======================================================================= -->
3878 <div class="doc_subsection">
3879 <a name="int_gc">Accurate Garbage Collection Intrinsics</a>
3882 <div class="doc_text">
3885 LLVM support for <a href="GarbageCollection.html">Accurate Garbage
3886 Collection</a> requires the implementation and generation of these intrinsics.
3887 These intrinsics allow identification of <a href="#int_gcroot">GC roots on the
3888 stack</a>, as well as garbage collector implementations that require <a
3889 href="#int_gcread">read</a> and <a href="#int_gcwrite">write</a> barriers.
3890 Front-ends for type-safe garbage collected languages should generate these
3891 intrinsics to make use of the LLVM garbage collectors. For more details, see <a
3892 href="GarbageCollection.html">Accurate Garbage Collection with LLVM</a>.
3896 <!-- _______________________________________________________________________ -->
3897 <div class="doc_subsubsection">
3898 <a name="int_gcroot">'<tt>llvm.gcroot</tt>' Intrinsic</a>
3901 <div class="doc_text">
3906 declare void @llvm.gcroot(<ty>** %ptrloc, <ty2>* %metadata)
3911 <p>The '<tt>llvm.gcroot</tt>' intrinsic declares the existence of a GC root to
3912 the code generator, and allows some metadata to be associated with it.</p>
3916 <p>The first argument specifies the address of a stack object that contains the
3917 root pointer. The second pointer (which must be either a constant or a global
3918 value address) contains the meta-data to be associated with the root.</p>
3922 <p>At runtime, a call to this intrinsics stores a null pointer into the "ptrloc"
3923 location. At compile-time, the code generator generates information to allow
3924 the runtime to find the pointer at GC safe points.
3930 <!-- _______________________________________________________________________ -->
3931 <div class="doc_subsubsection">
3932 <a name="int_gcread">'<tt>llvm.gcread</tt>' Intrinsic</a>
3935 <div class="doc_text">
3940 declare i8 * @llvm.gcread(i8 * %ObjPtr, i8 ** %Ptr)
3945 <p>The '<tt>llvm.gcread</tt>' intrinsic identifies reads of references from heap
3946 locations, allowing garbage collector implementations that require read
3951 <p>The second argument is the address to read from, which should be an address
3952 allocated from the garbage collector. The first object is a pointer to the
3953 start of the referenced object, if needed by the language runtime (otherwise
3958 <p>The '<tt>llvm.gcread</tt>' intrinsic has the same semantics as a load
3959 instruction, but may be replaced with substantially more complex code by the
3960 garbage collector runtime, as needed.</p>
3965 <!-- _______________________________________________________________________ -->
3966 <div class="doc_subsubsection">
3967 <a name="int_gcwrite">'<tt>llvm.gcwrite</tt>' Intrinsic</a>
3970 <div class="doc_text">
3975 declare void @llvm.gcwrite(i8 * %P1, i8 * %Obj, i8 ** %P2)
3980 <p>The '<tt>llvm.gcwrite</tt>' intrinsic identifies writes of references to heap
3981 locations, allowing garbage collector implementations that require write
3982 barriers (such as generational or reference counting collectors).</p>
3986 <p>The first argument is the reference to store, the second is the start of the
3987 object to store it to, and the third is the address of the field of Obj to
3988 store to. If the runtime does not require a pointer to the object, Obj may be
3993 <p>The '<tt>llvm.gcwrite</tt>' intrinsic has the same semantics as a store
3994 instruction, but may be replaced with substantially more complex code by the
3995 garbage collector runtime, as needed.</p>
4001 <!-- ======================================================================= -->
4002 <div class="doc_subsection">
4003 <a name="int_codegen">Code Generator Intrinsics</a>
4006 <div class="doc_text">
4008 These intrinsics are provided by LLVM to expose special features that may only
4009 be implemented with code generator support.
4014 <!-- _______________________________________________________________________ -->
4015 <div class="doc_subsubsection">
4016 <a name="int_returnaddress">'<tt>llvm.returnaddress</tt>' Intrinsic</a>
4019 <div class="doc_text">
4023 declare i8 *@llvm.returnaddress(i32 <level>)
4029 The '<tt>llvm.returnaddress</tt>' intrinsic attempts to compute a
4030 target-specific value indicating the return address of the current function
4031 or one of its callers.
4037 The argument to this intrinsic indicates which function to return the address
4038 for. Zero indicates the calling function, one indicates its caller, etc. The
4039 argument is <b>required</b> to be a constant integer value.
4045 The '<tt>llvm.returnaddress</tt>' intrinsic either returns a pointer indicating
4046 the return address of the specified call frame, or zero if it cannot be
4047 identified. The value returned by this intrinsic is likely to be incorrect or 0
4048 for arguments other than zero, so it should only be used for debugging purposes.
4052 Note that calling this intrinsic does not prevent function inlining or other
4053 aggressive transformations, so the value returned may not be that of the obvious
4054 source-language caller.
4059 <!-- _______________________________________________________________________ -->
4060 <div class="doc_subsubsection">
4061 <a name="int_frameaddress">'<tt>llvm.frameaddress</tt>' Intrinsic</a>
4064 <div class="doc_text">
4068 declare i8 *@llvm.frameaddress(i32 <level>)
4074 The '<tt>llvm.frameaddress</tt>' intrinsic attempts to return the
4075 target-specific frame pointer value for the specified stack frame.
4081 The argument to this intrinsic indicates which function to return the frame
4082 pointer for. Zero indicates the calling function, one indicates its caller,
4083 etc. The argument is <b>required</b> to be a constant integer value.
4089 The '<tt>llvm.frameaddress</tt>' intrinsic either returns a pointer indicating
4090 the frame address of the specified call frame, or zero if it cannot be
4091 identified. The value returned by this intrinsic is likely to be incorrect or 0
4092 for arguments other than zero, so it should only be used for debugging purposes.
4096 Note that calling this intrinsic does not prevent function inlining or other
4097 aggressive transformations, so the value returned may not be that of the obvious
4098 source-language caller.
4102 <!-- _______________________________________________________________________ -->
4103 <div class="doc_subsubsection">
4104 <a name="int_stacksave">'<tt>llvm.stacksave</tt>' Intrinsic</a>
4107 <div class="doc_text">
4111 declare i8 *@llvm.stacksave()
4117 The '<tt>llvm.stacksave</tt>' intrinsic is used to remember the current state of
4118 the function stack, for use with <a href="#int_stackrestore">
4119 <tt>llvm.stackrestore</tt></a>. This is useful for implementing language
4120 features like scoped automatic variable sized arrays in C99.
4126 This intrinsic returns a opaque pointer value that can be passed to <a
4127 href="#int_stackrestore"><tt>llvm.stackrestore</tt></a>. When an
4128 <tt>llvm.stackrestore</tt> intrinsic is executed with a value saved from
4129 <tt>llvm.stacksave</tt>, it effectively restores the state of the stack to the
4130 state it was in when the <tt>llvm.stacksave</tt> intrinsic executed. In
4131 practice, this pops any <a href="#i_alloca">alloca</a> blocks from the stack
4132 that were allocated after the <tt>llvm.stacksave</tt> was executed.
4137 <!-- _______________________________________________________________________ -->
4138 <div class="doc_subsubsection">
4139 <a name="int_stackrestore">'<tt>llvm.stackrestore</tt>' Intrinsic</a>
4142 <div class="doc_text">
4146 declare void @llvm.stackrestore(i8 * %ptr)
4152 The '<tt>llvm.stackrestore</tt>' intrinsic is used to restore the state of
4153 the function stack to the state it was in when the corresponding <a
4154 href="#int_stacksave"><tt>llvm.stacksave</tt></a> intrinsic executed. This is
4155 useful for implementing language features like scoped automatic variable sized
4162 See the description for <a href="#int_stacksave"><tt>llvm.stacksave</tt></a>.
4168 <!-- _______________________________________________________________________ -->
4169 <div class="doc_subsubsection">
4170 <a name="int_prefetch">'<tt>llvm.prefetch</tt>' Intrinsic</a>
4173 <div class="doc_text">
4177 declare void @llvm.prefetch(i8 * <address>,
4178 i32 <rw>, i32 <locality>)
4185 The '<tt>llvm.prefetch</tt>' intrinsic is a hint to the code generator to insert
4186 a prefetch instruction if supported; otherwise, it is a noop. Prefetches have
4188 effect on the behavior of the program but can change its performance
4195 <tt>address</tt> is the address to be prefetched, <tt>rw</tt> is the specifier
4196 determining if the fetch should be for a read (0) or write (1), and
4197 <tt>locality</tt> is a temporal locality specifier ranging from (0) - no
4198 locality, to (3) - extremely local keep in cache. The <tt>rw</tt> and
4199 <tt>locality</tt> arguments must be constant integers.
4205 This intrinsic does not modify the behavior of the program. In particular,
4206 prefetches cannot trap and do not produce a value. On targets that support this
4207 intrinsic, the prefetch can provide hints to the processor cache for better
4213 <!-- _______________________________________________________________________ -->
4214 <div class="doc_subsubsection">
4215 <a name="int_pcmarker">'<tt>llvm.pcmarker</tt>' Intrinsic</a>
4218 <div class="doc_text">
4222 declare void @llvm.pcmarker( i32 <id> )
4229 The '<tt>llvm.pcmarker</tt>' intrinsic is a method to export a Program Counter
4231 code to simulators and other tools. The method is target specific, but it is
4232 expected that the marker will use exported symbols to transmit the PC of the marker.
4233 The marker makes no guarantees that it will remain with any specific instruction
4234 after optimizations. It is possible that the presence of a marker will inhibit
4235 optimizations. The intended use is to be inserted after optimizations to allow
4236 correlations of simulation runs.
4242 <tt>id</tt> is a numerical id identifying the marker.
4248 This intrinsic does not modify the behavior of the program. Backends that do not
4249 support this intrinisic may ignore it.
4254 <!-- _______________________________________________________________________ -->
4255 <div class="doc_subsubsection">
4256 <a name="int_readcyclecounter">'<tt>llvm.readcyclecounter</tt>' Intrinsic</a>
4259 <div class="doc_text">
4263 declare i64 @llvm.readcyclecounter( )
4270 The '<tt>llvm.readcyclecounter</tt>' intrinsic provides access to the cycle
4271 counter register (or similar low latency, high accuracy clocks) on those targets
4272 that support it. On X86, it should map to RDTSC. On Alpha, it should map to RPCC.
4273 As the backing counters overflow quickly (on the order of 9 seconds on alpha), this
4274 should only be used for small timings.
4280 When directly supported, reading the cycle counter should not modify any memory.
4281 Implementations are allowed to either return a application specific value or a
4282 system wide value. On backends without support, this is lowered to a constant 0.
4287 <!-- ======================================================================= -->
4288 <div class="doc_subsection">
4289 <a name="int_libc">Standard C Library Intrinsics</a>
4292 <div class="doc_text">
4294 LLVM provides intrinsics for a few important standard C library functions.
4295 These intrinsics allow source-language front-ends to pass information about the
4296 alignment of the pointer arguments to the code generator, providing opportunity
4297 for more efficient code generation.
4302 <!-- _______________________________________________________________________ -->
4303 <div class="doc_subsubsection">
4304 <a name="int_memcpy">'<tt>llvm.memcpy</tt>' Intrinsic</a>
4307 <div class="doc_text">
4311 declare void @llvm.memcpy.i32(i8 * <dest>, i8 * <src>,
4312 i32 <len>, i32 <align>)
4313 declare void @llvm.memcpy.i64(i8 * <dest>, i8 * <src>,
4314 i64 <len>, i32 <align>)
4320 The '<tt>llvm.memcpy.*</tt>' intrinsics copy a block of memory from the source
4321 location to the destination location.
4325 Note that, unlike the standard libc function, the <tt>llvm.memcpy.*</tt>
4326 intrinsics do not return a value, and takes an extra alignment argument.
4332 The first argument is a pointer to the destination, the second is a pointer to
4333 the source. The third argument is an integer argument
4334 specifying the number of bytes to copy, and the fourth argument is the alignment
4335 of the source and destination locations.
4339 If the call to this intrinisic has an alignment value that is not 0 or 1, then
4340 the caller guarantees that both the source and destination pointers are aligned
4347 The '<tt>llvm.memcpy.*</tt>' intrinsics copy a block of memory from the source
4348 location to the destination location, which are not allowed to overlap. It
4349 copies "len" bytes of memory over. If the argument is known to be aligned to
4350 some boundary, this can be specified as the fourth argument, otherwise it should
4356 <!-- _______________________________________________________________________ -->
4357 <div class="doc_subsubsection">
4358 <a name="int_memmove">'<tt>llvm.memmove</tt>' Intrinsic</a>
4361 <div class="doc_text">
4365 declare void @llvm.memmove.i32(i8 * <dest>, i8 * <src>,
4366 i32 <len>, i32 <align>)
4367 declare void @llvm.memmove.i64(i8 * <dest>, i8 * <src>,
4368 i64 <len>, i32 <align>)
4374 The '<tt>llvm.memmove.*</tt>' intrinsics move a block of memory from the source
4375 location to the destination location. It is similar to the
4376 '<tt>llvm.memcmp</tt>' intrinsic but allows the two memory locations to overlap.
4380 Note that, unlike the standard libc function, the <tt>llvm.memmove.*</tt>
4381 intrinsics do not return a value, and takes an extra alignment argument.
4387 The first argument is a pointer to the destination, the second is a pointer to
4388 the source. The third argument is an integer argument
4389 specifying the number of bytes to copy, and the fourth argument is the alignment
4390 of the source and destination locations.
4394 If the call to this intrinisic has an alignment value that is not 0 or 1, then
4395 the caller guarantees that the source and destination pointers are aligned to
4402 The '<tt>llvm.memmove.*</tt>' intrinsics copy a block of memory from the source
4403 location to the destination location, which may overlap. It
4404 copies "len" bytes of memory over. If the argument is known to be aligned to
4405 some boundary, this can be specified as the fourth argument, otherwise it should
4411 <!-- _______________________________________________________________________ -->
4412 <div class="doc_subsubsection">
4413 <a name="int_memset">'<tt>llvm.memset.*</tt>' Intrinsics</a>
4416 <div class="doc_text">
4420 declare void @llvm.memset.i32(i8 * <dest>, i8 <val>,
4421 i32 <len>, i32 <align>)
4422 declare void @llvm.memset.i64(i8 * <dest>, i8 <val>,
4423 i64 <len>, i32 <align>)
4429 The '<tt>llvm.memset.*</tt>' intrinsics fill a block of memory with a particular
4434 Note that, unlike the standard libc function, the <tt>llvm.memset</tt> intrinsic
4435 does not return a value, and takes an extra alignment argument.
4441 The first argument is a pointer to the destination to fill, the second is the
4442 byte value to fill it with, the third argument is an integer
4443 argument specifying the number of bytes to fill, and the fourth argument is the
4444 known alignment of destination location.
4448 If the call to this intrinisic has an alignment value that is not 0 or 1, then
4449 the caller guarantees that the destination pointer is aligned to that boundary.
4455 The '<tt>llvm.memset.*</tt>' intrinsics fill "len" bytes of memory starting at
4457 destination location. If the argument is known to be aligned to some boundary,
4458 this can be specified as the fourth argument, otherwise it should be set to 0 or
4464 <!-- _______________________________________________________________________ -->
4465 <div class="doc_subsubsection">
4466 <a name="int_sqrt">'<tt>llvm.sqrt.*</tt>' Intrinsic</a>
4469 <div class="doc_text">
4473 declare float @llvm.sqrt.f32(float %Val)
4474 declare double @llvm.sqrt.f64(double %Val)
4480 The '<tt>llvm.sqrt</tt>' intrinsics return the sqrt of the specified operand,
4481 returning the same value as the libm '<tt>sqrt</tt>' function would. Unlike
4482 <tt>sqrt</tt> in libm, however, <tt>llvm.sqrt</tt> has undefined behavior for
4483 negative numbers (which allows for better optimization).
4489 The argument and return value are floating point numbers of the same type.
4495 This function returns the sqrt of the specified operand if it is a nonnegative
4496 floating point number.
4500 <!-- _______________________________________________________________________ -->
4501 <div class="doc_subsubsection">
4502 <a name="int_powi">'<tt>llvm.powi.*</tt>' Intrinsic</a>
4505 <div class="doc_text">
4509 declare float @llvm.powi.f32(float %Val, i32 %power)
4510 declare double @llvm.powi.f64(double %Val, i32 %power)
4516 The '<tt>llvm.powi.*</tt>' intrinsics return the first operand raised to the
4517 specified (positive or negative) power. The order of evaluation of
4518 multiplications is not defined.
4524 The second argument is an integer power, and the first is a value to raise to
4531 This function returns the first value raised to the second power with an
4532 unspecified sequence of rounding operations.</p>
4536 <!-- ======================================================================= -->
4537 <div class="doc_subsection">
4538 <a name="int_manip">Bit Manipulation Intrinsics</a>
4541 <div class="doc_text">
4543 LLVM provides intrinsics for a few important bit manipulation operations.
4544 These allow efficient code generation for some algorithms.
4549 <!-- _______________________________________________________________________ -->
4550 <div class="doc_subsubsection">
4551 <a name="int_bswap">'<tt>llvm.bswap.*</tt>' Intrinsics</a>
4554 <div class="doc_text">
4557 <p>This is an overloaded intrinsic function. You can use bswap on any integer
4558 type that is an even number of bytes (i.e. BitWidth % 16 == 0). Note the suffix
4559 that includes the type for the result and the operand.
4561 declare i16 @llvm.bswap.i16.i16(i16 <id>)
4562 declare i32 @llvm.bswap.i32.i32(i32 <id>)
4563 declare i64 @llvm.bswap.i64.i64(i64 <id>)
4569 The '<tt>llvm.bswap</tt>' family of intrinsics is used to byte swap integer
4570 values with an even number of bytes (positive multiple of 16 bits). These are
4571 useful for performing operations on data that is not in the target's native
4578 The <tt>llvm.bswap.16.i16</tt> intrinsic returns an i16 value that has the high
4579 and low byte of the input i16 swapped. Similarly, the <tt>llvm.bswap.i32</tt>
4580 intrinsic returns an i32 value that has the four bytes of the input i32
4581 swapped, so that if the input bytes are numbered 0, 1, 2, 3 then the returned
4582 i32 will have its bytes in 3, 2, 1, 0 order. The <tt>llvm.bswap.i48.i48</tt>,
4583 <tt>llvm.bswap.i64.i64</tt> and other intrinsics extend this concept to
4584 additional even-byte lengths (6 bytes, 8 bytes and more, respectively).
4589 <!-- _______________________________________________________________________ -->
4590 <div class="doc_subsubsection">
4591 <a name="int_ctpop">'<tt>llvm.ctpop.*</tt>' Intrinsic</a>
4594 <div class="doc_text">
4597 <p>This is an overloaded intrinsic. You can use llvm.ctpop on any integer bit
4598 width. Not all targets support all bit widths however.
4600 declare i32 @llvm.ctpop.i8 (i8 <src>)
4601 declare i32 @llvm.ctpop.i16(i16 <src>)
4602 declare i32 @llvm.ctpop.i32(i32 <src>)
4603 declare i32 @llvm.ctpop.i64(i64 <src>)
4604 declare i32 @llvm.ctpop.i256(i256 <src>)
4610 The '<tt>llvm.ctpop</tt>' family of intrinsics counts the number of bits set in a
4617 The only argument is the value to be counted. The argument may be of any
4618 integer type. The return type must match the argument type.
4624 The '<tt>llvm.ctpop</tt>' intrinsic counts the 1's in a variable.
4628 <!-- _______________________________________________________________________ -->
4629 <div class="doc_subsubsection">
4630 <a name="int_ctlz">'<tt>llvm.ctlz.*</tt>' Intrinsic</a>
4633 <div class="doc_text">
4636 <p>This is an overloaded intrinsic. You can use <tt>llvm.ctlz</tt> on any
4637 integer bit width. Not all targets support all bit widths however.
4639 declare i32 @llvm.ctlz.i8 (i8 <src>)
4640 declare i32 @llvm.ctlz.i16(i16 <src>)
4641 declare i32 @llvm.ctlz.i32(i32 <src>)
4642 declare i32 @llvm.ctlz.i64(i64 <src>)
4643 declare i32 @llvm.ctlz.i256(i256 <src>)
4649 The '<tt>llvm.ctlz</tt>' family of intrinsic functions counts the number of
4650 leading zeros in a variable.
4656 The only argument is the value to be counted. The argument may be of any
4657 integer type. The return type must match the argument type.
4663 The '<tt>llvm.ctlz</tt>' intrinsic counts the leading (most significant) zeros
4664 in a variable. If the src == 0 then the result is the size in bits of the type
4665 of src. For example, <tt>llvm.ctlz(i32 2) = 30</tt>.
4671 <!-- _______________________________________________________________________ -->
4672 <div class="doc_subsubsection">
4673 <a name="int_cttz">'<tt>llvm.cttz.*</tt>' Intrinsic</a>
4676 <div class="doc_text">
4679 <p>This is an overloaded intrinsic. You can use <tt>llvm.cttz</tt> on any
4680 integer bit width. Not all targets support all bit widths however.
4682 declare i32 @llvm.cttz.i8 (i8 <src>)
4683 declare i32 @llvm.cttz.i16(i16 <src>)
4684 declare i32 @llvm.cttz.i32(i32 <src>)
4685 declare i32 @llvm.cttz.i64(i64 <src>)
4686 declare i32 @llvm.cttz.i256(i256 <src>)
4692 The '<tt>llvm.cttz</tt>' family of intrinsic functions counts the number of
4699 The only argument is the value to be counted. The argument may be of any
4700 integer type. The return type must match the argument type.
4706 The '<tt>llvm.cttz</tt>' intrinsic counts the trailing (least significant) zeros
4707 in a variable. If the src == 0 then the result is the size in bits of the type
4708 of src. For example, <tt>llvm.cttz(2) = 1</tt>.
4712 <!-- _______________________________________________________________________ -->
4713 <div class="doc_subsubsection">
4714 <a name="int_part_select">'<tt>llvm.part.select.*</tt>' Intrinsic</a>
4717 <div class="doc_text">
4720 <p>This is an overloaded intrinsic. You can use <tt>llvm.part.select</tt>
4721 on any integer bit width.
4723 declare i17 @llvm.part.select.i17.i17 (i17 %val, i32 %loBit, i32 %hiBit)
4724 declare i29 @llvm.part.select.i29.i29 (i29 %val, i32 %loBit, i32 %hiBit)
4728 <p>The '<tt>llvm.part.select</tt>' family of intrinsic functions selects a
4729 range of bits from an integer value and returns them in the same bit width as
4730 the original value.</p>
4733 <p>The first argument, <tt>%val</tt> and the result may be integer types of
4734 any bit width but they must have the same bit width. The second and third
4735 arguments must be <tt>i32</tt> type since they specify only a bit index.</p>
4738 <p>The operation of the '<tt>llvm.part.select</tt>' intrinsic has two modes
4739 of operation: forwards and reverse. If <tt>%loBit</tt> is greater than
4740 <tt>%hiBits</tt> then the intrinsic operates in reverse mode. Otherwise it
4741 operates in forward mode.</p>
4742 <p>In forward mode, this intrinsic is the equivalent of shifting <tt>%val</tt>
4743 right by <tt>%loBit</tt> bits and then ANDing it with a mask with
4744 only the <tt>%hiBit - %loBit</tt> bits set, as follows:</p>
4746 <li>The <tt>%val</tt> is shifted right (LSHR) by the number of bits specified
4747 by <tt>%loBits</tt>. This normalizes the value to the low order bits.</li>
4748 <li>The <tt>%loBits</tt> value is subtracted from the <tt>%hiBits</tt> value
4749 to determine the number of bits to retain.</li>
4750 <li>A mask of the retained bits is created by shifting a -1 value.</li>
4751 <li>The mask is ANDed with <tt>%val</tt> to produce the result.
4753 <p>In reverse mode, a similar computation is made except that the bits are
4754 returned in the reverse order. So, for example, if <tt>X</tt> has the value
4755 <tt>i16 0x0ACF (101011001111)</tt> and we apply
4756 <tt>part.select(i16 X, 8, 3)</tt> to it, we get back the value
4757 <tt>i16 0x0026 (000000100110)</tt>.</p>
4760 <div class="doc_subsubsection">
4761 <a name="int_part_set">'<tt>llvm.part.set.*</tt>' Intrinsic</a>
4764 <div class="doc_text">
4767 <p>This is an overloaded intrinsic. You can use <tt>llvm.part.set</tt>
4768 on any integer bit width.
4770 declare i17 @llvm.part.set.i17.i17.i9 (i17 %val, i9 %repl, i32 %lo, i32 %hi)
4771 declare i29 @llvm.part.set.i29.i29.i9 (i29 %val, i9 %repl, i32 %lo, i32 %hi)
4775 <p>The '<tt>llvm.part.set</tt>' family of intrinsic functions replaces a range
4776 of bits in an integer value with another integer value. It returns the integer
4777 with the replaced bits.</p>
4780 <p>The first argument, <tt>%val</tt> and the result may be integer types of
4781 any bit width but they must have the same bit width. <tt>%val</tt> is the value
4782 whose bits will be replaced. The second argument, <tt>%repl</tt> may be an
4783 integer of any bit width. The third and fourth arguments must be <tt>i32</tt>
4784 type since they specify only a bit index.</p>
4787 <p>The operation of the '<tt>llvm.part.set</tt>' intrinsic has two modes
4788 of operation: forwards and reverse. If <tt>%lo</tt> is greater than
4789 <tt>%hi</tt> then the intrinsic operates in reverse mode. Otherwise it
4790 operates in forward mode.</p>
4791 <p>For both modes, the <tt>%repl</tt> value is prepared for use by either
4792 truncating it down to the size of the replacement area or zero extending it
4793 up to that size.</p>
4794 <p>In forward mode, the bits between <tt>%lo</tt> and <tt>%hi</tt> (inclusive)
4795 are replaced with corresponding bits from <tt>%repl</tt>. That is the 0th bit
4796 in <tt>%repl</tt> replaces the <tt>%lo</tt>th bit in <tt>%val</tt> and etc. up
4797 to the <tt>%hi</tt>th bit.
4798 <p>In reverse mode, a similar computation is made except that the bits are
4799 reversed. That is, the <tt>0</tt>th bit in <tt>%repl</tt> replaces the
4800 <tt>%hi</tt> bit in <tt>%val</tt> and etc. down to the <tt>%lo</tt>th bit.
4803 llvm.part.set(0xFFFF, 0, 4, 7) -> 0xFF0F
4804 llvm.part.set(0xFFFF, 0, 7, 4) -> 0xFF0F
4805 llvm.part.set(0xFFFF, 1, 7, 4) -> 0xFF8F
4806 llvm.part.set(0xFFFF, F, 8, 3) -> 0xFFE7
4807 llvm.part.set(0xFFFF, 0, 3, 8) -> 0xFE07
4811 <!-- ======================================================================= -->
4812 <div class="doc_subsection">
4813 <a name="int_debugger">Debugger Intrinsics</a>
4816 <div class="doc_text">
4818 The LLVM debugger intrinsics (which all start with <tt>llvm.dbg.</tt> prefix),
4819 are described in the <a
4820 href="SourceLevelDebugging.html#format_common_intrinsics">LLVM Source Level
4821 Debugging</a> document.
4826 <!-- ======================================================================= -->
4827 <div class="doc_subsection">
4828 <a name="int_eh">Exception Handling Intrinsics</a>
4831 <div class="doc_text">
4832 <p> The LLVM exception handling intrinsics (which all start with
4833 <tt>llvm.eh.</tt> prefix), are described in the <a
4834 href="ExceptionHandling.html#format_common_intrinsics">LLVM Exception
4835 Handling</a> document. </p>
4838 <!-- ======================================================================= -->
4839 <div class="doc_subsection">
4840 <a name="int_atomics">Atomic Operations and Synchronization Intrinsics</a>
4843 <div class="doc_text">
4845 These intrinsic functions expand the "universal IR" of LLVM to represent
4846 hardware constructs for atomic operations and memory synchronization. This
4847 provides an interface to the hardware, not an interface to the programmer. It
4848 is aimed at a low enough level to allow any programming models or APIs which
4849 need atomic behaviors to map cleanly onto it. It is also modeled primarily on
4850 hardware behavior. Just as hardware provides a "universal IR" for source
4851 languages, it also provides a starting point for developing a "universal"
4852 atomic operation and synchronization IR.
4855 These do <em>not</em> form an API such as high-level threading libraries,
4856 software transaction memory systems, atomic primitives, and intrinsic
4857 functions as found in BSD, GNU libc, atomic_ops, APR, and other system and
4858 application libraries. The hardware interface provided by LLVM should allow
4859 a clean implementation of all of these APIs and parallel programming models.
4860 No one model or paradigm should be selected above others unless the hardware
4861 itself ubiquitously does so.
4865 <!-- _______________________________________________________________________ -->
4866 <div class="doc_subsubsection">
4867 <a name="int_lcs">'<tt>llvm.atomic.lcs.*</tt>' Intrinsic</a>
4869 <div class="doc_text">
4872 This is an overloaded intrinsic. You can use <tt>llvm.atomic.lcs</tt> on any
4873 integer bit width. Not all targets support all bit widths however.</p>
4875 declare i8 @llvm.atomic.lcs.i8.i8p.i8.i8( i8* <ptr>, i8 <cmp>, i8 <val> )
4876 declare i16 @llvm.atomic.lcs.i16.i16p.i16.i16( i16* <ptr>, i16 <cmp>, i16 <val> )
4877 declare i32 @llvm.atomic.lcs.i32.i32p.i32.i32( i32* <ptr>, i32 <cmp>, i32 <val> )
4878 declare i64 @llvm.atomic.lcs.i64.i64p.i64.i64( i64* <ptr>, i64 <cmp>, i64 <val> )
4882 This loads a value in memory and compares it to a given value. If they are
4883 equal, it stores a new value into the memory.
4887 The <tt>llvm.atomic.lcs</tt> intrinsic takes three arguments. The result as
4888 well as both <tt>cmp</tt> and <tt>val</tt> must be integer values with the
4889 same bit width. The <tt>ptr</tt> argument must be a pointer to a value of
4890 this integer type. While any bit width integer may be used, targets may only
4891 lower representations they support in hardware.
4895 This entire intrinsic must be executed atomically. It first loads the value
4896 in memory pointed to by <tt>ptr</tt> and compares it with the value
4897 <tt>cmp</tt>. If they are equal, <tt>val</tt> is stored into the memory. The
4898 loaded value is yielded in all cases. This provides the equivalent of an
4899 atomic compare-and-swap operation within the SSA framework.
4906 %val1 = add i32 4, 4
4907 %result1 = call i32 @llvm.atomic.lcs( i32* %ptr, i32 4, %val1 )
4908 <i>; yields {i32}:result1 = 4</i>
4909 %stored1 = icmp eq i32 %result1, 4 <i>; yields {i1}:stored1 = true</i>
4910 %memval1 = load i32* %ptr <i>; yields {i32}:memval1 = 8</i>
4912 %val2 = add i32 1, 1
4913 %result2 = call i32 @llvm.atomic.lcs( i32* %ptr, i32 5, %val2 )
4914 <i>; yields {i32}:result2 = 8</i>
4915 %stored2 = icmp eq i32 %result2, 5 <i>; yields {i1}:stored2 = false</i>
4916 %memval2 = load i32* %ptr <i>; yields {i32}:memval2 = 8</i>
4920 <!-- _______________________________________________________________________ -->
4921 <div class="doc_subsubsection">
4922 <a name="int_ls">'<tt>llvm.atomic.ls.*</tt>' Intrinsic</a>
4924 <div class="doc_text">
4927 This is an overloaded intrinsic. You can use <tt>llvm.atomic.ls</tt> on any
4928 integer bit width. Not all targets support all bit widths however.</p>
4930 declare i8 @llvm.atomic.ls.i8.i8p.i8( i8* <ptr>, i8 <val> )
4931 declare i16 @llvm.atomic.ls.i16.i16p.i16( i16* <ptr>, i16 <val> )
4932 declare i32 @llvm.atomic.ls.i32.i32p.i32( i32* <ptr>, i32 <val> )
4933 declare i64 @llvm.atomic.ls.i64.i64p.i64( i64* <ptr>, i64 <val> )
4937 This intrinsic loads the value stored in memory at <tt>ptr</tt> and yields
4938 the value from memory. It then stores the value in <tt>val</tt> in the memory
4943 The <tt>llvm.atomic.ls</tt> intrinsic takes two arguments. Both the
4944 <tt>val</tt> argument and the result must be integers of the same bit width.
4945 The first argument, <tt>ptr</tt>, must be a pointer to a value of this
4946 integer type. The targets may only lower integer representations they
4951 This intrinsic loads the value pointed to by <tt>ptr</tt>, yields it, and
4952 stores <tt>val</tt> back into <tt>ptr</tt> atomically. This provides the
4953 equivalent of an atomic swap operation within the SSA framework.
4960 %val1 = add i32 4, 4
4961 %result1 = call i32 @llvm.atomic.ls( i32* %ptr, i32 %val1 )
4962 <i>; yields {i32}:result1 = 4</i>
4963 %stored1 = icmp eq i32 %result1, 4 <i>; yields {i1}:stored1 = true</i>
4964 %memval1 = load i32* %ptr <i>; yields {i32}:memval1 = 8</i>
4966 %val2 = add i32 1, 1
4967 %result2 = call i32 @llvm.atomic.ls( i32* %ptr, i32 %val2 )
4968 <i>; yields {i32}:result2 = 8</i>
4969 %stored2 = icmp eq i32 %result2, 8 <i>; yields {i1}:stored2 = true</i>
4970 %memval2 = load i32* %ptr <i>; yields {i32}:memval2 = 2</i>
4974 <!-- _______________________________________________________________________ -->
4975 <div class="doc_subsubsection">
4976 <a name="int_las">'<tt>llvm.atomic.las.*</tt>' Intrinsic</a>
4978 <div class="doc_text">
4981 This is an overloaded intrinsic. You can use <tt>llvm.atomic.las</tt> on any
4982 integer bit width. Not all targets support all bit widths however.</p>
4984 declare i8 @llvm.atomic.las.i8.i8p.i8( i8* <ptr>, i8 <delta> )
4985 declare i16 @llvm.atomic.las.i16.i16p.i16( i16* <ptr>, i16 <delta> )
4986 declare i32 @llvm.atomic.las.i32.i32p.i32( i32* <ptr>, i32 <delta> )
4987 declare i64 @llvm.atomic.las.i64.i64p.i64( i64* <ptr>, i64 <delta> )
4991 This intrinsic adds <tt>delta</tt> to the value stored in memory at
4992 <tt>ptr</tt>. It yields the original value at <tt>ptr</tt>.
4996 The intrinsic takes two arguments, the first a pointer to an integer value
4997 and the second an integer value. The result is also an integer value. These
4998 integer types can have any bit width, but they must all have the same bit
4999 width. The targets may only lower integer representations they support.
5003 This intrinsic does a series of operations atomically. It first loads the
5004 value stored at <tt>ptr</tt>. It then adds <tt>delta</tt>, stores the result
5005 to <tt>ptr</tt>. It yields the original value stored at <tt>ptr</tt>.
5011 %result1 = call i32 @llvm.atomic.las( i32* %ptr, i32 4 )
5012 <i>; yields {i32}:result1 = 4</i>
5013 %result2 = call i32 @llvm.atomic.las( i32* %ptr, i32 2 )
5014 <i>; yields {i32}:result2 = 8</i>
5015 %result3 = call i32 @llvm.atomic.las( i32* %ptr, i32 5 )
5016 <i>; yields {i32}:result3 = 10</i>
5017 %memval = load i32* %ptr <i>; yields {i32}:memval1 = 15</i>
5021 <!-- _______________________________________________________________________ -->
5022 <div class="doc_subsubsection">
5023 <a name="int_lss">'<tt>llvm.atomic.lss.*</tt>' Intrinsic</a>
5025 <div class="doc_text">
5028 This is an overloaded intrinsic. You can use <tt>llvm.atomic.lss</tt> on any
5029 integer bit width. Not all targets support all bit widths however.</p>
5031 declare i8 @llvm.atomic.lss.i8.i8.i8( i8* <ptr>, i8 <delta> )
5032 declare i16 @llvm.atomic.lss.i16.i16.i16( i16* <ptr>, i16 <delta> )
5033 declare i32 @llvm.atomic.lss.i32.i32.i32( i32* <ptr>, i32 <delta> )
5034 declare i64 @llvm.atomic.lss.i64.i64.i64( i64* <ptr>, i64 <delta> )
5038 This intrinsic subtracts <tt>delta</tt> from the value stored in memory at
5039 <tt>ptr</tt>. It yields the original value at <tt>ptr</tt>.
5043 The intrinsic takes two arguments, the first a pointer to an integer value
5044 and the second an integer value. The result is also an integer value. These
5045 integer types can have any bit width, but they must all have the same bit
5046 width. The targets may only lower integer representations they support.
5050 This intrinsic does a series of operations atomically. It first loads the
5051 value stored at <tt>ptr</tt>. It then subtracts <tt>delta</tt>,
5052 stores the result to <tt>ptr</tt>. It yields the original value stored
5059 %result1 = call i32 @llvm.atomic.lss( i32* %ptr, i32 4 )
5060 <i>; yields {i32}:result1 = 32</i>
5061 %result2 = call i32 @llvm.atomic.lss( i32* %ptr, i32 2 )
5062 <i>; yields {i32}:result2 = 28</i>
5063 %result3 = call i32 @llvm.atomic.lss( i32* %ptr, i32 5 )
5064 <i>; yields {i32}:result3 = 26</i>
5065 %memval = load i32* %ptr <i>; yields {i32}:memval1 = 21</i>
5069 <!-- _______________________________________________________________________ -->
5070 <div class="doc_subsubsection">
5071 <a name="int_memory_barrier">'<tt>llvm.memory.barrier</tt>' Intrinsic</a>
5073 <div class="doc_text">
5076 declare void @llvm.memory.barrier( i1 <ll>, i1 <ls>, i1 <sl>, i1 <ss> )
5080 The <tt>llvm.memory.barrier</tt> intrinsic guarantees ordering between
5081 specific pairs of memory access types.
5085 The <tt>llvm.memory.barrier</tt> intrinsic requires four boolean arguments.
5086 Each argument enables a specific barrier as listed below.
5089 <li><tt>ll</tt>: load-load barrier</li>
5090 <li><tt>ls</tt>: load-store barrier</li>
5091 <li><tt>sl</tt>: store-load barrier</li>
5092 <li><tt>ss</tt>: store-store barrier</li>
5096 This intrinsic causes the system to enforce some ordering constraints upon
5097 the loads and stores of the program. This barrier does not indicate
5098 <em>when</em> any events will occur, it only enforces an <em>order</em> in
5099 which they occur. For any of the specified pairs of load and store operations
5100 (f.ex. load-load, or store-load), all of the first operations preceding the
5101 barrier will complete before any of the second operations succeeding the
5102 barrier begin. Specifically the semantics for each pairing is as follows:
5105 <li><tt>ll</tt>: All loads before the barrier must complete before any load
5106 after the barrier begins.</li>
5107 <li><tt>ls</tt>: All loads before the barrier must complete before any
5108 store after the barrier begins.</li>
5109 <li><tt>ss</tt>: All stores before the barrier must complete before any
5110 store after the barrier begins.</li>
5111 <li><tt>sl</tt>: All stores before the barrier must complete before any
5112 load after the barrier begins.</li>
5115 These semantics are applied with a logical "and" behavior when more than one
5116 is enabled in a single memory barrier intrinsic.
5123 %result1 = load i32* %ptr <i>; yields {i32}:result1 = 4</i>
5124 call void @llvm.memory.barrier( i1 false, i1 true, i1 false, i1 false )
5125 <i>; guarantee the above finishes</i>
5126 store i32 8, %ptr <i>; before this begins</i>
5130 <!-- ======================================================================= -->
5131 <div class="doc_subsection">
5132 <a name="int_trampoline">Trampoline Intrinsics</a>
5135 <div class="doc_text">
5137 These intrinsics make it possible to excise one parameter, marked with
5138 the <tt>nest</tt> attribute, from a function. The result is a callable
5139 function pointer lacking the nest parameter - the caller does not need
5140 to provide a value for it. Instead, the value to use is stored in
5141 advance in a "trampoline", a block of memory usually allocated
5142 on the stack, which also contains code to splice the nest value into the
5143 argument list. This is used to implement the GCC nested function address
5147 For example, if the function is
5148 <tt>i32 f(i8* nest %c, i32 %x, i32 %y)</tt> then the resulting function
5149 pointer has signature <tt>i32 (i32, i32)*</tt>. It can be created as follows:
5151 %tramp1 = alloca [10 x i8], align 4 ; size and alignment only correct for X86
5152 %tramp = getelementptr [10 x i8]* %tramp1, i32 0, i32 0
5153 call void @llvm.init.trampoline( i8* %tramp, i8* bitcast (i32 (i8* nest , i32, i32)* @f to i8*), i8* %nval )
5154 %adj = call i8* @llvm.adjust.trampoline( i8* %tramp )
5155 %fp = bitcast i8* %adj to i32 (i32, i32)*
5157 The call <tt>%val = call i32 %fp( i32 %x, i32 %y )</tt> is then equivalent to
5158 <tt>%val = call i32 %f( i8* %nval, i32 %x, i32 %y )</tt>.
5161 Trampolines are currently only supported on the X86 architecture.
5165 <!-- _______________________________________________________________________ -->
5166 <div class="doc_subsubsection">
5167 <a name="int_it">'<tt>llvm.init.trampoline</tt>' Intrinsic</a>
5169 <div class="doc_text">
5172 declare void @llvm.init.trampoline(i8* <tramp>, i8* <func>, i8* <nval>)
5176 This initializes the memory pointed to by <tt>tramp</tt> as a trampoline.
5180 The <tt>llvm.init.trampoline</tt> intrinsic takes three arguments, all
5181 pointers. The <tt>tramp</tt> argument must point to a sufficiently large
5182 and sufficiently aligned block of memory; this memory is written to by the
5183 intrinsic. Currently LLVM provides no help in determining just how big and
5184 aligned the memory needs to be. The <tt>func</tt> argument must hold a
5185 function bitcast to an <tt>i8*</tt>.
5189 The block of memory pointed to by <tt>tramp</tt> is filled with target
5190 dependent code, turning it into a function.
5191 The new function's signature is the same as that of <tt>func</tt> with
5192 any arguments marked with the <tt>nest</tt> attribute removed. At most
5193 one such <tt>nest</tt> argument is allowed, and it must be of pointer
5194 type. Calling the new function is equivalent to calling <tt>func</tt>
5195 with the same argument list, but with <tt>nval</tt> used for the missing
5196 <tt>nest</tt> argument.
5200 <!-- _______________________________________________________________________ -->
5201 <div class="doc_subsubsection">
5202 <a name="int_at">'<tt>llvm.adjust.trampoline</tt>' Intrinsic</a>
5204 <div class="doc_text">
5207 declare i8* @llvm.adjust.trampoline(i8* <tramp>)
5211 This intrinsic returns a function pointer suitable for executing
5212 the trampoline code pointed to by <tt>tramp</tt>.
5216 The <tt>llvm.adjust.trampoline</tt> takes one argument, a pointer to a
5217 trampoline initialized by the
5218 <a href="#int_it">'<tt>llvm.init.trampoline</tt>' intrinsic</a>.
5222 A function pointer that can be used to execute the trampoline code in
5223 <tt>tramp</tt> is returned. The returned value should be bitcast to an
5224 <a href="#int_trampoline">appropriate function pointer type</a>
5225 before being called.
5229 <!-- ======================================================================= -->
5230 <div class="doc_subsection">
5231 <a name="int_general">General Intrinsics</a>
5234 <div class="doc_text">
5235 <p> This class of intrinsics is designed to be generic and has
5236 no specific purpose. </p>
5239 <!-- _______________________________________________________________________ -->
5240 <div class="doc_subsubsection">
5241 <a name="int_var_annotation">'<tt>llvm.var.annotation</tt>' Intrinsic</a>
5244 <div class="doc_text">
5248 declare void @llvm.var.annotation(i8* <val>, i8* <str>, i8* <str>, i32 <int> )
5254 The '<tt>llvm.var.annotation</tt>' intrinsic
5260 The first argument is a pointer to a value, the second is a pointer to a
5261 global string, the third is a pointer to a global string which is the source
5262 file name, and the last argument is the line number.
5268 This intrinsic allows annotation of local variables with arbitrary strings.
5269 This can be useful for special purpose optimizations that want to look for these
5270 annotations. These have no other defined use, they are ignored by code
5271 generation and optimization.
5275 <!-- *********************************************************************** -->
5278 <a href="http://jigsaw.w3.org/css-validator/check/referer"><img
5279 src="http://jigsaw.w3.org/css-validator/images/vcss" alt="Valid CSS!"></a>
5280 <a href="http://validator.w3.org/check/referer"><img
5281 src="http://www.w3.org/Icons/valid-html401" alt="Valid HTML 4.01!" /></a>
5283 <a href="mailto:sabre@nondot.org">Chris Lattner</a><br>
5284 <a href="http://llvm.org">The LLVM Compiler Infrastructure</a><br>
5285 Last modified: $Date$