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="#paramattrs">Parameter Attributes</a></li>
28 <li><a href="#moduleasm">Module-Level Inline Assembly</a></li>
31 <li><a href="#typesystem">Type System</a>
33 <li><a href="#t_primitive">Primitive Types</a>
35 <li><a href="#t_classifications">Type Classifications</a></li>
38 <li><a href="#t_derived">Derived Types</a>
40 <li><a href="#t_array">Array Type</a></li>
41 <li><a href="#t_function">Function Type</a></li>
42 <li><a href="#t_pointer">Pointer Type</a></li>
43 <li><a href="#t_struct">Structure Type</a></li>
44 <li><a href="#t_pstruct">Packed Structure Type</a></li>
45 <li><a href="#t_packed">Packed Type</a></li>
46 <li><a href="#t_opaque">Opaque Type</a></li>
51 <li><a href="#constants">Constants</a>
53 <li><a href="#simpleconstants">Simple Constants</a>
54 <li><a href="#aggregateconstants">Aggregate Constants</a>
55 <li><a href="#globalconstants">Global Variable and Function Addresses</a>
56 <li><a href="#undefvalues">Undefined Values</a>
57 <li><a href="#constantexprs">Constant Expressions</a>
60 <li><a href="#othervalues">Other Values</a>
62 <li><a href="#inlineasm">Inline Assembler Expressions</a>
65 <li><a href="#instref">Instruction Reference</a>
67 <li><a href="#terminators">Terminator Instructions</a>
69 <li><a href="#i_ret">'<tt>ret</tt>' Instruction</a></li>
70 <li><a href="#i_br">'<tt>br</tt>' Instruction</a></li>
71 <li><a href="#i_switch">'<tt>switch</tt>' Instruction</a></li>
72 <li><a href="#i_invoke">'<tt>invoke</tt>' Instruction</a></li>
73 <li><a href="#i_unwind">'<tt>unwind</tt>' Instruction</a></li>
74 <li><a href="#i_unreachable">'<tt>unreachable</tt>' Instruction</a></li>
77 <li><a href="#binaryops">Binary Operations</a>
79 <li><a href="#i_add">'<tt>add</tt>' Instruction</a></li>
80 <li><a href="#i_sub">'<tt>sub</tt>' Instruction</a></li>
81 <li><a href="#i_mul">'<tt>mul</tt>' Instruction</a></li>
82 <li><a href="#i_udiv">'<tt>udiv</tt>' Instruction</a></li>
83 <li><a href="#i_sdiv">'<tt>sdiv</tt>' Instruction</a></li>
84 <li><a href="#i_fdiv">'<tt>fdiv</tt>' Instruction</a></li>
85 <li><a href="#i_urem">'<tt>urem</tt>' Instruction</a></li>
86 <li><a href="#i_srem">'<tt>srem</tt>' Instruction</a></li>
87 <li><a href="#i_frem">'<tt>frem</tt>' Instruction</a></li>
90 <li><a href="#bitwiseops">Bitwise Binary Operations</a>
92 <li><a href="#i_and">'<tt>and</tt>' Instruction</a></li>
93 <li><a href="#i_or">'<tt>or</tt>' Instruction</a></li>
94 <li><a href="#i_xor">'<tt>xor</tt>' Instruction</a></li>
95 <li><a href="#i_shl">'<tt>shl</tt>' Instruction</a></li>
96 <li><a href="#i_lshr">'<tt>lshr</tt>' Instruction</a></li>
97 <li><a href="#i_ashr">'<tt>ashr</tt>' Instruction</a></li>
100 <li><a href="#vectorops">Vector Operations</a>
102 <li><a href="#i_extractelement">'<tt>extractelement</tt>' Instruction</a></li>
103 <li><a href="#i_insertelement">'<tt>insertelement</tt>' Instruction</a></li>
104 <li><a href="#i_shufflevector">'<tt>shufflevector</tt>' Instruction</a></li>
107 <li><a href="#memoryops">Memory Access and Addressing Operations</a>
109 <li><a href="#i_malloc">'<tt>malloc</tt>' Instruction</a></li>
110 <li><a href="#i_free">'<tt>free</tt>' Instruction</a></li>
111 <li><a href="#i_alloca">'<tt>alloca</tt>' Instruction</a></li>
112 <li><a href="#i_load">'<tt>load</tt>' Instruction</a></li>
113 <li><a href="#i_store">'<tt>store</tt>' Instruction</a></li>
114 <li><a href="#i_getelementptr">'<tt>getelementptr</tt>' Instruction</a></li>
117 <li><a href="#convertops">Conversion Operations</a>
119 <li><a href="#i_trunc">'<tt>trunc .. to</tt>' Instruction</a></li>
120 <li><a href="#i_zext">'<tt>zext .. to</tt>' Instruction</a></li>
121 <li><a href="#i_sext">'<tt>sext .. to</tt>' Instruction</a></li>
122 <li><a href="#i_fptrunc">'<tt>fptrunc .. to</tt>' Instruction</a></li>
123 <li><a href="#i_fpext">'<tt>fpext .. to</tt>' Instruction</a></li>
124 <li><a href="#i_fptoui">'<tt>fptoui .. to</tt>' Instruction</a></li>
125 <li><a href="#i_fptosi">'<tt>fptosi .. to</tt>' Instruction</a></li>
126 <li><a href="#i_uitofp">'<tt>uitofp .. to</tt>' Instruction</a></li>
127 <li><a href="#i_sitofp">'<tt>sitofp .. to</tt>' Instruction</a></li>
128 <li><a href="#i_ptrtoint">'<tt>ptrtoint .. to</tt>' Instruction</a></li>
129 <li><a href="#i_inttoptr">'<tt>inttoptr .. to</tt>' Instruction</a></li>
130 <li><a href="#i_bitcast">'<tt>bitcast .. to</tt>' Instruction</a></li>
132 <li><a href="#otherops">Other Operations</a>
134 <li><a href="#i_icmp">'<tt>icmp</tt>' Instruction</a></li>
135 <li><a href="#i_fcmp">'<tt>fcmp</tt>' Instruction</a></li>
136 <li><a href="#i_phi">'<tt>phi</tt>' Instruction</a></li>
137 <li><a href="#i_select">'<tt>select</tt>' Instruction</a></li>
138 <li><a href="#i_call">'<tt>call</tt>' Instruction</a></li>
139 <li><a href="#i_va_arg">'<tt>va_arg</tt>' Instruction</a></li>
144 <li><a href="#intrinsics">Intrinsic Functions</a>
146 <li><a href="#int_varargs">Variable Argument Handling Intrinsics</a>
148 <li><a href="#i_va_start">'<tt>llvm.va_start</tt>' Intrinsic</a></li>
149 <li><a href="#i_va_end">'<tt>llvm.va_end</tt>' Intrinsic</a></li>
150 <li><a href="#i_va_copy">'<tt>llvm.va_copy</tt>' Intrinsic</a></li>
153 <li><a href="#int_gc">Accurate Garbage Collection Intrinsics</a>
155 <li><a href="#i_gcroot">'<tt>llvm.gcroot</tt>' Intrinsic</a></li>
156 <li><a href="#i_gcread">'<tt>llvm.gcread</tt>' Intrinsic</a></li>
157 <li><a href="#i_gcwrite">'<tt>llvm.gcwrite</tt>' Intrinsic</a></li>
160 <li><a href="#int_codegen">Code Generator Intrinsics</a>
162 <li><a href="#i_returnaddress">'<tt>llvm.returnaddress</tt>' Intrinsic</a></li>
163 <li><a href="#i_frameaddress">'<tt>llvm.frameaddress</tt>' Intrinsic</a></li>
164 <li><a href="#i_stacksave">'<tt>llvm.stacksave</tt>' Intrinsic</a></li>
165 <li><a href="#i_stackrestore">'<tt>llvm.stackrestore</tt>' Intrinsic</a></li>
166 <li><a href="#i_prefetch">'<tt>llvm.prefetch</tt>' Intrinsic</a></li>
167 <li><a href="#i_pcmarker">'<tt>llvm.pcmarker</tt>' Intrinsic</a></li>
168 <li><a href="#i_readcyclecounter"><tt>llvm.readcyclecounter</tt>' Intrinsic</a></li>
171 <li><a href="#int_libc">Standard C Library Intrinsics</a>
173 <li><a href="#i_memcpy">'<tt>llvm.memcpy.*</tt>' Intrinsic</a></li>
174 <li><a href="#i_memmove">'<tt>llvm.memmove.*</tt>' Intrinsic</a></li>
175 <li><a href="#i_memset">'<tt>llvm.memset.*</tt>' Intrinsic</a></li>
176 <li><a href="#i_sqrt">'<tt>llvm.sqrt.*</tt>' Intrinsic</a></li>
177 <li><a href="#i_powi">'<tt>llvm.powi.*</tt>' Intrinsic</a></li>
180 <li><a href="#int_manip">Bit Manipulation Intrinsics</a>
182 <li><a href="#i_bswap">'<tt>llvm.bswap.*</tt>' Intrinsics</a></li>
183 <li><a href="#int_ctpop">'<tt>llvm.ctpop.*</tt>' Intrinsic </a></li>
184 <li><a href="#int_ctlz">'<tt>llvm.ctlz.*</tt>' Intrinsic </a></li>
185 <li><a href="#int_cttz">'<tt>llvm.cttz.*</tt>' Intrinsic </a></li>
188 <li><a href="#int_debugger">Debugger intrinsics</a></li>
193 <div class="doc_author">
194 <p>Written by <a href="mailto:sabre@nondot.org">Chris Lattner</a>
195 and <a href="mailto:vadve@cs.uiuc.edu">Vikram Adve</a></p>
198 <!-- *********************************************************************** -->
199 <div class="doc_section"> <a name="abstract">Abstract </a></div>
200 <!-- *********************************************************************** -->
202 <div class="doc_text">
203 <p>This document is a reference manual for the LLVM assembly language.
204 LLVM is an SSA based representation that provides type safety,
205 low-level operations, flexibility, and the capability of representing
206 'all' high-level languages cleanly. It is the common code
207 representation used throughout all phases of the LLVM compilation
211 <!-- *********************************************************************** -->
212 <div class="doc_section"> <a name="introduction">Introduction</a> </div>
213 <!-- *********************************************************************** -->
215 <div class="doc_text">
217 <p>The LLVM code representation is designed to be used in three
218 different forms: as an in-memory compiler IR, as an on-disk bytecode
219 representation (suitable for fast loading by a Just-In-Time compiler),
220 and as a human readable assembly language representation. This allows
221 LLVM to provide a powerful intermediate representation for efficient
222 compiler transformations and analysis, while providing a natural means
223 to debug and visualize the transformations. The three different forms
224 of LLVM are all equivalent. This document describes the human readable
225 representation and notation.</p>
227 <p>The LLVM representation aims to be light-weight and low-level
228 while being expressive, typed, and extensible at the same time. It
229 aims to be a "universal IR" of sorts, by being at a low enough level
230 that high-level ideas may be cleanly mapped to it (similar to how
231 microprocessors are "universal IR's", allowing many source languages to
232 be mapped to them). By providing type information, LLVM can be used as
233 the target of optimizations: for example, through pointer analysis, it
234 can be proven that a C automatic variable is never accessed outside of
235 the current function... allowing it to be promoted to a simple SSA
236 value instead of a memory location.</p>
240 <!-- _______________________________________________________________________ -->
241 <div class="doc_subsubsection"> <a name="wellformed">Well-Formedness</a> </div>
243 <div class="doc_text">
245 <p>It is important to note that this document describes 'well formed'
246 LLVM assembly language. There is a difference between what the parser
247 accepts and what is considered 'well formed'. For example, the
248 following instruction is syntactically okay, but not well formed:</p>
251 %x = <a href="#i_add">add</a> i32 1, %x
254 <p>...because the definition of <tt>%x</tt> does not dominate all of
255 its uses. The LLVM infrastructure provides a verification pass that may
256 be used to verify that an LLVM module is well formed. This pass is
257 automatically run by the parser after parsing input assembly and by
258 the optimizer before it outputs bytecode. The violations pointed out
259 by the verifier pass indicate bugs in transformation passes or input to
262 <!-- Describe the typesetting conventions here. --> </div>
264 <!-- *********************************************************************** -->
265 <div class="doc_section"> <a name="identifiers">Identifiers</a> </div>
266 <!-- *********************************************************************** -->
268 <div class="doc_text">
270 <p>LLVM uses three different forms of identifiers, for different
274 <li>Named values are represented as a string of characters with a '%' prefix.
275 For example, %foo, %DivisionByZero, %a.really.long.identifier. The actual
276 regular expression used is '<tt>%[a-zA-Z$._][a-zA-Z$._0-9]*</tt>'.
277 Identifiers which require other characters in their names can be surrounded
278 with quotes. In this way, anything except a <tt>"</tt> character can be used
281 <li>Unnamed values are represented as an unsigned numeric value with a '%'
282 prefix. For example, %12, %2, %44.</li>
284 <li>Constants, which are described in a <a href="#constants">section about
285 constants</a>, below.</li>
288 <p>LLVM requires that values start with a '%' sign for two reasons: Compilers
289 don't need to worry about name clashes with reserved words, and the set of
290 reserved words may be expanded in the future without penalty. Additionally,
291 unnamed identifiers allow a compiler to quickly come up with a temporary
292 variable without having to avoid symbol table conflicts.</p>
294 <p>Reserved words in LLVM are very similar to reserved words in other
295 languages. There are keywords for different opcodes
296 ('<tt><a href="#i_add">add</a></tt>',
297 '<tt><a href="#i_bitcast">bitcast</a></tt>',
298 '<tt><a href="#i_ret">ret</a></tt>', etc...), for primitive type names ('<tt><a
299 href="#t_void">void</a></tt>', '<tt><a href="#t_primitive">i32</a></tt>', etc...),
300 and others. These reserved words cannot conflict with variable names, because
301 none of them start with a '%' character.</p>
303 <p>Here is an example of LLVM code to multiply the integer variable
304 '<tt>%X</tt>' by 8:</p>
309 %result = <a href="#i_mul">mul</a> i32 %X, 8
312 <p>After strength reduction:</p>
315 %result = <a href="#i_shl">shl</a> i32 %X, i8 3
318 <p>And the hard way:</p>
321 <a href="#i_add">add</a> i32 %X, %X <i>; yields {i32}:%0</i>
322 <a href="#i_add">add</a> i32 %0, %0 <i>; yields {i32}:%1</i>
323 %result = <a href="#i_add">add</a> i32 %1, %1
326 <p>This last way of multiplying <tt>%X</tt> by 8 illustrates several
327 important lexical features of LLVM:</p>
331 <li>Comments are delimited with a '<tt>;</tt>' and go until the end of
334 <li>Unnamed temporaries are created when the result of a computation is not
335 assigned to a named value.</li>
337 <li>Unnamed temporaries are numbered sequentially</li>
341 <p>...and it also shows a convention that we follow in this document. When
342 demonstrating instructions, we will follow an instruction with a comment that
343 defines the type and name of value produced. Comments are shown in italic
348 <!-- *********************************************************************** -->
349 <div class="doc_section"> <a name="highlevel">High Level Structure</a> </div>
350 <!-- *********************************************************************** -->
352 <!-- ======================================================================= -->
353 <div class="doc_subsection"> <a name="modulestructure">Module Structure</a>
356 <div class="doc_text">
358 <p>LLVM programs are composed of "Module"s, each of which is a
359 translation unit of the input programs. Each module consists of
360 functions, global variables, and symbol table entries. Modules may be
361 combined together with the LLVM linker, which merges function (and
362 global variable) definitions, resolves forward declarations, and merges
363 symbol table entries. Here is an example of the "hello world" module:</p>
365 <pre><i>; Declare the string constant as a global constant...</i>
366 <a href="#identifiers">%.LC0</a> = <a href="#linkage_internal">internal</a> <a
367 href="#globalvars">constant</a> <a href="#t_array">[13 x i8 ]</a> c"hello world\0A\00" <i>; [13 x i8 ]*</i>
369 <i>; External declaration of the puts function</i>
370 <a href="#functionstructure">declare</a> i32 %puts(i8 *) <i>; i32(i8 *)* </i>
372 <i>; Global variable / Function body section separator</i>
375 <i>; Definition of main function</i>
376 define i32 %main() { <i>; i32()* </i>
377 <i>; Convert [13x i8 ]* to i8 *...</i>
379 href="#i_getelementptr">getelementptr</a> [13 x i8 ]* %.LC0, i64 0, i64 0 <i>; i8 *</i>
381 <i>; Call puts function to write out the string to stdout...</i>
383 href="#i_call">call</a> i32 %puts(i8 * %cast210) <i>; i32</i>
385 href="#i_ret">ret</a> i32 0<br>}<br></pre>
387 <p>This example is made up of a <a href="#globalvars">global variable</a>
388 named "<tt>.LC0</tt>", an external declaration of the "<tt>puts</tt>"
389 function, and a <a href="#functionstructure">function definition</a>
390 for "<tt>main</tt>".</p>
392 <p>In general, a module is made up of a list of global values,
393 where both functions and global variables are global values. Global values are
394 represented by a pointer to a memory location (in this case, a pointer to an
395 array of char, and a pointer to a function), and have one of the following <a
396 href="#linkage">linkage types</a>.</p>
398 <p>Due to a limitation in the current LLVM assembly parser (it is limited by
399 one-token lookahead), modules are split into two pieces by the "implementation"
400 keyword. Global variable prototypes and definitions must occur before the
401 keyword, and function definitions must occur after it. Function prototypes may
402 occur either before or after it. In the future, the implementation keyword may
403 become a noop, if the parser gets smarter.</p>
407 <!-- ======================================================================= -->
408 <div class="doc_subsection">
409 <a name="linkage">Linkage Types</a>
412 <div class="doc_text">
415 All Global Variables and Functions have one of the following types of linkage:
420 <dt><tt><b><a name="linkage_internal">internal</a></b></tt> </dt>
422 <dd>Global values with internal linkage are only directly accessible by
423 objects in the current module. In particular, linking code into a module with
424 an internal global value may cause the internal to be renamed as necessary to
425 avoid collisions. Because the symbol is internal to the module, all
426 references can be updated. This corresponds to the notion of the
427 '<tt>static</tt>' keyword in C, or the idea of "anonymous namespaces" in C++.
430 <dt><tt><b><a name="linkage_linkonce">linkonce</a></b></tt>: </dt>
432 <dd>"<tt>linkonce</tt>" linkage is similar to <tt>internal</tt> linkage, with
433 the twist that linking together two modules defining the same
434 <tt>linkonce</tt> globals will cause one of the globals to be discarded. This
435 is typically used to implement inline functions. Unreferenced
436 <tt>linkonce</tt> globals are allowed to be discarded.
439 <dt><tt><b><a name="linkage_weak">weak</a></b></tt>: </dt>
441 <dd>"<tt>weak</tt>" linkage is exactly the same as <tt>linkonce</tt> linkage,
442 except that unreferenced <tt>weak</tt> globals may not be discarded. This is
443 used to implement constructs in C such as "<tt>int X;</tt>" at global scope.
446 <dt><tt><b><a name="linkage_appending">appending</a></b></tt>: </dt>
448 <dd>"<tt>appending</tt>" linkage may only be applied to global variables of
449 pointer to array type. When two global variables with appending linkage are
450 linked together, the two global arrays are appended together. This is the
451 LLVM, typesafe, equivalent of having the system linker append together
452 "sections" with identical names when .o files are linked.
455 <dt><tt><b><a name="linkage_externweak">extern_weak</a></b></tt>: </dt>
456 <dd>The semantics of this linkage follow the ELF model: the symbol is weak
457 until linked, if not linked, the symbol becomes null instead of being an
462 <dt><tt><b><a name="linkage_external">externally visible</a></b></tt>:</dt>
464 <dd>If none of the above identifiers are used, the global is externally
465 visible, meaning that it participates in linkage and can be used to resolve
466 external symbol references.
470 The next two types of linkage are targeted for Microsoft Windows platform
471 only. They are designed to support importing (exporting) symbols from (to)
476 <dt><tt><b><a name="linkage_dllimport">dllimport</a></b></tt>: </dt>
478 <dd>"<tt>dllimport</tt>" linkage causes the compiler to reference a function
479 or variable via a global pointer to a pointer that is set up by the DLL
480 exporting the symbol. On Microsoft Windows targets, the pointer name is
481 formed by combining <code>_imp__</code> and the function or variable name.
484 <dt><tt><b><a name="linkage_dllexport">dllexport</a></b></tt>: </dt>
486 <dd>"<tt>dllexport</tt>" linkage causes the compiler to provide a global
487 pointer to a pointer in a DLL, so that it can be referenced with the
488 <tt>dllimport</tt> attribute. On Microsoft Windows targets, the pointer
489 name is formed by combining <code>_imp__</code> and the function or variable
495 <p><a name="linkage_external"></a>For example, since the "<tt>.LC0</tt>"
496 variable is defined to be internal, if another module defined a "<tt>.LC0</tt>"
497 variable and was linked with this one, one of the two would be renamed,
498 preventing a collision. Since "<tt>main</tt>" and "<tt>puts</tt>" are
499 external (i.e., lacking any linkage declarations), they are accessible
500 outside of the current module.</p>
501 <p>It is illegal for a function <i>declaration</i>
502 to have any linkage type other than "externally visible", <tt>dllimport</tt>,
503 or <tt>extern_weak</tt>.</p>
507 <!-- ======================================================================= -->
508 <div class="doc_subsection">
509 <a name="callingconv">Calling Conventions</a>
512 <div class="doc_text">
514 <p>LLVM <a href="#functionstructure">functions</a>, <a href="#i_call">calls</a>
515 and <a href="#i_invoke">invokes</a> can all have an optional calling convention
516 specified for the call. The calling convention of any pair of dynamic
517 caller/callee must match, or the behavior of the program is undefined. The
518 following calling conventions are supported by LLVM, and more may be added in
522 <dt><b>"<tt>ccc</tt>" - The C calling convention</b>:</dt>
524 <dd>This calling convention (the default if no other calling convention is
525 specified) matches the target C calling conventions. This calling convention
526 supports varargs function calls and tolerates some mismatch in the declared
527 prototype and implemented declaration of the function (as does normal C).
530 <dt><b>"<tt>csretcc</tt>" - The C struct return calling convention</b>:</dt>
532 <dd>This calling convention matches the target C calling conventions, except
533 that functions with this convention are required to take a pointer as their
534 first argument, and the return type of the function must be void. This is
535 used for C functions that return aggregates by-value. In this case, the
536 function has been transformed to take a pointer to the struct as the first
537 argument to the function. For targets where the ABI specifies specific
538 behavior for structure-return calls, the calling convention can be used to
539 distinguish between struct return functions and other functions that take a
540 pointer to a struct as the first argument.
543 <dt><b>"<tt>fastcc</tt>" - The fast calling convention</b>:</dt>
545 <dd>This calling convention attempts to make calls as fast as possible
546 (e.g. by passing things in registers). This calling convention allows the
547 target to use whatever tricks it wants to produce fast code for the target,
548 without having to conform to an externally specified ABI. Implementations of
549 this convention should allow arbitrary tail call optimization to be supported.
550 This calling convention does not support varargs and requires the prototype of
551 all callees to exactly match the prototype of the function definition.
554 <dt><b>"<tt>coldcc</tt>" - The cold calling convention</b>:</dt>
556 <dd>This calling convention attempts to make code in the caller as efficient
557 as possible under the assumption that the call is not commonly executed. As
558 such, these calls often preserve all registers so that the call does not break
559 any live ranges in the caller side. This calling convention does not support
560 varargs and requires the prototype of all callees to exactly match the
561 prototype of the function definition.
564 <dt><b>"<tt>cc <<em>n</em>></tt>" - Numbered convention</b>:</dt>
566 <dd>Any calling convention may be specified by number, allowing
567 target-specific calling conventions to be used. Target specific calling
568 conventions start at 64.
572 <p>More calling conventions can be added/defined on an as-needed basis, to
573 support pascal conventions or any other well-known target-independent
578 <!-- ======================================================================= -->
579 <div class="doc_subsection">
580 <a name="globalvars">Global Variables</a>
583 <div class="doc_text">
585 <p>Global variables define regions of memory allocated at compilation time
586 instead of run-time. Global variables may optionally be initialized, may have
587 an explicit section to be placed in, and may
588 have an optional explicit alignment specified. A
589 variable may be defined as a global "constant," which indicates that the
590 contents of the variable will <b>never</b> be modified (enabling better
591 optimization, allowing the global data to be placed in the read-only section of
592 an executable, etc). Note that variables that need runtime initialization
593 cannot be marked "constant" as there is a store to the variable.</p>
596 LLVM explicitly allows <em>declarations</em> of global variables to be marked
597 constant, even if the final definition of the global is not. This capability
598 can be used to enable slightly better optimization of the program, but requires
599 the language definition to guarantee that optimizations based on the
600 'constantness' are valid for the translation units that do not include the
604 <p>As SSA values, global variables define pointer values that are in
605 scope (i.e. they dominate) all basic blocks in the program. Global
606 variables always define a pointer to their "content" type because they
607 describe a region of memory, and all memory objects in LLVM are
608 accessed through pointers.</p>
610 <p>LLVM allows an explicit section to be specified for globals. If the target
611 supports it, it will emit globals to the section specified.</p>
613 <p>An explicit alignment may be specified for a global. If not present, or if
614 the alignment is set to zero, the alignment of the global is set by the target
615 to whatever it feels convenient. If an explicit alignment is specified, the
616 global is forced to have at least that much alignment. All alignments must be
619 <p>For example, the following defines a global with an initializer, section,
623 %G = constant float 1.0, section "foo", align 4
629 <!-- ======================================================================= -->
630 <div class="doc_subsection">
631 <a name="functionstructure">Functions</a>
634 <div class="doc_text">
636 <p>LLVM function definitions consist of the "<tt>define</tt>" keyord,
637 an optional <a href="#linkage">linkage type</a>, an optional
638 <a href="#callingconv">calling convention</a>, a return type, an optional
639 <a href="#paramattrs">parameter attribute</a> for the return type, a function
640 name, a (possibly empty) argument list (each with optional
641 <a href="#paramattrs">parameter attributes</a>), an optional section, an
642 optional alignment, an opening curly brace, a list of basic blocks, and a
643 closing curly brace. LLVM function declarations
644 consist of the "<tt>declare</tt>" keyword, an optional <a
645 href="#callingconv">calling convention</a>, a return type, an optional
646 <a href="#paramattrs">parameter attribute</a> for the return type, a function
647 name, a possibly empty list of arguments, and an optional alignment.</p>
649 <p>A function definition contains a list of basic blocks, forming the CFG for
650 the function. Each basic block may optionally start with a label (giving the
651 basic block a symbol table entry), contains a list of instructions, and ends
652 with a <a href="#terminators">terminator</a> instruction (such as a branch or
653 function return).</p>
655 <p>The first basic block in a program is special in two ways: it is immediately
656 executed on entrance to the function, and it is not allowed to have predecessor
657 basic blocks (i.e. there can not be any branches to the entry block of a
658 function). Because the block can have no predecessors, it also cannot have any
659 <a href="#i_phi">PHI nodes</a>.</p>
661 <p>LLVM functions are identified by their name and type signature. Hence, two
662 functions with the same name but different parameter lists or return values are
663 considered different functions, and LLVM will resolve references to each
666 <p>LLVM allows an explicit section to be specified for functions. If the target
667 supports it, it will emit functions to the section specified.</p>
669 <p>An explicit alignment may be specified for a function. If not present, or if
670 the alignment is set to zero, the alignment of the function is set by the target
671 to whatever it feels convenient. If an explicit alignment is specified, the
672 function is forced to have at least that much alignment. All alignments must be
677 <!-- ======================================================================= -->
678 <div class="doc_subsection"><a name="paramattrs">Parameter Attributes</a></div>
679 <div class="doc_text">
680 <p>The return type and each parameter of a function type may have a set of
681 <i>parameter attributes</i> associated with them. Parameter attributes are
682 used to communicate additional information about the result or parameters of
683 a function. Parameter attributes are considered to be part of the function
684 type so two functions types that differ only by the parameter attributes
685 are different function types.</p>
687 <p>Parameter attributes consist of an at sign (@) followed by either a single
688 keyword or a comma separate list of keywords enclosed in parentheses. For
690 %someFunc = i16 @zext (i8 @(sext) %someParam)
691 %someFunc = i16 @zext (i8 @zext %someParam)</pre>
692 <p>Note that the two function types above are unique because the parameter has
693 a different attribute (@sext in the first one, @zext in the second).</p>
695 <p>Currently, only the following parameter attributes are defined:</p>
697 <dt><tt>@zext</tt></dt>
698 <dd>This indicates that the parameter should be zero extended just before
699 a call to this function.</dd>
700 <dt><tt>@sext</tt></dt>
701 <dd>This indicates that the parameter should be sign extended just before
702 a call to this function.</dd>
705 <p>The current motivation for parameter attributes is to enable the sign and
706 zero extend information necessary for the C calling convention to be passed
707 from the front end to LLVM. The <tt>@zext</tt> and <tt>@sext</tt> attributes
708 are used by the code generator to perform the required extension. However,
709 parameter attributes are an orthogonal feature to calling conventions and
710 may be used for other purposes in the future.</p>
713 <!-- ======================================================================= -->
714 <div class="doc_subsection">
715 <a name="moduleasm">Module-Level Inline Assembly</a>
718 <div class="doc_text">
720 Modules may contain "module-level inline asm" blocks, which corresponds to the
721 GCC "file scope inline asm" blocks. These blocks are internally concatenated by
722 LLVM and treated as a single unit, but may be separated in the .ll file if
723 desired. The syntax is very simple:
726 <div class="doc_code"><pre>
727 module asm "inline asm code goes here"
728 module asm "more can go here"
731 <p>The strings can contain any character by escaping non-printable characters.
732 The escape sequence used is simply "\xx" where "xx" is the two digit hex code
737 The inline asm code is simply printed to the machine code .s file when
738 assembly code is generated.
743 <!-- *********************************************************************** -->
744 <div class="doc_section"> <a name="typesystem">Type System</a> </div>
745 <!-- *********************************************************************** -->
747 <div class="doc_text">
749 <p>The LLVM type system is one of the most important features of the
750 intermediate representation. Being typed enables a number of
751 optimizations to be performed on the IR directly, without having to do
752 extra analyses on the side before the transformation. A strong type
753 system makes it easier to read the generated code and enables novel
754 analyses and transformations that are not feasible to perform on normal
755 three address code representations.</p>
759 <!-- ======================================================================= -->
760 <div class="doc_subsection"> <a name="t_primitive">Primitive Types</a> </div>
761 <div class="doc_text">
762 <p>The primitive types are the fundamental building blocks of the LLVM
763 system. The current set of primitive types is as follows:</p>
765 <table class="layout">
770 <tr><th>Type</th><th>Description</th></tr>
771 <tr><td><tt>void</tt></td><td>No value</td></tr>
772 <tr><td><tt>i8</tt></td><td>Signless 8-bit value</td></tr>
773 <tr><td><tt>i32</tt></td><td>Signless 32-bit value</td></tr>
774 <tr><td><tt>float</tt></td><td>32-bit floating point value</td></tr>
775 <tr><td><tt>label</tt></td><td>Branch destination</td></tr>
782 <tr><th>Type</th><th>Description</th></tr>
783 <tr><td><tt>i1</tt></td><td>True or False value</td></tr>
784 <tr><td><tt>i16</tt></td><td>Signless 16-bit value</td></tr>
785 <tr><td><tt>i64</tt></td><td>Signless 64-bit value</td></tr>
786 <tr><td><tt>double</tt></td><td>64-bit floating point value</td></tr>
794 <!-- _______________________________________________________________________ -->
795 <div class="doc_subsubsection"> <a name="t_classifications">Type
796 Classifications</a> </div>
797 <div class="doc_text">
798 <p>These different primitive types fall into a few useful
801 <table border="1" cellspacing="0" cellpadding="4">
803 <tr><th>Classification</th><th>Types</th></tr>
805 <td><a name="t_integer">integer</a></td>
806 <td><tt>i8, i16, i32, i64</tt></td>
809 <td><a name="t_integral">integral</a></td>
810 <td><tt>i1, i8, i16, i32, i64</tt>
814 <td><a name="t_floating">floating point</a></td>
815 <td><tt>float, double</tt></td>
818 <td><a name="t_firstclass">first class</a></td>
819 <td><tt>i1, i8, i16, i32, i64, float, double, <br/>
820 <a href="#t_pointer">pointer</a>,<a href="#t_packed">packed</a></tt>
826 <p>The <a href="#t_firstclass">first class</a> types are perhaps the
827 most important. Values of these types are the only ones which can be
828 produced by instructions, passed as arguments, or used as operands to
829 instructions. This means that all structures and arrays must be
830 manipulated either by pointer or by component.</p>
833 <!-- ======================================================================= -->
834 <div class="doc_subsection"> <a name="t_derived">Derived Types</a> </div>
836 <div class="doc_text">
838 <p>The real power in LLVM comes from the derived types in the system.
839 This is what allows a programmer to represent arrays, functions,
840 pointers, and other useful types. Note that these derived types may be
841 recursive: For example, it is possible to have a two dimensional array.</p>
845 <!-- _______________________________________________________________________ -->
846 <div class="doc_subsubsection"> <a name="t_array">Array Type</a> </div>
848 <div class="doc_text">
852 <p>The array type is a very simple derived type that arranges elements
853 sequentially in memory. The array type requires a size (number of
854 elements) and an underlying data type.</p>
859 [<# elements> x <elementtype>]
862 <p>The number of elements is a constant integer value; elementtype may
863 be any type with a size.</p>
866 <table class="layout">
869 <tt>[40 x i32 ]</tt><br/>
870 <tt>[41 x i32 ]</tt><br/>
871 <tt>[40 x i8]</tt><br/>
874 Array of 40 32-bit integer values.<br/>
875 Array of 41 32-bit integer values.<br/>
876 Array of 40 8-bit integer values.<br/>
880 <p>Here are some examples of multidimensional arrays:</p>
881 <table class="layout">
884 <tt>[3 x [4 x i32]]</tt><br/>
885 <tt>[12 x [10 x float]]</tt><br/>
886 <tt>[2 x [3 x [4 x i16]]]</tt><br/>
889 3x4 array of 32-bit integer values.<br/>
890 12x10 array of single precision floating point values.<br/>
891 2x3x4 array of 16-bit integer values.<br/>
896 <p>Note that 'variable sized arrays' can be implemented in LLVM with a zero
897 length array. Normally, accesses past the end of an array are undefined in
898 LLVM (e.g. it is illegal to access the 5th element of a 3 element array).
899 As a special case, however, zero length arrays are recognized to be variable
900 length. This allows implementation of 'pascal style arrays' with the LLVM
901 type "{ i32, [0 x float]}", for example.</p>
905 <!-- _______________________________________________________________________ -->
906 <div class="doc_subsubsection"> <a name="t_function">Function Type</a> </div>
907 <div class="doc_text">
909 <p>The function type can be thought of as a function signature. It
910 consists of a return type and a list of formal parameter types.
911 Function types are usually used to build virtual function tables
912 (which are structures of pointers to functions), for indirect function
913 calls, and when defining a function.</p>
915 The return type of a function type cannot be an aggregate type.
918 <pre> <returntype> (<parameter list>)<br></pre>
919 <p>...where '<tt><parameter list></tt>' is a comma-separated list of type
920 specifiers. Optionally, the parameter list may include a type <tt>...</tt>,
921 which indicates that the function takes a variable number of arguments.
922 Variable argument functions can access their arguments with the <a
923 href="#int_varargs">variable argument handling intrinsic</a> functions.</p>
925 <table class="layout">
927 <td class="left"><tt>i32 (i32)</tt></td>
928 <td class="left">function taking an <tt>i32</tt>, returning an <tt>i32</tt>
930 </tr><tr class="layout">
931 <td class="left"><tt>float (i16 @sext, i32 *) *
933 <td class="left"><a href="#t_pointer">Pointer</a> to a function that takes
934 an <tt>i16</tt> that should be sign extended and a
935 <a href="#t_pointer">pointer</a> to <tt>i32</tt>, returning
938 </tr><tr class="layout">
939 <td class="left"><tt>i32 (i8*, ...)</tt></td>
940 <td class="left">A vararg function that takes at least one
941 <a href="#t_pointer">pointer</a> to <tt>i8 </tt> (char in C),
942 which returns an integer. This is the signature for <tt>printf</tt> in
949 <!-- _______________________________________________________________________ -->
950 <div class="doc_subsubsection"> <a name="t_struct">Structure Type</a> </div>
951 <div class="doc_text">
953 <p>The structure type is used to represent a collection of data members
954 together in memory. The packing of the field types is defined to match
955 the ABI of the underlying processor. The elements of a structure may
956 be any type that has a size.</p>
957 <p>Structures are accessed using '<tt><a href="#i_load">load</a></tt>
958 and '<tt><a href="#i_store">store</a></tt>' by getting a pointer to a
959 field with the '<tt><a href="#i_getelementptr">getelementptr</a></tt>'
962 <pre> { <type list> }<br></pre>
964 <table class="layout">
967 <tt>{ i32, i32, i32 }</tt><br/>
968 <tt>{ float, i32 (i32) * }</tt><br/>
971 a triple of three <tt>i32</tt> values<br/>
972 A pair, where the first element is a <tt>float</tt> and the second element
973 is a <a href="#t_pointer">pointer</a> to a <a href="#t_function">function</a>
974 that takes an <tt>i32</tt>, returning an <tt>i32</tt>.<br/>
980 <!-- _______________________________________________________________________ -->
981 <div class="doc_subsubsection"> <a name="t_pstruct">Packed Structure Type</a>
983 <div class="doc_text">
985 <p>The packed structure type is used to represent a collection of data members
986 together in memory. There is no padding between fields. Further, the alignment
987 of a packed structure is 1 byte. The elements of a packed structure may
988 be any type that has a size.</p>
989 <p>Structures are accessed using '<tt><a href="#i_load">load</a></tt>
990 and '<tt><a href="#i_store">store</a></tt>' by getting a pointer to a
991 field with the '<tt><a href="#i_getelementptr">getelementptr</a></tt>'
994 <pre> < { <type list> } > <br></pre>
996 <table class="layout">
999 <tt> < { i32, i32, i32 } > </tt><br/>
1000 <tt> < { float, i32 (i32) * } > </tt><br/>
1003 a triple of three <tt>i32</tt> values<br/>
1004 A pair, where the first element is a <tt>float</tt> and the second element
1005 is a <a href="#t_pointer">pointer</a> to a <a href="#t_function">function</a>
1006 that takes an <tt>i32</tt>, returning an <tt>i32</tt>.<br/>
1012 <!-- _______________________________________________________________________ -->
1013 <div class="doc_subsubsection"> <a name="t_pointer">Pointer Type</a> </div>
1014 <div class="doc_text">
1016 <p>As in many languages, the pointer type represents a pointer or
1017 reference to another object, which must live in memory.</p>
1019 <pre> <type> *<br></pre>
1021 <table class="layout">
1024 <tt>[4x i32]*</tt><br/>
1025 <tt>i32 (i32 *) *</tt><br/>
1028 A <a href="#t_pointer">pointer</a> to <a href="#t_array">array</a> of
1029 four <tt>i32</tt> values<br/>
1030 A <a href="#t_pointer">pointer</a> to a <a
1031 href="#t_function">function</a> that takes an <tt>i32*</tt>, returning an
1038 <!-- _______________________________________________________________________ -->
1039 <div class="doc_subsubsection"> <a name="t_packed">Packed Type</a> </div>
1040 <div class="doc_text">
1044 <p>A packed type is a simple derived type that represents a vector
1045 of elements. Packed types are used when multiple primitive data
1046 are operated in parallel using a single instruction (SIMD).
1047 A packed type requires a size (number of
1048 elements) and an underlying primitive data type. Vectors must have a power
1049 of two length (1, 2, 4, 8, 16 ...). Packed types are
1050 considered <a href="#t_firstclass">first class</a>.</p>
1055 < <# elements> x <elementtype> >
1058 <p>The number of elements is a constant integer value; elementtype may
1059 be any integral or floating point type.</p>
1063 <table class="layout">
1066 <tt><4 x i32></tt><br/>
1067 <tt><8 x float></tt><br/>
1068 <tt><2 x i64></tt><br/>
1071 Packed vector of 4 32-bit integer values.<br/>
1072 Packed vector of 8 floating-point values.<br/>
1073 Packed vector of 2 64-bit integer values.<br/>
1079 <!-- _______________________________________________________________________ -->
1080 <div class="doc_subsubsection"> <a name="t_opaque">Opaque Type</a> </div>
1081 <div class="doc_text">
1085 <p>Opaque types are used to represent unknown types in the system. This
1086 corresponds (for example) to the C notion of a foward declared structure type.
1087 In LLVM, opaque types can eventually be resolved to any type (not just a
1088 structure type).</p>
1098 <table class="layout">
1104 An opaque type.<br/>
1111 <!-- *********************************************************************** -->
1112 <div class="doc_section"> <a name="constants">Constants</a> </div>
1113 <!-- *********************************************************************** -->
1115 <div class="doc_text">
1117 <p>LLVM has several different basic types of constants. This section describes
1118 them all and their syntax.</p>
1122 <!-- ======================================================================= -->
1123 <div class="doc_subsection"><a name="simpleconstants">Simple Constants</a></div>
1125 <div class="doc_text">
1128 <dt><b>Boolean constants</b></dt>
1130 <dd>The two strings '<tt>true</tt>' and '<tt>false</tt>' are both valid
1131 constants of the <tt><a href="#t_primitive">i1</a></tt> type.
1134 <dt><b>Integer constants</b></dt>
1136 <dd>Standard integers (such as '4') are constants of the <a
1137 href="#t_integer">integer</a> type. Negative numbers may be used with
1141 <dt><b>Floating point constants</b></dt>
1143 <dd>Floating point constants use standard decimal notation (e.g. 123.421),
1144 exponential notation (e.g. 1.23421e+2), or a more precise hexadecimal
1145 notation (see below). Floating point constants must have a <a
1146 href="#t_floating">floating point</a> type. </dd>
1148 <dt><b>Null pointer constants</b></dt>
1150 <dd>The identifier '<tt>null</tt>' is recognized as a null pointer constant
1151 and must be of <a href="#t_pointer">pointer type</a>.</dd>
1155 <p>The one non-intuitive notation for constants is the optional hexadecimal form
1156 of floating point constants. For example, the form '<tt>double
1157 0x432ff973cafa8000</tt>' is equivalent to (but harder to read than) '<tt>double
1158 4.5e+15</tt>'. The only time hexadecimal floating point constants are required
1159 (and the only time that they are generated by the disassembler) is when a
1160 floating point constant must be emitted but it cannot be represented as a
1161 decimal floating point number. For example, NaN's, infinities, and other
1162 special values are represented in their IEEE hexadecimal format so that
1163 assembly and disassembly do not cause any bits to change in the constants.</p>
1167 <!-- ======================================================================= -->
1168 <div class="doc_subsection"><a name="aggregateconstants">Aggregate Constants</a>
1171 <div class="doc_text">
1172 <p>Aggregate constants arise from aggregation of simple constants
1173 and smaller aggregate constants.</p>
1176 <dt><b>Structure constants</b></dt>
1178 <dd>Structure constants are represented with notation similar to structure
1179 type definitions (a comma separated list of elements, surrounded by braces
1180 (<tt>{}</tt>)). For example: "<tt>{ i32 4, float 17.0, i32* %G }</tt>",
1181 where "<tt>%G</tt>" is declared as "<tt>%G = external global i32</tt>". Structure constants
1182 must have <a href="#t_struct">structure type</a>, and the number and
1183 types of elements must match those specified by the type.
1186 <dt><b>Array constants</b></dt>
1188 <dd>Array constants are represented with notation similar to array type
1189 definitions (a comma separated list of elements, surrounded by square brackets
1190 (<tt>[]</tt>)). For example: "<tt>[ i32 42, i32 11, i32 74 ]</tt>". Array
1191 constants must have <a href="#t_array">array type</a>, and the number and
1192 types of elements must match those specified by the type.
1195 <dt><b>Packed constants</b></dt>
1197 <dd>Packed constants are represented with notation similar to packed type
1198 definitions (a comma separated list of elements, surrounded by
1199 less-than/greater-than's (<tt><></tt>)). For example: "<tt>< i32 42,
1200 i32 11, i32 74, i32 100 ></tt>". Packed constants must have <a
1201 href="#t_packed">packed type</a>, and the number and types of elements must
1202 match those specified by the type.
1205 <dt><b>Zero initialization</b></dt>
1207 <dd>The string '<tt>zeroinitializer</tt>' can be used to zero initialize a
1208 value to zero of <em>any</em> type, including scalar and aggregate types.
1209 This is often used to avoid having to print large zero initializers (e.g. for
1210 large arrays) and is always exactly equivalent to using explicit zero
1217 <!-- ======================================================================= -->
1218 <div class="doc_subsection">
1219 <a name="globalconstants">Global Variable and Function Addresses</a>
1222 <div class="doc_text">
1224 <p>The addresses of <a href="#globalvars">global variables</a> and <a
1225 href="#functionstructure">functions</a> are always implicitly valid (link-time)
1226 constants. These constants are explicitly referenced when the <a
1227 href="#identifiers">identifier for the global</a> is used and always have <a
1228 href="#t_pointer">pointer</a> type. For example, the following is a legal LLVM
1234 %Z = global [2 x i32*] [ i32* %X, i32* %Y ]
1239 <!-- ======================================================================= -->
1240 <div class="doc_subsection"><a name="undefvalues">Undefined Values</a></div>
1241 <div class="doc_text">
1242 <p>The string '<tt>undef</tt>' is recognized as a type-less constant that has
1243 no specific value. Undefined values may be of any type and be used anywhere
1244 a constant is permitted.</p>
1246 <p>Undefined values indicate to the compiler that the program is well defined
1247 no matter what value is used, giving the compiler more freedom to optimize.
1251 <!-- ======================================================================= -->
1252 <div class="doc_subsection"><a name="constantexprs">Constant Expressions</a>
1255 <div class="doc_text">
1257 <p>Constant expressions are used to allow expressions involving other constants
1258 to be used as constants. Constant expressions may be of any <a
1259 href="#t_firstclass">first class</a> type and may involve any LLVM operation
1260 that does not have side effects (e.g. load and call are not supported). The
1261 following is the syntax for constant expressions:</p>
1264 <dt><b><tt>trunc ( CST to TYPE )</tt></b></dt>
1265 <dd>Truncate a constant to another type. The bit size of CST must be larger
1266 than the bit size of TYPE. Both types must be integral.</dd>
1268 <dt><b><tt>zext ( CST to TYPE )</tt></b></dt>
1269 <dd>Zero extend a constant to another type. The bit size of CST must be
1270 smaller or equal to the bit size of TYPE. Both types must be integral.</dd>
1272 <dt><b><tt>sext ( CST to TYPE )</tt></b></dt>
1273 <dd>Sign extend a constant to another type. The bit size of CST must be
1274 smaller or equal to the bit size of TYPE. Both types must be integral.</dd>
1276 <dt><b><tt>fptrunc ( CST to TYPE )</tt></b></dt>
1277 <dd>Truncate a floating point constant to another floating point type. The
1278 size of CST must be larger than the size of TYPE. Both types must be
1279 floating point.</dd>
1281 <dt><b><tt>fpext ( CST to TYPE )</tt></b></dt>
1282 <dd>Floating point extend a constant to another type. The size of CST must be
1283 smaller or equal to the size of TYPE. Both types must be floating point.</dd>
1285 <dt><b><tt>fp2uint ( CST to TYPE )</tt></b></dt>
1286 <dd>Convert a floating point constant to the corresponding unsigned integer
1287 constant. TYPE must be an integer type. CST must be floating point. If the
1288 value won't fit in the integer type, the results are undefined.</dd>
1290 <dt><b><tt>fptosi ( CST to TYPE )</tt></b></dt>
1291 <dd>Convert a floating point constant to the corresponding signed integer
1292 constant. TYPE must be an integer type. CST must be floating point. If the
1293 value won't fit in the integer type, the results are undefined.</dd>
1295 <dt><b><tt>uitofp ( CST to TYPE )</tt></b></dt>
1296 <dd>Convert an unsigned integer constant to the corresponding floating point
1297 constant. TYPE must be floating point. CST must be of integer type. If the
1298 value won't fit in the floating point type, the results are undefined.</dd>
1300 <dt><b><tt>sitofp ( CST to TYPE )</tt></b></dt>
1301 <dd>Convert a signed integer constant to the corresponding floating point
1302 constant. TYPE must be floating point. CST must be of integer type. If the
1303 value won't fit in the floating point type, the results are undefined.</dd>
1305 <dt><b><tt>ptrtoint ( CST to TYPE )</tt></b></dt>
1306 <dd>Convert a pointer typed constant to the corresponding integer constant
1307 TYPE must be an integer type. CST must be of pointer type. The CST value is
1308 zero extended, truncated, or unchanged to make it fit in TYPE.</dd>
1310 <dt><b><tt>inttoptr ( CST to TYPE )</tt></b></dt>
1311 <dd>Convert a integer constant to a pointer constant. TYPE must be a
1312 pointer type. CST must be of integer type. The CST value is zero extended,
1313 truncated, or unchanged to make it fit in a pointer size. This one is
1314 <i>really</i> dangerous!</dd>
1316 <dt><b><tt>bitcast ( CST to TYPE )</tt></b></dt>
1317 <dd>Convert a constant, CST, to another TYPE. The size of CST and TYPE must be
1318 identical (same number of bits). The conversion is done as if the CST value
1319 was stored to memory and read back as TYPE. In other words, no bits change
1320 with this operator, just the type. This can be used for conversion of
1321 packed types to any other type, as long as they have the same bit width. For
1322 pointers it is only valid to cast to another pointer type.
1325 <dt><b><tt>getelementptr ( CSTPTR, IDX0, IDX1, ... )</tt></b></dt>
1327 <dd>Perform the <a href="#i_getelementptr">getelementptr operation</a> on
1328 constants. As with the <a href="#i_getelementptr">getelementptr</a>
1329 instruction, the index list may have zero or more indexes, which are required
1330 to make sense for the type of "CSTPTR".</dd>
1332 <dt><b><tt>select ( COND, VAL1, VAL2 )</tt></b></dt>
1334 <dd>Perform the <a href="#i_select">select operation</a> on
1337 <dt><b><tt>icmp COND ( VAL1, VAL2 )</tt></b></dt>
1338 <dd>Performs the <a href="#i_icmp">icmp operation</a> on constants.</dd>
1340 <dt><b><tt>fcmp COND ( VAL1, VAL2 )</tt></b></dt>
1341 <dd>Performs the <a href="#i_fcmp">fcmp operation</a> on constants.</dd>
1343 <dt><b><tt>extractelement ( VAL, IDX )</tt></b></dt>
1345 <dd>Perform the <a href="#i_extractelement">extractelement
1346 operation</a> on constants.
1348 <dt><b><tt>insertelement ( VAL, ELT, IDX )</tt></b></dt>
1350 <dd>Perform the <a href="#i_insertelement">insertelement
1351 operation</a> on constants.</dd>
1354 <dt><b><tt>shufflevector ( VEC1, VEC2, IDXMASK )</tt></b></dt>
1356 <dd>Perform the <a href="#i_shufflevector">shufflevector
1357 operation</a> on constants.</dd>
1359 <dt><b><tt>OPCODE ( LHS, RHS )</tt></b></dt>
1361 <dd>Perform the specified operation of the LHS and RHS constants. OPCODE may
1362 be any of the <a href="#binaryops">binary</a> or <a href="#bitwiseops">bitwise
1363 binary</a> operations. The constraints on operands are the same as those for
1364 the corresponding instruction (e.g. no bitwise operations on floating point
1365 values are allowed).</dd>
1369 <!-- *********************************************************************** -->
1370 <div class="doc_section"> <a name="othervalues">Other Values</a> </div>
1371 <!-- *********************************************************************** -->
1373 <!-- ======================================================================= -->
1374 <div class="doc_subsection">
1375 <a name="inlineasm">Inline Assembler Expressions</a>
1378 <div class="doc_text">
1381 LLVM supports inline assembler expressions (as opposed to <a href="#moduleasm">
1382 Module-Level Inline Assembly</a>) through the use of a special value. This
1383 value represents the inline assembler as a string (containing the instructions
1384 to emit), a list of operand constraints (stored as a string), and a flag that
1385 indicates whether or not the inline asm expression has side effects. An example
1386 inline assembler expression is:
1390 i32 (i32) asm "bswap $0", "=r,r"
1394 Inline assembler expressions may <b>only</b> be used as the callee operand of
1395 a <a href="#i_call"><tt>call</tt> instruction</a>. Thus, typically we have:
1399 %X = call i32 asm "<a href="#i_bswap">bswap</a> $0", "=r,r"(i32 %Y)
1403 Inline asms with side effects not visible in the constraint list must be marked
1404 as having side effects. This is done through the use of the
1405 '<tt>sideeffect</tt>' keyword, like so:
1409 call void asm sideeffect "eieio", ""()
1412 <p>TODO: The format of the asm and constraints string still need to be
1413 documented here. Constraints on what can be done (e.g. duplication, moving, etc
1414 need to be documented).
1419 <!-- *********************************************************************** -->
1420 <div class="doc_section"> <a name="instref">Instruction Reference</a> </div>
1421 <!-- *********************************************************************** -->
1423 <div class="doc_text">
1425 <p>The LLVM instruction set consists of several different
1426 classifications of instructions: <a href="#terminators">terminator
1427 instructions</a>, <a href="#binaryops">binary instructions</a>,
1428 <a href="#bitwiseops">bitwise binary instructions</a>, <a
1429 href="#memoryops">memory instructions</a>, and <a href="#otherops">other
1430 instructions</a>.</p>
1434 <!-- ======================================================================= -->
1435 <div class="doc_subsection"> <a name="terminators">Terminator
1436 Instructions</a> </div>
1438 <div class="doc_text">
1440 <p>As mentioned <a href="#functionstructure">previously</a>, every
1441 basic block in a program ends with a "Terminator" instruction, which
1442 indicates which block should be executed after the current block is
1443 finished. These terminator instructions typically yield a '<tt>void</tt>'
1444 value: they produce control flow, not values (the one exception being
1445 the '<a href="#i_invoke"><tt>invoke</tt></a>' instruction).</p>
1446 <p>There are six different terminator instructions: the '<a
1447 href="#i_ret"><tt>ret</tt></a>' instruction, the '<a href="#i_br"><tt>br</tt></a>'
1448 instruction, the '<a href="#i_switch"><tt>switch</tt></a>' instruction,
1449 the '<a href="#i_invoke"><tt>invoke</tt></a>' instruction, the '<a
1450 href="#i_unwind"><tt>unwind</tt></a>' instruction, and the '<a
1451 href="#i_unreachable"><tt>unreachable</tt></a>' instruction.</p>
1455 <!-- _______________________________________________________________________ -->
1456 <div class="doc_subsubsection"> <a name="i_ret">'<tt>ret</tt>'
1457 Instruction</a> </div>
1458 <div class="doc_text">
1460 <pre> ret <type> <value> <i>; Return a value from a non-void function</i>
1461 ret void <i>; Return from void function</i>
1464 <p>The '<tt>ret</tt>' instruction is used to return control flow (and a
1465 value) from a function back to the caller.</p>
1466 <p>There are two forms of the '<tt>ret</tt>' instruction: one that
1467 returns a value and then causes control flow, and one that just causes
1468 control flow to occur.</p>
1470 <p>The '<tt>ret</tt>' instruction may return any '<a
1471 href="#t_firstclass">first class</a>' type. Notice that a function is
1472 not <a href="#wellformed">well formed</a> if there exists a '<tt>ret</tt>'
1473 instruction inside of the function that returns a value that does not
1474 match the return type of the function.</p>
1476 <p>When the '<tt>ret</tt>' instruction is executed, control flow
1477 returns back to the calling function's context. If the caller is a "<a
1478 href="#i_call"><tt>call</tt></a>" instruction, execution continues at
1479 the instruction after the call. If the caller was an "<a
1480 href="#i_invoke"><tt>invoke</tt></a>" instruction, execution continues
1481 at the beginning of the "normal" destination block. If the instruction
1482 returns a value, that value shall set the call or invoke instruction's
1485 <pre> ret i32 5 <i>; Return an integer value of 5</i>
1486 ret void <i>; Return from a void function</i>
1489 <!-- _______________________________________________________________________ -->
1490 <div class="doc_subsubsection"> <a name="i_br">'<tt>br</tt>' Instruction</a> </div>
1491 <div class="doc_text">
1493 <pre> br i1 <cond>, label <iftrue>, label <iffalse><br> br label <dest> <i>; Unconditional branch</i>
1496 <p>The '<tt>br</tt>' instruction is used to cause control flow to
1497 transfer to a different basic block in the current function. There are
1498 two forms of this instruction, corresponding to a conditional branch
1499 and an unconditional branch.</p>
1501 <p>The conditional branch form of the '<tt>br</tt>' instruction takes a
1502 single '<tt>i1</tt>' value and two '<tt>label</tt>' values. The
1503 unconditional form of the '<tt>br</tt>' instruction takes a single '<tt>label</tt>'
1504 value as a target.</p>
1506 <p>Upon execution of a conditional '<tt>br</tt>' instruction, the '<tt>i1</tt>'
1507 argument is evaluated. If the value is <tt>true</tt>, control flows
1508 to the '<tt>iftrue</tt>' <tt>label</tt> argument. If "cond" is <tt>false</tt>,
1509 control flows to the '<tt>iffalse</tt>' <tt>label</tt> argument.</p>
1511 <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
1512 href="#i_ret">ret</a> i32 1<br>IfUnequal:<br> <a href="#i_ret">ret</a> i32 0<br></pre>
1514 <!-- _______________________________________________________________________ -->
1515 <div class="doc_subsubsection">
1516 <a name="i_switch">'<tt>switch</tt>' Instruction</a>
1519 <div class="doc_text">
1523 switch <intty> <value>, label <defaultdest> [ <intty> <val>, label <dest> ... ]
1528 <p>The '<tt>switch</tt>' instruction is used to transfer control flow to one of
1529 several different places. It is a generalization of the '<tt>br</tt>'
1530 instruction, allowing a branch to occur to one of many possible
1536 <p>The '<tt>switch</tt>' instruction uses three parameters: an integer
1537 comparison value '<tt>value</tt>', a default '<tt>label</tt>' destination, and
1538 an array of pairs of comparison value constants and '<tt>label</tt>'s. The
1539 table is not allowed to contain duplicate constant entries.</p>
1543 <p>The <tt>switch</tt> instruction specifies a table of values and
1544 destinations. When the '<tt>switch</tt>' instruction is executed, this
1545 table is searched for the given value. If the value is found, control flow is
1546 transfered to the corresponding destination; otherwise, control flow is
1547 transfered to the default destination.</p>
1549 <h5>Implementation:</h5>
1551 <p>Depending on properties of the target machine and the particular
1552 <tt>switch</tt> instruction, this instruction may be code generated in different
1553 ways. For example, it could be generated as a series of chained conditional
1554 branches or with a lookup table.</p>
1559 <i>; Emulate a conditional br instruction</i>
1560 %Val = <a href="#i_zext">zext</a> i1 %value to i32
1561 switch i32 %Val, label %truedest [i32 0, label %falsedest ]
1563 <i>; Emulate an unconditional br instruction</i>
1564 switch i32 0, label %dest [ ]
1566 <i>; Implement a jump table:</i>
1567 switch i32 %val, label %otherwise [ i32 0, label %onzero
1569 i32 2, label %ontwo ]
1573 <!-- _______________________________________________________________________ -->
1574 <div class="doc_subsubsection">
1575 <a name="i_invoke">'<tt>invoke</tt>' Instruction</a>
1578 <div class="doc_text">
1583 <result> = invoke [<a href="#callingconv">cconv</a>] <ptr to function ty> %<function ptr val>(<function args>)
1584 to label <normal label> unwind label <exception label>
1589 <p>The '<tt>invoke</tt>' instruction causes control to transfer to a specified
1590 function, with the possibility of control flow transfer to either the
1591 '<tt>normal</tt>' label or the
1592 '<tt>exception</tt>' label. If the callee function returns with the
1593 "<tt><a href="#i_ret">ret</a></tt>" instruction, control flow will return to the
1594 "normal" label. If the callee (or any indirect callees) returns with the "<a
1595 href="#i_unwind"><tt>unwind</tt></a>" instruction, control is interrupted and
1596 continued at the dynamically nearest "exception" label.</p>
1600 <p>This instruction requires several arguments:</p>
1604 The optional "cconv" marker indicates which <a href="callingconv">calling
1605 convention</a> the call should use. If none is specified, the call defaults
1606 to using C calling conventions.
1608 <li>'<tt>ptr to function ty</tt>': shall be the signature of the pointer to
1609 function value being invoked. In most cases, this is a direct function
1610 invocation, but indirect <tt>invoke</tt>s are just as possible, branching off
1611 an arbitrary pointer to function value.
1614 <li>'<tt>function ptr val</tt>': An LLVM value containing a pointer to a
1615 function to be invoked. </li>
1617 <li>'<tt>function args</tt>': argument list whose types match the function
1618 signature argument types. If the function signature indicates the function
1619 accepts a variable number of arguments, the extra arguments can be
1622 <li>'<tt>normal label</tt>': the label reached when the called function
1623 executes a '<tt><a href="#i_ret">ret</a></tt>' instruction. </li>
1625 <li>'<tt>exception label</tt>': the label reached when a callee returns with
1626 the <a href="#i_unwind"><tt>unwind</tt></a> instruction. </li>
1632 <p>This instruction is designed to operate as a standard '<tt><a
1633 href="#i_call">call</a></tt>' instruction in most regards. The primary
1634 difference is that it establishes an association with a label, which is used by
1635 the runtime library to unwind the stack.</p>
1637 <p>This instruction is used in languages with destructors to ensure that proper
1638 cleanup is performed in the case of either a <tt>longjmp</tt> or a thrown
1639 exception. Additionally, this is important for implementation of
1640 '<tt>catch</tt>' clauses in high-level languages that support them.</p>
1644 %retval = invoke i32 %Test(i32 15) to label %Continue
1645 unwind label %TestCleanup <i>; {i32}:retval set</i>
1646 %retval = invoke <a href="#callingconv">coldcc</a> i32 %Test(i32 15) to label %Continue
1647 unwind label %TestCleanup <i>; {i32}:retval set</i>
1652 <!-- _______________________________________________________________________ -->
1654 <div class="doc_subsubsection"> <a name="i_unwind">'<tt>unwind</tt>'
1655 Instruction</a> </div>
1657 <div class="doc_text">
1666 <p>The '<tt>unwind</tt>' instruction unwinds the stack, continuing control flow
1667 at the first callee in the dynamic call stack which used an <a
1668 href="#i_invoke"><tt>invoke</tt></a> instruction to perform the call. This is
1669 primarily used to implement exception handling.</p>
1673 <p>The '<tt>unwind</tt>' intrinsic causes execution of the current function to
1674 immediately halt. The dynamic call stack is then searched for the first <a
1675 href="#i_invoke"><tt>invoke</tt></a> instruction on the call stack. Once found,
1676 execution continues at the "exceptional" destination block specified by the
1677 <tt>invoke</tt> instruction. If there is no <tt>invoke</tt> instruction in the
1678 dynamic call chain, undefined behavior results.</p>
1681 <!-- _______________________________________________________________________ -->
1683 <div class="doc_subsubsection"> <a name="i_unreachable">'<tt>unreachable</tt>'
1684 Instruction</a> </div>
1686 <div class="doc_text">
1695 <p>The '<tt>unreachable</tt>' instruction has no defined semantics. This
1696 instruction is used to inform the optimizer that a particular portion of the
1697 code is not reachable. This can be used to indicate that the code after a
1698 no-return function cannot be reached, and other facts.</p>
1702 <p>The '<tt>unreachable</tt>' instruction has no defined semantics.</p>
1707 <!-- ======================================================================= -->
1708 <div class="doc_subsection"> <a name="binaryops">Binary Operations</a> </div>
1709 <div class="doc_text">
1710 <p>Binary operators are used to do most of the computation in a
1711 program. They require two operands, execute an operation on them, and
1712 produce a single value. The operands might represent
1713 multiple data, as is the case with the <a href="#t_packed">packed</a> data type.
1714 The result value of a binary operator is not
1715 necessarily the same type as its operands.</p>
1716 <p>There are several different binary operators:</p>
1718 <!-- _______________________________________________________________________ -->
1719 <div class="doc_subsubsection"> <a name="i_add">'<tt>add</tt>'
1720 Instruction</a> </div>
1721 <div class="doc_text">
1723 <pre> <result> = add <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1726 <p>The '<tt>add</tt>' instruction returns the sum of its two operands.</p>
1728 <p>The two arguments to the '<tt>add</tt>' instruction must be either <a
1729 href="#t_integer">integer</a> or <a href="#t_floating">floating point</a> values.
1730 This instruction can also take <a href="#t_packed">packed</a> versions of the values.
1731 Both arguments must have identical types.</p>
1733 <p>The value produced is the integer or floating point sum of the two
1736 <pre> <result> = add i32 4, %var <i>; yields {i32}:result = 4 + %var</i>
1739 <!-- _______________________________________________________________________ -->
1740 <div class="doc_subsubsection"> <a name="i_sub">'<tt>sub</tt>'
1741 Instruction</a> </div>
1742 <div class="doc_text">
1744 <pre> <result> = sub <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1747 <p>The '<tt>sub</tt>' instruction returns the difference of its two
1749 <p>Note that the '<tt>sub</tt>' instruction is used to represent the '<tt>neg</tt>'
1750 instruction present in most other intermediate representations.</p>
1752 <p>The two arguments to the '<tt>sub</tt>' instruction must be either <a
1753 href="#t_integer">integer</a> or <a href="#t_floating">floating point</a>
1755 This instruction can also take <a href="#t_packed">packed</a> versions of the values.
1756 Both arguments must have identical types.</p>
1758 <p>The value produced is the integer or floating point difference of
1759 the two operands.</p>
1761 <pre> <result> = sub i32 4, %var <i>; yields {i32}:result = 4 - %var</i>
1762 <result> = sub i32 0, %val <i>; yields {i32}:result = -%var</i>
1765 <!-- _______________________________________________________________________ -->
1766 <div class="doc_subsubsection"> <a name="i_mul">'<tt>mul</tt>'
1767 Instruction</a> </div>
1768 <div class="doc_text">
1770 <pre> <result> = mul <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1773 <p>The '<tt>mul</tt>' instruction returns the product of its two
1776 <p>The two arguments to the '<tt>mul</tt>' instruction must be either <a
1777 href="#t_integer">integer</a> or <a href="#t_floating">floating point</a>
1779 This instruction can also take <a href="#t_packed">packed</a> versions of the values.
1780 Both arguments must have identical types.</p>
1782 <p>The value produced is the integer or floating point product of the
1784 <p>Because the operands are the same width, the result of an integer
1785 multiplication is the same whether the operands should be deemed unsigned or
1788 <pre> <result> = mul i32 4, %var <i>; yields {i32}:result = 4 * %var</i>
1791 <!-- _______________________________________________________________________ -->
1792 <div class="doc_subsubsection"> <a name="i_udiv">'<tt>udiv</tt>' Instruction
1794 <div class="doc_text">
1796 <pre> <result> = udiv <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1799 <p>The '<tt>udiv</tt>' instruction returns the quotient of its two
1802 <p>The two arguments to the '<tt>udiv</tt>' instruction must be
1803 <a href="#t_integer">integer</a> values. Both arguments must have identical
1804 types. This instruction can also take <a href="#t_packed">packed</a> versions
1805 of the values in which case the elements must be integers.</p>
1807 <p>The value produced is the unsigned integer quotient of the two operands. This
1808 instruction always performs an unsigned division operation, regardless of
1809 whether the arguments are unsigned or not.</p>
1811 <pre> <result> = udiv i32 4, %var <i>; yields {i32}:result = 4 / %var</i>
1814 <!-- _______________________________________________________________________ -->
1815 <div class="doc_subsubsection"> <a name="i_sdiv">'<tt>sdiv</tt>' Instruction
1817 <div class="doc_text">
1819 <pre> <result> = sdiv <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1822 <p>The '<tt>sdiv</tt>' instruction returns the quotient of its two
1825 <p>The two arguments to the '<tt>sdiv</tt>' instruction must be
1826 <a href="#t_integer">integer</a> values. Both arguments must have identical
1827 types. This instruction can also take <a href="#t_packed">packed</a> versions
1828 of the values in which case the elements must be integers.</p>
1830 <p>The value produced is the signed integer quotient of the two operands. This
1831 instruction always performs a signed division operation, regardless of whether
1832 the arguments are signed or not.</p>
1834 <pre> <result> = sdiv i32 4, %var <i>; yields {i32}:result = 4 / %var</i>
1837 <!-- _______________________________________________________________________ -->
1838 <div class="doc_subsubsection"> <a name="i_fdiv">'<tt>fdiv</tt>'
1839 Instruction</a> </div>
1840 <div class="doc_text">
1842 <pre> <result> = fdiv <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1845 <p>The '<tt>fdiv</tt>' instruction returns the quotient of its two
1848 <p>The two arguments to the '<tt>div</tt>' instruction must be
1849 <a href="#t_floating">floating point</a> values. Both arguments must have
1850 identical types. This instruction can also take <a href="#t_packed">packed</a>
1851 versions of the values in which case the elements must be floating point.</p>
1853 <p>The value produced is the floating point quotient of the two operands.</p>
1855 <pre> <result> = fdiv float 4.0, %var <i>; yields {float}:result = 4.0 / %var</i>
1858 <!-- _______________________________________________________________________ -->
1859 <div class="doc_subsubsection"> <a name="i_urem">'<tt>urem</tt>' Instruction</a>
1861 <div class="doc_text">
1863 <pre> <result> = urem <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1866 <p>The '<tt>urem</tt>' instruction returns the remainder from the
1867 unsigned division of its two arguments.</p>
1869 <p>The two arguments to the '<tt>urem</tt>' instruction must be
1870 <a href="#t_integer">integer</a> values. Both arguments must have identical
1873 <p>This instruction returns the unsigned integer <i>remainder</i> of a division.
1874 This instruction always performs an unsigned division to get the remainder,
1875 regardless of whether the arguments are unsigned or not.</p>
1877 <pre> <result> = urem i32 4, %var <i>; yields {i32}:result = 4 % %var</i>
1881 <!-- _______________________________________________________________________ -->
1882 <div class="doc_subsubsection"> <a name="i_srem">'<tt>srem</tt>'
1883 Instruction</a> </div>
1884 <div class="doc_text">
1886 <pre> <result> = srem <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1889 <p>The '<tt>srem</tt>' instruction returns the remainder from the
1890 signed division of its two operands.</p>
1892 <p>The two arguments to the '<tt>srem</tt>' instruction must be
1893 <a href="#t_integer">integer</a> values. Both arguments must have identical
1896 <p>This instruction returns the <i>remainder</i> of a division (where the result
1897 has the same sign as the divisor), not the <i>modulus</i> (where the
1898 result has the same sign as the dividend) of a value. For more
1899 information about the difference, see <a
1900 href="http://mathforum.org/dr.math/problems/anne.4.28.99.html">The
1903 <pre> <result> = srem i32 4, %var <i>; yields {i32}:result = 4 % %var</i>
1907 <!-- _______________________________________________________________________ -->
1908 <div class="doc_subsubsection"> <a name="i_frem">'<tt>frem</tt>'
1909 Instruction</a> </div>
1910 <div class="doc_text">
1912 <pre> <result> = frem <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1915 <p>The '<tt>frem</tt>' instruction returns the remainder from the
1916 division of its two operands.</p>
1918 <p>The two arguments to the '<tt>frem</tt>' instruction must be
1919 <a href="#t_floating">floating point</a> values. Both arguments must have
1920 identical types.</p>
1922 <p>This instruction returns the <i>remainder</i> of a division.</p>
1924 <pre> <result> = frem float 4.0, %var <i>; yields {float}:result = 4.0 % %var</i>
1928 <!-- ======================================================================= -->
1929 <div class="doc_subsection"> <a name="bitwiseops">Bitwise Binary
1930 Operations</a> </div>
1931 <div class="doc_text">
1932 <p>Bitwise binary operators are used to do various forms of
1933 bit-twiddling in a program. They are generally very efficient
1934 instructions and can commonly be strength reduced from other
1935 instructions. They require two operands, execute an operation on them,
1936 and produce a single value. The resulting value of the bitwise binary
1937 operators is always the same type as its first operand.</p>
1939 <!-- _______________________________________________________________________ -->
1940 <div class="doc_subsubsection"> <a name="i_and">'<tt>and</tt>'
1941 Instruction</a> </div>
1942 <div class="doc_text">
1944 <pre> <result> = and <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1947 <p>The '<tt>and</tt>' instruction returns the bitwise logical and of
1948 its two operands.</p>
1950 <p>The two arguments to the '<tt>and</tt>' instruction must be <a
1951 href="#t_integral">integral</a> values. Both arguments must have
1952 identical types.</p>
1954 <p>The truth table used for the '<tt>and</tt>' instruction is:</p>
1956 <div style="align: center">
1957 <table border="1" cellspacing="0" cellpadding="4">
1988 <pre> <result> = and i32 4, %var <i>; yields {i32}:result = 4 & %var</i>
1989 <result> = and i32 15, 40 <i>; yields {i32}:result = 8</i>
1990 <result> = and i32 4, 8 <i>; yields {i32}:result = 0</i>
1993 <!-- _______________________________________________________________________ -->
1994 <div class="doc_subsubsection"> <a name="i_or">'<tt>or</tt>' Instruction</a> </div>
1995 <div class="doc_text">
1997 <pre> <result> = or <ty> <var1>, <var2> <i>; yields {ty}:result</i>
2000 <p>The '<tt>or</tt>' instruction returns the bitwise logical inclusive
2001 or of its two operands.</p>
2003 <p>The two arguments to the '<tt>or</tt>' instruction must be <a
2004 href="#t_integral">integral</a> values. Both arguments must have
2005 identical types.</p>
2007 <p>The truth table used for the '<tt>or</tt>' instruction is:</p>
2009 <div style="align: center">
2010 <table border="1" cellspacing="0" cellpadding="4">
2041 <pre> <result> = or i32 4, %var <i>; yields {i32}:result = 4 | %var</i>
2042 <result> = or i32 15, 40 <i>; yields {i32}:result = 47</i>
2043 <result> = or i32 4, 8 <i>; yields {i32}:result = 12</i>
2046 <!-- _______________________________________________________________________ -->
2047 <div class="doc_subsubsection"> <a name="i_xor">'<tt>xor</tt>'
2048 Instruction</a> </div>
2049 <div class="doc_text">
2051 <pre> <result> = xor <ty> <var1>, <var2> <i>; yields {ty}:result</i>
2054 <p>The '<tt>xor</tt>' instruction returns the bitwise logical exclusive
2055 or of its two operands. The <tt>xor</tt> is used to implement the
2056 "one's complement" operation, which is the "~" operator in C.</p>
2058 <p>The two arguments to the '<tt>xor</tt>' instruction must be <a
2059 href="#t_integral">integral</a> values. Both arguments must have
2060 identical types.</p>
2062 <p>The truth table used for the '<tt>xor</tt>' instruction is:</p>
2064 <div style="align: center">
2065 <table border="1" cellspacing="0" cellpadding="4">
2097 <pre> <result> = xor i32 4, %var <i>; yields {i32}:result = 4 ^ %var</i>
2098 <result> = xor i32 15, 40 <i>; yields {i32}:result = 39</i>
2099 <result> = xor i32 4, 8 <i>; yields {i32}:result = 12</i>
2100 <result> = xor i32 %V, -1 <i>; yields {i32}:result = ~%V</i>
2103 <!-- _______________________________________________________________________ -->
2104 <div class="doc_subsubsection"> <a name="i_shl">'<tt>shl</tt>'
2105 Instruction</a> </div>
2106 <div class="doc_text">
2108 <pre> <result> = shl <ty> <var1>, i8 <var2> <i>; yields {ty}:result</i>
2111 <p>The '<tt>shl</tt>' instruction returns the first operand shifted to
2112 the left a specified number of bits.</p>
2114 <p>The first argument to the '<tt>shl</tt>' instruction must be an <a
2115 href="#t_integer">integer</a> type. The second argument must be an '<tt>i8</tt>'
2118 <p>The value produced is <tt>var1</tt> * 2<sup><tt>var2</tt></sup>.</p>
2120 <pre> <result> = shl i32 4, i8 %var <i>; yields {i32}:result = 4 << %var</i>
2121 <result> = shl i32 4, i8 2 <i>; yields {i32}:result = 16</i>
2122 <result> = shl i32 1, i8 10 <i>; yields {i32}:result = 1024</i>
2125 <!-- _______________________________________________________________________ -->
2126 <div class="doc_subsubsection"> <a name="i_lshr">'<tt>lshr</tt>'
2127 Instruction</a> </div>
2128 <div class="doc_text">
2130 <pre> <result> = lshr <ty> <var1>, i8 <var2> <i>; yields {ty}:result</i>
2134 <p>The '<tt>lshr</tt>' instruction (logical shift right) returns the first
2135 operand shifted to the right a specified number of bits.</p>
2138 <p>The first argument to the '<tt>lshr</tt>' instruction must be an <a
2139 href="#t_integer">integer</a> type. The second argument must be an '<tt>i8</tt>' type.</p>
2142 <p>This instruction always performs a logical shift right operation. The
2143 <tt>var2</tt> most significant bits will be filled with zero bits after the
2148 <result> = lshr i32 4, i8 1 <i>; yields {i32}:result = 2</i>
2149 <result> = lshr i32 4, i8 2 <i>; yields {i32}:result = 1</i>
2150 <result> = lshr i8 4, i8 3 <i>; yields {i8 }:result = 0</i>
2151 <result> = lshr i8 -2, i8 1 <i>; yields {i8 }:result = 0x7FFFFFFF </i>
2155 <!-- ======================================================================= -->
2156 <div class="doc_subsubsection"> <a name="i_ashr">'<tt>ashr</tt>'
2157 Instruction</a> </div>
2158 <div class="doc_text">
2161 <pre> <result> = ashr <ty> <var1>, i8 <var2> <i>; yields {ty}:result</i>
2165 <p>The '<tt>ashr</tt>' instruction (arithmetic shift right) returns the first
2166 operand shifted to the right a specified number of bits.</p>
2169 <p>The first argument to the '<tt>ashr</tt>' instruction must be an
2170 <a href="#t_integer">integer</a> type. The second argument must be an
2171 '<tt>i8</tt>' type.</p>
2174 <p>This instruction always performs an arithmetic shift right operation,
2175 regardless of whether the arguments are signed or not. The <tt>var2</tt> most
2176 significant bits will be filled with the sign bit of <tt>var1</tt>.</p>
2180 <result> = ashr i32 4, i8 1 <i>; yields {i32}:result = 2</i>
2181 <result> = ashr i32 4, i8 2 <i>; yields {i32}:result = 1</i>
2182 <result> = ashr i8 4, i8 3 <i>; yields {i8}:result = 0</i>
2183 <result> = ashr i8 -2, i8 1 <i>; yields {i8 }:result = -1</i>
2187 <!-- ======================================================================= -->
2188 <div class="doc_subsection">
2189 <a name="vectorops">Vector Operations</a>
2192 <div class="doc_text">
2194 <p>LLVM supports several instructions to represent vector operations in a
2195 target-independent manner. This instructions cover the element-access and
2196 vector-specific operations needed to process vectors effectively. While LLVM
2197 does directly support these vector operations, many sophisticated algorithms
2198 will want to use target-specific intrinsics to take full advantage of a specific
2203 <!-- _______________________________________________________________________ -->
2204 <div class="doc_subsubsection">
2205 <a name="i_extractelement">'<tt>extractelement</tt>' Instruction</a>
2208 <div class="doc_text">
2213 <result> = extractelement <n x <ty>> <val>, i32 <idx> <i>; yields <ty></i>
2219 The '<tt>extractelement</tt>' instruction extracts a single scalar
2220 element from a packed vector at a specified index.
2227 The first operand of an '<tt>extractelement</tt>' instruction is a
2228 value of <a href="#t_packed">packed</a> type. The second operand is
2229 an index indicating the position from which to extract the element.
2230 The index may be a variable.</p>
2235 The result is a scalar of the same type as the element type of
2236 <tt>val</tt>. Its value is the value at position <tt>idx</tt> of
2237 <tt>val</tt>. If <tt>idx</tt> exceeds the length of <tt>val</tt>, the
2238 results are undefined.
2244 %result = extractelement <4 x i32> %vec, i32 0 <i>; yields i32</i>
2249 <!-- _______________________________________________________________________ -->
2250 <div class="doc_subsubsection">
2251 <a name="i_insertelement">'<tt>insertelement</tt>' Instruction</a>
2254 <div class="doc_text">
2259 <result> = insertelement <n x <ty>> <val>, <ty> <elt>, i32 <idx> <i>; yields <n x <ty>></i>
2265 The '<tt>insertelement</tt>' instruction inserts a scalar
2266 element into a packed vector at a specified index.
2273 The first operand of an '<tt>insertelement</tt>' instruction is a
2274 value of <a href="#t_packed">packed</a> type. The second operand is a
2275 scalar value whose type must equal the element type of the first
2276 operand. The third operand is an index indicating the position at
2277 which to insert the value. The index may be a variable.</p>
2282 The result is a packed vector of the same type as <tt>val</tt>. Its
2283 element values are those of <tt>val</tt> except at position
2284 <tt>idx</tt>, where it gets the value <tt>elt</tt>. If <tt>idx</tt>
2285 exceeds the length of <tt>val</tt>, the results are undefined.
2291 %result = insertelement <4 x i32> %vec, i32 1, i32 0 <i>; yields <4 x i32></i>
2295 <!-- _______________________________________________________________________ -->
2296 <div class="doc_subsubsection">
2297 <a name="i_shufflevector">'<tt>shufflevector</tt>' Instruction</a>
2300 <div class="doc_text">
2305 <result> = shufflevector <n x <ty>> <v1>, <n x <ty>> <v2>, <n x i32> <mask> <i>; yields <n x <ty>></i>
2311 The '<tt>shufflevector</tt>' instruction constructs a permutation of elements
2312 from two input vectors, returning a vector of the same type.
2318 The first two operands of a '<tt>shufflevector</tt>' instruction are vectors
2319 with types that match each other and types that match the result of the
2320 instruction. The third argument is a shuffle mask, which has the same number
2321 of elements as the other vector type, but whose element type is always 'i32'.
2325 The shuffle mask operand is required to be a constant vector with either
2326 constant integer or undef values.
2332 The elements of the two input vectors are numbered from left to right across
2333 both of the vectors. The shuffle mask operand specifies, for each element of
2334 the result vector, which element of the two input registers the result element
2335 gets. The element selector may be undef (meaning "don't care") and the second
2336 operand may be undef if performing a shuffle from only one vector.
2342 %result = shufflevector <4 x i32> %v1, <4 x i32> %v2,
2343 <4 x i32> <i32 0, i32 4, i32 1, i32 5> <i>; yields <4 x i32></i>
2344 %result = shufflevector <4 x i32> %v1, <4 x i32> undef,
2345 <4 x i32> <i32 0, i32 1, i32 2, i32 3> <i>; yields <4 x i32></i> - Identity shuffle.
2350 <!-- ======================================================================= -->
2351 <div class="doc_subsection">
2352 <a name="memoryops">Memory Access and Addressing Operations</a>
2355 <div class="doc_text">
2357 <p>A key design point of an SSA-based representation is how it
2358 represents memory. In LLVM, no memory locations are in SSA form, which
2359 makes things very simple. This section describes how to read, write,
2360 allocate, and free memory in LLVM.</p>
2364 <!-- _______________________________________________________________________ -->
2365 <div class="doc_subsubsection">
2366 <a name="i_malloc">'<tt>malloc</tt>' Instruction</a>
2369 <div class="doc_text">
2374 <result> = malloc <type>[, i32 <NumElements>][, align <alignment>] <i>; yields {type*}:result</i>
2379 <p>The '<tt>malloc</tt>' instruction allocates memory from the system
2380 heap and returns a pointer to it.</p>
2384 <p>The '<tt>malloc</tt>' instruction allocates
2385 <tt>sizeof(<type>)*NumElements</tt>
2386 bytes of memory from the operating system and returns a pointer of the
2387 appropriate type to the program. If "NumElements" is specified, it is the
2388 number of elements allocated. If an alignment is specified, the value result
2389 of the allocation is guaranteed to be aligned to at least that boundary. If
2390 not specified, or if zero, the target can choose to align the allocation on any
2391 convenient boundary.</p>
2393 <p>'<tt>type</tt>' must be a sized type.</p>
2397 <p>Memory is allocated using the system "<tt>malloc</tt>" function, and
2398 a pointer is returned.</p>
2403 %array = malloc [4 x i8 ] <i>; yields {[%4 x i8]*}:array</i>
2405 %size = <a href="#i_add">add</a> i32 2, 2 <i>; yields {i32}:size = i32 4</i>
2406 %array1 = malloc i8, i32 4 <i>; yields {i8*}:array1</i>
2407 %array2 = malloc [12 x i8], i32 %size <i>; yields {[12 x i8]*}:array2</i>
2408 %array3 = malloc i32, i32 4, align 1024 <i>; yields {i32*}:array3</i>
2409 %array4 = malloc i32, align 1024 <i>; yields {i32*}:array4</i>
2413 <!-- _______________________________________________________________________ -->
2414 <div class="doc_subsubsection">
2415 <a name="i_free">'<tt>free</tt>' Instruction</a>
2418 <div class="doc_text">
2423 free <type> <value> <i>; yields {void}</i>
2428 <p>The '<tt>free</tt>' instruction returns memory back to the unused
2429 memory heap to be reallocated in the future.</p>
2433 <p>'<tt>value</tt>' shall be a pointer value that points to a value
2434 that was allocated with the '<tt><a href="#i_malloc">malloc</a></tt>'
2439 <p>Access to the memory pointed to by the pointer is no longer defined
2440 after this instruction executes.</p>
2445 %array = <a href="#i_malloc">malloc</a> [4 x i8] <i>; yields {[4 x i8]*}:array</i>
2446 free [4 x i8]* %array
2450 <!-- _______________________________________________________________________ -->
2451 <div class="doc_subsubsection">
2452 <a name="i_alloca">'<tt>alloca</tt>' Instruction</a>
2455 <div class="doc_text">
2460 <result> = alloca <type>[, i32 <NumElements>][, align <alignment>] <i>; yields {type*}:result</i>
2465 <p>The '<tt>alloca</tt>' instruction allocates memory on the current
2466 stack frame of the procedure that is live until the current function
2467 returns to its caller.</p>
2471 <p>The '<tt>alloca</tt>' instruction allocates <tt>sizeof(<type>)*NumElements</tt>
2472 bytes of memory on the runtime stack, returning a pointer of the
2473 appropriate type to the program. If "NumElements" is specified, it is the
2474 number of elements allocated. If an alignment is specified, the value result
2475 of the allocation is guaranteed to be aligned to at least that boundary. If
2476 not specified, or if zero, the target can choose to align the allocation on any
2477 convenient boundary.</p>
2479 <p>'<tt>type</tt>' may be any sized type.</p>
2483 <p>Memory is allocated; a pointer is returned. '<tt>alloca</tt>'d
2484 memory is automatically released when the function returns. The '<tt>alloca</tt>'
2485 instruction is commonly used to represent automatic variables that must
2486 have an address available. When the function returns (either with the <tt><a
2487 href="#i_ret">ret</a></tt> or <tt><a href="#i_unwind">unwind</a></tt>
2488 instructions), the memory is reclaimed.</p>
2493 %ptr = alloca i32 <i>; yields {i32*}:ptr</i>
2494 %ptr = alloca i32, i32 4 <i>; yields {i32*}:ptr</i>
2495 %ptr = alloca i32, i32 4, align 1024 <i>; yields {i32*}:ptr</i>
2496 %ptr = alloca i32, align 1024 <i>; yields {i32*}:ptr</i>
2500 <!-- _______________________________________________________________________ -->
2501 <div class="doc_subsubsection"> <a name="i_load">'<tt>load</tt>'
2502 Instruction</a> </div>
2503 <div class="doc_text">
2505 <pre> <result> = load <ty>* <pointer><br> <result> = volatile load <ty>* <pointer><br></pre>
2507 <p>The '<tt>load</tt>' instruction is used to read from memory.</p>
2509 <p>The argument to the '<tt>load</tt>' instruction specifies the memory
2510 address from which to load. The pointer must point to a <a
2511 href="#t_firstclass">first class</a> type. If the <tt>load</tt> is
2512 marked as <tt>volatile</tt>, then the optimizer is not allowed to modify
2513 the number or order of execution of this <tt>load</tt> with other
2514 volatile <tt>load</tt> and <tt><a href="#i_store">store</a></tt>
2517 <p>The location of memory pointed to is loaded.</p>
2519 <pre> %ptr = <a href="#i_alloca">alloca</a> i32 <i>; yields {i32*}:ptr</i>
2521 href="#i_store">store</a> i32 3, i32* %ptr <i>; yields {void}</i>
2522 %val = load i32* %ptr <i>; yields {i32}:val = i32 3</i>
2525 <!-- _______________________________________________________________________ -->
2526 <div class="doc_subsubsection"> <a name="i_store">'<tt>store</tt>'
2527 Instruction</a> </div>
2528 <div class="doc_text">
2530 <pre> store <ty> <value>, <ty>* <pointer> <i>; yields {void}</i>
2531 volatile store <ty> <value>, <ty>* <pointer> <i>; yields {void}</i>
2534 <p>The '<tt>store</tt>' instruction is used to write to memory.</p>
2536 <p>There are two arguments to the '<tt>store</tt>' instruction: a value
2537 to store and an address in which to store it. The type of the '<tt><pointer></tt>'
2538 operand must be a pointer to the type of the '<tt><value></tt>'
2539 operand. If the <tt>store</tt> is marked as <tt>volatile</tt>, then the
2540 optimizer is not allowed to modify the number or order of execution of
2541 this <tt>store</tt> with other volatile <tt>load</tt> and <tt><a
2542 href="#i_store">store</a></tt> instructions.</p>
2544 <p>The contents of memory are updated to contain '<tt><value></tt>'
2545 at the location specified by the '<tt><pointer></tt>' operand.</p>
2547 <pre> %ptr = <a href="#i_alloca">alloca</a> i32 <i>; yields {i32*}:ptr</i>
2549 href="#i_store">store</a> i32 3, i32* %ptr <i>; yields {void}</i>
2550 %val = load i32* %ptr <i>; yields {i32}:val = i32 3</i>
2554 <!-- _______________________________________________________________________ -->
2555 <div class="doc_subsubsection">
2556 <a name="i_getelementptr">'<tt>getelementptr</tt>' Instruction</a>
2559 <div class="doc_text">
2562 <result> = getelementptr <ty>* <ptrval>{, <ty> <idx>}*
2568 The '<tt>getelementptr</tt>' instruction is used to get the address of a
2569 subelement of an aggregate data structure.</p>
2573 <p>This instruction takes a list of integer operands that indicate what
2574 elements of the aggregate object to index to. The actual types of the arguments
2575 provided depend on the type of the first pointer argument. The
2576 '<tt>getelementptr</tt>' instruction is used to index down through the type
2577 levels of a structure or to a specific index in an array. When indexing into a
2578 structure, only <tt>i32</tt> integer constants are allowed. When indexing
2579 into an array or pointer, only integers of 32 or 64 bits are allowed, and will
2580 be sign extended to 64-bit values.</p>
2582 <p>For example, let's consider a C code fragment and how it gets
2583 compiled to LLVM:</p>
2597 define i32 *foo(struct ST *s) {
2598 return &s[1].Z.B[5][13];
2602 <p>The LLVM code generated by the GCC frontend is:</p>
2605 %RT = type { i8 , [10 x [20 x i32]], i8 }
2606 %ST = type { i32, double, %RT }
2610 define i32* %foo(%ST* %s) {
2612 %reg = getelementptr %ST* %s, i32 1, i32 2, i32 1, i32 5, i32 13
2619 <p>The index types specified for the '<tt>getelementptr</tt>' instruction depend
2620 on the pointer type that is being indexed into. <a href="#t_pointer">Pointer</a>
2621 and <a href="#t_array">array</a> types can use a 32-bit or 64-bit
2622 <a href="#t_integer">integer</a> type but the value will always be sign extended
2623 to 64-bits. <a href="#t_struct">Structure</a> types, require <tt>i32</tt>
2624 <b>constants</b>.</p>
2626 <p>In the example above, the first index is indexing into the '<tt>%ST*</tt>'
2627 type, which is a pointer, yielding a '<tt>%ST</tt>' = '<tt>{ i32, double, %RT
2628 }</tt>' type, a structure. The second index indexes into the third element of
2629 the structure, yielding a '<tt>%RT</tt>' = '<tt>{ i8 , [10 x [20 x i32]],
2630 i8 }</tt>' type, another structure. The third index indexes into the second
2631 element of the structure, yielding a '<tt>[10 x [20 x i32]]</tt>' type, an
2632 array. The two dimensions of the array are subscripted into, yielding an
2633 '<tt>i32</tt>' type. The '<tt>getelementptr</tt>' instruction returns a pointer
2634 to this element, thus computing a value of '<tt>i32*</tt>' type.</p>
2636 <p>Note that it is perfectly legal to index partially through a
2637 structure, returning a pointer to an inner element. Because of this,
2638 the LLVM code for the given testcase is equivalent to:</p>
2641 define i32* %foo(%ST* %s) {
2642 %t1 = getelementptr %ST* %s, i32 1 <i>; yields %ST*:%t1</i>
2643 %t2 = getelementptr %ST* %t1, i32 0, i32 2 <i>; yields %RT*:%t2</i>
2644 %t3 = getelementptr %RT* %t2, i32 0, i32 1 <i>; yields [10 x [20 x i32]]*:%t3</i>
2645 %t4 = getelementptr [10 x [20 x i32]]* %t3, i32 0, i32 5 <i>; yields [20 x i32]*:%t4</i>
2646 %t5 = getelementptr [20 x i32]* %t4, i32 0, i32 13 <i>; yields i32*:%t5</i>
2651 <p>Note that it is undefined to access an array out of bounds: array and
2652 pointer indexes must always be within the defined bounds of the array type.
2653 The one exception for this rules is zero length arrays. These arrays are
2654 defined to be accessible as variable length arrays, which requires access
2655 beyond the zero'th element.</p>
2657 <p>The getelementptr instruction is often confusing. For some more insight
2658 into how it works, see <a href="GetElementPtr.html">the getelementptr
2664 <i>; yields [12 x i8]*:aptr</i>
2665 %aptr = getelementptr {i32, [12 x i8]}* %sptr, i64 0, i32 1
2669 <!-- ======================================================================= -->
2670 <div class="doc_subsection"> <a name="convertops">Conversion Operations</a>
2672 <div class="doc_text">
2673 <p>The instructions in this category are the conversion instructions (casting)
2674 which all take a single operand and a type. They perform various bit conversions
2678 <!-- _______________________________________________________________________ -->
2679 <div class="doc_subsubsection">
2680 <a name="i_trunc">'<tt>trunc .. to</tt>' Instruction</a>
2682 <div class="doc_text">
2686 <result> = trunc <ty> <value> to <ty2> <i>; yields ty2</i>
2691 The '<tt>trunc</tt>' instruction truncates its operand to the type <tt>ty2</tt>.
2696 The '<tt>trunc</tt>' instruction takes a <tt>value</tt> to trunc, which must
2697 be an <a href="#t_integer">integer</a> type, and a type that specifies the size
2698 and type of the result, which must be an <a href="#t_integral">integral</a>
2699 type. The bit size of <tt>value</tt> must be larger than the bit size of
2700 <tt>ty2</tt>. Equal sized types are not allowed.</p>
2704 The '<tt>trunc</tt>' instruction truncates the high order bits in <tt>value</tt>
2705 and converts the remaining bits to <tt>ty2</tt>. Since the source size must be
2706 larger than the destination size, <tt>trunc</tt> cannot be a <i>no-op cast</i>.
2707 It will always truncate bits.</p>
2711 %X = trunc i32 257 to i8 <i>; yields i8:1</i>
2712 %Y = trunc i32 123 to i1 <i>; yields i1:true</i>
2713 %Y = trunc i32 122 to i1 <i>; yields i1:false</i>
2717 <!-- _______________________________________________________________________ -->
2718 <div class="doc_subsubsection">
2719 <a name="i_zext">'<tt>zext .. to</tt>' Instruction</a>
2721 <div class="doc_text">
2725 <result> = zext <ty> <value> to <ty2> <i>; yields ty2</i>
2729 <p>The '<tt>zext</tt>' instruction zero extends its operand to type
2734 <p>The '<tt>zext</tt>' instruction takes a value to cast, which must be of
2735 <a href="#t_integral">integral</a> type, and a type to cast it to, which must
2736 also be of <a href="#t_integral">integral</a> type. The bit size of the
2737 <tt>value</tt> must be smaller than the bit size of the destination type,
2741 <p>The <tt>zext</tt> fills the high order bits of the <tt>value</tt> with zero
2742 bits until it reaches the size of the destination type, <tt>ty2</tt>. When the
2743 the operand and the type are the same size, no bit filling is done and the
2744 cast is considered a <i>no-op cast</i> because no bits change (only the type
2747 <p>When zero extending from i1, the result will always be either 0 or 1.</p>
2751 %X = zext i32 257 to i64 <i>; yields i64:257</i>
2752 %Y = zext i1 true to i32 <i>; yields i32:1</i>
2756 <!-- _______________________________________________________________________ -->
2757 <div class="doc_subsubsection">
2758 <a name="i_sext">'<tt>sext .. to</tt>' Instruction</a>
2760 <div class="doc_text">
2764 <result> = sext <ty> <value> to <ty2> <i>; yields ty2</i>
2768 <p>The '<tt>sext</tt>' sign extends <tt>value</tt> to the type <tt>ty2</tt>.</p>
2772 The '<tt>sext</tt>' instruction takes a value to cast, which must be of
2773 <a href="#t_integral">integral</a> type, and a type to cast it to, which must
2774 also be of <a href="#t_integral">integral</a> type. The bit size of the
2775 <tt>value</tt> must be smaller than the bit size of the destination type,
2780 The '<tt>sext</tt>' instruction performs a sign extension by copying the sign
2781 bit (highest order bit) of the <tt>value</tt> until it reaches the bit size of
2782 the type <tt>ty2</tt>. When the the operand and the type are the same size,
2783 no bit filling is done and the cast is considered a <i>no-op cast</i> because
2784 no bits change (only the type changes).</p>
2786 <p>When sign extending from i1, the extension always results in -1 or 0.</p>
2790 %X = sext i8 -1 to i16 <i>; yields i16 :65535</i>
2791 %Y = sext i1 true to i32 <i>; yields i32:-1</i>
2795 <!-- _______________________________________________________________________ -->
2796 <div class="doc_subsubsection">
2797 <a name="i_fptrunc">'<tt>fptrunc .. to</tt>' Instruction</a>
2800 <div class="doc_text">
2805 <result> = fptrunc <ty> <value> to <ty2> <i>; yields ty2</i>
2809 <p>The '<tt>fptrunc</tt>' instruction truncates <tt>value</tt> to type
2814 <p>The '<tt>fptrunc</tt>' instruction takes a <a href="#t_floating">floating
2815 point</a> value to cast and a <a href="#t_floating">floating point</a> type to
2816 cast it to. The size of <tt>value</tt> must be larger than the size of
2817 <tt>ty2</tt>. This implies that <tt>fptrunc</tt> cannot be used to make a
2818 <i>no-op cast</i>.</p>
2821 <p> The '<tt>fptrunc</tt>' instruction truncates a <tt>value</tt> from a larger
2822 <a href="#t_floating">floating point</a> type to a smaller
2823 <a href="#t_floating">floating point</a> type. If the value cannot fit within
2824 the destination type, <tt>ty2</tt>, then the results are undefined.</p>
2828 %X = fptrunc double 123.0 to float <i>; yields float:123.0</i>
2829 %Y = fptrunc double 1.0E+300 to float <i>; yields undefined</i>
2833 <!-- _______________________________________________________________________ -->
2834 <div class="doc_subsubsection">
2835 <a name="i_fpext">'<tt>fpext .. to</tt>' Instruction</a>
2837 <div class="doc_text">
2841 <result> = fpext <ty> <value> to <ty2> <i>; yields ty2</i>
2845 <p>The '<tt>fpext</tt>' extends a floating point <tt>value</tt> to a larger
2846 floating point value.</p>
2849 <p>The '<tt>fpext</tt>' instruction takes a
2850 <a href="#t_floating">floating point</a> <tt>value</tt> to cast,
2851 and a <a href="#t_floating">floating point</a> type to cast it to. The source
2852 type must be smaller than the destination type.</p>
2855 <p>The '<tt>fpext</tt>' instruction extends the <tt>value</tt> from a smaller
2856 <a href="t_floating">floating point</a> type to a larger
2857 <a href="t_floating">floating point</a> type. The <tt>fpext</tt> cannot be
2858 used to make a <i>no-op cast</i> because it always changes bits. Use
2859 <tt>bitcast</tt> to make a <i>no-op cast</i> for a floating point cast.</p>
2863 %X = fpext float 3.1415 to double <i>; yields double:3.1415</i>
2864 %Y = fpext float 1.0 to float <i>; yields float:1.0 (no-op)</i>
2868 <!-- _______________________________________________________________________ -->
2869 <div class="doc_subsubsection">
2870 <a name="i_fp2uint">'<tt>fptoui .. to</tt>' Instruction</a>
2872 <div class="doc_text">
2876 <result> = fp2uint <ty> <value> to <ty2> <i>; yields ty2</i>
2880 <p>The '<tt>fp2uint</tt>' converts a floating point <tt>value</tt> to its
2881 unsigned integer equivalent of type <tt>ty2</tt>.
2885 <p>The '<tt>fp2uint</tt>' instruction takes a value to cast, which must be a
2886 <a href="#t_floating">floating point</a> value, and a type to cast it to, which
2887 must be an <a href="#t_integral">integral</a> type.</p>
2890 <p> The '<tt>fp2uint</tt>' instruction converts its
2891 <a href="#t_floating">floating point</a> operand into the nearest (rounding
2892 towards zero) unsigned integer value. If the value cannot fit in <tt>ty2</tt>,
2893 the results are undefined.</p>
2895 <p>When converting to i1, the conversion is done as a comparison against
2896 zero. If the <tt>value</tt> was zero, the i1 result will be <tt>false</tt>.
2897 If the <tt>value</tt> was non-zero, the i1 result will be <tt>true</tt>.</p>
2901 %X = fp2uint double 123.0 to i32 <i>; yields i32:123</i>
2902 %Y = fp2uint float 1.0E+300 to i1 <i>; yields i1:true</i>
2903 %X = fp2uint float 1.04E+17 to i8 <i>; yields undefined:1</i>
2907 <!-- _______________________________________________________________________ -->
2908 <div class="doc_subsubsection">
2909 <a name="i_fptosi">'<tt>fptosi .. to</tt>' Instruction</a>
2911 <div class="doc_text">
2915 <result> = fptosi <ty> <value> to <ty2> <i>; yields ty2</i>
2919 <p>The '<tt>fptosi</tt>' instruction converts
2920 <a href="#t_floating">floating point</a> <tt>value</tt> to type <tt>ty2</tt>.
2925 <p> The '<tt>fptosi</tt>' instruction takes a value to cast, which must be a
2926 <a href="#t_floating">floating point</a> value, and a type to cast it to, which
2927 must also be an <a href="#t_integral">integral</a> type.</p>
2930 <p>The '<tt>fptosi</tt>' instruction converts its
2931 <a href="#t_floating">floating point</a> operand into the nearest (rounding
2932 towards zero) signed integer value. If the value cannot fit in <tt>ty2</tt>,
2933 the results are undefined.</p>
2935 <p>When converting to i1, the conversion is done as a comparison against
2936 zero. If the <tt>value</tt> was zero, the i1 result will be <tt>false</tt>.
2937 If the <tt>value</tt> was non-zero, the i1 result will be <tt>true</tt>.</p>
2941 %X = fptosi double -123.0 to i32 <i>; yields i32:-123</i>
2942 %Y = fptosi float 1.0E-247 to i1 <i>; yields i1:true</i>
2943 %X = fptosi float 1.04E+17 to i8 <i>; yields undefined:1</i>
2947 <!-- _______________________________________________________________________ -->
2948 <div class="doc_subsubsection">
2949 <a name="i_uitofp">'<tt>uitofp .. to</tt>' Instruction</a>
2951 <div class="doc_text">
2955 <result> = uitofp <ty> <value> to <ty2> <i>; yields ty2</i>
2959 <p>The '<tt>uitofp</tt>' instruction regards <tt>value</tt> as an unsigned
2960 integer and converts that value to the <tt>ty2</tt> type.</p>
2964 <p>The '<tt>uitofp</tt>' instruction takes a value to cast, which must be an
2965 <a href="#t_integral">integral</a> value, and a type to cast it to, which must
2966 be a <a href="#t_floating">floating point</a> type.</p>
2969 <p>The '<tt>uitofp</tt>' instruction interprets its operand as an unsigned
2970 integer quantity and converts it to the corresponding floating point value. If
2971 the value cannot fit in the floating point value, the results are undefined.</p>
2976 %X = uitofp i32 257 to float <i>; yields float:257.0</i>
2977 %Y = uitofp i8 -1 to double <i>; yields double:255.0</i>
2981 <!-- _______________________________________________________________________ -->
2982 <div class="doc_subsubsection">
2983 <a name="i_sitofp">'<tt>sitofp .. to</tt>' Instruction</a>
2985 <div class="doc_text">
2989 <result> = sitofp <ty> <value> to <ty2> <i>; yields ty2</i>
2993 <p>The '<tt>sitofp</tt>' instruction regards <tt>value</tt> as a signed
2994 integer and converts that value to the <tt>ty2</tt> type.</p>
2997 <p>The '<tt>sitofp</tt>' instruction takes a value to cast, which must be an
2998 <a href="#t_integral">integral</a> value, and a type to cast it to, which must be
2999 a <a href="#t_floating">floating point</a> type.</p>
3002 <p>The '<tt>sitofp</tt>' instruction interprets its operand as a signed
3003 integer quantity and converts it to the corresponding floating point value. If
3004 the value cannot fit in the floating point value, the results are undefined.</p>
3008 %X = sitofp i32 257 to float <i>; yields float:257.0</i>
3009 %Y = sitofp i8 -1 to double <i>; yields double:-1.0</i>
3013 <!-- _______________________________________________________________________ -->
3014 <div class="doc_subsubsection">
3015 <a name="i_ptrtoint">'<tt>ptrtoint .. to</tt>' Instruction</a>
3017 <div class="doc_text">
3021 <result> = ptrtoint <ty> <value> to <ty2> <i>; yields ty2</i>
3025 <p>The '<tt>ptrtoint</tt>' instruction converts the pointer <tt>value</tt> to
3026 the integer type <tt>ty2</tt>.</p>
3029 <p>The '<tt>ptrtoint</tt>' instruction takes a <tt>value</tt> to cast, which
3030 must be a <a href="t_pointer">pointer</a> value, and a type to cast it to
3031 <tt>ty2</tt>, which must be an <a href="#t_integer">integer</a> type.
3034 <p>The '<tt>ptrtoint</tt>' instruction converts <tt>value</tt> to integer type
3035 <tt>ty2</tt> by interpreting the pointer value as an integer and either
3036 truncating or zero extending that value to the size of the integer type. If
3037 <tt>value</tt> is smaller than <tt>ty2</tt> then a zero extension is done. If
3038 <tt>value</tt> is larger than <tt>ty2</tt> then a truncation is done. If they
3039 are the same size, then nothing is done (<i>no-op cast</i>).</p>
3043 %X = ptrtoint i32* %X to i8 <i>; yields truncation on 32-bit</i>
3044 %Y = ptrtoint i32* %x to i64 <i>; yields zero extend on 32-bit</i>
3048 <!-- _______________________________________________________________________ -->
3049 <div class="doc_subsubsection">
3050 <a name="i_inttoptr">'<tt>inttoptr .. to</tt>' Instruction</a>
3052 <div class="doc_text">
3056 <result> = inttoptr <ty> <value> to <ty2> <i>; yields ty2</i>
3060 <p>The '<tt>inttoptr</tt>' instruction converts an integer <tt>value</tt> to
3061 a pointer type, <tt>ty2</tt>.</p>
3064 <p>The '<tt>inttoptr</tt>' instruction takes an <a href="i_integer">integer</a>
3065 value to cast, and a type to cast it to, which must be a
3066 <a href="#t_pointer">pointer</a> type.
3069 <p>The '<tt>inttoptr</tt>' instruction converts <tt>value</tt> to type
3070 <tt>ty2</tt> by applying either a zero extension or a truncation depending on
3071 the size of the integer <tt>value</tt>. If <tt>value</tt> is larger than the
3072 size of a pointer then a truncation is done. If <tt>value</tt> is smaller than
3073 the size of a pointer then a zero extension is done. If they are the same size,
3074 nothing is done (<i>no-op cast</i>).</p>
3078 %X = inttoptr i32 255 to i32* <i>; yields zero extend on 64-bit</i>
3079 %X = inttoptr i32 255 to i32* <i>; yields no-op on 32-bit </i>
3080 %Y = inttoptr i16 0 to i32* <i>; yields zero extend on 32-bit</i>
3084 <!-- _______________________________________________________________________ -->
3085 <div class="doc_subsubsection">
3086 <a name="i_bitcast">'<tt>bitcast .. to</tt>' Instruction</a>
3088 <div class="doc_text">
3092 <result> = bitcast <ty> <value> to <ty2> <i>; yields ty2</i>
3096 <p>The '<tt>bitcast</tt>' instruction converts <tt>value</tt> to type
3097 <tt>ty2</tt> without changing any bits.</p>
3100 <p>The '<tt>bitcast</tt>' instruction takes a value to cast, which must be
3101 a first class value, and a type to cast it to, which must also be a <a
3102 href="#t_firstclass">first class</a> type. The bit sizes of <tt>value</tt>
3103 and the destination type, <tt>ty2</tt>, must be identical. If the source
3104 type is a pointer, the destination type must also be a pointer.</p>
3107 <p>The '<tt>bitcast</tt>' instruction converts <tt>value</tt> to type
3108 <tt>ty2</tt>. It is always a <i>no-op cast</i> because no bits change with
3109 this conversion. The conversion is done as if the <tt>value</tt> had been
3110 stored to memory and read back as type <tt>ty2</tt>. Pointer types may only be
3111 converted to other pointer types with this instruction. To convert pointers to
3112 other types, use the <a href="#i_inttoptr">inttoptr</a> or
3113 <a href="#i_ptrtoint">ptrtoint</a> instructions first.</p>
3117 %X = bitcast i8 255 to i8 <i>; yields i8 :-1</i>
3118 %Y = bitcast i32* %x to sint* <i>; yields sint*:%x</i>
3119 %Z = bitcast <2xint> %V to i64; <i>; yields i64: %V</i>
3123 <!-- ======================================================================= -->
3124 <div class="doc_subsection"> <a name="otherops">Other Operations</a> </div>
3125 <div class="doc_text">
3126 <p>The instructions in this category are the "miscellaneous"
3127 instructions, which defy better classification.</p>
3130 <!-- _______________________________________________________________________ -->
3131 <div class="doc_subsubsection"><a name="i_icmp">'<tt>icmp</tt>' Instruction</a>
3133 <div class="doc_text">
3135 <pre> <result> = icmp <cond> <ty> <var1>, <var2>
3136 <i>; yields {i1}:result</i>
3139 <p>The '<tt>icmp</tt>' instruction returns a boolean value based on comparison
3140 of its two integer operands.</p>
3142 <p>The '<tt>icmp</tt>' instruction takes three operands. The first operand is
3143 the condition code which indicates the kind of comparison to perform. It is not
3144 a value, just a keyword. The possibilities for the condition code are:
3146 <li><tt>eq</tt>: equal</li>
3147 <li><tt>ne</tt>: not equal </li>
3148 <li><tt>ugt</tt>: unsigned greater than</li>
3149 <li><tt>uge</tt>: unsigned greater or equal</li>
3150 <li><tt>ult</tt>: unsigned less than</li>
3151 <li><tt>ule</tt>: unsigned less or equal</li>
3152 <li><tt>sgt</tt>: signed greater than</li>
3153 <li><tt>sge</tt>: signed greater or equal</li>
3154 <li><tt>slt</tt>: signed less than</li>
3155 <li><tt>sle</tt>: signed less or equal</li>
3157 <p>The remaining two arguments must be <a href="#t_integral">integral</a> or
3158 <a href="#t_pointer">pointer</a> typed. They must also be identical types.</p>
3160 <p>The '<tt>icmp</tt>' compares <tt>var1</tt> and <tt>var2</tt> according to
3161 the condition code given as <tt>cond</tt>. The comparison performed always
3162 yields a <a href="#t_primitive">i1</a> result, as follows:
3164 <li><tt>eq</tt>: yields <tt>true</tt> if the operands are equal,
3165 <tt>false</tt> otherwise. No sign interpretation is necessary or performed.
3167 <li><tt>ne</tt>: yields <tt>true</tt> if the operands are unequal,
3168 <tt>false</tt> otherwise. No sign interpretation is necessary or performed.
3169 <li><tt>ugt</tt>: interprets the operands as unsigned values and yields
3170 <tt>true</tt> if <tt>var1</tt> is greater than <tt>var2</tt>.</li>
3171 <li><tt>uge</tt>: interprets the operands as unsigned values and yields
3172 <tt>true</tt> if <tt>var1</tt> is greater than or equal to <tt>var2</tt>.</li>
3173 <li><tt>ult</tt>: interprets the operands as unsigned values and yields
3174 <tt>true</tt> if <tt>var1</tt> is less than <tt>var2</tt>.</li>
3175 <li><tt>ule</tt>: interprets the operands as unsigned values and yields
3176 <tt>true</tt> if <tt>var1</tt> is less than or equal to <tt>var2</tt>.</li>
3177 <li><tt>sgt</tt>: interprets the operands as signed values and yields
3178 <tt>true</tt> if <tt>var1</tt> is greater than <tt>var2</tt>.</li>
3179 <li><tt>sge</tt>: interprets the operands as signed values and yields
3180 <tt>true</tt> if <tt>var1</tt> is greater than or equal to <tt>var2</tt>.</li>
3181 <li><tt>slt</tt>: interprets the operands as signed values and yields
3182 <tt>true</tt> if <tt>var1</tt> is less than <tt>var2</tt>.</li>
3183 <li><tt>sle</tt>: interprets the operands as signed values and yields
3184 <tt>true</tt> if <tt>var1</tt> is less than or equal to <tt>var2</tt>.</li>
3186 <p>If the operands are <a href="#t_pointer">pointer</a> typed, the pointer
3187 values are treated as integers and then compared.</p>
3188 <p>If the operands are <a href="#t_packed">packed</a> typed, the elements of
3189 the vector are compared in turn and the predicate must hold for all
3193 <pre> <result> = icmp eq i32 4, 5 <i>; yields: result=false</i>
3194 <result> = icmp ne float* %X, %X <i>; yields: result=false</i>
3195 <result> = icmp ult i16 4, 5 <i>; yields: result=true</i>
3196 <result> = icmp sgt i16 4, 5 <i>; yields: result=false</i>
3197 <result> = icmp ule i16 -4, 5 <i>; yields: result=false</i>
3198 <result> = icmp sge i16 4, 5 <i>; yields: result=false</i>
3202 <!-- _______________________________________________________________________ -->
3203 <div class="doc_subsubsection"><a name="i_fcmp">'<tt>fcmp</tt>' Instruction</a>
3205 <div class="doc_text">
3207 <pre> <result> = fcmp <cond> <ty> <var1>, <var2>
3208 <i>; yields {i1}:result</i>
3211 <p>The '<tt>fcmp</tt>' instruction returns a boolean value based on comparison
3212 of its floating point operands.</p>
3214 <p>The '<tt>fcmp</tt>' instruction takes three operands. The first operand is
3215 the condition code which indicates the kind of comparison to perform. It is not
3216 a value, just a keyword. The possibilities for the condition code are:
3218 <li><tt>false</tt>: no comparison, always returns false</li>
3219 <li><tt>oeq</tt>: ordered and equal</li>
3220 <li><tt>ogt</tt>: ordered and greater than </li>
3221 <li><tt>oge</tt>: ordered and greater than or equal</li>
3222 <li><tt>olt</tt>: ordered and less than </li>
3223 <li><tt>ole</tt>: ordered and less than or equal</li>
3224 <li><tt>one</tt>: ordered and not equal</li>
3225 <li><tt>ord</tt>: ordered (no nans)</li>
3226 <li><tt>ueq</tt>: unordered or equal</li>
3227 <li><tt>ugt</tt>: unordered or greater than </li>
3228 <li><tt>uge</tt>: unordered or greater than or equal</li>
3229 <li><tt>ult</tt>: unordered or less than </li>
3230 <li><tt>ule</tt>: unordered or less than or equal</li>
3231 <li><tt>une</tt>: unordered or not equal</li>
3232 <li><tt>uno</tt>: unordered (either nans)</li>
3233 <li><tt>true</tt>: no comparison, always returns true</li>
3235 <p>In the preceding, <i>ordered</i> means that neither operand is a QNAN while
3236 <i>unordered</i> means that either operand may be a QNAN.</p>
3237 <p>The <tt>val1</tt> and <tt>val2</tt> arguments must be
3238 <a href="#t_floating">floating point</a> typed. They must have identical
3240 <p>In the foregoing, <i>ordered</i> means that neither operand is a QNAN and
3241 <i>unordered</i> means that either operand is a QNAN.</p>
3243 <p>The '<tt>fcmp</tt>' compares <tt>var1</tt> and <tt>var2</tt> according to
3244 the condition code given as <tt>cond</tt>. The comparison performed always
3245 yields a <a href="#t_primitive">i1</a> result, as follows:
3247 <li><tt>false</tt>: always yields <tt>false</tt>, regardless of operands.</li>
3248 <li><tt>oeq</tt>: yields <tt>true</tt> if both operands are not a QNAN and
3249 <tt>var1</tt> is equal to <tt>var2</tt>.</li>
3250 <li><tt>ogt</tt>: yields <tt>true</tt> if both operands are not a QNAN and
3251 <tt>var1</tt> is greather than <tt>var2</tt>.</li>
3252 <li><tt>oge</tt>: yields <tt>true</tt> if both operands are not a QNAN and
3253 <tt>var1</tt> is greater than or equal to <tt>var2</tt>.</li>
3254 <li><tt>olt</tt>: yields <tt>true</tt> if both operands are not a QNAN and
3255 <tt>var1</tt> is less than <tt>var2</tt>.</li>
3256 <li><tt>ole</tt>: yields <tt>true</tt> if both operands are not a QNAN and
3257 <tt>var1</tt> is less than or equal to <tt>var2</tt>.</li>
3258 <li><tt>one</tt>: yields <tt>true</tt> if both operands are not a QNAN and
3259 <tt>var1</tt> is not equal to <tt>var2</tt>.</li>
3260 <li><tt>ord</tt>: yields <tt>true</tt> if both operands are not a QNAN.</li>
3261 <li><tt>ueq</tt>: yields <tt>true</tt> if either operand is a QNAN or
3262 <tt>var1</tt> is equal to <tt>var2</tt>.</li>
3263 <li><tt>ugt</tt>: yields <tt>true</tt> if either operand is a QNAN or
3264 <tt>var1</tt> is greater than <tt>var2</tt>.</li>
3265 <li><tt>uge</tt>: yields <tt>true</tt> if either operand is a QNAN or
3266 <tt>var1</tt> is greater than or equal to <tt>var2</tt>.</li>
3267 <li><tt>ult</tt>: yields <tt>true</tt> if either operand is a QNAN or
3268 <tt>var1</tt> is less than <tt>var2</tt>.</li>
3269 <li><tt>ule</tt>: yields <tt>true</tt> if either operand is a QNAN or
3270 <tt>var1</tt> is less than or equal to <tt>var2</tt>.</li>
3271 <li><tt>une</tt>: yields <tt>true</tt> if either operand is a QNAN or
3272 <tt>var1</tt> is not equal to <tt>var2</tt>.</li>
3273 <li><tt>uno</tt>: yields <tt>true</tt> if either operand is a QNAN.</li>
3274 <li><tt>true</tt>: always yields <tt>true</tt>, regardless of operands.</li>
3276 <p>If the operands are <a href="#t_packed">packed</a> typed, the elements of
3277 the vector are compared in turn and the predicate must hold for all elements.
3281 <pre> <result> = fcmp oeq float 4.0, 5.0 <i>; yields: result=false</i>
3282 <result> = icmp one float 4.0, 5.0 <i>; yields: result=true</i>
3283 <result> = icmp olt float 4.0, 5.0 <i>; yields: result=true</i>
3284 <result> = icmp ueq double 1.0, 2.0 <i>; yields: result=false</i>
3288 <!-- _______________________________________________________________________ -->
3289 <div class="doc_subsubsection"> <a name="i_phi">'<tt>phi</tt>'
3290 Instruction</a> </div>
3291 <div class="doc_text">
3293 <pre> <result> = phi <ty> [ <val0>, <label0>], ...<br></pre>
3295 <p>The '<tt>phi</tt>' instruction is used to implement the φ node in
3296 the SSA graph representing the function.</p>
3298 <p>The type of the incoming values are specified with the first type
3299 field. After this, the '<tt>phi</tt>' instruction takes a list of pairs
3300 as arguments, with one pair for each predecessor basic block of the
3301 current block. Only values of <a href="#t_firstclass">first class</a>
3302 type may be used as the value arguments to the PHI node. Only labels
3303 may be used as the label arguments.</p>
3304 <p>There must be no non-phi instructions between the start of a basic
3305 block and the PHI instructions: i.e. PHI instructions must be first in
3308 <p>At runtime, the '<tt>phi</tt>' instruction logically takes on the
3309 value specified by the parameter, depending on which basic block we
3310 came from in the last <a href="#terminators">terminator</a> instruction.</p>
3312 <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>
3315 <!-- _______________________________________________________________________ -->
3316 <div class="doc_subsubsection">
3317 <a name="i_select">'<tt>select</tt>' Instruction</a>
3320 <div class="doc_text">
3325 <result> = select i1 <cond>, <ty> <val1>, <ty> <val2> <i>; yields ty</i>
3331 The '<tt>select</tt>' instruction is used to choose one value based on a
3332 condition, without branching.
3339 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.
3345 If the boolean condition evaluates to true, the instruction returns the first
3346 value argument; otherwise, it returns the second value argument.
3352 %X = select i1 true, i8 17, i8 42 <i>; yields i8:17</i>
3357 <!-- _______________________________________________________________________ -->
3358 <div class="doc_subsubsection">
3359 <a name="i_call">'<tt>call</tt>' Instruction</a>
3362 <div class="doc_text">
3366 <result> = [tail] call [<a href="#callingconv">cconv</a>] <ty>* <fnptrval>(<param list>)
3371 <p>The '<tt>call</tt>' instruction represents a simple function call.</p>
3375 <p>This instruction requires several arguments:</p>
3379 <p>The optional "tail" marker indicates whether the callee function accesses
3380 any allocas or varargs in the caller. If the "tail" marker is present, the
3381 function call is eligible for tail call optimization. Note that calls may
3382 be marked "tail" even if they do not occur before a <a
3383 href="#i_ret"><tt>ret</tt></a> instruction.
3386 <p>The optional "cconv" marker indicates which <a href="callingconv">calling
3387 convention</a> the call should use. If none is specified, the call defaults
3388 to using C calling conventions.
3391 <p>'<tt>ty</tt>': shall be the signature of the pointer to function value
3392 being invoked. The argument types must match the types implied by this
3393 signature. This type can be omitted if the function is not varargs and
3394 if the function type does not return a pointer to a function.</p>
3397 <p>'<tt>fnptrval</tt>': An LLVM value containing a pointer to a function to
3398 be invoked. In most cases, this is a direct function invocation, but
3399 indirect <tt>call</tt>s are just as possible, calling an arbitrary pointer
3400 to function value.</p>
3403 <p>'<tt>function args</tt>': argument list whose types match the
3404 function signature argument types. All arguments must be of
3405 <a href="#t_firstclass">first class</a> type. If the function signature
3406 indicates the function accepts a variable number of arguments, the extra
3407 arguments can be specified.</p>
3413 <p>The '<tt>call</tt>' instruction is used to cause control flow to
3414 transfer to a specified function, with its incoming arguments bound to
3415 the specified values. Upon a '<tt><a href="#i_ret">ret</a></tt>'
3416 instruction in the called function, control flow continues with the
3417 instruction after the function call, and the return value of the
3418 function is bound to the result argument. This is a simpler case of
3419 the <a href="#i_invoke">invoke</a> instruction.</p>
3424 %retval = call i32 %test(i32 %argc)
3425 call i32(i8 *, ...) *%printf(i8 * %msg, i32 12, i8 42);
3426 %X = tail call i32 %foo()
3427 %Y = tail call <a href="#callingconv">fastcc</a> i32 %foo()
3432 <!-- _______________________________________________________________________ -->
3433 <div class="doc_subsubsection">
3434 <a name="i_va_arg">'<tt>va_arg</tt>' Instruction</a>
3437 <div class="doc_text">
3442 <resultval> = va_arg <va_list*> <arglist>, <argty>
3447 <p>The '<tt>va_arg</tt>' instruction is used to access arguments passed through
3448 the "variable argument" area of a function call. It is used to implement the
3449 <tt>va_arg</tt> macro in C.</p>
3453 <p>This instruction takes a <tt>va_list*</tt> value and the type of
3454 the argument. It returns a value of the specified argument type and
3455 increments the <tt>va_list</tt> to point to the next argument. Again, the
3456 actual type of <tt>va_list</tt> is target specific.</p>
3460 <p>The '<tt>va_arg</tt>' instruction loads an argument of the specified
3461 type from the specified <tt>va_list</tt> and causes the
3462 <tt>va_list</tt> to point to the next argument. For more information,
3463 see the variable argument handling <a href="#int_varargs">Intrinsic
3466 <p>It is legal for this instruction to be called in a function which does not
3467 take a variable number of arguments, for example, the <tt>vfprintf</tt>
3470 <p><tt>va_arg</tt> is an LLVM instruction instead of an <a
3471 href="#intrinsics">intrinsic function</a> because it takes a type as an
3476 <p>See the <a href="#int_varargs">variable argument processing</a> section.</p>
3480 <!-- *********************************************************************** -->
3481 <div class="doc_section"> <a name="intrinsics">Intrinsic Functions</a> </div>
3482 <!-- *********************************************************************** -->
3484 <div class="doc_text">
3486 <p>LLVM supports the notion of an "intrinsic function". These functions have
3487 well known names and semantics and are required to follow certain
3488 restrictions. Overall, these instructions represent an extension mechanism for
3489 the LLVM language that does not require changing all of the transformations in
3490 LLVM to add to the language (or the bytecode reader/writer, the parser,
3493 <p>Intrinsic function names must all start with an "<tt>llvm.</tt>" prefix. This
3494 prefix is reserved in LLVM for intrinsic names; thus, functions may not be named
3495 this. Intrinsic functions must always be external functions: you cannot define
3496 the body of intrinsic functions. Intrinsic functions may only be used in call
3497 or invoke instructions: it is illegal to take the address of an intrinsic
3498 function. Additionally, because intrinsic functions are part of the LLVM
3499 language, it is required that they all be documented here if any are added.</p>
3502 <p>To learn how to add an intrinsic function, please see the <a
3503 href="ExtendingLLVM.html">Extending LLVM Guide</a>.
3508 <!-- ======================================================================= -->
3509 <div class="doc_subsection">
3510 <a name="int_varargs">Variable Argument Handling Intrinsics</a>
3513 <div class="doc_text">
3515 <p>Variable argument support is defined in LLVM with the <a
3516 href="#i_va_arg"><tt>va_arg</tt></a> instruction and these three
3517 intrinsic functions. These functions are related to the similarly
3518 named macros defined in the <tt><stdarg.h></tt> header file.</p>
3520 <p>All of these functions operate on arguments that use a
3521 target-specific value type "<tt>va_list</tt>". The LLVM assembly
3522 language reference manual does not define what this type is, so all
3523 transformations should be prepared to handle intrinsics with any type
3526 <p>This example shows how the <a href="#i_va_arg"><tt>va_arg</tt></a>
3527 instruction and the variable argument handling intrinsic functions are
3531 define i32 %test(i32 %X, ...) {
3532 ; Initialize variable argument processing
3534 %ap2 = bitcast i8** %ap to i8*
3535 call void %<a href="#i_va_start">llvm.va_start</a>(i8* %ap2)
3537 ; Read a single integer argument
3538 %tmp = va_arg i8 ** %ap, i32
3540 ; Demonstrate usage of llvm.va_copy and llvm.va_end
3542 %aq2 = bitcast i8** %aq to i8*
3543 call void %<a href="#i_va_copy">llvm.va_copy</a>(i8 *%aq2, i8* %ap2)
3544 call void %<a href="#i_va_end">llvm.va_end</a>(i8* %aq2)
3546 ; Stop processing of arguments.
3547 call void %<a href="#i_va_end">llvm.va_end</a>(i8* %ap2)
3553 <!-- _______________________________________________________________________ -->
3554 <div class="doc_subsubsection">
3555 <a name="i_va_start">'<tt>llvm.va_start</tt>' Intrinsic</a>
3559 <div class="doc_text">
3561 <pre> declare void %llvm.va_start(i8* <arglist>)<br></pre>
3563 <P>The '<tt>llvm.va_start</tt>' intrinsic initializes
3564 <tt>*<arglist></tt> for subsequent use by <tt><a
3565 href="#i_va_arg">va_arg</a></tt>.</p>
3569 <P>The argument is a pointer to a <tt>va_list</tt> element to initialize.</p>
3573 <P>The '<tt>llvm.va_start</tt>' intrinsic works just like the <tt>va_start</tt>
3574 macro available in C. In a target-dependent way, it initializes the
3575 <tt>va_list</tt> element the argument points to, so that the next call to
3576 <tt>va_arg</tt> will produce the first variable argument passed to the function.
3577 Unlike the C <tt>va_start</tt> macro, this intrinsic does not need to know the
3578 last argument of the function, the compiler can figure that out.</p>
3582 <!-- _______________________________________________________________________ -->
3583 <div class="doc_subsubsection">
3584 <a name="i_va_end">'<tt>llvm.va_end</tt>' Intrinsic</a>
3587 <div class="doc_text">
3589 <pre> declare void %llvm.va_end(i8* <arglist>)<br></pre>
3592 <p>The '<tt>llvm.va_end</tt>' intrinsic destroys <tt><arglist></tt>
3593 which has been initialized previously with <tt><a href="#i_va_start">llvm.va_start</a></tt>
3594 or <tt><a href="#i_va_copy">llvm.va_copy</a></tt>.</p>
3598 <p>The argument is a <tt>va_list</tt> to destroy.</p>
3602 <p>The '<tt>llvm.va_end</tt>' intrinsic works just like the <tt>va_end</tt>
3603 macro available in C. In a target-dependent way, it destroys the <tt>va_list</tt>.
3604 Calls to <a href="#i_va_start"><tt>llvm.va_start</tt></a> and <a
3605 href="#i_va_copy"><tt>llvm.va_copy</tt></a> must be matched exactly
3606 with calls to <tt>llvm.va_end</tt>.</p>
3610 <!-- _______________________________________________________________________ -->
3611 <div class="doc_subsubsection">
3612 <a name="i_va_copy">'<tt>llvm.va_copy</tt>' Intrinsic</a>
3615 <div class="doc_text">
3620 declare void %llvm.va_copy(i8* <destarglist>, i8* <srcarglist>)
3625 <p>The '<tt>llvm.va_copy</tt>' intrinsic copies the current argument position from
3626 the source argument list to the destination argument list.</p>
3630 <p>The first argument is a pointer to a <tt>va_list</tt> element to initialize.
3631 The second argument is a pointer to a <tt>va_list</tt> element to copy from.</p>
3636 <p>The '<tt>llvm.va_copy</tt>' intrinsic works just like the <tt>va_copy</tt> macro
3637 available in C. In a target-dependent way, it copies the source
3638 <tt>va_list</tt> element into the destination list. This intrinsic is necessary
3639 because the <tt><a href="i_va_begin">llvm.va_begin</a></tt> intrinsic may be
3640 arbitrarily complex and require memory allocation, for example.</p>
3644 <!-- ======================================================================= -->
3645 <div class="doc_subsection">
3646 <a name="int_gc">Accurate Garbage Collection Intrinsics</a>
3649 <div class="doc_text">
3652 LLVM support for <a href="GarbageCollection.html">Accurate Garbage
3653 Collection</a> requires the implementation and generation of these intrinsics.
3654 These intrinsics allow identification of <a href="#i_gcroot">GC roots on the
3655 stack</a>, as well as garbage collector implementations that require <a
3656 href="#i_gcread">read</a> and <a href="#i_gcwrite">write</a> barriers.
3657 Front-ends for type-safe garbage collected languages should generate these
3658 intrinsics to make use of the LLVM garbage collectors. For more details, see <a
3659 href="GarbageCollection.html">Accurate Garbage Collection with LLVM</a>.
3663 <!-- _______________________________________________________________________ -->
3664 <div class="doc_subsubsection">
3665 <a name="i_gcroot">'<tt>llvm.gcroot</tt>' Intrinsic</a>
3668 <div class="doc_text">
3673 declare void %llvm.gcroot(<ty>** %ptrloc, <ty2>* %metadata)
3678 <p>The '<tt>llvm.gcroot</tt>' intrinsic declares the existence of a GC root to
3679 the code generator, and allows some metadata to be associated with it.</p>
3683 <p>The first argument specifies the address of a stack object that contains the
3684 root pointer. The second pointer (which must be either a constant or a global
3685 value address) contains the meta-data to be associated with the root.</p>
3689 <p>At runtime, a call to this intrinsics stores a null pointer into the "ptrloc"
3690 location. At compile-time, the code generator generates information to allow
3691 the runtime to find the pointer at GC safe points.
3697 <!-- _______________________________________________________________________ -->
3698 <div class="doc_subsubsection">
3699 <a name="i_gcread">'<tt>llvm.gcread</tt>' Intrinsic</a>
3702 <div class="doc_text">
3707 declare i8 * %llvm.gcread(i8 * %ObjPtr, i8 ** %Ptr)
3712 <p>The '<tt>llvm.gcread</tt>' intrinsic identifies reads of references from heap
3713 locations, allowing garbage collector implementations that require read
3718 <p>The second argument is the address to read from, which should be an address
3719 allocated from the garbage collector. The first object is a pointer to the
3720 start of the referenced object, if needed by the language runtime (otherwise
3725 <p>The '<tt>llvm.gcread</tt>' intrinsic has the same semantics as a load
3726 instruction, but may be replaced with substantially more complex code by the
3727 garbage collector runtime, as needed.</p>
3732 <!-- _______________________________________________________________________ -->
3733 <div class="doc_subsubsection">
3734 <a name="i_gcwrite">'<tt>llvm.gcwrite</tt>' Intrinsic</a>
3737 <div class="doc_text">
3742 declare void %llvm.gcwrite(i8 * %P1, i8 * %Obj, i8 ** %P2)
3747 <p>The '<tt>llvm.gcwrite</tt>' intrinsic identifies writes of references to heap
3748 locations, allowing garbage collector implementations that require write
3749 barriers (such as generational or reference counting collectors).</p>
3753 <p>The first argument is the reference to store, the second is the start of the
3754 object to store it to, and the third is the address of the field of Obj to
3755 store to. If the runtime does not require a pointer to the object, Obj may be
3760 <p>The '<tt>llvm.gcwrite</tt>' intrinsic has the same semantics as a store
3761 instruction, but may be replaced with substantially more complex code by the
3762 garbage collector runtime, as needed.</p>
3768 <!-- ======================================================================= -->
3769 <div class="doc_subsection">
3770 <a name="int_codegen">Code Generator Intrinsics</a>
3773 <div class="doc_text">
3775 These intrinsics are provided by LLVM to expose special features that may only
3776 be implemented with code generator support.
3781 <!-- _______________________________________________________________________ -->
3782 <div class="doc_subsubsection">
3783 <a name="i_returnaddress">'<tt>llvm.returnaddress</tt>' Intrinsic</a>
3786 <div class="doc_text">
3790 declare i8 *%llvm.returnaddress(i32 <level>)
3796 The '<tt>llvm.returnaddress</tt>' intrinsic attempts to compute a
3797 target-specific value indicating the return address of the current function
3798 or one of its callers.
3804 The argument to this intrinsic indicates which function to return the address
3805 for. Zero indicates the calling function, one indicates its caller, etc. The
3806 argument is <b>required</b> to be a constant integer value.
3812 The '<tt>llvm.returnaddress</tt>' intrinsic either returns a pointer indicating
3813 the return address of the specified call frame, or zero if it cannot be
3814 identified. The value returned by this intrinsic is likely to be incorrect or 0
3815 for arguments other than zero, so it should only be used for debugging purposes.
3819 Note that calling this intrinsic does not prevent function inlining or other
3820 aggressive transformations, so the value returned may not be that of the obvious
3821 source-language caller.
3826 <!-- _______________________________________________________________________ -->
3827 <div class="doc_subsubsection">
3828 <a name="i_frameaddress">'<tt>llvm.frameaddress</tt>' Intrinsic</a>
3831 <div class="doc_text">
3835 declare i8 *%llvm.frameaddress(i32 <level>)
3841 The '<tt>llvm.frameaddress</tt>' intrinsic attempts to return the
3842 target-specific frame pointer value for the specified stack frame.
3848 The argument to this intrinsic indicates which function to return the frame
3849 pointer for. Zero indicates the calling function, one indicates its caller,
3850 etc. The argument is <b>required</b> to be a constant integer value.
3856 The '<tt>llvm.frameaddress</tt>' intrinsic either returns a pointer indicating
3857 the frame address of the specified call frame, or zero if it cannot be
3858 identified. The value returned by this intrinsic is likely to be incorrect or 0
3859 for arguments other than zero, so it should only be used for debugging purposes.
3863 Note that calling this intrinsic does not prevent function inlining or other
3864 aggressive transformations, so the value returned may not be that of the obvious
3865 source-language caller.
3869 <!-- _______________________________________________________________________ -->
3870 <div class="doc_subsubsection">
3871 <a name="i_stacksave">'<tt>llvm.stacksave</tt>' Intrinsic</a>
3874 <div class="doc_text">
3878 declare i8 *%llvm.stacksave()
3884 The '<tt>llvm.stacksave</tt>' intrinsic is used to remember the current state of
3885 the function stack, for use with <a href="#i_stackrestore">
3886 <tt>llvm.stackrestore</tt></a>. This is useful for implementing language
3887 features like scoped automatic variable sized arrays in C99.
3893 This intrinsic returns a opaque pointer value that can be passed to <a
3894 href="#i_stackrestore"><tt>llvm.stackrestore</tt></a>. When an
3895 <tt>llvm.stackrestore</tt> intrinsic is executed with a value saved from
3896 <tt>llvm.stacksave</tt>, it effectively restores the state of the stack to the
3897 state it was in when the <tt>llvm.stacksave</tt> intrinsic executed. In
3898 practice, this pops any <a href="#i_alloca">alloca</a> blocks from the stack
3899 that were allocated after the <tt>llvm.stacksave</tt> was executed.
3904 <!-- _______________________________________________________________________ -->
3905 <div class="doc_subsubsection">
3906 <a name="i_stackrestore">'<tt>llvm.stackrestore</tt>' Intrinsic</a>
3909 <div class="doc_text">
3913 declare void %llvm.stackrestore(i8 * %ptr)
3919 The '<tt>llvm.stackrestore</tt>' intrinsic is used to restore the state of
3920 the function stack to the state it was in when the corresponding <a
3921 href="#llvm.stacksave"><tt>llvm.stacksave</tt></a> intrinsic executed. This is
3922 useful for implementing language features like scoped automatic variable sized
3929 See the description for <a href="#i_stacksave"><tt>llvm.stacksave</tt></a>.
3935 <!-- _______________________________________________________________________ -->
3936 <div class="doc_subsubsection">
3937 <a name="i_prefetch">'<tt>llvm.prefetch</tt>' Intrinsic</a>
3940 <div class="doc_text">
3944 declare void %llvm.prefetch(i8 * <address>,
3945 i32 <rw>, i32 <locality>)
3952 The '<tt>llvm.prefetch</tt>' intrinsic is a hint to the code generator to insert
3953 a prefetch instruction if supported; otherwise, it is a noop. Prefetches have
3955 effect on the behavior of the program but can change its performance
3962 <tt>address</tt> is the address to be prefetched, <tt>rw</tt> is the specifier
3963 determining if the fetch should be for a read (0) or write (1), and
3964 <tt>locality</tt> is a temporal locality specifier ranging from (0) - no
3965 locality, to (3) - extremely local keep in cache. The <tt>rw</tt> and
3966 <tt>locality</tt> arguments must be constant integers.
3972 This intrinsic does not modify the behavior of the program. In particular,
3973 prefetches cannot trap and do not produce a value. On targets that support this
3974 intrinsic, the prefetch can provide hints to the processor cache for better
3980 <!-- _______________________________________________________________________ -->
3981 <div class="doc_subsubsection">
3982 <a name="i_pcmarker">'<tt>llvm.pcmarker</tt>' Intrinsic</a>
3985 <div class="doc_text">
3989 declare void %llvm.pcmarker( i32 <id> )
3996 The '<tt>llvm.pcmarker</tt>' intrinsic is a method to export a Program Counter
3998 code to simulators and other tools. The method is target specific, but it is
3999 expected that the marker will use exported symbols to transmit the PC of the marker.
4000 The marker makes no guarantees that it will remain with any specific instruction
4001 after optimizations. It is possible that the presence of a marker will inhibit
4002 optimizations. The intended use is to be inserted after optimizations to allow
4003 correlations of simulation runs.
4009 <tt>id</tt> is a numerical id identifying the marker.
4015 This intrinsic does not modify the behavior of the program. Backends that do not
4016 support this intrinisic may ignore it.
4021 <!-- _______________________________________________________________________ -->
4022 <div class="doc_subsubsection">
4023 <a name="i_readcyclecounter">'<tt>llvm.readcyclecounter</tt>' Intrinsic</a>
4026 <div class="doc_text">
4030 declare i64 %llvm.readcyclecounter( )
4037 The '<tt>llvm.readcyclecounter</tt>' intrinsic provides access to the cycle
4038 counter register (or similar low latency, high accuracy clocks) on those targets
4039 that support it. On X86, it should map to RDTSC. On Alpha, it should map to RPCC.
4040 As the backing counters overflow quickly (on the order of 9 seconds on alpha), this
4041 should only be used for small timings.
4047 When directly supported, reading the cycle counter should not modify any memory.
4048 Implementations are allowed to either return a application specific value or a
4049 system wide value. On backends without support, this is lowered to a constant 0.
4054 <!-- ======================================================================= -->
4055 <div class="doc_subsection">
4056 <a name="int_libc">Standard C Library Intrinsics</a>
4059 <div class="doc_text">
4061 LLVM provides intrinsics for a few important standard C library functions.
4062 These intrinsics allow source-language front-ends to pass information about the
4063 alignment of the pointer arguments to the code generator, providing opportunity
4064 for more efficient code generation.
4069 <!-- _______________________________________________________________________ -->
4070 <div class="doc_subsubsection">
4071 <a name="i_memcpy">'<tt>llvm.memcpy</tt>' Intrinsic</a>
4074 <div class="doc_text">
4078 declare void %llvm.memcpy.i32(i8 * <dest>, i8 * <src>,
4079 i32 <len>, i32 <align>)
4080 declare void %llvm.memcpy.i64(i8 * <dest>, i8 * <src>,
4081 i64 <len>, i32 <align>)
4087 The '<tt>llvm.memcpy.*</tt>' intrinsics copy a block of memory from the source
4088 location to the destination location.
4092 Note that, unlike the standard libc function, the <tt>llvm.memcpy.*</tt>
4093 intrinsics do not return a value, and takes an extra alignment argument.
4099 The first argument is a pointer to the destination, the second is a pointer to
4100 the source. The third argument is an integer argument
4101 specifying the number of bytes to copy, and the fourth argument is the alignment
4102 of the source and destination locations.
4106 If the call to this intrinisic has an alignment value that is not 0 or 1, then
4107 the caller guarantees that both the source and destination pointers are aligned
4114 The '<tt>llvm.memcpy.*</tt>' intrinsics copy a block of memory from the source
4115 location to the destination location, which are not allowed to overlap. It
4116 copies "len" bytes of memory over. If the argument is known to be aligned to
4117 some boundary, this can be specified as the fourth argument, otherwise it should
4123 <!-- _______________________________________________________________________ -->
4124 <div class="doc_subsubsection">
4125 <a name="i_memmove">'<tt>llvm.memmove</tt>' Intrinsic</a>
4128 <div class="doc_text">
4132 declare void %llvm.memmove.i32(i8 * <dest>, i8 * <src>,
4133 i32 <len>, i32 <align>)
4134 declare void %llvm.memmove.i64(i8 * <dest>, i8 * <src>,
4135 i64 <len>, i32 <align>)
4141 The '<tt>llvm.memmove.*</tt>' intrinsics move a block of memory from the source
4142 location to the destination location. It is similar to the
4143 '<tt>llvm.memcmp</tt>' intrinsic but allows the two memory locations to overlap.
4147 Note that, unlike the standard libc function, the <tt>llvm.memmove.*</tt>
4148 intrinsics do not return a value, and takes an extra alignment argument.
4154 The first argument is a pointer to the destination, the second is a pointer to
4155 the source. The third argument is an integer argument
4156 specifying the number of bytes to copy, and the fourth argument is the alignment
4157 of the source and destination locations.
4161 If the call to this intrinisic has an alignment value that is not 0 or 1, then
4162 the caller guarantees that the source and destination pointers are aligned to
4169 The '<tt>llvm.memmove.*</tt>' intrinsics copy a block of memory from the source
4170 location to the destination location, which may overlap. It
4171 copies "len" bytes of memory over. If the argument is known to be aligned to
4172 some boundary, this can be specified as the fourth argument, otherwise it should
4178 <!-- _______________________________________________________________________ -->
4179 <div class="doc_subsubsection">
4180 <a name="i_memset">'<tt>llvm.memset.*</tt>' Intrinsics</a>
4183 <div class="doc_text">
4187 declare void %llvm.memset.i32(i8 * <dest>, i8 <val>,
4188 i32 <len>, i32 <align>)
4189 declare void %llvm.memset.i64(i8 * <dest>, i8 <val>,
4190 i64 <len>, i32 <align>)
4196 The '<tt>llvm.memset.*</tt>' intrinsics fill a block of memory with a particular
4201 Note that, unlike the standard libc function, the <tt>llvm.memset</tt> intrinsic
4202 does not return a value, and takes an extra alignment argument.
4208 The first argument is a pointer to the destination to fill, the second is the
4209 byte value to fill it with, the third argument is an integer
4210 argument specifying the number of bytes to fill, and the fourth argument is the
4211 known alignment of destination location.
4215 If the call to this intrinisic has an alignment value that is not 0 or 1, then
4216 the caller guarantees that the destination pointer is aligned to that boundary.
4222 The '<tt>llvm.memset.*</tt>' intrinsics fill "len" bytes of memory starting at
4224 destination location. If the argument is known to be aligned to some boundary,
4225 this can be specified as the fourth argument, otherwise it should be set to 0 or
4231 <!-- _______________________________________________________________________ -->
4232 <div class="doc_subsubsection">
4233 <a name="i_sqrt">'<tt>llvm.sqrt.*</tt>' Intrinsic</a>
4236 <div class="doc_text">
4240 declare float %llvm.sqrt.f32(float %Val)
4241 declare double %llvm.sqrt.f64(double %Val)
4247 The '<tt>llvm.sqrt</tt>' intrinsics return the sqrt of the specified operand,
4248 returning the same value as the libm '<tt>sqrt</tt>' function would. Unlike
4249 <tt>sqrt</tt> in libm, however, <tt>llvm.sqrt</tt> has undefined behavior for
4250 negative numbers (which allows for better optimization).
4256 The argument and return value are floating point numbers of the same type.
4262 This function returns the sqrt of the specified operand if it is a positive
4263 floating point number.
4267 <!-- _______________________________________________________________________ -->
4268 <div class="doc_subsubsection">
4269 <a name="i_powi">'<tt>llvm.powi.*</tt>' Intrinsic</a>
4272 <div class="doc_text">
4276 declare float %llvm.powi.f32(float %Val, i32 %power)
4277 declare double %llvm.powi.f64(double %Val, i32 %power)
4283 The '<tt>llvm.powi.*</tt>' intrinsics return the first operand raised to the
4284 specified (positive or negative) power. The order of evaluation of
4285 multiplications is not defined.
4291 The second argument is an integer power, and the first is a value to raise to
4298 This function returns the first value raised to the second power with an
4299 unspecified sequence of rounding operations.</p>
4303 <!-- ======================================================================= -->
4304 <div class="doc_subsection">
4305 <a name="int_manip">Bit Manipulation Intrinsics</a>
4308 <div class="doc_text">
4310 LLVM provides intrinsics for a few important bit manipulation operations.
4311 These allow efficient code generation for some algorithms.
4316 <!-- _______________________________________________________________________ -->
4317 <div class="doc_subsubsection">
4318 <a name="i_bswap">'<tt>llvm.bswap.*</tt>' Intrinsics</a>
4321 <div class="doc_text">
4325 declare i16 %llvm.bswap.i16(i16 <id>)
4326 declare i32 %llvm.bswap.i32(i32 <id>)
4327 declare i64 %llvm.bswap.i64(i64 <id>)
4333 The '<tt>llvm.bwsap</tt>' family of intrinsics is used to byteswap a 16, 32 or
4334 64 bit quantity. These are useful for performing operations on data that is not
4335 in the target's native byte order.
4341 The <tt>llvm.bswap.16</tt> intrinsic returns an i16 value that has the high
4342 and low byte of the input i16 swapped. Similarly, the <tt>llvm.bswap.i32</tt>
4343 intrinsic returns an i32 value that has the four bytes of the input i32
4344 swapped, so that if the input bytes are numbered 0, 1, 2, 3 then the returned
4345 i32 will have its bytes in 3, 2, 1, 0 order. The <tt>llvm.bswap.i64</tt>
4346 intrinsic extends this concept to 64 bits.
4351 <!-- _______________________________________________________________________ -->
4352 <div class="doc_subsubsection">
4353 <a name="int_ctpop">'<tt>llvm.ctpop.*</tt>' Intrinsic</a>
4356 <div class="doc_text">
4360 declare i8 %llvm.ctpop.i8 (i8 <src>)
4361 declare i16 %llvm.ctpop.i16(i16 <src>)
4362 declare i32 %llvm.ctpop.i32(i32 <src>)
4363 declare i64 %llvm.ctpop.i64(i64 <src>)
4369 The '<tt>llvm.ctpop</tt>' family of intrinsics counts the number of bits set in a
4376 The only argument is the value to be counted. The argument may be of any
4377 integer type. The return type must match the argument type.
4383 The '<tt>llvm.ctpop</tt>' intrinsic counts the 1's in a variable.
4387 <!-- _______________________________________________________________________ -->
4388 <div class="doc_subsubsection">
4389 <a name="int_ctlz">'<tt>llvm.ctlz.*</tt>' Intrinsic</a>
4392 <div class="doc_text">
4396 declare i8 %llvm.ctlz.i8 (i8 <src>)
4397 declare i16 %llvm.ctlz.i16(i16 <src>)
4398 declare i32 %llvm.ctlz.i32(i32 <src>)
4399 declare i64 %llvm.ctlz.i64(i64 <src>)
4405 The '<tt>llvm.ctlz</tt>' family of intrinsic functions counts the number of
4406 leading zeros in a variable.
4412 The only argument is the value to be counted. The argument may be of any
4413 integer type. The return type must match the argument type.
4419 The '<tt>llvm.ctlz</tt>' intrinsic counts the leading (most significant) zeros
4420 in a variable. If the src == 0 then the result is the size in bits of the type
4421 of src. For example, <tt>llvm.ctlz(i32 2) = 30</tt>.
4427 <!-- _______________________________________________________________________ -->
4428 <div class="doc_subsubsection">
4429 <a name="int_cttz">'<tt>llvm.cttz.*</tt>' Intrinsic</a>
4432 <div class="doc_text">
4436 declare i8 %llvm.cttz.i8 (i8 <src>)
4437 declare i16 %llvm.cttz.i16(i16 <src>)
4438 declare i32 %llvm.cttz.i32(i32 <src>)
4439 declare i64 %llvm.cttz.i64(i64 <src>)
4445 The '<tt>llvm.cttz</tt>' family of intrinsic functions counts the number of
4452 The only argument is the value to be counted. The argument may be of any
4453 integer type. The return type must match the argument type.
4459 The '<tt>llvm.cttz</tt>' intrinsic counts the trailing (least significant) zeros
4460 in a variable. If the src == 0 then the result is the size in bits of the type
4461 of src. For example, <tt>llvm.cttz(2) = 1</tt>.
4465 <!-- ======================================================================= -->
4466 <div class="doc_subsection">
4467 <a name="int_debugger">Debugger Intrinsics</a>
4470 <div class="doc_text">
4472 The LLVM debugger intrinsics (which all start with <tt>llvm.dbg.</tt> prefix),
4473 are described in the <a
4474 href="SourceLevelDebugging.html#format_common_intrinsics">LLVM Source Level
4475 Debugging</a> document.
4480 <!-- *********************************************************************** -->
4483 <a href="http://jigsaw.w3.org/css-validator/check/referer"><img
4484 src="http://jigsaw.w3.org/css-validator/images/vcss" alt="Valid CSS!"></a>
4485 <a href="http://validator.w3.org/check/referer"><img
4486 src="http://www.w3.org/Icons/valid-html401" alt="Valid HTML 4.01!" /></a>
4488 <a href="mailto:sabre@nondot.org">Chris Lattner</a><br>
4489 <a href="http://llvm.org">The LLVM Compiler Infrastructure</a><br>
4490 Last modified: $Date$