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>i32 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_external">externally visible</a></b></tt>:</dt>
457 <dd>If none of the above identifiers are used, the global is externally
458 visible, meaning that it participates in linkage and can be used to resolve
459 external symbol references.
462 <dt><tt><b><a name="linkage_externweak">extern_weak</a></b></tt>: </dt>
464 <dd>"<tt>extern_weak</tt>" TBD
468 The next two types of linkage are targeted for Microsoft Windows platform
469 only. They are designed to support importing (exporting) symbols from (to)
473 <dt><tt><b><a name="linkage_dllimport">dllimport</a></b></tt>: </dt>
475 <dd>"<tt>dllimport</tt>" linkage causes the compiler to reference a function
476 or variable via a global pointer to a pointer that is set up by the DLL
477 exporting the symbol. On Microsoft Windows targets, the pointer name is
478 formed by combining <code>_imp__</code> and the function or variable name.
481 <dt><tt><b><a name="linkage_dllexport">dllexport</a></b></tt>: </dt>
483 <dd>"<tt>dllexport</tt>" linkage causes the compiler to provide a global
484 pointer to a pointer in a DLL, so that it can be referenced with the
485 <tt>dllimport</tt> attribute. On Microsoft Windows targets, the pointer
486 name is formed by combining <code>_imp__</code> and the function or variable
492 <p><a name="linkage_external">For example, since the "<tt>.LC0</tt>"
493 variable is defined to be internal, if another module defined a "<tt>.LC0</tt>"
494 variable and was linked with this one, one of the two would be renamed,
495 preventing a collision. Since "<tt>main</tt>" and "<tt>puts</tt>" are
496 external (i.e., lacking any linkage declarations), they are accessible
497 outside of the current module.</p>
498 <p>It is illegal for a function <i>declaration</i>
499 to have any linkage type other than "externally visible", <tt>dllimport</tt>,
500 or <tt>extern_weak</tt>.</a></p>
504 <!-- ======================================================================= -->
505 <div class="doc_subsection">
506 <a name="callingconv">Calling Conventions</a>
509 <div class="doc_text">
511 <p>LLVM <a href="#functionstructure">functions</a>, <a href="#i_call">calls</a>
512 and <a href="#i_invoke">invokes</a> can all have an optional calling convention
513 specified for the call. The calling convention of any pair of dynamic
514 caller/callee must match, or the behavior of the program is undefined. The
515 following calling conventions are supported by LLVM, and more may be added in
519 <dt><b>"<tt>ccc</tt>" - The C calling convention</b>:</dt>
521 <dd>This calling convention (the default if no other calling convention is
522 specified) matches the target C calling conventions. This calling convention
523 supports varargs function calls and tolerates some mismatch in the declared
524 prototype and implemented declaration of the function (as does normal C).
527 <dt><b>"<tt>csretcc</tt>" - The C struct return calling convention</b>:</dt>
529 <dd>This calling convention matches the target C calling conventions, except
530 that functions with this convention are required to take a pointer as their
531 first argument, and the return type of the function must be void. This is
532 used for C functions that return aggregates by-value. In this case, the
533 function has been transformed to take a pointer to the struct as the first
534 argument to the function. For targets where the ABI specifies specific
535 behavior for structure-return calls, the calling convention can be used to
536 distinguish between struct return functions and other functions that take a
537 pointer to a struct as the first argument.
540 <dt><b>"<tt>fastcc</tt>" - The fast calling convention</b>:</dt>
542 <dd>This calling convention attempts to make calls as fast as possible
543 (e.g. by passing things in registers). This calling convention allows the
544 target to use whatever tricks it wants to produce fast code for the target,
545 without having to conform to an externally specified ABI. Implementations of
546 this convention should allow arbitrary tail call optimization to be supported.
547 This calling convention does not support varargs and requires the prototype of
548 all callees to exactly match the prototype of the function definition.
551 <dt><b>"<tt>coldcc</tt>" - The cold calling convention</b>:</dt>
553 <dd>This calling convention attempts to make code in the caller as efficient
554 as possible under the assumption that the call is not commonly executed. As
555 such, these calls often preserve all registers so that the call does not break
556 any live ranges in the caller side. This calling convention does not support
557 varargs and requires the prototype of all callees to exactly match the
558 prototype of the function definition.
561 <dt><b>"<tt>cc <<em>n</em>></tt>" - Numbered convention</b>:</dt>
563 <dd>Any calling convention may be specified by number, allowing
564 target-specific calling conventions to be used. Target specific calling
565 conventions start at 64.
569 <p>More calling conventions can be added/defined on an as-needed basis, to
570 support pascal conventions or any other well-known target-independent
575 <!-- ======================================================================= -->
576 <div class="doc_subsection">
577 <a name="globalvars">Global Variables</a>
580 <div class="doc_text">
582 <p>Global variables define regions of memory allocated at compilation time
583 instead of run-time. Global variables may optionally be initialized, may have
584 an explicit section to be placed in, and may
585 have an optional explicit alignment specified. A
586 variable may be defined as a global "constant," which indicates that the
587 contents of the variable will <b>never</b> be modified (enabling better
588 optimization, allowing the global data to be placed in the read-only section of
589 an executable, etc). Note that variables that need runtime initialization
590 cannot be marked "constant" as there is a store to the variable.</p>
593 LLVM explicitly allows <em>declarations</em> of global variables to be marked
594 constant, even if the final definition of the global is not. This capability
595 can be used to enable slightly better optimization of the program, but requires
596 the language definition to guarantee that optimizations based on the
597 'constantness' are valid for the translation units that do not include the
601 <p>As SSA values, global variables define pointer values that are in
602 scope (i.e. they dominate) all basic blocks in the program. Global
603 variables always define a pointer to their "content" type because they
604 describe a region of memory, and all memory objects in LLVM are
605 accessed through pointers.</p>
607 <p>LLVM allows an explicit section to be specified for globals. If the target
608 supports it, it will emit globals to the section specified.</p>
610 <p>An explicit alignment may be specified for a global. If not present, or if
611 the alignment is set to zero, the alignment of the global is set by the target
612 to whatever it feels convenient. If an explicit alignment is specified, the
613 global is forced to have at least that much alignment. All alignments must be
619 <!-- ======================================================================= -->
620 <div class="doc_subsection">
621 <a name="functionstructure">Functions</a>
624 <div class="doc_text">
626 <p>LLVM function definitions consist of the "<tt>define</tt>" keyord,
627 an optional <a href="#linkage">linkage type</a>, an optional
628 <a href="#callingconv">calling convention</a>, a return type, an optional
629 <a href="#paramattrs">parameter attribute</a> for the return type, a function
630 name, a (possibly empty) argument list (each with optional
631 <a href="#paramattrs">parameter attributes</a>), an optional section, an
632 optional alignment, an opening curly brace, a list of basic blocks, and a
633 closing curly brace. LLVM function declarations
634 consist of the "<tt>declare</tt>" keyword, an optional <a
635 href="#callingconv">calling convention</a>, a return type, an optional
636 <a href="#paramattrs">parameter attribute</a> for the return type, a function
637 name, a possibly empty list of arguments, and an optional alignment.</p>
639 <p>A function definition contains a list of basic blocks, forming the CFG for
640 the function. Each basic block may optionally start with a label (giving the
641 basic block a symbol table entry), contains a list of instructions, and ends
642 with a <a href="#terminators">terminator</a> instruction (such as a branch or
643 function return).</p>
645 <p>The first basic block in a program is special in two ways: it is immediately
646 executed on entrance to the function, and it is not allowed to have predecessor
647 basic blocks (i.e. there can not be any branches to the entry block of a
648 function). Because the block can have no predecessors, it also cannot have any
649 <a href="#i_phi">PHI nodes</a>.</p>
651 <p>LLVM functions are identified by their name and type signature. Hence, two
652 functions with the same name but different parameter lists or return values are
653 considered different functions, and LLVM will resolve references to each
656 <p>LLVM allows an explicit section to be specified for functions. If the target
657 supports it, it will emit functions to the section specified.</p>
659 <p>An explicit alignment may be specified for a function. If not present, or if
660 the alignment is set to zero, the alignment of the function is set by the target
661 to whatever it feels convenient. If an explicit alignment is specified, the
662 function is forced to have at least that much alignment. All alignments must be
667 <!-- ======================================================================= -->
668 <div class="doc_subsection"><a name="paramattrs">Parameter Attributes</a></div>
669 <div class="doc_text">
670 <p>The return type and each parameter of a function type may have a set of
671 <i>parameter attributes</i> associated with them. Parameter attributes are
672 used to communicate additional information about the result or parameters of
673 a function. Parameter attributes are considered to be part of the function
674 type so two functions types that differ only by the parameter attributes
675 are different function types.</p>
677 <p>Parameter attributes consist of an at sign (@) followed by either a single
678 keyword or a comma separate list of keywords enclosed in parentheses. For
680 %someFunc = i16 @zext (i8 @(sext) %someParam)
681 %someFunc = i16 @zext (i8 @zext %someParam)</pre>
682 Note that the two function types above are unique because the parameter
683 has a different attribute (@sext in the first one, @zext in the second).</p>
685 <p>Currently, only the following parameter attributes are defined:
687 <dt><tt>@zext</tt></dt>
688 <dd>This indicates that the parameter should be zero extended just before
689 a call to this function.</dd>
690 <dt><tt>@sext</tt></dt>
691 <dd>This indicates that the parameter should be sign extended just before
692 a call to this function.</dd>
695 <p>The current motivation for parameter attributes is to enable the sign and
696 zero extend information necessary for the C calling convention to be passed
697 from the front end to LLVM. The <tt>@zext</tt> and <tt>@sext</tt> attributes
698 are used by the code generator to perform the required extension. However,
699 parameter attributes are an orthogonal feature to calling conventions and
700 may be used for other purposes in the future.</p>
703 <!-- ======================================================================= -->
704 <div class="doc_subsection">
705 <a name="moduleasm">Module-Level Inline Assembly</a>
708 <div class="doc_text">
710 Modules may contain "module-level inline asm" blocks, which corresponds to the
711 GCC "file scope inline asm" blocks. These blocks are internally concatenated by
712 LLVM and treated as a single unit, but may be separated in the .ll file if
713 desired. The syntax is very simple:
716 <div class="doc_code"><pre>
717 module asm "inline asm code goes here"
718 module asm "more can go here"
721 <p>The strings can contain any character by escaping non-printable characters.
722 The escape sequence used is simply "\xx" where "xx" is the two digit hex code
727 The inline asm code is simply printed to the machine code .s file when
728 assembly code is generated.
733 <!-- *********************************************************************** -->
734 <div class="doc_section"> <a name="typesystem">Type System</a> </div>
735 <!-- *********************************************************************** -->
737 <div class="doc_text">
739 <p>The LLVM type system is one of the most important features of the
740 intermediate representation. Being typed enables a number of
741 optimizations to be performed on the IR directly, without having to do
742 extra analyses on the side before the transformation. A strong type
743 system makes it easier to read the generated code and enables novel
744 analyses and transformations that are not feasible to perform on normal
745 three address code representations.</p>
749 <!-- ======================================================================= -->
750 <div class="doc_subsection"> <a name="t_primitive">Primitive Types</a> </div>
751 <div class="doc_text">
752 <p>The primitive types are the fundamental building blocks of the LLVM
753 system. The current set of primitive types is as follows:</p>
755 <table class="layout">
760 <tr><th>Type</th><th>Description</th></tr>
761 <tr><td><tt>void</tt></td><td>No value</td></tr>
762 <tr><td><tt>i8</tt></td><td>Signless 8-bit value</td></tr>
763 <tr><td><tt>i32</tt></td><td>Signless 32-bit value</td></tr>
764 <tr><td><tt>float</tt></td><td>32-bit floating point value</td></tr>
765 <tr><td><tt>label</tt></td><td>Branch destination</td></tr>
772 <tr><th>Type</th><th>Description</th></tr>
773 <tr><td><tt>bool</tt></td><td>True or False value</td></tr>
774 <tr><td><tt>i16</tt></td><td>Signless 16-bit value</td></tr>
775 <tr><td><tt>i64</tt></td><td>Signless 64-bit value</td></tr>
776 <tr><td><tt>double</tt></td><td>64-bit floating point value</td></tr>
784 <!-- _______________________________________________________________________ -->
785 <div class="doc_subsubsection"> <a name="t_classifications">Type
786 Classifications</a> </div>
787 <div class="doc_text">
788 <p>These different primitive types fall into a few useful
791 <table border="1" cellspacing="0" cellpadding="4">
793 <tr><th>Classification</th><th>Types</th></tr>
795 <td><a name="t_integer">integer</a></td>
796 <td><tt>i8, i16, i32, i64</tt></td>
799 <td><a name="t_integral">integral</a></td>
800 <td><tt>bool, i8, i16, i32, i64</tt>
804 <td><a name="t_floating">floating point</a></td>
805 <td><tt>float, double</tt></td>
808 <td><a name="t_firstclass">first class</a></td>
809 <td><tt>bool, i8, i16, i32, i64, float, double, <br/>
810 <a href="#t_pointer">pointer</a>,<a href="#t_packed">packed</a></tt>
816 <p>The <a href="#t_firstclass">first class</a> types are perhaps the
817 most important. Values of these types are the only ones which can be
818 produced by instructions, passed as arguments, or used as operands to
819 instructions. This means that all structures and arrays must be
820 manipulated either by pointer or by component.</p>
823 <!-- ======================================================================= -->
824 <div class="doc_subsection"> <a name="t_derived">Derived Types</a> </div>
826 <div class="doc_text">
828 <p>The real power in LLVM comes from the derived types in the system.
829 This is what allows a programmer to represent arrays, functions,
830 pointers, and other useful types. Note that these derived types may be
831 recursive: For example, it is possible to have a two dimensional array.</p>
835 <!-- _______________________________________________________________________ -->
836 <div class="doc_subsubsection"> <a name="t_array">Array Type</a> </div>
838 <div class="doc_text">
842 <p>The array type is a very simple derived type that arranges elements
843 sequentially in memory. The array type requires a size (number of
844 elements) and an underlying data type.</p>
849 [<# elements> x <elementtype>]
852 <p>The number of elements is a constant integer value; elementtype may
853 be any type with a size.</p>
856 <table class="layout">
859 <tt>[40 x i32 ]</tt><br/>
860 <tt>[41 x i32 ]</tt><br/>
861 <tt>[40 x i8]</tt><br/>
864 Array of 40 32-bit integer values.<br/>
865 Array of 41 32-bit integer values.<br/>
866 Array of 40 8-bit integer values.<br/>
870 <p>Here are some examples of multidimensional arrays:</p>
871 <table class="layout">
874 <tt>[3 x [4 x i32]]</tt><br/>
875 <tt>[12 x [10 x float]]</tt><br/>
876 <tt>[2 x [3 x [4 x i16]]]</tt><br/>
879 3x4 array of 32-bit integer values.<br/>
880 12x10 array of single precision floating point values.<br/>
881 2x3x4 array of 16-bit integer values.<br/>
886 <p>Note that 'variable sized arrays' can be implemented in LLVM with a zero
887 length array. Normally, accesses past the end of an array are undefined in
888 LLVM (e.g. it is illegal to access the 5th element of a 3 element array).
889 As a special case, however, zero length arrays are recognized to be variable
890 length. This allows implementation of 'pascal style arrays' with the LLVM
891 type "{ i32, [0 x float]}", for example.</p>
895 <!-- _______________________________________________________________________ -->
896 <div class="doc_subsubsection"> <a name="t_function">Function Type</a> </div>
897 <div class="doc_text">
899 <p>The function type can be thought of as a function signature. It
900 consists of a return type and a list of formal parameter types.
901 Function types are usually used to build virtual function tables
902 (which are structures of pointers to functions), for indirect function
903 calls, and when defining a function.</p>
905 The return type of a function type cannot be an aggregate type.
908 <pre> <returntype> (<parameter list>)<br></pre>
909 <p>...where '<tt><parameter list></tt>' is a comma-separated list of type
910 specifiers. Optionally, the parameter list may include a type <tt>...</tt>,
911 which indicates that the function takes a variable number of arguments.
912 Variable argument functions can access their arguments with the <a
913 href="#int_varargs">variable argument handling intrinsic</a> functions.</p>
915 <table class="layout">
917 <td class="left"><tt>i32 (i32)</tt></td>
918 <td class="left">function taking an <tt>i32</tt>, returning an <tt>i32</tt>
920 </tr><tr class="layout">
921 <td class="left"><tt>float (i16 @sext, i32 *) *
923 <td class="left"><a href="#t_pointer">Pointer</a> to a function that takes
924 an <tt>i16</tt> that should be sign extended and a
925 <a href="#t_pointer">pointer</a> to <tt>i32</tt>, returning
928 </tr><tr class="layout">
929 <td class="left"><tt>i32 (i8*, ...)</tt></td>
930 <td class="left">A vararg function that takes at least one
931 <a href="#t_pointer">pointer</a> to <tt>i8 </tt> (char in C),
932 which returns an integer. This is the signature for <tt>printf</tt> in
939 <!-- _______________________________________________________________________ -->
940 <div class="doc_subsubsection"> <a name="t_struct">Structure Type</a> </div>
941 <div class="doc_text">
943 <p>The structure type is used to represent a collection of data members
944 together in memory. The packing of the field types is defined to match
945 the ABI of the underlying processor. The elements of a structure may
946 be any type that has a size.</p>
947 <p>Structures are accessed using '<tt><a href="#i_load">load</a></tt>
948 and '<tt><a href="#i_store">store</a></tt>' by getting a pointer to a
949 field with the '<tt><a href="#i_getelementptr">getelementptr</a></tt>'
952 <pre> { <type list> }<br></pre>
954 <table class="layout">
957 <tt>{ i32, i32, i32 }</tt><br/>
958 <tt>{ float, i32 (i32) * }</tt><br/>
961 a triple of three <tt>i32</tt> values<br/>
962 A pair, where the first element is a <tt>float</tt> and the second element
963 is a <a href="#t_pointer">pointer</a> to a <a href="#t_function">function</a>
964 that takes an <tt>i32</tt>, returning an <tt>i32</tt>.<br/>
970 <!-- _______________________________________________________________________ -->
971 <div class="doc_subsubsection"> <a name="t_pstruct">Packed Structure Type</a>
973 <div class="doc_text">
975 <p>The packed structure type is used to represent a collection of data members
976 together in memory. There is no padding between fields. Further, the alignment
977 of a packed structure is 1 byte. The elements of a packed structure may
978 be any type that has a size.</p>
979 <p>Structures are accessed using '<tt><a href="#i_load">load</a></tt>
980 and '<tt><a href="#i_store">store</a></tt>' by getting a pointer to a
981 field with the '<tt><a href="#i_getelementptr">getelementptr</a></tt>'
984 <pre> < { <type list> } > <br></pre>
986 <table class="layout">
989 <tt> < { i32, i32, i32 } > </tt><br/>
990 <tt> < { float, i32 (i32) * } > </tt><br/>
993 a triple of three <tt>i32</tt> values<br/>
994 A pair, where the first element is a <tt>float</tt> and the second element
995 is a <a href="#t_pointer">pointer</a> to a <a href="#t_function">function</a>
996 that takes an <tt>i32</tt>, returning an <tt>i32</tt>.<br/>
1002 <!-- _______________________________________________________________________ -->
1003 <div class="doc_subsubsection"> <a name="t_pointer">Pointer Type</a> </div>
1004 <div class="doc_text">
1006 <p>As in many languages, the pointer type represents a pointer or
1007 reference to another object, which must live in memory.</p>
1009 <pre> <type> *<br></pre>
1011 <table class="layout">
1014 <tt>[4x i32]*</tt><br/>
1015 <tt>i32 (i32 *) *</tt><br/>
1018 A <a href="#t_pointer">pointer</a> to <a href="#t_array">array</a> of
1019 four <tt>i32</tt> values<br/>
1020 A <a href="#t_pointer">pointer</a> to a <a
1021 href="#t_function">function</a> that takes an <tt>i32*</tt>, returning an
1028 <!-- _______________________________________________________________________ -->
1029 <div class="doc_subsubsection"> <a name="t_packed">Packed Type</a> </div>
1030 <div class="doc_text">
1034 <p>A packed type is a simple derived type that represents a vector
1035 of elements. Packed types are used when multiple primitive data
1036 are operated in parallel using a single instruction (SIMD).
1037 A packed type requires a size (number of
1038 elements) and an underlying primitive data type. Vectors must have a power
1039 of two length (1, 2, 4, 8, 16 ...). Packed types are
1040 considered <a href="#t_firstclass">first class</a>.</p>
1045 < <# elements> x <elementtype> >
1048 <p>The number of elements is a constant integer value; elementtype may
1049 be any integral or floating point type.</p>
1053 <table class="layout">
1056 <tt><4 x i32></tt><br/>
1057 <tt><8 x float></tt><br/>
1058 <tt><2 x i64></tt><br/>
1061 Packed vector of 4 32-bit integer values.<br/>
1062 Packed vector of 8 floating-point values.<br/>
1063 Packed vector of 2 64-bit integer values.<br/>
1069 <!-- _______________________________________________________________________ -->
1070 <div class="doc_subsubsection"> <a name="t_opaque">Opaque Type</a> </div>
1071 <div class="doc_text">
1075 <p>Opaque types are used to represent unknown types in the system. This
1076 corresponds (for example) to the C notion of a foward declared structure type.
1077 In LLVM, opaque types can eventually be resolved to any type (not just a
1078 structure type).</p>
1088 <table class="layout">
1094 An opaque type.<br/>
1101 <!-- *********************************************************************** -->
1102 <div class="doc_section"> <a name="constants">Constants</a> </div>
1103 <!-- *********************************************************************** -->
1105 <div class="doc_text">
1107 <p>LLVM has several different basic types of constants. This section describes
1108 them all and their syntax.</p>
1112 <!-- ======================================================================= -->
1113 <div class="doc_subsection"><a name="simpleconstants">Simple Constants</a></div>
1115 <div class="doc_text">
1118 <dt><b>Boolean constants</b></dt>
1120 <dd>The two strings '<tt>true</tt>' and '<tt>false</tt>' are both valid
1121 constants of the <tt><a href="#t_primitive">bool</a></tt> type.
1124 <dt><b>Integer constants</b></dt>
1126 <dd>Standard integers (such as '4') are constants of the <a
1127 href="#t_integer">integer</a> type. Negative numbers may be used with
1131 <dt><b>Floating point constants</b></dt>
1133 <dd>Floating point constants use standard decimal notation (e.g. 123.421),
1134 exponential notation (e.g. 1.23421e+2), or a more precise hexadecimal
1135 notation (see below). Floating point constants must have a <a
1136 href="#t_floating">floating point</a> type. </dd>
1138 <dt><b>Null pointer constants</b></dt>
1140 <dd>The identifier '<tt>null</tt>' is recognized as a null pointer constant
1141 and must be of <a href="#t_pointer">pointer type</a>.</dd>
1145 <p>The one non-intuitive notation for constants is the optional hexadecimal form
1146 of floating point constants. For example, the form '<tt>double
1147 0x432ff973cafa8000</tt>' is equivalent to (but harder to read than) '<tt>double
1148 4.5e+15</tt>'. The only time hexadecimal floating point constants are required
1149 (and the only time that they are generated by the disassembler) is when a
1150 floating point constant must be emitted but it cannot be represented as a
1151 decimal floating point number. For example, NaN's, infinities, and other
1152 special values are represented in their IEEE hexadecimal format so that
1153 assembly and disassembly do not cause any bits to change in the constants.</p>
1157 <!-- ======================================================================= -->
1158 <div class="doc_subsection"><a name="aggregateconstants">Aggregate Constants</a>
1161 <div class="doc_text">
1162 <p>Aggregate constants arise from aggregation of simple constants
1163 and smaller aggregate constants.</p>
1166 <dt><b>Structure constants</b></dt>
1168 <dd>Structure constants are represented with notation similar to structure
1169 type definitions (a comma separated list of elements, surrounded by braces
1170 (<tt>{}</tt>)). For example: "<tt>{ i32 4, float 17.0, i32* %G }</tt>",
1171 where "<tt>%G</tt>" is declared as "<tt>%G = external global i32</tt>". Structure constants
1172 must have <a href="#t_struct">structure type</a>, and the number and
1173 types of elements must match those specified by the type.
1176 <dt><b>Array constants</b></dt>
1178 <dd>Array constants are represented with notation similar to array type
1179 definitions (a comma separated list of elements, surrounded by square brackets
1180 (<tt>[]</tt>)). For example: "<tt>[ i32 42, i32 11, i32 74 ]</tt>". Array
1181 constants must have <a href="#t_array">array type</a>, and the number and
1182 types of elements must match those specified by the type.
1185 <dt><b>Packed constants</b></dt>
1187 <dd>Packed constants are represented with notation similar to packed type
1188 definitions (a comma separated list of elements, surrounded by
1189 less-than/greater-than's (<tt><></tt>)). For example: "<tt>< i32 42,
1190 i32 11, i32 74, i32 100 ></tt>". Packed constants must have <a
1191 href="#t_packed">packed type</a>, and the number and types of elements must
1192 match those specified by the type.
1195 <dt><b>Zero initialization</b></dt>
1197 <dd>The string '<tt>zeroinitializer</tt>' can be used to zero initialize a
1198 value to zero of <em>any</em> type, including scalar and aggregate types.
1199 This is often used to avoid having to print large zero initializers (e.g. for
1200 large arrays) and is always exactly equivalent to using explicit zero
1207 <!-- ======================================================================= -->
1208 <div class="doc_subsection">
1209 <a name="globalconstants">Global Variable and Function Addresses</a>
1212 <div class="doc_text">
1214 <p>The addresses of <a href="#globalvars">global variables</a> and <a
1215 href="#functionstructure">functions</a> are always implicitly valid (link-time)
1216 constants. These constants are explicitly referenced when the <a
1217 href="#identifiers">identifier for the global</a> is used and always have <a
1218 href="#t_pointer">pointer</a> type. For example, the following is a legal LLVM
1224 %Z = global [2 x i32*] [ i32* %X, i32* %Y ]
1229 <!-- ======================================================================= -->
1230 <div class="doc_subsection"><a name="undefvalues">Undefined Values</a></div>
1231 <div class="doc_text">
1232 <p>The string '<tt>undef</tt>' is recognized as a type-less constant that has
1233 no specific value. Undefined values may be of any type and be used anywhere
1234 a constant is permitted.</p>
1236 <p>Undefined values indicate to the compiler that the program is well defined
1237 no matter what value is used, giving the compiler more freedom to optimize.
1241 <!-- ======================================================================= -->
1242 <div class="doc_subsection"><a name="constantexprs">Constant Expressions</a>
1245 <div class="doc_text">
1247 <p>Constant expressions are used to allow expressions involving other constants
1248 to be used as constants. Constant expressions may be of any <a
1249 href="#t_firstclass">first class</a> type and may involve any LLVM operation
1250 that does not have side effects (e.g. load and call are not supported). The
1251 following is the syntax for constant expressions:</p>
1254 <dt><b><tt>trunc ( CST to TYPE )</tt></b></dt>
1255 <dd>Truncate a constant to another type. The bit size of CST must be larger
1256 than the bit size of TYPE. Both types must be integral.</dd>
1258 <dt><b><tt>zext ( CST to TYPE )</tt></b></dt>
1259 <dd>Zero extend a constant to another type. The bit size of CST must be
1260 smaller or equal to the bit size of TYPE. Both types must be integral.</dd>
1262 <dt><b><tt>sext ( CST to TYPE )</tt></b></dt>
1263 <dd>Sign extend a constant to another type. The bit size of CST must be
1264 smaller or equal to the bit size of TYPE. Both types must be integral.</dd>
1266 <dt><b><tt>fptrunc ( CST to TYPE )</tt></b></dt>
1267 <dd>Truncate a floating point constant to another floating point type. The
1268 size of CST must be larger than the size of TYPE. Both types must be
1269 floating point.</dd>
1271 <dt><b><tt>fpext ( CST to TYPE )</tt></b></dt>
1272 <dd>Floating point extend a constant to another type. The size of CST must be
1273 smaller or equal to the size of TYPE. Both types must be floating point.</dd>
1275 <dt><b><tt>fp2uint ( CST to TYPE )</tt></b></dt>
1276 <dd>Convert a floating point constant to the corresponding unsigned integer
1277 constant. TYPE must be an integer type. CST must be floating point. If the
1278 value won't fit in the integer type, the results are undefined.</dd>
1280 <dt><b><tt>fptosi ( CST to TYPE )</tt></b></dt>
1281 <dd>Convert a floating point constant to the corresponding signed integer
1282 constant. TYPE must be an integer type. CST must be floating point. If the
1283 value won't fit in the integer type, the results are undefined.</dd>
1285 <dt><b><tt>uitofp ( CST to TYPE )</tt></b></dt>
1286 <dd>Convert an unsigned integer constant to the corresponding floating point
1287 constant. TYPE must be floating point. CST must be of integer type. If the
1288 value won't fit in the floating point type, the results are undefined.</dd>
1290 <dt><b><tt>sitofp ( CST to TYPE )</tt></b></dt>
1291 <dd>Convert a signed integer constant to the corresponding floating point
1292 constant. TYPE must be floating point. CST must be of integer type. If the
1293 value won't fit in the floating point type, the results are undefined.</dd>
1295 <dt><b><tt>ptrtoint ( CST to TYPE )</tt></b></dt>
1296 <dd>Convert a pointer typed constant to the corresponding integer constant
1297 TYPE must be an integer type. CST must be of pointer type. The CST value is
1298 zero extended, truncated, or unchanged to make it fit in TYPE.</dd>
1300 <dt><b><tt>inttoptr ( CST to TYPE )</tt></b></dt>
1301 <dd>Convert a integer constant to a pointer constant. TYPE must be a
1302 pointer type. CST must be of integer type. The CST value is zero extended,
1303 truncated, or unchanged to make it fit in a pointer size. This one is
1304 <i>really</i> dangerous!</dd>
1306 <dt><b><tt>bitcast ( CST to TYPE )</tt></b></dt>
1307 <dd>Convert a constant, CST, to another TYPE. The size of CST and TYPE must be
1308 identical (same number of bits). The conversion is done as if the CST value
1309 was stored to memory and read back as TYPE. In other words, no bits change
1310 with this operator, just the type. This can be used for conversion of
1311 packed types to any other type, as long as they have the same bit width. For
1312 pointers it is only valid to cast to another pointer type.
1315 <dt><b><tt>getelementptr ( CSTPTR, IDX0, IDX1, ... )</tt></b></dt>
1317 <dd>Perform the <a href="#i_getelementptr">getelementptr operation</a> on
1318 constants. As with the <a href="#i_getelementptr">getelementptr</a>
1319 instruction, the index list may have zero or more indexes, which are required
1320 to make sense for the type of "CSTPTR".</dd>
1322 <dt><b><tt>select ( COND, VAL1, VAL2 )</tt></b></dt>
1324 <dd>Perform the <a href="#i_select">select operation</a> on
1327 <dt><b><tt>icmp COND ( VAL1, VAL2 )</tt></b></dt>
1328 <dd>Performs the <a href="#i_icmp">icmp operation</a> on constants.</dd>
1330 <dt><b><tt>fcmp COND ( VAL1, VAL2 )</tt></b></dt>
1331 <dd>Performs the <a href="#i_fcmp">fcmp operation</a> on constants.</dd>
1333 <dt><b><tt>extractelement ( VAL, IDX )</tt></b></dt>
1335 <dd>Perform the <a href="#i_extractelement">extractelement
1336 operation</a> on constants.
1338 <dt><b><tt>insertelement ( VAL, ELT, IDX )</tt></b></dt>
1340 <dd>Perform the <a href="#i_insertelement">insertelement
1341 operation</a> on constants.</dd>
1344 <dt><b><tt>shufflevector ( VEC1, VEC2, IDXMASK )</tt></b></dt>
1346 <dd>Perform the <a href="#i_shufflevector">shufflevector
1347 operation</a> on constants.</dd>
1349 <dt><b><tt>OPCODE ( LHS, RHS )</tt></b></dt>
1351 <dd>Perform the specified operation of the LHS and RHS constants. OPCODE may
1352 be any of the <a href="#binaryops">binary</a> or <a href="#bitwiseops">bitwise
1353 binary</a> operations. The constraints on operands are the same as those for
1354 the corresponding instruction (e.g. no bitwise operations on floating point
1355 values are allowed).</dd>
1359 <!-- *********************************************************************** -->
1360 <div class="doc_section"> <a name="othervalues">Other Values</a> </div>
1361 <!-- *********************************************************************** -->
1363 <!-- ======================================================================= -->
1364 <div class="doc_subsection">
1365 <a name="inlineasm">Inline Assembler Expressions</a>
1368 <div class="doc_text">
1371 LLVM supports inline assembler expressions (as opposed to <a href="#moduleasm">
1372 Module-Level Inline Assembly</a>) through the use of a special value. This
1373 value represents the inline assembler as a string (containing the instructions
1374 to emit), a list of operand constraints (stored as a string), and a flag that
1375 indicates whether or not the inline asm expression has side effects. An example
1376 inline assembler expression is:
1380 i32 (i32) asm "bswap $0", "=r,r"
1384 Inline assembler expressions may <b>only</b> be used as the callee operand of
1385 a <a href="#i_call"><tt>call</tt> instruction</a>. Thus, typically we have:
1389 %X = call i32 asm "<a href="#i_bswap">bswap</a> $0", "=r,r"(i32 %Y)
1393 Inline asms with side effects not visible in the constraint list must be marked
1394 as having side effects. This is done through the use of the
1395 '<tt>sideeffect</tt>' keyword, like so:
1399 call void asm sideeffect "eieio", ""()
1402 <p>TODO: The format of the asm and constraints string still need to be
1403 documented here. Constraints on what can be done (e.g. duplication, moving, etc
1404 need to be documented).
1409 <!-- *********************************************************************** -->
1410 <div class="doc_section"> <a name="instref">Instruction Reference</a> </div>
1411 <!-- *********************************************************************** -->
1413 <div class="doc_text">
1415 <p>The LLVM instruction set consists of several different
1416 classifications of instructions: <a href="#terminators">terminator
1417 instructions</a>, <a href="#binaryops">binary instructions</a>,
1418 <a href="#bitwiseops">bitwise binary instructions</a>, <a
1419 href="#memoryops">memory instructions</a>, and <a href="#otherops">other
1420 instructions</a>.</p>
1424 <!-- ======================================================================= -->
1425 <div class="doc_subsection"> <a name="terminators">Terminator
1426 Instructions</a> </div>
1428 <div class="doc_text">
1430 <p>As mentioned <a href="#functionstructure">previously</a>, every
1431 basic block in a program ends with a "Terminator" instruction, which
1432 indicates which block should be executed after the current block is
1433 finished. These terminator instructions typically yield a '<tt>void</tt>'
1434 value: they produce control flow, not values (the one exception being
1435 the '<a href="#i_invoke"><tt>invoke</tt></a>' instruction).</p>
1436 <p>There are six different terminator instructions: the '<a
1437 href="#i_ret"><tt>ret</tt></a>' instruction, the '<a href="#i_br"><tt>br</tt></a>'
1438 instruction, the '<a href="#i_switch"><tt>switch</tt></a>' instruction,
1439 the '<a href="#i_invoke"><tt>invoke</tt></a>' instruction, the '<a
1440 href="#i_unwind"><tt>unwind</tt></a>' instruction, and the '<a
1441 href="#i_unreachable"><tt>unreachable</tt></a>' instruction.</p>
1445 <!-- _______________________________________________________________________ -->
1446 <div class="doc_subsubsection"> <a name="i_ret">'<tt>ret</tt>'
1447 Instruction</a> </div>
1448 <div class="doc_text">
1450 <pre> ret <type> <value> <i>; Return a value from a non-void function</i>
1451 ret void <i>; Return from void function</i>
1454 <p>The '<tt>ret</tt>' instruction is used to return control flow (and a
1455 value) from a function back to the caller.</p>
1456 <p>There are two forms of the '<tt>ret</tt>' instruction: one that
1457 returns a value and then causes control flow, and one that just causes
1458 control flow to occur.</p>
1460 <p>The '<tt>ret</tt>' instruction may return any '<a
1461 href="#t_firstclass">first class</a>' type. Notice that a function is
1462 not <a href="#wellformed">well formed</a> if there exists a '<tt>ret</tt>'
1463 instruction inside of the function that returns a value that does not
1464 match the return type of the function.</p>
1466 <p>When the '<tt>ret</tt>' instruction is executed, control flow
1467 returns back to the calling function's context. If the caller is a "<a
1468 href="#i_call"><tt>call</tt></a>" instruction, execution continues at
1469 the instruction after the call. If the caller was an "<a
1470 href="#i_invoke"><tt>invoke</tt></a>" instruction, execution continues
1471 at the beginning of the "normal" destination block. If the instruction
1472 returns a value, that value shall set the call or invoke instruction's
1475 <pre> ret i32 5 <i>; Return an integer value of 5</i>
1476 ret void <i>; Return from a void function</i>
1479 <!-- _______________________________________________________________________ -->
1480 <div class="doc_subsubsection"> <a name="i_br">'<tt>br</tt>' Instruction</a> </div>
1481 <div class="doc_text">
1483 <pre> br bool <cond>, label <iftrue>, label <iffalse><br> br label <dest> <i>; Unconditional branch</i>
1486 <p>The '<tt>br</tt>' instruction is used to cause control flow to
1487 transfer to a different basic block in the current function. There are
1488 two forms of this instruction, corresponding to a conditional branch
1489 and an unconditional branch.</p>
1491 <p>The conditional branch form of the '<tt>br</tt>' instruction takes a
1492 single '<tt>bool</tt>' value and two '<tt>label</tt>' values. The
1493 unconditional form of the '<tt>br</tt>' instruction takes a single '<tt>label</tt>'
1494 value as a target.</p>
1496 <p>Upon execution of a conditional '<tt>br</tt>' instruction, the '<tt>bool</tt>'
1497 argument is evaluated. If the value is <tt>true</tt>, control flows
1498 to the '<tt>iftrue</tt>' <tt>label</tt> argument. If "cond" is <tt>false</tt>,
1499 control flows to the '<tt>iffalse</tt>' <tt>label</tt> argument.</p>
1501 <pre>Test:<br> %cond = <a href="#i_icmp">icmp</a> eq, i32 %a, %b<br> br bool %cond, label %IfEqual, label %IfUnequal<br>IfEqual:<br> <a
1502 href="#i_ret">ret</a> i32 1<br>IfUnequal:<br> <a href="#i_ret">ret</a> i32 0<br></pre>
1504 <!-- _______________________________________________________________________ -->
1505 <div class="doc_subsubsection">
1506 <a name="i_switch">'<tt>switch</tt>' Instruction</a>
1509 <div class="doc_text">
1513 switch <intty> <value>, label <defaultdest> [ <intty> <val>, label <dest> ... ]
1518 <p>The '<tt>switch</tt>' instruction is used to transfer control flow to one of
1519 several different places. It is a generalization of the '<tt>br</tt>'
1520 instruction, allowing a branch to occur to one of many possible
1526 <p>The '<tt>switch</tt>' instruction uses three parameters: an integer
1527 comparison value '<tt>value</tt>', a default '<tt>label</tt>' destination, and
1528 an array of pairs of comparison value constants and '<tt>label</tt>'s. The
1529 table is not allowed to contain duplicate constant entries.</p>
1533 <p>The <tt>switch</tt> instruction specifies a table of values and
1534 destinations. When the '<tt>switch</tt>' instruction is executed, this
1535 table is searched for the given value. If the value is found, control flow is
1536 transfered to the corresponding destination; otherwise, control flow is
1537 transfered to the default destination.</p>
1539 <h5>Implementation:</h5>
1541 <p>Depending on properties of the target machine and the particular
1542 <tt>switch</tt> instruction, this instruction may be code generated in different
1543 ways. For example, it could be generated as a series of chained conditional
1544 branches or with a lookup table.</p>
1549 <i>; Emulate a conditional br instruction</i>
1550 %Val = <a href="#i_zext">zext</a> bool %value to i32
1551 switch i32 %Val, label %truedest [i32 0, label %falsedest ]
1553 <i>; Emulate an unconditional br instruction</i>
1554 switch i32 0, label %dest [ ]
1556 <i>; Implement a jump table:</i>
1557 switch i32 %val, label %otherwise [ i32 0, label %onzero
1559 i32 2, label %ontwo ]
1563 <!-- _______________________________________________________________________ -->
1564 <div class="doc_subsubsection">
1565 <a name="i_invoke">'<tt>invoke</tt>' Instruction</a>
1568 <div class="doc_text">
1573 <result> = invoke [<a href="#callingconv">cconv</a>] <ptr to function ty> %<function ptr val>(<function args>)
1574 to label <normal label> unwind label <exception label>
1579 <p>The '<tt>invoke</tt>' instruction causes control to transfer to a specified
1580 function, with the possibility of control flow transfer to either the
1581 '<tt>normal</tt>' label or the
1582 '<tt>exception</tt>' label. If the callee function returns with the
1583 "<tt><a href="#i_ret">ret</a></tt>" instruction, control flow will return to the
1584 "normal" label. If the callee (or any indirect callees) returns with the "<a
1585 href="#i_unwind"><tt>unwind</tt></a>" instruction, control is interrupted and
1586 continued at the dynamically nearest "exception" label.</p>
1590 <p>This instruction requires several arguments:</p>
1594 The optional "cconv" marker indicates which <a href="callingconv">calling
1595 convention</a> the call should use. If none is specified, the call defaults
1596 to using C calling conventions.
1598 <li>'<tt>ptr to function ty</tt>': shall be the signature of the pointer to
1599 function value being invoked. In most cases, this is a direct function
1600 invocation, but indirect <tt>invoke</tt>s are just as possible, branching off
1601 an arbitrary pointer to function value.
1604 <li>'<tt>function ptr val</tt>': An LLVM value containing a pointer to a
1605 function to be invoked. </li>
1607 <li>'<tt>function args</tt>': argument list whose types match the function
1608 signature argument types. If the function signature indicates the function
1609 accepts a variable number of arguments, the extra arguments can be
1612 <li>'<tt>normal label</tt>': the label reached when the called function
1613 executes a '<tt><a href="#i_ret">ret</a></tt>' instruction. </li>
1615 <li>'<tt>exception label</tt>': the label reached when a callee returns with
1616 the <a href="#i_unwind"><tt>unwind</tt></a> instruction. </li>
1622 <p>This instruction is designed to operate as a standard '<tt><a
1623 href="#i_call">call</a></tt>' instruction in most regards. The primary
1624 difference is that it establishes an association with a label, which is used by
1625 the runtime library to unwind the stack.</p>
1627 <p>This instruction is used in languages with destructors to ensure that proper
1628 cleanup is performed in the case of either a <tt>longjmp</tt> or a thrown
1629 exception. Additionally, this is important for implementation of
1630 '<tt>catch</tt>' clauses in high-level languages that support them.</p>
1634 %retval = invoke i32 %Test(i32 15) to label %Continue
1635 unwind label %TestCleanup <i>; {i32}:retval set</i>
1636 %retval = invoke <a href="#callingconv">coldcc</a> i32 %Test(i32 15) to label %Continue
1637 unwind label %TestCleanup <i>; {i32}:retval set</i>
1642 <!-- _______________________________________________________________________ -->
1644 <div class="doc_subsubsection"> <a name="i_unwind">'<tt>unwind</tt>'
1645 Instruction</a> </div>
1647 <div class="doc_text">
1656 <p>The '<tt>unwind</tt>' instruction unwinds the stack, continuing control flow
1657 at the first callee in the dynamic call stack which used an <a
1658 href="#i_invoke"><tt>invoke</tt></a> instruction to perform the call. This is
1659 primarily used to implement exception handling.</p>
1663 <p>The '<tt>unwind</tt>' intrinsic causes execution of the current function to
1664 immediately halt. The dynamic call stack is then searched for the first <a
1665 href="#i_invoke"><tt>invoke</tt></a> instruction on the call stack. Once found,
1666 execution continues at the "exceptional" destination block specified by the
1667 <tt>invoke</tt> instruction. If there is no <tt>invoke</tt> instruction in the
1668 dynamic call chain, undefined behavior results.</p>
1671 <!-- _______________________________________________________________________ -->
1673 <div class="doc_subsubsection"> <a name="i_unreachable">'<tt>unreachable</tt>'
1674 Instruction</a> </div>
1676 <div class="doc_text">
1685 <p>The '<tt>unreachable</tt>' instruction has no defined semantics. This
1686 instruction is used to inform the optimizer that a particular portion of the
1687 code is not reachable. This can be used to indicate that the code after a
1688 no-return function cannot be reached, and other facts.</p>
1692 <p>The '<tt>unreachable</tt>' instruction has no defined semantics.</p>
1697 <!-- ======================================================================= -->
1698 <div class="doc_subsection"> <a name="binaryops">Binary Operations</a> </div>
1699 <div class="doc_text">
1700 <p>Binary operators are used to do most of the computation in a
1701 program. They require two operands, execute an operation on them, and
1702 produce a single value. The operands might represent
1703 multiple data, as is the case with the <a href="#t_packed">packed</a> data type.
1704 The result value of a binary operator is not
1705 necessarily the same type as its operands.</p>
1706 <p>There are several different binary operators:</p>
1708 <!-- _______________________________________________________________________ -->
1709 <div class="doc_subsubsection"> <a name="i_add">'<tt>add</tt>'
1710 Instruction</a> </div>
1711 <div class="doc_text">
1713 <pre> <result> = add <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1716 <p>The '<tt>add</tt>' instruction returns the sum of its two operands.</p>
1718 <p>The two arguments to the '<tt>add</tt>' instruction must be either <a
1719 href="#t_integer">integer</a> or <a href="#t_floating">floating point</a> values.
1720 This instruction can also take <a href="#t_packed">packed</a> versions of the values.
1721 Both arguments must have identical types.</p>
1723 <p>The value produced is the integer or floating point sum of the two
1726 <pre> <result> = add i32 4, %var <i>; yields {i32}:result = 4 + %var</i>
1729 <!-- _______________________________________________________________________ -->
1730 <div class="doc_subsubsection"> <a name="i_sub">'<tt>sub</tt>'
1731 Instruction</a> </div>
1732 <div class="doc_text">
1734 <pre> <result> = sub <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1737 <p>The '<tt>sub</tt>' instruction returns the difference of its two
1739 <p>Note that the '<tt>sub</tt>' instruction is used to represent the '<tt>neg</tt>'
1740 instruction present in most other intermediate representations.</p>
1742 <p>The two arguments to the '<tt>sub</tt>' instruction must be either <a
1743 href="#t_integer">integer</a> or <a href="#t_floating">floating point</a>
1745 This instruction can also take <a href="#t_packed">packed</a> versions of the values.
1746 Both arguments must have identical types.</p>
1748 <p>The value produced is the integer or floating point difference of
1749 the two operands.</p>
1751 <pre> <result> = sub i32 4, %var <i>; yields {i32}:result = 4 - %var</i>
1752 <result> = sub i32 0, %val <i>; yields {i32}:result = -%var</i>
1755 <!-- _______________________________________________________________________ -->
1756 <div class="doc_subsubsection"> <a name="i_mul">'<tt>mul</tt>'
1757 Instruction</a> </div>
1758 <div class="doc_text">
1760 <pre> <result> = mul <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1763 <p>The '<tt>mul</tt>' instruction returns the product of its two
1766 <p>The two arguments to the '<tt>mul</tt>' instruction must be either <a
1767 href="#t_integer">integer</a> or <a href="#t_floating">floating point</a>
1769 This instruction can also take <a href="#t_packed">packed</a> versions of the values.
1770 Both arguments must have identical types.</p>
1772 <p>The value produced is the integer or floating point product of the
1774 <p>Because the operands are the same width, the result of an integer
1775 multiplication is the same whether the operands should be deemed unsigned or
1778 <pre> <result> = mul i32 4, %var <i>; yields {i32}:result = 4 * %var</i>
1781 <!-- _______________________________________________________________________ -->
1782 <div class="doc_subsubsection"> <a name="i_udiv">'<tt>udiv</tt>' Instruction
1784 <div class="doc_text">
1786 <pre> <result> = udiv <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1789 <p>The '<tt>udiv</tt>' instruction returns the quotient of its two
1792 <p>The two arguments to the '<tt>udiv</tt>' instruction must be
1793 <a href="#t_integer">integer</a> values. Both arguments must have identical
1794 types. This instruction can also take <a href="#t_packed">packed</a> versions
1795 of the values in which case the elements must be integers.</p>
1797 <p>The value produced is the unsigned integer quotient of the two operands. This
1798 instruction always performs an unsigned division operation, regardless of
1799 whether the arguments are unsigned or not.</p>
1801 <pre> <result> = udiv i32 4, %var <i>; yields {i32}:result = 4 / %var</i>
1804 <!-- _______________________________________________________________________ -->
1805 <div class="doc_subsubsection"> <a name="i_sdiv">'<tt>sdiv</tt>' Instruction
1807 <div class="doc_text">
1809 <pre> <result> = sdiv <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1812 <p>The '<tt>sdiv</tt>' instruction returns the quotient of its two
1815 <p>The two arguments to the '<tt>sdiv</tt>' instruction must be
1816 <a href="#t_integer">integer</a> values. Both arguments must have identical
1817 types. This instruction can also take <a href="#t_packed">packed</a> versions
1818 of the values in which case the elements must be integers.</p>
1820 <p>The value produced is the signed integer quotient of the two operands. This
1821 instruction always performs a signed division operation, regardless of whether
1822 the arguments are signed or not.</p>
1824 <pre> <result> = sdiv i32 4, %var <i>; yields {i32}:result = 4 / %var</i>
1827 <!-- _______________________________________________________________________ -->
1828 <div class="doc_subsubsection"> <a name="i_fdiv">'<tt>fdiv</tt>'
1829 Instruction</a> </div>
1830 <div class="doc_text">
1832 <pre> <result> = fdiv <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1835 <p>The '<tt>fdiv</tt>' instruction returns the quotient of its two
1838 <p>The two arguments to the '<tt>div</tt>' instruction must be
1839 <a href="#t_floating">floating point</a> values. Both arguments must have
1840 identical types. This instruction can also take <a href="#t_packed">packed</a>
1841 versions of the values in which case the elements must be floating point.</p>
1843 <p>The value produced is the floating point quotient of the two operands.</p>
1845 <pre> <result> = fdiv float 4.0, %var <i>; yields {float}:result = 4.0 / %var</i>
1848 <!-- _______________________________________________________________________ -->
1849 <div class="doc_subsubsection"> <a name="i_urem">'<tt>urem</tt>' Instruction</a>
1851 <div class="doc_text">
1853 <pre> <result> = urem <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1856 <p>The '<tt>urem</tt>' instruction returns the remainder from the
1857 unsigned division of its two arguments.</p>
1859 <p>The two arguments to the '<tt>urem</tt>' instruction must be
1860 <a href="#t_integer">integer</a> values. Both arguments must have identical
1863 <p>This instruction returns the unsigned integer <i>remainder</i> of a division.
1864 This instruction always performs an unsigned division to get the remainder,
1865 regardless of whether the arguments are unsigned or not.</p>
1867 <pre> <result> = urem i32 4, %var <i>; yields {i32}:result = 4 % %var</i>
1871 <!-- _______________________________________________________________________ -->
1872 <div class="doc_subsubsection"> <a name="i_srem">'<tt>srem</tt>'
1873 Instruction</a> </div>
1874 <div class="doc_text">
1876 <pre> <result> = srem <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1879 <p>The '<tt>srem</tt>' instruction returns the remainder from the
1880 signed division of its two operands.</p>
1882 <p>The two arguments to the '<tt>srem</tt>' instruction must be
1883 <a href="#t_integer">integer</a> values. Both arguments must have identical
1886 <p>This instruction returns the <i>remainder</i> of a division (where the result
1887 has the same sign as the divisor), not the <i>modulus</i> (where the
1888 result has the same sign as the dividend) of a value. For more
1889 information about the difference, see <a
1890 href="http://mathforum.org/dr.math/problems/anne.4.28.99.html">The
1893 <pre> <result> = srem i32 4, %var <i>; yields {i32}:result = 4 % %var</i>
1897 <!-- _______________________________________________________________________ -->
1898 <div class="doc_subsubsection"> <a name="i_frem">'<tt>frem</tt>'
1899 Instruction</a> </div>
1900 <div class="doc_text">
1902 <pre> <result> = frem <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1905 <p>The '<tt>frem</tt>' instruction returns the remainder from the
1906 division of its two operands.</p>
1908 <p>The two arguments to the '<tt>frem</tt>' instruction must be
1909 <a href="#t_floating">floating point</a> values. Both arguments must have
1910 identical types.</p>
1912 <p>This instruction returns the <i>remainder</i> of a division.</p>
1914 <pre> <result> = frem float 4.0, %var <i>; yields {float}:result = 4.0 % %var</i>
1918 <!-- ======================================================================= -->
1919 <div class="doc_subsection"> <a name="bitwiseops">Bitwise Binary
1920 Operations</a> </div>
1921 <div class="doc_text">
1922 <p>Bitwise binary operators are used to do various forms of
1923 bit-twiddling in a program. They are generally very efficient
1924 instructions and can commonly be strength reduced from other
1925 instructions. They require two operands, execute an operation on them,
1926 and produce a single value. The resulting value of the bitwise binary
1927 operators is always the same type as its first operand.</p>
1929 <!-- _______________________________________________________________________ -->
1930 <div class="doc_subsubsection"> <a name="i_and">'<tt>and</tt>'
1931 Instruction</a> </div>
1932 <div class="doc_text">
1934 <pre> <result> = and <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1937 <p>The '<tt>and</tt>' instruction returns the bitwise logical and of
1938 its two operands.</p>
1940 <p>The two arguments to the '<tt>and</tt>' instruction must be <a
1941 href="#t_integral">integral</a> values. Both arguments must have
1942 identical types.</p>
1944 <p>The truth table used for the '<tt>and</tt>' instruction is:</p>
1946 <div style="align: center">
1947 <table border="1" cellspacing="0" cellpadding="4">
1978 <pre> <result> = and i32 4, %var <i>; yields {i32}:result = 4 & %var</i>
1979 <result> = and i32 15, 40 <i>; yields {i32}:result = 8</i>
1980 <result> = and i32 4, 8 <i>; yields {i32}:result = 0</i>
1983 <!-- _______________________________________________________________________ -->
1984 <div class="doc_subsubsection"> <a name="i_or">'<tt>or</tt>' Instruction</a> </div>
1985 <div class="doc_text">
1987 <pre> <result> = or <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1990 <p>The '<tt>or</tt>' instruction returns the bitwise logical inclusive
1991 or of its two operands.</p>
1993 <p>The two arguments to the '<tt>or</tt>' instruction must be <a
1994 href="#t_integral">integral</a> values. Both arguments must have
1995 identical types.</p>
1997 <p>The truth table used for the '<tt>or</tt>' instruction is:</p>
1999 <div style="align: center">
2000 <table border="1" cellspacing="0" cellpadding="4">
2031 <pre> <result> = or i32 4, %var <i>; yields {i32}:result = 4 | %var</i>
2032 <result> = or i32 15, 40 <i>; yields {i32}:result = 47</i>
2033 <result> = or i32 4, 8 <i>; yields {i32}:result = 12</i>
2036 <!-- _______________________________________________________________________ -->
2037 <div class="doc_subsubsection"> <a name="i_xor">'<tt>xor</tt>'
2038 Instruction</a> </div>
2039 <div class="doc_text">
2041 <pre> <result> = xor <ty> <var1>, <var2> <i>; yields {ty}:result</i>
2044 <p>The '<tt>xor</tt>' instruction returns the bitwise logical exclusive
2045 or of its two operands. The <tt>xor</tt> is used to implement the
2046 "one's complement" operation, which is the "~" operator in C.</p>
2048 <p>The two arguments to the '<tt>xor</tt>' instruction must be <a
2049 href="#t_integral">integral</a> values. Both arguments must have
2050 identical types.</p>
2052 <p>The truth table used for the '<tt>xor</tt>' instruction is:</p>
2054 <div style="align: center">
2055 <table border="1" cellspacing="0" cellpadding="4">
2087 <pre> <result> = xor i32 4, %var <i>; yields {i32}:result = 4 ^ %var</i>
2088 <result> = xor i32 15, 40 <i>; yields {i32}:result = 39</i>
2089 <result> = xor i32 4, 8 <i>; yields {i32}:result = 12</i>
2090 <result> = xor i32 %V, -1 <i>; yields {i32}:result = ~%V</i>
2093 <!-- _______________________________________________________________________ -->
2094 <div class="doc_subsubsection"> <a name="i_shl">'<tt>shl</tt>'
2095 Instruction</a> </div>
2096 <div class="doc_text">
2098 <pre> <result> = shl <ty> <var1>, i8 <var2> <i>; yields {ty}:result</i>
2101 <p>The '<tt>shl</tt>' instruction returns the first operand shifted to
2102 the left a specified number of bits.</p>
2104 <p>The first argument to the '<tt>shl</tt>' instruction must be an <a
2105 href="#t_integer">integer</a> type. The second argument must be an '<tt>i8</tt>'
2108 <p>The value produced is <tt>var1</tt> * 2<sup><tt>var2</tt></sup>.</p>
2110 <pre> <result> = shl i32 4, i8 %var <i>; yields {i32}:result = 4 << %var</i>
2111 <result> = shl i32 4, i8 2 <i>; yields {i32}:result = 16</i>
2112 <result> = shl i32 1, i8 10 <i>; yields {i32}:result = 1024</i>
2115 <!-- _______________________________________________________________________ -->
2116 <div class="doc_subsubsection"> <a name="i_lshr">'<tt>lshr</tt>'
2117 Instruction</a> </div>
2118 <div class="doc_text">
2120 <pre> <result> = lshr <ty> <var1>, i8 <var2> <i>; yields {ty}:result</i>
2124 <p>The '<tt>lshr</tt>' instruction (logical shift right) returns the first
2125 operand shifted to the right a specified number of bits.</p>
2128 <p>The first argument to the '<tt>lshr</tt>' instruction must be an <a
2129 href="#t_integer">integer</a> type. The second argument must be an '<tt>i8</tt>' type.</p>
2132 <p>This instruction always performs a logical shift right operation. The
2133 <tt>var2</tt> most significant bits will be filled with zero bits after the
2138 <result> = lshr i32 4, i8 1 <i>; yields {i32}:result = 2</i>
2139 <result> = lshr i32 4, i8 2 <i>; yields {i32}:result = 1</i>
2140 <result> = lshr i8 4, i8 3 <i>; yields {i8 }:result = 0</i>
2141 <result> = lshr i8 -2, i8 1 <i>; yields {i8 }:result = 0x7FFFFFFF </i>
2145 <!-- ======================================================================= -->
2146 <div class="doc_subsubsection"> <a name="i_ashr">'<tt>ashr</tt>'
2147 Instruction</a> </div>
2148 <div class="doc_text">
2151 <pre> <result> = ashr <ty> <var1>, i8 <var2> <i>; yields {ty}:result</i>
2155 <p>The '<tt>ashr</tt>' instruction (arithmetic shift right) returns the first
2156 operand shifted to the right a specified number of bits.</p>
2159 <p>The first argument to the '<tt>ashr</tt>' instruction must be an
2160 <a href="#t_integer">integer</a> type. The second argument must be an
2161 '<tt>i8</tt>' type.</p>
2164 <p>This instruction always performs an arithmetic shift right operation,
2165 regardless of whether the arguments are signed or not. The <tt>var2</tt> most
2166 significant bits will be filled with the sign bit of <tt>var1</tt>.</p>
2170 <result> = ashr i32 4, i8 1 <i>; yields {i32}:result = 2</i>
2171 <result> = ashr i32 4, i8 2 <i>; yields {i32}:result = 1</i>
2172 <result> = ashr i8 4, i8 3 <i>; yields {i8}:result = 0</i>
2173 <result> = ashr i8 -2, i8 1 <i>; yields {i8 }:result = -1</i>
2177 <!-- ======================================================================= -->
2178 <div class="doc_subsection">
2179 <a name="vectorops">Vector Operations</a>
2182 <div class="doc_text">
2184 <p>LLVM supports several instructions to represent vector operations in a
2185 target-independent manner. This instructions cover the element-access and
2186 vector-specific operations needed to process vectors effectively. While LLVM
2187 does directly support these vector operations, many sophisticated algorithms
2188 will want to use target-specific intrinsics to take full advantage of a specific
2193 <!-- _______________________________________________________________________ -->
2194 <div class="doc_subsubsection">
2195 <a name="i_extractelement">'<tt>extractelement</tt>' Instruction</a>
2198 <div class="doc_text">
2203 <result> = extractelement <n x <ty>> <val>, i32 <idx> <i>; yields <ty></i>
2209 The '<tt>extractelement</tt>' instruction extracts a single scalar
2210 element from a packed vector at a specified index.
2217 The first operand of an '<tt>extractelement</tt>' instruction is a
2218 value of <a href="#t_packed">packed</a> type. The second operand is
2219 an index indicating the position from which to extract the element.
2220 The index may be a variable.</p>
2225 The result is a scalar of the same type as the element type of
2226 <tt>val</tt>. Its value is the value at position <tt>idx</tt> of
2227 <tt>val</tt>. If <tt>idx</tt> exceeds the length of <tt>val</tt>, the
2228 results are undefined.
2234 %result = extractelement <4 x i32> %vec, i32 0 <i>; yields i32</i>
2239 <!-- _______________________________________________________________________ -->
2240 <div class="doc_subsubsection">
2241 <a name="i_insertelement">'<tt>insertelement</tt>' Instruction</a>
2244 <div class="doc_text">
2249 <result> = insertelement <n x <ty>> <val>, <ty> <elt>, i32 <idx> <i>; yields <n x <ty>></i>
2255 The '<tt>insertelement</tt>' instruction inserts a scalar
2256 element into a packed vector at a specified index.
2263 The first operand of an '<tt>insertelement</tt>' instruction is a
2264 value of <a href="#t_packed">packed</a> type. The second operand is a
2265 scalar value whose type must equal the element type of the first
2266 operand. The third operand is an index indicating the position at
2267 which to insert the value. The index may be a variable.</p>
2272 The result is a packed vector of the same type as <tt>val</tt>. Its
2273 element values are those of <tt>val</tt> except at position
2274 <tt>idx</tt>, where it gets the value <tt>elt</tt>. If <tt>idx</tt>
2275 exceeds the length of <tt>val</tt>, the results are undefined.
2281 %result = insertelement <4 x i32> %vec, i32 1, i32 0 <i>; yields <4 x i32></i>
2285 <!-- _______________________________________________________________________ -->
2286 <div class="doc_subsubsection">
2287 <a name="i_shufflevector">'<tt>shufflevector</tt>' Instruction</a>
2290 <div class="doc_text">
2295 <result> = shufflevector <n x <ty>> <v1>, <n x <ty>> <v2>, <n x i32> <mask> <i>; yields <n x <ty>></i>
2301 The '<tt>shufflevector</tt>' instruction constructs a permutation of elements
2302 from two input vectors, returning a vector of the same type.
2308 The first two operands of a '<tt>shufflevector</tt>' instruction are vectors
2309 with types that match each other and types that match the result of the
2310 instruction. The third argument is a shuffle mask, which has the same number
2311 of elements as the other vector type, but whose element type is always 'i32'.
2315 The shuffle mask operand is required to be a constant vector with either
2316 constant integer or undef values.
2322 The elements of the two input vectors are numbered from left to right across
2323 both of the vectors. The shuffle mask operand specifies, for each element of
2324 the result vector, which element of the two input registers the result element
2325 gets. The element selector may be undef (meaning "don't care") and the second
2326 operand may be undef if performing a shuffle from only one vector.
2332 %result = shufflevector <4 x i32> %v1, <4 x i32> %v2,
2333 <4 x i32> <i32 0, i32 4, i32 1, i32 5> <i>; yields <4 x i32></i>
2334 %result = shufflevector <4 x i32> %v1, <4 x i32> undef,
2335 <4 x i32> <i32 0, i32 1, i32 2, i32 3> <i>; yields <4 x i32></i> - Identity shuffle.
2340 <!-- ======================================================================= -->
2341 <div class="doc_subsection">
2342 <a name="memoryops">Memory Access and Addressing Operations</a>
2345 <div class="doc_text">
2347 <p>A key design point of an SSA-based representation is how it
2348 represents memory. In LLVM, no memory locations are in SSA form, which
2349 makes things very simple. This section describes how to read, write,
2350 allocate, and free memory in LLVM.</p>
2354 <!-- _______________________________________________________________________ -->
2355 <div class="doc_subsubsection">
2356 <a name="i_malloc">'<tt>malloc</tt>' Instruction</a>
2359 <div class="doc_text">
2364 <result> = malloc <type>[, i32 <NumElements>][, align <alignment>] <i>; yields {type*}:result</i>
2369 <p>The '<tt>malloc</tt>' instruction allocates memory from the system
2370 heap and returns a pointer to it.</p>
2374 <p>The '<tt>malloc</tt>' instruction allocates
2375 <tt>sizeof(<type>)*NumElements</tt>
2376 bytes of memory from the operating system and returns a pointer of the
2377 appropriate type to the program. If "NumElements" is specified, it is the
2378 number of elements allocated. If an alignment is specified, the value result
2379 of the allocation is guaranteed to be aligned to at least that boundary. If
2380 not specified, or if zero, the target can choose to align the allocation on any
2381 convenient boundary.</p>
2383 <p>'<tt>type</tt>' must be a sized type.</p>
2387 <p>Memory is allocated using the system "<tt>malloc</tt>" function, and
2388 a pointer is returned.</p>
2393 %array = malloc [4 x i8 ] <i>; yields {[%4 x i8]*}:array</i>
2395 %size = <a href="#i_add">add</a> i32 2, 2 <i>; yields {i32}:size = i32 4</i>
2396 %array1 = malloc i8, i32 4 <i>; yields {i8*}:array1</i>
2397 %array2 = malloc [12 x i8], i32 %size <i>; yields {[12 x i8]*}:array2</i>
2398 %array3 = malloc i32, i32 4, align 1024 <i>; yields {i32*}:array3</i>
2399 %array4 = malloc i32, align 1024 <i>; yields {i32*}:array4</i>
2403 <!-- _______________________________________________________________________ -->
2404 <div class="doc_subsubsection">
2405 <a name="i_free">'<tt>free</tt>' Instruction</a>
2408 <div class="doc_text">
2413 free <type> <value> <i>; yields {void}</i>
2418 <p>The '<tt>free</tt>' instruction returns memory back to the unused
2419 memory heap to be reallocated in the future.</p>
2423 <p>'<tt>value</tt>' shall be a pointer value that points to a value
2424 that was allocated with the '<tt><a href="#i_malloc">malloc</a></tt>'
2429 <p>Access to the memory pointed to by the pointer is no longer defined
2430 after this instruction executes.</p>
2435 %array = <a href="#i_malloc">malloc</a> [4 x i8] <i>; yields {[4 x i8]*}:array</i>
2436 free [4 x i8]* %array
2440 <!-- _______________________________________________________________________ -->
2441 <div class="doc_subsubsection">
2442 <a name="i_alloca">'<tt>alloca</tt>' Instruction</a>
2445 <div class="doc_text">
2450 <result> = alloca <type>[, i32 <NumElements>][, align <alignment>] <i>; yields {type*}:result</i>
2455 <p>The '<tt>alloca</tt>' instruction allocates memory on the current
2456 stack frame of the procedure that is live until the current function
2457 returns to its caller.</p>
2461 <p>The '<tt>alloca</tt>' instruction allocates <tt>sizeof(<type>)*NumElements</tt>
2462 bytes of memory on the runtime stack, returning a pointer of the
2463 appropriate type to the program. If "NumElements" is specified, it is the
2464 number of elements allocated. If an alignment is specified, the value result
2465 of the allocation is guaranteed to be aligned to at least that boundary. If
2466 not specified, or if zero, the target can choose to align the allocation on any
2467 convenient boundary.</p>
2469 <p>'<tt>type</tt>' may be any sized type.</p>
2473 <p>Memory is allocated; a pointer is returned. '<tt>alloca</tt>'d
2474 memory is automatically released when the function returns. The '<tt>alloca</tt>'
2475 instruction is commonly used to represent automatic variables that must
2476 have an address available. When the function returns (either with the <tt><a
2477 href="#i_ret">ret</a></tt> or <tt><a href="#i_unwind">unwind</a></tt>
2478 instructions), the memory is reclaimed.</p>
2483 %ptr = alloca i32 <i>; yields {i32*}:ptr</i>
2484 %ptr = alloca i32, i32 4 <i>; yields {i32*}:ptr</i>
2485 %ptr = alloca i32, i32 4, align 1024 <i>; yields {i32*}:ptr</i>
2486 %ptr = alloca i32, align 1024 <i>; yields {i32*}:ptr</i>
2490 <!-- _______________________________________________________________________ -->
2491 <div class="doc_subsubsection"> <a name="i_load">'<tt>load</tt>'
2492 Instruction</a> </div>
2493 <div class="doc_text">
2495 <pre> <result> = load <ty>* <pointer><br> <result> = volatile load <ty>* <pointer><br></pre>
2497 <p>The '<tt>load</tt>' instruction is used to read from memory.</p>
2499 <p>The argument to the '<tt>load</tt>' instruction specifies the memory
2500 address from which to load. The pointer must point to a <a
2501 href="#t_firstclass">first class</a> type. If the <tt>load</tt> is
2502 marked as <tt>volatile</tt>, then the optimizer is not allowed to modify
2503 the number or order of execution of this <tt>load</tt> with other
2504 volatile <tt>load</tt> and <tt><a href="#i_store">store</a></tt>
2507 <p>The location of memory pointed to is loaded.</p>
2509 <pre> %ptr = <a href="#i_alloca">alloca</a> i32 <i>; yields {i32*}:ptr</i>
2511 href="#i_store">store</a> i32 3, i32* %ptr <i>; yields {void}</i>
2512 %val = load i32* %ptr <i>; yields {i32}:val = i32 3</i>
2515 <!-- _______________________________________________________________________ -->
2516 <div class="doc_subsubsection"> <a name="i_store">'<tt>store</tt>'
2517 Instruction</a> </div>
2518 <div class="doc_text">
2520 <pre> store <ty> <value>, <ty>* <pointer> <i>; yields {void}</i>
2521 volatile store <ty> <value>, <ty>* <pointer> <i>; yields {void}</i>
2524 <p>The '<tt>store</tt>' instruction is used to write to memory.</p>
2526 <p>There are two arguments to the '<tt>store</tt>' instruction: a value
2527 to store and an address in which to store it. The type of the '<tt><pointer></tt>'
2528 operand must be a pointer to the type of the '<tt><value></tt>'
2529 operand. If the <tt>store</tt> is marked as <tt>volatile</tt>, then the
2530 optimizer is not allowed to modify the number or order of execution of
2531 this <tt>store</tt> with other volatile <tt>load</tt> and <tt><a
2532 href="#i_store">store</a></tt> instructions.</p>
2534 <p>The contents of memory are updated to contain '<tt><value></tt>'
2535 at the location specified by the '<tt><pointer></tt>' operand.</p>
2537 <pre> %ptr = <a href="#i_alloca">alloca</a> i32 <i>; yields {i32*}:ptr</i>
2539 href="#i_store">store</a> i32 3, i32* %ptr <i>; yields {void}</i>
2540 %val = load i32* %ptr <i>; yields {i32}:val = i32 3</i>
2544 <!-- _______________________________________________________________________ -->
2545 <div class="doc_subsubsection">
2546 <a name="i_getelementptr">'<tt>getelementptr</tt>' Instruction</a>
2549 <div class="doc_text">
2552 <result> = getelementptr <ty>* <ptrval>{, <ty> <idx>}*
2558 The '<tt>getelementptr</tt>' instruction is used to get the address of a
2559 subelement of an aggregate data structure.</p>
2563 <p>This instruction takes a list of integer operands that indicate what
2564 elements of the aggregate object to index to. The actual types of the arguments
2565 provided depend on the type of the first pointer argument. The
2566 '<tt>getelementptr</tt>' instruction is used to index down through the type
2567 levels of a structure or to a specific index in an array. When indexing into a
2568 structure, only <tt>i32</tt> integer constants are allowed. When indexing
2569 into an array or pointer, only integers of 32 or 64 bits are allowed, and will
2570 be sign extended to 64-bit values.</p>
2572 <p>For example, let's consider a C code fragment and how it gets
2573 compiled to LLVM:</p>
2587 define i32 *foo(struct ST *s) {
2588 return &s[1].Z.B[5][13];
2592 <p>The LLVM code generated by the GCC frontend is:</p>
2595 %RT = type { i8 , [10 x [20 x i32]], i8 }
2596 %ST = type { i32, double, %RT }
2600 define i32* %foo(%ST* %s) {
2602 %reg = getelementptr %ST* %s, i32 1, i32 2, i32 1, i32 5, i32 13
2609 <p>The index types specified for the '<tt>getelementptr</tt>' instruction depend
2610 on the pointer type that is being indexed into. <a href="#t_pointer">Pointer</a>
2611 and <a href="#t_array">array</a> types can use a 32-bit or 64-bit
2612 <a href="#t_integer">integer</a> type but the value will always be sign extended
2613 to 64-bits. <a href="#t_struct">Structure</a> types, require <tt>i32</tt>
2614 <b>constants</b>.</p>
2616 <p>In the example above, the first index is indexing into the '<tt>%ST*</tt>'
2617 type, which is a pointer, yielding a '<tt>%ST</tt>' = '<tt>{ i32, double, %RT
2618 }</tt>' type, a structure. The second index indexes into the third element of
2619 the structure, yielding a '<tt>%RT</tt>' = '<tt>{ i8 , [10 x [20 x i32]],
2620 i8 }</tt>' type, another structure. The third index indexes into the second
2621 element of the structure, yielding a '<tt>[10 x [20 x i32]]</tt>' type, an
2622 array. The two dimensions of the array are subscripted into, yielding an
2623 '<tt>i32</tt>' type. The '<tt>getelementptr</tt>' instruction returns a pointer
2624 to this element, thus computing a value of '<tt>i32*</tt>' type.</p>
2626 <p>Note that it is perfectly legal to index partially through a
2627 structure, returning a pointer to an inner element. Because of this,
2628 the LLVM code for the given testcase is equivalent to:</p>
2631 define i32* %foo(%ST* %s) {
2632 %t1 = getelementptr %ST* %s, i32 1 <i>; yields %ST*:%t1</i>
2633 %t2 = getelementptr %ST* %t1, i32 0, i32 2 <i>; yields %RT*:%t2</i>
2634 %t3 = getelementptr %RT* %t2, i32 0, i32 1 <i>; yields [10 x [20 x i32]]*:%t3</i>
2635 %t4 = getelementptr [10 x [20 x i32]]* %t3, i32 0, i32 5 <i>; yields [20 x i32]*:%t4</i>
2636 %t5 = getelementptr [20 x i32]* %t4, i32 0, i32 13 <i>; yields i32*:%t5</i>
2641 <p>Note that it is undefined to access an array out of bounds: array and
2642 pointer indexes must always be within the defined bounds of the array type.
2643 The one exception for this rules is zero length arrays. These arrays are
2644 defined to be accessible as variable length arrays, which requires access
2645 beyond the zero'th element.</p>
2647 <p>The getelementptr instruction is often confusing. For some more insight
2648 into how it works, see <a href="GetElementPtr.html">the getelementptr
2654 <i>; yields [12 x i8]*:aptr</i>
2655 %aptr = getelementptr {i32, [12 x i8]}* %sptr, i64 0, i32 1
2659 <!-- ======================================================================= -->
2660 <div class="doc_subsection"> <a name="convertops">Conversion Operations</a>
2662 <div class="doc_text">
2663 <p>The instructions in this category are the conversion instructions (casting)
2664 which all take a single operand and a type. They perform various bit conversions
2668 <!-- _______________________________________________________________________ -->
2669 <div class="doc_subsubsection">
2670 <a name="i_trunc">'<tt>trunc .. to</tt>' Instruction</a>
2672 <div class="doc_text">
2676 <result> = trunc <ty> <value> to <ty2> <i>; yields ty2</i>
2681 The '<tt>trunc</tt>' instruction truncates its operand to the type <tt>ty2</tt>.
2686 The '<tt>trunc</tt>' instruction takes a <tt>value</tt> to trunc, which must
2687 be an <a href="#t_integer">integer</a> type, and a type that specifies the size
2688 and type of the result, which must be an <a href="#t_integral">integral</a>
2689 type. The bit size of <tt>value</tt> must be larger than the bit size of
2690 <tt>ty2</tt>. Equal sized types are not allowed.</p>
2694 The '<tt>trunc</tt>' instruction truncates the high order bits in <tt>value</tt>
2695 and converts the remaining bits to <tt>ty2</tt>. Since the source size must be
2696 larger than the destination size, <tt>trunc</tt> cannot be a <i>no-op cast</i>.
2697 It will always truncate bits.</p>
2701 %X = trunc i32 257 to i8 <i>; yields i8:1</i>
2702 %Y = trunc i32 123 to bool <i>; yields bool:true</i>
2703 %Y = trunc i32 122 to bool <i>; yields bool:false</i>
2707 <!-- _______________________________________________________________________ -->
2708 <div class="doc_subsubsection">
2709 <a name="i_zext">'<tt>zext .. to</tt>' Instruction</a>
2711 <div class="doc_text">
2715 <result> = zext <ty> <value> to <ty2> <i>; yields ty2</i>
2719 <p>The '<tt>zext</tt>' instruction zero extends its operand to type
2724 <p>The '<tt>zext</tt>' instruction takes a value to cast, which must be of
2725 <a href="#t_integral">integral</a> type, and a type to cast it to, which must
2726 also be of <a href="#t_integral">integral</a> type. The bit size of the
2727 <tt>value</tt> must be smaller than the bit size of the destination type,
2731 <p>The <tt>zext</tt> fills the high order bits of the <tt>value</tt> with zero
2732 bits until it reaches the size of the destination type, <tt>ty2</tt>. When the
2733 the operand and the type are the same size, no bit filling is done and the
2734 cast is considered a <i>no-op cast</i> because no bits change (only the type
2737 <p>When zero extending from bool, the result will alwasy be either 0 or 1.</p>
2741 %X = zext i32 257 to i64 <i>; yields i64:257</i>
2742 %Y = zext bool true to i32 <i>; yields i32:1</i>
2746 <!-- _______________________________________________________________________ -->
2747 <div class="doc_subsubsection">
2748 <a name="i_sext">'<tt>sext .. to</tt>' Instruction</a>
2750 <div class="doc_text">
2754 <result> = sext <ty> <value> to <ty2> <i>; yields ty2</i>
2758 <p>The '<tt>sext</tt>' sign extends <tt>value</tt> to the type <tt>ty2</tt>.</p>
2762 The '<tt>sext</tt>' instruction takes a value to cast, which must be of
2763 <a href="#t_integral">integral</a> type, and a type to cast it to, which must
2764 also be of <a href="#t_integral">integral</a> type. The bit size of the
2765 <tt>value</tt> must be smaller than the bit size of the destination type,
2770 The '<tt>sext</tt>' instruction performs a sign extension by copying the sign
2771 bit (highest order bit) of the <tt>value</tt> until it reaches the bit size of
2772 the type <tt>ty2</tt>. When the the operand and the type are the same size,
2773 no bit filling is done and the cast is considered a <i>no-op cast</i> because
2774 no bits change (only the type changes).</p>
2776 <p>When sign extending from bool, the extension always results in -1 or 0.</p>
2780 %X = sext i8 -1 to i16 <i>; yields i16 :65535</i>
2781 %Y = sext bool true to i32 <i>; yields i32:-1</i>
2785 <!-- _______________________________________________________________________ -->
2786 <div class="doc_subsubsection">
2787 <a name="i_fptrunc">'<tt>fptrunc .. to</tt>' Instruction</a>
2790 <div class="doc_text">
2795 <result> = fptrunc <ty> <value> to <ty2> <i>; yields ty2</i>
2799 <p>The '<tt>fptrunc</tt>' instruction truncates <tt>value</tt> to type
2804 <p>The '<tt>fptrunc</tt>' instruction takes a <a href="#t_floating">floating
2805 point</a> value to cast and a <a href="#t_floating">floating point</a> type to
2806 cast it to. The size of <tt>value</tt> must be larger than the size of
2807 <tt>ty2</tt>. This implies that <tt>fptrunc</tt> cannot be used to make a
2808 <i>no-op cast</i>.</p>
2811 <p> The '<tt>fptrunc</tt>' instruction truncates a <tt>value</tt> from a larger
2812 <a href="#t_floating">floating point</a> type to a smaller
2813 <a href="#t_floating">floating point</a> type. If the value cannot fit within
2814 the destination type, <tt>ty2</tt>, then the results are undefined.</p>
2818 %X = fptrunc double 123.0 to float <i>; yields float:123.0</i>
2819 %Y = fptrunc double 1.0E+300 to float <i>; yields undefined</i>
2823 <!-- _______________________________________________________________________ -->
2824 <div class="doc_subsubsection">
2825 <a name="i_fpext">'<tt>fpext .. to</tt>' Instruction</a>
2827 <div class="doc_text">
2831 <result> = fpext <ty> <value> to <ty2> <i>; yields ty2</i>
2835 <p>The '<tt>fpext</tt>' extends a floating point <tt>value</tt> to a larger
2836 floating point value.</p>
2839 <p>The '<tt>fpext</tt>' instruction takes a
2840 <a href="#t_floating">floating point</a> <tt>value</tt> to cast,
2841 and a <a href="#t_floating">floating point</a> type to cast it to. The source
2842 type must be smaller than the destination type.</p>
2845 <p>The '<tt>fpext</tt>' instruction extends the <tt>value</tt> from a smaller
2846 <a href="t_floating">floating point</a> type to a larger
2847 <a href="t_floating">floating point</a> type. The <tt>fpext</tt> cannot be
2848 used to make a <i>no-op cast</i> because it always changes bits. Use
2849 <tt>bitcast</tt> to make a <i>no-op cast</i> for a floating point cast.</p>
2853 %X = fpext float 3.1415 to double <i>; yields double:3.1415</i>
2854 %Y = fpext float 1.0 to float <i>; yields float:1.0 (no-op)</i>
2858 <!-- _______________________________________________________________________ -->
2859 <div class="doc_subsubsection">
2860 <a name="i_fp2uint">'<tt>fptoui .. to</tt>' Instruction</a>
2862 <div class="doc_text">
2866 <result> = fp2uint <ty> <value> to <ty2> <i>; yields ty2</i>
2870 <p>The '<tt>fp2uint</tt>' converts a floating point <tt>value</tt> to its
2871 unsigned integer equivalent of type <tt>ty2</tt>.
2875 <p>The '<tt>fp2uint</tt>' instruction takes a value to cast, which must be a
2876 <a href="#t_floating">floating point</a> value, and a type to cast it to, which
2877 must be an <a href="#t_integral">integral</a> type.</p>
2880 <p> The '<tt>fp2uint</tt>' instruction converts its
2881 <a href="#t_floating">floating point</a> operand into the nearest (rounding
2882 towards zero) unsigned integer value. If the value cannot fit in <tt>ty2</tt>,
2883 the results are undefined.</p>
2885 <p>When converting to bool, the conversion is done as a comparison against
2886 zero. If the <tt>value</tt> was zero, the bool result will be <tt>false</tt>.
2887 If the <tt>value</tt> was non-zero, the bool result will be <tt>true</tt>.</p>
2891 %X = fp2uint double 123.0 to i32 <i>; yields i32:123</i>
2892 %Y = fp2uint float 1.0E+300 to bool <i>; yields bool:true</i>
2893 %X = fp2uint float 1.04E+17 to i8 <i>; yields undefined:1</i>
2897 <!-- _______________________________________________________________________ -->
2898 <div class="doc_subsubsection">
2899 <a name="i_fptosi">'<tt>fptosi .. to</tt>' Instruction</a>
2901 <div class="doc_text">
2905 <result> = fptosi <ty> <value> to <ty2> <i>; yields ty2</i>
2909 <p>The '<tt>fptosi</tt>' instruction converts
2910 <a href="#t_floating">floating point</a> <tt>value</tt> to type <tt>ty2</tt>.
2915 <p> The '<tt>fptosi</tt>' instruction takes a value to cast, which must be a
2916 <a href="#t_floating">floating point</a> value, and a type to cast it to, which
2917 must also be an <a href="#t_integral">integral</a> type.</p>
2920 <p>The '<tt>fptosi</tt>' instruction converts its
2921 <a href="#t_floating">floating point</a> operand into the nearest (rounding
2922 towards zero) signed integer value. If the value cannot fit in <tt>ty2</tt>,
2923 the results are undefined.</p>
2925 <p>When converting to bool, the conversion is done as a comparison against
2926 zero. If the <tt>value</tt> was zero, the bool result will be <tt>false</tt>.
2927 If the <tt>value</tt> was non-zero, the bool result will be <tt>true</tt>.</p>
2931 %X = fptosi double -123.0 to i32 <i>; yields i32:-123</i>
2932 %Y = fptosi float 1.0E-247 to bool <i>; yields bool:true</i>
2933 %X = fptosi float 1.04E+17 to i8 <i>; yields undefined:1</i>
2937 <!-- _______________________________________________________________________ -->
2938 <div class="doc_subsubsection">
2939 <a name="i_uitofp">'<tt>uitofp .. to</tt>' Instruction</a>
2941 <div class="doc_text">
2945 <result> = uitofp <ty> <value> to <ty2> <i>; yields ty2</i>
2949 <p>The '<tt>uitofp</tt>' instruction regards <tt>value</tt> as an unsigned
2950 integer and converts that value to the <tt>ty2</tt> type.</p>
2954 <p>The '<tt>uitofp</tt>' instruction takes a value to cast, which must be an
2955 <a href="#t_integral">integral</a> value, and a type to cast it to, which must
2956 be a <a href="#t_floating">floating point</a> type.</p>
2959 <p>The '<tt>uitofp</tt>' instruction interprets its operand as an unsigned
2960 integer quantity and converts it to the corresponding floating point value. If
2961 the value cannot fit in the floating point value, the results are undefined.</p>
2966 %X = uitofp i32 257 to float <i>; yields float:257.0</i>
2967 %Y = uitofp i8 -1 to double <i>; yields double:255.0</i>
2971 <!-- _______________________________________________________________________ -->
2972 <div class="doc_subsubsection">
2973 <a name="i_sitofp">'<tt>sitofp .. to</tt>' Instruction</a>
2975 <div class="doc_text">
2979 <result> = sitofp <ty> <value> to <ty2> <i>; yields ty2</i>
2983 <p>The '<tt>sitofp</tt>' instruction regards <tt>value</tt> as a signed
2984 integer and converts that value to the <tt>ty2</tt> type.</p>
2987 <p>The '<tt>sitofp</tt>' instruction takes a value to cast, which must be an
2988 <a href="#t_integral">integral</a> value, and a type to cast it to, which must be
2989 a <a href="#t_floating">floating point</a> type.</p>
2992 <p>The '<tt>sitofp</tt>' instruction interprets its operand as a signed
2993 integer quantity and converts it to the corresponding floating point value. If
2994 the value cannot fit in the floating point value, the results are undefined.</p>
2998 %X = sitofp i32 257 to float <i>; yields float:257.0</i>
2999 %Y = sitofp i8 -1 to double <i>; yields double:-1.0</i>
3003 <!-- _______________________________________________________________________ -->
3004 <div class="doc_subsubsection">
3005 <a name="i_ptrtoint">'<tt>ptrtoint .. to</tt>' Instruction</a>
3007 <div class="doc_text">
3011 <result> = ptrtoint <ty> <value> to <ty2> <i>; yields ty2</i>
3015 <p>The '<tt>ptrtoint</tt>' instruction converts the pointer <tt>value</tt> to
3016 the integer type <tt>ty2</tt>.</p>
3019 <p>The '<tt>ptrtoint</tt>' instruction takes a <tt>value</tt> to cast, which
3020 must be a <a href="t_pointer">pointer</a> value, and a type to cast it to
3021 <tt>ty2</tt>, which must be an <a href="#t_integer">integer</a> type.
3024 <p>The '<tt>ptrtoint</tt>' instruction converts <tt>value</tt> to integer type
3025 <tt>ty2</tt> by interpreting the pointer value as an integer and either
3026 truncating or zero extending that value to the size of the integer type. If
3027 <tt>value</tt> is smaller than <tt>ty2</tt> then a zero extension is done. If
3028 <tt>value</tt> is larger than <tt>ty2</tt> then a truncation is done. If they
3029 are the same size, then nothing is done (<i>no-op cast</i>).</p>
3033 %X = ptrtoint i32* %X to i8 <i>; yields truncation on 32-bit</i>
3034 %Y = ptrtoint i32* %x to i64 <i>; yields zero extend on 32-bit</i>
3038 <!-- _______________________________________________________________________ -->
3039 <div class="doc_subsubsection">
3040 <a name="i_inttoptr">'<tt>inttoptr .. to</tt>' Instruction</a>
3042 <div class="doc_text">
3046 <result> = inttoptr <ty> <value> to <ty2> <i>; yields ty2</i>
3050 <p>The '<tt>inttoptr</tt>' instruction converts an integer <tt>value</tt> to
3051 a pointer type, <tt>ty2</tt>.</p>
3054 <p>The '<tt>inttoptr</tt>' instruction takes an <a href="i_integer">integer</a>
3055 value to cast, and a type to cast it to, which must be a
3056 <a href="#t_pointer">pointer</a> type. </tt>
3059 <p>The '<tt>inttoptr</tt>' instruction converts <tt>value</tt> to type
3060 <tt>ty2</tt> by applying either a zero extension or a truncation depending on
3061 the size of the integer <tt>value</tt>. If <tt>value</tt> is larger than the
3062 size of a pointer then a truncation is done. If <tt>value</tt> is smaller than
3063 the size of a pointer then a zero extension is done. If they are the same size,
3064 nothing is done (<i>no-op cast</i>).</p>
3068 %X = inttoptr i32 255 to i32* <i>; yields zero extend on 64-bit</i>
3069 %X = inttoptr i32 255 to i32* <i>; yields no-op on 32-bit </i>
3070 %Y = inttoptr i16 0 to i32* <i>; yields zero extend on 32-bit</i>
3074 <!-- _______________________________________________________________________ -->
3075 <div class="doc_subsubsection">
3076 <a name="i_bitcast">'<tt>bitcast .. to</tt>' Instruction</a>
3078 <div class="doc_text">
3082 <result> = bitcast <ty> <value> to <ty2> <i>; yields ty2</i>
3086 <p>The '<tt>bitcast</tt>' instruction converts <tt>value</tt> to type
3087 <tt>ty2</tt> without changing any bits.</p>
3090 <p>The '<tt>bitcast</tt>' instruction takes a value to cast, which must be
3091 a first class value, and a type to cast it to, which must also be a <a
3092 href="#t_firstclass">first class</a> type. The bit sizes of <tt>value</tt>
3093 and the destination type, <tt>ty2</tt>, must be identical. If the source
3094 type is a pointer, the destination type must also be a pointer.</p>
3097 <p>The '<tt>bitcast</tt>' instruction converts <tt>value</tt> to type
3098 <tt>ty2</tt>. It is always a <i>no-op cast</i> because no bits change with
3099 this conversion. The conversion is done as if the <tt>value</tt> had been
3100 stored to memory and read back as type <tt>ty2</tt>. Pointer types may only be
3101 converted to other pointer types with this instruction. To convert pointers to
3102 other types, use the <a href="#i_inttoptr">inttoptr</a> or
3103 <a href="#i_ptrtoint">ptrtoint</a> instructions first.</p>
3107 %X = bitcast i8 255 to i8 <i>; yields i8 :-1</i>
3108 %Y = bitcast i32* %x to sint* <i>; yields sint*:%x</i>
3109 %Z = bitcast <2xint> %V to i64; <i>; yields i64: %V</i>
3113 <!-- ======================================================================= -->
3114 <div class="doc_subsection"> <a name="otherops">Other Operations</a> </div>
3115 <div class="doc_text">
3116 <p>The instructions in this category are the "miscellaneous"
3117 instructions, which defy better classification.</p>
3120 <!-- _______________________________________________________________________ -->
3121 <div class="doc_subsubsection"><a name="i_icmp">'<tt>icmp</tt>' Instruction</a>
3123 <div class="doc_text">
3125 <pre> <result> = icmp <cond> <ty> <var1>, <var2> <i>; yields {bool}:result</i>
3128 <p>The '<tt>icmp</tt>' instruction returns a boolean value based on comparison
3129 of its two integer operands.</p>
3131 <p>The '<tt>icmp</tt>' instruction takes three operands. The first operand is
3132 the condition code which indicates the kind of comparison to perform. It is not
3133 a value, just a keyword. The possibilities for the condition code are:
3135 <li><tt>eq</tt>: equal</li>
3136 <li><tt>ne</tt>: not equal </li>
3137 <li><tt>ugt</tt>: unsigned greater than</li>
3138 <li><tt>uge</tt>: unsigned greater or equal</li>
3139 <li><tt>ult</tt>: unsigned less than</li>
3140 <li><tt>ule</tt>: unsigned less or equal</li>
3141 <li><tt>sgt</tt>: signed greater than</li>
3142 <li><tt>sge</tt>: signed greater or equal</li>
3143 <li><tt>slt</tt>: signed less than</li>
3144 <li><tt>sle</tt>: signed less or equal</li>
3146 <p>The remaining two arguments must be <a href="#t_integral">integral</a> or
3147 <a href="#t_pointer">pointer</a> typed. They must also be identical types.</p>
3149 <p>The '<tt>icmp</tt>' compares <tt>var1</tt> and <tt>var2</tt> according to
3150 the condition code given as <tt>cond</tt>. The comparison performed always
3151 yields a <a href="#t_bool">bool</a> result, as follows:
3153 <li><tt>eq</tt>: yields <tt>true</tt> if the operands are equal,
3154 <tt>false</tt> otherwise. No sign interpretation is necessary or performed.
3156 <li><tt>ne</tt>: yields <tt>true</tt> if the operands are unequal,
3157 <tt>false</tt> otherwise. No sign interpretation is necessary or performed.
3158 <li><tt>ugt</tt>: interprets the operands as unsigned values and yields
3159 <tt>true</tt> if <tt>var1</tt> is greater than <tt>var2</tt>.</li>
3160 <li><tt>uge</tt>: interprets the operands as unsigned values and yields
3161 <tt>true</tt> if <tt>var1</tt> is greater than or equal to <tt>var2</tt>.</li>
3162 <li><tt>ult</tt>: interprets the operands as unsigned values and yields
3163 <tt>true</tt> if <tt>var1</tt> is less than <tt>var2</tt>.</li>
3164 <li><tt>ule</tt>: interprets the operands as unsigned values and yields
3165 <tt>true</tt> if <tt>var1</tt> is less than or equal to <tt>var2</tt>.</li>
3166 <li><tt>sgt</tt>: interprets the operands as signed values and yields
3167 <tt>true</tt> if <tt>var1</tt> is greater than <tt>var2</tt>.</li>
3168 <li><tt>sge</tt>: interprets the operands as signed values and yields
3169 <tt>true</tt> if <tt>var1</tt> is greater than or equal to <tt>var2</tt>.</li>
3170 <li><tt>slt</tt>: interprets the operands as signed values and yields
3171 <tt>true</tt> if <tt>var1</tt> is less than <tt>var2</tt>.</li>
3172 <li><tt>sle</tt>: interprets the operands as signed values and yields
3173 <tt>true</tt> if <tt>var1</tt> is less than or equal to <tt>var2</tt>.</li>
3176 <p>If the operands are <a href="#t_pointer">pointer</a> typed, the pointer
3177 values are treated as integers and then compared.</p>
3178 <p>If the operands are <a href="#t_packed">packed</a> typed, the elements of
3179 the vector are compared in turn and the predicate must hold for all
3183 <pre> <result> = icmp eq i32 4, 5 <i>; yields: result=false</i>
3184 <result> = icmp ne float* %X, %X <i>; yields: result=false</i>
3185 <result> = icmp ult i16 4, 5 <i>; yields: result=true</i>
3186 <result> = icmp sgt i16 4, 5 <i>; yields: result=false</i>
3187 <result> = icmp ule i16 -4, 5 <i>; yields: result=false</i>
3188 <result> = icmp sge i16 4, 5 <i>; yields: result=false</i>
3192 <!-- _______________________________________________________________________ -->
3193 <div class="doc_subsubsection"><a name="i_fcmp">'<tt>fcmp</tt>' Instruction</a>
3195 <div class="doc_text">
3197 <pre> <result> = fcmp <cond> <ty> <var1>, <var2> <i>; yields {bool}:result</i>
3200 <p>The '<tt>fcmp</tt>' instruction returns a boolean value based on comparison
3201 of its floating point operands.</p>
3203 <p>The '<tt>fcmp</tt>' instruction takes three operands. The first operand is
3204 the condition code which indicates the kind of comparison to perform. It is not
3205 a value, just a keyword. The possibilities for the condition code are:
3207 <li><tt>false</tt>: no comparison, always returns false</li>
3208 <li><tt>oeq</tt>: ordered and equal</li>
3209 <li><tt>ogt</tt>: ordered and greater than </li>
3210 <li><tt>oge</tt>: ordered and greater than or equal</li>
3211 <li><tt>olt</tt>: ordered and less than </li>
3212 <li><tt>ole</tt>: ordered and less than or equal</li>
3213 <li><tt>one</tt>: ordered and not equal</li>
3214 <li><tt>ord</tt>: ordered (no nans)</li>
3215 <li><tt>ueq</tt>: unordered or equal</li>
3216 <li><tt>ugt</tt>: unordered or greater than </li>
3217 <li><tt>uge</tt>: unordered or greater than or equal</li>
3218 <li><tt>ult</tt>: unordered or less than </li>
3219 <li><tt>ule</tt>: unordered or less than or equal</li>
3220 <li><tt>une</tt>: unordered or not equal</li>
3221 <li><tt>uno</tt>: unordered (either nans)</li>
3222 <li><tt>true</tt>: no comparison, always returns true</li>
3224 <p>In the preceding, <i>ordered</i> means that neither operand is a QNAN while
3225 <i>unordered</i> means that either operand may be a QNAN.</p>
3226 <p>The <tt>val1</tt> and <tt>val2</tt> arguments must be
3227 <a href="#t_floating">floating point</a> typed. They must have identical
3229 <p>In the foregoing, <i>ordered</i> means that neither operand is a QNAN and
3230 <i>unordered</i> means that either operand is a QNAN.</p>
3232 <p>The '<tt>fcmp</tt>' compares <tt>var1</tt> and <tt>var2</tt> according to
3233 the condition code given as <tt>cond</tt>. The comparison performed always
3234 yields a <a href="#t_bool">bool</a> result, as follows:
3236 <li><tt>false</tt>: always yields <tt>false</tt>, regardless of operands.</li>
3237 <li><tt>oeq</tt>: yields <tt>true</tt> if both operands are not a QNAN and
3238 <tt>var1</tt> is equal to <tt>var2</tt>.</li>
3239 <li><tt>ogt</tt>: yields <tt>true</tt> if both operands are not a QNAN and
3240 <tt>var1</tt> is greather than <tt>var2</tt>.</li>
3241 <li><tt>oge</tt>: yields <tt>true</tt> if both operands are not a QNAN and
3242 <tt>var1</tt> is greater than or equal to <tt>var2</tt>.</li>
3243 <li><tt>olt</tt>: yields <tt>true</tt> if both operands are not a QNAN and
3244 <tt>var1</tt> is less than <tt>var2</tt>.</li>
3245 <li><tt>ole</tt>: yields <tt>true</tt> if both operands are not a QNAN and
3246 <tt>var1</tt> is less than or equal to <tt>var2</tt>.</li>
3247 <li><tt>one</tt>: yields <tt>true</tt> if both operands are not a QNAN and
3248 <tt>var1</tt> is not equal to <tt>var2</tt>.</li>
3249 <li><tt>ord</tt>: yields <tt>true</tt> if both operands are not a QNAN.</li>
3250 <li><tt>ueq</tt>: yields <tt>true</tt> if either operand is a QNAN or
3251 <tt>var1</tt> is equal to <tt>var2</tt>.</li>
3252 <li><tt>ugt</tt>: yields <tt>true</tt> if either operand is a QNAN or
3253 <tt>var1</tt> is greater than <tt>var2</tt>.</li>
3254 <li><tt>uge</tt>: yields <tt>true</tt> if either operand is a QNAN or
3255 <tt>var1</tt> is greater than or equal to <tt>var2</tt>.</li>
3256 <li><tt>ult</tt>: yields <tt>true</tt> if either operand is a QNAN or
3257 <tt>var1</tt> is less than <tt>var2</tt>.</li>
3258 <li><tt>ule</tt>: yields <tt>true</tt> if either operand is a QNAN or
3259 <tt>var1</tt> is less than or equal to <tt>var2</tt>.</li>
3260 <li><tt>une</tt>: yields <tt>true</tt> if either operand is a QNAN or
3261 <tt>var1</tt> is not equal to <tt>var2</tt>.</li>
3262 <li><tt>uno</tt>: yields <tt>true</tt> if either operand is a QNAN.</li>
3263 <li><tt>true</tt>: always yields <tt>true</tt>, regardless of operands.</li>
3265 <p>If the operands are <a href="#t_packed">packed</a> typed, the elements of
3266 the vector are compared in turn and the predicate must hold for all elements.
3270 <pre> <result> = fcmp oeq float 4.0, 5.0 <i>; yields: result=false</i>
3271 <result> = icmp one float 4.0, 5.0 <i>; yields: result=true</i>
3272 <result> = icmp olt float 4.0, 5.0 <i>; yields: result=true</i>
3273 <result> = icmp ueq double 1.0, 2.0 <i>; yields: result=false</i>
3277 <!-- _______________________________________________________________________ -->
3278 <div class="doc_subsubsection"> <a name="i_phi">'<tt>phi</tt>'
3279 Instruction</a> </div>
3280 <div class="doc_text">
3282 <pre> <result> = phi <ty> [ <val0>, <label0>], ...<br></pre>
3284 <p>The '<tt>phi</tt>' instruction is used to implement the φ node in
3285 the SSA graph representing the function.</p>
3287 <p>The type of the incoming values are specified with the first type
3288 field. After this, the '<tt>phi</tt>' instruction takes a list of pairs
3289 as arguments, with one pair for each predecessor basic block of the
3290 current block. Only values of <a href="#t_firstclass">first class</a>
3291 type may be used as the value arguments to the PHI node. Only labels
3292 may be used as the label arguments.</p>
3293 <p>There must be no non-phi instructions between the start of a basic
3294 block and the PHI instructions: i.e. PHI instructions must be first in
3297 <p>At runtime, the '<tt>phi</tt>' instruction logically takes on the
3298 value specified by the parameter, depending on which basic block we
3299 came from in the last <a href="#terminators">terminator</a> instruction.</p>
3301 <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>
3304 <!-- _______________________________________________________________________ -->
3305 <div class="doc_subsubsection">
3306 <a name="i_select">'<tt>select</tt>' Instruction</a>
3309 <div class="doc_text">
3314 <result> = select bool <cond>, <ty> <val1>, <ty> <val2> <i>; yields ty</i>
3320 The '<tt>select</tt>' instruction is used to choose one value based on a
3321 condition, without branching.
3328 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.
3334 If the boolean condition evaluates to true, the instruction returns the first
3335 value argument; otherwise, it returns the second value argument.
3341 %X = select bool true, i8 17, i8 42 <i>; yields i8:17</i>
3346 <!-- _______________________________________________________________________ -->
3347 <div class="doc_subsubsection">
3348 <a name="i_call">'<tt>call</tt>' Instruction</a>
3351 <div class="doc_text">
3355 <result> = [tail] call [<a href="#callingconv">cconv</a>] <ty>* <fnptrval>(<param list>)
3360 <p>The '<tt>call</tt>' instruction represents a simple function call.</p>
3364 <p>This instruction requires several arguments:</p>
3368 <p>The optional "tail" marker indicates whether the callee function accesses
3369 any allocas or varargs in the caller. If the "tail" marker is present, the
3370 function call is eligible for tail call optimization. Note that calls may
3371 be marked "tail" even if they do not occur before a <a
3372 href="#i_ret"><tt>ret</tt></a> instruction.
3375 <p>The optional "cconv" marker indicates which <a href="callingconv">calling
3376 convention</a> the call should use. If none is specified, the call defaults
3377 to using C calling conventions.
3380 <p>'<tt>ty</tt>': shall be the signature of the pointer to function value
3381 being invoked. The argument types must match the types implied by this
3382 signature. This type can be omitted if the function is not varargs and
3383 if the function type does not return a pointer to a function.</p>
3386 <p>'<tt>fnptrval</tt>': An LLVM value containing a pointer to a function to
3387 be invoked. In most cases, this is a direct function invocation, but
3388 indirect <tt>call</tt>s are just as possible, calling an arbitrary pointer
3389 to function value.</p>
3392 <p>'<tt>function args</tt>': argument list whose types match the
3393 function signature argument types. All arguments must be of
3394 <a href="#t_firstclass">first class</a> type. If the function signature
3395 indicates the function accepts a variable number of arguments, the extra
3396 arguments can be specified.</p>
3402 <p>The '<tt>call</tt>' instruction is used to cause control flow to
3403 transfer to a specified function, with its incoming arguments bound to
3404 the specified values. Upon a '<tt><a href="#i_ret">ret</a></tt>'
3405 instruction in the called function, control flow continues with the
3406 instruction after the function call, and the return value of the
3407 function is bound to the result argument. This is a simpler case of
3408 the <a href="#i_invoke">invoke</a> instruction.</p>
3413 %retval = call i32 %test(i32 %argc)
3414 call i32(i8 *, ...) *%printf(i8 * %msg, i32 12, i8 42);
3415 %X = tail call i32 %foo()
3416 %Y = tail call <a href="#callingconv">fastcc</a> i32 %foo()
3421 <!-- _______________________________________________________________________ -->
3422 <div class="doc_subsubsection">
3423 <a name="i_va_arg">'<tt>va_arg</tt>' Instruction</a>
3426 <div class="doc_text">
3431 <resultval> = va_arg <va_list*> <arglist>, <argty>
3436 <p>The '<tt>va_arg</tt>' instruction is used to access arguments passed through
3437 the "variable argument" area of a function call. It is used to implement the
3438 <tt>va_arg</tt> macro in C.</p>
3442 <p>This instruction takes a <tt>va_list*</tt> value and the type of
3443 the argument. It returns a value of the specified argument type and
3444 increments the <tt>va_list</tt> to point to the next argument. Again, the
3445 actual type of <tt>va_list</tt> is target specific.</p>
3449 <p>The '<tt>va_arg</tt>' instruction loads an argument of the specified
3450 type from the specified <tt>va_list</tt> and causes the
3451 <tt>va_list</tt> to point to the next argument. For more information,
3452 see the variable argument handling <a href="#int_varargs">Intrinsic
3455 <p>It is legal for this instruction to be called in a function which does not
3456 take a variable number of arguments, for example, the <tt>vfprintf</tt>
3459 <p><tt>va_arg</tt> is an LLVM instruction instead of an <a
3460 href="#intrinsics">intrinsic function</a> because it takes a type as an
3465 <p>See the <a href="#int_varargs">variable argument processing</a> section.</p>
3469 <!-- *********************************************************************** -->
3470 <div class="doc_section"> <a name="intrinsics">Intrinsic Functions</a> </div>
3471 <!-- *********************************************************************** -->
3473 <div class="doc_text">
3475 <p>LLVM supports the notion of an "intrinsic function". These functions have
3476 well known names and semantics and are required to follow certain
3477 restrictions. Overall, these instructions represent an extension mechanism for
3478 the LLVM language that does not require changing all of the transformations in
3479 LLVM to add to the language (or the bytecode reader/writer, the parser,
3482 <p>Intrinsic function names must all start with an "<tt>llvm.</tt>" prefix. This
3483 prefix is reserved in LLVM for intrinsic names; thus, functions may not be named
3484 this. Intrinsic functions must always be external functions: you cannot define
3485 the body of intrinsic functions. Intrinsic functions may only be used in call
3486 or invoke instructions: it is illegal to take the address of an intrinsic
3487 function. Additionally, because intrinsic functions are part of the LLVM
3488 language, it is required that they all be documented here if any are added.</p>
3491 <p>To learn how to add an intrinsic function, please see the <a
3492 href="ExtendingLLVM.html">Extending LLVM Guide</a>.
3497 <!-- ======================================================================= -->
3498 <div class="doc_subsection">
3499 <a name="int_varargs">Variable Argument Handling Intrinsics</a>
3502 <div class="doc_text">
3504 <p>Variable argument support is defined in LLVM with the <a
3505 href="#i_va_arg"><tt>va_arg</tt></a> instruction and these three
3506 intrinsic functions. These functions are related to the similarly
3507 named macros defined in the <tt><stdarg.h></tt> header file.</p>
3509 <p>All of these functions operate on arguments that use a
3510 target-specific value type "<tt>va_list</tt>". The LLVM assembly
3511 language reference manual does not define what this type is, so all
3512 transformations should be prepared to handle intrinsics with any type
3515 <p>This example shows how the <a href="#i_va_arg"><tt>va_arg</tt></a>
3516 instruction and the variable argument handling intrinsic functions are
3520 define i32 %test(i32 %X, ...) {
3521 ; Initialize variable argument processing
3523 %ap2 = bitcast i8** %ap to i8*
3524 call void %<a href="#i_va_start">llvm.va_start</a>(i8* %ap2)
3526 ; Read a single integer argument
3527 %tmp = va_arg i8 ** %ap, i32
3529 ; Demonstrate usage of llvm.va_copy and llvm.va_end
3531 %aq2 = bitcast i8** %aq to i8*
3532 call void %<a href="#i_va_copy">llvm.va_copy</a>(i8 *%aq2, i8* %ap2)
3533 call void %<a href="#i_va_end">llvm.va_end</a>(i8* %aq2)
3535 ; Stop processing of arguments.
3536 call void %<a href="#i_va_end">llvm.va_end</a>(i8* %ap2)
3542 <!-- _______________________________________________________________________ -->
3543 <div class="doc_subsubsection">
3544 <a name="i_va_start">'<tt>llvm.va_start</tt>' Intrinsic</a>
3548 <div class="doc_text">
3550 <pre> declare void %llvm.va_start(i8* <arglist>)<br></pre>
3552 <P>The '<tt>llvm.va_start</tt>' intrinsic initializes
3553 <tt>*<arglist></tt> for subsequent use by <tt><a
3554 href="#i_va_arg">va_arg</a></tt>.</p>
3558 <P>The argument is a pointer to a <tt>va_list</tt> element to initialize.</p>
3562 <P>The '<tt>llvm.va_start</tt>' intrinsic works just like the <tt>va_start</tt>
3563 macro available in C. In a target-dependent way, it initializes the
3564 <tt>va_list</tt> element the argument points to, so that the next call to
3565 <tt>va_arg</tt> will produce the first variable argument passed to the function.
3566 Unlike the C <tt>va_start</tt> macro, this intrinsic does not need to know the
3567 last argument of the function, the compiler can figure that out.</p>
3571 <!-- _______________________________________________________________________ -->
3572 <div class="doc_subsubsection">
3573 <a name="i_va_end">'<tt>llvm.va_end</tt>' Intrinsic</a>
3576 <div class="doc_text">
3578 <pre> declare void %llvm.va_end(i8* <arglist>)<br></pre>
3581 <p>The '<tt>llvm.va_end</tt>' intrinsic destroys <tt><arglist></tt>
3582 which has been initialized previously with <tt><a href="#i_va_start">llvm.va_start</a></tt>
3583 or <tt><a href="#i_va_copy">llvm.va_copy</a></tt>.</p>
3587 <p>The argument is a <tt>va_list</tt> to destroy.</p>
3591 <p>The '<tt>llvm.va_end</tt>' intrinsic works just like the <tt>va_end</tt>
3592 macro available in C. In a target-dependent way, it destroys the <tt>va_list</tt>.
3593 Calls to <a href="#i_va_start"><tt>llvm.va_start</tt></a> and <a
3594 href="#i_va_copy"><tt>llvm.va_copy</tt></a> must be matched exactly
3595 with calls to <tt>llvm.va_end</tt>.</p>
3599 <!-- _______________________________________________________________________ -->
3600 <div class="doc_subsubsection">
3601 <a name="i_va_copy">'<tt>llvm.va_copy</tt>' Intrinsic</a>
3604 <div class="doc_text">
3609 declare void %llvm.va_copy(i8* <destarglist>, i8* <srcarglist>)
3614 <p>The '<tt>llvm.va_copy</tt>' intrinsic copies the current argument position from
3615 the source argument list to the destination argument list.</p>
3619 <p>The first argument is a pointer to a <tt>va_list</tt> element to initialize.
3620 The second argument is a pointer to a <tt>va_list</tt> element to copy from.</p>
3625 <p>The '<tt>llvm.va_copy</tt>' intrinsic works just like the <tt>va_copy</tt> macro
3626 available in C. In a target-dependent way, it copies the source
3627 <tt>va_list</tt> element into the destination list. This intrinsic is necessary
3628 because the <tt><a href="i_va_begin">llvm.va_begin</a></tt> intrinsic may be
3629 arbitrarily complex and require memory allocation, for example.</p>
3633 <!-- ======================================================================= -->
3634 <div class="doc_subsection">
3635 <a name="int_gc">Accurate Garbage Collection Intrinsics</a>
3638 <div class="doc_text">
3641 LLVM support for <a href="GarbageCollection.html">Accurate Garbage
3642 Collection</a> requires the implementation and generation of these intrinsics.
3643 These intrinsics allow identification of <a href="#i_gcroot">GC roots on the
3644 stack</a>, as well as garbage collector implementations that require <a
3645 href="#i_gcread">read</a> and <a href="#i_gcwrite">write</a> barriers.
3646 Front-ends for type-safe garbage collected languages should generate these
3647 intrinsics to make use of the LLVM garbage collectors. For more details, see <a
3648 href="GarbageCollection.html">Accurate Garbage Collection with LLVM</a>.
3652 <!-- _______________________________________________________________________ -->
3653 <div class="doc_subsubsection">
3654 <a name="i_gcroot">'<tt>llvm.gcroot</tt>' Intrinsic</a>
3657 <div class="doc_text">
3662 declare void %llvm.gcroot(<ty>** %ptrloc, <ty2>* %metadata)
3667 <p>The '<tt>llvm.gcroot</tt>' intrinsic declares the existence of a GC root to
3668 the code generator, and allows some metadata to be associated with it.</p>
3672 <p>The first argument specifies the address of a stack object that contains the
3673 root pointer. The second pointer (which must be either a constant or a global
3674 value address) contains the meta-data to be associated with the root.</p>
3678 <p>At runtime, a call to this intrinsics stores a null pointer into the "ptrloc"
3679 location. At compile-time, the code generator generates information to allow
3680 the runtime to find the pointer at GC safe points.
3686 <!-- _______________________________________________________________________ -->
3687 <div class="doc_subsubsection">
3688 <a name="i_gcread">'<tt>llvm.gcread</tt>' Intrinsic</a>
3691 <div class="doc_text">
3696 declare i8 * %llvm.gcread(i8 * %ObjPtr, i8 ** %Ptr)
3701 <p>The '<tt>llvm.gcread</tt>' intrinsic identifies reads of references from heap
3702 locations, allowing garbage collector implementations that require read
3707 <p>The second argument is the address to read from, which should be an address
3708 allocated from the garbage collector. The first object is a pointer to the
3709 start of the referenced object, if needed by the language runtime (otherwise
3714 <p>The '<tt>llvm.gcread</tt>' intrinsic has the same semantics as a load
3715 instruction, but may be replaced with substantially more complex code by the
3716 garbage collector runtime, as needed.</p>
3721 <!-- _______________________________________________________________________ -->
3722 <div class="doc_subsubsection">
3723 <a name="i_gcwrite">'<tt>llvm.gcwrite</tt>' Intrinsic</a>
3726 <div class="doc_text">
3731 declare void %llvm.gcwrite(i8 * %P1, i8 * %Obj, i8 ** %P2)
3736 <p>The '<tt>llvm.gcwrite</tt>' intrinsic identifies writes of references to heap
3737 locations, allowing garbage collector implementations that require write
3738 barriers (such as generational or reference counting collectors).</p>
3742 <p>The first argument is the reference to store, the second is the start of the
3743 object to store it to, and the third is the address of the field of Obj to
3744 store to. If the runtime does not require a pointer to the object, Obj may be
3749 <p>The '<tt>llvm.gcwrite</tt>' intrinsic has the same semantics as a store
3750 instruction, but may be replaced with substantially more complex code by the
3751 garbage collector runtime, as needed.</p>
3757 <!-- ======================================================================= -->
3758 <div class="doc_subsection">
3759 <a name="int_codegen">Code Generator Intrinsics</a>
3762 <div class="doc_text">
3764 These intrinsics are provided by LLVM to expose special features that may only
3765 be implemented with code generator support.
3770 <!-- _______________________________________________________________________ -->
3771 <div class="doc_subsubsection">
3772 <a name="i_returnaddress">'<tt>llvm.returnaddress</tt>' Intrinsic</a>
3775 <div class="doc_text">
3779 declare i8 *%llvm.returnaddress(i32 <level>)
3785 The '<tt>llvm.returnaddress</tt>' intrinsic attempts to compute a
3786 target-specific value indicating the return address of the current function
3787 or one of its callers.
3793 The argument to this intrinsic indicates which function to return the address
3794 for. Zero indicates the calling function, one indicates its caller, etc. The
3795 argument is <b>required</b> to be a constant integer value.
3801 The '<tt>llvm.returnaddress</tt>' intrinsic either returns a pointer indicating
3802 the return address of the specified call frame, or zero if it cannot be
3803 identified. The value returned by this intrinsic is likely to be incorrect or 0
3804 for arguments other than zero, so it should only be used for debugging purposes.
3808 Note that calling this intrinsic does not prevent function inlining or other
3809 aggressive transformations, so the value returned may not be that of the obvious
3810 source-language caller.
3815 <!-- _______________________________________________________________________ -->
3816 <div class="doc_subsubsection">
3817 <a name="i_frameaddress">'<tt>llvm.frameaddress</tt>' Intrinsic</a>
3820 <div class="doc_text">
3824 declare i8 *%llvm.frameaddress(i32 <level>)
3830 The '<tt>llvm.frameaddress</tt>' intrinsic attempts to return the
3831 target-specific frame pointer value for the specified stack frame.
3837 The argument to this intrinsic indicates which function to return the frame
3838 pointer for. Zero indicates the calling function, one indicates its caller,
3839 etc. The argument is <b>required</b> to be a constant integer value.
3845 The '<tt>llvm.frameaddress</tt>' intrinsic either returns a pointer indicating
3846 the frame address of the specified call frame, or zero if it cannot be
3847 identified. The value returned by this intrinsic is likely to be incorrect or 0
3848 for arguments other than zero, so it should only be used for debugging purposes.
3852 Note that calling this intrinsic does not prevent function inlining or other
3853 aggressive transformations, so the value returned may not be that of the obvious
3854 source-language caller.
3858 <!-- _______________________________________________________________________ -->
3859 <div class="doc_subsubsection">
3860 <a name="i_stacksave">'<tt>llvm.stacksave</tt>' Intrinsic</a>
3863 <div class="doc_text">
3867 declare i8 *%llvm.stacksave()
3873 The '<tt>llvm.stacksave</tt>' intrinsic is used to remember the current state of
3874 the function stack, for use with <a href="#i_stackrestore">
3875 <tt>llvm.stackrestore</tt></a>. This is useful for implementing language
3876 features like scoped automatic variable sized arrays in C99.
3882 This intrinsic returns a opaque pointer value that can be passed to <a
3883 href="#i_stackrestore"><tt>llvm.stackrestore</tt></a>. When an
3884 <tt>llvm.stackrestore</tt> intrinsic is executed with a value saved from
3885 <tt>llvm.stacksave</tt>, it effectively restores the state of the stack to the
3886 state it was in when the <tt>llvm.stacksave</tt> intrinsic executed. In
3887 practice, this pops any <a href="#i_alloca">alloca</a> blocks from the stack
3888 that were allocated after the <tt>llvm.stacksave</tt> was executed.
3893 <!-- _______________________________________________________________________ -->
3894 <div class="doc_subsubsection">
3895 <a name="i_stackrestore">'<tt>llvm.stackrestore</tt>' Intrinsic</a>
3898 <div class="doc_text">
3902 declare void %llvm.stackrestore(i8 * %ptr)
3908 The '<tt>llvm.stackrestore</tt>' intrinsic is used to restore the state of
3909 the function stack to the state it was in when the corresponding <a
3910 href="#llvm.stacksave"><tt>llvm.stacksave</tt></a> intrinsic executed. This is
3911 useful for implementing language features like scoped automatic variable sized
3918 See the description for <a href="#i_stacksave"><tt>llvm.stacksave</tt></a>.
3924 <!-- _______________________________________________________________________ -->
3925 <div class="doc_subsubsection">
3926 <a name="i_prefetch">'<tt>llvm.prefetch</tt>' Intrinsic</a>
3929 <div class="doc_text">
3933 declare void %llvm.prefetch(i8 * <address>,
3934 i32 <rw>, i32 <locality>)
3941 The '<tt>llvm.prefetch</tt>' intrinsic is a hint to the code generator to insert
3942 a prefetch instruction if supported; otherwise, it is a noop. Prefetches have
3944 effect on the behavior of the program but can change its performance
3951 <tt>address</tt> is the address to be prefetched, <tt>rw</tt> is the specifier
3952 determining if the fetch should be for a read (0) or write (1), and
3953 <tt>locality</tt> is a temporal locality specifier ranging from (0) - no
3954 locality, to (3) - extremely local keep in cache. The <tt>rw</tt> and
3955 <tt>locality</tt> arguments must be constant integers.
3961 This intrinsic does not modify the behavior of the program. In particular,
3962 prefetches cannot trap and do not produce a value. On targets that support this
3963 intrinsic, the prefetch can provide hints to the processor cache for better
3969 <!-- _______________________________________________________________________ -->
3970 <div class="doc_subsubsection">
3971 <a name="i_pcmarker">'<tt>llvm.pcmarker</tt>' Intrinsic</a>
3974 <div class="doc_text">
3978 declare void %llvm.pcmarker( i32 <id> )
3985 The '<tt>llvm.pcmarker</tt>' intrinsic is a method to export a Program Counter
3987 code to simulators and other tools. The method is target specific, but it is
3988 expected that the marker will use exported symbols to transmit the PC of the marker.
3989 The marker makes no guarantees that it will remain with any specific instruction
3990 after optimizations. It is possible that the presence of a marker will inhibit
3991 optimizations. The intended use is to be inserted after optimizations to allow
3992 correlations of simulation runs.
3998 <tt>id</tt> is a numerical id identifying the marker.
4004 This intrinsic does not modify the behavior of the program. Backends that do not
4005 support this intrinisic may ignore it.
4010 <!-- _______________________________________________________________________ -->
4011 <div class="doc_subsubsection">
4012 <a name="i_readcyclecounter">'<tt>llvm.readcyclecounter</tt>' Intrinsic</a>
4015 <div class="doc_text">
4019 declare i64 %llvm.readcyclecounter( )
4026 The '<tt>llvm.readcyclecounter</tt>' intrinsic provides access to the cycle
4027 counter register (or similar low latency, high accuracy clocks) on those targets
4028 that support it. On X86, it should map to RDTSC. On Alpha, it should map to RPCC.
4029 As the backing counters overflow quickly (on the order of 9 seconds on alpha), this
4030 should only be used for small timings.
4036 When directly supported, reading the cycle counter should not modify any memory.
4037 Implementations are allowed to either return a application specific value or a
4038 system wide value. On backends without support, this is lowered to a constant 0.
4043 <!-- ======================================================================= -->
4044 <div class="doc_subsection">
4045 <a name="int_libc">Standard C Library Intrinsics</a>
4048 <div class="doc_text">
4050 LLVM provides intrinsics for a few important standard C library functions.
4051 These intrinsics allow source-language front-ends to pass information about the
4052 alignment of the pointer arguments to the code generator, providing opportunity
4053 for more efficient code generation.
4058 <!-- _______________________________________________________________________ -->
4059 <div class="doc_subsubsection">
4060 <a name="i_memcpy">'<tt>llvm.memcpy</tt>' Intrinsic</a>
4063 <div class="doc_text">
4067 declare void %llvm.memcpy.i32(i8 * <dest>, i8 * <src>,
4068 i32 <len>, i32 <align>)
4069 declare void %llvm.memcpy.i64(i8 * <dest>, i8 * <src>,
4070 i64 <len>, i32 <align>)
4076 The '<tt>llvm.memcpy.*</tt>' intrinsics copy a block of memory from the source
4077 location to the destination location.
4081 Note that, unlike the standard libc function, the <tt>llvm.memcpy.*</tt>
4082 intrinsics do not return a value, and takes an extra alignment argument.
4088 The first argument is a pointer to the destination, the second is a pointer to
4089 the source. The third argument is an integer argument
4090 specifying the number of bytes to copy, and the fourth argument is the alignment
4091 of the source and destination locations.
4095 If the call to this intrinisic has an alignment value that is not 0 or 1, then
4096 the caller guarantees that both the source and destination pointers are aligned
4103 The '<tt>llvm.memcpy.*</tt>' intrinsics copy a block of memory from the source
4104 location to the destination location, which are not allowed to overlap. It
4105 copies "len" bytes of memory over. If the argument is known to be aligned to
4106 some boundary, this can be specified as the fourth argument, otherwise it should
4112 <!-- _______________________________________________________________________ -->
4113 <div class="doc_subsubsection">
4114 <a name="i_memmove">'<tt>llvm.memmove</tt>' Intrinsic</a>
4117 <div class="doc_text">
4121 declare void %llvm.memmove.i32(i8 * <dest>, i8 * <src>,
4122 i32 <len>, i32 <align>)
4123 declare void %llvm.memmove.i64(i8 * <dest>, i8 * <src>,
4124 i64 <len>, i32 <align>)
4130 The '<tt>llvm.memmove.*</tt>' intrinsics move a block of memory from the source
4131 location to the destination location. It is similar to the
4132 '<tt>llvm.memcmp</tt>' intrinsic but allows the two memory locations to overlap.
4136 Note that, unlike the standard libc function, the <tt>llvm.memmove.*</tt>
4137 intrinsics do not return a value, and takes an extra alignment argument.
4143 The first argument is a pointer to the destination, the second is a pointer to
4144 the source. The third argument is an integer argument
4145 specifying the number of bytes to copy, and the fourth argument is the alignment
4146 of the source and destination locations.
4150 If the call to this intrinisic has an alignment value that is not 0 or 1, then
4151 the caller guarantees that the source and destination pointers are aligned to
4158 The '<tt>llvm.memmove.*</tt>' intrinsics copy a block of memory from the source
4159 location to the destination location, which may overlap. It
4160 copies "len" bytes of memory over. If the argument is known to be aligned to
4161 some boundary, this can be specified as the fourth argument, otherwise it should
4167 <!-- _______________________________________________________________________ -->
4168 <div class="doc_subsubsection">
4169 <a name="i_memset">'<tt>llvm.memset.*</tt>' Intrinsics</a>
4172 <div class="doc_text">
4176 declare void %llvm.memset.i32(i8 * <dest>, i8 <val>,
4177 i32 <len>, i32 <align>)
4178 declare void %llvm.memset.i64(i8 * <dest>, i8 <val>,
4179 i64 <len>, i32 <align>)
4185 The '<tt>llvm.memset.*</tt>' intrinsics fill a block of memory with a particular
4190 Note that, unlike the standard libc function, the <tt>llvm.memset</tt> intrinsic
4191 does not return a value, and takes an extra alignment argument.
4197 The first argument is a pointer to the destination to fill, the second is the
4198 byte value to fill it with, the third argument is an integer
4199 argument specifying the number of bytes to fill, and the fourth argument is the
4200 known alignment of destination location.
4204 If the call to this intrinisic has an alignment value that is not 0 or 1, then
4205 the caller guarantees that the destination pointer is aligned to that boundary.
4211 The '<tt>llvm.memset.*</tt>' intrinsics fill "len" bytes of memory starting at
4213 destination location. If the argument is known to be aligned to some boundary,
4214 this can be specified as the fourth argument, otherwise it should be set to 0 or
4220 <!-- _______________________________________________________________________ -->
4221 <div class="doc_subsubsection">
4222 <a name="i_sqrt">'<tt>llvm.sqrt.*</tt>' Intrinsic</a>
4225 <div class="doc_text">
4229 declare float %llvm.sqrt.f32(float %Val)
4230 declare double %llvm.sqrt.f64(double %Val)
4236 The '<tt>llvm.sqrt</tt>' intrinsics return the sqrt of the specified operand,
4237 returning the same value as the libm '<tt>sqrt</tt>' function would. Unlike
4238 <tt>sqrt</tt> in libm, however, <tt>llvm.sqrt</tt> has undefined behavior for
4239 negative numbers (which allows for better optimization).
4245 The argument and return value are floating point numbers of the same type.
4251 This function returns the sqrt of the specified operand if it is a positive
4252 floating point number.
4256 <!-- _______________________________________________________________________ -->
4257 <div class="doc_subsubsection">
4258 <a name="i_powi">'<tt>llvm.powi.*</tt>' Intrinsic</a>
4261 <div class="doc_text">
4265 declare float %llvm.powi.f32(float %Val, i32 %power)
4266 declare double %llvm.powi.f64(double %Val, i32 %power)
4272 The '<tt>llvm.powi.*</tt>' intrinsics return the first operand raised to the
4273 specified (positive or negative) power. The order of evaluation of
4274 multiplications is not defined.
4280 The second argument is an integer power, and the first is a value to raise to
4287 This function returns the first value raised to the second power with an
4288 unspecified sequence of rounding operations.</p>
4292 <!-- ======================================================================= -->
4293 <div class="doc_subsection">
4294 <a name="int_manip">Bit Manipulation Intrinsics</a>
4297 <div class="doc_text">
4299 LLVM provides intrinsics for a few important bit manipulation operations.
4300 These allow efficient code generation for some algorithms.
4305 <!-- _______________________________________________________________________ -->
4306 <div class="doc_subsubsection">
4307 <a name="i_bswap">'<tt>llvm.bswap.*</tt>' Intrinsics</a>
4310 <div class="doc_text">
4314 declare i16 %llvm.bswap.i16(i16 <id>)
4315 declare i32 %llvm.bswap.i32(i32 <id>)
4316 declare i64 %llvm.bswap.i64(i64 <id>)
4322 The '<tt>llvm.bwsap</tt>' family of intrinsics is used to byteswap a 16, 32 or
4323 64 bit quantity. These are useful for performing operations on data that is not
4324 in the target's native byte order.
4330 The <tt>llvm.bswap.16</tt> intrinsic returns an i16 value that has the high
4331 and low byte of the input i16 swapped. Similarly, the <tt>llvm.bswap.i32</tt>
4332 intrinsic returns an i32 value that has the four bytes of the input i32
4333 swapped, so that if the input bytes are numbered 0, 1, 2, 3 then the returned
4334 i32 will have its bytes in 3, 2, 1, 0 order. The <tt>llvm.bswap.i64</tt>
4335 intrinsic extends this concept to 64 bits.
4340 <!-- _______________________________________________________________________ -->
4341 <div class="doc_subsubsection">
4342 <a name="int_ctpop">'<tt>llvm.ctpop.*</tt>' Intrinsic</a>
4345 <div class="doc_text">
4349 declare i8 %llvm.ctpop.i8 (i8 <src>)
4350 declare i16 %llvm.ctpop.i16(i16 <src>)
4351 declare i32 %llvm.ctpop.i32(i32 <src>)
4352 declare i64 %llvm.ctpop.i64(i64 <src>)
4358 The '<tt>llvm.ctpop</tt>' family of intrinsics counts the number of bits set in a
4365 The only argument is the value to be counted. The argument may be of any
4366 integer type. The return type must match the argument type.
4372 The '<tt>llvm.ctpop</tt>' intrinsic counts the 1's in a variable.
4376 <!-- _______________________________________________________________________ -->
4377 <div class="doc_subsubsection">
4378 <a name="int_ctlz">'<tt>llvm.ctlz.*</tt>' Intrinsic</a>
4381 <div class="doc_text">
4385 declare i8 %llvm.ctlz.i8 (i8 <src>)
4386 declare i16 %llvm.ctlz.i16(i16 <src>)
4387 declare i32 %llvm.ctlz.i32(i32 <src>)
4388 declare i64 %llvm.ctlz.i64(i64 <src>)
4394 The '<tt>llvm.ctlz</tt>' family of intrinsic functions counts the number of
4395 leading zeros in a variable.
4401 The only argument is the value to be counted. The argument may be of any
4402 integer type. The return type must match the argument type.
4408 The '<tt>llvm.ctlz</tt>' intrinsic counts the leading (most significant) zeros
4409 in a variable. If the src == 0 then the result is the size in bits of the type
4410 of src. For example, <tt>llvm.ctlz(i32 2) = 30</tt>.
4416 <!-- _______________________________________________________________________ -->
4417 <div class="doc_subsubsection">
4418 <a name="int_cttz">'<tt>llvm.cttz.*</tt>' Intrinsic</a>
4421 <div class="doc_text">
4425 declare i8 %llvm.cttz.i8 (i8 <src>)
4426 declare i16 %llvm.cttz.i16(i16 <src>)
4427 declare i32 %llvm.cttz.i32(i32 <src>)
4428 declare i64 %llvm.cttz.i64(i64 <src>)
4434 The '<tt>llvm.cttz</tt>' family of intrinsic functions counts the number of
4441 The only argument is the value to be counted. The argument may be of any
4442 integer type. The return type must match the argument type.
4448 The '<tt>llvm.cttz</tt>' intrinsic counts the trailing (least significant) zeros
4449 in a variable. If the src == 0 then the result is the size in bits of the type
4450 of src. For example, <tt>llvm.cttz(2) = 1</tt>.
4454 <!-- ======================================================================= -->
4455 <div class="doc_subsection">
4456 <a name="int_debugger">Debugger Intrinsics</a>
4459 <div class="doc_text">
4461 The LLVM debugger intrinsics (which all start with <tt>llvm.dbg.</tt> prefix),
4462 are described in the <a
4463 href="SourceLevelDebugging.html#format_common_intrinsics">LLVM Source Level
4464 Debugging</a> document.
4469 <!-- *********************************************************************** -->
4472 <a href="http://jigsaw.w3.org/css-validator/check/referer"><img
4473 src="http://jigsaw.w3.org/css-validator/images/vcss" alt="Valid CSS!"></a>
4474 <a href="http://validator.w3.org/check/referer"><img
4475 src="http://www.w3.org/Icons/valid-html401" alt="Valid HTML 4.01!" /></a>
4477 <a href="mailto:sabre@nondot.org">Chris Lattner</a><br>
4478 <a href="http://llvm.org">The LLVM Compiler Infrastructure</a><br>
4479 Last modified: $Date$