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_vaarg">'<tt>vaarg</tt>' Instruction</a></li>
109 <li><a href="#intrinsics">Intrinsic Functions</a>
111 <li><a href="#int_varargs">Variable Argument Handling Intrinsics</a>
113 <li><a href="#i_va_start">'<tt>llvm.va_start</tt>' Intrinsic</a></li>
114 <li><a href="#i_va_end">'<tt>llvm.va_end</tt>' Intrinsic</a></li>
115 <li><a href="#i_va_copy">'<tt>llvm.va_copy</tt>' Intrinsic</a></li>
118 <li><a href="#int_gc">Accurate Garbage Collection Intrinsics</a>
120 <li><a href="#i_gcroot">'<tt>llvm.gcroot</tt>' Intrinsic</a></li>
121 <li><a href="#i_gcread">'<tt>llvm.gcread</tt>' Intrinsic</a></li>
122 <li><a href="#i_gcwrite">'<tt>llvm.gcwrite</tt>' Intrinsic</a></li>
125 <li><a href="#int_codegen">Code Generator Intrinsics</a>
127 <li><a href="#i_returnaddress">'<tt>llvm.returnaddress</tt>' Intrinsic</a></li>
128 <li><a href="#i_frameaddress">'<tt>llvm.frameaddress</tt>' Intrinsic</a></li>
129 <li><a href="#i_prefetch">'<tt>llvm.prefetch</tt>' Intrinsic</a></li>
130 <li><a href="#i_pcmarker">'<tt>llvm.pcmarker</tt>' Intrinsic</a></li>
133 <li><a href="#int_os">Operating System Intrinsics</a>
135 <li><a href="#i_readport">'<tt>llvm.readport</tt>' Intrinsic</a></li>
136 <li><a href="#i_writeport">'<tt>llvm.writeport</tt>' Intrinsic</a></li>
137 <li><a href="#i_readio">'<tt>llvm.readio</tt>' Intrinsic</a></li>
138 <li><a href="#i_writeio">'<tt>llvm.writeio</tt>' Intrinsic</a></li>
140 <li><a href="#int_libc">Standard C Library Intrinsics</a>
142 <li><a href="#i_memcpy">'<tt>llvm.memcpy</tt>' Intrinsic</a></li>
143 <li><a href="#i_memmove">'<tt>llvm.memmove</tt>' Intrinsic</a></li>
144 <li><a href="#i_memset">'<tt>llvm.memset</tt>' Intrinsic</a></li>
145 <li><a href="#i_isunordered">'<tt>llvm.isunordered</tt>' Intrinsic</a></li>
148 <li><a href="#int_count">Bit counting Intrinsics</a>
150 <li><a href="#int_ctpop">'<tt>llvm.ctpop</tt>' Intrinsic </a></li>
151 <li><a href="#int_ctlz">'<tt>llvm.ctlz</tt>' Intrinsic </a></li>
152 <li><a href="#int_cttz">'<tt>llvm.cttz</tt>' Intrinsic </a></li>
155 <li><a href="#int_debugger">Debugger intrinsics</a></li>
160 <div class="doc_author">
161 <p>Written by <a href="mailto:sabre@nondot.org">Chris Lattner</a>
162 and <a href="mailto:vadve@cs.uiuc.edu">Vikram Adve</a></p>
165 <!-- *********************************************************************** -->
166 <div class="doc_section"> <a name="abstract">Abstract </a></div>
167 <!-- *********************************************************************** -->
169 <div class="doc_text">
170 <p>This document is a reference manual for the LLVM assembly language.
171 LLVM is an SSA based representation that provides type safety,
172 low-level operations, flexibility, and the capability of representing
173 'all' high-level languages cleanly. It is the common code
174 representation used throughout all phases of the LLVM compilation
178 <!-- *********************************************************************** -->
179 <div class="doc_section"> <a name="introduction">Introduction</a> </div>
180 <!-- *********************************************************************** -->
182 <div class="doc_text">
184 <p>The LLVM code representation is designed to be used in three
185 different forms: as an in-memory compiler IR, as an on-disk bytecode
186 representation (suitable for fast loading by a Just-In-Time compiler),
187 and as a human readable assembly language representation. This allows
188 LLVM to provide a powerful intermediate representation for efficient
189 compiler transformations and analysis, while providing a natural means
190 to debug and visualize the transformations. The three different forms
191 of LLVM are all equivalent. This document describes the human readable
192 representation and notation.</p>
194 <p>The LLVM representation aims to be light-weight and low-level
195 while being expressive, typed, and extensible at the same time. It
196 aims to be a "universal IR" of sorts, by being at a low enough level
197 that high-level ideas may be cleanly mapped to it (similar to how
198 microprocessors are "universal IR's", allowing many source languages to
199 be mapped to them). By providing type information, LLVM can be used as
200 the target of optimizations: for example, through pointer analysis, it
201 can be proven that a C automatic variable is never accessed outside of
202 the current function... allowing it to be promoted to a simple SSA
203 value instead of a memory location.</p>
207 <!-- _______________________________________________________________________ -->
208 <div class="doc_subsubsection"> <a name="wellformed">Well-Formedness</a> </div>
210 <div class="doc_text">
212 <p>It is important to note that this document describes 'well formed'
213 LLVM assembly language. There is a difference between what the parser
214 accepts and what is considered 'well formed'. For example, the
215 following instruction is syntactically okay, but not well formed:</p>
218 %x = <a href="#i_add">add</a> int 1, %x
221 <p>...because the definition of <tt>%x</tt> does not dominate all of
222 its uses. The LLVM infrastructure provides a verification pass that may
223 be used to verify that an LLVM module is well formed. This pass is
224 automatically run by the parser after parsing input assembly and by
225 the optimizer before it outputs bytecode. The violations pointed out
226 by the verifier pass indicate bugs in transformation passes or input to
229 <!-- Describe the typesetting conventions here. --> </div>
231 <!-- *********************************************************************** -->
232 <div class="doc_section"> <a name="identifiers">Identifiers</a> </div>
233 <!-- *********************************************************************** -->
235 <div class="doc_text">
237 <p>LLVM uses three different forms of identifiers, for different
241 <li>Named values are represented as a string of characters with a '%' prefix.
242 For example, %foo, %DivisionByZero, %a.really.long.identifier. The actual
243 regular expression used is '<tt>%[a-zA-Z$._][a-zA-Z$._0-9]*</tt>'.
244 Identifiers which require other characters in their names can be surrounded
245 with quotes. In this way, anything except a <tt>"</tt> character can be used
248 <li>Unnamed values are represented as an unsigned numeric value with a '%'
249 prefix. For example, %12, %2, %44.</li>
251 <li>Constants, which are described in a <a href="#constants">section about
252 constants</a>, below.</li>
255 <p>LLVM requires that values start with a '%' sign for two reasons: Compilers
256 don't need to worry about name clashes with reserved words, and the set of
257 reserved words may be expanded in the future without penalty. Additionally,
258 unnamed identifiers allow a compiler to quickly come up with a temporary
259 variable without having to avoid symbol table conflicts.</p>
261 <p>Reserved words in LLVM are very similar to reserved words in other
262 languages. There are keywords for different opcodes ('<tt><a
263 href="#i_add">add</a></tt>', '<tt><a href="#i_cast">cast</a></tt>', '<tt><a
264 href="#i_ret">ret</a></tt>', etc...), for primitive type names ('<tt><a
265 href="#t_void">void</a></tt>', '<tt><a href="#t_uint">uint</a></tt>', etc...),
266 and others. These reserved words cannot conflict with variable names, because
267 none of them start with a '%' character.</p>
269 <p>Here is an example of LLVM code to multiply the integer variable
270 '<tt>%X</tt>' by 8:</p>
275 %result = <a href="#i_mul">mul</a> uint %X, 8
278 <p>After strength reduction:</p>
281 %result = <a href="#i_shl">shl</a> uint %X, ubyte 3
284 <p>And the hard way:</p>
287 <a href="#i_add">add</a> uint %X, %X <i>; yields {uint}:%0</i>
288 <a href="#i_add">add</a> uint %0, %0 <i>; yields {uint}:%1</i>
289 %result = <a href="#i_add">add</a> uint %1, %1
292 <p>This last way of multiplying <tt>%X</tt> by 8 illustrates several
293 important lexical features of LLVM:</p>
297 <li>Comments are delimited with a '<tt>;</tt>' and go until the end of
300 <li>Unnamed temporaries are created when the result of a computation is not
301 assigned to a named value.</li>
303 <li>Unnamed temporaries are numbered sequentially</li>
307 <p>...and it also shows a convention that we follow in this document. When
308 demonstrating instructions, we will follow an instruction with a comment that
309 defines the type and name of value produced. Comments are shown in italic
314 <!-- *********************************************************************** -->
315 <div class="doc_section"> <a name="highlevel">High Level Structure</a> </div>
316 <!-- *********************************************************************** -->
318 <!-- ======================================================================= -->
319 <div class="doc_subsection"> <a name="modulestructure">Module Structure</a>
322 <div class="doc_text">
324 <p>LLVM programs are composed of "Module"s, each of which is a
325 translation unit of the input programs. Each module consists of
326 functions, global variables, and symbol table entries. Modules may be
327 combined together with the LLVM linker, which merges function (and
328 global variable) definitions, resolves forward declarations, and merges
329 symbol table entries. Here is an example of the "hello world" module:</p>
331 <pre><i>; Declare the string constant as a global constant...</i>
332 <a href="#identifiers">%.LC0</a> = <a href="#linkage_internal">internal</a> <a
333 href="#globalvars">constant</a> <a href="#t_array">[13 x sbyte]</a> c"hello world\0A\00" <i>; [13 x sbyte]*</i>
335 <i>; External declaration of the puts function</i>
336 <a href="#functionstructure">declare</a> int %puts(sbyte*) <i>; int(sbyte*)* </i>
338 <i>; Definition of main function</i>
339 int %main() { <i>; int()* </i>
340 <i>; Convert [13x sbyte]* to sbyte *...</i>
342 href="#i_getelementptr">getelementptr</a> [13 x sbyte]* %.LC0, long 0, long 0 <i>; sbyte*</i>
344 <i>; Call puts function to write out the string to stdout...</i>
346 href="#i_call">call</a> int %puts(sbyte* %cast210) <i>; int</i>
348 href="#i_ret">ret</a> int 0<br>}<br></pre>
350 <p>This example is made up of a <a href="#globalvars">global variable</a>
351 named "<tt>.LC0</tt>", an external declaration of the "<tt>puts</tt>"
352 function, and a <a href="#functionstructure">function definition</a>
353 for "<tt>main</tt>".</p>
355 <p>In general, a module is made up of a list of global values,
356 where both functions and global variables are global values. Global values are
357 represented by a pointer to a memory location (in this case, a pointer to an
358 array of char, and a pointer to a function), and have one of the following <a
359 href="#linkage">linkage types</a>.</p>
363 <!-- ======================================================================= -->
364 <div class="doc_subsection">
365 <a name="linkage">Linkage Types</a>
368 <div class="doc_text">
371 All Global Variables and Functions have one of the following types of linkage:
376 <dt><tt><b><a name="linkage_internal">internal</a></b></tt> </dt>
378 <dd>Global values with internal linkage are only directly accessible by
379 objects in the current module. In particular, linking code into a module with
380 an internal global value may cause the internal to be renamed as necessary to
381 avoid collisions. Because the symbol is internal to the module, all
382 references can be updated. This corresponds to the notion of the
383 '<tt>static</tt>' keyword in C, or the idea of "anonymous namespaces" in C++.
386 <dt><tt><b><a name="linkage_linkonce">linkonce</a></b></tt>: </dt>
388 <dd>"<tt>linkonce</tt>" linkage is similar to <tt>internal</tt> linkage, with
389 the twist that linking together two modules defining the same
390 <tt>linkonce</tt> globals will cause one of the globals to be discarded. This
391 is typically used to implement inline functions. Unreferenced
392 <tt>linkonce</tt> globals are allowed to be discarded.
395 <dt><tt><b><a name="linkage_weak">weak</a></b></tt>: </dt>
397 <dd>"<tt>weak</tt>" linkage is exactly the same as <tt>linkonce</tt> linkage,
398 except that unreferenced <tt>weak</tt> globals may not be discarded. This is
399 used to implement constructs in C such as "<tt>int X;</tt>" at global scope.
402 <dt><tt><b><a name="linkage_appending">appending</a></b></tt>: </dt>
404 <dd>"<tt>appending</tt>" linkage may only be applied to global variables of
405 pointer to array type. When two global variables with appending linkage are
406 linked together, the two global arrays are appended together. This is the
407 LLVM, typesafe, equivalent of having the system linker append together
408 "sections" with identical names when .o files are linked.
411 <dt><tt><b><a name="linkage_external">externally visible</a></b></tt>:</dt>
413 <dd>If none of the above identifiers are used, the global is externally
414 visible, meaning that it participates in linkage and can be used to resolve
415 external symbol references.
419 <p><a name="linkage_external">For example, since the "<tt>.LC0</tt>"
420 variable is defined to be internal, if another module defined a "<tt>.LC0</tt>"
421 variable and was linked with this one, one of the two would be renamed,
422 preventing a collision. Since "<tt>main</tt>" and "<tt>puts</tt>" are
423 external (i.e., lacking any linkage declarations), they are accessible
424 outside of the current module. It is illegal for a function <i>declaration</i>
425 to have any linkage type other than "externally visible".</a></p>
429 <!-- ======================================================================= -->
430 <div class="doc_subsection">
431 <a name="callingconv">Calling Conventions</a>
434 <div class="doc_text">
436 <p>LLVM <a href="#functionstructure">functions</a>, <a href="#i_call">calls</a>
437 and <a href="#i_invoke">invokes</a> can all have an optional calling convention
438 specified for the call. The calling convention of any pair of dynamic
439 caller/callee must match, or the behavior of the program is undefined. The
440 following calling conventions are supported by LLVM, and more may be added in
444 <dt><b>"<tt>ccc</tt>" - The C calling convention</b>:</dt>
446 <dd>This calling convention (the default if no other calling convention is
447 specified) matches the target C calling conventions. This calling convention
448 supports varargs function calls and tolerates some mismatch in the declared
449 prototype and implemented declaration of the function (as does normal C).
452 <dt><b>"<tt>fastcc</tt>" - The fast calling convention</b>:</dt>
454 <dd>This calling convention attempts to make calls as fast as possible
455 (e.g. by passing things in registers). This calling convention allows the
456 target to use whatever tricks it wants to produce fast code for the target,
457 without having to conform to an externally specified ABI. Implementations of
458 this convention should allow arbitrary tail call optimization to be supported.
459 This calling convention does not support varargs and requires the prototype of
460 all callees to exactly match the prototype of the function definition.
463 <dt><b>"<tt>coldcc</tt>" - The cold calling convention</b>:</dt>
465 <dd>This calling convention attempts to make code in the caller as efficient
466 as possible under the assumption that the call is not commonly executed. As
467 such, these calls often preserve all registers so that the call does not break
468 any live ranges in the caller side. This calling convention does not support
469 varargs and requires the prototype of all callees to exactly match the
470 prototype of the function definition.
473 <dt><b>"<tt>cc <<em>n</em>></tt>" - Numbered convention</b>:</dt>
475 <dd>Any calling convention may be specified by number, allowing
476 target-specific calling conventions to be used. Target specific calling
477 conventions start at 64.
481 <p>More calling conventions can be added/defined on an as-needed basis, to
482 support pascal conventions or any other well-known target-independent
487 <!-- ======================================================================= -->
488 <div class="doc_subsection">
489 <a name="globalvars">Global Variables</a>
492 <div class="doc_text">
494 <p>Global variables define regions of memory allocated at compilation time
495 instead of run-time. Global variables may optionally be initialized. A
496 variable may be defined as a global "constant", which indicates that the
497 contents of the variable will <b>never</b> be modified (enabling better
498 optimization, allowing the global data to be placed in the read-only section of
499 an executable, etc). Note that variables that need runtime initialization
500 cannot be marked "constant", as there is a store to the variable.</p>
503 LLVM explicitly allows <em>declarations</em> of global variables to be marked
504 constant, even if the final definition of the global is not. This capability
505 can be used to enable slightly better optimization of the program, but requires
506 the language definition to guarantee that optimizations based on the
507 'constantness' are valid for the translation units that do not include the
511 <p>As SSA values, global variables define pointer values that are in
512 scope (i.e. they dominate) all basic blocks in the program. Global
513 variables always define a pointer to their "content" type because they
514 describe a region of memory, and all memory objects in LLVM are
515 accessed through pointers.</p>
520 <!-- ======================================================================= -->
521 <div class="doc_subsection">
522 <a name="functionstructure">Functions</a>
525 <div class="doc_text">
527 <p>LLVM function definitions consist of an optional <a href="#linkage">linkage
528 type</a>, an optional <a href="#callingconv">calling convention</a>, a return
529 type, a function name, a (possibly empty) argument list, an opening curly brace,
530 a list of basic blocks, and a closing curly brace. LLVM function declarations
531 are defined with the "<tt>declare</tt>" keyword, an optional <a
532 href="#callingconv">calling convention</a>, a return type, a function name, and
533 a possibly empty list of arguments.</p>
535 <p>A function definition contains a list of basic blocks, forming the CFG for
536 the function. Each basic block may optionally start with a label (giving the
537 basic block a symbol table entry), contains a list of instructions, and ends
538 with a <a href="#terminators">terminator</a> instruction (such as a branch or
539 function return).</p>
541 <p>The first basic block in a program is special in two ways: it is immediately
542 executed on entrance to the function, and it is not allowed to have predecessor
543 basic blocks (i.e. there can not be any branches to the entry block of a
544 function). Because the block can have no predecessors, it also cannot have any
545 <a href="#i_phi">PHI nodes</a>.</p>
547 <p>LLVM functions are identified by their name and type signature. Hence, two
548 functions with the same name but different parameter lists or return values are
549 considered different functions, and LLVM will resolve references to each
556 <!-- *********************************************************************** -->
557 <div class="doc_section"> <a name="typesystem">Type System</a> </div>
558 <!-- *********************************************************************** -->
560 <div class="doc_text">
562 <p>The LLVM type system is one of the most important features of the
563 intermediate representation. Being typed enables a number of
564 optimizations to be performed on the IR directly, without having to do
565 extra analyses on the side before the transformation. A strong type
566 system makes it easier to read the generated code and enables novel
567 analyses and transformations that are not feasible to perform on normal
568 three address code representations.</p>
572 <!-- ======================================================================= -->
573 <div class="doc_subsection"> <a name="t_primitive">Primitive Types</a> </div>
574 <div class="doc_text">
575 <p>The primitive types are the fundamental building blocks of the LLVM
576 system. The current set of primitive types is as follows:</p>
578 <table class="layout">
583 <tr><th>Type</th><th>Description</th></tr>
584 <tr><td><tt>void</tt></td><td>No value</td></tr>
585 <tr><td><tt>ubyte</tt></td><td>Unsigned 8-bit value</td></tr>
586 <tr><td><tt>ushort</tt></td><td>Unsigned 16-bit value</td></tr>
587 <tr><td><tt>uint</tt></td><td>Unsigned 32-bit value</td></tr>
588 <tr><td><tt>ulong</tt></td><td>Unsigned 64-bit value</td></tr>
589 <tr><td><tt>float</tt></td><td>32-bit floating point value</td></tr>
590 <tr><td><tt>label</tt></td><td>Branch destination</td></tr>
597 <tr><th>Type</th><th>Description</th></tr>
598 <tr><td><tt>bool</tt></td><td>True or False value</td></tr>
599 <tr><td><tt>sbyte</tt></td><td>Signed 8-bit value</td></tr>
600 <tr><td><tt>short</tt></td><td>Signed 16-bit value</td></tr>
601 <tr><td><tt>int</tt></td><td>Signed 32-bit value</td></tr>
602 <tr><td><tt>long</tt></td><td>Signed 64-bit value</td></tr>
603 <tr><td><tt>double</tt></td><td>64-bit floating point value</td></tr>
611 <!-- _______________________________________________________________________ -->
612 <div class="doc_subsubsection"> <a name="t_classifications">Type
613 Classifications</a> </div>
614 <div class="doc_text">
615 <p>These different primitive types fall into a few useful
618 <table border="1" cellspacing="0" cellpadding="4">
620 <tr><th>Classification</th><th>Types</th></tr>
622 <td><a name="t_signed">signed</a></td>
623 <td><tt>sbyte, short, int, long, float, double</tt></td>
626 <td><a name="t_unsigned">unsigned</a></td>
627 <td><tt>ubyte, ushort, uint, ulong</tt></td>
630 <td><a name="t_integer">integer</a></td>
631 <td><tt>ubyte, sbyte, ushort, short, uint, int, ulong, long</tt></td>
634 <td><a name="t_integral">integral</a></td>
635 <td><tt>bool, ubyte, sbyte, ushort, short, uint, int, ulong, long</tt>
639 <td><a name="t_floating">floating point</a></td>
640 <td><tt>float, double</tt></td>
643 <td><a name="t_firstclass">first class</a></td>
644 <td><tt>bool, ubyte, sbyte, ushort, short, uint, int, ulong, long,<br>
645 float, double, <a href="#t_pointer">pointer</a>,
646 <a href="#t_packed">packed</a></tt></td>
651 <p>The <a href="#t_firstclass">first class</a> types are perhaps the
652 most important. Values of these types are the only ones which can be
653 produced by instructions, passed as arguments, or used as operands to
654 instructions. This means that all structures and arrays must be
655 manipulated either by pointer or by component.</p>
658 <!-- ======================================================================= -->
659 <div class="doc_subsection"> <a name="t_derived">Derived Types</a> </div>
661 <div class="doc_text">
663 <p>The real power in LLVM comes from the derived types in the system.
664 This is what allows a programmer to represent arrays, functions,
665 pointers, and other useful types. Note that these derived types may be
666 recursive: For example, it is possible to have a two dimensional array.</p>
670 <!-- _______________________________________________________________________ -->
671 <div class="doc_subsubsection"> <a name="t_array">Array Type</a> </div>
673 <div class="doc_text">
677 <p>The array type is a very simple derived type that arranges elements
678 sequentially in memory. The array type requires a size (number of
679 elements) and an underlying data type.</p>
684 [<# elements> x <elementtype>]
687 <p>The number of elements is a constant integer value; elementtype may
688 be any type with a size.</p>
691 <table class="layout">
694 <tt>[40 x int ]</tt><br/>
695 <tt>[41 x int ]</tt><br/>
696 <tt>[40 x uint]</tt><br/>
699 Array of 40 integer values.<br/>
700 Array of 41 integer values.<br/>
701 Array of 40 unsigned integer values.<br/>
705 <p>Here are some examples of multidimensional arrays:</p>
706 <table class="layout">
709 <tt>[3 x [4 x int]]</tt><br/>
710 <tt>[12 x [10 x float]]</tt><br/>
711 <tt>[2 x [3 x [4 x uint]]]</tt><br/>
714 3x4 array of integer values.<br/>
715 12x10 array of single precision floating point values.<br/>
716 2x3x4 array of unsigned integer values.<br/>
721 <p>Note that 'variable sized arrays' can be implemented in LLVM With a zero
722 length array. Normally accesses past the end of an array are undefined in
723 LLVM (e.g. it is illegal to access the 5th element of a 3 element array).
724 As a special case, however, zero length arrays are recognized to be variable
725 length. This allows implementation of 'pascal style arrays' with the LLVM
726 type "{ int, [0 x float]}", for example.</p>
730 <!-- _______________________________________________________________________ -->
731 <div class="doc_subsubsection"> <a name="t_function">Function Type</a> </div>
732 <div class="doc_text">
734 <p>The function type can be thought of as a function signature. It
735 consists of a return type and a list of formal parameter types.
736 Function types are usually used to build virtual function tables
737 (which are structures of pointers to functions), for indirect function
738 calls, and when defining a function.</p>
740 The return type of a function type cannot be an aggregate type.
743 <pre> <returntype> (<parameter list>)<br></pre>
744 <p>Where '<tt><parameter list></tt>' is a comma-separated list of type
745 specifiers. Optionally, the parameter list may include a type <tt>...</tt>,
746 which indicates that the function takes a variable number of arguments.
747 Variable argument functions can access their arguments with the <a
748 href="#int_varargs">variable argument handling intrinsic</a> functions.</p>
750 <table class="layout">
753 <tt>int (int)</tt> <br/>
754 <tt>float (int, int *) *</tt><br/>
755 <tt>int (sbyte *, ...)</tt><br/>
758 function taking an <tt>int</tt>, returning an <tt>int</tt><br/>
759 <a href="#t_pointer">Pointer</a> to a function that takes an
760 <tt>int</tt> and a <a href="#t_pointer">pointer</a> to <tt>int</tt>,
761 returning <tt>float</tt>.<br/>
762 A vararg function that takes at least one <a href="#t_pointer">pointer</a>
763 to <tt>sbyte</tt> (signed char in C), which returns an integer. This is
764 the signature for <tt>printf</tt> in LLVM.<br/>
770 <!-- _______________________________________________________________________ -->
771 <div class="doc_subsubsection"> <a name="t_struct">Structure Type</a> </div>
772 <div class="doc_text">
774 <p>The structure type is used to represent a collection of data members
775 together in memory. The packing of the field types is defined to match
776 the ABI of the underlying processor. The elements of a structure may
777 be any type that has a size.</p>
778 <p>Structures are accessed using '<tt><a href="#i_load">load</a></tt>
779 and '<tt><a href="#i_store">store</a></tt>' by getting a pointer to a
780 field with the '<tt><a href="#i_getelementptr">getelementptr</a></tt>'
783 <pre> { <type list> }<br></pre>
785 <table class="layout">
788 <tt>{ int, int, int }</tt><br/>
789 <tt>{ float, int (int) * }</tt><br/>
792 a triple of three <tt>int</tt> values<br/>
793 A pair, where the first element is a <tt>float</tt> and the second element
794 is a <a href="#t_pointer">pointer</a> to a <a href="#t_function">function</a>
795 that takes an <tt>int</tt>, returning an <tt>int</tt>.<br/>
801 <!-- _______________________________________________________________________ -->
802 <div class="doc_subsubsection"> <a name="t_pointer">Pointer Type</a> </div>
803 <div class="doc_text">
805 <p>As in many languages, the pointer type represents a pointer or
806 reference to another object, which must live in memory.</p>
808 <pre> <type> *<br></pre>
810 <table class="layout">
813 <tt>[4x int]*</tt><br/>
814 <tt>int (int *) *</tt><br/>
817 A <a href="#t_pointer">pointer</a> to <a href="#t_array">array</a> of
818 four <tt>int</tt> values<br/>
819 A <a href="#t_pointer">pointer</a> to a <a
820 href="#t_function">function</a> that takes an <tt>int*</tt>, returning an
827 <!-- _______________________________________________________________________ -->
828 <div class="doc_subsubsection"> <a name="t_packed">Packed Type</a> </div>
829 <div class="doc_text">
833 <p>A packed type is a simple derived type that represents a vector
834 of elements. Packed types are used when multiple primitive data
835 are operated in parallel using a single instruction (SIMD).
836 A packed type requires a size (number of
837 elements) and an underlying primitive data type. Packed types are
838 considered <a href="#t_firstclass">first class</a>.</p>
843 < <# elements> x <elementtype> >
846 <p>The number of elements is a constant integer value; elementtype may
847 be any integral or floating point type.</p>
851 <table class="layout">
854 <tt><4 x int></tt><br/>
855 <tt><8 x float></tt><br/>
856 <tt><2 x uint></tt><br/>
859 Packed vector of 4 integer values.<br/>
860 Packed vector of 8 floating-point values.<br/>
861 Packed vector of 2 unsigned integer values.<br/>
867 <!-- _______________________________________________________________________ -->
868 <div class="doc_subsubsection"> <a name="t_opaque">Opaque Type</a> </div>
869 <div class="doc_text">
873 <p>Opaque types are used to represent unknown types in the system. This
874 corresponds (for example) to the C notion of a foward declared structure type.
875 In LLVM, opaque types can eventually be resolved to any type (not just a
886 <table class="layout">
899 <!-- *********************************************************************** -->
900 <div class="doc_section"> <a name="constants">Constants</a> </div>
901 <!-- *********************************************************************** -->
903 <div class="doc_text">
905 <p>LLVM has several different basic types of constants. This section describes
906 them all and their syntax.</p>
910 <!-- ======================================================================= -->
911 <div class="doc_subsection"><a name="simpleconstants">Simple Constants</a></div>
913 <div class="doc_text">
916 <dt><b>Boolean constants</b></dt>
918 <dd>The two strings '<tt>true</tt>' and '<tt>false</tt>' are both valid
919 constants of the <tt><a href="#t_primitive">bool</a></tt> type.
922 <dt><b>Integer constants</b></dt>
924 <dd>Standard integers (such as '4') are constants of the <a
925 href="#t_integer">integer</a> type. Negative numbers may be used with signed
929 <dt><b>Floating point constants</b></dt>
931 <dd>Floating point constants use standard decimal notation (e.g. 123.421),
932 exponential notation (e.g. 1.23421e+2), or a more precise hexadecimal
933 notation (see below). Floating point constants must have a <a
934 href="#t_floating">floating point</a> type. </dd>
936 <dt><b>Null pointer constants</b></dt>
938 <dd>The identifier '<tt>null</tt>' is recognized as a null pointer constant
939 and must be of <a href="#t_pointer">pointer type</a>.</dd>
943 <p>The one non-intuitive notation for constants is the optional hexadecimal form
944 of floating point constants. For example, the form '<tt>double
945 0x432ff973cafa8000</tt>' is equivalent to (but harder to read than) '<tt>double
946 4.5e+15</tt>'. The only time hexadecimal floating point constants are required
947 (and the only time that they are generated by the disassembler) is when a
948 floating point constant must be emitted but it cannot be represented as a
949 decimal floating point number. For example, NaN's, infinities, and other
950 special values are represented in their IEEE hexadecimal format so that
951 assembly and disassembly do not cause any bits to change in the constants.</p>
955 <!-- ======================================================================= -->
956 <div class="doc_subsection"><a name="aggregateconstants">Aggregate Constants</a>
959 <div class="doc_text">
960 <p>Aggregate constants arise from aggregation of simple constants
961 and smaller aggregate constants.</p>
964 <dt><b>Structure constants</b></dt>
966 <dd>Structure constants are represented with notation similar to structure
967 type definitions (a comma separated list of elements, surrounded by braces
968 (<tt>{}</tt>)). For example: "<tt>{ int 4, float 17.0, int* %G }</tt>",
969 where "<tt>%G</tt>" is declared as "<tt>%G = external global int</tt>". Structure constants
970 must have <a href="#t_struct">structure type</a>, and the number and
971 types of elements must match those specified by the type.
974 <dt><b>Array constants</b></dt>
976 <dd>Array constants are represented with notation similar to array type
977 definitions (a comma separated list of elements, surrounded by square brackets
978 (<tt>[]</tt>)). For example: "<tt>[ int 42, int 11, int 74 ]</tt>". Array
979 constants must have <a href="#t_array">array type</a>, and the number and
980 types of elements must match those specified by the type.
983 <dt><b>Packed constants</b></dt>
985 <dd>Packed constants are represented with notation similar to packed type
986 definitions (a comma separated list of elements, surrounded by
987 less-than/greater-than's (<tt><></tt>)). For example: "<tt>< int 42,
988 int 11, int 74, int 100 ></tt>". Packed constants must have <a
989 href="#t_packed">packed type</a>, and the number and types of elements must
990 match those specified by the type.
993 <dt><b>Zero initialization</b></dt>
995 <dd>The string '<tt>zeroinitializer</tt>' can be used to zero initialize a
996 value to zero of <em>any</em> type, including scalar and aggregate types.
997 This is often used to avoid having to print large zero initializers (e.g. for
998 large arrays), and is always exactly equivalent to using explicit zero
1005 <!-- ======================================================================= -->
1006 <div class="doc_subsection">
1007 <a name="globalconstants">Global Variable and Function Addresses</a>
1010 <div class="doc_text">
1012 <p>The addresses of <a href="#globalvars">global variables</a> and <a
1013 href="#functionstructure">functions</a> are always implicitly valid (link-time)
1014 constants. These constants are explicitly referenced when the <a
1015 href="#identifiers">identifier for the global</a> is used and always have <a
1016 href="#t_pointer">pointer</a> type. For example, the following is a legal LLVM
1022 %Z = global [2 x int*] [ int* %X, int* %Y ]
1027 <!-- ======================================================================= -->
1028 <div class="doc_subsection"><a name="undefvalues">Undefined Values</a></div>
1029 <div class="doc_text">
1030 <p>The string '<tt>undef</tt>' is recognized as a type-less constant that has
1031 no specific value. Undefined values may be of any type and be used anywhere
1032 a constant is permitted.</p>
1034 <p>Undefined values indicate to the compiler that the program is well defined
1035 no matter what value is used, giving the compiler more freedom to optimize.
1039 <!-- ======================================================================= -->
1040 <div class="doc_subsection"><a name="constantexprs">Constant Expressions</a>
1043 <div class="doc_text">
1045 <p>Constant expressions are used to allow expressions involving other constants
1046 to be used as constants. Constant expressions may be of any <a
1047 href="#t_firstclass">first class</a> type and may involve any LLVM operation
1048 that does not have side effects (e.g. load and call are not supported). The
1049 following is the syntax for constant expressions:</p>
1052 <dt><b><tt>cast ( CST to TYPE )</tt></b></dt>
1054 <dd>Cast a constant to another type.</dd>
1056 <dt><b><tt>getelementptr ( CSTPTR, IDX0, IDX1, ... )</tt></b></dt>
1058 <dd>Perform the <a href="#i_getelementptr">getelementptr operation</a> on
1059 constants. As with the <a href="#i_getelementptr">getelementptr</a>
1060 instruction, the index list may have zero or more indexes, which are required
1061 to make sense for the type of "CSTPTR".</dd>
1063 <dt><b><tt>OPCODE ( LHS, RHS )</tt></b></dt>
1065 <dd>Perform the specified operation of the LHS and RHS constants. OPCODE may
1066 be any of the <a href="#binaryops">binary</a> or <a href="#bitwiseops">bitwise
1067 binary</a> operations. The constraints on operands are the same as those for
1068 the corresponding instruction (e.g. no bitwise operations on floating point
1069 values are allowed).</dd>
1073 <!-- *********************************************************************** -->
1074 <div class="doc_section"> <a name="instref">Instruction Reference</a> </div>
1075 <!-- *********************************************************************** -->
1077 <div class="doc_text">
1079 <p>The LLVM instruction set consists of several different
1080 classifications of instructions: <a href="#terminators">terminator
1081 instructions</a>, <a href="#binaryops">binary instructions</a>,
1082 <a href="#bitwiseops">bitwise binary instructions</a>, <a
1083 href="#memoryops">memory instructions</a>, and <a href="#otherops">other
1084 instructions</a>.</p>
1088 <!-- ======================================================================= -->
1089 <div class="doc_subsection"> <a name="terminators">Terminator
1090 Instructions</a> </div>
1092 <div class="doc_text">
1094 <p>As mentioned <a href="#functionstructure">previously</a>, every
1095 basic block in a program ends with a "Terminator" instruction, which
1096 indicates which block should be executed after the current block is
1097 finished. These terminator instructions typically yield a '<tt>void</tt>'
1098 value: they produce control flow, not values (the one exception being
1099 the '<a href="#i_invoke"><tt>invoke</tt></a>' instruction).</p>
1100 <p>There are six different terminator instructions: the '<a
1101 href="#i_ret"><tt>ret</tt></a>' instruction, the '<a href="#i_br"><tt>br</tt></a>'
1102 instruction, the '<a href="#i_switch"><tt>switch</tt></a>' instruction,
1103 the '<a href="#i_invoke"><tt>invoke</tt></a>' instruction, the '<a
1104 href="#i_unwind"><tt>unwind</tt></a>' instruction, and the '<a
1105 href="#i_unreachable"><tt>unreachable</tt></a>' instruction.</p>
1109 <!-- _______________________________________________________________________ -->
1110 <div class="doc_subsubsection"> <a name="i_ret">'<tt>ret</tt>'
1111 Instruction</a> </div>
1112 <div class="doc_text">
1114 <pre> ret <type> <value> <i>; Return a value from a non-void function</i>
1115 ret void <i>; Return from void function</i>
1118 <p>The '<tt>ret</tt>' instruction is used to return control flow (and a
1119 value) from a function back to the caller.</p>
1120 <p>There are two forms of the '<tt>ret</tt>' instruction: one that
1121 returns a value and then causes control flow, and one that just causes
1122 control flow to occur.</p>
1124 <p>The '<tt>ret</tt>' instruction may return any '<a
1125 href="#t_firstclass">first class</a>' type. Notice that a function is
1126 not <a href="#wellformed">well formed</a> if there exists a '<tt>ret</tt>'
1127 instruction inside of the function that returns a value that does not
1128 match the return type of the function.</p>
1130 <p>When the '<tt>ret</tt>' instruction is executed, control flow
1131 returns back to the calling function's context. If the caller is a "<a
1132 href="#i_call"><tt>call</tt></a>" instruction, execution continues at
1133 the instruction after the call. If the caller was an "<a
1134 href="#i_invoke"><tt>invoke</tt></a>" instruction, execution continues
1135 at the beginning of the "normal" destination block. If the instruction
1136 returns a value, that value shall set the call or invoke instruction's
1139 <pre> ret int 5 <i>; Return an integer value of 5</i>
1140 ret void <i>; Return from a void function</i>
1143 <!-- _______________________________________________________________________ -->
1144 <div class="doc_subsubsection"> <a name="i_br">'<tt>br</tt>' Instruction</a> </div>
1145 <div class="doc_text">
1147 <pre> br bool <cond>, label <iftrue>, label <iffalse><br> br label <dest> <i>; Unconditional branch</i>
1150 <p>The '<tt>br</tt>' instruction is used to cause control flow to
1151 transfer to a different basic block in the current function. There are
1152 two forms of this instruction, corresponding to a conditional branch
1153 and an unconditional branch.</p>
1155 <p>The conditional branch form of the '<tt>br</tt>' instruction takes a
1156 single '<tt>bool</tt>' value and two '<tt>label</tt>' values. The
1157 unconditional form of the '<tt>br</tt>' instruction takes a single '<tt>label</tt>'
1158 value as a target.</p>
1160 <p>Upon execution of a conditional '<tt>br</tt>' instruction, the '<tt>bool</tt>'
1161 argument is evaluated. If the value is <tt>true</tt>, control flows
1162 to the '<tt>iftrue</tt>' <tt>label</tt> argument. If "cond" is <tt>false</tt>,
1163 control flows to the '<tt>iffalse</tt>' <tt>label</tt> argument.</p>
1165 <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
1166 href="#i_ret">ret</a> int 1<br>IfUnequal:<br> <a href="#i_ret">ret</a> int 0<br></pre>
1168 <!-- _______________________________________________________________________ -->
1169 <div class="doc_subsubsection">
1170 <a name="i_switch">'<tt>switch</tt>' Instruction</a>
1173 <div class="doc_text">
1177 switch <intty> <value>, label <defaultdest> [ <intty> <val>, label <dest> ... ]
1182 <p>The '<tt>switch</tt>' instruction is used to transfer control flow to one of
1183 several different places. It is a generalization of the '<tt>br</tt>'
1184 instruction, allowing a branch to occur to one of many possible
1190 <p>The '<tt>switch</tt>' instruction uses three parameters: an integer
1191 comparison value '<tt>value</tt>', a default '<tt>label</tt>' destination, and
1192 an array of pairs of comparison value constants and '<tt>label</tt>'s. The
1193 table is not allowed to contain duplicate constant entries.</p>
1197 <p>The <tt>switch</tt> instruction specifies a table of values and
1198 destinations. When the '<tt>switch</tt>' instruction is executed, this
1199 table is searched for the given value. If the value is found, control flow is
1200 transfered to the corresponding destination; otherwise, control flow is
1201 transfered to the default destination.</p>
1203 <h5>Implementation:</h5>
1205 <p>Depending on properties of the target machine and the particular
1206 <tt>switch</tt> instruction, this instruction may be code generated in different
1207 ways. For example, it could be generated as a series of chained conditional
1208 branches or with a lookup table.</p>
1213 <i>; Emulate a conditional br instruction</i>
1214 %Val = <a href="#i_cast">cast</a> bool %value to int
1215 switch int %Val, label %truedest [int 0, label %falsedest ]
1217 <i>; Emulate an unconditional br instruction</i>
1218 switch uint 0, label %dest [ ]
1220 <i>; Implement a jump table:</i>
1221 switch uint %val, label %otherwise [ uint 0, label %onzero
1222 uint 1, label %onone
1223 uint 2, label %ontwo ]
1227 <!-- _______________________________________________________________________ -->
1228 <div class="doc_subsubsection">
1229 <a name="i_invoke">'<tt>invoke</tt>' Instruction</a>
1232 <div class="doc_text">
1237 <result> = invoke [<a href="#callingconv">cconv</a>] <ptr to function ty> %<function ptr val>(<function args>)
1238 to label <normal label> except label <exception label>
1243 <p>The '<tt>invoke</tt>' instruction causes control to transfer to a specified
1244 function, with the possibility of control flow transfer to either the
1245 '<tt>normal</tt>' label or the
1246 '<tt>exception</tt>' label. If the callee function returns with the
1247 "<tt><a href="#i_ret">ret</a></tt>" instruction, control flow will return to the
1248 "normal" label. If the callee (or any indirect callees) returns with the "<a
1249 href="#i_unwind"><tt>unwind</tt></a>" instruction, control is interrupted and
1250 continued at the dynamically nearest "exception" label.</p>
1254 <p>This instruction requires several arguments:</p>
1258 The optional "cconv" marker indicates which <a href="callingconv">calling
1259 convention</a> the call should use. If none is specified, the call defaults
1260 to using C calling conventions.
1262 <li>'<tt>ptr to function ty</tt>': shall be the signature of the pointer to
1263 function value being invoked. In most cases, this is a direct function
1264 invocation, but indirect <tt>invoke</tt>s are just as possible, branching off
1265 an arbitrary pointer to function value.
1268 <li>'<tt>function ptr val</tt>': An LLVM value containing a pointer to a
1269 function to be invoked. </li>
1271 <li>'<tt>function args</tt>': argument list whose types match the function
1272 signature argument types. If the function signature indicates the function
1273 accepts a variable number of arguments, the extra arguments can be
1276 <li>'<tt>normal label</tt>': the label reached when the called function
1277 executes a '<tt><a href="#i_ret">ret</a></tt>' instruction. </li>
1279 <li>'<tt>exception label</tt>': the label reached when a callee returns with
1280 the <a href="#i_unwind"><tt>unwind</tt></a> instruction. </li>
1286 <p>This instruction is designed to operate as a standard '<tt><a
1287 href="#i_call">call</a></tt>' instruction in most regards. The primary
1288 difference is that it establishes an association with a label, which is used by
1289 the runtime library to unwind the stack.</p>
1291 <p>This instruction is used in languages with destructors to ensure that proper
1292 cleanup is performed in the case of either a <tt>longjmp</tt> or a thrown
1293 exception. Additionally, this is important for implementation of
1294 '<tt>catch</tt>' clauses in high-level languages that support them.</p>
1298 %retval = invoke int %Test(int 15) to label %Continue
1299 except label %TestCleanup <i>; {int}:retval set</i>
1300 %retval = invoke <a href="#callingconv">coldcc</a> int %Test(int 15) to label %Continue
1301 except label %TestCleanup <i>; {int}:retval set</i>
1306 <!-- _______________________________________________________________________ -->
1308 <div class="doc_subsubsection"> <a name="i_unwind">'<tt>unwind</tt>'
1309 Instruction</a> </div>
1311 <div class="doc_text">
1320 <p>The '<tt>unwind</tt>' instruction unwinds the stack, continuing control flow
1321 at the first callee in the dynamic call stack which used an <a
1322 href="#i_invoke"><tt>invoke</tt></a> instruction to perform the call. This is
1323 primarily used to implement exception handling.</p>
1327 <p>The '<tt>unwind</tt>' intrinsic causes execution of the current function to
1328 immediately halt. The dynamic call stack is then searched for the first <a
1329 href="#i_invoke"><tt>invoke</tt></a> instruction on the call stack. Once found,
1330 execution continues at the "exceptional" destination block specified by the
1331 <tt>invoke</tt> instruction. If there is no <tt>invoke</tt> instruction in the
1332 dynamic call chain, undefined behavior results.</p>
1335 <!-- _______________________________________________________________________ -->
1337 <div class="doc_subsubsection"> <a name="i_unreachable">'<tt>unreachable</tt>'
1338 Instruction</a> </div>
1340 <div class="doc_text">
1349 <p>The '<tt>unreachable</tt>' instruction has no defined semantics. This
1350 instruction is used to inform the optimizer that a particular portion of the
1351 code is not reachable. This can be used to indicate that the code after a
1352 no-return function cannot be reached, and other facts.</p>
1356 <p>The '<tt>unreachable</tt>' instruction has no defined semantics.</p>
1361 <!-- ======================================================================= -->
1362 <div class="doc_subsection"> <a name="binaryops">Binary Operations</a> </div>
1363 <div class="doc_text">
1364 <p>Binary operators are used to do most of the computation in a
1365 program. They require two operands, execute an operation on them, and
1366 produce a single value. The operands might represent
1367 multiple data, as is the case with the <a href="#t_packed">packed</a> data type.
1368 The result value of a binary operator is not
1369 necessarily the same type as its operands.</p>
1370 <p>There are several different binary operators:</p>
1372 <!-- _______________________________________________________________________ -->
1373 <div class="doc_subsubsection"> <a name="i_add">'<tt>add</tt>'
1374 Instruction</a> </div>
1375 <div class="doc_text">
1377 <pre> <result> = add <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1380 <p>The '<tt>add</tt>' instruction returns the sum of its two operands.</p>
1382 <p>The two arguments to the '<tt>add</tt>' instruction must be either <a
1383 href="#t_integer">integer</a> or <a href="#t_floating">floating point</a> values.
1384 This instruction can also take <a href="#t_packed">packed</a> versions of the values.
1385 Both arguments must have identical types.</p>
1387 <p>The value produced is the integer or floating point sum of the two
1390 <pre> <result> = add int 4, %var <i>; yields {int}:result = 4 + %var</i>
1393 <!-- _______________________________________________________________________ -->
1394 <div class="doc_subsubsection"> <a name="i_sub">'<tt>sub</tt>'
1395 Instruction</a> </div>
1396 <div class="doc_text">
1398 <pre> <result> = sub <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1401 <p>The '<tt>sub</tt>' instruction returns the difference of its two
1403 <p>Note that the '<tt>sub</tt>' instruction is used to represent the '<tt>neg</tt>'
1404 instruction present in most other intermediate representations.</p>
1406 <p>The two arguments to the '<tt>sub</tt>' instruction must be either <a
1407 href="#t_integer">integer</a> or <a href="#t_floating">floating point</a>
1409 This instruction can also take <a href="#t_packed">packed</a> versions of the values.
1410 Both arguments must have identical types.</p>
1412 <p>The value produced is the integer or floating point difference of
1413 the two operands.</p>
1415 <pre> <result> = sub int 4, %var <i>; yields {int}:result = 4 - %var</i>
1416 <result> = sub int 0, %val <i>; yields {int}:result = -%var</i>
1419 <!-- _______________________________________________________________________ -->
1420 <div class="doc_subsubsection"> <a name="i_mul">'<tt>mul</tt>'
1421 Instruction</a> </div>
1422 <div class="doc_text">
1424 <pre> <result> = mul <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1427 <p>The '<tt>mul</tt>' instruction returns the product of its two
1430 <p>The two arguments to the '<tt>mul</tt>' instruction must be either <a
1431 href="#t_integer">integer</a> or <a href="#t_floating">floating point</a>
1433 This instruction can also take <a href="#t_packed">packed</a> versions of the values.
1434 Both arguments must have identical types.</p>
1436 <p>The value produced is the integer or floating point product of the
1438 <p>There is no signed vs unsigned multiplication. The appropriate
1439 action is taken based on the type of the operand.</p>
1441 <pre> <result> = mul int 4, %var <i>; yields {int}:result = 4 * %var</i>
1444 <!-- _______________________________________________________________________ -->
1445 <div class="doc_subsubsection"> <a name="i_div">'<tt>div</tt>'
1446 Instruction</a> </div>
1447 <div class="doc_text">
1449 <pre> <result> = div <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1452 <p>The '<tt>div</tt>' instruction returns the quotient of its two
1455 <p>The two arguments to the '<tt>div</tt>' instruction must be either <a
1456 href="#t_integer">integer</a> or <a href="#t_floating">floating point</a>
1458 This instruction can also take <a href="#t_packed">packed</a> versions of the values.
1459 Both arguments must have identical types.</p>
1461 <p>The value produced is the integer or floating point quotient of the
1464 <pre> <result> = div int 4, %var <i>; yields {int}:result = 4 / %var</i>
1467 <!-- _______________________________________________________________________ -->
1468 <div class="doc_subsubsection"> <a name="i_rem">'<tt>rem</tt>'
1469 Instruction</a> </div>
1470 <div class="doc_text">
1472 <pre> <result> = rem <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1475 <p>The '<tt>rem</tt>' instruction returns the remainder from the
1476 division of its two operands.</p>
1478 <p>The two arguments to the '<tt>rem</tt>' instruction must be either <a
1479 href="#t_integer">integer</a> or <a href="#t_floating">floating point</a>
1481 This instruction can also take <a href="#t_packed">packed</a> versions of the values.
1482 Both arguments must have identical types.</p>
1484 <p>This returns the <i>remainder</i> of a division (where the result
1485 has the same sign as the divisor), not the <i>modulus</i> (where the
1486 result has the same sign as the dividend) of a value. For more
1487 information about the difference, see: <a
1488 href="http://mathforum.org/dr.math/problems/anne.4.28.99.html">The
1491 <pre> <result> = rem int 4, %var <i>; yields {int}:result = 4 % %var</i>
1494 <!-- _______________________________________________________________________ -->
1495 <div class="doc_subsubsection"> <a name="i_setcc">'<tt>set<i>cc</i></tt>'
1496 Instructions</a> </div>
1497 <div class="doc_text">
1499 <pre> <result> = seteq <ty> <var1>, <var2> <i>; yields {bool}:result</i>
1500 <result> = setne <ty> <var1>, <var2> <i>; yields {bool}:result</i>
1501 <result> = setlt <ty> <var1>, <var2> <i>; yields {bool}:result</i>
1502 <result> = setgt <ty> <var1>, <var2> <i>; yields {bool}:result</i>
1503 <result> = setle <ty> <var1>, <var2> <i>; yields {bool}:result</i>
1504 <result> = setge <ty> <var1>, <var2> <i>; yields {bool}:result</i>
1507 <p>The '<tt>set<i>cc</i></tt>' family of instructions returns a boolean
1508 value based on a comparison of their two operands.</p>
1510 <p>The two arguments to the '<tt>set<i>cc</i></tt>' instructions must
1511 be of <a href="#t_firstclass">first class</a> type (it is not possible
1512 to compare '<tt>label</tt>'s, '<tt>array</tt>'s, '<tt>structure</tt>'
1513 or '<tt>void</tt>' values, etc...). Both arguments must have identical
1516 <p>The '<tt>seteq</tt>' instruction yields a <tt>true</tt> '<tt>bool</tt>'
1517 value if both operands are equal.<br>
1518 The '<tt>setne</tt>' instruction yields a <tt>true</tt> '<tt>bool</tt>'
1519 value if both operands are unequal.<br>
1520 The '<tt>setlt</tt>' instruction yields a <tt>true</tt> '<tt>bool</tt>'
1521 value if the first operand is less than the second operand.<br>
1522 The '<tt>setgt</tt>' instruction yields a <tt>true</tt> '<tt>bool</tt>'
1523 value if the first operand is greater than the second operand.<br>
1524 The '<tt>setle</tt>' instruction yields a <tt>true</tt> '<tt>bool</tt>'
1525 value if the first operand is less than or equal to the second operand.<br>
1526 The '<tt>setge</tt>' instruction yields a <tt>true</tt> '<tt>bool</tt>'
1527 value if the first operand is greater than or equal to the second
1530 <pre> <result> = seteq int 4, 5 <i>; yields {bool}:result = false</i>
1531 <result> = setne float 4, 5 <i>; yields {bool}:result = true</i>
1532 <result> = setlt uint 4, 5 <i>; yields {bool}:result = true</i>
1533 <result> = setgt sbyte 4, 5 <i>; yields {bool}:result = false</i>
1534 <result> = setle sbyte 4, 5 <i>; yields {bool}:result = true</i>
1535 <result> = setge sbyte 4, 5 <i>; yields {bool}:result = false</i>
1538 <!-- ======================================================================= -->
1539 <div class="doc_subsection"> <a name="bitwiseops">Bitwise Binary
1540 Operations</a> </div>
1541 <div class="doc_text">
1542 <p>Bitwise binary operators are used to do various forms of
1543 bit-twiddling in a program. They are generally very efficient
1544 instructions and can commonly be strength reduced from other
1545 instructions. They require two operands, execute an operation on them,
1546 and produce a single value. The resulting value of the bitwise binary
1547 operators is always the same type as its first operand.</p>
1549 <!-- _______________________________________________________________________ -->
1550 <div class="doc_subsubsection"> <a name="i_and">'<tt>and</tt>'
1551 Instruction</a> </div>
1552 <div class="doc_text">
1554 <pre> <result> = and <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1557 <p>The '<tt>and</tt>' instruction returns the bitwise logical and of
1558 its two operands.</p>
1560 <p>The two arguments to the '<tt>and</tt>' instruction must be <a
1561 href="#t_integral">integral</a> values. Both arguments must have
1562 identical types.</p>
1564 <p>The truth table used for the '<tt>and</tt>' instruction is:</p>
1566 <div style="align: center">
1567 <table border="1" cellspacing="0" cellpadding="4">
1598 <pre> <result> = and int 4, %var <i>; yields {int}:result = 4 & %var</i>
1599 <result> = and int 15, 40 <i>; yields {int}:result = 8</i>
1600 <result> = and int 4, 8 <i>; yields {int}:result = 0</i>
1603 <!-- _______________________________________________________________________ -->
1604 <div class="doc_subsubsection"> <a name="i_or">'<tt>or</tt>' Instruction</a> </div>
1605 <div class="doc_text">
1607 <pre> <result> = or <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1610 <p>The '<tt>or</tt>' instruction returns the bitwise logical inclusive
1611 or of its two operands.</p>
1613 <p>The two arguments to the '<tt>or</tt>' instruction must be <a
1614 href="#t_integral">integral</a> values. Both arguments must have
1615 identical types.</p>
1617 <p>The truth table used for the '<tt>or</tt>' instruction is:</p>
1619 <div style="align: center">
1620 <table border="1" cellspacing="0" cellpadding="4">
1651 <pre> <result> = or int 4, %var <i>; yields {int}:result = 4 | %var</i>
1652 <result> = or int 15, 40 <i>; yields {int}:result = 47</i>
1653 <result> = or int 4, 8 <i>; yields {int}:result = 12</i>
1656 <!-- _______________________________________________________________________ -->
1657 <div class="doc_subsubsection"> <a name="i_xor">'<tt>xor</tt>'
1658 Instruction</a> </div>
1659 <div class="doc_text">
1661 <pre> <result> = xor <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1664 <p>The '<tt>xor</tt>' instruction returns the bitwise logical exclusive
1665 or of its two operands. The <tt>xor</tt> is used to implement the
1666 "one's complement" operation, which is the "~" operator in C.</p>
1668 <p>The two arguments to the '<tt>xor</tt>' instruction must be <a
1669 href="#t_integral">integral</a> values. Both arguments must have
1670 identical types.</p>
1672 <p>The truth table used for the '<tt>xor</tt>' instruction is:</p>
1674 <div style="align: center">
1675 <table border="1" cellspacing="0" cellpadding="4">
1707 <pre> <result> = xor int 4, %var <i>; yields {int}:result = 4 ^ %var</i>
1708 <result> = xor int 15, 40 <i>; yields {int}:result = 39</i>
1709 <result> = xor int 4, 8 <i>; yields {int}:result = 12</i>
1710 <result> = xor int %V, -1 <i>; yields {int}:result = ~%V</i>
1713 <!-- _______________________________________________________________________ -->
1714 <div class="doc_subsubsection"> <a name="i_shl">'<tt>shl</tt>'
1715 Instruction</a> </div>
1716 <div class="doc_text">
1718 <pre> <result> = shl <ty> <var1>, ubyte <var2> <i>; yields {ty}:result</i>
1721 <p>The '<tt>shl</tt>' instruction returns the first operand shifted to
1722 the left a specified number of bits.</p>
1724 <p>The first argument to the '<tt>shl</tt>' instruction must be an <a
1725 href="#t_integer">integer</a> type. The second argument must be an '<tt>ubyte</tt>'
1728 <p>The value produced is <tt>var1</tt> * 2<sup><tt>var2</tt></sup>.</p>
1730 <pre> <result> = shl int 4, ubyte %var <i>; yields {int}:result = 4 << %var</i>
1731 <result> = shl int 4, ubyte 2 <i>; yields {int}:result = 16</i>
1732 <result> = shl int 1, ubyte 10 <i>; yields {int}:result = 1024</i>
1735 <!-- _______________________________________________________________________ -->
1736 <div class="doc_subsubsection"> <a name="i_shr">'<tt>shr</tt>'
1737 Instruction</a> </div>
1738 <div class="doc_text">
1740 <pre> <result> = shr <ty> <var1>, ubyte <var2> <i>; yields {ty}:result</i>
1743 <p>The '<tt>shr</tt>' instruction returns the first operand shifted to
1744 the right a specified number of bits.</p>
1746 <p>The first argument to the '<tt>shr</tt>' instruction must be an <a
1747 href="#t_integer">integer</a> type. The second argument must be an '<tt>ubyte</tt>'
1750 <p>If the first argument is a <a href="#t_signed">signed</a> type, the
1751 most significant bit is duplicated in the newly free'd bit positions.
1752 If the first argument is unsigned, zero bits shall fill the empty
1755 <pre> <result> = shr int 4, ubyte %var <i>; yields {int}:result = 4 >> %var</i>
1756 <result> = shr uint 4, ubyte 1 <i>; yields {uint}:result = 2</i>
1757 <result> = shr int 4, ubyte 2 <i>; yields {int}:result = 1</i>
1758 <result> = shr sbyte 4, ubyte 3 <i>; yields {sbyte}:result = 0</i>
1759 <result> = shr sbyte -2, ubyte 1 <i>; yields {sbyte}:result = -1</i>
1762 <!-- ======================================================================= -->
1763 <div class="doc_subsection"> <a name="memoryops">Memory Access
1764 Operations</a></div>
1765 <div class="doc_text">
1766 <p>A key design point of an SSA-based representation is how it
1767 represents memory. In LLVM, no memory locations are in SSA form, which
1768 makes things very simple. This section describes how to read, write,
1769 allocate, and free memory in LLVM.</p>
1771 <!-- _______________________________________________________________________ -->
1772 <div class="doc_subsubsection"> <a name="i_malloc">'<tt>malloc</tt>'
1773 Instruction</a> </div>
1774 <div class="doc_text">
1776 <pre> <result> = malloc <type>, uint <NumElements> <i>; yields {type*}:result</i>
1777 <result> = malloc <type> <i>; yields {type*}:result</i>
1780 <p>The '<tt>malloc</tt>' instruction allocates memory from the system
1781 heap and returns a pointer to it.</p>
1783 <p>The '<tt>malloc</tt>' instruction allocates <tt>sizeof(<type>)*NumElements</tt>
1784 bytes of memory from the operating system and returns a pointer of the
1785 appropriate type to the program. The second form of the instruction is
1786 a shorter version of the first instruction that defaults to allocating
1788 <p>'<tt>type</tt>' must be a sized type.</p>
1790 <p>Memory is allocated using the system "<tt>malloc</tt>" function, and
1791 a pointer is returned.</p>
1793 <pre> %array = malloc [4 x ubyte ] <i>; yields {[%4 x ubyte]*}:array</i>
1796 href="#i_add">add</a> uint 2, 2 <i>; yields {uint}:size = uint 4</i>
1797 %array1 = malloc ubyte, uint 4 <i>; yields {ubyte*}:array1</i>
1798 %array2 = malloc [12 x ubyte], uint %size <i>; yields {[12 x ubyte]*}:array2</i>
1801 <!-- _______________________________________________________________________ -->
1802 <div class="doc_subsubsection"> <a name="i_free">'<tt>free</tt>'
1803 Instruction</a> </div>
1804 <div class="doc_text">
1806 <pre> free <type> <value> <i>; yields {void}</i>
1809 <p>The '<tt>free</tt>' instruction returns memory back to the unused
1810 memory heap to be reallocated in the future.</p>
1813 <p>'<tt>value</tt>' shall be a pointer value that points to a value
1814 that was allocated with the '<tt><a href="#i_malloc">malloc</a></tt>'
1817 <p>Access to the memory pointed to by the pointer is no longer defined
1818 after this instruction executes.</p>
1820 <pre> %array = <a href="#i_malloc">malloc</a> [4 x ubyte] <i>; yields {[4 x ubyte]*}:array</i>
1821 free [4 x ubyte]* %array
1824 <!-- _______________________________________________________________________ -->
1825 <div class="doc_subsubsection"> <a name="i_alloca">'<tt>alloca</tt>'
1826 Instruction</a> </div>
1827 <div class="doc_text">
1829 <pre> <result> = alloca <type>, uint <NumElements> <i>; yields {type*}:result</i>
1830 <result> = alloca <type> <i>; yields {type*}:result</i>
1833 <p>The '<tt>alloca</tt>' instruction allocates memory on the current
1834 stack frame of the procedure that is live until the current function
1835 returns to its caller.</p>
1837 <p>The '<tt>alloca</tt>' instruction allocates <tt>sizeof(<type>)*NumElements</tt>
1838 bytes of memory on the runtime stack, returning a pointer of the
1839 appropriate type to the program. The second form of the instruction is
1840 a shorter version of the first that defaults to allocating one element.</p>
1841 <p>'<tt>type</tt>' may be any sized type.</p>
1843 <p>Memory is allocated; a pointer is returned. '<tt>alloca</tt>'d
1844 memory is automatically released when the function returns. The '<tt>alloca</tt>'
1845 instruction is commonly used to represent automatic variables that must
1846 have an address available. When the function returns (either with the <tt><a
1847 href="#i_ret">ret</a></tt> or <tt><a href="#i_unwind">unwind</a></tt>
1848 instructions), the memory is reclaimed.</p>
1850 <pre> %ptr = alloca int <i>; yields {int*}:ptr</i>
1851 %ptr = alloca int, uint 4 <i>; yields {int*}:ptr</i>
1854 <!-- _______________________________________________________________________ -->
1855 <div class="doc_subsubsection"> <a name="i_load">'<tt>load</tt>'
1856 Instruction</a> </div>
1857 <div class="doc_text">
1859 <pre> <result> = load <ty>* <pointer><br> <result> = volatile load <ty>* <pointer><br></pre>
1861 <p>The '<tt>load</tt>' instruction is used to read from memory.</p>
1863 <p>The argument to the '<tt>load</tt>' instruction specifies the memory
1864 address to load from. The pointer must point to a <a
1865 href="#t_firstclass">first class</a> type. If the <tt>load</tt> is
1866 marked as <tt>volatile</tt> then the optimizer is not allowed to modify
1867 the number or order of execution of this <tt>load</tt> with other
1868 volatile <tt>load</tt> and <tt><a href="#i_store">store</a></tt>
1871 <p>The location of memory pointed to is loaded.</p>
1873 <pre> %ptr = <a href="#i_alloca">alloca</a> int <i>; yields {int*}:ptr</i>
1875 href="#i_store">store</a> int 3, int* %ptr <i>; yields {void}</i>
1876 %val = load int* %ptr <i>; yields {int}:val = int 3</i>
1879 <!-- _______________________________________________________________________ -->
1880 <div class="doc_subsubsection"> <a name="i_store">'<tt>store</tt>'
1881 Instruction</a> </div>
1883 <pre> store <ty> <value>, <ty>* <pointer> <i>; yields {void}</i>
1884 volatile store <ty> <value>, <ty>* <pointer> <i>; yields {void}</i>
1887 <p>The '<tt>store</tt>' instruction is used to write to memory.</p>
1889 <p>There are two arguments to the '<tt>store</tt>' instruction: a value
1890 to store and an address to store it into. The type of the '<tt><pointer></tt>'
1891 operand must be a pointer to the type of the '<tt><value></tt>'
1892 operand. If the <tt>store</tt> is marked as <tt>volatile</tt>, then the
1893 optimizer is not allowed to modify the number or order of execution of
1894 this <tt>store</tt> with other volatile <tt>load</tt> and <tt><a
1895 href="#i_store">store</a></tt> instructions.</p>
1897 <p>The contents of memory are updated to contain '<tt><value></tt>'
1898 at the location specified by the '<tt><pointer></tt>' operand.</p>
1900 <pre> %ptr = <a href="#i_alloca">alloca</a> int <i>; yields {int*}:ptr</i>
1902 href="#i_store">store</a> int 3, int* %ptr <i>; yields {void}</i>
1903 %val = load int* %ptr <i>; yields {int}:val = int 3</i>
1905 <!-- _______________________________________________________________________ -->
1906 <div class="doc_subsubsection">
1907 <a name="i_getelementptr">'<tt>getelementptr</tt>' Instruction</a>
1910 <div class="doc_text">
1913 <result> = getelementptr <ty>* <ptrval>{, <ty> <idx>}*
1919 The '<tt>getelementptr</tt>' instruction is used to get the address of a
1920 subelement of an aggregate data structure.</p>
1924 <p>This instruction takes a list of integer constants that indicate what
1925 elements of the aggregate object to index to. The actual types of the arguments
1926 provided depend on the type of the first pointer argument. The
1927 '<tt>getelementptr</tt>' instruction is used to index down through the type
1928 levels of a structure or to a specific index in an array. When indexing into a
1929 structure, only <tt>uint</tt>
1930 integer constants are allowed. When indexing into an array or pointer,
1931 <tt>int</tt> and <tt>long</tt> indexes are allowed of any sign.</p>
1933 <p>For example, let's consider a C code fragment and how it gets
1934 compiled to LLVM:</p>
1948 int *foo(struct ST *s) {
1949 return &s[1].Z.B[5][13];
1953 <p>The LLVM code generated by the GCC frontend is:</p>
1956 %RT = type { sbyte, [10 x [20 x int]], sbyte }
1957 %ST = type { int, double, %RT }
1961 int* %foo(%ST* %s) {
1963 %reg = getelementptr %ST* %s, int 1, uint 2, uint 1, int 5, int 13
1970 <p>The index types specified for the '<tt>getelementptr</tt>' instruction depend
1971 on the pointer type that is being indexed into. <a href="#t_pointer">Pointer</a>
1972 and <a href="#t_array">array</a> types require <tt>uint</tt>, <tt>int</tt>,
1973 <tt>ulong</tt>, or <tt>long</tt> values, and <a href="#t_struct">structure</a>
1974 types require <tt>uint</tt> <b>constants</b>.</p>
1976 <p>In the example above, the first index is indexing into the '<tt>%ST*</tt>'
1977 type, which is a pointer, yielding a '<tt>%ST</tt>' = '<tt>{ int, double, %RT
1978 }</tt>' type, a structure. The second index indexes into the third element of
1979 the structure, yielding a '<tt>%RT</tt>' = '<tt>{ sbyte, [10 x [20 x int]],
1980 sbyte }</tt>' type, another structure. The third index indexes into the second
1981 element of the structure, yielding a '<tt>[10 x [20 x int]]</tt>' type, an
1982 array. The two dimensions of the array are subscripted into, yielding an
1983 '<tt>int</tt>' type. The '<tt>getelementptr</tt>' instruction returns a pointer
1984 to this element, thus computing a value of '<tt>int*</tt>' type.</p>
1986 <p>Note that it is perfectly legal to index partially through a
1987 structure, returning a pointer to an inner element. Because of this,
1988 the LLVM code for the given testcase is equivalent to:</p>
1991 int* %foo(%ST* %s) {
1992 %t1 = getelementptr %ST* %s, int 1 <i>; yields %ST*:%t1</i>
1993 %t2 = getelementptr %ST* %t1, int 0, uint 2 <i>; yields %RT*:%t2</i>
1994 %t3 = getelementptr %RT* %t2, int 0, uint 1 <i>; yields [10 x [20 x int]]*:%t3</i>
1995 %t4 = getelementptr [10 x [20 x int]]* %t3, int 0, int 5 <i>; yields [20 x int]*:%t4</i>
1996 %t5 = getelementptr [20 x int]* %t4, int 0, int 13 <i>; yields int*:%t5</i>
2001 <p>Note that it is undefined to access an array out of bounds: array and
2002 pointer indexes must always be within the defined bounds of the array type.
2003 The one exception for this rules is zero length arrays. These arrays are
2004 defined to be accessible as variable length arrays, which requires access
2005 beyond the zero'th element.</p>
2010 <i>; yields [12 x ubyte]*:aptr</i>
2011 %aptr = getelementptr {int, [12 x ubyte]}* %sptr, long 0, uint 1
2015 <!-- ======================================================================= -->
2016 <div class="doc_subsection"> <a name="otherops">Other Operations</a> </div>
2017 <div class="doc_text">
2018 <p>The instructions in this category are the "miscellaneous"
2019 instructions, which defy better classification.</p>
2021 <!-- _______________________________________________________________________ -->
2022 <div class="doc_subsubsection"> <a name="i_phi">'<tt>phi</tt>'
2023 Instruction</a> </div>
2024 <div class="doc_text">
2026 <pre> <result> = phi <ty> [ <val0>, <label0>], ...<br></pre>
2028 <p>The '<tt>phi</tt>' instruction is used to implement the φ node in
2029 the SSA graph representing the function.</p>
2031 <p>The type of the incoming values are specified with the first type
2032 field. After this, the '<tt>phi</tt>' instruction takes a list of pairs
2033 as arguments, with one pair for each predecessor basic block of the
2034 current block. Only values of <a href="#t_firstclass">first class</a>
2035 type may be used as the value arguments to the PHI node. Only labels
2036 may be used as the label arguments.</p>
2037 <p>There must be no non-phi instructions between the start of a basic
2038 block and the PHI instructions: i.e. PHI instructions must be first in
2041 <p>At runtime, the '<tt>phi</tt>' instruction logically takes on the
2042 value specified by the parameter, depending on which basic block we
2043 came from in the last <a href="#terminators">terminator</a> instruction.</p>
2045 <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>
2048 <!-- _______________________________________________________________________ -->
2049 <div class="doc_subsubsection">
2050 <a name="i_cast">'<tt>cast .. to</tt>' Instruction</a>
2053 <div class="doc_text">
2058 <result> = cast <ty> <value> to <ty2> <i>; yields ty2</i>
2064 The '<tt>cast</tt>' instruction is used as the primitive means to convert
2065 integers to floating point, change data type sizes, and break type safety (by
2073 The '<tt>cast</tt>' instruction takes a value to cast, which must be a first
2074 class value, and a type to cast it to, which must also be a <a
2075 href="#t_firstclass">first class</a> type.
2081 This instruction follows the C rules for explicit casts when determining how the
2082 data being cast must change to fit in its new container.
2086 When casting to bool, any value that would be considered true in the context of
2087 a C '<tt>if</tt>' condition is converted to the boolean '<tt>true</tt>' values,
2088 all else are '<tt>false</tt>'.
2092 When extending an integral value from a type of one signness to another (for
2093 example '<tt>sbyte</tt>' to '<tt>ulong</tt>'), the value is sign-extended if the
2094 <b>source</b> value is signed, and zero-extended if the source value is
2095 unsigned. <tt>bool</tt> values are always zero extended into either zero or
2102 %X = cast int 257 to ubyte <i>; yields ubyte:1</i>
2103 %Y = cast int 123 to bool <i>; yields bool:true</i>
2107 <!-- _______________________________________________________________________ -->
2108 <div class="doc_subsubsection">
2109 <a name="i_select">'<tt>select</tt>' Instruction</a>
2112 <div class="doc_text">
2117 <result> = select bool <cond>, <ty> <val1>, <ty> <val2> <i>; yields ty</i>
2123 The '<tt>select</tt>' instruction is used to choose one value based on a
2124 condition, without branching.
2131 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.
2137 If the boolean condition evaluates to true, the instruction returns the first
2138 value argument; otherwise, it returns the second value argument.
2144 %X = select bool true, ubyte 17, ubyte 42 <i>; yields ubyte:17</i>
2152 <!-- _______________________________________________________________________ -->
2153 <div class="doc_subsubsection">
2154 <a name="i_call">'<tt>call</tt>' Instruction</a>
2157 <div class="doc_text">
2161 <result> = [tail] call [<a href="#callingconv">cconv</a>] <ty>* <fnptrval>(<param list>)
2166 <p>The '<tt>call</tt>' instruction represents a simple function call.</p>
2170 <p>This instruction requires several arguments:</p>
2174 <p>The optional "tail" marker indicates whether the callee function accesses
2175 any allocas or varargs in the caller. If the "tail" marker is present, the
2176 function call is eligible for tail call optimization. Note that calls may
2177 be marked "tail" even if they do not occur before a <a
2178 href="#i_ret"><tt>ret</tt></a> instruction.
2181 <p>The optional "cconv" marker indicates which <a href="callingconv">calling
2182 convention</a> the call should use. If none is specified, the call defaults
2183 to using C calling conventions.
2186 <p>'<tt>ty</tt>': shall be the signature of the pointer to function value
2187 being invoked. The argument types must match the types implied by this
2188 signature. This type can be omitted if the function is not varargs and
2189 if the function type does not return a pointer to a function.</p>
2192 <p>'<tt>fnptrval</tt>': An LLVM value containing a pointer to a function to
2193 be invoked. In most cases, this is a direct function invocation, but
2194 indirect <tt>call</tt>s are just as possible, calling an arbitrary pointer
2195 to function value.</p>
2198 <p>'<tt>function args</tt>': argument list whose types match the
2199 function signature argument types. All arguments must be of
2200 <a href="#t_firstclass">first class</a> type. If the function signature
2201 indicates the function accepts a variable number of arguments, the extra
2202 arguments can be specified.</p>
2208 <p>The '<tt>call</tt>' instruction is used to cause control flow to
2209 transfer to a specified function, with its incoming arguments bound to
2210 the specified values. Upon a '<tt><a href="#i_ret">ret</a></tt>'
2211 instruction in the called function, control flow continues with the
2212 instruction after the function call, and the return value of the
2213 function is bound to the result argument. This is a simpler case of
2214 the <a href="#i_invoke">invoke</a> instruction.</p>
2219 %retval = call int %test(int %argc)
2220 call int(sbyte*, ...) *%printf(sbyte* %msg, int 12, sbyte 42);
2221 %X = tail call int %foo()
2222 %Y = tail call <a href="#callingconv">fastcc</a> int %foo()
2227 <!-- _______________________________________________________________________ -->
2228 <div class="doc_subsubsection">
2229 <a name="i_vaarg">'<tt>vaarg</tt>' Instruction</a>
2232 <div class="doc_text">
2237 <resultval> = va_arg <va_list*> <arglist>, <argty>
2242 <p>The '<tt>va_arg</tt>' instruction is used to access arguments passed through
2243 the "variable argument" area of a function call. It is used to implement the
2244 <tt>va_arg</tt> macro in C.</p>
2248 <p>This instruction takes a <tt>va_list*</tt> value and the type of
2249 the argument. It returns a value of the specified argument type and
2250 increments the <tt>va_list</tt> to poin to the next argument. Again, the
2251 actual type of <tt>va_list</tt> is target specific.</p>
2255 <p>The '<tt>va_arg</tt>' instruction loads an argument of the specified
2256 type from the specified <tt>va_list</tt> and causes the
2257 <tt>va_list</tt> to point to the next argument. For more information,
2258 see the variable argument handling <a href="#int_varargs">Intrinsic
2261 <p>It is legal for this instruction to be called in a function which does not
2262 take a variable number of arguments, for example, the <tt>vfprintf</tt>
2265 <p><tt>va_arg</tt> is an LLVM instruction instead of an <a
2266 href="#intrinsics">intrinsic function</a> because it takes a type as an
2271 <p>See the <a href="#int_varargs">variable argument processing</a> section.</p>
2275 <!-- *********************************************************************** -->
2276 <div class="doc_section"> <a name="intrinsics">Intrinsic Functions</a> </div>
2277 <!-- *********************************************************************** -->
2279 <div class="doc_text">
2281 <p>LLVM supports the notion of an "intrinsic function". These functions have
2282 well known names and semantics and are required to follow certain
2283 restrictions. Overall, these instructions represent an extension mechanism for
2284 the LLVM language that does not require changing all of the transformations in
2285 LLVM to add to the language (or the bytecode reader/writer, the parser,
2288 <p>Intrinsic function names must all start with an "<tt>llvm.</tt>" prefix. This
2289 prefix is reserved in LLVM for intrinsic names; thus, functions may not be named
2290 this. Intrinsic functions must always be external functions: you cannot define
2291 the body of intrinsic functions. Intrinsic functions may only be used in call
2292 or invoke instructions: it is illegal to take the address of an intrinsic
2293 function. Additionally, because intrinsic functions are part of the LLVM
2294 language, it is required that they all be documented here if any are added.</p>
2297 <p>To learn how to add an intrinsic function, please see the <a
2298 href="ExtendingLLVM.html">Extending LLVM Guide</a>.
2303 <!-- ======================================================================= -->
2304 <div class="doc_subsection">
2305 <a name="int_varargs">Variable Argument Handling Intrinsics</a>
2308 <div class="doc_text">
2310 <p>Variable argument support is defined in LLVM with the <a
2311 href="#i_vanext"><tt>vanext</tt></a> instruction and these three
2312 intrinsic functions. These functions are related to the similarly
2313 named macros defined in the <tt><stdarg.h></tt> header file.</p>
2315 <p>All of these functions operate on arguments that use a
2316 target-specific value type "<tt>va_list</tt>". The LLVM assembly
2317 language reference manual does not define what this type is, so all
2318 transformations should be prepared to handle intrinsics with any type
2321 <p>This example shows how the <a href="#i_vanext"><tt>vanext</tt></a>
2322 instruction and the variable argument handling intrinsic functions are
2326 int %test(int %X, ...) {
2327 ; Initialize variable argument processing
2329 call void %<a href="#i_va_start">llvm.va_start</a>(sbyte** %ap)
2331 ; Read a single integer argument
2332 %tmp = va_arg sbyte** %ap, int
2334 ; Demonstrate usage of llvm.va_copy and llvm.va_end
2336 call void %<a href="#i_va_copy">llvm.va_copy</a>(sbyte** %aq, sbyte** %ap)
2337 call void %<a href="#i_va_end">llvm.va_end</a>(sbyte** %aq)
2339 ; Stop processing of arguments.
2340 call void %<a href="#i_va_end">llvm.va_end</a>(sbyte** %ap)
2346 <!-- _______________________________________________________________________ -->
2347 <div class="doc_subsubsection">
2348 <a name="i_va_start">'<tt>llvm.va_start</tt>' Intrinsic</a>
2352 <div class="doc_text">
2354 <pre> declare void %llvm.va_start(<va_list>* <arglist>)<br></pre>
2356 <P>The '<tt>llvm.va_start</tt>' intrinsic initializes
2357 <tt>*<arglist></tt> for subsequent use by <tt><a
2358 href="#i_va_arg">va_arg</a></tt>.</p>
2362 <P>The argument is a pointer to a <tt>va_list</tt> element to initialize.</p>
2366 <P>The '<tt>llvm.va_start</tt>' intrinsic works just like the <tt>va_start</tt>
2367 macro available in C. In a target-dependent way, it initializes the
2368 <tt>va_list</tt> element the argument points to, so that the next call to
2369 <tt>va_arg</tt> will produce the first variable argument passed to the function.
2370 Unlike the C <tt>va_start</tt> macro, this intrinsic does not need to know the
2371 last argument of the function, the compiler can figure that out.</p>
2375 <!-- _______________________________________________________________________ -->
2376 <div class="doc_subsubsection">
2377 <a name="i_va_end">'<tt>llvm.va_end</tt>' Intrinsic</a>
2380 <div class="doc_text">
2382 <pre> declare void %llvm.va_end(<va_list*> <arglist>)<br></pre>
2384 <p>The '<tt>llvm.va_end</tt>' intrinsic destroys <tt><arglist></tt>
2385 which has been initialized previously with <tt><a href="#i_va_start">llvm.va_start</a></tt>
2386 or <tt><a href="#i_va_copy">llvm.va_copy</a></tt>.</p>
2388 <p>The argument is a <tt>va_list</tt> to destroy.</p>
2390 <p>The '<tt>llvm.va_end</tt>' intrinsic works just like the <tt>va_end</tt>
2391 macro available in C. In a target-dependent way, it destroys the <tt>va_list</tt>.
2392 Calls to <a href="#i_va_start"><tt>llvm.va_start</tt></a> and <a
2393 href="#i_va_copy"><tt>llvm.va_copy</tt></a> must be matched exactly
2394 with calls to <tt>llvm.va_end</tt>.</p>
2397 <!-- _______________________________________________________________________ -->
2398 <div class="doc_subsubsection">
2399 <a name="i_va_copy">'<tt>llvm.va_copy</tt>' Intrinsic</a>
2402 <div class="doc_text">
2407 declare void %llvm.va_copy(<va_list>* <destarglist>,
2408 <va_list>* <srcarglist>)
2413 <p>The '<tt>llvm.va_copy</tt>' intrinsic copies the current argument position from
2414 the source argument list to the destination argument list.</p>
2418 <p>The first argument is a pointer to a <tt>va_list</tt> element to initialize.
2419 The second argument is a pointer to a <tt>va_list</tt> element to copy from.</p>
2424 <p>The '<tt>llvm.va_copy</tt>' intrinsic works just like the <tt>va_copy</tt> macro
2425 available in C. In a target-dependent way, it copies the source
2426 <tt>va_list</tt> element into the destination list. This intrinsic is necessary
2427 because the <tt><a href="i_va_begin">llvm.va_begin</a></tt> intrinsic may be
2428 arbitrarily complex and require memory allocation, for example.</p>
2432 <!-- ======================================================================= -->
2433 <div class="doc_subsection">
2434 <a name="int_gc">Accurate Garbage Collection Intrinsics</a>
2437 <div class="doc_text">
2440 LLVM support for <a href="GarbageCollection.html">Accurate Garbage
2441 Collection</a> requires the implementation and generation of these intrinsics.
2442 These intrinsics allow identification of <a href="#i_gcroot">GC roots on the
2443 stack</a>, as well as garbage collector implementations that require <a
2444 href="#i_gcread">read</a> and <a href="#i_gcwrite">write</a> barriers.
2445 Front-ends for type-safe garbage collected languages should generate these
2446 intrinsics to make use of the LLVM garbage collectors. For more details, see <a
2447 href="GarbageCollection.html">Accurate Garbage Collection with LLVM</a>.
2451 <!-- _______________________________________________________________________ -->
2452 <div class="doc_subsubsection">
2453 <a name="i_gcroot">'<tt>llvm.gcroot</tt>' Intrinsic</a>
2456 <div class="doc_text">
2461 declare void %llvm.gcroot(<ty>** %ptrloc, <ty2>* %metadata)
2466 <p>The '<tt>llvm.gcroot</tt>' intrinsic declares the existence of a GC root to
2467 the code generator, and allows some metadata to be associated with it.</p>
2471 <p>The first argument specifies the address of a stack object that contains the
2472 root pointer. The second pointer (which must be either a constant or a global
2473 value address) contains the meta-data to be associated with the root.</p>
2477 <p>At runtime, a call to this intrinsics stores a null pointer into the "ptrloc"
2478 location. At compile-time, the code generator generates information to allow
2479 the runtime to find the pointer at GC safe points.
2485 <!-- _______________________________________________________________________ -->
2486 <div class="doc_subsubsection">
2487 <a name="i_gcread">'<tt>llvm.gcread</tt>' Intrinsic</a>
2490 <div class="doc_text">
2495 declare sbyte* %llvm.gcread(sbyte** %Ptr)
2500 <p>The '<tt>llvm.gcread</tt>' intrinsic identifies reads of references from heap
2501 locations, allowing garbage collector implementations that require read
2506 <p>The argument is the address to read from, which should be an address
2507 allocated from the garbage collector.</p>
2511 <p>The '<tt>llvm.gcread</tt>' intrinsic has the same semantics as a load
2512 instruction, but may be replaced with substantially more complex code by the
2513 garbage collector runtime, as needed.</p>
2518 <!-- _______________________________________________________________________ -->
2519 <div class="doc_subsubsection">
2520 <a name="i_gcwrite">'<tt>llvm.gcwrite</tt>' Intrinsic</a>
2523 <div class="doc_text">
2528 declare void %llvm.gcwrite(sbyte* %P1, sbyte** %P2)
2533 <p>The '<tt>llvm.gcwrite</tt>' intrinsic identifies writes of references to heap
2534 locations, allowing garbage collector implementations that require write
2535 barriers (such as generational or reference counting collectors).</p>
2539 <p>The first argument is the reference to store, and the second is the heap
2540 location to store to.</p>
2544 <p>The '<tt>llvm.gcwrite</tt>' intrinsic has the same semantics as a store
2545 instruction, but may be replaced with substantially more complex code by the
2546 garbage collector runtime, as needed.</p>
2552 <!-- ======================================================================= -->
2553 <div class="doc_subsection">
2554 <a name="int_codegen">Code Generator Intrinsics</a>
2557 <div class="doc_text">
2559 These intrinsics are provided by LLVM to expose special features that may only
2560 be implemented with code generator support.
2565 <!-- _______________________________________________________________________ -->
2566 <div class="doc_subsubsection">
2567 <a name="i_returnaddress">'<tt>llvm.returnaddress</tt>' Intrinsic</a>
2570 <div class="doc_text">
2574 declare void* %llvm.returnaddress(uint <level>)
2580 The '<tt>llvm.returnaddress</tt>' intrinsic returns a target-specific value
2581 indicating the return address of the current function or one of its callers.
2587 The argument to this intrinsic indicates which function to return the address
2588 for. Zero indicates the calling function, one indicates its caller, etc. The
2589 argument is <b>required</b> to be a constant integer value.
2595 The '<tt>llvm.returnaddress</tt>' intrinsic either returns a pointer indicating
2596 the return address of the specified call frame, or zero if it cannot be
2597 identified. The value returned by this intrinsic is likely to be incorrect or 0
2598 for arguments other than zero, so it should only be used for debugging purposes.
2602 Note that calling this intrinsic does not prevent function inlining or other
2603 aggressive transformations, so the value returned may not be that of the obvious
2604 source-language caller.
2609 <!-- _______________________________________________________________________ -->
2610 <div class="doc_subsubsection">
2611 <a name="i_frameaddress">'<tt>llvm.frameaddress</tt>' Intrinsic</a>
2614 <div class="doc_text">
2618 declare void* %llvm.frameaddress(uint <level>)
2624 The '<tt>llvm.frameaddress</tt>' intrinsic returns the target-specific frame
2625 pointer value for the specified stack frame.
2631 The argument to this intrinsic indicates which function to return the frame
2632 pointer for. Zero indicates the calling function, one indicates its caller,
2633 etc. The argument is <b>required</b> to be a constant integer value.
2639 The '<tt>llvm.frameaddress</tt>' intrinsic either returns a pointer indicating
2640 the frame address of the specified call frame, or zero if it cannot be
2641 identified. The value returned by this intrinsic is likely to be incorrect or 0
2642 for arguments other than zero, so it should only be used for debugging purposes.
2646 Note that calling this intrinsic does not prevent function inlining or other
2647 aggressive transformations, so the value returned may not be that of the obvious
2648 source-language caller.
2652 <!-- _______________________________________________________________________ -->
2653 <div class="doc_subsubsection">
2654 <a name="i_prefetch">'<tt>llvm.prefetch</tt>' Intrinsic</a>
2657 <div class="doc_text">
2661 declare void %llvm.prefetch(sbyte * <address>,
2662 uint <rw>, uint <locality>)
2669 The '<tt>llvm.prefetch</tt>' intrinsic is a hint to the code generator to insert
2670 a prefetch instruction if supported; otherwise, it is a noop. Prefetches have
2672 effect on the behavior of the program but can change its performance
2679 <tt>address</tt> is the address to be prefetched, <tt>rw</tt> is the specifier
2680 determining if the fetch should be for a read (0) or write (1), and
2681 <tt>locality</tt> is a temporal locality specifier ranging from (0) - no
2682 locality, to (3) - extremely local keep in cache. The <tt>rw</tt> and
2683 <tt>locality</tt> arguments must be constant integers.
2689 This intrinsic does not modify the behavior of the program. In particular,
2690 prefetches cannot trap and do not produce a value. On targets that support this
2691 intrinsic, the prefetch can provide hints to the processor cache for better
2697 <!-- _______________________________________________________________________ -->
2698 <div class="doc_subsubsection">
2699 <a name="i_pcmarker">'<tt>llvm.pcmarker</tt>' Intrinsic</a>
2702 <div class="doc_text">
2706 declare void %llvm.pcmarker( uint <id> )
2713 The '<tt>llvm.pcmarker</tt>' intrinsic is a method to export a Program Counter
2715 code to simulators and other tools. The method is target specific, but it is
2716 expected that the marker will use exported symbols to transmit the PC of the marker.
2717 The marker makes no guaranties that it will remain with any specific instruction
2718 after optimizations. It is possible that the presense of a marker will inhibit
2719 optimizations. The intended use is to be inserted after optmizations to allow
2720 correlations of simulation runs.
2726 <tt>id</tt> is a numerical id identifying the marker.
2732 This intrinsic does not modify the behavior of the program. Backends that do not
2733 support this intrinisic may ignore it.
2739 <!-- ======================================================================= -->
2740 <div class="doc_subsection">
2741 <a name="int_os">Operating System Intrinsics</a>
2744 <div class="doc_text">
2746 These intrinsics are provided by LLVM to support the implementation of
2747 operating system level code.
2752 <!-- _______________________________________________________________________ -->
2753 <div class="doc_subsubsection">
2754 <a name="i_readport">'<tt>llvm.readport</tt>' Intrinsic</a>
2757 <div class="doc_text">
2761 declare <integer type> %llvm.readport (<integer type> <address>)
2767 The '<tt>llvm.readport</tt>' intrinsic reads data from the specified hardware
2774 The argument to this intrinsic indicates the hardware I/O address from which
2775 to read the data. The address is in the hardware I/O address namespace (as
2776 opposed to being a memory location for memory mapped I/O).
2782 The '<tt>llvm.readport</tt>' intrinsic reads data from the hardware I/O port
2783 specified by <i>address</i> and returns the value. The address and return
2784 value must be integers, but the size is dependent upon the platform upon which
2785 the program is code generated. For example, on x86, the address must be an
2786 unsigned 16-bit value, and the return value must be 8, 16, or 32 bits.
2791 <!-- _______________________________________________________________________ -->
2792 <div class="doc_subsubsection">
2793 <a name="i_writeport">'<tt>llvm.writeport</tt>' Intrinsic</a>
2796 <div class="doc_text">
2800 call void (<integer type>, <integer type>)*
2801 %llvm.writeport (<integer type> <value>,
2802 <integer type> <address>)
2808 The '<tt>llvm.writeport</tt>' intrinsic writes data to the specified hardware
2815 The first argument is the value to write to the I/O port.
2819 The second argument indicates the hardware I/O address to which data should be
2820 written. The address is in the hardware I/O address namespace (as opposed to
2821 being a memory location for memory mapped I/O).
2827 The '<tt>llvm.writeport</tt>' intrinsic writes <i>value</i> to the I/O port
2828 specified by <i>address</i>. The address and value must be integers, but the
2829 size is dependent upon the platform upon which the program is code generated.
2830 For example, on x86, the address must be an unsigned 16-bit value, and the
2831 value written must be 8, 16, or 32 bits in length.
2836 <!-- _______________________________________________________________________ -->
2837 <div class="doc_subsubsection">
2838 <a name="i_readio">'<tt>llvm.readio</tt>' Intrinsic</a>
2841 <div class="doc_text">
2845 declare <result> %llvm.readio (<ty> * <pointer>)
2851 The '<tt>llvm.readio</tt>' intrinsic reads data from a memory mapped I/O
2858 The argument to this intrinsic is a pointer indicating the memory address from
2859 which to read the data. The data must be a
2860 <a href="#t_firstclass">first class</a> type.
2866 The '<tt>llvm.readio</tt>' intrinsic reads data from a memory mapped I/O
2867 location specified by <i>pointer</i> and returns the value. The argument must
2868 be a pointer, and the return value must be a
2869 <a href="#t_firstclass">first class</a> type. However, certain architectures
2870 may not support I/O on all first class types. For example, 32-bit processors
2871 may only support I/O on data types that are 32 bits or less.
2875 This intrinsic enforces an in-order memory model for llvm.readio and
2876 llvm.writeio calls on machines that use dynamic scheduling. Dynamically
2877 scheduled processors may execute loads and stores out of order, re-ordering at
2878 run time accesses to memory mapped I/O registers. Using these intrinsics
2879 ensures that accesses to memory mapped I/O registers occur in program order.
2884 <!-- _______________________________________________________________________ -->
2885 <div class="doc_subsubsection">
2886 <a name="i_writeio">'<tt>llvm.writeio</tt>' Intrinsic</a>
2889 <div class="doc_text">
2893 declare void %llvm.writeio (<ty1> <value>, <ty2> * <pointer>)
2899 The '<tt>llvm.writeio</tt>' intrinsic writes data to the specified memory
2906 The first argument is the value to write to the memory mapped I/O location.
2907 The second argument is a pointer indicating the memory address to which the
2908 data should be written.
2914 The '<tt>llvm.writeio</tt>' intrinsic writes <i>value</i> to the memory mapped
2915 I/O address specified by <i>pointer</i>. The value must be a
2916 <a href="#t_firstclass">first class</a> type. However, certain architectures
2917 may not support I/O on all first class types. For example, 32-bit processors
2918 may only support I/O on data types that are 32 bits or less.
2922 This intrinsic enforces an in-order memory model for llvm.readio and
2923 llvm.writeio calls on machines that use dynamic scheduling. Dynamically
2924 scheduled processors may execute loads and stores out of order, re-ordering at
2925 run time accesses to memory mapped I/O registers. Using these intrinsics
2926 ensures that accesses to memory mapped I/O registers occur in program order.
2931 <!-- ======================================================================= -->
2932 <div class="doc_subsection">
2933 <a name="int_libc">Standard C Library Intrinsics</a>
2936 <div class="doc_text">
2938 LLVM provides intrinsics for a few important standard C library functions.
2939 These intrinsics allow source-language front-ends to pass information about the
2940 alignment of the pointer arguments to the code generator, providing opportunity
2941 for more efficient code generation.
2946 <!-- _______________________________________________________________________ -->
2947 <div class="doc_subsubsection">
2948 <a name="i_memcpy">'<tt>llvm.memcpy</tt>' Intrinsic</a>
2951 <div class="doc_text">
2955 declare void %llvm.memcpy(sbyte* <dest>, sbyte* <src>,
2956 uint <len>, uint <align>)
2962 The '<tt>llvm.memcpy</tt>' intrinsic copies a block of memory from the source
2963 location to the destination location.
2967 Note that, unlike the standard libc function, the <tt>llvm.memcpy</tt> intrinsic
2968 does not return a value, and takes an extra alignment argument.
2974 The first argument is a pointer to the destination, the second is a pointer to
2975 the source. The third argument is an (arbitrarily sized) integer argument
2976 specifying the number of bytes to copy, and the fourth argument is the alignment
2977 of the source and destination locations.
2981 If the call to this intrinisic has an alignment value that is not 0 or 1, then
2982 the caller guarantees that the size of the copy is a multiple of the alignment
2983 and that both the source and destination pointers are aligned to that boundary.
2989 The '<tt>llvm.memcpy</tt>' intrinsic copies a block of memory from the source
2990 location to the destination location, which are not allowed to overlap. It
2991 copies "len" bytes of memory over. If the argument is known to be aligned to
2992 some boundary, this can be specified as the fourth argument, otherwise it should
2998 <!-- _______________________________________________________________________ -->
2999 <div class="doc_subsubsection">
3000 <a name="i_memmove">'<tt>llvm.memmove</tt>' Intrinsic</a>
3003 <div class="doc_text">
3007 declare void %llvm.memmove(sbyte* <dest>, sbyte* <src>,
3008 uint <len>, uint <align>)
3014 The '<tt>llvm.memmove</tt>' intrinsic moves a block of memory from the source
3015 location to the destination location. It is similar to the '<tt>llvm.memcpy</tt>'
3016 intrinsic but allows the two memory locations to overlap.
3020 Note that, unlike the standard libc function, the <tt>llvm.memmove</tt> intrinsic
3021 does not return a value, and takes an extra alignment argument.
3027 The first argument is a pointer to the destination, the second is a pointer to
3028 the source. The third argument is an (arbitrarily sized) integer argument
3029 specifying the number of bytes to copy, and the fourth argument is the alignment
3030 of the source and destination locations.
3034 If the call to this intrinisic has an alignment value that is not 0 or 1, then
3035 the caller guarantees that the size of the copy is a multiple of the alignment
3036 and that both the source and destination pointers are aligned to that boundary.
3042 The '<tt>llvm.memmove</tt>' intrinsic copies a block of memory from the source
3043 location to the destination location, which may overlap. It
3044 copies "len" bytes of memory over. If the argument is known to be aligned to
3045 some boundary, this can be specified as the fourth argument, otherwise it should
3051 <!-- _______________________________________________________________________ -->
3052 <div class="doc_subsubsection">
3053 <a name="i_memset">'<tt>llvm.memset</tt>' Intrinsic</a>
3056 <div class="doc_text">
3060 declare void %llvm.memset(sbyte* <dest>, ubyte <val>,
3061 uint <len>, uint <align>)
3067 The '<tt>llvm.memset</tt>' intrinsic fills a block of memory with a particular
3072 Note that, unlike the standard libc function, the <tt>llvm.memset</tt> intrinsic
3073 does not return a value, and takes an extra alignment argument.
3079 The first argument is a pointer to the destination to fill, the second is the
3080 byte value to fill it with, the third argument is an (arbitrarily sized) integer
3081 argument specifying the number of bytes to fill, and the fourth argument is the
3082 known alignment of destination location.
3086 If the call to this intrinisic has an alignment value that is not 0 or 1, then
3087 the caller guarantees that the size of the copy is a multiple of the alignment
3088 and that the destination pointer is aligned to that boundary.
3094 The '<tt>llvm.memset</tt>' intrinsic fills "len" bytes of memory starting at the
3095 destination location. If the argument is known to be aligned to some boundary,
3096 this can be specified as the fourth argument, otherwise it should be set to 0 or
3102 <!-- _______________________________________________________________________ -->
3103 <div class="doc_subsubsection">
3104 <a name="i_isunordered">'<tt>llvm.isunordered</tt>' Intrinsic</a>
3107 <div class="doc_text">
3111 declare bool %llvm.isunordered(<float or double> Val1, <float or double> Val2)
3117 The '<tt>llvm.isunordered</tt>' intrinsic returns true if either or both of the
3118 specified floating point values is a NAN.
3124 The arguments are floating point numbers of the same type.
3130 If either or both of the arguments is a SNAN or QNAN, it returns true, otherwise
3136 <!-- ======================================================================= -->
3137 <div class="doc_subsection">
3138 <a name="int_count">Bit Counting Intrinsics</a>
3141 <div class="doc_text">
3143 LLVM provides intrinsics for a few important bit counting operations.
3144 These allow efficient code generation for some algorithms.
3149 <!-- _______________________________________________________________________ -->
3150 <div class="doc_subsubsection">
3151 <a name="int_ctpop">'<tt>llvm.ctpop</tt>' Intrinsic</a>
3154 <div class="doc_text">
3158 declare int %llvm.ctpop(int <src>)
3165 The '<tt>llvm.ctpop</tt>' intrinsic counts the number of ones in a variable.
3171 The only argument is the value to be counted. The argument may be of any
3172 integer type. The return type must match the argument type.
3178 The '<tt>llvm.ctpop</tt>' intrinsic counts the 1's in a variable.
3182 <!-- _______________________________________________________________________ -->
3183 <div class="doc_subsubsection">
3184 <a name="int_ctlz">'<tt>llvm.ctlz</tt>' Intrinsic</a>
3187 <div class="doc_text">
3191 declare int %llvm.ctlz(int <src>)
3198 The '<tt>llvm.ctlz</tt>' intrinsic counts the number of leading zeros in a
3205 The only argument is the value to be counted. The argument may be of any
3206 integer type. The return type must match the argument type.
3212 The '<tt>llvm.ctlz</tt>' intrinsic counts the leading (most significant) zeros
3213 in a variable. If the src == 0 then the result is the size in bits of the type
3214 of src. For example, <tt>llvm.cttz(int 2) = 30</tt>.
3220 <!-- _______________________________________________________________________ -->
3221 <div class="doc_subsubsection">
3222 <a name="int_cttz">'<tt>llvm.cttz</tt>' Intrinsic</a>
3225 <div class="doc_text">
3229 declare int %llvm.cttz(int <src>)
3236 The '<tt>llvm.cttz</tt>' intrinsic counts the number of trailing zeros.
3242 The only argument is the value to be counted. The argument may be of any
3243 integer type. The return type must match the argument type.
3249 The '<tt>llvm.cttz</tt>' intrinsic counts the trailing (least significant) zeros
3250 in a variable. If the src == 0 then the result is the size in bits of the type
3251 of src. For example, <tt>llvm.cttz(2) = 1</tt>.
3255 <!-- ======================================================================= -->
3256 <div class="doc_subsection">
3257 <a name="int_debugger">Debugger Intrinsics</a>
3260 <div class="doc_text">
3262 The LLVM debugger intrinsics (which all start with <tt>llvm.dbg.</tt> prefix),
3263 are described in the <a
3264 href="SourceLevelDebugging.html#format_common_intrinsics">LLVM Source Level
3265 Debugging</a> document.
3270 <!-- *********************************************************************** -->
3273 <a href="http://jigsaw.w3.org/css-validator/check/referer"><img
3274 src="http://jigsaw.w3.org/css-validator/images/vcss" alt="Valid CSS!"></a>
3275 <a href="http://validator.w3.org/check/referer"><img
3276 src="http://www.w3.org/Icons/valid-html401" alt="Valid HTML 4.01!" /></a>
3278 <a href="mailto:sabre@nondot.org">Chris Lattner</a><br>
3279 <a href="http://llvm.cs.uiuc.edu">The LLVM Compiler Infrastructure</a><br>
3280 Last modified: $Date$