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="#moduleasm">Module-Level Inline Assembly</a></li>
30 <li><a href="#typesystem">Type System</a>
32 <li><a href="#t_primitive">Primitive Types</a>
34 <li><a href="#t_classifications">Type Classifications</a></li>
37 <li><a href="#t_derived">Derived Types</a>
39 <li><a href="#t_array">Array Type</a></li>
40 <li><a href="#t_function">Function Type</a></li>
41 <li><a href="#t_pointer">Pointer Type</a></li>
42 <li><a href="#t_struct">Structure Type</a></li>
43 <li><a href="#t_packed">Packed Type</a></li>
44 <li><a href="#t_opaque">Opaque Type</a></li>
49 <li><a href="#constants">Constants</a>
51 <li><a href="#simpleconstants">Simple Constants</a>
52 <li><a href="#aggregateconstants">Aggregate Constants</a>
53 <li><a href="#globalconstants">Global Variable and Function Addresses</a>
54 <li><a href="#undefvalues">Undefined Values</a>
55 <li><a href="#constantexprs">Constant Expressions</a>
58 <li><a href="#othervalues">Other Values</a>
60 <li><a href="#inlineasm">Inline Assembler Expressions</a>
63 <li><a href="#instref">Instruction Reference</a>
65 <li><a href="#terminators">Terminator Instructions</a>
67 <li><a href="#i_ret">'<tt>ret</tt>' Instruction</a></li>
68 <li><a href="#i_br">'<tt>br</tt>' Instruction</a></li>
69 <li><a href="#i_switch">'<tt>switch</tt>' Instruction</a></li>
70 <li><a href="#i_invoke">'<tt>invoke</tt>' Instruction</a></li>
71 <li><a href="#i_unwind">'<tt>unwind</tt>' Instruction</a></li>
72 <li><a href="#i_unreachable">'<tt>unreachable</tt>' Instruction</a></li>
75 <li><a href="#binaryops">Binary Operations</a>
77 <li><a href="#i_add">'<tt>add</tt>' Instruction</a></li>
78 <li><a href="#i_sub">'<tt>sub</tt>' Instruction</a></li>
79 <li><a href="#i_mul">'<tt>mul</tt>' Instruction</a></li>
80 <li><a href="#i_udiv">'<tt>udiv</tt>' Instruction</a></li>
81 <li><a href="#i_sdiv">'<tt>sdiv</tt>' Instruction</a></li>
82 <li><a href="#i_fdiv">'<tt>fdiv</tt>' Instruction</a></li>
83 <li><a href="#i_urem">'<tt>urem</tt>' Instruction</a></li>
84 <li><a href="#i_srem">'<tt>srem</tt>' Instruction</a></li>
85 <li><a href="#i_frem">'<tt>frem</tt>' Instruction</a></li>
88 <li><a href="#bitwiseops">Bitwise Binary Operations</a>
90 <li><a href="#i_and">'<tt>and</tt>' Instruction</a></li>
91 <li><a href="#i_or">'<tt>or</tt>' Instruction</a></li>
92 <li><a href="#i_xor">'<tt>xor</tt>' Instruction</a></li>
93 <li><a href="#i_shl">'<tt>shl</tt>' Instruction</a></li>
94 <li><a href="#i_lshr">'<tt>lshr</tt>' Instruction</a></li>
95 <li><a href="#i_ashr">'<tt>ashr</tt>' Instruction</a></li>
99 <li><a href="#vectorops">Vector Operations</a>
101 <li><a href="#i_extractelement">'<tt>extractelement</tt>' Instruction</a></li>
102 <li><a href="#i_insertelement">'<tt>insertelement</tt>' Instruction</a></li>
103 <li><a href="#i_shufflevector">'<tt>shufflevector</tt>' Instruction</a></li>
106 <li><a href="#memoryops">Memory Access and Addressing Operations</a>
108 <li><a href="#i_malloc">'<tt>malloc</tt>' Instruction</a></li>
109 <li><a href="#i_free">'<tt>free</tt>' Instruction</a></li>
110 <li><a href="#i_alloca">'<tt>alloca</tt>' Instruction</a></li>
111 <li><a href="#i_load">'<tt>load</tt>' Instruction</a></li>
112 <li><a href="#i_store">'<tt>store</tt>' Instruction</a></li>
113 <li><a href="#i_getelementptr">'<tt>getelementptr</tt>' Instruction</a></li>
116 <li><a href="#convertops">Conversion Operations</a>
118 <li><a href="#i_trunc">'<tt>trunc .. to</tt>' Instruction</a></li>
119 <li><a href="#i_zext">'<tt>zext .. to</tt>' Instruction</a></li>
120 <li><a href="#i_sext">'<tt>sext .. to</tt>' Instruction</a></li>
121 <li><a href="#i_fptrunc">'<tt>fptrunc .. to</tt>' Instruction</a></li>
122 <li><a href="#i_fpext">'<tt>fpext .. to</tt>' Instruction</a></li>
123 <li><a href="#i_fptoui">'<tt>fptoui .. to</tt>' Instruction</a></li>
124 <li><a href="#i_fptosi">'<tt>fptosi .. to</tt>' Instruction</a></li>
125 <li><a href="#i_uitofp">'<tt>uitofp .. to</tt>' Instruction</a></li>
126 <li><a href="#i_sitofp">'<tt>sitofp .. to</tt>' Instruction</a></li>
127 <li><a href="#i_ptrtoint">'<tt>ptrtoint .. to</tt>' Instruction</a></li>
128 <li><a href="#i_inttoptr">'<tt>inttoptr .. to</tt>' Instruction</a></li>
129 <li><a href="#i_bitcast">'<tt>bitcast .. to</tt>' Instruction</a></li>
131 <li><a href="#otherops">Other Operations</a>
133 <li><a href="#i_icmp">'<tt>icmp</tt>' Instruction</a></li>
134 <li><a href="#i_fcmp">'<tt>fcmp</tt>' Instruction</a></li>
135 <li><a href="#i_phi">'<tt>phi</tt>' Instruction</a></li>
136 <li><a href="#i_select">'<tt>select</tt>' Instruction</a></li>
137 <li><a href="#i_call">'<tt>call</tt>' Instruction</a></li>
138 <li><a href="#i_va_arg">'<tt>va_arg</tt>' Instruction</a></li>
143 <li><a href="#intrinsics">Intrinsic Functions</a>
145 <li><a href="#int_varargs">Variable Argument Handling Intrinsics</a>
147 <li><a href="#i_va_start">'<tt>llvm.va_start</tt>' Intrinsic</a></li>
148 <li><a href="#i_va_end">'<tt>llvm.va_end</tt>' Intrinsic</a></li>
149 <li><a href="#i_va_copy">'<tt>llvm.va_copy</tt>' Intrinsic</a></li>
152 <li><a href="#int_gc">Accurate Garbage Collection Intrinsics</a>
154 <li><a href="#i_gcroot">'<tt>llvm.gcroot</tt>' Intrinsic</a></li>
155 <li><a href="#i_gcread">'<tt>llvm.gcread</tt>' Intrinsic</a></li>
156 <li><a href="#i_gcwrite">'<tt>llvm.gcwrite</tt>' Intrinsic</a></li>
159 <li><a href="#int_codegen">Code Generator Intrinsics</a>
161 <li><a href="#i_returnaddress">'<tt>llvm.returnaddress</tt>' Intrinsic</a></li>
162 <li><a href="#i_frameaddress">'<tt>llvm.frameaddress</tt>' Intrinsic</a></li>
163 <li><a href="#i_stacksave">'<tt>llvm.stacksave</tt>' Intrinsic</a></li>
164 <li><a href="#i_stackrestore">'<tt>llvm.stackrestore</tt>' Intrinsic</a></li>
165 <li><a href="#i_prefetch">'<tt>llvm.prefetch</tt>' Intrinsic</a></li>
166 <li><a href="#i_pcmarker">'<tt>llvm.pcmarker</tt>' Intrinsic</a></li>
167 <li><a href="#i_readcyclecounter"><tt>llvm.readcyclecounter</tt>' Intrinsic</a></li>
170 <li><a href="#int_libc">Standard C Library Intrinsics</a>
172 <li><a href="#i_memcpy">'<tt>llvm.memcpy.*</tt>' Intrinsic</a></li>
173 <li><a href="#i_memmove">'<tt>llvm.memmove.*</tt>' Intrinsic</a></li>
174 <li><a href="#i_memset">'<tt>llvm.memset.*</tt>' Intrinsic</a></li>
175 <li><a href="#i_isunordered">'<tt>llvm.isunordered.*</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> int 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_uint">uint</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> uint %X, 8
312 <p>After strength reduction:</p>
315 %result = <a href="#i_shl">shl</a> uint %X, ubyte 3
318 <p>And the hard way:</p>
321 <a href="#i_add">add</a> uint %X, %X <i>; yields {uint}:%0</i>
322 <a href="#i_add">add</a> uint %0, %0 <i>; yields {uint}:%1</i>
323 %result = <a href="#i_add">add</a> uint %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 sbyte]</a> c"hello world\0A\00" <i>; [13 x sbyte]*</i>
369 <i>; External declaration of the puts function</i>
370 <a href="#functionstructure">declare</a> int %puts(sbyte*) <i>; int(sbyte*)* </i>
372 <i>; Global variable / Function body section separator</i>
375 <i>; Definition of main function</i>
376 int %main() { <i>; int()* </i>
377 <i>; Convert [13x sbyte]* to sbyte *...</i>
379 href="#i_getelementptr">getelementptr</a> [13 x sbyte]* %.LC0, long 0, long 0 <i>; sbyte*</i>
381 <i>; Call puts function to write out the string to stdout...</i>
383 href="#i_call">call</a> int %puts(sbyte* %cast210) <i>; int</i>
385 href="#i_ret">ret</a> int 0<br>}<br></pre>
387 <p>This example is made up of a <a href="#globalvars">global variable</a>
388 named "<tt>.LC0</tt>", an external declaration of the "<tt>puts</tt>"
389 function, and a <a href="#functionstructure">function definition</a>
390 for "<tt>main</tt>".</p>
392 <p>In general, a module is made up of a list of global values,
393 where both functions and global variables are global values. Global values are
394 represented by a pointer to a memory location (in this case, a pointer to an
395 array of char, and a pointer to a function), and have one of the following <a
396 href="#linkage">linkage types</a>.</p>
398 <p>Due to a limitation in the current LLVM assembly parser (it is limited by
399 one-token lookahead), modules are split into two pieces by the "implementation"
400 keyword. Global variable prototypes and definitions must occur before the
401 keyword, and function definitions must occur after it. Function prototypes may
402 occur either before or after it. In the future, the implementation keyword may
403 become a noop, if the parser gets smarter.</p>
407 <!-- ======================================================================= -->
408 <div class="doc_subsection">
409 <a name="linkage">Linkage Types</a>
412 <div class="doc_text">
415 All Global Variables and Functions have one of the following types of linkage:
420 <dt><tt><b><a name="linkage_internal">internal</a></b></tt> </dt>
422 <dd>Global values with internal linkage are only directly accessible by
423 objects in the current module. In particular, linking code into a module with
424 an internal global value may cause the internal to be renamed as necessary to
425 avoid collisions. Because the symbol is internal to the module, all
426 references can be updated. This corresponds to the notion of the
427 '<tt>static</tt>' keyword in C, or the idea of "anonymous namespaces" in C++.
430 <dt><tt><b><a name="linkage_linkonce">linkonce</a></b></tt>: </dt>
432 <dd>"<tt>linkonce</tt>" linkage is similar to <tt>internal</tt> linkage, with
433 the twist that linking together two modules defining the same
434 <tt>linkonce</tt> globals will cause one of the globals to be discarded. This
435 is typically used to implement inline functions. Unreferenced
436 <tt>linkonce</tt> globals are allowed to be discarded.
439 <dt><tt><b><a name="linkage_weak">weak</a></b></tt>: </dt>
441 <dd>"<tt>weak</tt>" linkage is exactly the same as <tt>linkonce</tt> linkage,
442 except that unreferenced <tt>weak</tt> globals may not be discarded. This is
443 used to implement constructs in C such as "<tt>int X;</tt>" at global scope.
446 <dt><tt><b><a name="linkage_appending">appending</a></b></tt>: </dt>
448 <dd>"<tt>appending</tt>" linkage may only be applied to global variables of
449 pointer to array type. When two global variables with appending linkage are
450 linked together, the two global arrays are appended together. This is the
451 LLVM, typesafe, equivalent of having the system linker append together
452 "sections" with identical names when .o files are linked.
455 <dt><tt><b><a name="linkage_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. It is illegal for a function <i>declaration</i>
498 to have any linkage type other than "externally visible".</a></p>
502 <!-- ======================================================================= -->
503 <div class="doc_subsection">
504 <a name="callingconv">Calling Conventions</a>
507 <div class="doc_text">
509 <p>LLVM <a href="#functionstructure">functions</a>, <a href="#i_call">calls</a>
510 and <a href="#i_invoke">invokes</a> can all have an optional calling convention
511 specified for the call. The calling convention of any pair of dynamic
512 caller/callee must match, or the behavior of the program is undefined. The
513 following calling conventions are supported by LLVM, and more may be added in
517 <dt><b>"<tt>ccc</tt>" - The C calling convention</b>:</dt>
519 <dd>This calling convention (the default if no other calling convention is
520 specified) matches the target C calling conventions. This calling convention
521 supports varargs function calls and tolerates some mismatch in the declared
522 prototype and implemented declaration of the function (as does normal C).
525 <dt><b>"<tt>csretcc</tt>" - The C struct return calling convention</b>:</dt>
527 <dd>This calling convention matches the target C calling conventions, except
528 that functions with this convention are required to take a pointer as their
529 first argument, and the return type of the function must be void. This is
530 used for C functions that return aggregates by-value. In this case, the
531 function has been transformed to take a pointer to the struct as the first
532 argument to the function. For targets where the ABI specifies specific
533 behavior for structure-return calls, the calling convention can be used to
534 distinguish between struct return functions and other functions that take a
535 pointer to a struct as the first argument.
538 <dt><b>"<tt>fastcc</tt>" - The fast calling convention</b>:</dt>
540 <dd>This calling convention attempts to make calls as fast as possible
541 (e.g. by passing things in registers). This calling convention allows the
542 target to use whatever tricks it wants to produce fast code for the target,
543 without having to conform to an externally specified ABI. Implementations of
544 this convention should allow arbitrary tail call optimization to be supported.
545 This calling convention does not support varargs and requires the prototype of
546 all callees to exactly match the prototype of the function definition.
549 <dt><b>"<tt>coldcc</tt>" - The cold calling convention</b>:</dt>
551 <dd>This calling convention attempts to make code in the caller as efficient
552 as possible under the assumption that the call is not commonly executed. As
553 such, these calls often preserve all registers so that the call does not break
554 any live ranges in the caller side. This calling convention does not support
555 varargs and requires the prototype of all callees to exactly match the
556 prototype of the function definition.
559 <dt><b>"<tt>cc <<em>n</em>></tt>" - Numbered convention</b>:</dt>
561 <dd>Any calling convention may be specified by number, allowing
562 target-specific calling conventions to be used. Target specific calling
563 conventions start at 64.
567 <p>More calling conventions can be added/defined on an as-needed basis, to
568 support pascal conventions or any other well-known target-independent
573 <!-- ======================================================================= -->
574 <div class="doc_subsection">
575 <a name="globalvars">Global Variables</a>
578 <div class="doc_text">
580 <p>Global variables define regions of memory allocated at compilation time
581 instead of run-time. Global variables may optionally be initialized, may have
582 an explicit section to be placed in, and may
583 have an optional explicit alignment specified. A
584 variable may be defined as a global "constant," which indicates that the
585 contents of the variable will <b>never</b> be modified (enabling better
586 optimization, allowing the global data to be placed in the read-only section of
587 an executable, etc). Note that variables that need runtime initialization
588 cannot be marked "constant" as there is a store to the variable.</p>
591 LLVM explicitly allows <em>declarations</em> of global variables to be marked
592 constant, even if the final definition of the global is not. This capability
593 can be used to enable slightly better optimization of the program, but requires
594 the language definition to guarantee that optimizations based on the
595 'constantness' are valid for the translation units that do not include the
599 <p>As SSA values, global variables define pointer values that are in
600 scope (i.e. they dominate) all basic blocks in the program. Global
601 variables always define a pointer to their "content" type because they
602 describe a region of memory, and all memory objects in LLVM are
603 accessed through pointers.</p>
605 <p>LLVM allows an explicit section to be specified for globals. If the target
606 supports it, it will emit globals to the section specified.</p>
608 <p>An explicit alignment may be specified for a global. If not present, or if
609 the alignment is set to zero, the alignment of the global is set by the target
610 to whatever it feels convenient. If an explicit alignment is specified, the
611 global is forced to have at least that much alignment. All alignments must be
617 <!-- ======================================================================= -->
618 <div class="doc_subsection">
619 <a name="functionstructure">Functions</a>
622 <div class="doc_text">
624 <p>LLVM function definitions consist of an optional <a href="#linkage">linkage
625 type</a>, an optional <a href="#callingconv">calling convention</a>, a return
626 type, a function name, a (possibly empty) argument list, an optional section,
627 an optional alignment, an opening curly brace,
628 a list of basic blocks, and a closing curly brace. LLVM function declarations
629 are defined with the "<tt>declare</tt>" keyword, an optional <a
630 href="#callingconv">calling convention</a>, a return type, a function name,
631 a possibly empty list of arguments, and an optional alignment.</p>
633 <p>A function definition contains a list of basic blocks, forming the CFG for
634 the function. Each basic block may optionally start with a label (giving the
635 basic block a symbol table entry), contains a list of instructions, and ends
636 with a <a href="#terminators">terminator</a> instruction (such as a branch or
637 function return).</p>
639 <p>The first basic block in a program is special in two ways: it is immediately
640 executed on entrance to the function, and it is not allowed to have predecessor
641 basic blocks (i.e. there can not be any branches to the entry block of a
642 function). Because the block can have no predecessors, it also cannot have any
643 <a href="#i_phi">PHI nodes</a>.</p>
645 <p>LLVM functions are identified by their name and type signature. Hence, two
646 functions with the same name but different parameter lists or return values are
647 considered different functions, and LLVM will resolve references to each
650 <p>LLVM allows an explicit section to be specified for functions. If the target
651 supports it, it will emit functions to the section specified.</p>
653 <p>An explicit alignment may be specified for a function. If not present, or if
654 the alignment is set to zero, the alignment of the function is set by the target
655 to whatever it feels convenient. If an explicit alignment is specified, the
656 function is forced to have at least that much alignment. All alignments must be
661 <!-- ======================================================================= -->
662 <div class="doc_subsection">
663 <a name="moduleasm">Module-Level Inline Assembly</a>
666 <div class="doc_text">
668 Modules may contain "module-level inline asm" blocks, which corresponds to the
669 GCC "file scope inline asm" blocks. These blocks are internally concatenated by
670 LLVM and treated as a single unit, but may be separated in the .ll file if
671 desired. The syntax is very simple:
674 <div class="doc_code"><pre>
675 module asm "inline asm code goes here"
676 module asm "more can go here"
679 <p>The strings can contain any character by escaping non-printable characters.
680 The escape sequence used is simply "\xx" where "xx" is the two digit hex code
685 The inline asm code is simply printed to the machine code .s file when
686 assembly code is generated.
691 <!-- *********************************************************************** -->
692 <div class="doc_section"> <a name="typesystem">Type System</a> </div>
693 <!-- *********************************************************************** -->
695 <div class="doc_text">
697 <p>The LLVM type system is one of the most important features of the
698 intermediate representation. Being typed enables a number of
699 optimizations to be performed on the IR directly, without having to do
700 extra analyses on the side before the transformation. A strong type
701 system makes it easier to read the generated code and enables novel
702 analyses and transformations that are not feasible to perform on normal
703 three address code representations.</p>
707 <!-- ======================================================================= -->
708 <div class="doc_subsection"> <a name="t_primitive">Primitive Types</a> </div>
709 <div class="doc_text">
710 <p>The primitive types are the fundamental building blocks of the LLVM
711 system. The current set of primitive types is as follows:</p>
713 <table class="layout">
718 <tr><th>Type</th><th>Description</th></tr>
719 <tr><td><tt>void</tt></td><td>No value</td></tr>
720 <tr><td><tt>ubyte</tt></td><td>Unsigned 8-bit value</td></tr>
721 <tr><td><tt>ushort</tt></td><td>Unsigned 16-bit value</td></tr>
722 <tr><td><tt>uint</tt></td><td>Unsigned 32-bit value</td></tr>
723 <tr><td><tt>ulong</tt></td><td>Unsigned 64-bit value</td></tr>
724 <tr><td><tt>float</tt></td><td>32-bit floating point value</td></tr>
725 <tr><td><tt>label</tt></td><td>Branch destination</td></tr>
732 <tr><th>Type</th><th>Description</th></tr>
733 <tr><td><tt>bool</tt></td><td>True or False value</td></tr>
734 <tr><td><tt>sbyte</tt></td><td>Signed 8-bit value</td></tr>
735 <tr><td><tt>short</tt></td><td>Signed 16-bit value</td></tr>
736 <tr><td><tt>int</tt></td><td>Signed 32-bit value</td></tr>
737 <tr><td><tt>long</tt></td><td>Signed 64-bit value</td></tr>
738 <tr><td><tt>double</tt></td><td>64-bit floating point value</td></tr>
746 <!-- _______________________________________________________________________ -->
747 <div class="doc_subsubsection"> <a name="t_classifications">Type
748 Classifications</a> </div>
749 <div class="doc_text">
750 <p>These different primitive types fall into a few useful
753 <table border="1" cellspacing="0" cellpadding="4">
755 <tr><th>Classification</th><th>Types</th></tr>
757 <td><a name="t_signed">signed</a></td>
758 <td><tt>sbyte, short, int, long, float, double</tt></td>
761 <td><a name="t_unsigned">unsigned</a></td>
762 <td><tt>ubyte, ushort, uint, ulong</tt></td>
765 <td><a name="t_integer">integer</a></td>
766 <td><tt>ubyte, sbyte, ushort, short, uint, int, ulong, long</tt></td>
769 <td><a name="t_integral">integral</a></td>
770 <td><tt>bool, ubyte, sbyte, ushort, short, uint, int, ulong, long</tt>
774 <td><a name="t_floating">floating point</a></td>
775 <td><tt>float, double</tt></td>
778 <td><a name="t_firstclass">first class</a></td>
779 <td><tt>bool, ubyte, sbyte, ushort, short, uint, int, ulong, long,<br>
780 float, double, <a href="#t_pointer">pointer</a>,
781 <a href="#t_packed">packed</a></tt></td>
786 <p>The <a href="#t_firstclass">first class</a> types are perhaps the
787 most important. Values of these types are the only ones which can be
788 produced by instructions, passed as arguments, or used as operands to
789 instructions. This means that all structures and arrays must be
790 manipulated either by pointer or by component.</p>
793 <!-- ======================================================================= -->
794 <div class="doc_subsection"> <a name="t_derived">Derived Types</a> </div>
796 <div class="doc_text">
798 <p>The real power in LLVM comes from the derived types in the system.
799 This is what allows a programmer to represent arrays, functions,
800 pointers, and other useful types. Note that these derived types may be
801 recursive: For example, it is possible to have a two dimensional array.</p>
805 <!-- _______________________________________________________________________ -->
806 <div class="doc_subsubsection"> <a name="t_array">Array Type</a> </div>
808 <div class="doc_text">
812 <p>The array type is a very simple derived type that arranges elements
813 sequentially in memory. The array type requires a size (number of
814 elements) and an underlying data type.</p>
819 [<# elements> x <elementtype>]
822 <p>The number of elements is a constant integer value; elementtype may
823 be any type with a size.</p>
826 <table class="layout">
829 <tt>[40 x int ]</tt><br/>
830 <tt>[41 x int ]</tt><br/>
831 <tt>[40 x uint]</tt><br/>
834 Array of 40 integer values.<br/>
835 Array of 41 integer values.<br/>
836 Array of 40 unsigned integer values.<br/>
840 <p>Here are some examples of multidimensional arrays:</p>
841 <table class="layout">
844 <tt>[3 x [4 x int]]</tt><br/>
845 <tt>[12 x [10 x float]]</tt><br/>
846 <tt>[2 x [3 x [4 x uint]]]</tt><br/>
849 3x4 array of integer values.<br/>
850 12x10 array of single precision floating point values.<br/>
851 2x3x4 array of unsigned integer values.<br/>
856 <p>Note that 'variable sized arrays' can be implemented in LLVM with a zero
857 length array. Normally, accesses past the end of an array are undefined in
858 LLVM (e.g. it is illegal to access the 5th element of a 3 element array).
859 As a special case, however, zero length arrays are recognized to be variable
860 length. This allows implementation of 'pascal style arrays' with the LLVM
861 type "{ int, [0 x float]}", for example.</p>
865 <!-- _______________________________________________________________________ -->
866 <div class="doc_subsubsection"> <a name="t_function">Function Type</a> </div>
867 <div class="doc_text">
869 <p>The function type can be thought of as a function signature. It
870 consists of a return type and a list of formal parameter types.
871 Function types are usually used to build virtual function tables
872 (which are structures of pointers to functions), for indirect function
873 calls, and when defining a function.</p>
875 The return type of a function type cannot be an aggregate type.
878 <pre> <returntype> (<parameter list>)<br></pre>
879 <p>...where '<tt><parameter list></tt>' is a comma-separated list of type
880 specifiers. Optionally, the parameter list may include a type <tt>...</tt>,
881 which indicates that the function takes a variable number of arguments.
882 Variable argument functions can access their arguments with the <a
883 href="#int_varargs">variable argument handling intrinsic</a> functions.</p>
885 <table class="layout">
888 <tt>int (int)</tt> <br/>
889 <tt>float (int, int *) *</tt><br/>
890 <tt>int (sbyte *, ...)</tt><br/>
893 function taking an <tt>int</tt>, returning an <tt>int</tt><br/>
894 <a href="#t_pointer">Pointer</a> to a function that takes an
895 <tt>int</tt> and a <a href="#t_pointer">pointer</a> to <tt>int</tt>,
896 returning <tt>float</tt>.<br/>
897 A vararg function that takes at least one <a href="#t_pointer">pointer</a>
898 to <tt>sbyte</tt> (signed char in C), which returns an integer. This is
899 the signature for <tt>printf</tt> in LLVM.<br/>
905 <!-- _______________________________________________________________________ -->
906 <div class="doc_subsubsection"> <a name="t_struct">Structure Type</a> </div>
907 <div class="doc_text">
909 <p>The structure type is used to represent a collection of data members
910 together in memory. The packing of the field types is defined to match
911 the ABI of the underlying processor. The elements of a structure may
912 be any type that has a size.</p>
913 <p>Structures are accessed using '<tt><a href="#i_load">load</a></tt>
914 and '<tt><a href="#i_store">store</a></tt>' by getting a pointer to a
915 field with the '<tt><a href="#i_getelementptr">getelementptr</a></tt>'
918 <pre> { <type list> }<br></pre>
920 <table class="layout">
923 <tt>{ int, int, int }</tt><br/>
924 <tt>{ float, int (int) * }</tt><br/>
927 a triple of three <tt>int</tt> values<br/>
928 A pair, where the first element is a <tt>float</tt> and the second element
929 is a <a href="#t_pointer">pointer</a> to a <a href="#t_function">function</a>
930 that takes an <tt>int</tt>, returning an <tt>int</tt>.<br/>
936 <!-- _______________________________________________________________________ -->
937 <div class="doc_subsubsection"> <a name="t_pointer">Pointer Type</a> </div>
938 <div class="doc_text">
940 <p>As in many languages, the pointer type represents a pointer or
941 reference to another object, which must live in memory.</p>
943 <pre> <type> *<br></pre>
945 <table class="layout">
948 <tt>[4x int]*</tt><br/>
949 <tt>int (int *) *</tt><br/>
952 A <a href="#t_pointer">pointer</a> to <a href="#t_array">array</a> of
953 four <tt>int</tt> values<br/>
954 A <a href="#t_pointer">pointer</a> to a <a
955 href="#t_function">function</a> that takes an <tt>int*</tt>, returning an
962 <!-- _______________________________________________________________________ -->
963 <div class="doc_subsubsection"> <a name="t_packed">Packed Type</a> </div>
964 <div class="doc_text">
968 <p>A packed type is a simple derived type that represents a vector
969 of elements. Packed types are used when multiple primitive data
970 are operated in parallel using a single instruction (SIMD).
971 A packed type requires a size (number of
972 elements) and an underlying primitive data type. Vectors must have a power
973 of two length (1, 2, 4, 8, 16 ...). Packed types are
974 considered <a href="#t_firstclass">first class</a>.</p>
979 < <# elements> x <elementtype> >
982 <p>The number of elements is a constant integer value; elementtype may
983 be any integral or floating point type.</p>
987 <table class="layout">
990 <tt><4 x int></tt><br/>
991 <tt><8 x float></tt><br/>
992 <tt><2 x uint></tt><br/>
995 Packed vector of 4 integer values.<br/>
996 Packed vector of 8 floating-point values.<br/>
997 Packed vector of 2 unsigned integer values.<br/>
1003 <!-- _______________________________________________________________________ -->
1004 <div class="doc_subsubsection"> <a name="t_opaque">Opaque Type</a> </div>
1005 <div class="doc_text">
1009 <p>Opaque types are used to represent unknown types in the system. This
1010 corresponds (for example) to the C notion of a foward declared structure type.
1011 In LLVM, opaque types can eventually be resolved to any type (not just a
1012 structure type).</p>
1022 <table class="layout">
1028 An opaque type.<br/>
1035 <!-- *********************************************************************** -->
1036 <div class="doc_section"> <a name="constants">Constants</a> </div>
1037 <!-- *********************************************************************** -->
1039 <div class="doc_text">
1041 <p>LLVM has several different basic types of constants. This section describes
1042 them all and their syntax.</p>
1046 <!-- ======================================================================= -->
1047 <div class="doc_subsection"><a name="simpleconstants">Simple Constants</a></div>
1049 <div class="doc_text">
1052 <dt><b>Boolean constants</b></dt>
1054 <dd>The two strings '<tt>true</tt>' and '<tt>false</tt>' are both valid
1055 constants of the <tt><a href="#t_primitive">bool</a></tt> type.
1058 <dt><b>Integer constants</b></dt>
1060 <dd>Standard integers (such as '4') are constants of the <a
1061 href="#t_integer">integer</a> type. Negative numbers may be used with signed
1065 <dt><b>Floating point constants</b></dt>
1067 <dd>Floating point constants use standard decimal notation (e.g. 123.421),
1068 exponential notation (e.g. 1.23421e+2), or a more precise hexadecimal
1069 notation (see below). Floating point constants must have a <a
1070 href="#t_floating">floating point</a> type. </dd>
1072 <dt><b>Null pointer constants</b></dt>
1074 <dd>The identifier '<tt>null</tt>' is recognized as a null pointer constant
1075 and must be of <a href="#t_pointer">pointer type</a>.</dd>
1079 <p>The one non-intuitive notation for constants is the optional hexadecimal form
1080 of floating point constants. For example, the form '<tt>double
1081 0x432ff973cafa8000</tt>' is equivalent to (but harder to read than) '<tt>double
1082 4.5e+15</tt>'. The only time hexadecimal floating point constants are required
1083 (and the only time that they are generated by the disassembler) is when a
1084 floating point constant must be emitted but it cannot be represented as a
1085 decimal floating point number. For example, NaN's, infinities, and other
1086 special values are represented in their IEEE hexadecimal format so that
1087 assembly and disassembly do not cause any bits to change in the constants.</p>
1091 <!-- ======================================================================= -->
1092 <div class="doc_subsection"><a name="aggregateconstants">Aggregate Constants</a>
1095 <div class="doc_text">
1096 <p>Aggregate constants arise from aggregation of simple constants
1097 and smaller aggregate constants.</p>
1100 <dt><b>Structure constants</b></dt>
1102 <dd>Structure constants are represented with notation similar to structure
1103 type definitions (a comma separated list of elements, surrounded by braces
1104 (<tt>{}</tt>)). For example: "<tt>{ int 4, float 17.0, int* %G }</tt>",
1105 where "<tt>%G</tt>" is declared as "<tt>%G = external global int</tt>". Structure constants
1106 must have <a href="#t_struct">structure type</a>, and the number and
1107 types of elements must match those specified by the type.
1110 <dt><b>Array constants</b></dt>
1112 <dd>Array constants are represented with notation similar to array type
1113 definitions (a comma separated list of elements, surrounded by square brackets
1114 (<tt>[]</tt>)). For example: "<tt>[ int 42, int 11, int 74 ]</tt>". Array
1115 constants must have <a href="#t_array">array type</a>, and the number and
1116 types of elements must match those specified by the type.
1119 <dt><b>Packed constants</b></dt>
1121 <dd>Packed constants are represented with notation similar to packed type
1122 definitions (a comma separated list of elements, surrounded by
1123 less-than/greater-than's (<tt><></tt>)). For example: "<tt>< int 42,
1124 int 11, int 74, int 100 ></tt>". Packed constants must have <a
1125 href="#t_packed">packed type</a>, and the number and types of elements must
1126 match those specified by the type.
1129 <dt><b>Zero initialization</b></dt>
1131 <dd>The string '<tt>zeroinitializer</tt>' can be used to zero initialize a
1132 value to zero of <em>any</em> type, including scalar and aggregate types.
1133 This is often used to avoid having to print large zero initializers (e.g. for
1134 large arrays) and is always exactly equivalent to using explicit zero
1141 <!-- ======================================================================= -->
1142 <div class="doc_subsection">
1143 <a name="globalconstants">Global Variable and Function Addresses</a>
1146 <div class="doc_text">
1148 <p>The addresses of <a href="#globalvars">global variables</a> and <a
1149 href="#functionstructure">functions</a> are always implicitly valid (link-time)
1150 constants. These constants are explicitly referenced when the <a
1151 href="#identifiers">identifier for the global</a> is used and always have <a
1152 href="#t_pointer">pointer</a> type. For example, the following is a legal LLVM
1158 %Z = global [2 x int*] [ int* %X, int* %Y ]
1163 <!-- ======================================================================= -->
1164 <div class="doc_subsection"><a name="undefvalues">Undefined Values</a></div>
1165 <div class="doc_text">
1166 <p>The string '<tt>undef</tt>' is recognized as a type-less constant that has
1167 no specific value. Undefined values may be of any type and be used anywhere
1168 a constant is permitted.</p>
1170 <p>Undefined values indicate to the compiler that the program is well defined
1171 no matter what value is used, giving the compiler more freedom to optimize.
1175 <!-- ======================================================================= -->
1176 <div class="doc_subsection"><a name="constantexprs">Constant Expressions</a>
1179 <div class="doc_text">
1181 <p>Constant expressions are used to allow expressions involving other constants
1182 to be used as constants. Constant expressions may be of any <a
1183 href="#t_firstclass">first class</a> type and may involve any LLVM operation
1184 that does not have side effects (e.g. load and call are not supported). The
1185 following is the syntax for constant expressions:</p>
1188 <dt><b><tt>trunc ( CST to TYPE )</tt></b></dt>
1189 <dd>Truncate a constant to another type. The bit size of CST must be larger
1190 than the bit size of TYPE. Both types must be integral.</dd>
1192 <dt><b><tt>zext ( CST to TYPE )</tt></b></dt>
1193 <dd>Zero extend a constant to another type. The bit size of CST must be
1194 smaller or equal to the bit size of TYPE. Both types must be integral.</dd>
1196 <dt><b><tt>sext ( CST to TYPE )</tt></b></dt>
1197 <dd>Sign extend a constant to another type. The bit size of CST must be
1198 smaller or equal to the bit size of TYPE. Both types must be integral.</dd>
1200 <dt><b><tt>fptrunc ( CST to TYPE )</tt></b></dt>
1201 <dd>Truncate a floating point constant to another floating point type. The
1202 size of CST must be larger than the size of TYPE. Both types must be
1203 floating point.</dd>
1205 <dt><b><tt>fpext ( CST to TYPE )</tt></b></dt>
1206 <dd>Floating point extend a constant to another type. The size of CST must be
1207 smaller or equal to the size of TYPE. Both types must be floating point.</dd>
1209 <dt><b><tt>fp2uint ( CST to TYPE )</tt></b></dt>
1210 <dd>Convert a floating point constant to the corresponding unsigned integer
1211 constant. TYPE must be an integer type. CST must be floating point. If the
1212 value won't fit in the integer type, the results are undefined.</dd>
1214 <dt><b><tt>fptosi ( CST to TYPE )</tt></b></dt>
1215 <dd>Convert a floating point constant to the corresponding signed integer
1216 constant. TYPE must be an integer type. CST must be floating point. If the
1217 value won't fit in the integer type, the results are undefined.</dd>
1219 <dt><b><tt>uitofp ( CST to TYPE )</tt></b></dt>
1220 <dd>Convert an unsigned integer constant to the corresponding floating point
1221 constant. TYPE must be floating point. CST must be of integer type. If the
1222 value won't fit in the floating point type, the results are undefined.</dd>
1224 <dt><b><tt>sitofp ( CST to TYPE )</tt></b></dt>
1225 <dd>Convert a signed integer constant to the corresponding floating point
1226 constant. TYPE must be floating point. CST must be of integer type. If the
1227 value won't fit in the floating point type, the results are undefined.</dd>
1229 <dt><b><tt>ptrtoint ( CST to TYPE )</tt></b></dt>
1230 <dd>Convert a pointer typed constant to the corresponding integer constant
1231 TYPE must be an integer type. CST must be of pointer type. The CST value is
1232 zero extended, truncated, or unchanged to make it fit in TYPE.</dd>
1234 <dt><b><tt>inttoptr ( CST to TYPE )</tt></b></dt>
1235 <dd>Convert a integer constant to a pointer constant. TYPE must be a
1236 pointer type. CST must be of integer type. The CST value is zero extended,
1237 truncated, or unchanged to make it fit in a pointer size. This one is
1238 <i>really</i> dangerous!</dd>
1240 <dt><b><tt>bitcast ( CST to TYPE )</tt></b></dt>
1241 <dd>Convert a constant, CST, to another TYPE. The size of CST and TYPE must be
1242 identical (same number of bits). The conversion is done as if the CST value
1243 was stored to memory and read back as TYPE. In other words, no bits change
1244 with this operator, just the type. This can be used for conversion of
1245 packed types to any other type, as long as they have the same bit width. For
1246 pointers it is only valid to cast to another pointer type.
1249 <dt><b><tt>getelementptr ( CSTPTR, IDX0, IDX1, ... )</tt></b></dt>
1251 <dd>Perform the <a href="#i_getelementptr">getelementptr operation</a> on
1252 constants. As with the <a href="#i_getelementptr">getelementptr</a>
1253 instruction, the index list may have zero or more indexes, which are required
1254 to make sense for the type of "CSTPTR".</dd>
1256 <dt><b><tt>select ( COND, VAL1, VAL2 )</tt></b></dt>
1258 <dd>Perform the <a href="#i_select">select operation</a> on
1261 <dt><b><tt>extractelement ( VAL, IDX )</tt></b></dt>
1263 <dd>Perform the <a href="#i_extractelement">extractelement
1264 operation</a> on constants.
1266 <dt><b><tt>insertelement ( VAL, ELT, IDX )</tt></b></dt>
1268 <dd>Perform the <a href="#i_insertelement">insertelement
1269 operation</a> on constants.
1272 <dt><b><tt>shufflevector ( VEC1, VEC2, IDXMASK )</tt></b></dt>
1274 <dd>Perform the <a href="#i_shufflevector">shufflevector
1275 operation</a> on constants.
1277 <dt><b><tt>OPCODE ( LHS, RHS )</tt></b></dt>
1279 <dd>Perform the specified operation of the LHS and RHS constants. OPCODE may
1280 be any of the <a href="#binaryops">binary</a> or <a href="#bitwiseops">bitwise
1281 binary</a> operations. The constraints on operands are the same as those for
1282 the corresponding instruction (e.g. no bitwise operations on floating point
1283 values are allowed).</dd>
1287 <!-- *********************************************************************** -->
1288 <div class="doc_section"> <a name="othervalues">Other Values</a> </div>
1289 <!-- *********************************************************************** -->
1291 <!-- ======================================================================= -->
1292 <div class="doc_subsection">
1293 <a name="inlineasm">Inline Assembler Expressions</a>
1296 <div class="doc_text">
1299 LLVM supports inline assembler expressions (as opposed to <a href="#moduleasm">
1300 Module-Level Inline Assembly</a>) through the use of a special value. This
1301 value represents the inline assembler as a string (containing the instructions
1302 to emit), a list of operand constraints (stored as a string), and a flag that
1303 indicates whether or not the inline asm expression has side effects. An example
1304 inline assembler expression is:
1308 int(int) asm "bswap $0", "=r,r"
1312 Inline assembler expressions may <b>only</b> be used as the callee operand of
1313 a <a href="#i_call"><tt>call</tt> instruction</a>. Thus, typically we have:
1317 %X = call int asm "<a href="#i_bswap">bswap</a> $0", "=r,r"(int %Y)
1321 Inline asms with side effects not visible in the constraint list must be marked
1322 as having side effects. This is done through the use of the
1323 '<tt>sideeffect</tt>' keyword, like so:
1327 call void asm sideeffect "eieio", ""()
1330 <p>TODO: The format of the asm and constraints string still need to be
1331 documented here. Constraints on what can be done (e.g. duplication, moving, etc
1332 need to be documented).
1337 <!-- *********************************************************************** -->
1338 <div class="doc_section"> <a name="instref">Instruction Reference</a> </div>
1339 <!-- *********************************************************************** -->
1341 <div class="doc_text">
1343 <p>The LLVM instruction set consists of several different
1344 classifications of instructions: <a href="#terminators">terminator
1345 instructions</a>, <a href="#binaryops">binary instructions</a>,
1346 <a href="#bitwiseops">bitwise binary instructions</a>, <a
1347 href="#memoryops">memory instructions</a>, and <a href="#otherops">other
1348 instructions</a>.</p>
1352 <!-- ======================================================================= -->
1353 <div class="doc_subsection"> <a name="terminators">Terminator
1354 Instructions</a> </div>
1356 <div class="doc_text">
1358 <p>As mentioned <a href="#functionstructure">previously</a>, every
1359 basic block in a program ends with a "Terminator" instruction, which
1360 indicates which block should be executed after the current block is
1361 finished. These terminator instructions typically yield a '<tt>void</tt>'
1362 value: they produce control flow, not values (the one exception being
1363 the '<a href="#i_invoke"><tt>invoke</tt></a>' instruction).</p>
1364 <p>There are six different terminator instructions: the '<a
1365 href="#i_ret"><tt>ret</tt></a>' instruction, the '<a href="#i_br"><tt>br</tt></a>'
1366 instruction, the '<a href="#i_switch"><tt>switch</tt></a>' instruction,
1367 the '<a href="#i_invoke"><tt>invoke</tt></a>' instruction, the '<a
1368 href="#i_unwind"><tt>unwind</tt></a>' instruction, and the '<a
1369 href="#i_unreachable"><tt>unreachable</tt></a>' instruction.</p>
1373 <!-- _______________________________________________________________________ -->
1374 <div class="doc_subsubsection"> <a name="i_ret">'<tt>ret</tt>'
1375 Instruction</a> </div>
1376 <div class="doc_text">
1378 <pre> ret <type> <value> <i>; Return a value from a non-void function</i>
1379 ret void <i>; Return from void function</i>
1382 <p>The '<tt>ret</tt>' instruction is used to return control flow (and a
1383 value) from a function back to the caller.</p>
1384 <p>There are two forms of the '<tt>ret</tt>' instruction: one that
1385 returns a value and then causes control flow, and one that just causes
1386 control flow to occur.</p>
1388 <p>The '<tt>ret</tt>' instruction may return any '<a
1389 href="#t_firstclass">first class</a>' type. Notice that a function is
1390 not <a href="#wellformed">well formed</a> if there exists a '<tt>ret</tt>'
1391 instruction inside of the function that returns a value that does not
1392 match the return type of the function.</p>
1394 <p>When the '<tt>ret</tt>' instruction is executed, control flow
1395 returns back to the calling function's context. If the caller is a "<a
1396 href="#i_call"><tt>call</tt></a>" instruction, execution continues at
1397 the instruction after the call. If the caller was an "<a
1398 href="#i_invoke"><tt>invoke</tt></a>" instruction, execution continues
1399 at the beginning of the "normal" destination block. If the instruction
1400 returns a value, that value shall set the call or invoke instruction's
1403 <pre> ret int 5 <i>; Return an integer value of 5</i>
1404 ret void <i>; Return from a void function</i>
1407 <!-- _______________________________________________________________________ -->
1408 <div class="doc_subsubsection"> <a name="i_br">'<tt>br</tt>' Instruction</a> </div>
1409 <div class="doc_text">
1411 <pre> br bool <cond>, label <iftrue>, label <iffalse><br> br label <dest> <i>; Unconditional branch</i>
1414 <p>The '<tt>br</tt>' instruction is used to cause control flow to
1415 transfer to a different basic block in the current function. There are
1416 two forms of this instruction, corresponding to a conditional branch
1417 and an unconditional branch.</p>
1419 <p>The conditional branch form of the '<tt>br</tt>' instruction takes a
1420 single '<tt>bool</tt>' value and two '<tt>label</tt>' values. The
1421 unconditional form of the '<tt>br</tt>' instruction takes a single '<tt>label</tt>'
1422 value as a target.</p>
1424 <p>Upon execution of a conditional '<tt>br</tt>' instruction, the '<tt>bool</tt>'
1425 argument is evaluated. If the value is <tt>true</tt>, control flows
1426 to the '<tt>iftrue</tt>' <tt>label</tt> argument. If "cond" is <tt>false</tt>,
1427 control flows to the '<tt>iffalse</tt>' <tt>label</tt> argument.</p>
1429 <pre>Test:<br> %cond = <a href="#i_icmp">icmp</a> eq, int %a, %b<br> br bool %cond, label %IfEqual, label %IfUnequal<br>IfEqual:<br> <a
1430 href="#i_ret">ret</a> int 1<br>IfUnequal:<br> <a href="#i_ret">ret</a> int 0<br></pre>
1432 <!-- _______________________________________________________________________ -->
1433 <div class="doc_subsubsection">
1434 <a name="i_switch">'<tt>switch</tt>' Instruction</a>
1437 <div class="doc_text">
1441 switch <intty> <value>, label <defaultdest> [ <intty> <val>, label <dest> ... ]
1446 <p>The '<tt>switch</tt>' instruction is used to transfer control flow to one of
1447 several different places. It is a generalization of the '<tt>br</tt>'
1448 instruction, allowing a branch to occur to one of many possible
1454 <p>The '<tt>switch</tt>' instruction uses three parameters: an integer
1455 comparison value '<tt>value</tt>', a default '<tt>label</tt>' destination, and
1456 an array of pairs of comparison value constants and '<tt>label</tt>'s. The
1457 table is not allowed to contain duplicate constant entries.</p>
1461 <p>The <tt>switch</tt> instruction specifies a table of values and
1462 destinations. When the '<tt>switch</tt>' instruction is executed, this
1463 table is searched for the given value. If the value is found, control flow is
1464 transfered to the corresponding destination; otherwise, control flow is
1465 transfered to the default destination.</p>
1467 <h5>Implementation:</h5>
1469 <p>Depending on properties of the target machine and the particular
1470 <tt>switch</tt> instruction, this instruction may be code generated in different
1471 ways. For example, it could be generated as a series of chained conditional
1472 branches or with a lookup table.</p>
1477 <i>; Emulate a conditional br instruction</i>
1478 %Val = <a href="#i_zext">zext</a> bool %value to int
1479 switch int %Val, label %truedest [int 0, label %falsedest ]
1481 <i>; Emulate an unconditional br instruction</i>
1482 switch uint 0, label %dest [ ]
1484 <i>; Implement a jump table:</i>
1485 switch uint %val, label %otherwise [ uint 0, label %onzero
1486 uint 1, label %onone
1487 uint 2, label %ontwo ]
1491 <!-- _______________________________________________________________________ -->
1492 <div class="doc_subsubsection">
1493 <a name="i_invoke">'<tt>invoke</tt>' Instruction</a>
1496 <div class="doc_text">
1501 <result> = invoke [<a href="#callingconv">cconv</a>] <ptr to function ty> %<function ptr val>(<function args>)
1502 to label <normal label> unwind label <exception label>
1507 <p>The '<tt>invoke</tt>' instruction causes control to transfer to a specified
1508 function, with the possibility of control flow transfer to either the
1509 '<tt>normal</tt>' label or the
1510 '<tt>exception</tt>' label. If the callee function returns with the
1511 "<tt><a href="#i_ret">ret</a></tt>" instruction, control flow will return to the
1512 "normal" label. If the callee (or any indirect callees) returns with the "<a
1513 href="#i_unwind"><tt>unwind</tt></a>" instruction, control is interrupted and
1514 continued at the dynamically nearest "exception" label.</p>
1518 <p>This instruction requires several arguments:</p>
1522 The optional "cconv" marker indicates which <a href="callingconv">calling
1523 convention</a> the call should use. If none is specified, the call defaults
1524 to using C calling conventions.
1526 <li>'<tt>ptr to function ty</tt>': shall be the signature of the pointer to
1527 function value being invoked. In most cases, this is a direct function
1528 invocation, but indirect <tt>invoke</tt>s are just as possible, branching off
1529 an arbitrary pointer to function value.
1532 <li>'<tt>function ptr val</tt>': An LLVM value containing a pointer to a
1533 function to be invoked. </li>
1535 <li>'<tt>function args</tt>': argument list whose types match the function
1536 signature argument types. If the function signature indicates the function
1537 accepts a variable number of arguments, the extra arguments can be
1540 <li>'<tt>normal label</tt>': the label reached when the called function
1541 executes a '<tt><a href="#i_ret">ret</a></tt>' instruction. </li>
1543 <li>'<tt>exception label</tt>': the label reached when a callee returns with
1544 the <a href="#i_unwind"><tt>unwind</tt></a> instruction. </li>
1550 <p>This instruction is designed to operate as a standard '<tt><a
1551 href="#i_call">call</a></tt>' instruction in most regards. The primary
1552 difference is that it establishes an association with a label, which is used by
1553 the runtime library to unwind the stack.</p>
1555 <p>This instruction is used in languages with destructors to ensure that proper
1556 cleanup is performed in the case of either a <tt>longjmp</tt> or a thrown
1557 exception. Additionally, this is important for implementation of
1558 '<tt>catch</tt>' clauses in high-level languages that support them.</p>
1562 %retval = invoke int %Test(int 15) to label %Continue
1563 unwind label %TestCleanup <i>; {int}:retval set</i>
1564 %retval = invoke <a href="#callingconv">coldcc</a> int %Test(int 15) to label %Continue
1565 unwind label %TestCleanup <i>; {int}:retval set</i>
1570 <!-- _______________________________________________________________________ -->
1572 <div class="doc_subsubsection"> <a name="i_unwind">'<tt>unwind</tt>'
1573 Instruction</a> </div>
1575 <div class="doc_text">
1584 <p>The '<tt>unwind</tt>' instruction unwinds the stack, continuing control flow
1585 at the first callee in the dynamic call stack which used an <a
1586 href="#i_invoke"><tt>invoke</tt></a> instruction to perform the call. This is
1587 primarily used to implement exception handling.</p>
1591 <p>The '<tt>unwind</tt>' intrinsic causes execution of the current function to
1592 immediately halt. The dynamic call stack is then searched for the first <a
1593 href="#i_invoke"><tt>invoke</tt></a> instruction on the call stack. Once found,
1594 execution continues at the "exceptional" destination block specified by the
1595 <tt>invoke</tt> instruction. If there is no <tt>invoke</tt> instruction in the
1596 dynamic call chain, undefined behavior results.</p>
1599 <!-- _______________________________________________________________________ -->
1601 <div class="doc_subsubsection"> <a name="i_unreachable">'<tt>unreachable</tt>'
1602 Instruction</a> </div>
1604 <div class="doc_text">
1613 <p>The '<tt>unreachable</tt>' instruction has no defined semantics. This
1614 instruction is used to inform the optimizer that a particular portion of the
1615 code is not reachable. This can be used to indicate that the code after a
1616 no-return function cannot be reached, and other facts.</p>
1620 <p>The '<tt>unreachable</tt>' instruction has no defined semantics.</p>
1625 <!-- ======================================================================= -->
1626 <div class="doc_subsection"> <a name="binaryops">Binary Operations</a> </div>
1627 <div class="doc_text">
1628 <p>Binary operators are used to do most of the computation in a
1629 program. They require two operands, execute an operation on them, and
1630 produce a single value. The operands might represent
1631 multiple data, as is the case with the <a href="#t_packed">packed</a> data type.
1632 The result value of a binary operator is not
1633 necessarily the same type as its operands.</p>
1634 <p>There are several different binary operators:</p>
1636 <!-- _______________________________________________________________________ -->
1637 <div class="doc_subsubsection"> <a name="i_add">'<tt>add</tt>'
1638 Instruction</a> </div>
1639 <div class="doc_text">
1641 <pre> <result> = add <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1644 <p>The '<tt>add</tt>' instruction returns the sum of its two operands.</p>
1646 <p>The two arguments to the '<tt>add</tt>' instruction must be either <a
1647 href="#t_integer">integer</a> or <a href="#t_floating">floating point</a> values.
1648 This instruction can also take <a href="#t_packed">packed</a> versions of the values.
1649 Both arguments must have identical types.</p>
1651 <p>The value produced is the integer or floating point sum of the two
1654 <pre> <result> = add int 4, %var <i>; yields {int}:result = 4 + %var</i>
1657 <!-- _______________________________________________________________________ -->
1658 <div class="doc_subsubsection"> <a name="i_sub">'<tt>sub</tt>'
1659 Instruction</a> </div>
1660 <div class="doc_text">
1662 <pre> <result> = sub <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1665 <p>The '<tt>sub</tt>' instruction returns the difference of its two
1667 <p>Note that the '<tt>sub</tt>' instruction is used to represent the '<tt>neg</tt>'
1668 instruction present in most other intermediate representations.</p>
1670 <p>The two arguments to the '<tt>sub</tt>' instruction must be either <a
1671 href="#t_integer">integer</a> or <a href="#t_floating">floating point</a>
1673 This instruction can also take <a href="#t_packed">packed</a> versions of the values.
1674 Both arguments must have identical types.</p>
1676 <p>The value produced is the integer or floating point difference of
1677 the two operands.</p>
1679 <pre> <result> = sub int 4, %var <i>; yields {int}:result = 4 - %var</i>
1680 <result> = sub int 0, %val <i>; yields {int}:result = -%var</i>
1683 <!-- _______________________________________________________________________ -->
1684 <div class="doc_subsubsection"> <a name="i_mul">'<tt>mul</tt>'
1685 Instruction</a> </div>
1686 <div class="doc_text">
1688 <pre> <result> = mul <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1691 <p>The '<tt>mul</tt>' instruction returns the product of its two
1694 <p>The two arguments to the '<tt>mul</tt>' instruction must be either <a
1695 href="#t_integer">integer</a> or <a href="#t_floating">floating point</a>
1697 This instruction can also take <a href="#t_packed">packed</a> versions of the values.
1698 Both arguments must have identical types.</p>
1700 <p>The value produced is the integer or floating point product of the
1702 <p>There is no signed vs unsigned multiplication. The appropriate
1703 action is taken based on the type of the operand.</p>
1705 <pre> <result> = mul int 4, %var <i>; yields {int}:result = 4 * %var</i>
1708 <!-- _______________________________________________________________________ -->
1709 <div class="doc_subsubsection"> <a name="i_udiv">'<tt>udiv</tt>' Instruction
1711 <div class="doc_text">
1713 <pre> <result> = udiv <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1716 <p>The '<tt>udiv</tt>' instruction returns the quotient of its two
1719 <p>The two arguments to the '<tt>udiv</tt>' instruction must be
1720 <a href="#t_integer">integer</a> values. Both arguments must have identical
1721 types. This instruction can also take <a href="#t_packed">packed</a> versions
1722 of the values in which case the elements must be integers.</p>
1724 <p>The value produced is the unsigned integer quotient of the two operands. This
1725 instruction always performs an unsigned division operation, regardless of
1726 whether the arguments are unsigned or not.</p>
1728 <pre> <result> = udiv uint 4, %var <i>; yields {uint}:result = 4 / %var</i>
1731 <!-- _______________________________________________________________________ -->
1732 <div class="doc_subsubsection"> <a name="i_sdiv">'<tt>sdiv</tt>' Instruction
1734 <div class="doc_text">
1736 <pre> <result> = sdiv <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1739 <p>The '<tt>sdiv</tt>' instruction returns the quotient of its two
1742 <p>The two arguments to the '<tt>sdiv</tt>' instruction must be
1743 <a href="#t_integer">integer</a> values. Both arguments must have identical
1744 types. This instruction can also take <a href="#t_packed">packed</a> versions
1745 of the values in which case the elements must be integers.</p>
1747 <p>The value produced is the signed integer quotient of the two operands. This
1748 instruction always performs a signed division operation, regardless of whether
1749 the arguments are signed or not.</p>
1751 <pre> <result> = sdiv int 4, %var <i>; yields {int}:result = 4 / %var</i>
1754 <!-- _______________________________________________________________________ -->
1755 <div class="doc_subsubsection"> <a name="i_fdiv">'<tt>fdiv</tt>'
1756 Instruction</a> </div>
1757 <div class="doc_text">
1759 <pre> <result> = fdiv <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1762 <p>The '<tt>fdiv</tt>' instruction returns the quotient of its two
1765 <p>The two arguments to the '<tt>div</tt>' instruction must be
1766 <a href="#t_floating">floating point</a> values. Both arguments must have
1767 identical types. This instruction can also take <a href="#t_packed">packed</a>
1768 versions of the values in which case the elements must be floating point.</p>
1770 <p>The value produced is the floating point quotient of the two operands.</p>
1772 <pre> <result> = fdiv float 4.0, %var <i>; yields {float}:result = 4.0 / %var</i>
1775 <!-- _______________________________________________________________________ -->
1776 <div class="doc_subsubsection"> <a name="i_urem">'<tt>urem</tt>' Instruction</a>
1778 <div class="doc_text">
1780 <pre> <result> = urem <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1783 <p>The '<tt>urem</tt>' instruction returns the remainder from the
1784 unsigned division of its two arguments.</p>
1786 <p>The two arguments to the '<tt>urem</tt>' instruction must be
1787 <a href="#t_integer">integer</a> values. Both arguments must have identical
1790 <p>This instruction returns the unsigned integer <i>remainder</i> of a division.
1791 This instruction always performs an unsigned division to get the remainder,
1792 regardless of whether the arguments are unsigned or not.</p>
1794 <pre> <result> = urem uint 4, %var <i>; yields {uint}:result = 4 % %var</i>
1798 <!-- _______________________________________________________________________ -->
1799 <div class="doc_subsubsection"> <a name="i_srem">'<tt>srem</tt>'
1800 Instruction</a> </div>
1801 <div class="doc_text">
1803 <pre> <result> = srem <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1806 <p>The '<tt>srem</tt>' instruction returns the remainder from the
1807 signed division of its two operands.</p>
1809 <p>The two arguments to the '<tt>srem</tt>' instruction must be
1810 <a href="#t_integer">integer</a> values. Both arguments must have identical
1813 <p>This instruction returns the <i>remainder</i> of a division (where the result
1814 has the same sign as the divisor), not the <i>modulus</i> (where the
1815 result has the same sign as the dividend) of a value. For more
1816 information about the difference, see <a
1817 href="http://mathforum.org/dr.math/problems/anne.4.28.99.html">The
1820 <pre> <result> = srem int 4, %var <i>; yields {int}:result = 4 % %var</i>
1824 <!-- _______________________________________________________________________ -->
1825 <div class="doc_subsubsection"> <a name="i_frem">'<tt>frem</tt>'
1826 Instruction</a> </div>
1827 <div class="doc_text">
1829 <pre> <result> = frem <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1832 <p>The '<tt>frem</tt>' instruction returns the remainder from the
1833 division of its two operands.</p>
1835 <p>The two arguments to the '<tt>frem</tt>' instruction must be
1836 <a href="#t_floating">floating point</a> values. Both arguments must have
1837 identical types.</p>
1839 <p>This instruction returns the <i>remainder</i> of a division.</p>
1841 <pre> <result> = frem float 4.0, %var <i>; yields {float}:result = 4.0 % %var</i>
1845 <!-- ======================================================================= -->
1846 <div class="doc_subsection"> <a name="bitwiseops">Bitwise Binary
1847 Operations</a> </div>
1848 <div class="doc_text">
1849 <p>Bitwise binary operators are used to do various forms of
1850 bit-twiddling in a program. They are generally very efficient
1851 instructions and can commonly be strength reduced from other
1852 instructions. They require two operands, execute an operation on them,
1853 and produce a single value. The resulting value of the bitwise binary
1854 operators is always the same type as its first operand.</p>
1856 <!-- _______________________________________________________________________ -->
1857 <div class="doc_subsubsection"> <a name="i_and">'<tt>and</tt>'
1858 Instruction</a> </div>
1859 <div class="doc_text">
1861 <pre> <result> = and <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1864 <p>The '<tt>and</tt>' instruction returns the bitwise logical and of
1865 its two operands.</p>
1867 <p>The two arguments to the '<tt>and</tt>' instruction must be <a
1868 href="#t_integral">integral</a> values. Both arguments must have
1869 identical types.</p>
1871 <p>The truth table used for the '<tt>and</tt>' instruction is:</p>
1873 <div style="align: center">
1874 <table border="1" cellspacing="0" cellpadding="4">
1905 <pre> <result> = and int 4, %var <i>; yields {int}:result = 4 & %var</i>
1906 <result> = and int 15, 40 <i>; yields {int}:result = 8</i>
1907 <result> = and int 4, 8 <i>; yields {int}:result = 0</i>
1910 <!-- _______________________________________________________________________ -->
1911 <div class="doc_subsubsection"> <a name="i_or">'<tt>or</tt>' Instruction</a> </div>
1912 <div class="doc_text">
1914 <pre> <result> = or <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1917 <p>The '<tt>or</tt>' instruction returns the bitwise logical inclusive
1918 or of its two operands.</p>
1920 <p>The two arguments to the '<tt>or</tt>' instruction must be <a
1921 href="#t_integral">integral</a> values. Both arguments must have
1922 identical types.</p>
1924 <p>The truth table used for the '<tt>or</tt>' instruction is:</p>
1926 <div style="align: center">
1927 <table border="1" cellspacing="0" cellpadding="4">
1958 <pre> <result> = or int 4, %var <i>; yields {int}:result = 4 | %var</i>
1959 <result> = or int 15, 40 <i>; yields {int}:result = 47</i>
1960 <result> = or int 4, 8 <i>; yields {int}:result = 12</i>
1963 <!-- _______________________________________________________________________ -->
1964 <div class="doc_subsubsection"> <a name="i_xor">'<tt>xor</tt>'
1965 Instruction</a> </div>
1966 <div class="doc_text">
1968 <pre> <result> = xor <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1971 <p>The '<tt>xor</tt>' instruction returns the bitwise logical exclusive
1972 or of its two operands. The <tt>xor</tt> is used to implement the
1973 "one's complement" operation, which is the "~" operator in C.</p>
1975 <p>The two arguments to the '<tt>xor</tt>' instruction must be <a
1976 href="#t_integral">integral</a> values. Both arguments must have
1977 identical types.</p>
1979 <p>The truth table used for the '<tt>xor</tt>' instruction is:</p>
1981 <div style="align: center">
1982 <table border="1" cellspacing="0" cellpadding="4">
2014 <pre> <result> = xor int 4, %var <i>; yields {int}:result = 4 ^ %var</i>
2015 <result> = xor int 15, 40 <i>; yields {int}:result = 39</i>
2016 <result> = xor int 4, 8 <i>; yields {int}:result = 12</i>
2017 <result> = xor int %V, -1 <i>; yields {int}:result = ~%V</i>
2020 <!-- _______________________________________________________________________ -->
2021 <div class="doc_subsubsection"> <a name="i_shl">'<tt>shl</tt>'
2022 Instruction</a> </div>
2023 <div class="doc_text">
2025 <pre> <result> = shl <ty> <var1>, ubyte <var2> <i>; yields {ty}:result</i>
2028 <p>The '<tt>shl</tt>' instruction returns the first operand shifted to
2029 the left a specified number of bits.</p>
2031 <p>The first argument to the '<tt>shl</tt>' instruction must be an <a
2032 href="#t_integer">integer</a> type. The second argument must be an '<tt>ubyte</tt>'
2035 <p>The value produced is <tt>var1</tt> * 2<sup><tt>var2</tt></sup>.</p>
2037 <pre> <result> = shl int 4, ubyte %var <i>; yields {int}:result = 4 << %var</i>
2038 <result> = shl int 4, ubyte 2 <i>; yields {int}:result = 16</i>
2039 <result> = shl int 1, ubyte 10 <i>; yields {int}:result = 1024</i>
2042 <!-- _______________________________________________________________________ -->
2043 <div class="doc_subsubsection"> <a name="i_lshr">'<tt>lshr</tt>'
2044 Instruction</a> </div>
2045 <div class="doc_text">
2047 <pre> <result> = lshr <ty> <var1>, ubyte <var2> <i>; yields {ty}:result</i>
2051 <p>The '<tt>lshr</tt>' instruction (logical shift right) returns the first
2052 operand shifted to the right a specified number of bits.</p>
2055 <p>The first argument to the '<tt>lshr</tt>' instruction must be an <a
2056 href="#t_integer">integer</a> type. The second argument must be an '<tt>ubyte</tt>' type.</p>
2059 <p>This instruction always performs a logical shift right operation, regardless
2060 of whether the arguments are unsigned or not. The <tt>var2</tt> most significant
2061 bits will be filled with zero bits after the shift.</p>
2065 <result> = lshr uint 4, ubyte 1 <i>; yields {uint}:result = 2</i>
2066 <result> = lshr int 4, ubyte 2 <i>; yields {uint}:result = 1</i>
2067 <result> = lshr sbyte 4, ubyte 3 <i>; yields {sbyte}:result = 0</i>
2068 <result> = lshr sbyte -2, ubyte 1 <i>; yields {sbyte}:result = 0x7FFFFFFF </i>
2072 <!-- ======================================================================= -->
2073 <div class="doc_subsubsection"> <a name="i_ashr">'<tt>ashr</tt>'
2074 Instruction</a> </div>
2075 <div class="doc_text">
2078 <pre> <result> = ashr <ty> <var1>, ubyte <var2> <i>; yields {ty}:result</i>
2082 <p>The '<tt>ashr</tt>' instruction (arithmetic shift right) returns the first
2083 operand shifted to the right a specified number of bits.</p>
2086 <p>The first argument to the '<tt>ashr</tt>' instruction must be an
2087 <a href="#t_integer">integer</a> type. The second argument must be an
2088 '<tt>ubyte</tt>' type.</p>
2091 <p>This instruction always performs an arithmetic shift right operation,
2092 regardless of whether the arguments are signed or not. The <tt>var2</tt> most
2093 significant bits will be filled with the sign bit of <tt>var1</tt>.</p>
2097 <result> = ashr uint 4, ubyte 1 <i>; yields {uint}:result = 2</i>
2098 <result> = ashr int 4, ubyte 2 <i>; yields {int}:result = 1</i>
2099 <result> = ashr ubyte 4, ubyte 3 <i>; yields {ubyte}:result = 0</i>
2100 <result> = ashr sbyte -2, ubyte 1 <i>; yields {sbyte}:result = -1</i>
2104 <!-- ======================================================================= -->
2105 <div class="doc_subsection">
2106 <a name="vectorops">Vector Operations</a>
2109 <div class="doc_text">
2111 <p>LLVM supports several instructions to represent vector operations in a
2112 target-independent manner. This instructions cover the element-access and
2113 vector-specific operations needed to process vectors effectively. While LLVM
2114 does directly support these vector operations, many sophisticated algorithms
2115 will want to use target-specific intrinsics to take full advantage of a specific
2120 <!-- _______________________________________________________________________ -->
2121 <div class="doc_subsubsection">
2122 <a name="i_extractelement">'<tt>extractelement</tt>' Instruction</a>
2125 <div class="doc_text">
2130 <result> = extractelement <n x <ty>> <val>, uint <idx> <i>; yields <ty></i>
2136 The '<tt>extractelement</tt>' instruction extracts a single scalar
2137 element from a packed vector at a specified index.
2144 The first operand of an '<tt>extractelement</tt>' instruction is a
2145 value of <a href="#t_packed">packed</a> type. The second operand is
2146 an index indicating the position from which to extract the element.
2147 The index may be a variable.</p>
2152 The result is a scalar of the same type as the element type of
2153 <tt>val</tt>. Its value is the value at position <tt>idx</tt> of
2154 <tt>val</tt>. If <tt>idx</tt> exceeds the length of <tt>val</tt>, the
2155 results are undefined.
2161 %result = extractelement <4 x int> %vec, uint 0 <i>; yields int</i>
2166 <!-- _______________________________________________________________________ -->
2167 <div class="doc_subsubsection">
2168 <a name="i_insertelement">'<tt>insertelement</tt>' Instruction</a>
2171 <div class="doc_text">
2176 <result> = insertelement <n x <ty>> <val>, <ty> <elt>, uint <idx> <i>; yields <n x <ty>></i>
2182 The '<tt>insertelement</tt>' instruction inserts a scalar
2183 element into a packed vector at a specified index.
2190 The first operand of an '<tt>insertelement</tt>' instruction is a
2191 value of <a href="#t_packed">packed</a> type. The second operand is a
2192 scalar value whose type must equal the element type of the first
2193 operand. The third operand is an index indicating the position at
2194 which to insert the value. The index may be a variable.</p>
2199 The result is a packed vector of the same type as <tt>val</tt>. Its
2200 element values are those of <tt>val</tt> except at position
2201 <tt>idx</tt>, where it gets the value <tt>elt</tt>. If <tt>idx</tt>
2202 exceeds the length of <tt>val</tt>, the results are undefined.
2208 %result = insertelement <4 x int> %vec, int 1, uint 0 <i>; yields <4 x int></i>
2212 <!-- _______________________________________________________________________ -->
2213 <div class="doc_subsubsection">
2214 <a name="i_shufflevector">'<tt>shufflevector</tt>' Instruction</a>
2217 <div class="doc_text">
2222 <result> = shufflevector <n x <ty>> <v1>, <n x <ty>> <v2>, <n x uint> <mask> <i>; yields <n x <ty>></i>
2228 The '<tt>shufflevector</tt>' instruction constructs a permutation of elements
2229 from two input vectors, returning a vector of the same type.
2235 The first two operands of a '<tt>shufflevector</tt>' instruction are vectors
2236 with types that match each other and types that match the result of the
2237 instruction. The third argument is a shuffle mask, which has the same number
2238 of elements as the other vector type, but whose element type is always 'uint'.
2242 The shuffle mask operand is required to be a constant vector with either
2243 constant integer or undef values.
2249 The elements of the two input vectors are numbered from left to right across
2250 both of the vectors. The shuffle mask operand specifies, for each element of
2251 the result vector, which element of the two input registers the result element
2252 gets. The element selector may be undef (meaning "don't care") and the second
2253 operand may be undef if performing a shuffle from only one vector.
2259 %result = shufflevector <4 x int> %v1, <4 x int> %v2,
2260 <4 x uint> <uint 0, uint 4, uint 1, uint 5> <i>; yields <4 x int></i>
2261 %result = shufflevector <4 x int> %v1, <4 x int> undef,
2262 <4 x uint> <uint 0, uint 1, uint 2, uint 3> <i>; yields <4 x int></i> - Identity shuffle.
2267 <!-- ======================================================================= -->
2268 <div class="doc_subsection">
2269 <a name="memoryops">Memory Access and Addressing Operations</a>
2272 <div class="doc_text">
2274 <p>A key design point of an SSA-based representation is how it
2275 represents memory. In LLVM, no memory locations are in SSA form, which
2276 makes things very simple. This section describes how to read, write,
2277 allocate, and free memory in LLVM.</p>
2281 <!-- _______________________________________________________________________ -->
2282 <div class="doc_subsubsection">
2283 <a name="i_malloc">'<tt>malloc</tt>' Instruction</a>
2286 <div class="doc_text">
2291 <result> = malloc <type>[, uint <NumElements>][, align <alignment>] <i>; yields {type*}:result</i>
2296 <p>The '<tt>malloc</tt>' instruction allocates memory from the system
2297 heap and returns a pointer to it.</p>
2301 <p>The '<tt>malloc</tt>' instruction allocates
2302 <tt>sizeof(<type>)*NumElements</tt>
2303 bytes of memory from the operating system and returns a pointer of the
2304 appropriate type to the program. If "NumElements" is specified, it is the
2305 number of elements allocated. If an alignment is specified, the value result
2306 of the allocation is guaranteed to be aligned to at least that boundary. If
2307 not specified, or if zero, the target can choose to align the allocation on any
2308 convenient boundary.</p>
2310 <p>'<tt>type</tt>' must be a sized type.</p>
2314 <p>Memory is allocated using the system "<tt>malloc</tt>" function, and
2315 a pointer is returned.</p>
2320 %array = malloc [4 x ubyte ] <i>; yields {[%4 x ubyte]*}:array</i>
2322 %size = <a href="#i_add">add</a> uint 2, 2 <i>; yields {uint}:size = uint 4</i>
2323 %array1 = malloc ubyte, uint 4 <i>; yields {ubyte*}:array1</i>
2324 %array2 = malloc [12 x ubyte], uint %size <i>; yields {[12 x ubyte]*}:array2</i>
2325 %array3 = malloc int, uint 4, align 1024 <i>; yields {int*}:array3</i>
2326 %array4 = malloc int, align 1024 <i>; yields {int*}:array4</i>
2330 <!-- _______________________________________________________________________ -->
2331 <div class="doc_subsubsection">
2332 <a name="i_free">'<tt>free</tt>' Instruction</a>
2335 <div class="doc_text">
2340 free <type> <value> <i>; yields {void}</i>
2345 <p>The '<tt>free</tt>' instruction returns memory back to the unused
2346 memory heap to be reallocated in the future.</p>
2350 <p>'<tt>value</tt>' shall be a pointer value that points to a value
2351 that was allocated with the '<tt><a href="#i_malloc">malloc</a></tt>'
2356 <p>Access to the memory pointed to by the pointer is no longer defined
2357 after this instruction executes.</p>
2362 %array = <a href="#i_malloc">malloc</a> [4 x ubyte] <i>; yields {[4 x ubyte]*}:array</i>
2363 free [4 x ubyte]* %array
2367 <!-- _______________________________________________________________________ -->
2368 <div class="doc_subsubsection">
2369 <a name="i_alloca">'<tt>alloca</tt>' Instruction</a>
2372 <div class="doc_text">
2377 <result> = alloca <type>[, uint <NumElements>][, align <alignment>] <i>; yields {type*}:result</i>
2382 <p>The '<tt>alloca</tt>' instruction allocates memory on the current
2383 stack frame of the procedure that is live until the current function
2384 returns to its caller.</p>
2388 <p>The '<tt>alloca</tt>' instruction allocates <tt>sizeof(<type>)*NumElements</tt>
2389 bytes of memory on the runtime stack, returning a pointer of the
2390 appropriate type to the program. If "NumElements" is specified, it is the
2391 number of elements allocated. If an alignment is specified, the value result
2392 of the allocation is guaranteed to be aligned to at least that boundary. If
2393 not specified, or if zero, the target can choose to align the allocation on any
2394 convenient boundary.</p>
2396 <p>'<tt>type</tt>' may be any sized type.</p>
2400 <p>Memory is allocated; a pointer is returned. '<tt>alloca</tt>'d
2401 memory is automatically released when the function returns. The '<tt>alloca</tt>'
2402 instruction is commonly used to represent automatic variables that must
2403 have an address available. When the function returns (either with the <tt><a
2404 href="#i_ret">ret</a></tt> or <tt><a href="#i_unwind">unwind</a></tt>
2405 instructions), the memory is reclaimed.</p>
2410 %ptr = alloca int <i>; yields {int*}:ptr</i>
2411 %ptr = alloca int, uint 4 <i>; yields {int*}:ptr</i>
2412 %ptr = alloca int, uint 4, align 1024 <i>; yields {int*}:ptr</i>
2413 %ptr = alloca int, align 1024 <i>; yields {int*}:ptr</i>
2417 <!-- _______________________________________________________________________ -->
2418 <div class="doc_subsubsection"> <a name="i_load">'<tt>load</tt>'
2419 Instruction</a> </div>
2420 <div class="doc_text">
2422 <pre> <result> = load <ty>* <pointer><br> <result> = volatile load <ty>* <pointer><br></pre>
2424 <p>The '<tt>load</tt>' instruction is used to read from memory.</p>
2426 <p>The argument to the '<tt>load</tt>' instruction specifies the memory
2427 address from which to load. The pointer must point to a <a
2428 href="#t_firstclass">first class</a> type. If the <tt>load</tt> is
2429 marked as <tt>volatile</tt>, then the optimizer is not allowed to modify
2430 the number or order of execution of this <tt>load</tt> with other
2431 volatile <tt>load</tt> and <tt><a href="#i_store">store</a></tt>
2434 <p>The location of memory pointed to is loaded.</p>
2436 <pre> %ptr = <a href="#i_alloca">alloca</a> int <i>; yields {int*}:ptr</i>
2438 href="#i_store">store</a> int 3, int* %ptr <i>; yields {void}</i>
2439 %val = load int* %ptr <i>; yields {int}:val = int 3</i>
2442 <!-- _______________________________________________________________________ -->
2443 <div class="doc_subsubsection"> <a name="i_store">'<tt>store</tt>'
2444 Instruction</a> </div>
2445 <div class="doc_text">
2447 <pre> store <ty> <value>, <ty>* <pointer> <i>; yields {void}</i>
2448 volatile store <ty> <value>, <ty>* <pointer> <i>; yields {void}</i>
2451 <p>The '<tt>store</tt>' instruction is used to write to memory.</p>
2453 <p>There are two arguments to the '<tt>store</tt>' instruction: a value
2454 to store and an address in which to store it. The type of the '<tt><pointer></tt>'
2455 operand must be a pointer to the type of the '<tt><value></tt>'
2456 operand. If the <tt>store</tt> is marked as <tt>volatile</tt>, then the
2457 optimizer is not allowed to modify the number or order of execution of
2458 this <tt>store</tt> with other volatile <tt>load</tt> and <tt><a
2459 href="#i_store">store</a></tt> instructions.</p>
2461 <p>The contents of memory are updated to contain '<tt><value></tt>'
2462 at the location specified by the '<tt><pointer></tt>' operand.</p>
2464 <pre> %ptr = <a href="#i_alloca">alloca</a> int <i>; yields {int*}:ptr</i>
2466 href="#i_store">store</a> int 3, int* %ptr <i>; yields {void}</i>
2467 %val = load int* %ptr <i>; yields {int}:val = int 3</i>
2471 <!-- _______________________________________________________________________ -->
2472 <div class="doc_subsubsection">
2473 <a name="i_getelementptr">'<tt>getelementptr</tt>' Instruction</a>
2476 <div class="doc_text">
2479 <result> = getelementptr <ty>* <ptrval>{, <ty> <idx>}*
2485 The '<tt>getelementptr</tt>' instruction is used to get the address of a
2486 subelement of an aggregate data structure.</p>
2490 <p>This instruction takes a list of integer constants that indicate what
2491 elements of the aggregate object to index to. The actual types of the arguments
2492 provided depend on the type of the first pointer argument. The
2493 '<tt>getelementptr</tt>' instruction is used to index down through the type
2494 levels of a structure or to a specific index in an array. When indexing into a
2495 structure, only <tt>uint</tt>
2496 integer constants are allowed. When indexing into an array or pointer,
2497 <tt>int</tt> and <tt>long</tt> and <tt>ulong</tt> indexes are allowed.</p>
2499 <p>For example, let's consider a C code fragment and how it gets
2500 compiled to LLVM:</p>
2514 int *foo(struct ST *s) {
2515 return &s[1].Z.B[5][13];
2519 <p>The LLVM code generated by the GCC frontend is:</p>
2522 %RT = type { sbyte, [10 x [20 x int]], sbyte }
2523 %ST = type { int, double, %RT }
2527 int* %foo(%ST* %s) {
2529 %reg = getelementptr %ST* %s, int 1, uint 2, uint 1, int 5, int 13
2536 <p>The index types specified for the '<tt>getelementptr</tt>' instruction depend
2537 on the pointer type that is being indexed into. <a href="#t_pointer">Pointer</a>
2538 and <a href="#t_array">array</a> types require <tt>int</tt>,
2539 <tt>ulong</tt>, or <tt>long</tt> values, and <a href="#t_struct">structure</a>
2540 types require <tt>uint</tt> <b>constants</b>.</p>
2542 <p>In the example above, the first index is indexing into the '<tt>%ST*</tt>'
2543 type, which is a pointer, yielding a '<tt>%ST</tt>' = '<tt>{ int, double, %RT
2544 }</tt>' type, a structure. The second index indexes into the third element of
2545 the structure, yielding a '<tt>%RT</tt>' = '<tt>{ sbyte, [10 x [20 x int]],
2546 sbyte }</tt>' type, another structure. The third index indexes into the second
2547 element of the structure, yielding a '<tt>[10 x [20 x int]]</tt>' type, an
2548 array. The two dimensions of the array are subscripted into, yielding an
2549 '<tt>int</tt>' type. The '<tt>getelementptr</tt>' instruction returns a pointer
2550 to this element, thus computing a value of '<tt>int*</tt>' type.</p>
2552 <p>Note that it is perfectly legal to index partially through a
2553 structure, returning a pointer to an inner element. Because of this,
2554 the LLVM code for the given testcase is equivalent to:</p>
2557 int* %foo(%ST* %s) {
2558 %t1 = getelementptr %ST* %s, int 1 <i>; yields %ST*:%t1</i>
2559 %t2 = getelementptr %ST* %t1, int 0, uint 2 <i>; yields %RT*:%t2</i>
2560 %t3 = getelementptr %RT* %t2, int 0, uint 1 <i>; yields [10 x [20 x int]]*:%t3</i>
2561 %t4 = getelementptr [10 x [20 x int]]* %t3, int 0, int 5 <i>; yields [20 x int]*:%t4</i>
2562 %t5 = getelementptr [20 x int]* %t4, int 0, int 13 <i>; yields int*:%t5</i>
2567 <p>Note that it is undefined to access an array out of bounds: array and
2568 pointer indexes must always be within the defined bounds of the array type.
2569 The one exception for this rules is zero length arrays. These arrays are
2570 defined to be accessible as variable length arrays, which requires access
2571 beyond the zero'th element.</p>
2573 <p>The getelementptr instruction is often confusing. For some more insight
2574 into how it works, see <a href="GetElementPtr.html">the getelementptr
2580 <i>; yields [12 x ubyte]*:aptr</i>
2581 %aptr = getelementptr {int, [12 x ubyte]}* %sptr, long 0, uint 1
2585 <!-- ======================================================================= -->
2586 <div class="doc_subsection"> <a name="convertops">Conversion Operations</a>
2588 <div class="doc_text">
2589 <p>The instructions in this category are the conversion instructions (casting)
2590 which all take a single operand and a type. They perform various bit conversions
2594 <!-- _______________________________________________________________________ -->
2595 <div class="doc_subsubsection">
2596 <a name="i_trunc">'<tt>trunc .. to</tt>' Instruction</a>
2598 <div class="doc_text">
2602 <result> = trunc <ty> <value> to <ty2> <i>; yields ty2</i>
2607 The '<tt>trunc</tt>' instruction truncates its operand to the type <tt>ty2</tt>.
2612 The '<tt>trunc</tt>' instruction takes a <tt>value</tt> to trunc, which must
2613 be an <a href="#t_integer">integer</a> type, and a type that specifies the size
2614 and type of the result, which must be an <a href="#t_integral">integral</a>
2615 type. The bit size of <tt>value</tt> must be larger than the bit size of
2616 <tt>ty2</tt>. Equal sized types are not allowed.</p>
2620 The '<tt>trunc</tt>' instruction truncates the high order bits in <tt>value</tt>
2621 and converts the remaining bits to <tt>ty2</tt>. Since the source size must be
2622 larger than the destination size, <tt>trunc</tt> cannot be a <i>no-op cast</i>.
2623 It will always truncate bits.</p>
2627 %X = trunc int 257 to ubyte <i>; yields ubyte:1</i>
2628 %Y = trunc int 123 to bool <i>; yields bool:true</i>
2632 <!-- _______________________________________________________________________ -->
2633 <div class="doc_subsubsection">
2634 <a name="i_zext">'<tt>zext .. to</tt>' Instruction</a>
2636 <div class="doc_text">
2640 <result> = zext <ty> <value> to <ty2> <i>; yields ty2</i>
2644 <p>The '<tt>zext</tt>' instruction zero extends its operand to type
2649 <p>The '<tt>zext</tt>' instruction takes a value to cast, which must be of
2650 <a href="#t_integral">integral</a> type, and a type to cast it to, which must
2651 also be of <a href="#t_integral">integral</a> type. The bit size of the
2652 <tt>value</tt> must be smaller than the bit size of the destination type,
2656 <p>The <tt>zext</tt> fills the high order bits of the <tt>value</tt> with zero
2657 bits until it reaches the size of the destination type, <tt>ty2</tt>. When the
2658 the operand and the type are the same size, no bit filling is done and the
2659 cast is considered a <i>no-op cast</i> because no bits change (only the type
2662 <p>When zero extending from bool, the result will alwasy be either 0 or 1.</p>
2666 %X = zext int 257 to ulong <i>; yields ulong:257</i>
2667 %Y = zext bool true to int <i>; yields int:1</i>
2671 <!-- _______________________________________________________________________ -->
2672 <div class="doc_subsubsection">
2673 <a name="i_sext">'<tt>sext .. to</tt>' Instruction</a>
2675 <div class="doc_text">
2679 <result> = sext <ty> <value> to <ty2> <i>; yields ty2</i>
2683 <p>The '<tt>sext</tt>' sign extends <tt>value</tt> to the type <tt>ty2</tt>.</p>
2687 The '<tt>sext</tt>' instruction takes a value to cast, which must be of
2688 <a href="#t_integral">integral</a> type, and a type to cast it to, which must
2689 also be of <a href="#t_integral">integral</a> type. The bit size of the
2690 <tt>value</tt> must be smaller than the bit size of the destination type,
2695 The '<tt>sext</tt>' instruction performs a sign extension by copying the sign
2696 bit (highest order bit) of the <tt>value</tt> until it reaches the bit size of
2697 the type <tt>ty2</tt>. When the the operand and the type are the same size,
2698 no bit filling is done and the cast is considered a <i>no-op cast</i> because
2699 no bits change (only the type changes).</p>
2701 <p>When sign extending from bool, the extension always results in -1 or 0.</p>
2705 %X = sext sbyte -1 to ushort <i>; yields ushort:65535</i>
2706 %Y = sext bool true to int <i>; yields int:-1</i>
2710 <!-- _______________________________________________________________________ -->
2711 <div class="doc_subsubsection">
2712 <a name="i_fptrunc">'<tt>fptrunc .. to</tt>' Instruction</a>
2715 <div class="doc_text">
2720 <result> = fptrunc <ty> <value> to <ty2> <i>; yields ty2</i>
2724 <p>The '<tt>fptrunc</tt>' instruction truncates <tt>value</tt> to type
2729 <p>The '<tt>fptrunc</tt>' instruction takes a <a href="#t_floating">floating
2730 point</a> value to cast and a <a href="#t_floating">floating point</a> type to
2731 cast it to. The size of <tt>value</tt> must be larger than the size of
2732 <tt>ty2</tt>. This implies that <tt>fptrunc</tt> cannot be used to make a
2733 <i>no-op cast</i>.</p>
2736 <p> The '<tt>fptrunc</tt>' instruction truncates a <tt>value</tt> from a larger
2737 <a href="#t_floating">floating point</a> type to a smaller
2738 <a href="#t_floating">floating point</a> type. If the value cannot fit within
2739 the destination type, <tt>ty2</tt>, then the results are undefined.</p>
2743 %X = fptrunc double 123.0 to float <i>; yields float:123.0</i>
2744 %Y = fptrunc double 1.0E+300 to float <i>; yields undefined</i>
2748 <!-- _______________________________________________________________________ -->
2749 <div class="doc_subsubsection">
2750 <a name="i_fpext">'<tt>fpext .. to</tt>' Instruction</a>
2752 <div class="doc_text">
2756 <result> = fpext <ty> <value> to <ty2> <i>; yields ty2</i>
2760 <p>The '<tt>fpext</tt>' extends a floating point <tt>value</tt> to a larger
2761 floating point value.</p>
2764 <p>The '<tt>fpext</tt>' instruction takes a
2765 <a href="#t_floating">floating point</a> <tt>value</tt> to cast,
2766 and a <a href="#t_floating">floating point</a> type to cast it to. The source
2767 type must be smaller than the destination type.</p>
2770 <p>The '<tt>fpext</tt>' instruction extends the <tt>value</tt> from a smaller
2771 <a href="t_floating">floating point</a> type to a larger
2772 <a href="t_floating">floating point</a> type. The <tt>fpext</tt> cannot be
2773 used to make a <i>no-op cast</i> because it always changes bits. Use
2774 <tt>bitcast</tt> to make a <i>no-op cast</i> for a floating point cast.</p>
2778 %X = fpext float 3.1415 to double <i>; yields double:3.1415</i>
2779 %Y = fpext float 1.0 to float <i>; yields float:1.0 (no-op)</i>
2783 <!-- _______________________________________________________________________ -->
2784 <div class="doc_subsubsection">
2785 <a name="i_fp2uint">'<tt>fptoui .. to</tt>' Instruction</a>
2787 <div class="doc_text">
2791 <result> = fp2uint <ty> <value> to <ty2> <i>; yields ty2</i>
2795 <p>The '<tt>fp2uint</tt>' converts a floating point <tt>value</tt> to its
2796 unsigned integer equivalent of type <tt>ty2</tt>.
2800 <p>The '<tt>fp2uint</tt>' instruction takes a value to cast, which must be a
2801 <a href="#t_floating">floating point</a> value, and a type to cast it to, which
2802 must be an <a href="#t_integral">integral</a> type.</p>
2805 <p> The '<tt>fp2uint</tt>' instruction converts its
2806 <a href="#t_floating">floating point</a> operand into the nearest (rounding
2807 towards zero) unsigned integer value. If the value cannot fit in <tt>ty2</tt>,
2808 the results are undefined.</p>
2810 <p>When converting to bool, the conversion is done as a comparison against
2811 zero. If the <tt>value</tt> was zero, the bool result will be <tt>false</tt>.
2812 If the <tt>value</tt> was non-zero, the bool result will be <tt>true</tt>.</p>
2816 %X = fp2uint double 123.0 to int <i>; yields int:123</i>
2817 %Y = fp2uint float 1.0E+300 to bool <i>; yields bool:true</i>
2818 %X = fp2uint float 1.04E+17 to ubyte <i>; yields undefined:1</i>
2822 <!-- _______________________________________________________________________ -->
2823 <div class="doc_subsubsection">
2824 <a name="i_fptosi">'<tt>fptosi .. to</tt>' Instruction</a>
2826 <div class="doc_text">
2830 <result> = fptosi <ty> <value> to <ty2> <i>; yields ty2</i>
2834 <p>The '<tt>fptosi</tt>' instruction converts
2835 <a href="#t_floating">floating point</a> <tt>value</tt> to type <tt>ty2</tt>.
2840 <p> The '<tt>fptosi</tt>' instruction takes a value to cast, which must be a
2841 <a href="#t_floating">floating point</a> value, and a type to cast it to, which
2842 must also be an <a href="#t_integral">integral</a> type.</p>
2845 <p>The '<tt>fptosi</tt>' instruction converts its
2846 <a href="#t_floating">floating point</a> operand into the nearest (rounding
2847 towards zero) signed integer value. If the value cannot fit in <tt>ty2</tt>,
2848 the results are undefined.</p>
2850 <p>When converting to bool, the conversion is done as a comparison against
2851 zero. If the <tt>value</tt> was zero, the bool result will be <tt>false</tt>.
2852 If the <tt>value</tt> was non-zero, the bool result will be <tt>true</tt>.</p>
2856 %X = fptosi double -123.0 to int <i>; yields int:-123</i>
2857 %Y = fptosi float 1.0E-247 to bool <i>; yields bool:true</i>
2858 %X = fptosi float 1.04E+17 to sbyte <i>; yields undefined:1</i>
2862 <!-- _______________________________________________________________________ -->
2863 <div class="doc_subsubsection">
2864 <a name="i_uitofp">'<tt>uitofp .. to</tt>' Instruction</a>
2866 <div class="doc_text">
2870 <result> = uitofp <ty> <value> to <ty2> <i>; yields ty2</i>
2874 <p>The '<tt>uitofp</tt>' instruction regards <tt>value</tt> as an unsigned
2875 integer and converts that value to the <tt>ty2</tt> type.</p>
2879 <p>The '<tt>uitofp</tt>' instruction takes a value to cast, which must be an
2880 <a href="#t_integral">integral</a> value, and a type to cast it to, which must
2881 be a <a href="#t_floating">floating point</a> type.</p>
2884 <p>The '<tt>uitofp</tt>' instruction interprets its operand as an unsigned
2885 integer quantity and converts it to the corresponding floating point value. If
2886 the value cannot fit in the floating point value, the results are undefined.</p>
2891 %X = uitofp int 257 to float <i>; yields float:257.0</i>
2892 %Y = uitofp sbyte -1 to double <i>; yields double:255.0</i>
2896 <!-- _______________________________________________________________________ -->
2897 <div class="doc_subsubsection">
2898 <a name="i_sitofp">'<tt>sitofp .. to</tt>' Instruction</a>
2900 <div class="doc_text">
2904 <result> = sitofp <ty> <value> to <ty2> <i>; yields ty2</i>
2908 <p>The '<tt>sitofp</tt>' instruction regards <tt>value</tt> as a signed
2909 integer and converts that value to the <tt>ty2</tt> type.</p>
2912 <p>The '<tt>sitofp</tt>' instruction takes a value to cast, which must be an
2913 <a href="#t_integral">integral</a> value, and a type to cast it to, which must be
2914 a <a href="#t_floating">floating point</a> type.</p>
2917 <p>The '<tt>sitofp</tt>' instruction interprets its operand as a signed
2918 integer quantity and converts it to the corresponding floating point value. If
2919 the value cannot fit in the floating point value, the results are undefined.</p>
2923 %X = sitofp int 257 to float <i>; yields float:257.0</i>
2924 %Y = sitofp sbyte -1 to double <i>; yields double:-1.0</i>
2928 <!-- _______________________________________________________________________ -->
2929 <div class="doc_subsubsection">
2930 <a name="i_ptrtoint">'<tt>ptrtoint .. to</tt>' Instruction</a>
2932 <div class="doc_text">
2936 <result> = ptrtoint <ty> <value> to <ty2> <i>; yields ty2</i>
2940 <p>The '<tt>ptrtoint</tt>' instruction converts the pointer <tt>value</tt> to
2941 the integer type <tt>ty2</tt>.</p>
2944 <p>The '<tt>ptrtoint</tt>' instruction takes a <tt>value</tt> to cast, which
2945 must be a <a href="t_pointer">pointer</a> value, and a type to cast it to
2946 <tt>ty2</tt>, which must be an <a href="#t_integer">integer</a> type.
2949 <p>The '<tt>ptrtoint</tt>' instruction converts <tt>value</tt> to integer type
2950 <tt>ty2</tt> by interpreting the pointer value as an integer and either
2951 truncating or zero extending that value to the size of the integer type. If
2952 <tt>value</tt> is smaller than <tt>ty2</tt> then a zero extension is done. If
2953 <tt>value</tt> is larger than <tt>ty2</tt> then a truncation is done. If they
2954 are the same size, then nothing is done (<i>no-op cast</i>).</p>
2958 %X = ptrtoint int* %X to sbyte <i>; yields truncation on 32-bit</i>
2959 %Y = ptrtoint int* %x to ulong <i>; yields zero extend on 32-bit</i>
2963 <!-- _______________________________________________________________________ -->
2964 <div class="doc_subsubsection">
2965 <a name="i_inttoptr">'<tt>inttoptr .. to</tt>' Instruction</a>
2967 <div class="doc_text">
2971 <result> = inttoptr <ty> <value> to <ty2> <i>; yields ty2</i>
2975 <p>The '<tt>inttoptr</tt>' instruction converts an integer <tt>value</tt> to
2976 a pointer type, <tt>ty2</tt>.</p>
2979 <p>The '<tt>inttoptr</tt>' instruction takes an <a href="i_integer">integer</a>
2980 value to cast, and a type to cast it to, which must be a
2981 <a href="#t_pointer">pointer</a> type. </tt>
2984 <p>The '<tt>inttoptr</tt>' instruction converts <tt>value</tt> to type
2985 <tt>ty2</tt> by applying either a zero extension or a truncation depending on
2986 the size of the integer <tt>value</tt>. If <tt>value</tt> is larger than the
2987 size of a pointer then a truncation is done. If <tt>value</tt> is smaller than
2988 the size of a pointer then a zero extension is done. If they are the same size,
2989 nothing is done (<i>no-op cast</i>).</p>
2993 %X = inttoptr int 255 to int* <i>; yields zero extend on 64-bit</i>
2994 %X = inttoptr int 255 to int* <i>; yields no-op on 32-bit </i>
2995 %Y = inttoptr short 0 to int* <i>; yields zero extend on 32-bit</i>
2999 <!-- _______________________________________________________________________ -->
3000 <div class="doc_subsubsection">
3001 <a name="i_bitcast">'<tt>bitcast .. to</tt>' Instruction</a>
3003 <div class="doc_text">
3007 <result> = bitcast <ty> <value> to <ty2> <i>; yields ty2</i>
3011 <p>The '<tt>bitcast</tt>' instruction converts <tt>value</tt> to type
3012 <tt>ty2</tt> without changing any bits.</p>
3015 <p>The '<tt>bitcast</tt>' instruction takes a value to cast, which must be
3016 a first class value, and a type to cast it to, which must also be a <a
3017 href="#t_firstclass">first class</a> type. The bit sizes of <tt>value</tt>
3018 and the destination type, <tt>ty2</tt>, must be identical.</p>
3021 <p>The '<tt>bitcast</tt>' instruction converts <tt>value</tt> to type
3022 <tt>ty2</tt>. It is always a <i>no-op cast</i> because no bits change with
3023 this conversion. The conversion is done as if the <tt>value</tt> had been
3024 stored to memory and read back as type <tt>ty2</tt>. Pointer types may only be
3025 converted to other pointer types with this instruction. To convert pointers to
3026 other types, use the <a href="#i_inttoptr">inttoptr</a> or
3027 <a href="#i_ptrtoint">ptrtoint</a> instructions first.</p>
3031 %X = bitcast ubyte 255 to sbyte <i>; yields sbyte:-1</i>
3032 %Y = bitcast uint* %x to sint* <i>; yields sint*:%x</i>
3033 %Z = bitcast <2xint> %V to long; <i>; yields long: %V</i>
3037 <!-- ======================================================================= -->
3038 <div class="doc_subsection"> <a name="otherops">Other Operations</a> </div>
3039 <div class="doc_text">
3040 <p>The instructions in this category are the "miscellaneous"
3041 instructions, which defy better classification.</p>
3044 <!-- _______________________________________________________________________ -->
3045 <div class="doc_subsubsection"><a name="i_icmp">'<tt>icmp</tt>' Instruction</a>
3047 <div class="doc_text">
3049 <pre> <result> = icmp <cond> <ty> <var1>, <var2> <i>; yields {bool}:result</i>
3052 <p>The '<tt>icmp</tt>' instruction returns a boolean value based on comparison
3053 of its two integer operands.</p>
3055 <p>The '<tt>icmp</tt>' instruction takes three operands. The first operand is
3056 the condition code which indicates the kind of comparison to perform. It is not
3057 a value, just a keyword. The possibilities for the condition code are:
3059 <li><tt>eq</tt>: equal</li>
3060 <li><tt>ne</tt>: not equal </li>
3061 <li><tt>ugt</tt>: unsigned greater than</li>
3062 <li><tt>uge</tt>: unsigned greater or equal</li>
3063 <li><tt>ult</tt>: unsigned less than</li>
3064 <li><tt>ule</tt>: unsigned less or equal</li>
3065 <li><tt>sgt</tt>: signed greater than</li>
3066 <li><tt>sge</tt>: signed greater or equal</li>
3067 <li><tt>slt</tt>: signed less than</li>
3068 <li><tt>sle</tt>: signed less or equal</li>
3070 <p>The remaining two arguments must be of <a href="#t_integral">integral</a>,
3071 <a href="#t_pointer">pointer</a> or a <a href="#t_packed">packed</a> integral
3072 type. They must have identical types.</p>
3074 <p>The '<tt>icmp</tt>' compares <tt>var1</tt> and <tt>var2</tt> according to
3075 the condition code given as <tt>cond</tt>. The comparison performed always
3076 yields a <a href="#t_bool">bool</a> result, as follows:
3078 <li><tt>eq</tt>: yields <tt>true</tt> if the operands are equal,
3079 <tt>false</tt> otherwise. No sign interpretation is necessary or performed.
3081 <li><tt>ne</tt>: yields <tt>true</tt> if the operands are unequal,
3082 <tt>false</tt> otherwise. No sign interpretation is necessary or performed.
3083 <li><tt>ugt</tt>: interprets the operands as unsigned values and yields
3084 <tt>true</tt> if <tt>var1</tt> is greater than <tt>var2</tt>.</li>
3085 <li><tt>uge</tt>: interprets the operands as unsigned values and yields
3086 <tt>true</tt> if <tt>var1</tt> is greater than or equal to <tt>var2</tt>.</li>
3087 <li><tt>ult</tt>: interprets the operands as unsigned values and yields
3088 <tt>true</tt> if <tt>var1</tt> is less than <tt>var2</tt>.</li>
3089 <li><tt>ule</tt>: interprets the operands as unsigned values and yields
3090 <tt>true</tt> if <tt>var1</tt> is less than or equal to <tt>var2</tt>.</li>
3091 <li><tt>sgt</tt>: interprets the operands as signed values and yields
3092 <tt>true</tt> if <tt>var1</tt> is greater than <tt>var2</tt>.</li>
3093 <li><tt>sge</tt>: interprets the operands as signed values and yields
3094 <tt>true</tt> if <tt>var1</tt> is greater than or equal to <tt>var2</tt>.</li>
3095 <li><tt>slt</tt>: interprets the operands as signed values and yields
3096 <tt>true</tt> if <tt>var1</tt> is less than <tt>var2</tt>.</li>
3097 <li><tt>sle</tt>: interprets the operands as signed values and yields
3098 <tt>true</tt> if <tt>var1</tt> is less than or equal to <tt>var2</tt>.</li>
3101 <p>If the operands are <a href="#t_pointer">pointer</a> typed, the pointer
3102 values are treated as integers and then compared.</p>
3103 <p>If the operands are <a href="#t_packed">packed</a> typed, the elements of
3104 the vector are compared in turn and the predicate must hold for all elements.
3105 While this is of dubious use for predicates other than <tt>eq</tt> and
3106 <tt>ne</tt>, the other predicates can be used with packed types.</p>
3109 <pre> <result> = icmp eq int 4, 5 <i>; yields: result=false</i>
3110 <result> = icmp ne float* %X, %X <i>; yields: result=false</i>
3111 <result> = icmp ult short 4, 5 <i>; yields: result=true</i>
3112 <result> = icmp sgt sbyte 4, 5 <i>; yields: result=false</i>
3113 <result> = icmp ule sbyte -4, 5 <i>; yields: result=false</i>
3114 <result> = icmp sge sbyte 4, 5 <i>; yields: result=false</i>
3118 <!-- _______________________________________________________________________ -->
3119 <div class="doc_subsubsection"><a name="i_fcmp">'<tt>fcmp</tt>' Instruction</a>
3121 <div class="doc_text">
3123 <pre> <result> = fcmp <cond> <ty> <var1>, <var2> <i>; yields {bool}:result</i>
3126 <p>The '<tt>fcmp</tt>' instruction returns a boolean value based on comparison
3127 of its floating point operands.</p>
3129 <p>The '<tt>fcmp</tt>' instruction takes three operands. The first operand is
3130 the condition code which indicates the kind of comparison to perform. It is not
3131 a value, just a keyword. The possibilities for the condition code are:
3133 <li><tt>false</tt>: no comparison, always false (always folded)</li>
3134 <li><tt>oeq</tt>: ordered and equal</li>
3135 <li><tt>ogt</tt>: ordered and greater than </li>
3136 <li><tt>oge</tt>: ordered and greater than or equal</li>
3137 <li><tt>olt</tt>: ordered and less than </li>
3138 <li><tt>ole</tt>: ordered and less than or equal</li>
3139 <li><tt>one</tt>: ordered and not equal</li>
3140 <li><tt>ord</tt>: ordered (no nans)</li>
3141 <li><tt>ueq</tt>: unordered or equal</li>
3142 <li><tt>ugt</tt>: unordered or greater than </li>
3143 <li><tt>uge</tt>: unordered or greater than or equal</li>
3144 <li><tt>ult</tt>: unordered or less than </li>
3145 <li><tt>ule</tt>: unordered or less than or equal</li>
3146 <li><tt>une</tt>: unordered or not equal</li>
3147 <li><tt>uno</tt>: unordered (either nans)</li>
3148 <li><tt>true</tt>: no comparison, always true (always folded)</li>
3150 <p>The <tt>val1</tt> and <tt>val2</tt> arguments must be of
3151 <a href="#t_floating">floating point</a>, or a <a href="#t_packed">packed</a>
3152 floating point type. They must have identical types.</p>
3154 <p>The '<tt>fcmp</tt>' compares <tt>var1</tt> and <tt>var2</tt> according to
3155 the condition code given as <tt>cond</tt>. The comparison performed always
3156 yields a <a href="#t_bool">bool</a> result, as follows:
3158 <li><tt>false</tt>: always yields <tt>false</tt>, regardless of operands.</li>
3159 <li><tt>oeq</tt>: yields <tt>true</tt> if both operands are ordered and
3160 <tt>var1</tt> is equal to <tt>var2</tt>.</li>
3161 <li><tt>ogt</tt>: yields <tt>true</tt> if both operands are ordered and
3162 <tt>var1</tt> is greather than <tt>var2</tt>.</li>
3163 <li><tt>oge</tt>: yields <tt>true</tt> if both operands are ordered and
3164 <tt>var1</tt> is greater than or equal to <tt>var2</tt>.</li>
3165 <li><tt>olt</tt>: yields <tt>true</tt> if both operands are ordered and
3166 <tt>var1</tt> is less than <tt>var2</tt>.</li>
3167 <li><tt>ole</tt>: yields <tt>true</tt> if both operands are ordered and
3168 <tt>var1</tt> is less than or equal to <tt>var2</tt>.</li>
3169 <li><tt>one</tt>: yields <tt>true</tt> if both operands are ordered and
3170 <tt>var1</tt> is not equal to <tt>var2</tt>.</li>
3171 <li><tt>ord</tt>: yields <tt>true</tt> if both operands are ordered.</li>
3172 <li><tt>ueq</tt>: yields <tt>true</tt> if either operand is unordered or
3173 <tt>var1</tt> is equal to <tt>var2</tt>.</li>
3174 <li><tt>ugt</tt>: yields <tt>true</tt> if either operand is unordered or
3175 <tt>var1</tt> is greater than <tt>var2</tt>.</li>
3176 <li><tt>uge</tt>: yields <tt>true</tt> if either operand is unordered or
3177 <tt>var1</tt> is greater than or equal to <tt>var2</tt>.</li>
3178 <li><tt>ult</tt>: yields <tt>true</tt> if either operand is unordered or
3179 <tt>var1</tt> is less than <tt>var2</tt>.</li>
3180 <li><tt>ule</tt>: yields <tt>true</tt> if either operand is unordered or
3181 <tt>var1</tt> is less than or equal to <tt>var2</tt>.</li>
3182 <li><tt>une</tt>: yields <tt>true</tt> if either operand is unordered or
3183 <tt>var1</tt> is not equal to <tt>var2</tt>.</li>
3184 <li><tt>uno</tt>: yields <tt>true</tt> if either operand is unordered.</li>
3185 <li><tt>true</tt>: always yields <tt>true</tt>, regardless of operands.</li>
3187 <p>If the operands are <a href="#t_packed">packed</a> typed, the elements of
3188 the vector are compared in turn and the predicate must hold for all elements.
3189 While this is of dubious use for predicates other than <tt>eq</tt> and
3190 <tt>ne</tt>, the other predicates can be used with packed types.</p>
3193 <pre> <result> = fcmp oeq float 4.0, 5.0 <i>; yields: result=false</i>
3194 <result> = icmp one float 4.0, 5.0 <i>; yields: result=true</i>
3195 <result> = icmp olt float 4.0, 5.0 <i>; yields: result=true</i>
3196 <result> = icmp ueq double 1.0, 2.0 <i>; yields: result=false</i>
3200 <!-- _______________________________________________________________________ -->
3201 <div class="doc_subsubsection"> <a name="i_phi">'<tt>phi</tt>'
3202 Instruction</a> </div>
3203 <div class="doc_text">
3205 <pre> <result> = phi <ty> [ <val0>, <label0>], ...<br></pre>
3207 <p>The '<tt>phi</tt>' instruction is used to implement the φ node in
3208 the SSA graph representing the function.</p>
3210 <p>The type of the incoming values are specified with the first type
3211 field. After this, the '<tt>phi</tt>' instruction takes a list of pairs
3212 as arguments, with one pair for each predecessor basic block of the
3213 current block. Only values of <a href="#t_firstclass">first class</a>
3214 type may be used as the value arguments to the PHI node. Only labels
3215 may be used as the label arguments.</p>
3216 <p>There must be no non-phi instructions between the start of a basic
3217 block and the PHI instructions: i.e. PHI instructions must be first in
3220 <p>At runtime, the '<tt>phi</tt>' instruction logically takes on the
3221 value specified by the parameter, depending on which basic block we
3222 came from in the last <a href="#terminators">terminator</a> instruction.</p>
3224 <pre>Loop: ; Infinite loop that counts from 0 on up...<br> %indvar = phi uint [ 0, %LoopHeader ], [ %nextindvar, %Loop ]<br> %nextindvar = add uint %indvar, 1<br> br label %Loop<br></pre>
3227 <!-- _______________________________________________________________________ -->
3228 <div class="doc_subsubsection">
3229 <a name="i_select">'<tt>select</tt>' Instruction</a>
3232 <div class="doc_text">
3237 <result> = select bool <cond>, <ty> <val1>, <ty> <val2> <i>; yields ty</i>
3243 The '<tt>select</tt>' instruction is used to choose one value based on a
3244 condition, without branching.
3251 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.
3257 If the boolean condition evaluates to true, the instruction returns the first
3258 value argument; otherwise, it returns the second value argument.
3264 %X = select bool true, ubyte 17, ubyte 42 <i>; yields ubyte:17</i>
3269 <!-- _______________________________________________________________________ -->
3270 <div class="doc_subsubsection">
3271 <a name="i_call">'<tt>call</tt>' Instruction</a>
3274 <div class="doc_text">
3278 <result> = [tail] call [<a href="#callingconv">cconv</a>] <ty>* <fnptrval>(<param list>)
3283 <p>The '<tt>call</tt>' instruction represents a simple function call.</p>
3287 <p>This instruction requires several arguments:</p>
3291 <p>The optional "tail" marker indicates whether the callee function accesses
3292 any allocas or varargs in the caller. If the "tail" marker is present, the
3293 function call is eligible for tail call optimization. Note that calls may
3294 be marked "tail" even if they do not occur before a <a
3295 href="#i_ret"><tt>ret</tt></a> instruction.
3298 <p>The optional "cconv" marker indicates which <a href="callingconv">calling
3299 convention</a> the call should use. If none is specified, the call defaults
3300 to using C calling conventions.
3303 <p>'<tt>ty</tt>': shall be the signature of the pointer to function value
3304 being invoked. The argument types must match the types implied by this
3305 signature. This type can be omitted if the function is not varargs and
3306 if the function type does not return a pointer to a function.</p>
3309 <p>'<tt>fnptrval</tt>': An LLVM value containing a pointer to a function to
3310 be invoked. In most cases, this is a direct function invocation, but
3311 indirect <tt>call</tt>s are just as possible, calling an arbitrary pointer
3312 to function value.</p>
3315 <p>'<tt>function args</tt>': argument list whose types match the
3316 function signature argument types. All arguments must be of
3317 <a href="#t_firstclass">first class</a> type. If the function signature
3318 indicates the function accepts a variable number of arguments, the extra
3319 arguments can be specified.</p>
3325 <p>The '<tt>call</tt>' instruction is used to cause control flow to
3326 transfer to a specified function, with its incoming arguments bound to
3327 the specified values. Upon a '<tt><a href="#i_ret">ret</a></tt>'
3328 instruction in the called function, control flow continues with the
3329 instruction after the function call, and the return value of the
3330 function is bound to the result argument. This is a simpler case of
3331 the <a href="#i_invoke">invoke</a> instruction.</p>
3336 %retval = call int %test(int %argc)
3337 call int(sbyte*, ...) *%printf(sbyte* %msg, int 12, sbyte 42);
3338 %X = tail call int %foo()
3339 %Y = tail call <a href="#callingconv">fastcc</a> int %foo()
3344 <!-- _______________________________________________________________________ -->
3345 <div class="doc_subsubsection">
3346 <a name="i_va_arg">'<tt>va_arg</tt>' Instruction</a>
3349 <div class="doc_text">
3354 <resultval> = va_arg <va_list*> <arglist>, <argty>
3359 <p>The '<tt>va_arg</tt>' instruction is used to access arguments passed through
3360 the "variable argument" area of a function call. It is used to implement the
3361 <tt>va_arg</tt> macro in C.</p>
3365 <p>This instruction takes a <tt>va_list*</tt> value and the type of
3366 the argument. It returns a value of the specified argument type and
3367 increments the <tt>va_list</tt> to point to the next argument. Again, the
3368 actual type of <tt>va_list</tt> is target specific.</p>
3372 <p>The '<tt>va_arg</tt>' instruction loads an argument of the specified
3373 type from the specified <tt>va_list</tt> and causes the
3374 <tt>va_list</tt> to point to the next argument. For more information,
3375 see the variable argument handling <a href="#int_varargs">Intrinsic
3378 <p>It is legal for this instruction to be called in a function which does not
3379 take a variable number of arguments, for example, the <tt>vfprintf</tt>
3382 <p><tt>va_arg</tt> is an LLVM instruction instead of an <a
3383 href="#intrinsics">intrinsic function</a> because it takes a type as an
3388 <p>See the <a href="#int_varargs">variable argument processing</a> section.</p>
3392 <!-- *********************************************************************** -->
3393 <div class="doc_section"> <a name="intrinsics">Intrinsic Functions</a> </div>
3394 <!-- *********************************************************************** -->
3396 <div class="doc_text">
3398 <p>LLVM supports the notion of an "intrinsic function". These functions have
3399 well known names and semantics and are required to follow certain
3400 restrictions. Overall, these instructions represent an extension mechanism for
3401 the LLVM language that does not require changing all of the transformations in
3402 LLVM to add to the language (or the bytecode reader/writer, the parser,
3405 <p>Intrinsic function names must all start with an "<tt>llvm.</tt>" prefix. This
3406 prefix is reserved in LLVM for intrinsic names; thus, functions may not be named
3407 this. Intrinsic functions must always be external functions: you cannot define
3408 the body of intrinsic functions. Intrinsic functions may only be used in call
3409 or invoke instructions: it is illegal to take the address of an intrinsic
3410 function. Additionally, because intrinsic functions are part of the LLVM
3411 language, it is required that they all be documented here if any are added.</p>
3414 <p>To learn how to add an intrinsic function, please see the <a
3415 href="ExtendingLLVM.html">Extending LLVM Guide</a>.
3420 <!-- ======================================================================= -->
3421 <div class="doc_subsection">
3422 <a name="int_varargs">Variable Argument Handling Intrinsics</a>
3425 <div class="doc_text">
3427 <p>Variable argument support is defined in LLVM with the <a
3428 href="#i_va_arg"><tt>va_arg</tt></a> instruction and these three
3429 intrinsic functions. These functions are related to the similarly
3430 named macros defined in the <tt><stdarg.h></tt> header file.</p>
3432 <p>All of these functions operate on arguments that use a
3433 target-specific value type "<tt>va_list</tt>". The LLVM assembly
3434 language reference manual does not define what this type is, so all
3435 transformations should be prepared to handle intrinsics with any type
3438 <p>This example shows how the <a href="#i_va_arg"><tt>va_arg</tt></a>
3439 instruction and the variable argument handling intrinsic functions are
3443 int %test(int %X, ...) {
3444 ; Initialize variable argument processing
3446 call void %<a href="#i_va_start">llvm.va_start</a>(sbyte** %ap)
3448 ; Read a single integer argument
3449 %tmp = va_arg sbyte** %ap, int
3451 ; Demonstrate usage of llvm.va_copy and llvm.va_end
3453 call void %<a href="#i_va_copy">llvm.va_copy</a>(sbyte** %aq, sbyte** %ap)
3454 call void %<a href="#i_va_end">llvm.va_end</a>(sbyte** %aq)
3456 ; Stop processing of arguments.
3457 call void %<a href="#i_va_end">llvm.va_end</a>(sbyte** %ap)
3463 <!-- _______________________________________________________________________ -->
3464 <div class="doc_subsubsection">
3465 <a name="i_va_start">'<tt>llvm.va_start</tt>' Intrinsic</a>
3469 <div class="doc_text">
3471 <pre> declare void %llvm.va_start(<va_list>* <arglist>)<br></pre>
3473 <P>The '<tt>llvm.va_start</tt>' intrinsic initializes
3474 <tt>*<arglist></tt> for subsequent use by <tt><a
3475 href="#i_va_arg">va_arg</a></tt>.</p>
3479 <P>The argument is a pointer to a <tt>va_list</tt> element to initialize.</p>
3483 <P>The '<tt>llvm.va_start</tt>' intrinsic works just like the <tt>va_start</tt>
3484 macro available in C. In a target-dependent way, it initializes the
3485 <tt>va_list</tt> element the argument points to, so that the next call to
3486 <tt>va_arg</tt> will produce the first variable argument passed to the function.
3487 Unlike the C <tt>va_start</tt> macro, this intrinsic does not need to know the
3488 last argument of the function, the compiler can figure that out.</p>
3492 <!-- _______________________________________________________________________ -->
3493 <div class="doc_subsubsection">
3494 <a name="i_va_end">'<tt>llvm.va_end</tt>' Intrinsic</a>
3497 <div class="doc_text">
3499 <pre> declare void %llvm.va_end(<va_list*> <arglist>)<br></pre>
3501 <p>The '<tt>llvm.va_end</tt>' intrinsic destroys <tt><arglist></tt>
3502 which has been initialized previously with <tt><a href="#i_va_start">llvm.va_start</a></tt>
3503 or <tt><a href="#i_va_copy">llvm.va_copy</a></tt>.</p>
3505 <p>The argument is a <tt>va_list</tt> to destroy.</p>
3507 <p>The '<tt>llvm.va_end</tt>' intrinsic works just like the <tt>va_end</tt>
3508 macro available in C. In a target-dependent way, it destroys the <tt>va_list</tt>.
3509 Calls to <a href="#i_va_start"><tt>llvm.va_start</tt></a> and <a
3510 href="#i_va_copy"><tt>llvm.va_copy</tt></a> must be matched exactly
3511 with calls to <tt>llvm.va_end</tt>.</p>
3514 <!-- _______________________________________________________________________ -->
3515 <div class="doc_subsubsection">
3516 <a name="i_va_copy">'<tt>llvm.va_copy</tt>' Intrinsic</a>
3519 <div class="doc_text">
3524 declare void %llvm.va_copy(<va_list>* <destarglist>,
3525 <va_list>* <srcarglist>)
3530 <p>The '<tt>llvm.va_copy</tt>' intrinsic copies the current argument position from
3531 the source argument list to the destination argument list.</p>
3535 <p>The first argument is a pointer to a <tt>va_list</tt> element to initialize.
3536 The second argument is a pointer to a <tt>va_list</tt> element to copy from.</p>
3541 <p>The '<tt>llvm.va_copy</tt>' intrinsic works just like the <tt>va_copy</tt> macro
3542 available in C. In a target-dependent way, it copies the source
3543 <tt>va_list</tt> element into the destination list. This intrinsic is necessary
3544 because the <tt><a href="i_va_begin">llvm.va_begin</a></tt> intrinsic may be
3545 arbitrarily complex and require memory allocation, for example.</p>
3549 <!-- ======================================================================= -->
3550 <div class="doc_subsection">
3551 <a name="int_gc">Accurate Garbage Collection Intrinsics</a>
3554 <div class="doc_text">
3557 LLVM support for <a href="GarbageCollection.html">Accurate Garbage
3558 Collection</a> requires the implementation and generation of these intrinsics.
3559 These intrinsics allow identification of <a href="#i_gcroot">GC roots on the
3560 stack</a>, as well as garbage collector implementations that require <a
3561 href="#i_gcread">read</a> and <a href="#i_gcwrite">write</a> barriers.
3562 Front-ends for type-safe garbage collected languages should generate these
3563 intrinsics to make use of the LLVM garbage collectors. For more details, see <a
3564 href="GarbageCollection.html">Accurate Garbage Collection with LLVM</a>.
3568 <!-- _______________________________________________________________________ -->
3569 <div class="doc_subsubsection">
3570 <a name="i_gcroot">'<tt>llvm.gcroot</tt>' Intrinsic</a>
3573 <div class="doc_text">
3578 declare void %llvm.gcroot(<ty>** %ptrloc, <ty2>* %metadata)
3583 <p>The '<tt>llvm.gcroot</tt>' intrinsic declares the existence of a GC root to
3584 the code generator, and allows some metadata to be associated with it.</p>
3588 <p>The first argument specifies the address of a stack object that contains the
3589 root pointer. The second pointer (which must be either a constant or a global
3590 value address) contains the meta-data to be associated with the root.</p>
3594 <p>At runtime, a call to this intrinsics stores a null pointer into the "ptrloc"
3595 location. At compile-time, the code generator generates information to allow
3596 the runtime to find the pointer at GC safe points.
3602 <!-- _______________________________________________________________________ -->
3603 <div class="doc_subsubsection">
3604 <a name="i_gcread">'<tt>llvm.gcread</tt>' Intrinsic</a>
3607 <div class="doc_text">
3612 declare sbyte* %llvm.gcread(sbyte* %ObjPtr, sbyte** %Ptr)
3617 <p>The '<tt>llvm.gcread</tt>' intrinsic identifies reads of references from heap
3618 locations, allowing garbage collector implementations that require read
3623 <p>The second argument is the address to read from, which should be an address
3624 allocated from the garbage collector. The first object is a pointer to the
3625 start of the referenced object, if needed by the language runtime (otherwise
3630 <p>The '<tt>llvm.gcread</tt>' intrinsic has the same semantics as a load
3631 instruction, but may be replaced with substantially more complex code by the
3632 garbage collector runtime, as needed.</p>
3637 <!-- _______________________________________________________________________ -->
3638 <div class="doc_subsubsection">
3639 <a name="i_gcwrite">'<tt>llvm.gcwrite</tt>' Intrinsic</a>
3642 <div class="doc_text">
3647 declare void %llvm.gcwrite(sbyte* %P1, sbyte* %Obj, sbyte** %P2)
3652 <p>The '<tt>llvm.gcwrite</tt>' intrinsic identifies writes of references to heap
3653 locations, allowing garbage collector implementations that require write
3654 barriers (such as generational or reference counting collectors).</p>
3658 <p>The first argument is the reference to store, the second is the start of the
3659 object to store it to, and the third is the address of the field of Obj to
3660 store to. If the runtime does not require a pointer to the object, Obj may be
3665 <p>The '<tt>llvm.gcwrite</tt>' intrinsic has the same semantics as a store
3666 instruction, but may be replaced with substantially more complex code by the
3667 garbage collector runtime, as needed.</p>
3673 <!-- ======================================================================= -->
3674 <div class="doc_subsection">
3675 <a name="int_codegen">Code Generator Intrinsics</a>
3678 <div class="doc_text">
3680 These intrinsics are provided by LLVM to expose special features that may only
3681 be implemented with code generator support.
3686 <!-- _______________________________________________________________________ -->
3687 <div class="doc_subsubsection">
3688 <a name="i_returnaddress">'<tt>llvm.returnaddress</tt>' Intrinsic</a>
3691 <div class="doc_text">
3695 declare sbyte *%llvm.returnaddress(uint <level>)
3701 The '<tt>llvm.returnaddress</tt>' intrinsic attempts to compute a
3702 target-specific value indicating the return address of the current function
3703 or one of its callers.
3709 The argument to this intrinsic indicates which function to return the address
3710 for. Zero indicates the calling function, one indicates its caller, etc. The
3711 argument is <b>required</b> to be a constant integer value.
3717 The '<tt>llvm.returnaddress</tt>' intrinsic either returns a pointer indicating
3718 the return address of the specified call frame, or zero if it cannot be
3719 identified. The value returned by this intrinsic is likely to be incorrect or 0
3720 for arguments other than zero, so it should only be used for debugging purposes.
3724 Note that calling this intrinsic does not prevent function inlining or other
3725 aggressive transformations, so the value returned may not be that of the obvious
3726 source-language caller.
3731 <!-- _______________________________________________________________________ -->
3732 <div class="doc_subsubsection">
3733 <a name="i_frameaddress">'<tt>llvm.frameaddress</tt>' Intrinsic</a>
3736 <div class="doc_text">
3740 declare sbyte *%llvm.frameaddress(uint <level>)
3746 The '<tt>llvm.frameaddress</tt>' intrinsic attempts to return the
3747 target-specific frame pointer value for the specified stack frame.
3753 The argument to this intrinsic indicates which function to return the frame
3754 pointer for. Zero indicates the calling function, one indicates its caller,
3755 etc. The argument is <b>required</b> to be a constant integer value.
3761 The '<tt>llvm.frameaddress</tt>' intrinsic either returns a pointer indicating
3762 the frame address of the specified call frame, or zero if it cannot be
3763 identified. The value returned by this intrinsic is likely to be incorrect or 0
3764 for arguments other than zero, so it should only be used for debugging purposes.
3768 Note that calling this intrinsic does not prevent function inlining or other
3769 aggressive transformations, so the value returned may not be that of the obvious
3770 source-language caller.
3774 <!-- _______________________________________________________________________ -->
3775 <div class="doc_subsubsection">
3776 <a name="i_stacksave">'<tt>llvm.stacksave</tt>' Intrinsic</a>
3779 <div class="doc_text">
3783 declare sbyte *%llvm.stacksave()
3789 The '<tt>llvm.stacksave</tt>' intrinsic is used to remember the current state of
3790 the function stack, for use with <a href="#i_stackrestore">
3791 <tt>llvm.stackrestore</tt></a>. This is useful for implementing language
3792 features like scoped automatic variable sized arrays in C99.
3798 This intrinsic returns a opaque pointer value that can be passed to <a
3799 href="#i_stackrestore"><tt>llvm.stackrestore</tt></a>. When an
3800 <tt>llvm.stackrestore</tt> intrinsic is executed with a value saved from
3801 <tt>llvm.stacksave</tt>, it effectively restores the state of the stack to the
3802 state it was in when the <tt>llvm.stacksave</tt> intrinsic executed. In
3803 practice, this pops any <a href="#i_alloca">alloca</a> blocks from the stack
3804 that were allocated after the <tt>llvm.stacksave</tt> was executed.
3809 <!-- _______________________________________________________________________ -->
3810 <div class="doc_subsubsection">
3811 <a name="i_stackrestore">'<tt>llvm.stackrestore</tt>' Intrinsic</a>
3814 <div class="doc_text">
3818 declare void %llvm.stackrestore(sbyte* %ptr)
3824 The '<tt>llvm.stackrestore</tt>' intrinsic is used to restore the state of
3825 the function stack to the state it was in when the corresponding <a
3826 href="#llvm.stacksave"><tt>llvm.stacksave</tt></a> intrinsic executed. This is
3827 useful for implementing language features like scoped automatic variable sized
3834 See the description for <a href="#i_stacksave"><tt>llvm.stacksave</tt></a>.
3840 <!-- _______________________________________________________________________ -->
3841 <div class="doc_subsubsection">
3842 <a name="i_prefetch">'<tt>llvm.prefetch</tt>' Intrinsic</a>
3845 <div class="doc_text">
3849 declare void %llvm.prefetch(sbyte * <address>,
3850 uint <rw>, uint <locality>)
3857 The '<tt>llvm.prefetch</tt>' intrinsic is a hint to the code generator to insert
3858 a prefetch instruction if supported; otherwise, it is a noop. Prefetches have
3860 effect on the behavior of the program but can change its performance
3867 <tt>address</tt> is the address to be prefetched, <tt>rw</tt> is the specifier
3868 determining if the fetch should be for a read (0) or write (1), and
3869 <tt>locality</tt> is a temporal locality specifier ranging from (0) - no
3870 locality, to (3) - extremely local keep in cache. The <tt>rw</tt> and
3871 <tt>locality</tt> arguments must be constant integers.
3877 This intrinsic does not modify the behavior of the program. In particular,
3878 prefetches cannot trap and do not produce a value. On targets that support this
3879 intrinsic, the prefetch can provide hints to the processor cache for better
3885 <!-- _______________________________________________________________________ -->
3886 <div class="doc_subsubsection">
3887 <a name="i_pcmarker">'<tt>llvm.pcmarker</tt>' Intrinsic</a>
3890 <div class="doc_text">
3894 declare void %llvm.pcmarker( uint <id> )
3901 The '<tt>llvm.pcmarker</tt>' intrinsic is a method to export a Program Counter
3903 code to simulators and other tools. The method is target specific, but it is
3904 expected that the marker will use exported symbols to transmit the PC of the marker.
3905 The marker makes no guarantees that it will remain with any specific instruction
3906 after optimizations. It is possible that the presence of a marker will inhibit
3907 optimizations. The intended use is to be inserted after optimizations to allow
3908 correlations of simulation runs.
3914 <tt>id</tt> is a numerical id identifying the marker.
3920 This intrinsic does not modify the behavior of the program. Backends that do not
3921 support this intrinisic may ignore it.
3926 <!-- _______________________________________________________________________ -->
3927 <div class="doc_subsubsection">
3928 <a name="i_readcyclecounter">'<tt>llvm.readcyclecounter</tt>' Intrinsic</a>
3931 <div class="doc_text">
3935 declare ulong %llvm.readcyclecounter( )
3942 The '<tt>llvm.readcyclecounter</tt>' intrinsic provides access to the cycle
3943 counter register (or similar low latency, high accuracy clocks) on those targets
3944 that support it. On X86, it should map to RDTSC. On Alpha, it should map to RPCC.
3945 As the backing counters overflow quickly (on the order of 9 seconds on alpha), this
3946 should only be used for small timings.
3952 When directly supported, reading the cycle counter should not modify any memory.
3953 Implementations are allowed to either return a application specific value or a
3954 system wide value. On backends without support, this is lowered to a constant 0.
3959 <!-- ======================================================================= -->
3960 <div class="doc_subsection">
3961 <a name="int_libc">Standard C Library Intrinsics</a>
3964 <div class="doc_text">
3966 LLVM provides intrinsics for a few important standard C library functions.
3967 These intrinsics allow source-language front-ends to pass information about the
3968 alignment of the pointer arguments to the code generator, providing opportunity
3969 for more efficient code generation.
3974 <!-- _______________________________________________________________________ -->
3975 <div class="doc_subsubsection">
3976 <a name="i_memcpy">'<tt>llvm.memcpy</tt>' Intrinsic</a>
3979 <div class="doc_text">
3983 declare void %llvm.memcpy.i32(sbyte* <dest>, sbyte* <src>,
3984 uint <len>, uint <align>)
3985 declare void %llvm.memcpy.i64(sbyte* <dest>, sbyte* <src>,
3986 ulong <len>, uint <align>)
3992 The '<tt>llvm.memcpy.*</tt>' intrinsics copy a block of memory from the source
3993 location to the destination location.
3997 Note that, unlike the standard libc function, the <tt>llvm.memcpy.*</tt>
3998 intrinsics do not return a value, and takes an extra alignment argument.
4004 The first argument is a pointer to the destination, the second is a pointer to
4005 the source. The third argument is an integer argument
4006 specifying the number of bytes to copy, and the fourth argument is the alignment
4007 of the source and destination locations.
4011 If the call to this intrinisic has an alignment value that is not 0 or 1, then
4012 the caller guarantees that both the source and destination pointers are aligned
4019 The '<tt>llvm.memcpy.*</tt>' intrinsics copy a block of memory from the source
4020 location to the destination location, which are not allowed to overlap. It
4021 copies "len" bytes of memory over. If the argument is known to be aligned to
4022 some boundary, this can be specified as the fourth argument, otherwise it should
4028 <!-- _______________________________________________________________________ -->
4029 <div class="doc_subsubsection">
4030 <a name="i_memmove">'<tt>llvm.memmove</tt>' Intrinsic</a>
4033 <div class="doc_text">
4037 declare void %llvm.memmove.i32(sbyte* <dest>, sbyte* <src>,
4038 uint <len>, uint <align>)
4039 declare void %llvm.memmove.i64(sbyte* <dest>, sbyte* <src>,
4040 ulong <len>, uint <align>)
4046 The '<tt>llvm.memmove.*</tt>' intrinsics move a block of memory from the source
4047 location to the destination location. It is similar to the
4048 '<tt>llvm.memcmp</tt>' intrinsic but allows the two memory locations to overlap.
4052 Note that, unlike the standard libc function, the <tt>llvm.memmove.*</tt>
4053 intrinsics do not return a value, and takes an extra alignment argument.
4059 The first argument is a pointer to the destination, the second is a pointer to
4060 the source. The third argument is an integer argument
4061 specifying the number of bytes to copy, and the fourth argument is the alignment
4062 of the source and destination locations.
4066 If the call to this intrinisic has an alignment value that is not 0 or 1, then
4067 the caller guarantees that the source and destination pointers are aligned to
4074 The '<tt>llvm.memmove.*</tt>' intrinsics copy a block of memory from the source
4075 location to the destination location, which may overlap. It
4076 copies "len" bytes of memory over. If the argument is known to be aligned to
4077 some boundary, this can be specified as the fourth argument, otherwise it should
4083 <!-- _______________________________________________________________________ -->
4084 <div class="doc_subsubsection">
4085 <a name="i_memset">'<tt>llvm.memset.*</tt>' Intrinsics</a>
4088 <div class="doc_text">
4092 declare void %llvm.memset.i32(sbyte* <dest>, ubyte <val>,
4093 uint <len>, uint <align>)
4094 declare void %llvm.memset.i64(sbyte* <dest>, ubyte <val>,
4095 ulong <len>, uint <align>)
4101 The '<tt>llvm.memset.*</tt>' intrinsics fill a block of memory with a particular
4106 Note that, unlike the standard libc function, the <tt>llvm.memset</tt> intrinsic
4107 does not return a value, and takes an extra alignment argument.
4113 The first argument is a pointer to the destination to fill, the second is the
4114 byte value to fill it with, the third argument is an integer
4115 argument specifying the number of bytes to fill, and the fourth argument is the
4116 known alignment of destination location.
4120 If the call to this intrinisic has an alignment value that is not 0 or 1, then
4121 the caller guarantees that the destination pointer is aligned to that boundary.
4127 The '<tt>llvm.memset.*</tt>' intrinsics fill "len" bytes of memory starting at
4129 destination location. If the argument is known to be aligned to some boundary,
4130 this can be specified as the fourth argument, otherwise it should be set to 0 or
4136 <!-- _______________________________________________________________________ -->
4137 <div class="doc_subsubsection">
4138 <a name="i_isunordered">'<tt>llvm.isunordered.*</tt>' Intrinsic</a>
4141 <div class="doc_text">
4145 declare bool %llvm.isunordered.f32(float Val1, float Val2)
4146 declare bool %llvm.isunordered.f64(double Val1, double Val2)
4152 The '<tt>llvm.isunordered</tt>' intrinsics return true if either or both of the
4153 specified floating point values is a NAN.
4159 The arguments are floating point numbers of the same type.
4165 If either or both of the arguments is a SNAN or QNAN, it returns true, otherwise
4171 <!-- _______________________________________________________________________ -->
4172 <div class="doc_subsubsection">
4173 <a name="i_sqrt">'<tt>llvm.sqrt.*</tt>' Intrinsic</a>
4176 <div class="doc_text">
4180 declare float %llvm.sqrt.f32(float %Val)
4181 declare double %llvm.sqrt.f64(double %Val)
4187 The '<tt>llvm.sqrt</tt>' intrinsics return the sqrt of the specified operand,
4188 returning the same value as the libm '<tt>sqrt</tt>' function would. Unlike
4189 <tt>sqrt</tt> in libm, however, <tt>llvm.sqrt</tt> has undefined behavior for
4190 negative numbers (which allows for better optimization).
4196 The argument and return value are floating point numbers of the same type.
4202 This function returns the sqrt of the specified operand if it is a positive
4203 floating point number.
4207 <!-- _______________________________________________________________________ -->
4208 <div class="doc_subsubsection">
4209 <a name="i_powi">'<tt>llvm.powi.*</tt>' Intrinsic</a>
4212 <div class="doc_text">
4216 declare float %llvm.powi.f32(float %Val, int %power)
4217 declare double %llvm.powi.f64(double %Val, int %power)
4223 The '<tt>llvm.powi.*</tt>' intrinsics return the first operand raised to the
4224 specified (positive or negative) power. The order of evaluation of
4225 multiplications is not defined.
4231 The second argument is an integer power, and the first is a value to raise to
4238 This function returns the first value raised to the second power with an
4239 unspecified sequence of rounding operations.</p>
4243 <!-- ======================================================================= -->
4244 <div class="doc_subsection">
4245 <a name="int_manip">Bit Manipulation Intrinsics</a>
4248 <div class="doc_text">
4250 LLVM provides intrinsics for a few important bit manipulation operations.
4251 These allow efficient code generation for some algorithms.
4256 <!-- _______________________________________________________________________ -->
4257 <div class="doc_subsubsection">
4258 <a name="i_bswap">'<tt>llvm.bswap.*</tt>' Intrinsics</a>
4261 <div class="doc_text">
4265 declare ushort %llvm.bswap.i16(ushort <id>)
4266 declare uint %llvm.bswap.i32(uint <id>)
4267 declare ulong %llvm.bswap.i64(ulong <id>)
4273 The '<tt>llvm.bwsap</tt>' family of intrinsics is used to byteswap a 16, 32 or
4274 64 bit quantity. These are useful for performing operations on data that is not
4275 in the target's native byte order.
4281 The <tt>llvm.bswap.16</tt> intrinsic returns a ushort value that has the high and low
4282 byte of the input ushort swapped. Similarly, the <tt>llvm.bswap.i32</tt> intrinsic
4283 returns a uint value that has the four bytes of the input uint swapped, so that
4284 if the input bytes are numbered 0, 1, 2, 3 then the returned uint will have its
4285 bytes in 3, 2, 1, 0 order. The <tt>llvm.bswap.i64</tt> intrinsic extends this concept
4291 <!-- _______________________________________________________________________ -->
4292 <div class="doc_subsubsection">
4293 <a name="int_ctpop">'<tt>llvm.ctpop.*</tt>' Intrinsic</a>
4296 <div class="doc_text">
4300 declare ubyte %llvm.ctpop.i8 (ubyte <src>)
4301 declare ushort %llvm.ctpop.i16(ushort <src>)
4302 declare uint %llvm.ctpop.i32(uint <src>)
4303 declare ulong %llvm.ctpop.i64(ulong <src>)
4309 The '<tt>llvm.ctpop</tt>' family of intrinsics counts the number of bits set in a
4316 The only argument is the value to be counted. The argument may be of any
4317 unsigned integer type. The return type must match the argument type.
4323 The '<tt>llvm.ctpop</tt>' intrinsic counts the 1's in a variable.
4327 <!-- _______________________________________________________________________ -->
4328 <div class="doc_subsubsection">
4329 <a name="int_ctlz">'<tt>llvm.ctlz.*</tt>' Intrinsic</a>
4332 <div class="doc_text">
4336 declare ubyte %llvm.ctlz.i8 (ubyte <src>)
4337 declare ushort %llvm.ctlz.i16(ushort <src>)
4338 declare uint %llvm.ctlz.i32(uint <src>)
4339 declare ulong %llvm.ctlz.i64(ulong <src>)
4345 The '<tt>llvm.ctlz</tt>' family of intrinsic functions counts the number of
4346 leading zeros in a variable.
4352 The only argument is the value to be counted. The argument may be of any
4353 unsigned integer type. The return type must match the argument type.
4359 The '<tt>llvm.ctlz</tt>' intrinsic counts the leading (most significant) zeros
4360 in a variable. If the src == 0 then the result is the size in bits of the type
4361 of src. For example, <tt>llvm.ctlz(int 2) = 30</tt>.
4367 <!-- _______________________________________________________________________ -->
4368 <div class="doc_subsubsection">
4369 <a name="int_cttz">'<tt>llvm.cttz.*</tt>' Intrinsic</a>
4372 <div class="doc_text">
4376 declare ubyte %llvm.cttz.i8 (ubyte <src>)
4377 declare ushort %llvm.cttz.i16(ushort <src>)
4378 declare uint %llvm.cttz.i32(uint <src>)
4379 declare ulong %llvm.cttz.i64(ulong <src>)
4385 The '<tt>llvm.cttz</tt>' family of intrinsic functions counts the number of
4392 The only argument is the value to be counted. The argument may be of any
4393 unsigned integer type. The return type must match the argument type.
4399 The '<tt>llvm.cttz</tt>' intrinsic counts the trailing (least significant) zeros
4400 in a variable. If the src == 0 then the result is the size in bits of the type
4401 of src. For example, <tt>llvm.cttz(2) = 1</tt>.
4405 <!-- ======================================================================= -->
4406 <div class="doc_subsection">
4407 <a name="int_debugger">Debugger Intrinsics</a>
4410 <div class="doc_text">
4412 The LLVM debugger intrinsics (which all start with <tt>llvm.dbg.</tt> prefix),
4413 are described in the <a
4414 href="SourceLevelDebugging.html#format_common_intrinsics">LLVM Source Level
4415 Debugging</a> document.
4420 <!-- *********************************************************************** -->
4423 <a href="http://jigsaw.w3.org/css-validator/check/referer"><img
4424 src="http://jigsaw.w3.org/css-validator/images/vcss" alt="Valid CSS!"></a>
4425 <a href="http://validator.w3.org/check/referer"><img
4426 src="http://www.w3.org/Icons/valid-html401" alt="Valid HTML 4.01!" /></a>
4428 <a href="mailto:sabre@nondot.org">Chris Lattner</a><br>
4429 <a href="http://llvm.org">The LLVM Compiler Infrastructure</a><br>
4430 Last modified: $Date$