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">Function Structure</a></li>
29 <li><a href="#typesystem">Type System</a>
31 <li><a href="#t_primitive">Primitive Types</a>
33 <li><a href="#t_classifications">Type Classifications</a></li>
36 <li><a href="#t_derived">Derived Types</a>
38 <li><a href="#t_array">Array Type</a></li>
39 <li><a href="#t_function">Function Type</a></li>
40 <li><a href="#t_pointer">Pointer Type</a></li>
41 <li><a href="#t_struct">Structure Type</a></li>
42 <li><a href="#t_packed">Packed Type</a></li>
43 <li><a href="#t_opaque">Opaque Type</a></li>
48 <li><a href="#constants">Constants</a>
50 <li><a href="#simpleconstants">Simple Constants</a>
51 <li><a href="#aggregateconstants">Aggregate Constants</a>
52 <li><a href="#globalconstants">Global Variable and Function Addresses</a>
53 <li><a href="#undefvalues">Undefined Values</a>
54 <li><a href="#constantexprs">Constant Expressions</a>
57 <li><a href="#instref">Instruction Reference</a>
59 <li><a href="#terminators">Terminator Instructions</a>
61 <li><a href="#i_ret">'<tt>ret</tt>' Instruction</a></li>
62 <li><a href="#i_br">'<tt>br</tt>' Instruction</a></li>
63 <li><a href="#i_switch">'<tt>switch</tt>' Instruction</a></li>
64 <li><a href="#i_invoke">'<tt>invoke</tt>' Instruction</a></li>
65 <li><a href="#i_unwind">'<tt>unwind</tt>' Instruction</a></li>
66 <li><a href="#i_unreachable">'<tt>unreachable</tt>' Instruction</a></li>
69 <li><a href="#binaryops">Binary Operations</a>
71 <li><a href="#i_add">'<tt>add</tt>' Instruction</a></li>
72 <li><a href="#i_sub">'<tt>sub</tt>' Instruction</a></li>
73 <li><a href="#i_mul">'<tt>mul</tt>' Instruction</a></li>
74 <li><a href="#i_div">'<tt>div</tt>' Instruction</a></li>
75 <li><a href="#i_rem">'<tt>rem</tt>' Instruction</a></li>
76 <li><a href="#i_setcc">'<tt>set<i>cc</i></tt>' Instructions</a></li>
79 <li><a href="#bitwiseops">Bitwise Binary Operations</a>
81 <li><a href="#i_and">'<tt>and</tt>' Instruction</a></li>
82 <li><a href="#i_or">'<tt>or</tt>' Instruction</a></li>
83 <li><a href="#i_xor">'<tt>xor</tt>' Instruction</a></li>
84 <li><a href="#i_shl">'<tt>shl</tt>' Instruction</a></li>
85 <li><a href="#i_shr">'<tt>shr</tt>' Instruction</a></li>
88 <li><a href="#memoryops">Memory Access Operations</a>
90 <li><a href="#i_malloc">'<tt>malloc</tt>' Instruction</a></li>
91 <li><a href="#i_free">'<tt>free</tt>' Instruction</a></li>
92 <li><a href="#i_alloca">'<tt>alloca</tt>' Instruction</a></li>
93 <li><a href="#i_load">'<tt>load</tt>' Instruction</a></li>
94 <li><a href="#i_store">'<tt>store</tt>' Instruction</a></li>
95 <li><a href="#i_getelementptr">'<tt>getelementptr</tt>' Instruction</a></li>
98 <li><a href="#otherops">Other Operations</a>
100 <li><a href="#i_phi">'<tt>phi</tt>' Instruction</a></li>
101 <li><a href="#i_cast">'<tt>cast .. to</tt>' Instruction</a></li>
102 <li><a href="#i_select">'<tt>select</tt>' Instruction</a></li>
103 <li><a href="#i_call">'<tt>call</tt>' Instruction</a></li>
104 <li><a href="#i_vanext">'<tt>vanext</tt>' Instruction</a></li>
105 <li><a href="#i_vaarg">'<tt>vaarg</tt>' Instruction</a></li>
110 <li><a href="#intrinsics">Intrinsic Functions</a>
112 <li><a href="#int_varargs">Variable Argument Handling Intrinsics</a>
114 <li><a href="#i_va_start">'<tt>llvm.va_start</tt>' Intrinsic</a></li>
115 <li><a href="#i_va_end">'<tt>llvm.va_end</tt>' Intrinsic</a></li>
116 <li><a href="#i_va_copy">'<tt>llvm.va_copy</tt>' Intrinsic</a></li>
119 <li><a href="#int_gc">Accurate Garbage Collection Intrinsics</a>
121 <li><a href="#i_gcroot">'<tt>llvm.gcroot</tt>' Intrinsic</a></li>
122 <li><a href="#i_gcread">'<tt>llvm.gcread</tt>' Intrinsic</a></li>
123 <li><a href="#i_gcwrite">'<tt>llvm.gcwrite</tt>' Intrinsic</a></li>
126 <li><a href="#int_codegen">Code Generator Intrinsics</a>
128 <li><a href="#i_returnaddress">'<tt>llvm.returnaddress</tt>' Intrinsic</a></li>
129 <li><a href="#i_frameaddress">'<tt>llvm.frameaddress</tt>' Intrinsic</a></li>
130 <li><a href="#i_prefetch">'<tt>llvm.prefetch</tt>' Intrinsic</a></li>
131 <li><a href="#i_pcmarker">'<tt>llvm.pcmarker</tt>' Intrinsic</a></li>
134 <li><a href="#int_os">Operating System Intrinsics</a>
136 <li><a href="#i_readport">'<tt>llvm.readport</tt>' Intrinsic</a></li>
137 <li><a href="#i_writeport">'<tt>llvm.writeport</tt>' Intrinsic</a></li>
138 <li><a href="#i_readio">'<tt>llvm.readio</tt>' Intrinsic</a></li>
139 <li><a href="#i_writeio">'<tt>llvm.writeio</tt>' Intrinsic</a></li>
141 <li><a href="#int_libc">Standard C Library Intrinsics</a>
143 <li><a href="#i_memcpy">'<tt>llvm.memcpy</tt>' Intrinsic</a></li>
144 <li><a href="#i_memmove">'<tt>llvm.memmove</tt>' Intrinsic</a></li>
145 <li><a href="#i_memset">'<tt>llvm.memset</tt>' Intrinsic</a></li>
146 <li><a href="#i_isunordered">'<tt>llvm.isunordered</tt>' Intrinsic</a></li>
149 <li><a href="#int_count">Bit counting Intrinsics</a>
151 <li><a href="#int_ctpop">'<tt>llvm.ctpop</tt>' Intrinsic </a></li>
152 <li><a href="#int_ctlz">'<tt>llvm.ctlz</tt>' Intrinsic </a></li>
153 <li><a href="#int_cttz">'<tt>llvm.cttz</tt>' Intrinsic </a></li>
156 <li><a href="#int_debugger">Debugger intrinsics</a></li>
161 <div class="doc_author">
162 <p>Written by <a href="mailto:sabre@nondot.org">Chris Lattner</a>
163 and <a href="mailto:vadve@cs.uiuc.edu">Vikram Adve</a></p>
166 <!-- *********************************************************************** -->
167 <div class="doc_section"> <a name="abstract">Abstract </a></div>
168 <!-- *********************************************************************** -->
170 <div class="doc_text">
171 <p>This document is a reference manual for the LLVM assembly language.
172 LLVM is an SSA based representation that provides type safety,
173 low-level operations, flexibility, and the capability of representing
174 'all' high-level languages cleanly. It is the common code
175 representation used throughout all phases of the LLVM compilation
179 <!-- *********************************************************************** -->
180 <div class="doc_section"> <a name="introduction">Introduction</a> </div>
181 <!-- *********************************************************************** -->
183 <div class="doc_text">
185 <p>The LLVM code representation is designed to be used in three
186 different forms: as an in-memory compiler IR, as an on-disk bytecode
187 representation (suitable for fast loading by a Just-In-Time compiler),
188 and as a human readable assembly language representation. This allows
189 LLVM to provide a powerful intermediate representation for efficient
190 compiler transformations and analysis, while providing a natural means
191 to debug and visualize the transformations. The three different forms
192 of LLVM are all equivalent. This document describes the human readable
193 representation and notation.</p>
195 <p>The LLVM representation aims to be light-weight and low-level
196 while being expressive, typed, and extensible at the same time. It
197 aims to be a "universal IR" of sorts, by being at a low enough level
198 that high-level ideas may be cleanly mapped to it (similar to how
199 microprocessors are "universal IR's", allowing many source languages to
200 be mapped to them). By providing type information, LLVM can be used as
201 the target of optimizations: for example, through pointer analysis, it
202 can be proven that a C automatic variable is never accessed outside of
203 the current function... allowing it to be promoted to a simple SSA
204 value instead of a memory location.</p>
208 <!-- _______________________________________________________________________ -->
209 <div class="doc_subsubsection"> <a name="wellformed">Well-Formedness</a> </div>
211 <div class="doc_text">
213 <p>It is important to note that this document describes 'well formed'
214 LLVM assembly language. There is a difference between what the parser
215 accepts and what is considered 'well formed'. For example, the
216 following instruction is syntactically okay, but not well formed:</p>
219 %x = <a href="#i_add">add</a> int 1, %x
222 <p>...because the definition of <tt>%x</tt> does not dominate all of
223 its uses. The LLVM infrastructure provides a verification pass that may
224 be used to verify that an LLVM module is well formed. This pass is
225 automatically run by the parser after parsing input assembly and by
226 the optimizer before it outputs bytecode. The violations pointed out
227 by the verifier pass indicate bugs in transformation passes or input to
230 <!-- Describe the typesetting conventions here. --> </div>
232 <!-- *********************************************************************** -->
233 <div class="doc_section"> <a name="identifiers">Identifiers</a> </div>
234 <!-- *********************************************************************** -->
236 <div class="doc_text">
238 <p>LLVM uses three different forms of identifiers, for different
242 <li>Named values are represented as a string of characters with a '%' prefix.
243 For example, %foo, %DivisionByZero, %a.really.long.identifier. The actual
244 regular expression used is '<tt>%[a-zA-Z$._][a-zA-Z$._0-9]*</tt>'.
245 Identifiers which require other characters in their names can be surrounded
246 with quotes. In this way, anything except a <tt>"</tt> character can be used
249 <li>Unnamed values are represented as an unsigned numeric value with a '%'
250 prefix. For example, %12, %2, %44.</li>
252 <li>Constants, which are described in a <a href="#constants">section about
253 constants</a>, below.</li>
256 <p>LLVM requires that values start with a '%' sign for two reasons: Compilers
257 don't need to worry about name clashes with reserved words, and the set of
258 reserved words may be expanded in the future without penalty. Additionally,
259 unnamed identifiers allow a compiler to quickly come up with a temporary
260 variable without having to avoid symbol table conflicts.</p>
262 <p>Reserved words in LLVM are very similar to reserved words in other
263 languages. There are keywords for different opcodes ('<tt><a
264 href="#i_add">add</a></tt>', '<tt><a href="#i_cast">cast</a></tt>', '<tt><a
265 href="#i_ret">ret</a></tt>', etc...), for primitive type names ('<tt><a
266 href="#t_void">void</a></tt>', '<tt><a href="#t_uint">uint</a></tt>', etc...),
267 and others. These reserved words cannot conflict with variable names, because
268 none of them start with a '%' character.</p>
270 <p>Here is an example of LLVM code to multiply the integer variable
271 '<tt>%X</tt>' by 8:</p>
276 %result = <a href="#i_mul">mul</a> uint %X, 8
279 <p>After strength reduction:</p>
282 %result = <a href="#i_shl">shl</a> uint %X, ubyte 3
285 <p>And the hard way:</p>
288 <a href="#i_add">add</a> uint %X, %X <i>; yields {uint}:%0</i>
289 <a href="#i_add">add</a> uint %0, %0 <i>; yields {uint}:%1</i>
290 %result = <a href="#i_add">add</a> uint %1, %1
293 <p>This last way of multiplying <tt>%X</tt> by 8 illustrates several
294 important lexical features of LLVM:</p>
298 <li>Comments are delimited with a '<tt>;</tt>' and go until the end of
301 <li>Unnamed temporaries are created when the result of a computation is not
302 assigned to a named value.</li>
304 <li>Unnamed temporaries are numbered sequentially</li>
308 <p>...and it also shows a convention that we follow in this document. When
309 demonstrating instructions, we will follow an instruction with a comment that
310 defines the type and name of value produced. Comments are shown in italic
315 <!-- *********************************************************************** -->
316 <div class="doc_section"> <a name="highlevel">High Level Structure</a> </div>
317 <!-- *********************************************************************** -->
319 <!-- ======================================================================= -->
320 <div class="doc_subsection"> <a name="modulestructure">Module Structure</a>
323 <div class="doc_text">
325 <p>LLVM programs are composed of "Module"s, each of which is a
326 translation unit of the input programs. Each module consists of
327 functions, global variables, and symbol table entries. Modules may be
328 combined together with the LLVM linker, which merges function (and
329 global variable) definitions, resolves forward declarations, and merges
330 symbol table entries. Here is an example of the "hello world" module:</p>
332 <pre><i>; Declare the string constant as a global constant...</i>
333 <a href="#identifiers">%.LC0</a> = <a href="#linkage_internal">internal</a> <a
334 href="#globalvars">constant</a> <a href="#t_array">[13 x sbyte]</a> c"hello world\0A\00" <i>; [13 x sbyte]*</i>
336 <i>; External declaration of the puts function</i>
337 <a href="#functionstructure">declare</a> int %puts(sbyte*) <i>; int(sbyte*)* </i>
339 <i>; Definition of main function</i>
340 int %main() { <i>; int()* </i>
341 <i>; Convert [13x sbyte]* to sbyte *...</i>
343 href="#i_getelementptr">getelementptr</a> [13 x sbyte]* %.LC0, long 0, long 0 <i>; sbyte*</i>
345 <i>; Call puts function to write out the string to stdout...</i>
347 href="#i_call">call</a> int %puts(sbyte* %cast210) <i>; int</i>
349 href="#i_ret">ret</a> int 0<br>}<br></pre>
351 <p>This example is made up of a <a href="#globalvars">global variable</a>
352 named "<tt>.LC0</tt>", an external declaration of the "<tt>puts</tt>"
353 function, and a <a href="#functionstructure">function definition</a>
354 for "<tt>main</tt>".</p>
356 <p>In general, a module is made up of a list of global values,
357 where both functions and global variables are global values. Global values are
358 represented by a pointer to a memory location (in this case, a pointer to an
359 array of char, and a pointer to a function), and have one of the following <a
360 href="#linkage">linkage types</a>.</p>
364 <!-- ======================================================================= -->
365 <div class="doc_subsection">
366 <a name="linkage">Linkage Types</a>
369 <div class="doc_text">
372 All Global Variables and Functions have one of the following types of linkage:
377 <dt><tt><b><a name="linkage_internal">internal</a></b></tt> </dt>
379 <dd>Global values with internal linkage are only directly accessible by
380 objects in the current module. In particular, linking code into a module with
381 an internal global value may cause the internal to be renamed as necessary to
382 avoid collisions. Because the symbol is internal to the module, all
383 references can be updated. This corresponds to the notion of the
384 '<tt>static</tt>' keyword in C, or the idea of "anonymous namespaces" in C++.
387 <dt><tt><b><a name="linkage_linkonce">linkonce</a></b></tt>: </dt>
389 <dd>"<tt>linkonce</tt>" linkage is similar to <tt>internal</tt> linkage, with
390 the twist that linking together two modules defining the same
391 <tt>linkonce</tt> globals will cause one of the globals to be discarded. This
392 is typically used to implement inline functions. Unreferenced
393 <tt>linkonce</tt> globals are allowed to be discarded.
396 <dt><tt><b><a name="linkage_weak">weak</a></b></tt>: </dt>
398 <dd>"<tt>weak</tt>" linkage is exactly the same as <tt>linkonce</tt> linkage,
399 except that unreferenced <tt>weak</tt> globals may not be discarded. This is
400 used to implement constructs in C such as "<tt>int X;</tt>" at global scope.
403 <dt><tt><b><a name="linkage_appending">appending</a></b></tt>: </dt>
405 <dd>"<tt>appending</tt>" linkage may only be applied to global variables of
406 pointer to array type. When two global variables with appending linkage are
407 linked together, the two global arrays are appended together. This is the
408 LLVM, typesafe, equivalent of having the system linker append together
409 "sections" with identical names when .o files are linked.
412 <dt><tt><b><a name="linkage_external">externally visible</a></b></tt>:</dt>
414 <dd>If none of the above identifiers are used, the global is externally
415 visible, meaning that it participates in linkage and can be used to resolve
416 external symbol references.
420 <p><a name="linkage_external">For example, since the "<tt>.LC0</tt>"
421 variable is defined to be internal, if another module defined a "<tt>.LC0</tt>"
422 variable and was linked with this one, one of the two would be renamed,
423 preventing a collision. Since "<tt>main</tt>" and "<tt>puts</tt>" are
424 external (i.e., lacking any linkage declarations), they are accessible
425 outside of the current module. It is illegal for a function <i>declaration</i>
426 to have any linkage type other than "externally visible".</a></p>
430 <!-- ======================================================================= -->
431 <div class="doc_subsection">
432 <a name="callingconv">Calling Conventions</a>
435 <div class="doc_text">
437 <p>LLVM <a href="#functionstructure">functions</a>, <a href="#i_call">calls</a>
438 and <a href="#i_invoke">invokes</a> can all have an optional calling convention
439 specified for the call. The calling convention of any pair of dynamic
440 caller/callee must match, or the behavior of the program is undefined. The
441 following calling conventions are supported by LLVM, and more may be added in
445 <dt><b>"<tt>ccc</tt>" - The C calling convention</b>:</dt>
447 <dd>This calling convention (the default if no other calling convention is
448 specified) matches the target C calling conventions. This calling convention
449 supports varargs function calls and tolerates some mismatch in the declared
450 prototype and implemented declaration of the function (as does normal C).
453 <dt><b>"<tt>fastcc</tt>" - The fast calling convention</b>:</dt>
455 <dd>This calling convention attempts to make calls as fast as possible
456 (e.g. by passing things in registers). This calling convention allows the
457 target to use whatever tricks it wants to produce fast code for the target,
458 without having to conform to an externally specified ABI. Implementations of
459 this convention should allow arbitrary tail call optimization to be supported.
460 This calling convention does not support varargs and requires the prototype of
461 all callees to exactly match the prototype of the function definition.
464 <dt><b>"<tt>coldcc</tt>" - The cold calling convention</b>:</dt>
466 <dd>This calling convention attempts to make code in the caller as efficient
467 as possible under the assumption that the call is not commonly executed. As
468 such, these calls often preserve all registers so that the call does not break
469 any live ranges in the caller side. This calling convention does not support
470 varargs and requires the prototype of all callees to exactly match the
471 prototype of the function definition.
474 <dt><b>"<tt>cc <<em>n</em>></tt>" - Numbered convention</b>:</dt>
476 <dd>Any calling convention may be specified by number, allowing
477 target-specific calling conventions to be used. Target specific calling
478 conventions start at 64.
482 <p>More calling conventions can be added/defined on an as-needed basis, to
483 support pascal conventions or any other well-known target-independent
488 <!-- ======================================================================= -->
489 <div class="doc_subsection">
490 <a name="globalvars">Global Variables</a>
493 <div class="doc_text">
495 <p>Global variables define regions of memory allocated at compilation time
496 instead of run-time. Global variables may optionally be initialized. A
497 variable may be defined as a global "constant", which indicates that the
498 contents of the variable will <b>never</b> be modified (enabling better
499 optimization, allowing the global data to be placed in the read-only section of
500 an executable, etc). Note that variables that need runtime initialization
501 cannot be marked "constant", as there is a store to the variable.</p>
504 LLVM explicitly allows <em>declarations</em> of global variables to be marked
505 constant, even if the final definition of the global is not. This capability
506 can be used to enable slightly better optimization of the program, but requires
507 the language definition to guarantee that optimizations based on the
508 'constantness' are valid for the translation units that do not include the
512 <p>As SSA values, global variables define pointer values that are in
513 scope (i.e. they dominate) all basic blocks in the program. Global
514 variables always define a pointer to their "content" type because they
515 describe a region of memory, and all memory objects in LLVM are
516 accessed through pointers.</p>
521 <!-- ======================================================================= -->
522 <div class="doc_subsection">
523 <a name="functionstructure">Functions</a>
526 <div class="doc_text">
528 <p>LLVM function definitions consist of an optional <a href="#linkage">linkage
529 type</a>, an optional <a href="#callingconv">calling convention</a>, a return
530 type, a function name, a (possibly empty) argument list, an opening curly brace,
531 a list of basic blocks, and a closing curly brace. LLVM function declarations
532 are defined with the "<tt>declare</tt>" keyword, an optional <a
533 href="#callingconv">calling convention</a>, a return type, a function name, and
534 a possibly empty list of arguments.</p>
536 <p>A function definition contains a list of basic blocks, forming the CFG for
537 the function. Each basic block may optionally start with a label (giving the
538 basic block a symbol table entry), contains a list of instructions, and ends
539 with a <a href="#terminators">terminator</a> instruction (such as a branch or
540 function return).</p>
542 <p>The first basic block in a program is special in two ways: it is immediately
543 executed on entrance to the function, and it is not allowed to have predecessor
544 basic blocks (i.e. there can not be any branches to the entry block of a
545 function). Because the block can have no predecessors, it also cannot have any
546 <a href="#i_phi">PHI nodes</a>.</p>
548 <p>LLVM functions are identified by their name and type signature. Hence, two
549 functions with the same name but different parameter lists or return values are
550 considered different functions, and LLVM will resolve references to each
557 <!-- *********************************************************************** -->
558 <div class="doc_section"> <a name="typesystem">Type System</a> </div>
559 <!-- *********************************************************************** -->
561 <div class="doc_text">
563 <p>The LLVM type system is one of the most important features of the
564 intermediate representation. Being typed enables a number of
565 optimizations to be performed on the IR directly, without having to do
566 extra analyses on the side before the transformation. A strong type
567 system makes it easier to read the generated code and enables novel
568 analyses and transformations that are not feasible to perform on normal
569 three address code representations.</p>
573 <!-- ======================================================================= -->
574 <div class="doc_subsection"> <a name="t_primitive">Primitive Types</a> </div>
575 <div class="doc_text">
576 <p>The primitive types are the fundamental building blocks of the LLVM
577 system. The current set of primitive types is as follows:</p>
579 <table class="layout">
584 <tr><th>Type</th><th>Description</th></tr>
585 <tr><td><tt>void</tt></td><td>No value</td></tr>
586 <tr><td><tt>ubyte</tt></td><td>Unsigned 8-bit value</td></tr>
587 <tr><td><tt>ushort</tt></td><td>Unsigned 16-bit value</td></tr>
588 <tr><td><tt>uint</tt></td><td>Unsigned 32-bit value</td></tr>
589 <tr><td><tt>ulong</tt></td><td>Unsigned 64-bit value</td></tr>
590 <tr><td><tt>float</tt></td><td>32-bit floating point value</td></tr>
591 <tr><td><tt>label</tt></td><td>Branch destination</td></tr>
598 <tr><th>Type</th><th>Description</th></tr>
599 <tr><td><tt>bool</tt></td><td>True or False value</td></tr>
600 <tr><td><tt>sbyte</tt></td><td>Signed 8-bit value</td></tr>
601 <tr><td><tt>short</tt></td><td>Signed 16-bit value</td></tr>
602 <tr><td><tt>int</tt></td><td>Signed 32-bit value</td></tr>
603 <tr><td><tt>long</tt></td><td>Signed 64-bit value</td></tr>
604 <tr><td><tt>double</tt></td><td>64-bit floating point value</td></tr>
612 <!-- _______________________________________________________________________ -->
613 <div class="doc_subsubsection"> <a name="t_classifications">Type
614 Classifications</a> </div>
615 <div class="doc_text">
616 <p>These different primitive types fall into a few useful
619 <table border="1" cellspacing="0" cellpadding="4">
621 <tr><th>Classification</th><th>Types</th></tr>
623 <td><a name="t_signed">signed</a></td>
624 <td><tt>sbyte, short, int, long, float, double</tt></td>
627 <td><a name="t_unsigned">unsigned</a></td>
628 <td><tt>ubyte, ushort, uint, ulong</tt></td>
631 <td><a name="t_integer">integer</a></td>
632 <td><tt>ubyte, sbyte, ushort, short, uint, int, ulong, long</tt></td>
635 <td><a name="t_integral">integral</a></td>
636 <td><tt>bool, ubyte, sbyte, ushort, short, uint, int, ulong, long</tt>
640 <td><a name="t_floating">floating point</a></td>
641 <td><tt>float, double</tt></td>
644 <td><a name="t_firstclass">first class</a></td>
645 <td><tt>bool, ubyte, sbyte, ushort, short, uint, int, ulong, long,<br>
646 float, double, <a href="#t_pointer">pointer</a>,
647 <a href="#t_packed">packed</a></tt></td>
652 <p>The <a href="#t_firstclass">first class</a> types are perhaps the
653 most important. Values of these types are the only ones which can be
654 produced by instructions, passed as arguments, or used as operands to
655 instructions. This means that all structures and arrays must be
656 manipulated either by pointer or by component.</p>
659 <!-- ======================================================================= -->
660 <div class="doc_subsection"> <a name="t_derived">Derived Types</a> </div>
662 <div class="doc_text">
664 <p>The real power in LLVM comes from the derived types in the system.
665 This is what allows a programmer to represent arrays, functions,
666 pointers, and other useful types. Note that these derived types may be
667 recursive: For example, it is possible to have a two dimensional array.</p>
671 <!-- _______________________________________________________________________ -->
672 <div class="doc_subsubsection"> <a name="t_array">Array Type</a> </div>
674 <div class="doc_text">
678 <p>The array type is a very simple derived type that arranges elements
679 sequentially in memory. The array type requires a size (number of
680 elements) and an underlying data type.</p>
685 [<# elements> x <elementtype>]
688 <p>The number of elements is a constant integer value; elementtype may
689 be any type with a size.</p>
692 <table class="layout">
695 <tt>[40 x int ]</tt><br/>
696 <tt>[41 x int ]</tt><br/>
697 <tt>[40 x uint]</tt><br/>
700 Array of 40 integer values.<br/>
701 Array of 41 integer values.<br/>
702 Array of 40 unsigned integer values.<br/>
706 <p>Here are some examples of multidimensional arrays:</p>
707 <table class="layout">
710 <tt>[3 x [4 x int]]</tt><br/>
711 <tt>[12 x [10 x float]]</tt><br/>
712 <tt>[2 x [3 x [4 x uint]]]</tt><br/>
715 3x4 array of integer values.<br/>
716 12x10 array of single precision floating point values.<br/>
717 2x3x4 array of unsigned integer values.<br/>
723 <!-- _______________________________________________________________________ -->
724 <div class="doc_subsubsection"> <a name="t_function">Function Type</a> </div>
725 <div class="doc_text">
727 <p>The function type can be thought of as a function signature. It
728 consists of a return type and a list of formal parameter types.
729 Function types are usually used to build virtual function tables
730 (which are structures of pointers to functions), for indirect function
731 calls, and when defining a function.</p>
733 The return type of a function type cannot be an aggregate type.
736 <pre> <returntype> (<parameter list>)<br></pre>
737 <p>Where '<tt><parameter list></tt>' is a comma-separated list of type
738 specifiers. Optionally, the parameter list may include a type <tt>...</tt>,
739 which indicates that the function takes a variable number of arguments.
740 Variable argument functions can access their arguments with the <a
741 href="#int_varargs">variable argument handling intrinsic</a> functions.</p>
743 <table class="layout">
746 <tt>int (int)</tt> <br/>
747 <tt>float (int, int *) *</tt><br/>
748 <tt>int (sbyte *, ...)</tt><br/>
751 function taking an <tt>int</tt>, returning an <tt>int</tt><br/>
752 <a href="#t_pointer">Pointer</a> to a function that takes an
753 <tt>int</tt> and a <a href="#t_pointer">pointer</a> to <tt>int</tt>,
754 returning <tt>float</tt>.<br/>
755 A vararg function that takes at least one <a href="#t_pointer">pointer</a>
756 to <tt>sbyte</tt> (signed char in C), which returns an integer. This is
757 the signature for <tt>printf</tt> in LLVM.<br/>
763 <!-- _______________________________________________________________________ -->
764 <div class="doc_subsubsection"> <a name="t_struct">Structure Type</a> </div>
765 <div class="doc_text">
767 <p>The structure type is used to represent a collection of data members
768 together in memory. The packing of the field types is defined to match
769 the ABI of the underlying processor. The elements of a structure may
770 be any type that has a size.</p>
771 <p>Structures are accessed using '<tt><a href="#i_load">load</a></tt>
772 and '<tt><a href="#i_store">store</a></tt>' by getting a pointer to a
773 field with the '<tt><a href="#i_getelementptr">getelementptr</a></tt>'
776 <pre> { <type list> }<br></pre>
778 <table class="layout">
781 <tt>{ int, int, int }</tt><br/>
782 <tt>{ float, int (int) * }</tt><br/>
785 a triple of three <tt>int</tt> values<br/>
786 A pair, where the first element is a <tt>float</tt> and the second element
787 is a <a href="#t_pointer">pointer</a> to a <a href="#t_function">function</a>
788 that takes an <tt>int</tt>, returning an <tt>int</tt>.<br/>
794 <!-- _______________________________________________________________________ -->
795 <div class="doc_subsubsection"> <a name="t_pointer">Pointer Type</a> </div>
796 <div class="doc_text">
798 <p>As in many languages, the pointer type represents a pointer or
799 reference to another object, which must live in memory.</p>
801 <pre> <type> *<br></pre>
803 <table class="layout">
806 <tt>[4x int]*</tt><br/>
807 <tt>int (int *) *</tt><br/>
810 A <a href="#t_pointer">pointer</a> to <a href="#t_array">array</a> of
811 four <tt>int</tt> values<br/>
812 A <a href="#t_pointer">pointer</a> to a <a
813 href="#t_function">function</a> that takes an <tt>int*</tt>, returning an
820 <!-- _______________________________________________________________________ -->
821 <div class="doc_subsubsection"> <a name="t_packed">Packed Type</a> </div>
822 <div class="doc_text">
826 <p>A packed type is a simple derived type that represents a vector
827 of elements. Packed types are used when multiple primitive data
828 are operated in parallel using a single instruction (SIMD).
829 A packed type requires a size (number of
830 elements) and an underlying primitive data type. Packed types are
831 considered <a href="#t_firstclass">first class</a>.</p>
836 < <# elements> x <elementtype> >
839 <p>The number of elements is a constant integer value; elementtype may
840 be any integral or floating point type.</p>
844 <table class="layout">
847 <tt><4 x int></tt><br/>
848 <tt><8 x float></tt><br/>
849 <tt><2 x uint></tt><br/>
852 Packed vector of 4 integer values.<br/>
853 Packed vector of 8 floating-point values.<br/>
854 Packed vector of 2 unsigned integer values.<br/>
860 <!-- _______________________________________________________________________ -->
861 <div class="doc_subsubsection"> <a name="t_opaque">Opaque Type</a> </div>
862 <div class="doc_text">
866 <p>Opaque types are used to represent unknown types in the system. This
867 corresponds (for example) to the C notion of a foward declared structure type.
868 In LLVM, opaque types can eventually be resolved to any type (not just a
879 <table class="layout">
892 <!-- *********************************************************************** -->
893 <div class="doc_section"> <a name="constants">Constants</a> </div>
894 <!-- *********************************************************************** -->
896 <div class="doc_text">
898 <p>LLVM has several different basic types of constants. This section describes
899 them all and their syntax.</p>
903 <!-- ======================================================================= -->
904 <div class="doc_subsection"><a name="simpleconstants">Simple Constants</a></div>
906 <div class="doc_text">
909 <dt><b>Boolean constants</b></dt>
911 <dd>The two strings '<tt>true</tt>' and '<tt>false</tt>' are both valid
912 constants of the <tt><a href="#t_primitive">bool</a></tt> type.
915 <dt><b>Integer constants</b></dt>
917 <dd>Standard integers (such as '4') are constants of the <a
918 href="#t_integer">integer</a> type. Negative numbers may be used with signed
922 <dt><b>Floating point constants</b></dt>
924 <dd>Floating point constants use standard decimal notation (e.g. 123.421),
925 exponential notation (e.g. 1.23421e+2), or a more precise hexadecimal
926 notation (see below). Floating point constants must have a <a
927 href="#t_floating">floating point</a> type. </dd>
929 <dt><b>Null pointer constants</b></dt>
931 <dd>The identifier '<tt>null</tt>' is recognized as a null pointer constant
932 and must be of <a href="#t_pointer">pointer type</a>.</dd>
936 <p>The one non-intuitive notation for constants is the optional hexadecimal form
937 of floating point constants. For example, the form '<tt>double
938 0x432ff973cafa8000</tt>' is equivalent to (but harder to read than) '<tt>double
939 4.5e+15</tt>'. The only time hexadecimal floating point constants are required
940 (and the only time that they are generated by the disassembler) is when a
941 floating point constant must be emitted but it cannot be represented as a
942 decimal floating point number. For example, NaN's, infinities, and other
943 special values are represented in their IEEE hexadecimal format so that
944 assembly and disassembly do not cause any bits to change in the constants.</p>
948 <!-- ======================================================================= -->
949 <div class="doc_subsection"><a name="aggregateconstants">Aggregate Constants</a>
952 <div class="doc_text">
953 <p>Aggregate constants arise from aggregation of simple constants
954 and smaller aggregate constants.</p>
957 <dt><b>Structure constants</b></dt>
959 <dd>Structure constants are represented with notation similar to structure
960 type definitions (a comma separated list of elements, surrounded by braces
961 (<tt>{}</tt>)). For example: "<tt>{ int 4, float 17.0, int* %G }</tt>",
962 where "<tt>%G</tt>" is declared as "<tt>%G = external global int</tt>". Structure constants
963 must have <a href="#t_struct">structure type</a>, and the number and
964 types of elements must match those specified by the type.
967 <dt><b>Array constants</b></dt>
969 <dd>Array constants are represented with notation similar to array type
970 definitions (a comma separated list of elements, surrounded by square brackets
971 (<tt>[]</tt>)). For example: "<tt>[ int 42, int 11, int 74 ]</tt>". Array
972 constants must have <a href="#t_array">array type</a>, and the number and
973 types of elements must match those specified by the type.
976 <dt><b>Packed constants</b></dt>
978 <dd>Packed constants are represented with notation similar to packed type
979 definitions (a comma separated list of elements, surrounded by
980 less-than/greater-than's (<tt><></tt>)). For example: "<tt>< int 42,
981 int 11, int 74, int 100 ></tt>". Packed constants must have <a
982 href="#t_packed">packed type</a>, and the number and types of elements must
983 match those specified by the type.
986 <dt><b>Zero initialization</b></dt>
988 <dd>The string '<tt>zeroinitializer</tt>' can be used to zero initialize a
989 value to zero of <em>any</em> type, including scalar and aggregate types.
990 This is often used to avoid having to print large zero initializers (e.g. for
991 large arrays), and is always exactly equivalent to using explicit zero
998 <!-- ======================================================================= -->
999 <div class="doc_subsection">
1000 <a name="globalconstants">Global Variable and Function Addresses</a>
1003 <div class="doc_text">
1005 <p>The addresses of <a href="#globalvars">global variables</a> and <a
1006 href="#functionstructure">functions</a> are always implicitly valid (link-time)
1007 constants. These constants are explicitly referenced when the <a
1008 href="#identifiers">identifier for the global</a> is used and always have <a
1009 href="#t_pointer">pointer</a> type. For example, the following is a legal LLVM
1015 %Z = global [2 x int*] [ int* %X, int* %Y ]
1020 <!-- ======================================================================= -->
1021 <div class="doc_subsection"><a name="undefvalues">Undefined Values</a></div>
1022 <div class="doc_text">
1023 <p>The string '<tt>undef</tt>' is recognized as a type-less constant that has
1024 no specific value. Undefined values may be of any type and be used anywhere
1025 a constant is permitted.</p>
1027 <p>Undefined values indicate to the compiler that the program is well defined
1028 no matter what value is used, giving the compiler more freedom to optimize.
1032 <!-- ======================================================================= -->
1033 <div class="doc_subsection"><a name="constantexprs">Constant Expressions</a>
1036 <div class="doc_text">
1038 <p>Constant expressions are used to allow expressions involving other constants
1039 to be used as constants. Constant expressions may be of any <a
1040 href="#t_firstclass">first class</a> type and may involve any LLVM operation
1041 that does not have side effects (e.g. load and call are not supported). The
1042 following is the syntax for constant expressions:</p>
1045 <dt><b><tt>cast ( CST to TYPE )</tt></b></dt>
1047 <dd>Cast a constant to another type.</dd>
1049 <dt><b><tt>getelementptr ( CSTPTR, IDX0, IDX1, ... )</tt></b></dt>
1051 <dd>Perform the <a href="#i_getelementptr">getelementptr operation</a> on
1052 constants. As with the <a href="#i_getelementptr">getelementptr</a>
1053 instruction, the index list may have zero or more indexes, which are required
1054 to make sense for the type of "CSTPTR".</dd>
1056 <dt><b><tt>OPCODE ( LHS, RHS )</tt></b></dt>
1058 <dd>Perform the specified operation of the LHS and RHS constants. OPCODE may
1059 be any of the <a href="#binaryops">binary</a> or <a href="#bitwiseops">bitwise
1060 binary</a> operations. The constraints on operands are the same as those for
1061 the corresponding instruction (e.g. no bitwise operations on floating point
1062 values are allowed).</dd>
1066 <!-- *********************************************************************** -->
1067 <div class="doc_section"> <a name="instref">Instruction Reference</a> </div>
1068 <!-- *********************************************************************** -->
1070 <div class="doc_text">
1072 <p>The LLVM instruction set consists of several different
1073 classifications of instructions: <a href="#terminators">terminator
1074 instructions</a>, <a href="#binaryops">binary instructions</a>,
1075 <a href="#bitwiseops">bitwise binary instructions</a>, <a
1076 href="#memoryops">memory instructions</a>, and <a href="#otherops">other
1077 instructions</a>.</p>
1081 <!-- ======================================================================= -->
1082 <div class="doc_subsection"> <a name="terminators">Terminator
1083 Instructions</a> </div>
1085 <div class="doc_text">
1087 <p>As mentioned <a href="#functionstructure">previously</a>, every
1088 basic block in a program ends with a "Terminator" instruction, which
1089 indicates which block should be executed after the current block is
1090 finished. These terminator instructions typically yield a '<tt>void</tt>'
1091 value: they produce control flow, not values (the one exception being
1092 the '<a href="#i_invoke"><tt>invoke</tt></a>' instruction).</p>
1093 <p>There are six different terminator instructions: the '<a
1094 href="#i_ret"><tt>ret</tt></a>' instruction, the '<a href="#i_br"><tt>br</tt></a>'
1095 instruction, the '<a href="#i_switch"><tt>switch</tt></a>' instruction,
1096 the '<a href="#i_invoke"><tt>invoke</tt></a>' instruction, the '<a
1097 href="#i_unwind"><tt>unwind</tt></a>' instruction, and the '<a
1098 href="#i_unreachable"><tt>unreachable</tt></a>' instruction.</p>
1102 <!-- _______________________________________________________________________ -->
1103 <div class="doc_subsubsection"> <a name="i_ret">'<tt>ret</tt>'
1104 Instruction</a> </div>
1105 <div class="doc_text">
1107 <pre> ret <type> <value> <i>; Return a value from a non-void function</i>
1108 ret void <i>; Return from void function</i>
1111 <p>The '<tt>ret</tt>' instruction is used to return control flow (and a
1112 value) from a function back to the caller.</p>
1113 <p>There are two forms of the '<tt>ret</tt>' instruction: one that
1114 returns a value and then causes control flow, and one that just causes
1115 control flow to occur.</p>
1117 <p>The '<tt>ret</tt>' instruction may return any '<a
1118 href="#t_firstclass">first class</a>' type. Notice that a function is
1119 not <a href="#wellformed">well formed</a> if there exists a '<tt>ret</tt>'
1120 instruction inside of the function that returns a value that does not
1121 match the return type of the function.</p>
1123 <p>When the '<tt>ret</tt>' instruction is executed, control flow
1124 returns back to the calling function's context. If the caller is a "<a
1125 href="#i_call"><tt>call</tt></a>" instruction, execution continues at
1126 the instruction after the call. If the caller was an "<a
1127 href="#i_invoke"><tt>invoke</tt></a>" instruction, execution continues
1128 at the beginning of the "normal" destination block. If the instruction
1129 returns a value, that value shall set the call or invoke instruction's
1132 <pre> ret int 5 <i>; Return an integer value of 5</i>
1133 ret void <i>; Return from a void function</i>
1136 <!-- _______________________________________________________________________ -->
1137 <div class="doc_subsubsection"> <a name="i_br">'<tt>br</tt>' Instruction</a> </div>
1138 <div class="doc_text">
1140 <pre> br bool <cond>, label <iftrue>, label <iffalse><br> br label <dest> <i>; Unconditional branch</i>
1143 <p>The '<tt>br</tt>' instruction is used to cause control flow to
1144 transfer to a different basic block in the current function. There are
1145 two forms of this instruction, corresponding to a conditional branch
1146 and an unconditional branch.</p>
1148 <p>The conditional branch form of the '<tt>br</tt>' instruction takes a
1149 single '<tt>bool</tt>' value and two '<tt>label</tt>' values. The
1150 unconditional form of the '<tt>br</tt>' instruction takes a single '<tt>label</tt>'
1151 value as a target.</p>
1153 <p>Upon execution of a conditional '<tt>br</tt>' instruction, the '<tt>bool</tt>'
1154 argument is evaluated. If the value is <tt>true</tt>, control flows
1155 to the '<tt>iftrue</tt>' <tt>label</tt> argument. If "cond" is <tt>false</tt>,
1156 control flows to the '<tt>iffalse</tt>' <tt>label</tt> argument.</p>
1158 <pre>Test:<br> %cond = <a href="#i_setcc">seteq</a> int %a, %b<br> br bool %cond, label %IfEqual, label %IfUnequal<br>IfEqual:<br> <a
1159 href="#i_ret">ret</a> int 1<br>IfUnequal:<br> <a href="#i_ret">ret</a> int 0<br></pre>
1161 <!-- _______________________________________________________________________ -->
1162 <div class="doc_subsubsection">
1163 <a name="i_switch">'<tt>switch</tt>' Instruction</a>
1166 <div class="doc_text">
1170 switch <intty> <value>, label <defaultdest> [ <intty> <val>, label <dest> ... ]
1175 <p>The '<tt>switch</tt>' instruction is used to transfer control flow to one of
1176 several different places. It is a generalization of the '<tt>br</tt>'
1177 instruction, allowing a branch to occur to one of many possible
1183 <p>The '<tt>switch</tt>' instruction uses three parameters: an integer
1184 comparison value '<tt>value</tt>', a default '<tt>label</tt>' destination, and
1185 an array of pairs of comparison value constants and '<tt>label</tt>'s. The
1186 table is not allowed to contain duplicate constant entries.</p>
1190 <p>The <tt>switch</tt> instruction specifies a table of values and
1191 destinations. When the '<tt>switch</tt>' instruction is executed, this
1192 table is searched for the given value. If the value is found, control flow is
1193 transfered to the corresponding destination; otherwise, control flow is
1194 transfered to the default destination.</p>
1196 <h5>Implementation:</h5>
1198 <p>Depending on properties of the target machine and the particular
1199 <tt>switch</tt> instruction, this instruction may be code generated in different
1200 ways. For example, it could be generated as a series of chained conditional
1201 branches or with a lookup table.</p>
1206 <i>; Emulate a conditional br instruction</i>
1207 %Val = <a href="#i_cast">cast</a> bool %value to int
1208 switch int %Val, label %truedest [int 0, label %falsedest ]
1210 <i>; Emulate an unconditional br instruction</i>
1211 switch uint 0, label %dest [ ]
1213 <i>; Implement a jump table:</i>
1214 switch uint %val, label %otherwise [ uint 0, label %onzero
1215 uint 1, label %onone
1216 uint 2, label %ontwo ]
1220 <!-- _______________________________________________________________________ -->
1221 <div class="doc_subsubsection">
1222 <a name="i_invoke">'<tt>invoke</tt>' Instruction</a>
1225 <div class="doc_text">
1230 <result> = invoke [<a href="#callingconv">cconv</a>] <ptr to function ty> %<function ptr val>(<function args>)
1231 to label <normal label> except label <exception label>
1236 <p>The '<tt>invoke</tt>' instruction causes control to transfer to a specified
1237 function, with the possibility of control flow transfer to either the
1238 '<tt>normal</tt>' label or the
1239 '<tt>exception</tt>' label. If the callee function returns with the
1240 "<tt><a href="#i_ret">ret</a></tt>" instruction, control flow will return to the
1241 "normal" label. If the callee (or any indirect callees) returns with the "<a
1242 href="#i_unwind"><tt>unwind</tt></a>" instruction, control is interrupted and
1243 continued at the dynamically nearest "exception" label.</p>
1247 <p>This instruction requires several arguments:</p>
1251 The optional "cconv" marker indicates which <a href="callingconv">calling
1252 convention</a> the call should use. If none is specified, the call defaults
1253 to using C calling conventions.
1255 <li>'<tt>ptr to function ty</tt>': shall be the signature of the pointer to
1256 function value being invoked. In most cases, this is a direct function
1257 invocation, but indirect <tt>invoke</tt>s are just as possible, branching off
1258 an arbitrary pointer to function value.
1261 <li>'<tt>function ptr val</tt>': An LLVM value containing a pointer to a
1262 function to be invoked. </li>
1264 <li>'<tt>function args</tt>': argument list whose types match the function
1265 signature argument types. If the function signature indicates the function
1266 accepts a variable number of arguments, the extra arguments can be
1269 <li>'<tt>normal label</tt>': the label reached when the called function
1270 executes a '<tt><a href="#i_ret">ret</a></tt>' instruction. </li>
1272 <li>'<tt>exception label</tt>': the label reached when a callee returns with
1273 the <a href="#i_unwind"><tt>unwind</tt></a> instruction. </li>
1279 <p>This instruction is designed to operate as a standard '<tt><a
1280 href="#i_call">call</a></tt>' instruction in most regards. The primary
1281 difference is that it establishes an association with a label, which is used by
1282 the runtime library to unwind the stack.</p>
1284 <p>This instruction is used in languages with destructors to ensure that proper
1285 cleanup is performed in the case of either a <tt>longjmp</tt> or a thrown
1286 exception. Additionally, this is important for implementation of
1287 '<tt>catch</tt>' clauses in high-level languages that support them.</p>
1291 %retval = invoke int %Test(int 15) to label %Continue
1292 except label %TestCleanup <i>; {int}:retval set</i>
1293 %retval = invoke <a href="#callingconv">coldcc</a> int %Test(int 15) to label %Continue
1294 except label %TestCleanup <i>; {int}:retval set</i>
1299 <!-- _______________________________________________________________________ -->
1301 <div class="doc_subsubsection"> <a name="i_unwind">'<tt>unwind</tt>'
1302 Instruction</a> </div>
1304 <div class="doc_text">
1313 <p>The '<tt>unwind</tt>' instruction unwinds the stack, continuing control flow
1314 at the first callee in the dynamic call stack which used an <a
1315 href="#i_invoke"><tt>invoke</tt></a> instruction to perform the call. This is
1316 primarily used to implement exception handling.</p>
1320 <p>The '<tt>unwind</tt>' intrinsic causes execution of the current function to
1321 immediately halt. The dynamic call stack is then searched for the first <a
1322 href="#i_invoke"><tt>invoke</tt></a> instruction on the call stack. Once found,
1323 execution continues at the "exceptional" destination block specified by the
1324 <tt>invoke</tt> instruction. If there is no <tt>invoke</tt> instruction in the
1325 dynamic call chain, undefined behavior results.</p>
1328 <!-- _______________________________________________________________________ -->
1330 <div class="doc_subsubsection"> <a name="i_unreachable">'<tt>unreachable</tt>'
1331 Instruction</a> </div>
1333 <div class="doc_text">
1342 <p>The '<tt>unreachable</tt>' instruction has no defined semantics. This
1343 instruction is used to inform the optimizer that a particular portion of the
1344 code is not reachable. This can be used to indicate that the code after a
1345 no-return function cannot be reached, and other facts.</p>
1349 <p>The '<tt>unreachable</tt>' instruction has no defined semantics.</p>
1354 <!-- ======================================================================= -->
1355 <div class="doc_subsection"> <a name="binaryops">Binary Operations</a> </div>
1356 <div class="doc_text">
1357 <p>Binary operators are used to do most of the computation in a
1358 program. They require two operands, execute an operation on them, and
1359 produce a single value. The operands might represent
1360 multiple data, as is the case with the <a href="#t_packed">packed</a> data type.
1361 The result value of a binary operator is not
1362 necessarily the same type as its operands.</p>
1363 <p>There are several different binary operators:</p>
1365 <!-- _______________________________________________________________________ -->
1366 <div class="doc_subsubsection"> <a name="i_add">'<tt>add</tt>'
1367 Instruction</a> </div>
1368 <div class="doc_text">
1370 <pre> <result> = add <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1373 <p>The '<tt>add</tt>' instruction returns the sum of its two operands.</p>
1375 <p>The two arguments to the '<tt>add</tt>' instruction must be either <a
1376 href="#t_integer">integer</a> or <a href="#t_floating">floating point</a> values.
1377 This instruction can also take <a href="#t_packed">packed</a> versions of the values.
1378 Both arguments must have identical types.</p>
1380 <p>The value produced is the integer or floating point sum of the two
1383 <pre> <result> = add int 4, %var <i>; yields {int}:result = 4 + %var</i>
1386 <!-- _______________________________________________________________________ -->
1387 <div class="doc_subsubsection"> <a name="i_sub">'<tt>sub</tt>'
1388 Instruction</a> </div>
1389 <div class="doc_text">
1391 <pre> <result> = sub <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1394 <p>The '<tt>sub</tt>' instruction returns the difference of its two
1396 <p>Note that the '<tt>sub</tt>' instruction is used to represent the '<tt>neg</tt>'
1397 instruction present in most other intermediate representations.</p>
1399 <p>The two arguments to the '<tt>sub</tt>' instruction must be either <a
1400 href="#t_integer">integer</a> or <a href="#t_floating">floating point</a>
1402 This instruction can also take <a href="#t_packed">packed</a> versions of the values.
1403 Both arguments must have identical types.</p>
1405 <p>The value produced is the integer or floating point difference of
1406 the two operands.</p>
1408 <pre> <result> = sub int 4, %var <i>; yields {int}:result = 4 - %var</i>
1409 <result> = sub int 0, %val <i>; yields {int}:result = -%var</i>
1412 <!-- _______________________________________________________________________ -->
1413 <div class="doc_subsubsection"> <a name="i_mul">'<tt>mul</tt>'
1414 Instruction</a> </div>
1415 <div class="doc_text">
1417 <pre> <result> = mul <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1420 <p>The '<tt>mul</tt>' instruction returns the product of its two
1423 <p>The two arguments to the '<tt>mul</tt>' instruction must be either <a
1424 href="#t_integer">integer</a> or <a href="#t_floating">floating point</a>
1426 This instruction can also take <a href="#t_packed">packed</a> versions of the values.
1427 Both arguments must have identical types.</p>
1429 <p>The value produced is the integer or floating point product of the
1431 <p>There is no signed vs unsigned multiplication. The appropriate
1432 action is taken based on the type of the operand.</p>
1434 <pre> <result> = mul int 4, %var <i>; yields {int}:result = 4 * %var</i>
1437 <!-- _______________________________________________________________________ -->
1438 <div class="doc_subsubsection"> <a name="i_div">'<tt>div</tt>'
1439 Instruction</a> </div>
1440 <div class="doc_text">
1442 <pre> <result> = div <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1445 <p>The '<tt>div</tt>' instruction returns the quotient of its two
1448 <p>The two arguments to the '<tt>div</tt>' instruction must be either <a
1449 href="#t_integer">integer</a> or <a href="#t_floating">floating point</a>
1451 This instruction can also take <a href="#t_packed">packed</a> versions of the values.
1452 Both arguments must have identical types.</p>
1454 <p>The value produced is the integer or floating point quotient of the
1457 <pre> <result> = div int 4, %var <i>; yields {int}:result = 4 / %var</i>
1460 <!-- _______________________________________________________________________ -->
1461 <div class="doc_subsubsection"> <a name="i_rem">'<tt>rem</tt>'
1462 Instruction</a> </div>
1463 <div class="doc_text">
1465 <pre> <result> = rem <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1468 <p>The '<tt>rem</tt>' instruction returns the remainder from the
1469 division of its two operands.</p>
1471 <p>The two arguments to the '<tt>rem</tt>' instruction must be either <a
1472 href="#t_integer">integer</a> or <a href="#t_floating">floating point</a>
1474 This instruction can also take <a href="#t_packed">packed</a> versions of the values.
1475 Both arguments must have identical types.</p>
1477 <p>This returns the <i>remainder</i> of a division (where the result
1478 has the same sign as the divisor), not the <i>modulus</i> (where the
1479 result has the same sign as the dividend) of a value. For more
1480 information about the difference, see: <a
1481 href="http://mathforum.org/dr.math/problems/anne.4.28.99.html">The
1484 <pre> <result> = rem int 4, %var <i>; yields {int}:result = 4 % %var</i>
1487 <!-- _______________________________________________________________________ -->
1488 <div class="doc_subsubsection"> <a name="i_setcc">'<tt>set<i>cc</i></tt>'
1489 Instructions</a> </div>
1490 <div class="doc_text">
1492 <pre> <result> = seteq <ty> <var1>, <var2> <i>; yields {bool}:result</i>
1493 <result> = setne <ty> <var1>, <var2> <i>; yields {bool}:result</i>
1494 <result> = setlt <ty> <var1>, <var2> <i>; yields {bool}:result</i>
1495 <result> = setgt <ty> <var1>, <var2> <i>; yields {bool}:result</i>
1496 <result> = setle <ty> <var1>, <var2> <i>; yields {bool}:result</i>
1497 <result> = setge <ty> <var1>, <var2> <i>; yields {bool}:result</i>
1500 <p>The '<tt>set<i>cc</i></tt>' family of instructions returns a boolean
1501 value based on a comparison of their two operands.</p>
1503 <p>The two arguments to the '<tt>set<i>cc</i></tt>' instructions must
1504 be of <a href="#t_firstclass">first class</a> type (it is not possible
1505 to compare '<tt>label</tt>'s, '<tt>array</tt>'s, '<tt>structure</tt>'
1506 or '<tt>void</tt>' values, etc...). Both arguments must have identical
1509 <p>The '<tt>seteq</tt>' instruction yields a <tt>true</tt> '<tt>bool</tt>'
1510 value if both operands are equal.<br>
1511 The '<tt>setne</tt>' instruction yields a <tt>true</tt> '<tt>bool</tt>'
1512 value if both operands are unequal.<br>
1513 The '<tt>setlt</tt>' instruction yields a <tt>true</tt> '<tt>bool</tt>'
1514 value if the first operand is less than the second operand.<br>
1515 The '<tt>setgt</tt>' instruction yields a <tt>true</tt> '<tt>bool</tt>'
1516 value if the first operand is greater than the second operand.<br>
1517 The '<tt>setle</tt>' instruction yields a <tt>true</tt> '<tt>bool</tt>'
1518 value if the first operand is less than or equal to the second operand.<br>
1519 The '<tt>setge</tt>' instruction yields a <tt>true</tt> '<tt>bool</tt>'
1520 value if the first operand is greater than or equal to the second
1523 <pre> <result> = seteq int 4, 5 <i>; yields {bool}:result = false</i>
1524 <result> = setne float 4, 5 <i>; yields {bool}:result = true</i>
1525 <result> = setlt uint 4, 5 <i>; yields {bool}:result = true</i>
1526 <result> = setgt sbyte 4, 5 <i>; yields {bool}:result = false</i>
1527 <result> = setle sbyte 4, 5 <i>; yields {bool}:result = true</i>
1528 <result> = setge sbyte 4, 5 <i>; yields {bool}:result = false</i>
1531 <!-- ======================================================================= -->
1532 <div class="doc_subsection"> <a name="bitwiseops">Bitwise Binary
1533 Operations</a> </div>
1534 <div class="doc_text">
1535 <p>Bitwise binary operators are used to do various forms of
1536 bit-twiddling in a program. They are generally very efficient
1537 instructions and can commonly be strength reduced from other
1538 instructions. They require two operands, execute an operation on them,
1539 and produce a single value. The resulting value of the bitwise binary
1540 operators is always the same type as its first operand.</p>
1542 <!-- _______________________________________________________________________ -->
1543 <div class="doc_subsubsection"> <a name="i_and">'<tt>and</tt>'
1544 Instruction</a> </div>
1545 <div class="doc_text">
1547 <pre> <result> = and <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1550 <p>The '<tt>and</tt>' instruction returns the bitwise logical and of
1551 its two operands.</p>
1553 <p>The two arguments to the '<tt>and</tt>' instruction must be <a
1554 href="#t_integral">integral</a> values. Both arguments must have
1555 identical types.</p>
1557 <p>The truth table used for the '<tt>and</tt>' instruction is:</p>
1559 <div style="align: center">
1560 <table border="1" cellspacing="0" cellpadding="4">
1591 <pre> <result> = and int 4, %var <i>; yields {int}:result = 4 & %var</i>
1592 <result> = and int 15, 40 <i>; yields {int}:result = 8</i>
1593 <result> = and int 4, 8 <i>; yields {int}:result = 0</i>
1596 <!-- _______________________________________________________________________ -->
1597 <div class="doc_subsubsection"> <a name="i_or">'<tt>or</tt>' Instruction</a> </div>
1598 <div class="doc_text">
1600 <pre> <result> = or <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1603 <p>The '<tt>or</tt>' instruction returns the bitwise logical inclusive
1604 or of its two operands.</p>
1606 <p>The two arguments to the '<tt>or</tt>' instruction must be <a
1607 href="#t_integral">integral</a> values. Both arguments must have
1608 identical types.</p>
1610 <p>The truth table used for the '<tt>or</tt>' instruction is:</p>
1612 <div style="align: center">
1613 <table border="1" cellspacing="0" cellpadding="4">
1644 <pre> <result> = or int 4, %var <i>; yields {int}:result = 4 | %var</i>
1645 <result> = or int 15, 40 <i>; yields {int}:result = 47</i>
1646 <result> = or int 4, 8 <i>; yields {int}:result = 12</i>
1649 <!-- _______________________________________________________________________ -->
1650 <div class="doc_subsubsection"> <a name="i_xor">'<tt>xor</tt>'
1651 Instruction</a> </div>
1652 <div class="doc_text">
1654 <pre> <result> = xor <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1657 <p>The '<tt>xor</tt>' instruction returns the bitwise logical exclusive
1658 or of its two operands. The <tt>xor</tt> is used to implement the
1659 "one's complement" operation, which is the "~" operator in C.</p>
1661 <p>The two arguments to the '<tt>xor</tt>' instruction must be <a
1662 href="#t_integral">integral</a> values. Both arguments must have
1663 identical types.</p>
1665 <p>The truth table used for the '<tt>xor</tt>' instruction is:</p>
1667 <div style="align: center">
1668 <table border="1" cellspacing="0" cellpadding="4">
1700 <pre> <result> = xor int 4, %var <i>; yields {int}:result = 4 ^ %var</i>
1701 <result> = xor int 15, 40 <i>; yields {int}:result = 39</i>
1702 <result> = xor int 4, 8 <i>; yields {int}:result = 12</i>
1703 <result> = xor int %V, -1 <i>; yields {int}:result = ~%V</i>
1706 <!-- _______________________________________________________________________ -->
1707 <div class="doc_subsubsection"> <a name="i_shl">'<tt>shl</tt>'
1708 Instruction</a> </div>
1709 <div class="doc_text">
1711 <pre> <result> = shl <ty> <var1>, ubyte <var2> <i>; yields {ty}:result</i>
1714 <p>The '<tt>shl</tt>' instruction returns the first operand shifted to
1715 the left a specified number of bits.</p>
1717 <p>The first argument to the '<tt>shl</tt>' instruction must be an <a
1718 href="#t_integer">integer</a> type. The second argument must be an '<tt>ubyte</tt>'
1721 <p>The value produced is <tt>var1</tt> * 2<sup><tt>var2</tt></sup>.</p>
1723 <pre> <result> = shl int 4, ubyte %var <i>; yields {int}:result = 4 << %var</i>
1724 <result> = shl int 4, ubyte 2 <i>; yields {int}:result = 16</i>
1725 <result> = shl int 1, ubyte 10 <i>; yields {int}:result = 1024</i>
1728 <!-- _______________________________________________________________________ -->
1729 <div class="doc_subsubsection"> <a name="i_shr">'<tt>shr</tt>'
1730 Instruction</a> </div>
1731 <div class="doc_text">
1733 <pre> <result> = shr <ty> <var1>, ubyte <var2> <i>; yields {ty}:result</i>
1736 <p>The '<tt>shr</tt>' instruction returns the first operand shifted to
1737 the right a specified number of bits.</p>
1739 <p>The first argument to the '<tt>shr</tt>' instruction must be an <a
1740 href="#t_integer">integer</a> type. The second argument must be an '<tt>ubyte</tt>'
1743 <p>If the first argument is a <a href="#t_signed">signed</a> type, the
1744 most significant bit is duplicated in the newly free'd bit positions.
1745 If the first argument is unsigned, zero bits shall fill the empty
1748 <pre> <result> = shr int 4, ubyte %var <i>; yields {int}:result = 4 >> %var</i>
1749 <result> = shr uint 4, ubyte 1 <i>; yields {uint}:result = 2</i>
1750 <result> = shr int 4, ubyte 2 <i>; yields {int}:result = 1</i>
1751 <result> = shr sbyte 4, ubyte 3 <i>; yields {sbyte}:result = 0</i>
1752 <result> = shr sbyte -2, ubyte 1 <i>; yields {sbyte}:result = -1</i>
1755 <!-- ======================================================================= -->
1756 <div class="doc_subsection"> <a name="memoryops">Memory Access
1757 Operations</a></div>
1758 <div class="doc_text">
1759 <p>A key design point of an SSA-based representation is how it
1760 represents memory. In LLVM, no memory locations are in SSA form, which
1761 makes things very simple. This section describes how to read, write,
1762 allocate, and free memory in LLVM.</p>
1764 <!-- _______________________________________________________________________ -->
1765 <div class="doc_subsubsection"> <a name="i_malloc">'<tt>malloc</tt>'
1766 Instruction</a> </div>
1767 <div class="doc_text">
1769 <pre> <result> = malloc <type>, uint <NumElements> <i>; yields {type*}:result</i>
1770 <result> = malloc <type> <i>; yields {type*}:result</i>
1773 <p>The '<tt>malloc</tt>' instruction allocates memory from the system
1774 heap and returns a pointer to it.</p>
1776 <p>The '<tt>malloc</tt>' instruction allocates <tt>sizeof(<type>)*NumElements</tt>
1777 bytes of memory from the operating system and returns a pointer of the
1778 appropriate type to the program. The second form of the instruction is
1779 a shorter version of the first instruction that defaults to allocating
1781 <p>'<tt>type</tt>' must be a sized type.</p>
1783 <p>Memory is allocated using the system "<tt>malloc</tt>" function, and
1784 a pointer is returned.</p>
1786 <pre> %array = malloc [4 x ubyte ] <i>; yields {[%4 x ubyte]*}:array</i>
1789 href="#i_add">add</a> uint 2, 2 <i>; yields {uint}:size = uint 4</i>
1790 %array1 = malloc ubyte, uint 4 <i>; yields {ubyte*}:array1</i>
1791 %array2 = malloc [12 x ubyte], uint %size <i>; yields {[12 x ubyte]*}:array2</i>
1794 <!-- _______________________________________________________________________ -->
1795 <div class="doc_subsubsection"> <a name="i_free">'<tt>free</tt>'
1796 Instruction</a> </div>
1797 <div class="doc_text">
1799 <pre> free <type> <value> <i>; yields {void}</i>
1802 <p>The '<tt>free</tt>' instruction returns memory back to the unused
1803 memory heap to be reallocated in the future.</p>
1806 <p>'<tt>value</tt>' shall be a pointer value that points to a value
1807 that was allocated with the '<tt><a href="#i_malloc">malloc</a></tt>'
1810 <p>Access to the memory pointed to by the pointer is no longer defined
1811 after this instruction executes.</p>
1813 <pre> %array = <a href="#i_malloc">malloc</a> [4 x ubyte] <i>; yields {[4 x ubyte]*}:array</i>
1814 free [4 x ubyte]* %array
1817 <!-- _______________________________________________________________________ -->
1818 <div class="doc_subsubsection"> <a name="i_alloca">'<tt>alloca</tt>'
1819 Instruction</a> </div>
1820 <div class="doc_text">
1822 <pre> <result> = alloca <type>, uint <NumElements> <i>; yields {type*}:result</i>
1823 <result> = alloca <type> <i>; yields {type*}:result</i>
1826 <p>The '<tt>alloca</tt>' instruction allocates memory on the current
1827 stack frame of the procedure that is live until the current function
1828 returns to its caller.</p>
1830 <p>The '<tt>alloca</tt>' instruction allocates <tt>sizeof(<type>)*NumElements</tt>
1831 bytes of memory on the runtime stack, returning a pointer of the
1832 appropriate type to the program. The second form of the instruction is
1833 a shorter version of the first that defaults to allocating one element.</p>
1834 <p>'<tt>type</tt>' may be any sized type.</p>
1836 <p>Memory is allocated; a pointer is returned. '<tt>alloca</tt>'d
1837 memory is automatically released when the function returns. The '<tt>alloca</tt>'
1838 instruction is commonly used to represent automatic variables that must
1839 have an address available. When the function returns (either with the <tt><a
1840 href="#i_ret">ret</a></tt> or <tt><a href="#i_unwind">unwind</a></tt>
1841 instructions), the memory is reclaimed.</p>
1843 <pre> %ptr = alloca int <i>; yields {int*}:ptr</i>
1844 %ptr = alloca int, uint 4 <i>; yields {int*}:ptr</i>
1847 <!-- _______________________________________________________________________ -->
1848 <div class="doc_subsubsection"> <a name="i_load">'<tt>load</tt>'
1849 Instruction</a> </div>
1850 <div class="doc_text">
1852 <pre> <result> = load <ty>* <pointer><br> <result> = volatile load <ty>* <pointer><br></pre>
1854 <p>The '<tt>load</tt>' instruction is used to read from memory.</p>
1856 <p>The argument to the '<tt>load</tt>' instruction specifies the memory
1857 address to load from. The pointer must point to a <a
1858 href="#t_firstclass">first class</a> type. If the <tt>load</tt> is
1859 marked as <tt>volatile</tt> then the optimizer is not allowed to modify
1860 the number or order of execution of this <tt>load</tt> with other
1861 volatile <tt>load</tt> and <tt><a href="#i_store">store</a></tt>
1864 <p>The location of memory pointed to is loaded.</p>
1866 <pre> %ptr = <a href="#i_alloca">alloca</a> int <i>; yields {int*}:ptr</i>
1868 href="#i_store">store</a> int 3, int* %ptr <i>; yields {void}</i>
1869 %val = load int* %ptr <i>; yields {int}:val = int 3</i>
1872 <!-- _______________________________________________________________________ -->
1873 <div class="doc_subsubsection"> <a name="i_store">'<tt>store</tt>'
1874 Instruction</a> </div>
1876 <pre> store <ty> <value>, <ty>* <pointer> <i>; yields {void}</i>
1877 volatile store <ty> <value>, <ty>* <pointer> <i>; yields {void}</i>
1880 <p>The '<tt>store</tt>' instruction is used to write to memory.</p>
1882 <p>There are two arguments to the '<tt>store</tt>' instruction: a value
1883 to store and an address to store it into. The type of the '<tt><pointer></tt>'
1884 operand must be a pointer to the type of the '<tt><value></tt>'
1885 operand. If the <tt>store</tt> is marked as <tt>volatile</tt>, then the
1886 optimizer is not allowed to modify the number or order of execution of
1887 this <tt>store</tt> with other volatile <tt>load</tt> and <tt><a
1888 href="#i_store">store</a></tt> instructions.</p>
1890 <p>The contents of memory are updated to contain '<tt><value></tt>'
1891 at the location specified by the '<tt><pointer></tt>' operand.</p>
1893 <pre> %ptr = <a href="#i_alloca">alloca</a> int <i>; yields {int*}:ptr</i>
1895 href="#i_store">store</a> int 3, int* %ptr <i>; yields {void}</i>
1896 %val = load int* %ptr <i>; yields {int}:val = int 3</i>
1898 <!-- _______________________________________________________________________ -->
1899 <div class="doc_subsubsection">
1900 <a name="i_getelementptr">'<tt>getelementptr</tt>' Instruction</a>
1903 <div class="doc_text">
1906 <result> = getelementptr <ty>* <ptrval>{, <ty> <idx>}*
1912 The '<tt>getelementptr</tt>' instruction is used to get the address of a
1913 subelement of an aggregate data structure.</p>
1917 <p>This instruction takes a list of integer constants that indicate what
1918 elements of the aggregate object to index to. The actual types of the arguments
1919 provided depend on the type of the first pointer argument. The
1920 '<tt>getelementptr</tt>' instruction is used to index down through the type
1921 levels of a structure or to a specific index in an array. When indexing into a
1922 structure, only <tt>uint</tt>
1923 integer constants are allowed. When indexing into an array or pointer,
1924 <tt>int</tt> and <tt>long</tt> indexes are allowed of any sign.</p>
1926 <p>For example, let's consider a C code fragment and how it gets
1927 compiled to LLVM:</p>
1941 int *foo(struct ST *s) {
1942 return &s[1].Z.B[5][13];
1946 <p>The LLVM code generated by the GCC frontend is:</p>
1949 %RT = type { sbyte, [10 x [20 x int]], sbyte }
1950 %ST = type { int, double, %RT }
1954 int* %foo(%ST* %s) {
1956 %reg = getelementptr %ST* %s, int 1, uint 2, uint 1, int 5, int 13
1963 <p>The index types specified for the '<tt>getelementptr</tt>' instruction depend
1964 on the pointer type that is being indexed into. <a href="#t_pointer">Pointer</a>
1965 and <a href="#t_array">array</a> types require <tt>uint</tt>, <tt>int</tt>,
1966 <tt>ulong</tt>, or <tt>long</tt> values, and <a href="#t_struct">structure</a>
1967 types require <tt>uint</tt> <b>constants</b>.</p>
1969 <p>In the example above, the first index is indexing into the '<tt>%ST*</tt>'
1970 type, which is a pointer, yielding a '<tt>%ST</tt>' = '<tt>{ int, double, %RT
1971 }</tt>' type, a structure. The second index indexes into the third element of
1972 the structure, yielding a '<tt>%RT</tt>' = '<tt>{ sbyte, [10 x [20 x int]],
1973 sbyte }</tt>' type, another structure. The third index indexes into the second
1974 element of the structure, yielding a '<tt>[10 x [20 x int]]</tt>' type, an
1975 array. The two dimensions of the array are subscripted into, yielding an
1976 '<tt>int</tt>' type. The '<tt>getelementptr</tt>' instruction returns a pointer
1977 to this element, thus computing a value of '<tt>int*</tt>' type.</p>
1979 <p>Note that it is perfectly legal to index partially through a
1980 structure, returning a pointer to an inner element. Because of this,
1981 the LLVM code for the given testcase is equivalent to:</p>
1984 int* %foo(%ST* %s) {
1985 %t1 = getelementptr %ST* %s, int 1 <i>; yields %ST*:%t1</i>
1986 %t2 = getelementptr %ST* %t1, int 0, uint 2 <i>; yields %RT*:%t2</i>
1987 %t3 = getelementptr %RT* %t2, int 0, uint 1 <i>; yields [10 x [20 x int]]*:%t3</i>
1988 %t4 = getelementptr [10 x [20 x int]]* %t3, int 0, int 5 <i>; yields [20 x int]*:%t4</i>
1989 %t5 = getelementptr [20 x int]* %t4, int 0, int 13 <i>; yields int*:%t5</i>
1995 <i>; yields [12 x ubyte]*:aptr</i>
1996 %aptr = getelementptr {int, [12 x ubyte]}* %sptr, long 0, uint 1
2000 <!-- ======================================================================= -->
2001 <div class="doc_subsection"> <a name="otherops">Other Operations</a> </div>
2002 <div class="doc_text">
2003 <p>The instructions in this category are the "miscellaneous"
2004 instructions, which defy better classification.</p>
2006 <!-- _______________________________________________________________________ -->
2007 <div class="doc_subsubsection"> <a name="i_phi">'<tt>phi</tt>'
2008 Instruction</a> </div>
2009 <div class="doc_text">
2011 <pre> <result> = phi <ty> [ <val0>, <label0>], ...<br></pre>
2013 <p>The '<tt>phi</tt>' instruction is used to implement the φ node in
2014 the SSA graph representing the function.</p>
2016 <p>The type of the incoming values are specified with the first type
2017 field. After this, the '<tt>phi</tt>' instruction takes a list of pairs
2018 as arguments, with one pair for each predecessor basic block of the
2019 current block. Only values of <a href="#t_firstclass">first class</a>
2020 type may be used as the value arguments to the PHI node. Only labels
2021 may be used as the label arguments.</p>
2022 <p>There must be no non-phi instructions between the start of a basic
2023 block and the PHI instructions: i.e. PHI instructions must be first in
2026 <p>At runtime, the '<tt>phi</tt>' instruction logically takes on the
2027 value specified by the parameter, depending on which basic block we
2028 came from in the last <a href="#terminators">terminator</a> instruction.</p>
2030 <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>
2033 <!-- _______________________________________________________________________ -->
2034 <div class="doc_subsubsection">
2035 <a name="i_cast">'<tt>cast .. to</tt>' Instruction</a>
2038 <div class="doc_text">
2043 <result> = cast <ty> <value> to <ty2> <i>; yields ty2</i>
2049 The '<tt>cast</tt>' instruction is used as the primitive means to convert
2050 integers to floating point, change data type sizes, and break type safety (by
2058 The '<tt>cast</tt>' instruction takes a value to cast, which must be a first
2059 class value, and a type to cast it to, which must also be a <a
2060 href="#t_firstclass">first class</a> type.
2066 This instruction follows the C rules for explicit casts when determining how the
2067 data being cast must change to fit in its new container.
2071 When casting to bool, any value that would be considered true in the context of
2072 a C '<tt>if</tt>' condition is converted to the boolean '<tt>true</tt>' values,
2073 all else are '<tt>false</tt>'.
2077 When extending an integral value from a type of one signness to another (for
2078 example '<tt>sbyte</tt>' to '<tt>ulong</tt>'), the value is sign-extended if the
2079 <b>source</b> value is signed, and zero-extended if the source value is
2080 unsigned. <tt>bool</tt> values are always zero extended into either zero or
2087 %X = cast int 257 to ubyte <i>; yields ubyte:1</i>
2088 %Y = cast int 123 to bool <i>; yields bool:true</i>
2092 <!-- _______________________________________________________________________ -->
2093 <div class="doc_subsubsection">
2094 <a name="i_select">'<tt>select</tt>' Instruction</a>
2097 <div class="doc_text">
2102 <result> = select bool <cond>, <ty> <val1>, <ty> <val2> <i>; yields ty</i>
2108 The '<tt>select</tt>' instruction is used to choose one value based on a
2109 condition, without branching.
2116 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.
2122 If the boolean condition evaluates to true, the instruction returns the first
2123 value argument; otherwise, it returns the second value argument.
2129 %X = select bool true, ubyte 17, ubyte 42 <i>; yields ubyte:17</i>
2137 <!-- _______________________________________________________________________ -->
2138 <div class="doc_subsubsection">
2139 <a name="i_call">'<tt>call</tt>' Instruction</a>
2142 <div class="doc_text">
2146 <result> = [tail] call [<a href="#callingconv">cconv</a>] <ty>* <fnptrval>(<param list>)
2151 <p>The '<tt>call</tt>' instruction represents a simple function call.</p>
2155 <p>This instruction requires several arguments:</p>
2159 <p>The optional "tail" marker indicates whether the callee function accesses
2160 any allocas or varargs in the caller. If the "tail" marker is present, the
2161 function call is eligible for tail call optimization. Note that calls may
2162 be marked "tail" even if they do not occur before a <a
2163 href="#i_ret"><tt>ret</tt></a> instruction.
2166 <p>The optional "cconv" marker indicates which <a href="callingconv">calling
2167 convention</a> the call should use. If none is specified, the call defaults
2168 to using C calling conventions.
2171 <p>'<tt>ty</tt>': shall be the signature of the pointer to function value
2172 being invoked. The argument types must match the types implied by this
2173 signature. This type can be omitted if the function is not varargs and
2174 if the function type does not return a pointer to a function.</p>
2177 <p>'<tt>fnptrval</tt>': An LLVM value containing a pointer to a function to
2178 be invoked. In most cases, this is a direct function invocation, but
2179 indirect <tt>call</tt>s are just as possible, calling an arbitrary pointer
2180 to function value.</p>
2183 <p>'<tt>function args</tt>': argument list whose types match the
2184 function signature argument types. All arguments must be of
2185 <a href="#t_firstclass">first class</a> type. If the function signature
2186 indicates the function accepts a variable number of arguments, the extra
2187 arguments can be specified.</p>
2193 <p>The '<tt>call</tt>' instruction is used to cause control flow to
2194 transfer to a specified function, with its incoming arguments bound to
2195 the specified values. Upon a '<tt><a href="#i_ret">ret</a></tt>'
2196 instruction in the called function, control flow continues with the
2197 instruction after the function call, and the return value of the
2198 function is bound to the result argument. This is a simpler case of
2199 the <a href="#i_invoke">invoke</a> instruction.</p>
2204 %retval = call int %test(int %argc)
2205 call int(sbyte*, ...) *%printf(sbyte* %msg, int 12, sbyte 42);
2206 %X = tail call int %foo()
2207 %Y = tail call <a href="#callingconv">fastcc</a> int %foo()
2212 <!-- _______________________________________________________________________ -->
2213 <div class="doc_subsubsection">
2214 <a name="i_vanext">'<tt>vanext</tt>' Instruction</a>
2217 <div class="doc_text">
2222 <resultarglist> = vanext <va_list> <arglist>, <argty>
2227 <p>The '<tt>vanext</tt>' instruction is used to access arguments passed
2228 through the "variable argument" area of a function call. It is used to
2229 implement the <tt>va_arg</tt> macro in C.</p>
2233 <p>This instruction takes a <tt>va_list</tt> value and the type of the
2234 argument. It returns another <tt>va_list</tt>. The actual type of
2235 <tt>va_list</tt> may be defined differently for different targets. Most targets
2236 use a <tt>va_list</tt> type of <tt>sbyte*</tt> or some other pointer type.</p>
2240 <p>The '<tt>vanext</tt>' instruction advances the specified <tt>va_list</tt>
2241 past an argument of the specified type. In conjunction with the <a
2242 href="#i_vaarg"><tt>vaarg</tt></a> instruction, it is used to implement
2243 the <tt>va_arg</tt> macro available in C. For more information, see
2244 the variable argument handling <a href="#int_varargs">Intrinsic
2247 <p>It is legal for this instruction to be called in a function which
2248 does not take a variable number of arguments, for example, the <tt>vfprintf</tt>
2251 <p><tt>vanext</tt> is an LLVM instruction instead of an <a
2252 href="#intrinsics">intrinsic function</a> because it takes a type as an
2253 argument. The type refers to the current argument in the <tt>va_list</tt>; it
2254 tells the compiler how far on the stack it needs to advance to find the next
2259 <p>See the <a href="#int_varargs">variable argument processing</a>
2264 <!-- _______________________________________________________________________ -->
2265 <div class="doc_subsubsection">
2266 <a name="i_vaarg">'<tt>vaarg</tt>' Instruction</a>
2269 <div class="doc_text">
2274 <resultval> = vaarg <va_list> <arglist>, <argty>
2279 <p>The '<tt>vaarg</tt>' instruction is used to access arguments passed through
2280 the "variable argument" area of a function call. It is used to implement the
2281 <tt>va_arg</tt> macro in C.</p>
2285 <p>This instruction takes a <tt>va_list</tt> value and the type of the
2286 argument. It returns a value of the specified argument type. Again, the actual
2287 type of <tt>va_list</tt> is target specific.</p>
2291 <p>The '<tt>vaarg</tt>' instruction loads an argument of the specified type from
2292 the specified <tt>va_list</tt>. In conjunction with the <a
2293 href="#i_vanext"><tt>vanext</tt></a> instruction, it is used to implement the
2294 <tt>va_arg</tt> macro available in C. For more information, see the variable
2295 argument handling <a href="#int_varargs">Intrinsic Functions</a>.</p>
2297 <p>It is legal for this instruction to be called in a function which does not
2298 take a variable number of arguments, for example, the <tt>vfprintf</tt>
2301 <p><tt>vaarg</tt> is an LLVM instruction instead of an <a
2302 href="#intrinsics">intrinsic function</a> because it takes a type as an
2307 <p>See the <a href="#int_varargs">variable argument processing</a> section.</p>
2311 <!-- *********************************************************************** -->
2312 <div class="doc_section"> <a name="intrinsics">Intrinsic Functions</a> </div>
2313 <!-- *********************************************************************** -->
2315 <div class="doc_text">
2317 <p>LLVM supports the notion of an "intrinsic function". These functions have
2318 well known names and semantics and are required to follow certain
2319 restrictions. Overall, these instructions represent an extension mechanism for
2320 the LLVM language that does not require changing all of the transformations in
2321 LLVM to add to the language (or the bytecode reader/writer, the parser,
2324 <p>Intrinsic function names must all start with an "<tt>llvm.</tt>" prefix. This
2325 prefix is reserved in LLVM for intrinsic names; thus, functions may not be named
2326 this. Intrinsic functions must always be external functions: you cannot define
2327 the body of intrinsic functions. Intrinsic functions may only be used in call
2328 or invoke instructions: it is illegal to take the address of an intrinsic
2329 function. Additionally, because intrinsic functions are part of the LLVM
2330 language, it is required that they all be documented here if any are added.</p>
2333 <p>To learn how to add an intrinsic function, please see the <a
2334 href="ExtendingLLVM.html">Extending LLVM Guide</a>.
2339 <!-- ======================================================================= -->
2340 <div class="doc_subsection">
2341 <a name="int_varargs">Variable Argument Handling Intrinsics</a>
2344 <div class="doc_text">
2346 <p>Variable argument support is defined in LLVM with the <a
2347 href="#i_vanext"><tt>vanext</tt></a> instruction and these three
2348 intrinsic functions. These functions are related to the similarly
2349 named macros defined in the <tt><stdarg.h></tt> header file.</p>
2351 <p>All of these functions operate on arguments that use a
2352 target-specific value type "<tt>va_list</tt>". The LLVM assembly
2353 language reference manual does not define what this type is, so all
2354 transformations should be prepared to handle intrinsics with any type
2357 <p>This example shows how the <a href="#i_vanext"><tt>vanext</tt></a>
2358 instruction and the variable argument handling intrinsic functions are
2362 int %test(int %X, ...) {
2363 ; Initialize variable argument processing
2364 %ap = call sbyte* %<a href="#i_va_start">llvm.va_start</a>()
2366 ; Read a single integer argument
2367 %tmp = vaarg sbyte* %ap, int
2369 ; Advance to the next argument
2370 %ap2 = vanext sbyte* %ap, int
2372 ; Demonstrate usage of llvm.va_copy and llvm.va_end
2373 %aq = call sbyte* %<a href="#i_va_copy">llvm.va_copy</a>(sbyte* %ap2)
2374 call void %<a href="#i_va_end">llvm.va_end</a>(sbyte* %aq)
2376 ; Stop processing of arguments.
2377 call void %<a href="#i_va_end">llvm.va_end</a>(sbyte* %ap2)
2383 <!-- _______________________________________________________________________ -->
2384 <div class="doc_subsubsection">
2385 <a name="i_va_start">'<tt>llvm.va_start</tt>' Intrinsic</a>
2389 <div class="doc_text">
2391 <pre> declare <va_list> %llvm.va_start()<br></pre>
2393 <p>The '<tt>llvm.va_start</tt>' intrinsic returns a new <tt><arglist></tt>
2394 for subsequent use by the variable argument intrinsics.</p>
2396 <p>The '<tt>llvm.va_start</tt>' intrinsic works just like the <tt>va_start</tt>
2397 macro available in C. In a target-dependent way, it initializes and
2398 returns a <tt>va_list</tt> element, so that the next <tt>vaarg</tt>
2399 will produce the first variable argument passed to the function. Unlike
2400 the C <tt>va_start</tt> macro, this intrinsic does not need to know the
2401 last argument of the function; the compiler can figure that out.</p>
2402 <p>Note that this intrinsic function is only legal to be called from
2403 within the body of a variable argument function.</p>
2406 <!-- _______________________________________________________________________ -->
2407 <div class="doc_subsubsection">
2408 <a name="i_va_end">'<tt>llvm.va_end</tt>' Intrinsic</a>
2411 <div class="doc_text">
2413 <pre> declare void %llvm.va_end(<va_list> <arglist>)<br></pre>
2415 <p>The '<tt>llvm.va_end</tt>' intrinsic destroys <tt><arglist></tt>
2416 which has been initialized previously with <tt><a href="#i_va_start">llvm.va_start</a></tt>
2417 or <tt><a href="#i_va_copy">llvm.va_copy</a></tt>.</p>
2419 <p>The argument is a <tt>va_list</tt> to destroy.</p>
2421 <p>The '<tt>llvm.va_end</tt>' intrinsic works just like the <tt>va_end</tt>
2422 macro available in C. In a target-dependent way, it destroys the <tt>va_list</tt>.
2423 Calls to <a href="#i_va_start"><tt>llvm.va_start</tt></a> and <a
2424 href="#i_va_copy"><tt>llvm.va_copy</tt></a> must be matched exactly
2425 with calls to <tt>llvm.va_end</tt>.</p>
2428 <!-- _______________________________________________________________________ -->
2429 <div class="doc_subsubsection">
2430 <a name="i_va_copy">'<tt>llvm.va_copy</tt>' Intrinsic</a>
2433 <div class="doc_text">
2438 declare <va_list> %llvm.va_copy(<va_list> <destarglist>)
2443 <p>The '<tt>llvm.va_copy</tt>' intrinsic copies the current argument position
2444 from the source argument list to the destination argument list.</p>
2448 <p>The argument is the <tt>va_list</tt> to copy.</p>
2452 <p>The '<tt>llvm.va_copy</tt>' intrinsic works just like the <tt>va_copy</tt>
2453 macro available in C. In a target-dependent way, it copies the source
2454 <tt>va_list</tt> element into the returned list. This intrinsic is necessary
2455 because the <tt><a href="#i_va_start">llvm.va_start</a></tt> intrinsic may be
2456 arbitrarily complex and require memory allocation, for example.</p>
2460 <!-- ======================================================================= -->
2461 <div class="doc_subsection">
2462 <a name="int_gc">Accurate Garbage Collection Intrinsics</a>
2465 <div class="doc_text">
2468 LLVM support for <a href="GarbageCollection.html">Accurate Garbage
2469 Collection</a> requires the implementation and generation of these intrinsics.
2470 These intrinsics allow identification of <a href="#i_gcroot">GC roots on the
2471 stack</a>, as well as garbage collector implementations that require <a
2472 href="#i_gcread">read</a> and <a href="#i_gcwrite">write</a> barriers.
2473 Front-ends for type-safe garbage collected languages should generate these
2474 intrinsics to make use of the LLVM garbage collectors. For more details, see <a
2475 href="GarbageCollection.html">Accurate Garbage Collection with LLVM</a>.
2479 <!-- _______________________________________________________________________ -->
2480 <div class="doc_subsubsection">
2481 <a name="i_gcroot">'<tt>llvm.gcroot</tt>' Intrinsic</a>
2484 <div class="doc_text">
2489 declare void %llvm.gcroot(<ty>** %ptrloc, <ty2>* %metadata)
2494 <p>The '<tt>llvm.gcroot</tt>' intrinsic declares the existence of a GC root to
2495 the code generator, and allows some metadata to be associated with it.</p>
2499 <p>The first argument specifies the address of a stack object that contains the
2500 root pointer. The second pointer (which must be either a constant or a global
2501 value address) contains the meta-data to be associated with the root.</p>
2505 <p>At runtime, a call to this intrinsics stores a null pointer into the "ptrloc"
2506 location. At compile-time, the code generator generates information to allow
2507 the runtime to find the pointer at GC safe points.
2513 <!-- _______________________________________________________________________ -->
2514 <div class="doc_subsubsection">
2515 <a name="i_gcread">'<tt>llvm.gcread</tt>' Intrinsic</a>
2518 <div class="doc_text">
2523 declare sbyte* %llvm.gcread(sbyte** %Ptr)
2528 <p>The '<tt>llvm.gcread</tt>' intrinsic identifies reads of references from heap
2529 locations, allowing garbage collector implementations that require read
2534 <p>The argument is the address to read from, which should be an address
2535 allocated from the garbage collector.</p>
2539 <p>The '<tt>llvm.gcread</tt>' intrinsic has the same semantics as a load
2540 instruction, but may be replaced with substantially more complex code by the
2541 garbage collector runtime, as needed.</p>
2546 <!-- _______________________________________________________________________ -->
2547 <div class="doc_subsubsection">
2548 <a name="i_gcwrite">'<tt>llvm.gcwrite</tt>' Intrinsic</a>
2551 <div class="doc_text">
2556 declare void %llvm.gcwrite(sbyte* %P1, sbyte** %P2)
2561 <p>The '<tt>llvm.gcwrite</tt>' intrinsic identifies writes of references to heap
2562 locations, allowing garbage collector implementations that require write
2563 barriers (such as generational or reference counting collectors).</p>
2567 <p>The first argument is the reference to store, and the second is the heap
2568 location to store to.</p>
2572 <p>The '<tt>llvm.gcwrite</tt>' intrinsic has the same semantics as a store
2573 instruction, but may be replaced with substantially more complex code by the
2574 garbage collector runtime, as needed.</p>
2580 <!-- ======================================================================= -->
2581 <div class="doc_subsection">
2582 <a name="int_codegen">Code Generator Intrinsics</a>
2585 <div class="doc_text">
2587 These intrinsics are provided by LLVM to expose special features that may only
2588 be implemented with code generator support.
2593 <!-- _______________________________________________________________________ -->
2594 <div class="doc_subsubsection">
2595 <a name="i_returnaddress">'<tt>llvm.returnaddress</tt>' Intrinsic</a>
2598 <div class="doc_text">
2602 declare void* %llvm.returnaddress(uint <level>)
2608 The '<tt>llvm.returnaddress</tt>' intrinsic returns a target-specific value
2609 indicating the return address of the current function or one of its callers.
2615 The argument to this intrinsic indicates which function to return the address
2616 for. Zero indicates the calling function, one indicates its caller, etc. The
2617 argument is <b>required</b> to be a constant integer value.
2623 The '<tt>llvm.returnaddress</tt>' intrinsic either returns a pointer indicating
2624 the return address of the specified call frame, or zero if it cannot be
2625 identified. The value returned by this intrinsic is likely to be incorrect or 0
2626 for arguments other than zero, so it should only be used for debugging purposes.
2630 Note that calling this intrinsic does not prevent function inlining or other
2631 aggressive transformations, so the value returned may not be that of the obvious
2632 source-language caller.
2637 <!-- _______________________________________________________________________ -->
2638 <div class="doc_subsubsection">
2639 <a name="i_frameaddress">'<tt>llvm.frameaddress</tt>' Intrinsic</a>
2642 <div class="doc_text">
2646 declare void* %llvm.frameaddress(uint <level>)
2652 The '<tt>llvm.frameaddress</tt>' intrinsic returns the target-specific frame
2653 pointer value for the specified stack frame.
2659 The argument to this intrinsic indicates which function to return the frame
2660 pointer for. Zero indicates the calling function, one indicates its caller,
2661 etc. The argument is <b>required</b> to be a constant integer value.
2667 The '<tt>llvm.frameaddress</tt>' intrinsic either returns a pointer indicating
2668 the frame address of the specified call frame, or zero if it cannot be
2669 identified. The value returned by this intrinsic is likely to be incorrect or 0
2670 for arguments other than zero, so it should only be used for debugging purposes.
2674 Note that calling this intrinsic does not prevent function inlining or other
2675 aggressive transformations, so the value returned may not be that of the obvious
2676 source-language caller.
2680 <!-- _______________________________________________________________________ -->
2681 <div class="doc_subsubsection">
2682 <a name="i_prefetch">'<tt>llvm.prefetch</tt>' Intrinsic</a>
2685 <div class="doc_text">
2689 declare void %llvm.prefetch(sbyte * <address>,
2690 uint <rw>, uint <locality>)
2697 The '<tt>llvm.prefetch</tt>' intrinsic is a hint to the code generator to insert
2698 a prefetch instruction if supported; otherwise, it is a noop. Prefetches have
2700 effect on the behavior of the program but can change its performance
2707 <tt>address</tt> is the address to be prefetched, <tt>rw</tt> is the specifier
2708 determining if the fetch should be for a read (0) or write (1), and
2709 <tt>locality</tt> is a temporal locality specifier ranging from (0) - no
2710 locality, to (3) - extremely local keep in cache. The <tt>rw</tt> and
2711 <tt>locality</tt> arguments must be constant integers.
2717 This intrinsic does not modify the behavior of the program. In particular,
2718 prefetches cannot trap and do not produce a value. On targets that support this
2719 intrinsic, the prefetch can provide hints to the processor cache for better
2725 <!-- _______________________________________________________________________ -->
2726 <div class="doc_subsubsection">
2727 <a name="i_pcmarker">'<tt>llvm.pcmarker</tt>' Intrinsic</a>
2730 <div class="doc_text">
2734 declare void %llvm.pcmarker( uint <id> )
2741 The '<tt>llvm.pcmarker</tt>' intrinsic is a method to export a Program Counter
2743 code to simulators and other tools. The method is target specific, but it is
2744 expected that the marker will use exported symbols to transmit the PC of the marker.
2745 The marker makes no guaranties that it will remain with any specific instruction
2746 after optimizations. It is possible that the presense of a marker will inhibit
2747 optimizations. The intended use is to be inserted after optmizations to allow
2748 correlations of simulation runs.
2754 <tt>id</tt> is a numerical id identifying the marker.
2760 This intrinsic does not modify the behavior of the program. Backends that do not
2761 support this intrinisic may ignore it.
2767 <!-- ======================================================================= -->
2768 <div class="doc_subsection">
2769 <a name="int_os">Operating System Intrinsics</a>
2772 <div class="doc_text">
2774 These intrinsics are provided by LLVM to support the implementation of
2775 operating system level code.
2780 <!-- _______________________________________________________________________ -->
2781 <div class="doc_subsubsection">
2782 <a name="i_readport">'<tt>llvm.readport</tt>' Intrinsic</a>
2785 <div class="doc_text">
2789 declare <integer type> %llvm.readport (<integer type> <address>)
2795 The '<tt>llvm.readport</tt>' intrinsic reads data from the specified hardware
2802 The argument to this intrinsic indicates the hardware I/O address from which
2803 to read the data. The address is in the hardware I/O address namespace (as
2804 opposed to being a memory location for memory mapped I/O).
2810 The '<tt>llvm.readport</tt>' intrinsic reads data from the hardware I/O port
2811 specified by <i>address</i> and returns the value. The address and return
2812 value must be integers, but the size is dependent upon the platform upon which
2813 the program is code generated. For example, on x86, the address must be an
2814 unsigned 16-bit value, and the return value must be 8, 16, or 32 bits.
2819 <!-- _______________________________________________________________________ -->
2820 <div class="doc_subsubsection">
2821 <a name="i_writeport">'<tt>llvm.writeport</tt>' Intrinsic</a>
2824 <div class="doc_text">
2828 call void (<integer type>, <integer type>)*
2829 %llvm.writeport (<integer type> <value>,
2830 <integer type> <address>)
2836 The '<tt>llvm.writeport</tt>' intrinsic writes data to the specified hardware
2843 The first argument is the value to write to the I/O port.
2847 The second argument indicates the hardware I/O address to which data should be
2848 written. The address is in the hardware I/O address namespace (as opposed to
2849 being a memory location for memory mapped I/O).
2855 The '<tt>llvm.writeport</tt>' intrinsic writes <i>value</i> to the I/O port
2856 specified by <i>address</i>. The address and value must be integers, but the
2857 size is dependent upon the platform upon which the program is code generated.
2858 For example, on x86, the address must be an unsigned 16-bit value, and the
2859 value written must be 8, 16, or 32 bits in length.
2864 <!-- _______________________________________________________________________ -->
2865 <div class="doc_subsubsection">
2866 <a name="i_readio">'<tt>llvm.readio</tt>' Intrinsic</a>
2869 <div class="doc_text">
2873 declare <result> %llvm.readio (<ty> * <pointer>)
2879 The '<tt>llvm.readio</tt>' intrinsic reads data from a memory mapped I/O
2886 The argument to this intrinsic is a pointer indicating the memory address from
2887 which to read the data. The data must be a
2888 <a href="#t_firstclass">first class</a> type.
2894 The '<tt>llvm.readio</tt>' intrinsic reads data from a memory mapped I/O
2895 location specified by <i>pointer</i> and returns the value. The argument must
2896 be a pointer, and the return value must be a
2897 <a href="#t_firstclass">first class</a> type. However, certain architectures
2898 may not support I/O on all first class types. For example, 32-bit processors
2899 may only support I/O on data types that are 32 bits or less.
2903 This intrinsic enforces an in-order memory model for llvm.readio and
2904 llvm.writeio calls on machines that use dynamic scheduling. Dynamically
2905 scheduled processors may execute loads and stores out of order, re-ordering at
2906 run time accesses to memory mapped I/O registers. Using these intrinsics
2907 ensures that accesses to memory mapped I/O registers occur in program order.
2912 <!-- _______________________________________________________________________ -->
2913 <div class="doc_subsubsection">
2914 <a name="i_writeio">'<tt>llvm.writeio</tt>' Intrinsic</a>
2917 <div class="doc_text">
2921 declare void %llvm.writeio (<ty1> <value>, <ty2> * <pointer>)
2927 The '<tt>llvm.writeio</tt>' intrinsic writes data to the specified memory
2934 The first argument is the value to write to the memory mapped I/O location.
2935 The second argument is a pointer indicating the memory address to which the
2936 data should be written.
2942 The '<tt>llvm.writeio</tt>' intrinsic writes <i>value</i> to the memory mapped
2943 I/O address specified by <i>pointer</i>. The value must be a
2944 <a href="#t_firstclass">first class</a> type. However, certain architectures
2945 may not support I/O on all first class types. For example, 32-bit processors
2946 may only support I/O on data types that are 32 bits or less.
2950 This intrinsic enforces an in-order memory model for llvm.readio and
2951 llvm.writeio calls on machines that use dynamic scheduling. Dynamically
2952 scheduled processors may execute loads and stores out of order, re-ordering at
2953 run time accesses to memory mapped I/O registers. Using these intrinsics
2954 ensures that accesses to memory mapped I/O registers occur in program order.
2959 <!-- ======================================================================= -->
2960 <div class="doc_subsection">
2961 <a name="int_libc">Standard C Library Intrinsics</a>
2964 <div class="doc_text">
2966 LLVM provides intrinsics for a few important standard C library functions.
2967 These intrinsics allow source-language front-ends to pass information about the
2968 alignment of the pointer arguments to the code generator, providing opportunity
2969 for more efficient code generation.
2974 <!-- _______________________________________________________________________ -->
2975 <div class="doc_subsubsection">
2976 <a name="i_memcpy">'<tt>llvm.memcpy</tt>' Intrinsic</a>
2979 <div class="doc_text">
2983 declare void %llvm.memcpy(sbyte* <dest>, sbyte* <src>,
2984 uint <len>, uint <align>)
2990 The '<tt>llvm.memcpy</tt>' intrinsic copies a block of memory from the source
2991 location to the destination location.
2995 Note that, unlike the standard libc function, the <tt>llvm.memcpy</tt> intrinsic
2996 does not return a value, and takes an extra alignment argument.
3002 The first argument is a pointer to the destination, the second is a pointer to
3003 the source. The third argument is an (arbitrarily sized) integer argument
3004 specifying the number of bytes to copy, and the fourth argument is the alignment
3005 of the source and destination locations.
3009 If the call to this intrinisic has an alignment value that is not 0 or 1, then
3010 the caller guarantees that the size of the copy is a multiple of the alignment
3011 and that both the source and destination pointers are aligned to that boundary.
3017 The '<tt>llvm.memcpy</tt>' intrinsic copies a block of memory from the source
3018 location to the destination location, which are not allowed to overlap. It
3019 copies "len" bytes of memory over. If the argument is known to be aligned to
3020 some boundary, this can be specified as the fourth argument, otherwise it should
3026 <!-- _______________________________________________________________________ -->
3027 <div class="doc_subsubsection">
3028 <a name="i_memmove">'<tt>llvm.memmove</tt>' Intrinsic</a>
3031 <div class="doc_text">
3035 declare void %llvm.memmove(sbyte* <dest>, sbyte* <src>,
3036 uint <len>, uint <align>)
3042 The '<tt>llvm.memmove</tt>' intrinsic moves a block of memory from the source
3043 location to the destination location. It is similar to the '<tt>llvm.memcpy</tt>'
3044 intrinsic but allows the two memory locations to overlap.
3048 Note that, unlike the standard libc function, the <tt>llvm.memmove</tt> intrinsic
3049 does not return a value, and takes an extra alignment argument.
3055 The first argument is a pointer to the destination, the second is a pointer to
3056 the source. The third argument is an (arbitrarily sized) integer argument
3057 specifying the number of bytes to copy, and the fourth argument is the alignment
3058 of the source and destination locations.
3062 If the call to this intrinisic has an alignment value that is not 0 or 1, then
3063 the caller guarantees that the size of the copy is a multiple of the alignment
3064 and that both the source and destination pointers are aligned to that boundary.
3070 The '<tt>llvm.memmove</tt>' intrinsic copies a block of memory from the source
3071 location to the destination location, which may overlap. It
3072 copies "len" bytes of memory over. If the argument is known to be aligned to
3073 some boundary, this can be specified as the fourth argument, otherwise it should
3079 <!-- _______________________________________________________________________ -->
3080 <div class="doc_subsubsection">
3081 <a name="i_memset">'<tt>llvm.memset</tt>' Intrinsic</a>
3084 <div class="doc_text">
3088 declare void %llvm.memset(sbyte* <dest>, ubyte <val>,
3089 uint <len>, uint <align>)
3095 The '<tt>llvm.memset</tt>' intrinsic fills a block of memory with a particular
3100 Note that, unlike the standard libc function, the <tt>llvm.memset</tt> intrinsic
3101 does not return a value, and takes an extra alignment argument.
3107 The first argument is a pointer to the destination to fill, the second is the
3108 byte value to fill it with, the third argument is an (arbitrarily sized) integer
3109 argument specifying the number of bytes to fill, and the fourth argument is the
3110 known alignment of destination location.
3114 If the call to this intrinisic has an alignment value that is not 0 or 1, then
3115 the caller guarantees that the size of the copy is a multiple of the alignment
3116 and that the destination pointer is aligned to that boundary.
3122 The '<tt>llvm.memset</tt>' intrinsic fills "len" bytes of memory starting at the
3123 destination location. If the argument is known to be aligned to some boundary,
3124 this can be specified as the fourth argument, otherwise it should be set to 0 or
3130 <!-- _______________________________________________________________________ -->
3131 <div class="doc_subsubsection">
3132 <a name="i_isunordered">'<tt>llvm.isunordered</tt>' Intrinsic</a>
3135 <div class="doc_text">
3139 declare bool %llvm.isunordered(<float or double> Val1, <float or double> Val2)
3145 The '<tt>llvm.isunordered</tt>' intrinsic returns true if either or both of the
3146 specified floating point values is a NAN.
3152 The arguments are floating point numbers of the same type.
3158 If either or both of the arguments is a SNAN or QNAN, it returns true, otherwise
3164 <!-- ======================================================================= -->
3165 <div class="doc_subsection">
3166 <a name="int_count">Bit Counting Intrinsics</a>
3169 <div class="doc_text">
3171 LLVM provides intrinsics for a few important bit counting operations.
3172 These allow efficient code generation for some algorithms.
3177 <!-- _______________________________________________________________________ -->
3178 <div class="doc_subsubsection">
3179 <a name="int_ctpop">'<tt>llvm.ctpop</tt>' Intrinsic</a>
3182 <div class="doc_text">
3186 declare int %llvm.ctpop(int <src>)
3193 The '<tt>llvm.ctpop</tt>' intrinsic counts the number of ones in a variable.
3199 The only argument is the value to be counted. The argument may be of any
3200 integer type. The return type must match the argument type.
3206 The '<tt>llvm.ctpop</tt>' intrinsic counts the 1's in a variable.
3210 <!-- _______________________________________________________________________ -->
3211 <div class="doc_subsubsection">
3212 <a name="int_ctlz">'<tt>llvm.ctlz</tt>' Intrinsic</a>
3215 <div class="doc_text">
3219 declare int %llvm.ctlz(int <src>)
3226 The '<tt>llvm.ctlz</tt>' intrinsic counts the number of leading zeros in a
3233 The only argument is the value to be counted. The argument may be of any
3234 integer type. The return type must match the argument type.
3240 The '<tt>llvm.ctlz</tt>' intrinsic counts the leading (most significant) zeros
3241 in a variable. If the src == 0 then the result is the size in bits of the type
3242 of src. For example, <tt>llvm.cttz(int 2) = 30</tt>.
3248 <!-- _______________________________________________________________________ -->
3249 <div class="doc_subsubsection">
3250 <a name="int_cttz">'<tt>llvm.cttz</tt>' Intrinsic</a>
3253 <div class="doc_text">
3257 declare int %llvm.cttz(int <src>)
3264 The '<tt>llvm.cttz</tt>' intrinsic counts the number of trailing zeros.
3270 The only argument is the value to be counted. The argument may be of any
3271 integer type. The return type must match the argument type.
3277 The '<tt>llvm.cttz</tt>' intrinsic counts the trailing (least significant) zeros
3278 in a variable. If the src == 0 then the result is the size in bits of the type
3279 of src. For example, <tt>llvm.cttz(2) = 1</tt>.
3283 <!-- ======================================================================= -->
3284 <div class="doc_subsection">
3285 <a name="int_debugger">Debugger Intrinsics</a>
3288 <div class="doc_text">
3290 The LLVM debugger intrinsics (which all start with <tt>llvm.dbg.</tt> prefix),
3291 are described in the <a
3292 href="SourceLevelDebugging.html#format_common_intrinsics">LLVM Source Level
3293 Debugging</a> document.
3298 <!-- *********************************************************************** -->
3301 <a href="http://jigsaw.w3.org/css-validator/check/referer"><img
3302 src="http://jigsaw.w3.org/css-validator/images/vcss" alt="Valid CSS!"></a>
3303 <a href="http://validator.w3.org/check/referer"><img
3304 src="http://www.w3.org/Icons/valid-html401" alt="Valid HTML 4.01!" /></a>
3306 <a href="mailto:sabre@nondot.org">Chris Lattner</a><br>
3307 <a href="http://llvm.cs.uiuc.edu">The LLVM Compiler Infrastructure</a><br>
3308 Last modified: $Date$