Blame - docs/LangRef.html - fp2-dev/platform/external/llvm

blob: 687c87f45f3a290b13b16a429b1ea0c9f680887f [file] [log] [blame]

Dan Gohman	f17a25c	2007-07-18 16:29:46 +0000	[diff] [blame]	1	<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01//EN"
				2	"http://www.w3.org/TR/html4/strict.dtd">
				3	<html>
				4	<head>
				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">
				11	</head>
				12
				13	<body>
				14
				15	<div class="doc_title"> LLVM Language Reference Manual </div>
				16	<ol>
				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>
				21	<ol>
				22	<li><a href="#modulestructure">Module Structure</a></li>
				23	<li><a href="#linkage">Linkage Types</a></li>
				24	<li><a href="#callingconv">Calling Conventions</a></li>
				25	<li><a href="#globalvars">Global Variables</a></li>
				26	<li><a href="#functionstructure">Functions</a></li>
				27	<li><a href="#aliasstructure">Aliases</a>
				28	<li><a href="#paramattrs">Parameter Attributes</a></li>
				29	<li><a href="#moduleasm">Module-Level Inline Assembly</a></li>
				30	<li><a href="#datalayout">Data Layout</a></li>
				31	</ol>
				32	</li>
				33	<li><a href="#typesystem">Type System</a>
				34	<ol>
				35	<li><a href="#t_primitive">Primitive Types</a>
				36	<ol>
				37	<li><a href="#t_classifications">Type Classifications</a></li>
				38	</ol>
				39	</li>
				40	<li><a href="#t_derived">Derived Types</a>
				41	<ol>
				42	<li><a href="#t_array">Array Type</a></li>
				43	<li><a href="#t_function">Function Type</a></li>
				44	<li><a href="#t_pointer">Pointer Type</a></li>
				45	<li><a href="#t_struct">Structure Type</a></li>
				46	<li><a href="#t_pstruct">Packed Structure Type</a></li>
				47	<li><a href="#t_vector">Vector Type</a></li>
				48	<li><a href="#t_opaque">Opaque Type</a></li>
				49	</ol>
				50	</li>
				51	</ol>
				52	</li>
				53	<li><a href="#constants">Constants</a>
				54	<ol>
				55	<li><a href="#simpleconstants">Simple Constants</a>
				56	<li><a href="#aggregateconstants">Aggregate Constants</a>
				57	<li><a href="#globalconstants">Global Variable and Function Addresses</a>
				58	<li><a href="#undefvalues">Undefined Values</a>
				59	<li><a href="#constantexprs">Constant Expressions</a>
				60	</ol>
				61	</li>
				62	<li><a href="#othervalues">Other Values</a>
				63	<ol>
				64	<li><a href="#inlineasm">Inline Assembler Expressions</a>
				65	</ol>
				66	</li>
				67	<li><a href="#instref">Instruction Reference</a>
				68	<ol>
				69	<li><a href="#terminators">Terminator Instructions</a>
				70	<ol>
				71	<li><a href="#i_ret">'<tt>ret</tt>' Instruction</a></li>
				72	<li><a href="#i_br">'<tt>br</tt>' Instruction</a></li>
				73	<li><a href="#i_switch">'<tt>switch</tt>' Instruction</a></li>
				74	<li><a href="#i_invoke">'<tt>invoke</tt>' Instruction</a></li>
				75	<li><a href="#i_unwind">'<tt>unwind</tt>' Instruction</a></li>
				76	<li><a href="#i_unreachable">'<tt>unreachable</tt>' Instruction</a></li>
				77	</ol>
				78	</li>
				79	<li><a href="#binaryops">Binary Operations</a>
				80	<ol>
				81	<li><a href="#i_add">'<tt>add</tt>' Instruction</a></li>
				82	<li><a href="#i_sub">'<tt>sub</tt>' Instruction</a></li>
				83	<li><a href="#i_mul">'<tt>mul</tt>' Instruction</a></li>
				84	<li><a href="#i_udiv">'<tt>udiv</tt>' Instruction</a></li>
				85	<li><a href="#i_sdiv">'<tt>sdiv</tt>' Instruction</a></li>
				86	<li><a href="#i_fdiv">'<tt>fdiv</tt>' Instruction</a></li>
				87	<li><a href="#i_urem">'<tt>urem</tt>' Instruction</a></li>
				88	<li><a href="#i_srem">'<tt>srem</tt>' Instruction</a></li>
				89	<li><a href="#i_frem">'<tt>frem</tt>' Instruction</a></li>
				90	</ol>
				91	</li>
				92	<li><a href="#bitwiseops">Bitwise Binary Operations</a>
				93	<ol>
				94	<li><a href="#i_shl">'<tt>shl</tt>' Instruction</a></li>
				95	<li><a href="#i_lshr">'<tt>lshr</tt>' Instruction</a></li>
				96	<li><a href="#i_ashr">'<tt>ashr</tt>' Instruction</a></li>
				97	<li><a href="#i_and">'<tt>and</tt>' Instruction</a></li>
				98	<li><a href="#i_or">'<tt>or</tt>' Instruction</a></li>
				99	<li><a href="#i_xor">'<tt>xor</tt>' Instruction</a></li>
				100	</ol>
				101	</li>
				102	<li><a href="#vectorops">Vector Operations</a>
				103	<ol>
				104	<li><a href="#i_extractelement">'<tt>extractelement</tt>' Instruction</a></li>
				105	<li><a href="#i_insertelement">'<tt>insertelement</tt>' Instruction</a></li>
				106	<li><a href="#i_shufflevector">'<tt>shufflevector</tt>' Instruction</a></li>
				107	</ol>
				108	</li>
				109	<li><a href="#memoryops">Memory Access and Addressing Operations</a>
				110	<ol>
				111	<li><a href="#i_malloc">'<tt>malloc</tt>' Instruction</a></li>
				112	<li><a href="#i_free">'<tt>free</tt>' Instruction</a></li>
				113	<li><a href="#i_alloca">'<tt>alloca</tt>' Instruction</a></li>
				114	<li><a href="#i_load">'<tt>load</tt>' Instruction</a></li>
				115	<li><a href="#i_store">'<tt>store</tt>' Instruction</a></li>
				116	<li><a href="#i_getelementptr">'<tt>getelementptr</tt>' Instruction</a></li>
				117	</ol>
				118	</li>
				119	<li><a href="#convertops">Conversion Operations</a>
				120	<ol>
				121	<li><a href="#i_trunc">'<tt>trunc .. to</tt>' Instruction</a></li>
				122	<li><a href="#i_zext">'<tt>zext .. to</tt>' Instruction</a></li>
				123	<li><a href="#i_sext">'<tt>sext .. to</tt>' Instruction</a></li>
				124	<li><a href="#i_fptrunc">'<tt>fptrunc .. to</tt>' Instruction</a></li>
				125	<li><a href="#i_fpext">'<tt>fpext .. to</tt>' Instruction</a></li>
				126	<li><a href="#i_fptoui">'<tt>fptoui .. to</tt>' Instruction</a></li>
				127	<li><a href="#i_fptosi">'<tt>fptosi .. to</tt>' Instruction</a></li>
				128	<li><a href="#i_uitofp">'<tt>uitofp .. to</tt>' Instruction</a></li>
				129	<li><a href="#i_sitofp">'<tt>sitofp .. to</tt>' Instruction</a></li>
				130	<li><a href="#i_ptrtoint">'<tt>ptrtoint .. to</tt>' Instruction</a></li>
				131	<li><a href="#i_inttoptr">'<tt>inttoptr .. to</tt>' Instruction</a></li>
				132	<li><a href="#i_bitcast">'<tt>bitcast .. to</tt>' Instruction</a></li>
				133	</ol>
				134	<li><a href="#otherops">Other Operations</a>
				135	<ol>
				136	<li><a href="#i_icmp">'<tt>icmp</tt>' Instruction</a></li>
				137	<li><a href="#i_fcmp">'<tt>fcmp</tt>' Instruction</a></li>
				138	<li><a href="#i_phi">'<tt>phi</tt>' Instruction</a></li>
				139	<li><a href="#i_select">'<tt>select</tt>' Instruction</a></li>
				140	<li><a href="#i_call">'<tt>call</tt>' Instruction</a></li>
				141	<li><a href="#i_va_arg">'<tt>va_arg</tt>' Instruction</a></li>
				142	</ol>
				143	</li>
				144	</ol>
				145	</li>
				146	<li><a href="#intrinsics">Intrinsic Functions</a>
				147	<ol>
				148	<li><a href="#int_varargs">Variable Argument Handling Intrinsics</a>
				149	<ol>
				150	<li><a href="#int_va_start">'<tt>llvm.va_start</tt>' Intrinsic</a></li>
				151	<li><a href="#int_va_end">'<tt>llvm.va_end</tt>' Intrinsic</a></li>
				152	<li><a href="#int_va_copy">'<tt>llvm.va_copy</tt>' Intrinsic</a></li>
				153	</ol>
				154	</li>
				155	<li><a href="#int_gc">Accurate Garbage Collection Intrinsics</a>
				156	<ol>
				157	<li><a href="#int_gcroot">'<tt>llvm.gcroot</tt>' Intrinsic</a></li>
				158	<li><a href="#int_gcread">'<tt>llvm.gcread</tt>' Intrinsic</a></li>
				159	<li><a href="#int_gcwrite">'<tt>llvm.gcwrite</tt>' Intrinsic</a></li>
				160	</ol>
				161	</li>
				162	<li><a href="#int_codegen">Code Generator Intrinsics</a>
				163	<ol>
				164	<li><a href="#int_returnaddress">'<tt>llvm.returnaddress</tt>' Intrinsic</a></li>
				165	<li><a href="#int_frameaddress">'<tt>llvm.frameaddress</tt>' Intrinsic</a></li>
				166	<li><a href="#int_stacksave">'<tt>llvm.stacksave</tt>' Intrinsic</a></li>
				167	<li><a href="#int_stackrestore">'<tt>llvm.stackrestore</tt>' Intrinsic</a></li>
				168	<li><a href="#int_prefetch">'<tt>llvm.prefetch</tt>' Intrinsic</a></li>
				169	<li><a href="#int_pcmarker">'<tt>llvm.pcmarker</tt>' Intrinsic</a></li>
				170	<li><a href="#int_readcyclecounter"><tt>llvm.readcyclecounter</tt>' Intrinsic</a></li>
				171	</ol>
				172	</li>
				173	<li><a href="#int_libc">Standard C Library Intrinsics</a>
				174	<ol>
				175	<li><a href="#int_memcpy">'<tt>llvm.memcpy.*</tt>' Intrinsic</a></li>
				176	<li><a href="#int_memmove">'<tt>llvm.memmove.*</tt>' Intrinsic</a></li>
				177	<li><a href="#int_memset">'<tt>llvm.memset.*</tt>' Intrinsic</a></li>
				178	<li><a href="#int_sqrt">'<tt>llvm.sqrt.*</tt>' Intrinsic</a></li>
				179	<li><a href="#int_powi">'<tt>llvm.powi.*</tt>' Intrinsic</a></li>
				180	</ol>
				181	</li>
				182	<li><a href="#int_manip">Bit Manipulation Intrinsics</a>
				183	<ol>
				184	<li><a href="#int_bswap">'<tt>llvm.bswap.*</tt>' Intrinsics</a></li>
				185	<li><a href="#int_ctpop">'<tt>llvm.ctpop.*</tt>' Intrinsic </a></li>
				186	<li><a href="#int_ctlz">'<tt>llvm.ctlz.*</tt>' Intrinsic </a></li>
				187	<li><a href="#int_cttz">'<tt>llvm.cttz.*</tt>' Intrinsic </a></li>
				188	<li><a href="#int_part_select">'<tt>llvm.part.select.*</tt>' Intrinsic </a></li>
				189	<li><a href="#int_part_set">'<tt>llvm.part.set.*</tt>' Intrinsic </a></li>
				190	</ol>
				191	</li>
				192	<li><a href="#int_debugger">Debugger intrinsics</a></li>
				193	<li><a href="#int_eh">Exception Handling intrinsics</a></li>
Chandler Carruth	c0470a8	2007-07-20 19:34:37 +0000	[diff] [blame]	194	<li><a href="#int_atomics">Atomic Operations and Synchronization Intrinsics</a>
				195	<ol>
				196	<li><a href="#int_lcs">'<tt>llvm.atomic.lcs.*</tt>' Intrinsic</a></li>
				197	<li><a href="#int_ls">'<tt>llvm.atomic.ls.*</tt>' Intrinsic</a></li>
				198	<li><a href="#int_las">'<tt>llvm.atomic.las.*</tt>' Intrinsic</a></li>
				199	<li><a href="#int_lss">'<tt>llvm.atomic.lss.*</tt>' Intrinsic</a></li>
				200	<li><a href="#int_memory_barrier">'<tt>llvm.memory.barrier</tt>' Intrinsic</a></li>
				201	</ol>
				202	</li>
Duncan Sands	38947cd	2007-07-27 12:58:54 +0000	[diff] [blame]	203	<li><a href="#int_trampoline">Trampoline Intrinsics</a>
				204	<ol>
				205	<li><a href="#int_it">'<tt>llvm.init.trampoline</tt>' Intrinsic</a></li>
				206	<li><a href="#int_at">'<tt>llvm.adjust.trampoline</tt>' Intrinsic</a></li>
				207	</ol>
				208	</li>
Reid Spencer	b043f67	2007-07-20 19:59:11 +0000	[diff] [blame]	209	<li><a href="#int_general">General intrinsics</a>
Dan Gohman	f17a25c	2007-07-18 16:29:46 +0000	[diff] [blame]	210	<ol>
Reid Spencer	b043f67	2007-07-20 19:59:11 +0000	[diff] [blame]	211	<li><a href="#int_var_annotation">
				212	<tt>llvm.var.annotation</tt>' Intrinsic</a></li>
				213	</ol>
Dan Gohman	f17a25c	2007-07-18 16:29:46 +0000	[diff] [blame]	214	</li>
				215	</ol>
				216	</li>
				217	</ol>
				218
				219	<div class="doc_author">
				220	<p>Written by <a href="mailto:sabre@nondot.org">Chris Lattner</a>
				221	and <a href="mailto:vadve@cs.uiuc.edu">Vikram Adve</a></p>
				222	</div>
				223
				224	<!-- *********************************************************************** -->
				225	<div class="doc_section"> <a name="abstract">Abstract </a></div>
				226	<!-- *********************************************************************** -->
				227
				228	<div class="doc_text">
				229	<p>This document is a reference manual for the LLVM assembly language.
				230	LLVM is an SSA based representation that provides type safety,
				231	low-level operations, flexibility, and the capability of representing
				232	'all' high-level languages cleanly. It is the common code
				233	representation used throughout all phases of the LLVM compilation
				234	strategy.</p>
				235	</div>
				236
				237	<!-- *********************************************************************** -->
				238	<div class="doc_section"> <a name="introduction">Introduction</a> </div>
				239	<!-- *********************************************************************** -->
				240
				241	<div class="doc_text">
				242
				243	<p>The LLVM code representation is designed to be used in three
				244	different forms: as an in-memory compiler IR, as an on-disk bitcode
				245	representation (suitable for fast loading by a Just-In-Time compiler),
				246	and as a human readable assembly language representation. This allows
				247	LLVM to provide a powerful intermediate representation for efficient
				248	compiler transformations and analysis, while providing a natural means
				249	to debug and visualize the transformations. The three different forms
				250	of LLVM are all equivalent. This document describes the human readable
				251	representation and notation.</p>
				252
				253	<p>The LLVM representation aims to be light-weight and low-level
				254	while being expressive, typed, and extensible at the same time. It
				255	aims to be a "universal IR" of sorts, by being at a low enough level
				256	that high-level ideas may be cleanly mapped to it (similar to how
				257	microprocessors are "universal IR's", allowing many source languages to
				258	be mapped to them). By providing type information, LLVM can be used as
				259	the target of optimizations: for example, through pointer analysis, it
				260	can be proven that a C automatic variable is never accessed outside of
				261	the current function... allowing it to be promoted to a simple SSA
				262	value instead of a memory location.</p>
				263
				264	</div>
				265
				266	<!-- _______________________________________________________________________ -->
				267	<div class="doc_subsubsection"> <a name="wellformed">Well-Formedness</a> </div>
				268
				269	<div class="doc_text">
				270
				271	<p>It is important to note that this document describes 'well formed'
				272	LLVM assembly language. There is a difference between what the parser
				273	accepts and what is considered 'well formed'. For example, the
				274	following instruction is syntactically okay, but not well formed:</p>
				275
				276	<div class="doc_code">
				277	<pre>
				278	%x = <a href="#i_add">add</a> i32 1, %x
				279	</pre>
				280	</div>
				281
				282	<p>...because the definition of <tt>%x</tt> does not dominate all of
				283	its uses. The LLVM infrastructure provides a verification pass that may
				284	be used to verify that an LLVM module is well formed. This pass is
				285	automatically run by the parser after parsing input assembly and by
				286	the optimizer before it outputs bitcode. The violations pointed out
				287	by the verifier pass indicate bugs in transformation passes or input to
				288	the parser.</p>
				289	</div>
				290
Reid Spencer	b043f67	2007-07-20 19:59:11 +0000	[diff] [blame]	291	<!-- Describe the typesetting conventions here. -->
Dan Gohman	f17a25c	2007-07-18 16:29:46 +0000	[diff] [blame]	292
				293	<!-- *********************************************************************** -->
				294	<div class="doc_section"> <a name="identifiers">Identifiers</a> </div>
				295	<!-- *********************************************************************** -->
				296
				297	<div class="doc_text">
				298
Reid Spencer	c8245b0	2007-08-07 14:34:28 +0000	[diff] [blame]	299	<p>LLVM identifiers come in two basic types: global and local. Global
				300	identifiers (functions, global variables) begin with the @ character. Local
				301	identifiers (register names, types) begin with the % character. Additionally,
				302	there are three different formats for identifiers, for different purposes:
Dan Gohman	f17a25c	2007-07-18 16:29:46 +0000	[diff] [blame]	303
				304	<ol>
Reid Spencer	c8245b0	2007-08-07 14:34:28 +0000	[diff] [blame]	305	<li>Named values are represented as a string of characters with their prefix.
				306	For example, %foo, @DivisionByZero, %a.really.long.identifier. The actual
				307	regular expression used is '<tt>[%@][a-zA-Z$._][a-zA-Z$._0-9]*</tt>'.
Dan Gohman	f17a25c	2007-07-18 16:29:46 +0000	[diff] [blame]	308	Identifiers which require other characters in their names can be surrounded
Reid Spencer	c8245b0	2007-08-07 14:34:28 +0000	[diff] [blame]	309	with quotes. In this way, anything except a <tt>"</tt> character can
				310	be used in a named value.</li>
Dan Gohman	f17a25c	2007-07-18 16:29:46 +0000	[diff] [blame]	311
Reid Spencer	c8245b0	2007-08-07 14:34:28 +0000	[diff] [blame]	312	<li>Unnamed values are represented as an unsigned numeric value with their
				313	prefix. For example, %12, @2, %44.</li>
Dan Gohman	f17a25c	2007-07-18 16:29:46 +0000	[diff] [blame]	314
				315	<li>Constants, which are described in a <a href="#constants">section about
				316	constants</a>, below.</li>
				317	</ol>
				318
Reid Spencer	c8245b0	2007-08-07 14:34:28 +0000	[diff] [blame]	319	<p>LLVM requires that values start with a prefix for two reasons: Compilers
Dan Gohman	f17a25c	2007-07-18 16:29:46 +0000	[diff] [blame]	320	don't need to worry about name clashes with reserved words, and the set of
				321	reserved words may be expanded in the future without penalty. Additionally,
				322	unnamed identifiers allow a compiler to quickly come up with a temporary
				323	variable without having to avoid symbol table conflicts.</p>
				324
				325	<p>Reserved words in LLVM are very similar to reserved words in other
				326	languages. There are keywords for different opcodes
				327	('<tt><a href="#i_add">add</a></tt>',
				328	'<tt><a href="#i_bitcast">bitcast</a></tt>',
				329	'<tt><a href="#i_ret">ret</a></tt>', etc...), for primitive type names ('<tt><a
				330	href="#t_void">void</a></tt>', '<tt><a href="#t_primitive">i32</a></tt>', etc...),
				331	and others. These reserved words cannot conflict with variable names, because
Reid Spencer	c8245b0	2007-08-07 14:34:28 +0000	[diff] [blame]	332	none of them start with a prefix character ('%' or '@').</p>
Dan Gohman	f17a25c	2007-07-18 16:29:46 +0000	[diff] [blame]	333
				334	<p>Here is an example of LLVM code to multiply the integer variable
				335	'<tt>%X</tt>' by 8:</p>
				336
				337	<p>The easy way:</p>
				338
				339	<div class="doc_code">
				340	<pre>
				341	%result = <a href="#i_mul">mul</a> i32 %X, 8
				342	</pre>
				343	</div>
				344
				345	<p>After strength reduction:</p>
				346
				347	<div class="doc_code">
				348	<pre>
				349	%result = <a href="#i_shl">shl</a> i32 %X, i8 3
				350	</pre>
				351	</div>
				352
				353	<p>And the hard way:</p>
				354
				355	<div class="doc_code">
				356	<pre>
				357	<a href="#i_add">add</a> i32 %X, %X <i>; yields {i32}:%0</i>
				358	<a href="#i_add">add</a> i32 %0, %0 <i>; yields {i32}:%1</i>
				359	%result = <a href="#i_add">add</a> i32 %1, %1
				360	</pre>
				361	</div>
				362
				363	<p>This last way of multiplying <tt>%X</tt> by 8 illustrates several
				364	important lexical features of LLVM:</p>
				365
				366	<ol>
				367
				368	<li>Comments are delimited with a '<tt>;</tt>' and go until the end of
				369	line.</li>
				370
				371	<li>Unnamed temporaries are created when the result of a computation is not
				372	assigned to a named value.</li>
				373
				374	<li>Unnamed temporaries are numbered sequentially</li>
				375
				376	</ol>
				377
				378	<p>...and it also shows a convention that we follow in this document. When
				379	demonstrating instructions, we will follow an instruction with a comment that
				380	defines the type and name of value produced. Comments are shown in italic
				381	text.</p>
				382
				383	</div>
				384
				385	<!-- *********************************************************************** -->
				386	<div class="doc_section"> <a name="highlevel">High Level Structure</a> </div>
				387	<!-- *********************************************************************** -->
				388
				389	<!-- ======================================================================= -->
				390	<div class="doc_subsection"> <a name="modulestructure">Module Structure</a>
				391	</div>
				392
				393	<div class="doc_text">
				394
				395	<p>LLVM programs are composed of "Module"s, each of which is a
				396	translation unit of the input programs. Each module consists of
				397	functions, global variables, and symbol table entries. Modules may be
				398	combined together with the LLVM linker, which merges function (and
				399	global variable) definitions, resolves forward declarations, and merges
				400	symbol table entries. Here is an example of the "hello world" module:</p>
				401
				402	<div class="doc_code">
				403	<pre><i>; Declare the string constant as a global constant...</i>
				404	<a href="#identifiers">@.LC0</a> = <a href="#linkage_internal">internal</a> <a
				405	href="#globalvars">constant</a> <a href="#t_array">[13 x i8]</a> c"hello world\0A\00" <i>; [13 x i8]*</i>
				406
				407	<i>; External declaration of the puts function</i>
				408	<a href="#functionstructure">declare</a> i32 @puts(i8 ) <i>; i32(i8 )* </i>
				409
				410	<i>; Definition of main function</i>
				411	define i32 @main() { <i>; i32()* </i>
				412	<i>; Convert [13x i8 ]* to i8 *...</i>
				413	%cast210 = <a
				414	href="#i_getelementptr">getelementptr</a> [13 x i8 ]* @.LC0, i64 0, i64 0 <i>; i8 *</i>
				415
				416	<i>; Call puts function to write out the string to stdout...</i>
				417	<a
				418	href="#i_call">call</a> i32 @puts(i8 * %cast210) <i>; i32</i>
				419	<a
				420	href="#i_ret">ret</a> i32 0<br>}<br>
				421	</pre>
				422	</div>
				423
				424	<p>This example is made up of a <a href="#globalvars">global variable</a>
				425	named "<tt>.LC0</tt>", an external declaration of the "<tt>puts</tt>"
				426	function, and a <a href="#functionstructure">function definition</a>
				427	for "<tt>main</tt>".</p>
				428
				429	<p>In general, a module is made up of a list of global values,
				430	where both functions and global variables are global values. Global values are
				431	represented by a pointer to a memory location (in this case, a pointer to an
				432	array of char, and a pointer to a function), and have one of the following <a
				433	href="#linkage">linkage types</a>.</p>
				434
				435	</div>
				436
				437	<!-- ======================================================================= -->
				438	<div class="doc_subsection">
				439	<a name="linkage">Linkage Types</a>
				440	</div>
				441
				442	<div class="doc_text">
				443
				444	<p>
				445	All Global Variables and Functions have one of the following types of linkage:
				446	</p>
				447
				448	<dl>
				449
				450	<dt><tt><b><a name="linkage_internal">internal</a></b></tt> </dt>
				451
				452	<dd>Global values with internal linkage are only directly accessible by
				453	objects in the current module. In particular, linking code into a module with
				454	an internal global value may cause the internal to be renamed as necessary to
				455	avoid collisions. Because the symbol is internal to the module, all
				456	references can be updated. This corresponds to the notion of the
				457	'<tt>static</tt>' keyword in C.
				458	</dd>
				459
				460	<dt><tt><b><a name="linkage_linkonce">linkonce</a></b></tt>: </dt>
				461
				462	<dd>Globals with "<tt>linkonce</tt>" linkage are merged with other globals of
				463	the same name when linkage occurs. This is typically used to implement
				464	inline functions, templates, or other code which must be generated in each
				465	translation unit that uses it. Unreferenced <tt>linkonce</tt> globals are
				466	allowed to be discarded.
				467	</dd>
				468
				469	<dt><tt><b><a name="linkage_weak">weak</a></b></tt>: </dt>
				470
				471	<dd>"<tt>weak</tt>" linkage is exactly the same as <tt>linkonce</tt> linkage,
				472	except that unreferenced <tt>weak</tt> globals may not be discarded. This is
				473	used for globals that may be emitted in multiple translation units, but that
				474	are not guaranteed to be emitted into every translation unit that uses them.
				475	One example of this are common globals in C, such as "<tt>int X;</tt>" at
				476	global scope.
				477	</dd>
				478
				479	<dt><tt><b><a name="linkage_appending">appending</a></b></tt>: </dt>
				480
				481	<dd>"<tt>appending</tt>" linkage may only be applied to global variables of
				482	pointer to array type. When two global variables with appending linkage are
				483	linked together, the two global arrays are appended together. This is the
				484	LLVM, typesafe, equivalent of having the system linker append together
				485	"sections" with identical names when .o files are linked.
				486	</dd>
				487
				488	<dt><tt><b><a name="linkage_externweak">extern_weak</a></b></tt>: </dt>
				489	<dd>The semantics of this linkage follow the ELF model: the symbol is weak
				490	until linked, if not linked, the symbol becomes null instead of being an
				491	undefined reference.
				492	</dd>
				493
				494	<dt><tt><b><a name="linkage_external">externally visible</a></b></tt>:</dt>
				495
				496	<dd>If none of the above identifiers are used, the global is externally
				497	visible, meaning that it participates in linkage and can be used to resolve
				498	external symbol references.
				499	</dd>
				500	</dl>
				501
				502	<p>
				503	The next two types of linkage are targeted for Microsoft Windows platform
				504	only. They are designed to support importing (exporting) symbols from (to)
				505	DLLs.
				506	</p>
				507
				508	<dl>
				509	<dt><tt><b><a name="linkage_dllimport">dllimport</a></b></tt>: </dt>
				510
				511	<dd>"<tt>dllimport</tt>" linkage causes the compiler to reference a function
				512	or variable via a global pointer to a pointer that is set up by the DLL
				513	exporting the symbol. On Microsoft Windows targets, the pointer name is
				514	formed by combining <code>_imp__</code> and the function or variable name.
				515	</dd>
				516
				517	<dt><tt><b><a name="linkage_dllexport">dllexport</a></b></tt>: </dt>
				518
				519	<dd>"<tt>dllexport</tt>" linkage causes the compiler to provide a global
				520	pointer to a pointer in a DLL, so that it can be referenced with the
				521	<tt>dllimport</tt> attribute. On Microsoft Windows targets, the pointer
				522	name is formed by combining <code>_imp__</code> and the function or variable
				523	name.
				524	</dd>
				525
				526	</dl>
				527
				528	<p><a name="linkage_external"></a>For example, since the "<tt>.LC0</tt>"
				529	variable is defined to be internal, if another module defined a "<tt>.LC0</tt>"
				530	variable and was linked with this one, one of the two would be renamed,
				531	preventing a collision. Since "<tt>main</tt>" and "<tt>puts</tt>" are
				532	external (i.e., lacking any linkage declarations), they are accessible
				533	outside of the current module.</p>
				534	<p>It is illegal for a function <i>declaration</i>
				535	to have any linkage type other than "externally visible", <tt>dllimport</tt>,
				536	or <tt>extern_weak</tt>.</p>
				537	<p>Aliases can have only <tt>external</tt>, <tt>internal</tt> and <tt>weak</tt>
				538	linkages.
				539	</div>
				540
				541	<!-- ======================================================================= -->
				542	<div class="doc_subsection">
				543	<a name="callingconv">Calling Conventions</a>
				544	</div>
				545
				546	<div class="doc_text">
				547
				548	<p>LLVM <a href="#functionstructure">functions</a>, <a href="#i_call">calls</a>
				549	and <a href="#i_invoke">invokes</a> can all have an optional calling convention
				550	specified for the call. The calling convention of any pair of dynamic
				551	caller/callee must match, or the behavior of the program is undefined. The
				552	following calling conventions are supported by LLVM, and more may be added in
				553	the future:</p>
				554
				555	<dl>
				556	<dt><b>"<tt>ccc</tt>" - The C calling convention</b>:</dt>
				557
				558	<dd>This calling convention (the default if no other calling convention is
				559	specified) matches the target C calling conventions. This calling convention
				560	supports varargs function calls and tolerates some mismatch in the declared
				561	prototype and implemented declaration of the function (as does normal C).
				562	</dd>
				563
				564	<dt><b>"<tt>fastcc</tt>" - The fast calling convention</b>:</dt>
				565
				566	<dd>This calling convention attempts to make calls as fast as possible
				567	(e.g. by passing things in registers). This calling convention allows the
				568	target to use whatever tricks it wants to produce fast code for the target,
				569	without having to conform to an externally specified ABI. Implementations of
				570	this convention should allow arbitrary tail call optimization to be supported.
				571	This calling convention does not support varargs and requires the prototype of
				572	all callees to exactly match the prototype of the function definition.
				573	</dd>
				574
				575	<dt><b>"<tt>coldcc</tt>" - The cold calling convention</b>:</dt>
				576
				577	<dd>This calling convention attempts to make code in the caller as efficient
				578	as possible under the assumption that the call is not commonly executed. As
				579	such, these calls often preserve all registers so that the call does not break
				580	any live ranges in the caller side. This calling convention does not support
				581	varargs and requires the prototype of all callees to exactly match the
				582	prototype of the function definition.
				583	</dd>
				584
				585	<dt><b>"<tt>cc <<em>n</em>></tt>" - Numbered convention</b>:</dt>
				586
				587	<dd>Any calling convention may be specified by number, allowing
				588	target-specific calling conventions to be used. Target specific calling
				589	conventions start at 64.
				590	</dd>
				591	</dl>
				592
				593	<p>More calling conventions can be added/defined on an as-needed basis, to
				594	support pascal conventions or any other well-known target-independent
				595	convention.</p>
				596
				597	</div>
				598
				599	<!-- ======================================================================= -->
				600	<div class="doc_subsection">
				601	<a name="visibility">Visibility Styles</a>
				602	</div>
				603
				604	<div class="doc_text">
				605
				606	<p>
				607	All Global Variables and Functions have one of the following visibility styles:
				608	</p>
				609
				610	<dl>
				611	<dt><b>"<tt>default</tt>" - Default style</b>:</dt>
				612
				613	<dd>On ELF, default visibility means that the declaration is visible to other
				614	modules and, in shared libraries, means that the declared entity may be
				615	overridden. On Darwin, default visibility means that the declaration is
				616	visible to other modules. Default visibility corresponds to "external
				617	linkage" in the language.
				618	</dd>
				619
				620	<dt><b>"<tt>hidden</tt>" - Hidden style</b>:</dt>
				621
				622	<dd>Two declarations of an object with hidden visibility refer to the same
				623	object if they are in the same shared object. Usually, hidden visibility
				624	indicates that the symbol will not be placed into the dynamic symbol table,
				625	so no other module (executable or shared library) can reference it
				626	directly.
				627	</dd>
				628
				629	<dt><b>"<tt>protected</tt>" - Protected style</b>:</dt>
				630
				631	<dd>On ELF, protected visibility indicates that the symbol will be placed in
				632	the dynamic symbol table, but that references within the defining module will
				633	bind to the local symbol. That is, the symbol cannot be overridden by another
				634	module.
				635	</dd>
				636	</dl>
				637
				638	</div>
				639
				640	<!-- ======================================================================= -->
				641	<div class="doc_subsection">
				642	<a name="globalvars">Global Variables</a>
				643	</div>
				644
				645	<div class="doc_text">
				646
				647	<p>Global variables define regions of memory allocated at compilation time
				648	instead of run-time. Global variables may optionally be initialized, may have
				649	an explicit section to be placed in, and may have an optional explicit alignment
				650	specified. A variable may be defined as "thread_local", which means that it
				651	will not be shared by threads (each thread will have a separated copy of the
				652	variable). A variable may be defined as a global "constant," which indicates
				653	that the contents of the variable will <b>never</b> be modified (enabling better
				654	optimization, allowing the global data to be placed in the read-only section of
				655	an executable, etc). Note that variables that need runtime initialization
				656	cannot be marked "constant" as there is a store to the variable.</p>
				657
				658	<p>
				659	LLVM explicitly allows <em>declarations</em> of global variables to be marked
				660	constant, even if the final definition of the global is not. This capability
				661	can be used to enable slightly better optimization of the program, but requires
				662	the language definition to guarantee that optimizations based on the
				663	'constantness' are valid for the translation units that do not include the
				664	definition.
				665	</p>
				666
				667	<p>As SSA values, global variables define pointer values that are in
				668	scope (i.e. they dominate) all basic blocks in the program. Global
				669	variables always define a pointer to their "content" type because they
				670	describe a region of memory, and all memory objects in LLVM are
				671	accessed through pointers.</p>
				672
				673	<p>LLVM allows an explicit section to be specified for globals. If the target
				674	supports it, it will emit globals to the section specified.</p>
				675
				676	<p>An explicit alignment may be specified for a global. If not present, or if
				677	the alignment is set to zero, the alignment of the global is set by the target
				678	to whatever it feels convenient. If an explicit alignment is specified, the
				679	global is forced to have at least that much alignment. All alignments must be
				680	a power of 2.</p>
				681
				682	<p>For example, the following defines a global with an initializer, section,
				683	and alignment:</p>
				684
				685	<div class="doc_code">
				686	<pre>
				687	@G = constant float 1.0, section "foo", align 4
				688	</pre>
				689	</div>
				690
				691	</div>
				692
				693
				694	<!-- ======================================================================= -->
				695	<div class="doc_subsection">
				696	<a name="functionstructure">Functions</a>
				697	</div>
				698
				699	<div class="doc_text">
				700
				701	<p>LLVM function definitions consist of the "<tt>define</tt>" keyord,
				702	an optional <a href="#linkage">linkage type</a>, an optional
				703	<a href="#visibility">visibility style</a>, an optional
				704	<a href="#callingconv">calling convention</a>, a return type, an optional
				705	<a href="#paramattrs">parameter attribute</a> for the return type, a function
				706	name, a (possibly empty) argument list (each with optional
				707	<a href="#paramattrs">parameter attributes</a>), an optional section, an
				708	optional alignment, an opening curly brace, a list of basic blocks, and a
				709	closing curly brace.
				710
				711	LLVM function declarations consist of the "<tt>declare</tt>" keyword, an
				712	optional <a href="#linkage">linkage type</a>, an optional
				713	<a href="#visibility">visibility style</a>, an optional
				714	<a href="#callingconv">calling convention</a>, a return type, an optional
				715	<a href="#paramattrs">parameter attribute</a> for the return type, a function
				716	name, a possibly empty list of arguments, and an optional alignment.</p>
				717
				718	<p>A function definition contains a list of basic blocks, forming the CFG for
				719	the function. Each basic block may optionally start with a label (giving the
				720	basic block a symbol table entry), contains a list of instructions, and ends
				721	with a <a href="#terminators">terminator</a> instruction (such as a branch or
				722	function return).</p>
				723
				724	<p>The first basic block in a function is special in two ways: it is immediately
				725	executed on entrance to the function, and it is not allowed to have predecessor
				726	basic blocks (i.e. there can not be any branches to the entry block of a
				727	function). Because the block can have no predecessors, it also cannot have any
				728	<a href="#i_phi">PHI nodes</a>.</p>
				729
				730	<p>LLVM allows an explicit section to be specified for functions. If the target
				731	supports it, it will emit functions to the section specified.</p>
				732
				733	<p>An explicit alignment may be specified for a function. If not present, or if
				734	the alignment is set to zero, the alignment of the function is set by the target
				735	to whatever it feels convenient. If an explicit alignment is specified, the
				736	function is forced to have at least that much alignment. All alignments must be
				737	a power of 2.</p>
				738
				739	</div>
				740
				741
				742	<!-- ======================================================================= -->
				743	<div class="doc_subsection">
				744	<a name="aliasstructure">Aliases</a>
				745	</div>
				746	<div class="doc_text">
				747	<p>Aliases act as "second name" for the aliasee value (which can be either
				748	function or global variable or bitcast of global value). Aliases may have an
				749	optional <a href="#linkage">linkage type</a>, and an
				750	optional <a href="#visibility">visibility style</a>.</p>
				751
				752	<h5>Syntax:</h5>
				753
				754	<div class="doc_code">
				755	<pre>
				756	@<Name> = [Linkage] [Visibility] alias <AliaseeTy> @<Aliasee>
				757	</pre>
				758	</div>
				759
				760	</div>
				761
				762
				763
				764	<!-- ======================================================================= -->
				765	<div class="doc_subsection"><a name="paramattrs">Parameter Attributes</a></div>
				766	<div class="doc_text">
				767	<p>The return type and each parameter of a function type may have a set of
				768	<i>parameter attributes</i> associated with them. Parameter attributes are
				769	used to communicate additional information about the result or parameters of
				770	a function. Parameter attributes are considered to be part of the function
				771	type so two functions types that differ only by the parameter attributes
				772	are different function types.</p>
				773
				774	<p>Parameter attributes are simple keywords that follow the type specified. If
				775	multiple parameter attributes are needed, they are space separated. For
				776	example:</p>
				777
				778	<div class="doc_code">
				779	<pre>
Reid Spencer	f234bed	2007-07-19 23:13:04 +0000	[diff] [blame]	780	%someFunc = i16 (i8 signext %someParam) zeroext
				781	%someFunc = i16 (i8 zeroext %someParam) zeroext
Dan Gohman	f17a25c	2007-07-18 16:29:46 +0000	[diff] [blame]	782	</pre>
				783	</div>
				784
				785	<p>Note that the two function types above are unique because the parameter has
Reid Spencer	f234bed	2007-07-19 23:13:04 +0000	[diff] [blame]	786	a different attribute (<tt>signext</tt> in the first one, <tt>zeroext</tt> in
				787	the second). Also note that the attribute for the function result
				788	(<tt>zeroext</tt>) comes immediately after the argument list.</p>
Dan Gohman	f17a25c	2007-07-18 16:29:46 +0000	[diff] [blame]	789
				790	<p>Currently, only the following parameter attributes are defined:</p>
				791	<dl>
Reid Spencer	f234bed	2007-07-19 23:13:04 +0000	[diff] [blame]	792	<dt><tt>zeroext</tt></dt>
Dan Gohman	f17a25c	2007-07-18 16:29:46 +0000	[diff] [blame]	793	<dd>This indicates that the parameter should be zero extended just before
				794	a call to this function.</dd>
Reid Spencer	f234bed	2007-07-19 23:13:04 +0000	[diff] [blame]	795	<dt><tt>signext</tt></dt>
Dan Gohman	f17a25c	2007-07-18 16:29:46 +0000	[diff] [blame]	796	<dd>This indicates that the parameter should be sign extended just before
				797	a call to this function.</dd>
				798	<dt><tt>inreg</tt></dt>
				799	<dd>This indicates that the parameter should be placed in register (if
				800	possible) during assembling function call. Support for this attribute is
				801	target-specific</dd>
				802	<dt><tt>sret</tt></dt>
				803	<dd>This indicates that the parameter specifies the address of a structure
				804	that is the return value of the function in the source program.</dd>
				805	<dt><tt>noalias</tt></dt>
				806	<dd>This indicates that the parameter not alias any other object or any
				807	other "noalias" objects during the function call.
				808	<dt><tt>noreturn</tt></dt>
				809	<dd>This function attribute indicates that the function never returns. This
				810	indicates to LLVM that every call to this function should be treated as if
				811	an <tt>unreachable</tt> instruction immediately followed the call.</dd>
				812	<dt><tt>nounwind</tt></dt>
				813	<dd>This function attribute indicates that the function type does not use
				814	the unwind instruction and does not allow stack unwinding to propagate
				815	through it.</dd>
Duncan Sands	4ee4681	2007-07-27 19:57:41 +0000	[diff] [blame]	816	<dt><tt>nest</tt></dt>
				817	<dd>This indicates that the parameter can be excised using the
				818	<a href="#int_trampoline">trampoline intrinsics</a>.</dd>
Dan Gohman	f17a25c	2007-07-18 16:29:46 +0000	[diff] [blame]	819	</dl>
				820
				821	</div>
				822
				823	<!-- ======================================================================= -->
				824	<div class="doc_subsection">
				825	<a name="moduleasm">Module-Level Inline Assembly</a>
				826	</div>
				827
				828	<div class="doc_text">
				829	<p>
				830	Modules may contain "module-level inline asm" blocks, which corresponds to the
				831	GCC "file scope inline asm" blocks. These blocks are internally concatenated by
				832	LLVM and treated as a single unit, but may be separated in the .ll file if
				833	desired. The syntax is very simple:
				834	</p>
				835
				836	<div class="doc_code">
				837	<pre>
				838	module asm "inline asm code goes here"
				839	module asm "more can go here"
				840	</pre>
				841	</div>
				842
				843	<p>The strings can contain any character by escaping non-printable characters.
				844	The escape sequence used is simply "\xx" where "xx" is the two digit hex code
				845	for the number.
				846	</p>
				847
				848	<p>
				849	The inline asm code is simply printed to the machine code .s file when
				850	assembly code is generated.
				851	</p>
				852	</div>
				853
				854	<!-- ======================================================================= -->
				855	<div class="doc_subsection">
				856	<a name="datalayout">Data Layout</a>
				857	</div>
				858
				859	<div class="doc_text">
				860	<p>A module may specify a target specific data layout string that specifies how
				861	data is to be laid out in memory. The syntax for the data layout is simply:</p>
				862	<pre> target datalayout = "<i>layout specification</i>"</pre>
				863	<p>The <i>layout specification</i> consists of a list of specifications
				864	separated by the minus sign character ('-'). Each specification starts with a
				865	letter and may include other information after the letter to define some
				866	aspect of the data layout. The specifications accepted are as follows: </p>
				867	<dl>
				868	<dt><tt>E</tt></dt>
				869	<dd>Specifies that the target lays out data in big-endian form. That is, the
				870	bits with the most significance have the lowest address location.</dd>
				871	<dt><tt>e</tt></dt>
				872	<dd>Specifies that hte target lays out data in little-endian form. That is,
				873	the bits with the least significance have the lowest address location.</dd>
				874	<dt><tt>p:<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
				875	<dd>This specifies the <i>size</i> of a pointer and its <i>abi</i> and
				876	<i>preferred</i> alignments. All sizes are in bits. Specifying the <i>pref</i>
				877	alignment is optional. If omitted, the preceding <tt>:</tt> should be omitted
				878	too.</dd>
				879	<dt><tt>i<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
				880	<dd>This specifies the alignment for an integer type of a given bit
				881	<i>size</i>. The value of <i>size</i> must be in the range [1,2^23).</dd>
				882	<dt><tt>v<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
				883	<dd>This specifies the alignment for a vector type of a given bit
				884	<i>size</i>.</dd>
				885	<dt><tt>f<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
				886	<dd>This specifies the alignment for a floating point type of a given bit
				887	<i>size</i>. The value of <i>size</i> must be either 32 (float) or 64
				888	(double).</dd>
				889	<dt><tt>a<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
				890	<dd>This specifies the alignment for an aggregate type of a given bit
				891	<i>size</i>.</dd>
				892	</dl>
				893	<p>When constructing the data layout for a given target, LLVM starts with a
				894	default set of specifications which are then (possibly) overriden by the
				895	specifications in the <tt>datalayout</tt> keyword. The default specifications
				896	are given in this list:</p>
				897	<ul>
				898	<li><tt>E</tt> - big endian</li>
				899	<li><tt>p:32:64:64</tt> - 32-bit pointers with 64-bit alignment</li>
				900	<li><tt>i1:8:8</tt> - i1 is 8-bit (byte) aligned</li>
				901	<li><tt>i8:8:8</tt> - i8 is 8-bit (byte) aligned</li>
				902	<li><tt>i16:16:16</tt> - i16 is 16-bit aligned</li>
				903	<li><tt>i32:32:32</tt> - i32 is 32-bit aligned</li>
				904	<li><tt>i64:32:64</tt> - i64 has abi alignment of 32-bits but preferred
				905	alignment of 64-bits</li>
				906	<li><tt>f32:32:32</tt> - float is 32-bit aligned</li>
				907	<li><tt>f64:64:64</tt> - double is 64-bit aligned</li>
				908	<li><tt>v64:64:64</tt> - 64-bit vector is 64-bit aligned</li>
				909	<li><tt>v128:128:128</tt> - 128-bit vector is 128-bit aligned</li>
				910	<li><tt>a0:0:1</tt> - aggregates are 8-bit aligned</li>
				911	</ul>
				912	<p>When llvm is determining the alignment for a given type, it uses the
				913	following rules:
				914	<ol>
				915	<li>If the type sought is an exact match for one of the specifications, that
				916	specification is used.</li>
				917	<li>If no match is found, and the type sought is an integer type, then the
				918	smallest integer type that is larger than the bitwidth of the sought type is
				919	used. If none of the specifications are larger than the bitwidth then the the
				920	largest integer type is used. For example, given the default specifications
				921	above, the i7 type will use the alignment of i8 (next largest) while both
				922	i65 and i256 will use the alignment of i64 (largest specified).</li>
				923	<li>If no match is found, and the type sought is a vector type, then the
				924	largest vector type that is smaller than the sought vector type will be used
				925	as a fall back. This happens because <128 x double> can be implemented in
				926	terms of 64 <2 x double>, for example.</li>
				927	</ol>
				928	</div>
				929
				930	<!-- *********************************************************************** -->
				931	<div class="doc_section"> <a name="typesystem">Type System</a> </div>
				932	<!-- *********************************************************************** -->
				933
				934	<div class="doc_text">
				935
				936	<p>The LLVM type system is one of the most important features of the
				937	intermediate representation. Being typed enables a number of
				938	optimizations to be performed on the IR directly, without having to do
				939	extra analyses on the side before the transformation. A strong type
				940	system makes it easier to read the generated code and enables novel
				941	analyses and transformations that are not feasible to perform on normal
				942	three address code representations.</p>
				943
				944	</div>
				945
				946	<!-- ======================================================================= -->
				947	<div class="doc_subsection"> <a name="t_primitive">Primitive Types</a> </div>
				948	<div class="doc_text">
				949	<p>The primitive types are the fundamental building blocks of the LLVM
				950	system. The current set of primitive types is as follows:</p>
				951
				952	<table class="layout">
				953	<tr class="layout">
				954	<td class="left">
				955	<table>
				956	<tbody>
				957	<tr><th>Type</th><th>Description</th></tr>
				958	<tr><td><tt><a name="t_void">void</a></tt></td><td>No value</td></tr>
				959	<tr><td><tt>label</tt></td><td>Branch destination</td></tr>
				960	</tbody>
				961	</table>
				962	</td>
				963	<td class="right">
				964	<table>
				965	<tbody>
				966	<tr><th>Type</th><th>Description</th></tr>
				967	<tr><td><tt>float</tt></td><td>32-bit floating point value</td></tr>
				968	<tr><td><tt>double</tt></td><td>64-bit floating point value</td></tr>
				969	</tbody>
				970	</table>
				971	</td>
				972	</tr>
				973	</table>
				974	</div>
				975
				976	<!-- _______________________________________________________________________ -->
				977	<div class="doc_subsubsection"> <a name="t_classifications">Type
				978	Classifications</a> </div>
				979	<div class="doc_text">
				980	<p>These different primitive types fall into a few useful
				981	classifications:</p>
				982
				983	<table border="1" cellspacing="0" cellpadding="4">
				984	<tbody>
				985	<tr><th>Classification</th><th>Types</th></tr>
				986	<tr>
				987	<td><a name="t_integer">integer</a></td>
				988	<td><tt>i1, i2, i3, ... i8, ... i16, ... i32, ... i64, ... </tt></td>
				989	</tr>
				990	<tr>
				991	<td><a name="t_floating">floating point</a></td>
				992	<td><tt>float, double</tt></td>
				993	</tr>
				994	<tr>
				995	<td><a name="t_firstclass">first class</a></td>
				996	<td><tt>i1, ..., float, double, <br/>
				997	<a href="#t_pointer">pointer</a>,<a href="#t_vector">vector</a></tt>
				998	</td>
				999	</tr>
				1000	</tbody>
				1001	</table>
				1002
				1003	<p>The <a href="#t_firstclass">first class</a> types are perhaps the
				1004	most important. Values of these types are the only ones which can be
				1005	produced by instructions, passed as arguments, or used as operands to
				1006	instructions. This means that all structures and arrays must be
				1007	manipulated either by pointer or by component.</p>
				1008	</div>
				1009
				1010	<!-- ======================================================================= -->
				1011	<div class="doc_subsection"> <a name="t_derived">Derived Types</a> </div>
				1012
				1013	<div class="doc_text">
				1014
				1015	<p>The real power in LLVM comes from the derived types in the system.
				1016	This is what allows a programmer to represent arrays, functions,
				1017	pointers, and other useful types. Note that these derived types may be
				1018	recursive: For example, it is possible to have a two dimensional array.</p>
				1019
				1020	</div>
				1021
				1022	<!-- _______________________________________________________________________ -->
				1023	<div class="doc_subsubsection"> <a name="t_integer">Integer Type</a> </div>
				1024
				1025	<div class="doc_text">
				1026
				1027	<h5>Overview:</h5>
				1028	<p>The integer type is a very simple derived type that simply specifies an
				1029	arbitrary bit width for the integer type desired. Any bit width from 1 bit to
				1030	2^23-1 (about 8 million) can be specified.</p>
				1031
				1032	<h5>Syntax:</h5>
				1033
				1034	<pre>
				1035	iN
				1036	</pre>
				1037
				1038	<p>The number of bits the integer will occupy is specified by the <tt>N</tt>
				1039	value.</p>
				1040
				1041	<h5>Examples:</h5>
				1042	<table class="layout">
				1043	<tr class="layout">
				1044	<td class="left">
				1045	<tt>i1</tt><br/>
				1046	<tt>i4</tt><br/>
				1047	<tt>i8</tt><br/>
				1048	<tt>i16</tt><br/>
				1049	<tt>i32</tt><br/>
				1050	<tt>i42</tt><br/>
				1051	<tt>i64</tt><br/>
				1052	<tt>i1942652</tt><br/>
				1053	</td>
				1054	<td class="left">
				1055	A boolean integer of 1 bit<br/>
				1056	A nibble sized integer of 4 bits.<br/>
				1057	A byte sized integer of 8 bits.<br/>
				1058	A half word sized integer of 16 bits.<br/>
				1059	A word sized integer of 32 bits.<br/>
				1060	An integer whose bit width is the answer. <br/>
				1061	A double word sized integer of 64 bits.<br/>
				1062	A really big integer of over 1 million bits.<br/>
				1063	</td>
				1064	</tr>
				1065	</table>
				1066	</div>
				1067
				1068	<!-- _______________________________________________________________________ -->
				1069	<div class="doc_subsubsection"> <a name="t_array">Array Type</a> </div>
				1070
				1071	<div class="doc_text">
				1072
				1073	<h5>Overview:</h5>
				1074
				1075	<p>The array type is a very simple derived type that arranges elements
				1076	sequentially in memory. The array type requires a size (number of
				1077	elements) and an underlying data type.</p>
				1078
				1079	<h5>Syntax:</h5>
				1080
				1081	<pre>
				1082	[<# elements> x <elementtype>]
				1083	</pre>
				1084
				1085	<p>The number of elements is a constant integer value; elementtype may
				1086	be any type with a size.</p>
				1087
				1088	<h5>Examples:</h5>
				1089	<table class="layout">
				1090	<tr class="layout">
				1091	<td class="left">
				1092	<tt>[40 x i32 ]</tt><br/>
				1093	<tt>[41 x i32 ]</tt><br/>
				1094	<tt>[40 x i8]</tt><br/>
				1095	</td>
				1096	<td class="left">
				1097	Array of 40 32-bit integer values.<br/>
				1098	Array of 41 32-bit integer values.<br/>
				1099	Array of 40 8-bit integer values.<br/>
				1100	</td>
				1101	</tr>
				1102	</table>
				1103	<p>Here are some examples of multidimensional arrays:</p>
				1104	<table class="layout">
				1105	<tr class="layout">
				1106	<td class="left">
				1107	<tt>[3 x [4 x i32]]</tt><br/>
				1108	<tt>[12 x [10 x float]]</tt><br/>
				1109	<tt>[2 x [3 x [4 x i16]]]</tt><br/>
				1110	</td>
				1111	<td class="left">
				1112	3x4 array of 32-bit integer values.<br/>
				1113	12x10 array of single precision floating point values.<br/>
				1114	2x3x4 array of 16-bit integer values.<br/>
				1115	</td>
				1116	</tr>
				1117	</table>
				1118
				1119	<p>Note that 'variable sized arrays' can be implemented in LLVM with a zero
				1120	length array. Normally, accesses past the end of an array are undefined in
				1121	LLVM (e.g. it is illegal to access the 5th element of a 3 element array).
				1122	As a special case, however, zero length arrays are recognized to be variable
				1123	length. This allows implementation of 'pascal style arrays' with the LLVM
				1124	type "{ i32, [0 x float]}", for example.</p>
				1125
				1126	</div>
				1127
				1128	<!-- _______________________________________________________________________ -->
				1129	<div class="doc_subsubsection"> <a name="t_function">Function Type</a> </div>
				1130	<div class="doc_text">
				1131	<h5>Overview:</h5>
				1132	<p>The function type can be thought of as a function signature. It
				1133	consists of a return type and a list of formal parameter types.
				1134	Function types are usually used to build virtual function tables
				1135	(which are structures of pointers to functions), for indirect function
				1136	calls, and when defining a function.</p>
				1137	<p>
				1138	The return type of a function type cannot be an aggregate type.
				1139	</p>
				1140	<h5>Syntax:</h5>
				1141	<pre> <returntype> (<parameter list>)<br></pre>
				1142	<p>...where '<tt><parameter list></tt>' is a comma-separated list of type
				1143	specifiers. Optionally, the parameter list may include a type <tt>...</tt>,
				1144	which indicates that the function takes a variable number of arguments.
				1145	Variable argument functions can access their arguments with the <a
				1146	href="#int_varargs">variable argument handling intrinsic</a> functions.</p>
				1147	<h5>Examples:</h5>
				1148	<table class="layout">
				1149	<tr class="layout">
				1150	<td class="left"><tt>i32 (i32)</tt></td>
				1151	<td class="left">function taking an <tt>i32</tt>, returning an <tt>i32</tt>
				1152	</td>
				1153	</tr><tr class="layout">
Reid Spencer	f234bed	2007-07-19 23:13:04 +0000	[diff] [blame]	1154	<td class="left"><tt>float (i16 signext, i32 )
Dan Gohman	f17a25c	2007-07-18 16:29:46 +0000	[diff] [blame]	1155	</tt></td>
				1156	<td class="left"><a href="#t_pointer">Pointer</a> to a function that takes
				1157	an <tt>i16</tt> that should be sign extended and a
				1158	<a href="#t_pointer">pointer</a> to <tt>i32</tt>, returning
				1159	<tt>float</tt>.
				1160	</td>
				1161	</tr><tr class="layout">
				1162	<td class="left"><tt>i32 (i8*, ...)</tt></td>
				1163	<td class="left">A vararg function that takes at least one
				1164	<a href="#t_pointer">pointer</a> to <tt>i8 </tt> (char in C),
				1165	which returns an integer. This is the signature for <tt>printf</tt> in
				1166	LLVM.
				1167	</td>
				1168	</tr>
				1169	</table>
				1170
				1171	</div>
				1172	<!-- _______________________________________________________________________ -->
				1173	<div class="doc_subsubsection"> <a name="t_struct">Structure Type</a> </div>
				1174	<div class="doc_text">
				1175	<h5>Overview:</h5>
				1176	<p>The structure type is used to represent a collection of data members
				1177	together in memory. The packing of the field types is defined to match
				1178	the ABI of the underlying processor. The elements of a structure may
				1179	be any type that has a size.</p>
				1180	<p>Structures are accessed using '<tt><a href="#i_load">load</a></tt>
				1181	and '<tt><a href="#i_store">store</a></tt>' by getting a pointer to a
				1182	field with the '<tt><a href="#i_getelementptr">getelementptr</a></tt>'
				1183	instruction.</p>
				1184	<h5>Syntax:</h5>
				1185	<pre> { <type list> }<br></pre>
				1186	<h5>Examples:</h5>
				1187	<table class="layout">
				1188	<tr class="layout">
				1189	<td class="left"><tt>{ i32, i32, i32 }</tt></td>
				1190	<td class="left">A triple of three <tt>i32</tt> values</td>
				1191	</tr><tr class="layout">
				1192	<td class="left"><tt>{ float, i32 (i32) * }</tt></td>
				1193	<td class="left">A pair, where the first element is a <tt>float</tt> and the
				1194	second element is a <a href="#t_pointer">pointer</a> to a
				1195	<a href="#t_function">function</a> that takes an <tt>i32</tt>, returning
				1196	an <tt>i32</tt>.</td>
				1197	</tr>
				1198	</table>
				1199	</div>
				1200
				1201	<!-- _______________________________________________________________________ -->
				1202	<div class="doc_subsubsection"> <a name="t_pstruct">Packed Structure Type</a>
				1203	</div>
				1204	<div class="doc_text">
				1205	<h5>Overview:</h5>
				1206	<p>The packed structure type is used to represent a collection of data members
				1207	together in memory. There is no padding between fields. Further, the alignment
				1208	of a packed structure is 1 byte. The elements of a packed structure may
				1209	be any type that has a size.</p>
				1210	<p>Structures are accessed using '<tt><a href="#i_load">load</a></tt>
				1211	and '<tt><a href="#i_store">store</a></tt>' by getting a pointer to a
				1212	field with the '<tt><a href="#i_getelementptr">getelementptr</a></tt>'
				1213	instruction.</p>
				1214	<h5>Syntax:</h5>
				1215	<pre> < { <type list> } > <br></pre>
				1216	<h5>Examples:</h5>
				1217	<table class="layout">
				1218	<tr class="layout">
				1219	<td class="left"><tt>< { i32, i32, i32 } ></tt></td>
				1220	<td class="left">A triple of three <tt>i32</tt> values</td>
				1221	</tr><tr class="layout">
				1222	<td class="left"><tt>< { float, i32 (i32) * } ></tt></td>
				1223	<td class="left">A pair, where the first element is a <tt>float</tt> and the
				1224	second element is a <a href="#t_pointer">pointer</a> to a
				1225	<a href="#t_function">function</a> that takes an <tt>i32</tt>, returning
				1226	an <tt>i32</tt>.</td>
				1227	</tr>
				1228	</table>
				1229	</div>
				1230
				1231	<!-- _______________________________________________________________________ -->
				1232	<div class="doc_subsubsection"> <a name="t_pointer">Pointer Type</a> </div>
				1233	<div class="doc_text">
				1234	<h5>Overview:</h5>
				1235	<p>As in many languages, the pointer type represents a pointer or
				1236	reference to another object, which must live in memory.</p>
				1237	<h5>Syntax:</h5>
				1238	<pre> <type> *<br></pre>
				1239	<h5>Examples:</h5>
				1240	<table class="layout">
				1241	<tr class="layout">
				1242	<td class="left">
				1243	<tt>[4x i32]*</tt><br/>
				1244	<tt>i32 (i32 ) </tt><br/>
				1245	</td>
				1246	<td class="left">
				1247	A <a href="#t_pointer">pointer</a> to <a href="#t_array">array</a> of
				1248	four <tt>i32</tt> values<br/>
				1249	A <a href="#t_pointer">pointer</a> to a <a
				1250	href="#t_function">function</a> that takes an <tt>i32*</tt>, returning an
				1251	<tt>i32</tt>.<br/>
				1252	</td>
				1253	</tr>
				1254	</table>
				1255	</div>
				1256
				1257	<!-- _______________________________________________________________________ -->
				1258	<div class="doc_subsubsection"> <a name="t_vector">Vector Type</a> </div>
				1259	<div class="doc_text">
				1260
				1261	<h5>Overview:</h5>
				1262
				1263	<p>A vector type is a simple derived type that represents a vector
				1264	of elements. Vector types are used when multiple primitive data
				1265	are operated in parallel using a single instruction (SIMD).
				1266	A vector type requires a size (number of
				1267	elements) and an underlying primitive data type. Vectors must have a power
				1268	of two length (1, 2, 4, 8, 16 ...). Vector types are
				1269	considered <a href="#t_firstclass">first class</a>.</p>
				1270
				1271	<h5>Syntax:</h5>
				1272
				1273	<pre>
				1274	< <# elements> x <elementtype> >
				1275	</pre>
				1276
				1277	<p>The number of elements is a constant integer value; elementtype may
				1278	be any integer or floating point type.</p>
				1279
				1280	<h5>Examples:</h5>
				1281
				1282	<table class="layout">
				1283	<tr class="layout">
				1284	<td class="left">
				1285	<tt><4 x i32></tt><br/>
				1286	<tt><8 x float></tt><br/>
				1287	<tt><2 x i64></tt><br/>
				1288	</td>
				1289	<td class="left">
				1290	Vector of 4 32-bit integer values.<br/>
				1291	Vector of 8 floating-point values.<br/>
				1292	Vector of 2 64-bit integer values.<br/>
				1293	</td>
				1294	</tr>
				1295	</table>
				1296	</div>
				1297
				1298	<!-- _______________________________________________________________________ -->
				1299	<div class="doc_subsubsection"> <a name="t_opaque">Opaque Type</a> </div>
				1300	<div class="doc_text">
				1301
				1302	<h5>Overview:</h5>
				1303
				1304	<p>Opaque types are used to represent unknown types in the system. This
				1305	corresponds (for example) to the C notion of a foward declared structure type.
				1306	In LLVM, opaque types can eventually be resolved to any type (not just a
				1307	structure type).</p>
				1308
				1309	<h5>Syntax:</h5>
				1310
				1311	<pre>
				1312	opaque
				1313	</pre>
				1314
				1315	<h5>Examples:</h5>
				1316
				1317	<table class="layout">
				1318	<tr class="layout">
				1319	<td class="left">
				1320	<tt>opaque</tt>
				1321	</td>
				1322	<td class="left">
				1323	An opaque type.<br/>
				1324	</td>
				1325	</tr>
				1326	</table>
				1327	</div>
				1328
				1329
				1330	<!-- *********************************************************************** -->
				1331	<div class="doc_section"> <a name="constants">Constants</a> </div>
				1332	<!-- *********************************************************************** -->
				1333
				1334	<div class="doc_text">
				1335
				1336	<p>LLVM has several different basic types of constants. This section describes
				1337	them all and their syntax.</p>
				1338
				1339	</div>
				1340
				1341	<!-- ======================================================================= -->
				1342	<div class="doc_subsection"><a name="simpleconstants">Simple Constants</a></div>
				1343
				1344	<div class="doc_text">
				1345
				1346	<dl>
				1347	<dt><b>Boolean constants</b></dt>
				1348
				1349	<dd>The two strings '<tt>true</tt>' and '<tt>false</tt>' are both valid
				1350	constants of the <tt><a href="#t_primitive">i1</a></tt> type.
				1351	</dd>
				1352
				1353	<dt><b>Integer constants</b></dt>
				1354
				1355	<dd>Standard integers (such as '4') are constants of the <a
				1356	href="#t_integer">integer</a> type. Negative numbers may be used with
				1357	integer types.
				1358	</dd>
				1359
				1360	<dt><b>Floating point constants</b></dt>
				1361
				1362	<dd>Floating point constants use standard decimal notation (e.g. 123.421),
				1363	exponential notation (e.g. 1.23421e+2), or a more precise hexadecimal
				1364	notation (see below). Floating point constants must have a <a
				1365	href="#t_floating">floating point</a> type. </dd>
				1366
				1367	<dt><b>Null pointer constants</b></dt>
				1368
				1369	<dd>The identifier '<tt>null</tt>' is recognized as a null pointer constant
				1370	and must be of <a href="#t_pointer">pointer type</a>.</dd>
				1371
				1372	</dl>
				1373
				1374	<p>The one non-intuitive notation for constants is the optional hexadecimal form
				1375	of floating point constants. For example, the form '<tt>double
				1376	0x432ff973cafa8000</tt>' is equivalent to (but harder to read than) '<tt>double
				1377	4.5e+15</tt>'. The only time hexadecimal floating point constants are required
				1378	(and the only time that they are generated by the disassembler) is when a
				1379	floating point constant must be emitted but it cannot be represented as a
				1380	decimal floating point number. For example, NaN's, infinities, and other
				1381	special values are represented in their IEEE hexadecimal format so that
				1382	assembly and disassembly do not cause any bits to change in the constants.</p>
				1383
				1384	</div>
				1385
				1386	<!-- ======================================================================= -->
				1387	<div class="doc_subsection"><a name="aggregateconstants">Aggregate Constants</a>
				1388	</div>
				1389
				1390	<div class="doc_text">
				1391	<p>Aggregate constants arise from aggregation of simple constants
				1392	and smaller aggregate constants.</p>
				1393
				1394	<dl>
				1395	<dt><b>Structure constants</b></dt>
				1396
				1397	<dd>Structure constants are represented with notation similar to structure
				1398	type definitions (a comma separated list of elements, surrounded by braces
				1399	(<tt>{}</tt>)). For example: "<tt>{ i32 4, float 17.0, i32* %G }</tt>",
				1400	where "<tt>%G</tt>" is declared as "<tt>@G = external global i32</tt>". Structure constants
				1401	must have <a href="#t_struct">structure type</a>, and the number and
				1402	types of elements must match those specified by the type.
				1403	</dd>
				1404
				1405	<dt><b>Array constants</b></dt>
				1406
				1407	<dd>Array constants are represented with notation similar to array type
				1408	definitions (a comma separated list of elements, surrounded by square brackets
				1409	(<tt>[]</tt>)). For example: "<tt>[ i32 42, i32 11, i32 74 ]</tt>". Array
				1410	constants must have <a href="#t_array">array type</a>, and the number and
				1411	types of elements must match those specified by the type.
				1412	</dd>
				1413
				1414	<dt><b>Vector constants</b></dt>
				1415
				1416	<dd>Vector constants are represented with notation similar to vector type
				1417	definitions (a comma separated list of elements, surrounded by
				1418	less-than/greater-than's (<tt><></tt>)). For example: "<tt>< i32 42,
				1419	i32 11, i32 74, i32 100 ></tt>". Vector constants must have <a
				1420	href="#t_vector">vector type</a>, and the number and types of elements must
				1421	match those specified by the type.
				1422	</dd>
				1423
				1424	<dt><b>Zero initialization</b></dt>
				1425
				1426	<dd>The string '<tt>zeroinitializer</tt>' can be used to zero initialize a
				1427	value to zero of <em>any</em> type, including scalar and aggregate types.
				1428	This is often used to avoid having to print large zero initializers (e.g. for
				1429	large arrays) and is always exactly equivalent to using explicit zero
				1430	initializers.
				1431	</dd>
				1432	</dl>
				1433
				1434	</div>
				1435
				1436	<!-- ======================================================================= -->
				1437	<div class="doc_subsection">
				1438	<a name="globalconstants">Global Variable and Function Addresses</a>
				1439	</div>
				1440
				1441	<div class="doc_text">
				1442
				1443	<p>The addresses of <a href="#globalvars">global variables</a> and <a
				1444	href="#functionstructure">functions</a> are always implicitly valid (link-time)
				1445	constants. These constants are explicitly referenced when the <a
				1446	href="#identifiers">identifier for the global</a> is used and always have <a
				1447	href="#t_pointer">pointer</a> type. For example, the following is a legal LLVM
				1448	file:</p>
				1449
				1450	<div class="doc_code">
				1451	<pre>
				1452	@X = global i32 17
				1453	@Y = global i32 42
				1454	@Z = global [2 x i32] [ i32 @X, i32* @Y ]
				1455	</pre>
				1456	</div>
				1457
				1458	</div>
				1459
				1460	<!-- ======================================================================= -->
				1461	<div class="doc_subsection"><a name="undefvalues">Undefined Values</a></div>
				1462	<div class="doc_text">
				1463	<p>The string '<tt>undef</tt>' is recognized as a type-less constant that has
				1464	no specific value. Undefined values may be of any type and be used anywhere
				1465	a constant is permitted.</p>
				1466
				1467	<p>Undefined values indicate to the compiler that the program is well defined
				1468	no matter what value is used, giving the compiler more freedom to optimize.
				1469	</p>
				1470	</div>
				1471
				1472	<!-- ======================================================================= -->
				1473	<div class="doc_subsection"><a name="constantexprs">Constant Expressions</a>
				1474	</div>
				1475
				1476	<div class="doc_text">
				1477
				1478	<p>Constant expressions are used to allow expressions involving other constants
				1479	to be used as constants. Constant expressions may be of any <a
				1480	href="#t_firstclass">first class</a> type and may involve any LLVM operation
				1481	that does not have side effects (e.g. load and call are not supported). The
				1482	following is the syntax for constant expressions:</p>
				1483
				1484	<dl>
				1485	<dt><b><tt>trunc ( CST to TYPE )</tt></b></dt>
				1486	<dd>Truncate a constant to another type. The bit size of CST must be larger
				1487	than the bit size of TYPE. Both types must be integers.</dd>
				1488
				1489	<dt><b><tt>zext ( CST to TYPE )</tt></b></dt>
				1490	<dd>Zero extend a constant to another type. The bit size of CST must be
				1491	smaller or equal to the bit size of TYPE. Both types must be integers.</dd>
				1492
				1493	<dt><b><tt>sext ( CST to TYPE )</tt></b></dt>
				1494	<dd>Sign extend a constant to another type. The bit size of CST must be
				1495	smaller or equal to the bit size of TYPE. Both types must be integers.</dd>
				1496
				1497	<dt><b><tt>fptrunc ( CST to TYPE )</tt></b></dt>
				1498	<dd>Truncate a floating point constant to another floating point type. The
				1499	size of CST must be larger than the size of TYPE. Both types must be
				1500	floating point.</dd>
				1501
				1502	<dt><b><tt>fpext ( CST to TYPE )</tt></b></dt>
				1503	<dd>Floating point extend a constant to another type. The size of CST must be
				1504	smaller or equal to the size of TYPE. Both types must be floating point.</dd>
				1505
Reid Spencer	e6adee8	2007-07-31 14:40:14 +0000	[diff] [blame]	1506	<dt><b><tt>fptoui ( CST to TYPE )</tt></b></dt>
Dan Gohman	f17a25c	2007-07-18 16:29:46 +0000	[diff] [blame]	1507	<dd>Convert a floating point constant to the corresponding unsigned integer
				1508	constant. TYPE must be an integer type. CST must be floating point. If the
				1509	value won't fit in the integer type, the results are undefined.</dd>
				1510
				1511	<dt><b><tt>fptosi ( CST to TYPE )</tt></b></dt>
				1512	<dd>Convert a floating point constant to the corresponding signed integer
				1513	constant. TYPE must be an integer type. CST must be floating point. If the
				1514	value won't fit in the integer type, the results are undefined.</dd>
				1515
				1516	<dt><b><tt>uitofp ( CST to TYPE )</tt></b></dt>
				1517	<dd>Convert an unsigned integer constant to the corresponding floating point
				1518	constant. TYPE must be floating point. CST must be of integer type. If the
				1519	value won't fit in the floating point type, the results are undefined.</dd>
				1520
				1521	<dt><b><tt>sitofp ( CST to TYPE )</tt></b></dt>
				1522	<dd>Convert a signed integer constant to the corresponding floating point
				1523	constant. TYPE must be floating point. CST must be of integer type. If the
				1524	value won't fit in the floating point type, the results are undefined.</dd>
				1525
				1526	<dt><b><tt>ptrtoint ( CST to TYPE )</tt></b></dt>
				1527	<dd>Convert a pointer typed constant to the corresponding integer constant
				1528	TYPE must be an integer type. CST must be of pointer type. The CST value is
				1529	zero extended, truncated, or unchanged to make it fit in TYPE.</dd>
				1530
				1531	<dt><b><tt>inttoptr ( CST to TYPE )</tt></b></dt>
				1532	<dd>Convert a integer constant to a pointer constant. TYPE must be a
				1533	pointer type. CST must be of integer type. The CST value is zero extended,
				1534	truncated, or unchanged to make it fit in a pointer size. This one is
				1535	<i>really</i> dangerous!</dd>
				1536
				1537	<dt><b><tt>bitcast ( CST to TYPE )</tt></b></dt>
				1538	<dd>Convert a constant, CST, to another TYPE. The size of CST and TYPE must be
				1539	identical (same number of bits). The conversion is done as if the CST value
				1540	was stored to memory and read back as TYPE. In other words, no bits change
				1541	with this operator, just the type. This can be used for conversion of
				1542	vector types to any other type, as long as they have the same bit width. For
				1543	pointers it is only valid to cast to another pointer type.
				1544	</dd>
				1545
				1546	<dt><b><tt>getelementptr ( CSTPTR, IDX0, IDX1, ... )</tt></b></dt>
				1547
				1548	<dd>Perform the <a href="#i_getelementptr">getelementptr operation</a> on
				1549	constants. As with the <a href="#i_getelementptr">getelementptr</a>
				1550	instruction, the index list may have zero or more indexes, which are required
				1551	to make sense for the type of "CSTPTR".</dd>
				1552
				1553	<dt><b><tt>select ( COND, VAL1, VAL2 )</tt></b></dt>
				1554
				1555	<dd>Perform the <a href="#i_select">select operation</a> on
				1556	constants.</dd>
				1557
				1558	<dt><b><tt>icmp COND ( VAL1, VAL2 )</tt></b></dt>
				1559	<dd>Performs the <a href="#i_icmp">icmp operation</a> on constants.</dd>
				1560
				1561	<dt><b><tt>fcmp COND ( VAL1, VAL2 )</tt></b></dt>
				1562	<dd>Performs the <a href="#i_fcmp">fcmp operation</a> on constants.</dd>
				1563
				1564	<dt><b><tt>extractelement ( VAL, IDX )</tt></b></dt>
				1565
				1566	<dd>Perform the <a href="#i_extractelement">extractelement
				1567	operation</a> on constants.
				1568
				1569	<dt><b><tt>insertelement ( VAL, ELT, IDX )</tt></b></dt>
				1570
				1571	<dd>Perform the <a href="#i_insertelement">insertelement
				1572	operation</a> on constants.</dd>
				1573
				1574
				1575	<dt><b><tt>shufflevector ( VEC1, VEC2, IDXMASK )</tt></b></dt>
				1576
				1577	<dd>Perform the <a href="#i_shufflevector">shufflevector
				1578	operation</a> on constants.</dd>
				1579
				1580	<dt><b><tt>OPCODE ( LHS, RHS )</tt></b></dt>
				1581
				1582	<dd>Perform the specified operation of the LHS and RHS constants. OPCODE may
				1583	be any of the <a href="#binaryops">binary</a> or <a href="#bitwiseops">bitwise
				1584	binary</a> operations. The constraints on operands are the same as those for
				1585	the corresponding instruction (e.g. no bitwise operations on floating point
				1586	values are allowed).</dd>
				1587	</dl>
				1588	</div>
				1589
				1590	<!-- *********************************************************************** -->
				1591	<div class="doc_section"> <a name="othervalues">Other Values</a> </div>
				1592	<!-- *********************************************************************** -->
				1593
				1594	<!-- ======================================================================= -->
				1595	<div class="doc_subsection">
				1596	<a name="inlineasm">Inline Assembler Expressions</a>
				1597	</div>
				1598
				1599	<div class="doc_text">
				1600
				1601	<p>
				1602	LLVM supports inline assembler expressions (as opposed to <a href="#moduleasm">
				1603	Module-Level Inline Assembly</a>) through the use of a special value. This
				1604	value represents the inline assembler as a string (containing the instructions
				1605	to emit), a list of operand constraints (stored as a string), and a flag that
				1606	indicates whether or not the inline asm expression has side effects. An example
				1607	inline assembler expression is:
				1608	</p>
				1609
				1610	<div class="doc_code">
				1611	<pre>
				1612	i32 (i32) asm "bswap $0", "=r,r"
				1613	</pre>
				1614	</div>
				1615
				1616	<p>
				1617	Inline assembler expressions may <b>only</b> be used as the callee operand of
				1618	a <a href="#i_call"><tt>call</tt> instruction</a>. Thus, typically we have:
				1619	</p>
				1620
				1621	<div class="doc_code">
				1622	<pre>
				1623	%X = call i32 asm "<a href="#int_bswap">bswap</a> $0", "=r,r"(i32 %Y)
				1624	</pre>
				1625	</div>
				1626
				1627	<p>
				1628	Inline asms with side effects not visible in the constraint list must be marked
				1629	as having side effects. This is done through the use of the
				1630	'<tt>sideeffect</tt>' keyword, like so:
				1631	</p>
				1632
				1633	<div class="doc_code">
				1634	<pre>
				1635	call void asm sideeffect "eieio", ""()
				1636	</pre>
				1637	</div>
				1638
				1639	<p>TODO: The format of the asm and constraints string still need to be
				1640	documented here. Constraints on what can be done (e.g. duplication, moving, etc
				1641	need to be documented).
				1642	</p>
				1643
				1644	</div>
				1645
				1646	<!-- *********************************************************************** -->
				1647	<div class="doc_section"> <a name="instref">Instruction Reference</a> </div>
				1648	<!-- *********************************************************************** -->
				1649
				1650	<div class="doc_text">
				1651
				1652	<p>The LLVM instruction set consists of several different
				1653	classifications of instructions: <a href="#terminators">terminator
				1654	instructions</a>, <a href="#binaryops">binary instructions</a>,
				1655	<a href="#bitwiseops">bitwise binary instructions</a>, <a
				1656	href="#memoryops">memory instructions</a>, and <a href="#otherops">other
				1657	instructions</a>.</p>
				1658
				1659	</div>
				1660
				1661	<!-- ======================================================================= -->
				1662	<div class="doc_subsection"> <a name="terminators">Terminator
				1663	Instructions</a> </div>
				1664
				1665	<div class="doc_text">
				1666
				1667	<p>As mentioned <a href="#functionstructure">previously</a>, every
				1668	basic block in a program ends with a "Terminator" instruction, which
				1669	indicates which block should be executed after the current block is
				1670	finished. These terminator instructions typically yield a '<tt>void</tt>'
				1671	value: they produce control flow, not values (the one exception being
				1672	the '<a href="#i_invoke"><tt>invoke</tt></a>' instruction).</p>
				1673	<p>There are six different terminator instructions: the '<a
				1674	href="#i_ret"><tt>ret</tt></a>' instruction, the '<a href="#i_br"><tt>br</tt></a>'
				1675	instruction, the '<a href="#i_switch"><tt>switch</tt></a>' instruction,
				1676	the '<a href="#i_invoke"><tt>invoke</tt></a>' instruction, the '<a
				1677	href="#i_unwind"><tt>unwind</tt></a>' instruction, and the '<a
				1678	href="#i_unreachable"><tt>unreachable</tt></a>' instruction.</p>
				1679
				1680	</div>
				1681
				1682	<!-- _______________________________________________________________________ -->
				1683	<div class="doc_subsubsection"> <a name="i_ret">'<tt>ret</tt>'
				1684	Instruction</a> </div>
				1685	<div class="doc_text">
				1686	<h5>Syntax:</h5>
				1687	<pre> ret <type> <value> <i>; Return a value from a non-void function</i>
				1688	ret void <i>; Return from void function</i>
				1689	</pre>
				1690	<h5>Overview:</h5>
				1691	<p>The '<tt>ret</tt>' instruction is used to return control flow (and a
				1692	value) from a function back to the caller.</p>
				1693	<p>There are two forms of the '<tt>ret</tt>' instruction: one that
				1694	returns a value and then causes control flow, and one that just causes
				1695	control flow to occur.</p>
				1696	<h5>Arguments:</h5>
				1697	<p>The '<tt>ret</tt>' instruction may return any '<a
				1698	href="#t_firstclass">first class</a>' type. Notice that a function is
				1699	not <a href="#wellformed">well formed</a> if there exists a '<tt>ret</tt>'
				1700	instruction inside of the function that returns a value that does not
				1701	match the return type of the function.</p>
				1702	<h5>Semantics:</h5>
				1703	<p>When the '<tt>ret</tt>' instruction is executed, control flow
				1704	returns back to the calling function's context. If the caller is a "<a
				1705	href="#i_call"><tt>call</tt></a>" instruction, execution continues at
				1706	the instruction after the call. If the caller was an "<a
				1707	href="#i_invoke"><tt>invoke</tt></a>" instruction, execution continues
				1708	at the beginning of the "normal" destination block. If the instruction
				1709	returns a value, that value shall set the call or invoke instruction's
				1710	return value.</p>
				1711	<h5>Example:</h5>
				1712	<pre> ret i32 5 <i>; Return an integer value of 5</i>
				1713	ret void <i>; Return from a void function</i>
				1714	</pre>
				1715	</div>
				1716	<!-- _______________________________________________________________________ -->
				1717	<div class="doc_subsubsection"> <a name="i_br">'<tt>br</tt>' Instruction</a> </div>
				1718	<div class="doc_text">
				1719	<h5>Syntax:</h5>
				1720	<pre> br i1 <cond>, label <iftrue>, label <iffalse><br> br label <dest> <i>; Unconditional branch</i>
				1721	</pre>
				1722	<h5>Overview:</h5>
				1723	<p>The '<tt>br</tt>' instruction is used to cause control flow to
				1724	transfer to a different basic block in the current function. There are
				1725	two forms of this instruction, corresponding to a conditional branch
				1726	and an unconditional branch.</p>
				1727	<h5>Arguments:</h5>
				1728	<p>The conditional branch form of the '<tt>br</tt>' instruction takes a
				1729	single '<tt>i1</tt>' value and two '<tt>label</tt>' values. The
				1730	unconditional form of the '<tt>br</tt>' instruction takes a single
				1731	'<tt>label</tt>' value as a target.</p>
				1732	<h5>Semantics:</h5>
				1733	<p>Upon execution of a conditional '<tt>br</tt>' instruction, the '<tt>i1</tt>'
				1734	argument is evaluated. If the value is <tt>true</tt>, control flows
				1735	to the '<tt>iftrue</tt>' <tt>label</tt> argument. If "cond" is <tt>false</tt>,
				1736	control flows to the '<tt>iffalse</tt>' <tt>label</tt> argument.</p>
				1737	<h5>Example:</h5>
				1738	<pre>Test:<br> %cond = <a href="#i_icmp">icmp</a> eq, i32 %a, %b<br> br i1 %cond, label %IfEqual, label %IfUnequal<br>IfEqual:<br> <a
				1739	href="#i_ret">ret</a> i32 1<br>IfUnequal:<br> <a href="#i_ret">ret</a> i32 0<br></pre>
				1740	</div>
				1741	<!-- _______________________________________________________________________ -->
				1742	<div class="doc_subsubsection">
				1743	<a name="i_switch">'<tt>switch</tt>' Instruction</a>
				1744	</div>
				1745
				1746	<div class="doc_text">
				1747	<h5>Syntax:</h5>
				1748
				1749	<pre>
				1750	switch <intty> <value>, label <defaultdest> [ <intty> <val>, label <dest> ... ]
				1751	</pre>
				1752
				1753	<h5>Overview:</h5>
				1754
				1755	<p>The '<tt>switch</tt>' instruction is used to transfer control flow to one of
				1756	several different places. It is a generalization of the '<tt>br</tt>'
				1757	instruction, allowing a branch to occur to one of many possible
				1758	destinations.</p>
				1759
				1760
				1761	<h5>Arguments:</h5>
				1762
				1763	<p>The '<tt>switch</tt>' instruction uses three parameters: an integer
				1764	comparison value '<tt>value</tt>', a default '<tt>label</tt>' destination, and
				1765	an array of pairs of comparison value constants and '<tt>label</tt>'s. The
				1766	table is not allowed to contain duplicate constant entries.</p>
				1767
				1768	<h5>Semantics:</h5>
				1769
				1770	<p>The <tt>switch</tt> instruction specifies a table of values and
				1771	destinations. When the '<tt>switch</tt>' instruction is executed, this
				1772	table is searched for the given value. If the value is found, control flow is
				1773	transfered to the corresponding destination; otherwise, control flow is
				1774	transfered to the default destination.</p>
				1775
				1776	<h5>Implementation:</h5>
				1777
				1778	<p>Depending on properties of the target machine and the particular
				1779	<tt>switch</tt> instruction, this instruction may be code generated in different
				1780	ways. For example, it could be generated as a series of chained conditional
				1781	branches or with a lookup table.</p>
				1782
				1783	<h5>Example:</h5>
				1784
				1785	<pre>
				1786	<i>; Emulate a conditional br instruction</i>
				1787	%Val = <a href="#i_zext">zext</a> i1 %value to i32
				1788	switch i32 %Val, label %truedest [i32 0, label %falsedest ]
				1789
				1790	<i>; Emulate an unconditional br instruction</i>
				1791	switch i32 0, label %dest [ ]
				1792
				1793	<i>; Implement a jump table:</i>
				1794	switch i32 %val, label %otherwise [ i32 0, label %onzero
				1795	i32 1, label %onone
				1796	i32 2, label %ontwo ]
				1797	</pre>
				1798	</div>
				1799
				1800	<!-- _______________________________________________________________________ -->
				1801	<div class="doc_subsubsection">
				1802	<a name="i_invoke">'<tt>invoke</tt>' Instruction</a>
				1803	</div>
				1804
				1805	<div class="doc_text">
				1806
				1807	<h5>Syntax:</h5>
				1808
				1809	<pre>
				1810	<result> = invoke [<a href="#callingconv">cconv</a>] <ptr to function ty> %<function ptr val>(<function args>)
				1811	to label <normal label> unwind label <exception label>
				1812	</pre>
				1813
				1814	<h5>Overview:</h5>
				1815
				1816	<p>The '<tt>invoke</tt>' instruction causes control to transfer to a specified
				1817	function, with the possibility of control flow transfer to either the
				1818	'<tt>normal</tt>' label or the
				1819	'<tt>exception</tt>' label. If the callee function returns with the
				1820	"<tt><a href="#i_ret">ret</a></tt>" instruction, control flow will return to the
				1821	"normal" label. If the callee (or any indirect callees) returns with the "<a
				1822	href="#i_unwind"><tt>unwind</tt></a>" instruction, control is interrupted and
				1823	continued at the dynamically nearest "exception" label.</p>
				1824
				1825	<h5>Arguments:</h5>
				1826
				1827	<p>This instruction requires several arguments:</p>
				1828
				1829	<ol>
				1830	<li>
				1831	The optional "cconv" marker indicates which <a href="#callingconv">calling
				1832	convention</a> the call should use. If none is specified, the call defaults
				1833	to using C calling conventions.
				1834	</li>
				1835	<li>'<tt>ptr to function ty</tt>': shall be the signature of the pointer to
				1836	function value being invoked. In most cases, this is a direct function
				1837	invocation, but indirect <tt>invoke</tt>s are just as possible, branching off
				1838	an arbitrary pointer to function value.
				1839	</li>
				1840
				1841	<li>'<tt>function ptr val</tt>': An LLVM value containing a pointer to a
				1842	function to be invoked. </li>
				1843
				1844	<li>'<tt>function args</tt>': argument list whose types match the function
				1845	signature argument types. If the function signature indicates the function
				1846	accepts a variable number of arguments, the extra arguments can be
				1847	specified. </li>
				1848
				1849	<li>'<tt>normal label</tt>': the label reached when the called function
				1850	executes a '<tt><a href="#i_ret">ret</a></tt>' instruction. </li>
				1851
				1852	<li>'<tt>exception label</tt>': the label reached when a callee returns with
				1853	the <a href="#i_unwind"><tt>unwind</tt></a> instruction. </li>
				1854
				1855	</ol>
				1856
				1857	<h5>Semantics:</h5>
				1858
				1859	<p>This instruction is designed to operate as a standard '<tt><a
				1860	href="#i_call">call</a></tt>' instruction in most regards. The primary
				1861	difference is that it establishes an association with a label, which is used by
				1862	the runtime library to unwind the stack.</p>
				1863
				1864	<p>This instruction is used in languages with destructors to ensure that proper
				1865	cleanup is performed in the case of either a <tt>longjmp</tt> or a thrown
				1866	exception. Additionally, this is important for implementation of
				1867	'<tt>catch</tt>' clauses in high-level languages that support them.</p>
				1868
				1869	<h5>Example:</h5>
				1870	<pre>
				1871	%retval = invoke i32 %Test(i32 15) to label %Continue
				1872	unwind label %TestCleanup <i>; {i32}:retval set</i>
				1873	%retval = invoke <a href="#callingconv">coldcc</a> i32 %Test(i32 15) to label %Continue
				1874	unwind label %TestCleanup <i>; {i32}:retval set</i>
				1875	</pre>
				1876	</div>
				1877
				1878
				1879	<!-- _______________________________________________________________________ -->
				1880
				1881	<div class="doc_subsubsection"> <a name="i_unwind">'<tt>unwind</tt>'
				1882	Instruction</a> </div>
				1883
				1884	<div class="doc_text">
				1885
				1886	<h5>Syntax:</h5>
				1887	<pre>
				1888	unwind
				1889	</pre>
				1890
				1891	<h5>Overview:</h5>
				1892
				1893	<p>The '<tt>unwind</tt>' instruction unwinds the stack, continuing control flow
				1894	at the first callee in the dynamic call stack which used an <a
				1895	href="#i_invoke"><tt>invoke</tt></a> instruction to perform the call. This is
				1896	primarily used to implement exception handling.</p>
				1897
				1898	<h5>Semantics:</h5>
				1899
				1900	<p>The '<tt>unwind</tt>' intrinsic causes execution of the current function to
				1901	immediately halt. The dynamic call stack is then searched for the first <a
				1902	href="#i_invoke"><tt>invoke</tt></a> instruction on the call stack. Once found,
				1903	execution continues at the "exceptional" destination block specified by the
				1904	<tt>invoke</tt> instruction. If there is no <tt>invoke</tt> instruction in the
				1905	dynamic call chain, undefined behavior results.</p>
				1906	</div>
				1907
				1908	<!-- _______________________________________________________________________ -->
				1909
				1910	<div class="doc_subsubsection"> <a name="i_unreachable">'<tt>unreachable</tt>'
				1911	Instruction</a> </div>
				1912
				1913	<div class="doc_text">
				1914
				1915	<h5>Syntax:</h5>
				1916	<pre>
				1917	unreachable
				1918	</pre>
				1919
				1920	<h5>Overview:</h5>
				1921
				1922	<p>The '<tt>unreachable</tt>' instruction has no defined semantics. This
				1923	instruction is used to inform the optimizer that a particular portion of the
				1924	code is not reachable. This can be used to indicate that the code after a
				1925	no-return function cannot be reached, and other facts.</p>
				1926
				1927	<h5>Semantics:</h5>
				1928
				1929	<p>The '<tt>unreachable</tt>' instruction has no defined semantics.</p>
				1930	</div>
				1931
				1932
				1933
				1934	<!-- ======================================================================= -->
				1935	<div class="doc_subsection"> <a name="binaryops">Binary Operations</a> </div>
				1936	<div class="doc_text">
				1937	<p>Binary operators are used to do most of the computation in a
				1938	program. They require two operands, execute an operation on them, and
				1939	produce a single value. The operands might represent
				1940	multiple data, as is the case with the <a href="#t_vector">vector</a> data type.
				1941	The result value of a binary operator is not
				1942	necessarily the same type as its operands.</p>
				1943	<p>There are several different binary operators:</p>
				1944	</div>
				1945	<!-- _______________________________________________________________________ -->
				1946	<div class="doc_subsubsection"> <a name="i_add">'<tt>add</tt>'
				1947	Instruction</a> </div>
				1948	<div class="doc_text">
				1949	<h5>Syntax:</h5>
				1950	<pre> <result> = add <ty> <var1>, <var2> <i>; yields {ty}:result</i>
				1951	</pre>
				1952	<h5>Overview:</h5>
				1953	<p>The '<tt>add</tt>' instruction returns the sum of its two operands.</p>
				1954	<h5>Arguments:</h5>
				1955	<p>The two arguments to the '<tt>add</tt>' instruction must be either <a
				1956	href="#t_integer">integer</a> or <a href="#t_floating">floating point</a> values.
				1957	This instruction can also take <a href="#t_vector">vector</a> versions of the values.
				1958	Both arguments must have identical types.</p>
				1959	<h5>Semantics:</h5>
				1960	<p>The value produced is the integer or floating point sum of the two
				1961	operands.</p>
				1962	<h5>Example:</h5>
				1963	<pre> <result> = add i32 4, %var <i>; yields {i32}:result = 4 + %var</i>
				1964	</pre>
				1965	</div>
				1966	<!-- _______________________________________________________________________ -->
				1967	<div class="doc_subsubsection"> <a name="i_sub">'<tt>sub</tt>'
				1968	Instruction</a> </div>
				1969	<div class="doc_text">
				1970	<h5>Syntax:</h5>
				1971	<pre> <result> = sub <ty> <var1>, <var2> <i>; yields {ty}:result</i>
				1972	</pre>
				1973	<h5>Overview:</h5>
				1974	<p>The '<tt>sub</tt>' instruction returns the difference of its two
				1975	operands.</p>
				1976	<p>Note that the '<tt>sub</tt>' instruction is used to represent the '<tt>neg</tt>'
				1977	instruction present in most other intermediate representations.</p>
				1978	<h5>Arguments:</h5>
				1979	<p>The two arguments to the '<tt>sub</tt>' instruction must be either <a
				1980	href="#t_integer">integer</a> or <a href="#t_floating">floating point</a>
				1981	values.
				1982	This instruction can also take <a href="#t_vector">vector</a> versions of the values.
				1983	Both arguments must have identical types.</p>
				1984	<h5>Semantics:</h5>
				1985	<p>The value produced is the integer or floating point difference of
				1986	the two operands.</p>
				1987	<h5>Example:</h5>
				1988	<pre>
				1989	<result> = sub i32 4, %var <i>; yields {i32}:result = 4 - %var</i>
				1990	<result> = sub i32 0, %val <i>; yields {i32}:result = -%var</i>
				1991	</pre>
				1992	</div>
				1993	<!-- _______________________________________________________________________ -->
				1994	<div class="doc_subsubsection"> <a name="i_mul">'<tt>mul</tt>'
				1995	Instruction</a> </div>
				1996	<div class="doc_text">
				1997	<h5>Syntax:</h5>
				1998	<pre> <result> = mul <ty> <var1>, <var2> <i>; yields {ty}:result</i>
				1999	</pre>
				2000	<h5>Overview:</h5>
				2001	<p>The '<tt>mul</tt>' instruction returns the product of its two
				2002	operands.</p>
				2003	<h5>Arguments:</h5>
				2004	<p>The two arguments to the '<tt>mul</tt>' instruction must be either <a
				2005	href="#t_integer">integer</a> or <a href="#t_floating">floating point</a>
				2006	values.
				2007	This instruction can also take <a href="#t_vector">vector</a> versions of the values.
				2008	Both arguments must have identical types.</p>
				2009	<h5>Semantics:</h5>
				2010	<p>The value produced is the integer or floating point product of the
				2011	two operands.</p>
				2012	<p>Because the operands are the same width, the result of an integer
				2013	multiplication is the same whether the operands should be deemed unsigned or
				2014	signed.</p>
				2015	<h5>Example:</h5>
				2016	<pre> <result> = mul i32 4, %var <i>; yields {i32}:result = 4 * %var</i>
				2017	</pre>
				2018	</div>
				2019	<!-- _______________________________________________________________________ -->
				2020	<div class="doc_subsubsection"> <a name="i_udiv">'<tt>udiv</tt>' Instruction
				2021	</a></div>
				2022	<div class="doc_text">
				2023	<h5>Syntax:</h5>
				2024	<pre> <result> = udiv <ty> <var1>, <var2> <i>; yields {ty}:result</i>
				2025	</pre>
				2026	<h5>Overview:</h5>
				2027	<p>The '<tt>udiv</tt>' instruction returns the quotient of its two
				2028	operands.</p>
				2029	<h5>Arguments:</h5>
				2030	<p>The two arguments to the '<tt>udiv</tt>' instruction must be
				2031	<a href="#t_integer">integer</a> values. Both arguments must have identical
				2032	types. This instruction can also take <a href="#t_vector">vector</a> versions
				2033	of the values in which case the elements must be integers.</p>
				2034	<h5>Semantics:</h5>
				2035	<p>The value produced is the unsigned integer quotient of the two operands. This
				2036	instruction always performs an unsigned division operation, regardless of
				2037	whether the arguments are unsigned or not.</p>
				2038	<h5>Example:</h5>
				2039	<pre> <result> = udiv i32 4, %var <i>; yields {i32}:result = 4 / %var</i>
				2040	</pre>
				2041	</div>
				2042	<!-- _______________________________________________________________________ -->
				2043	<div class="doc_subsubsection"> <a name="i_sdiv">'<tt>sdiv</tt>' Instruction
				2044	</a> </div>
				2045	<div class="doc_text">
				2046	<h5>Syntax:</h5>
				2047	<pre> <result> = sdiv <ty> <var1>, <var2> <i>; yields {ty}:result</i>
				2048	</pre>
				2049	<h5>Overview:</h5>
				2050	<p>The '<tt>sdiv</tt>' instruction returns the quotient of its two
				2051	operands.</p>
				2052	<h5>Arguments:</h5>
				2053	<p>The two arguments to the '<tt>sdiv</tt>' instruction must be
				2054	<a href="#t_integer">integer</a> values. Both arguments must have identical
				2055	types. This instruction can also take <a href="#t_vector">vector</a> versions
				2056	of the values in which case the elements must be integers.</p>
				2057	<h5>Semantics:</h5>
				2058	<p>The value produced is the signed integer quotient of the two operands. This
				2059	instruction always performs a signed division operation, regardless of whether
				2060	the arguments are signed or not.</p>
				2061	<h5>Example:</h5>
				2062	<pre> <result> = sdiv i32 4, %var <i>; yields {i32}:result = 4 / %var</i>
				2063	</pre>
				2064	</div>
				2065	<!-- _______________________________________________________________________ -->
				2066	<div class="doc_subsubsection"> <a name="i_fdiv">'<tt>fdiv</tt>'
				2067	Instruction</a> </div>
				2068	<div class="doc_text">
				2069	<h5>Syntax:</h5>
				2070	<pre> <result> = fdiv <ty> <var1>, <var2> <i>; yields {ty}:result</i>
				2071	</pre>
				2072	<h5>Overview:</h5>
				2073	<p>The '<tt>fdiv</tt>' instruction returns the quotient of its two
				2074	operands.</p>
				2075	<h5>Arguments:</h5>
				2076	<p>The two arguments to the '<tt>fdiv</tt>' instruction must be
				2077	<a href="#t_floating">floating point</a> values. Both arguments must have
				2078	identical types. This instruction can also take <a href="#t_vector">vector</a>
				2079	versions of floating point values.</p>
				2080	<h5>Semantics:</h5>
				2081	<p>The value produced is the floating point quotient of the two operands.</p>
				2082	<h5>Example:</h5>
				2083	<pre> <result> = fdiv float 4.0, %var <i>; yields {float}:result = 4.0 / %var</i>
				2084	</pre>
				2085	</div>
				2086	<!-- _______________________________________________________________________ -->
				2087	<div class="doc_subsubsection"> <a name="i_urem">'<tt>urem</tt>' Instruction</a>
				2088	</div>
				2089	<div class="doc_text">
				2090	<h5>Syntax:</h5>
				2091	<pre> <result> = urem <ty> <var1>, <var2> <i>; yields {ty}:result</i>
				2092	</pre>
				2093	<h5>Overview:</h5>
				2094	<p>The '<tt>urem</tt>' instruction returns the remainder from the
				2095	unsigned division of its two arguments.</p>
				2096	<h5>Arguments:</h5>
				2097	<p>The two arguments to the '<tt>urem</tt>' instruction must be
				2098	<a href="#t_integer">integer</a> values. Both arguments must have identical
				2099	types.</p>
				2100	<h5>Semantics:</h5>
				2101	<p>This instruction returns the unsigned integer <i>remainder</i> of a division.
				2102	This instruction always performs an unsigned division to get the remainder,
				2103	regardless of whether the arguments are unsigned or not.</p>
				2104	<h5>Example:</h5>
				2105	<pre> <result> = urem i32 4, %var <i>; yields {i32}:result = 4 % %var</i>
				2106	</pre>
				2107
				2108	</div>
				2109	<!-- _______________________________________________________________________ -->
				2110	<div class="doc_subsubsection"> <a name="i_srem">'<tt>srem</tt>'
				2111	Instruction</a> </div>
				2112	<div class="doc_text">
				2113	<h5>Syntax:</h5>
				2114	<pre> <result> = srem <ty> <var1>, <var2> <i>; yields {ty}:result</i>
				2115	</pre>
				2116	<h5>Overview:</h5>
				2117	<p>The '<tt>srem</tt>' instruction returns the remainder from the
				2118	signed division of its two operands.</p>
				2119	<h5>Arguments:</h5>
				2120	<p>The two arguments to the '<tt>srem</tt>' instruction must be
				2121	<a href="#t_integer">integer</a> values. Both arguments must have identical
				2122	types.</p>
				2123	<h5>Semantics:</h5>
				2124	<p>This instruction returns the <i>remainder</i> of a division (where the result
				2125	has the same sign as the dividend, <tt>var1</tt>), not the <i>modulo</i>
				2126	operator (where the result has the same sign as the divisor, <tt>var2</tt>) of
				2127	a value. For more information about the difference, see <a
				2128	href="http://mathforum.org/dr.math/problems/anne.4.28.99.html">The
				2129	Math Forum</a>. For a table of how this is implemented in various languages,
				2130	please see <a href="http://en.wikipedia.org/wiki/Modulo_operation">
				2131	Wikipedia: modulo operation</a>.</p>
				2132	<h5>Example:</h5>
				2133	<pre> <result> = srem i32 4, %var <i>; yields {i32}:result = 4 % %var</i>
				2134	</pre>
				2135
				2136	</div>
				2137	<!-- _______________________________________________________________________ -->
				2138	<div class="doc_subsubsection"> <a name="i_frem">'<tt>frem</tt>'
				2139	Instruction</a> </div>
				2140	<div class="doc_text">
				2141	<h5>Syntax:</h5>
				2142	<pre> <result> = frem <ty> <var1>, <var2> <i>; yields {ty}:result</i>
				2143	</pre>
				2144	<h5>Overview:</h5>
				2145	<p>The '<tt>frem</tt>' instruction returns the remainder from the
				2146	division of its two operands.</p>
				2147	<h5>Arguments:</h5>
				2148	<p>The two arguments to the '<tt>frem</tt>' instruction must be
				2149	<a href="#t_floating">floating point</a> values. Both arguments must have
				2150	identical types.</p>
				2151	<h5>Semantics:</h5>
				2152	<p>This instruction returns the <i>remainder</i> of a division.</p>
				2153	<h5>Example:</h5>
				2154	<pre> <result> = frem float 4.0, %var <i>; yields {float}:result = 4.0 % %var</i>
				2155	</pre>
				2156	</div>
				2157
				2158	<!-- ======================================================================= -->
				2159	<div class="doc_subsection"> <a name="bitwiseops">Bitwise Binary
				2160	Operations</a> </div>
				2161	<div class="doc_text">
				2162	<p>Bitwise binary operators are used to do various forms of
				2163	bit-twiddling in a program. They are generally very efficient
				2164	instructions and can commonly be strength reduced from other
				2165	instructions. They require two operands, execute an operation on them,
				2166	and produce a single value. The resulting value of the bitwise binary
				2167	operators is always the same type as its first operand.</p>
				2168	</div>
				2169
				2170	<!-- _______________________________________________________________________ -->
				2171	<div class="doc_subsubsection"> <a name="i_shl">'<tt>shl</tt>'
				2172	Instruction</a> </div>
				2173	<div class="doc_text">
				2174	<h5>Syntax:</h5>
				2175	<pre> <result> = shl <ty> <var1>, <var2> <i>; yields {ty}:result</i>
				2176	</pre>
				2177	<h5>Overview:</h5>
				2178	<p>The '<tt>shl</tt>' instruction returns the first operand shifted to
				2179	the left a specified number of bits.</p>
				2180	<h5>Arguments:</h5>
				2181	<p>Both arguments to the '<tt>shl</tt>' instruction must be the same <a
				2182	href="#t_integer">integer</a> type.</p>
				2183	<h5>Semantics:</h5>
				2184	<p>The value produced is <tt>var1</tt> * 2<sup><tt>var2</tt></sup>.</p>
				2185	<h5>Example:</h5><pre>
				2186	<result> = shl i32 4, %var <i>; yields {i32}: 4 << %var</i>
				2187	<result> = shl i32 4, 2 <i>; yields {i32}: 16</i>
				2188	<result> = shl i32 1, 10 <i>; yields {i32}: 1024</i>
				2189	</pre>
				2190	</div>
				2191	<!-- _______________________________________________________________________ -->
				2192	<div class="doc_subsubsection"> <a name="i_lshr">'<tt>lshr</tt>'
				2193	Instruction</a> </div>
				2194	<div class="doc_text">
				2195	<h5>Syntax:</h5>
				2196	<pre> <result> = lshr <ty> <var1>, <var2> <i>; yields {ty}:result</i>
				2197	</pre>
				2198
				2199	<h5>Overview:</h5>
				2200	<p>The '<tt>lshr</tt>' instruction (logical shift right) returns the first
				2201	operand shifted to the right a specified number of bits with zero fill.</p>
				2202
				2203	<h5>Arguments:</h5>
				2204	<p>Both arguments to the '<tt>lshr</tt>' instruction must be the same
				2205	<a href="#t_integer">integer</a> type.</p>
				2206
				2207	<h5>Semantics:</h5>
				2208	<p>This instruction always performs a logical shift right operation. The most
				2209	significant bits of the result will be filled with zero bits after the
				2210	shift.</p>
				2211
				2212	<h5>Example:</h5>
				2213	<pre>
				2214	<result> = lshr i32 4, 1 <i>; yields {i32}:result = 2</i>
				2215	<result> = lshr i32 4, 2 <i>; yields {i32}:result = 1</i>
				2216	<result> = lshr i8 4, 3 <i>; yields {i8}:result = 0</i>
				2217	<result> = lshr i8 -2, 1 <i>; yields {i8}:result = 0x7FFFFFFF </i>
				2218	</pre>
				2219	</div>
				2220
				2221	<!-- _______________________________________________________________________ -->
				2222	<div class="doc_subsubsection"> <a name="i_ashr">'<tt>ashr</tt>'
				2223	Instruction</a> </div>
				2224	<div class="doc_text">
				2225
				2226	<h5>Syntax:</h5>
				2227	<pre> <result> = ashr <ty> <var1>, <var2> <i>; yields {ty}:result</i>
				2228	</pre>
				2229
				2230	<h5>Overview:</h5>
				2231	<p>The '<tt>ashr</tt>' instruction (arithmetic shift right) returns the first
				2232	operand shifted to the right a specified number of bits with sign extension.</p>
				2233
				2234	<h5>Arguments:</h5>
				2235	<p>Both arguments to the '<tt>ashr</tt>' instruction must be the same
				2236	<a href="#t_integer">integer</a> type.</p>
				2237
				2238	<h5>Semantics:</h5>
				2239	<p>This instruction always performs an arithmetic shift right operation,
				2240	The most significant bits of the result will be filled with the sign bit
				2241	of <tt>var1</tt>.</p>
				2242
				2243	<h5>Example:</h5>
				2244	<pre>
				2245	<result> = ashr i32 4, 1 <i>; yields {i32}:result = 2</i>
				2246	<result> = ashr i32 4, 2 <i>; yields {i32}:result = 1</i>
				2247	<result> = ashr i8 4, 3 <i>; yields {i8}:result = 0</i>
				2248	<result> = ashr i8 -2, 1 <i>; yields {i8}:result = -1</i>
				2249	</pre>
				2250	</div>
				2251
				2252	<!-- _______________________________________________________________________ -->
				2253	<div class="doc_subsubsection"> <a name="i_and">'<tt>and</tt>'
				2254	Instruction</a> </div>
				2255	<div class="doc_text">
				2256	<h5>Syntax:</h5>
				2257	<pre> <result> = and <ty> <var1>, <var2> <i>; yields {ty}:result</i>
				2258	</pre>
				2259	<h5>Overview:</h5>
				2260	<p>The '<tt>and</tt>' instruction returns the bitwise logical and of
				2261	its two operands.</p>
				2262	<h5>Arguments:</h5>
				2263	<p>The two arguments to the '<tt>and</tt>' instruction must be <a
				2264	href="#t_integer">integer</a> values. Both arguments must have
				2265	identical types.</p>
				2266	<h5>Semantics:</h5>
				2267	<p>The truth table used for the '<tt>and</tt>' instruction is:</p>
				2268	<p> </p>
				2269	<div style="align: center">
				2270	<table border="1" cellspacing="0" cellpadding="4">
				2271	<tbody>
				2272	<tr>
				2273	<td>In0</td>
				2274	<td>In1</td>
				2275	<td>Out</td>
				2276	</tr>
				2277	<tr>
				2278	<td>0</td>
				2279	<td>0</td>
				2280	<td>0</td>
				2281	</tr>
				2282	<tr>
				2283	<td>0</td>
				2284	<td>1</td>
				2285	<td>0</td>
				2286	</tr>
				2287	<tr>
				2288	<td>1</td>
				2289	<td>0</td>
				2290	<td>0</td>
				2291	</tr>
				2292	<tr>
				2293	<td>1</td>
				2294	<td>1</td>
				2295	<td>1</td>
				2296	</tr>
				2297	</tbody>
				2298	</table>
				2299	</div>
				2300	<h5>Example:</h5>
				2301	<pre> <result> = and i32 4, %var <i>; yields {i32}:result = 4 & %var</i>
				2302	<result> = and i32 15, 40 <i>; yields {i32}:result = 8</i>
				2303	<result> = and i32 4, 8 <i>; yields {i32}:result = 0</i>
				2304	</pre>
				2305	</div>
				2306	<!-- _______________________________________________________________________ -->
				2307	<div class="doc_subsubsection"> <a name="i_or">'<tt>or</tt>' Instruction</a> </div>
				2308	<div class="doc_text">
				2309	<h5>Syntax:</h5>
				2310	<pre> <result> = or <ty> <var1>, <var2> <i>; yields {ty}:result</i>
				2311	</pre>
				2312	<h5>Overview:</h5>
				2313	<p>The '<tt>or</tt>' instruction returns the bitwise logical inclusive
				2314	or of its two operands.</p>
				2315	<h5>Arguments:</h5>
				2316	<p>The two arguments to the '<tt>or</tt>' instruction must be <a
				2317	href="#t_integer">integer</a> values. Both arguments must have
				2318	identical types.</p>
				2319	<h5>Semantics:</h5>
				2320	<p>The truth table used for the '<tt>or</tt>' instruction is:</p>
				2321	<p> </p>
				2322	<div style="align: center">
				2323	<table border="1" cellspacing="0" cellpadding="4">
				2324	<tbody>
				2325	<tr>
				2326	<td>In0</td>
				2327	<td>In1</td>
				2328	<td>Out</td>
				2329	</tr>
				2330	<tr>
				2331	<td>0</td>
				2332	<td>0</td>
				2333	<td>0</td>
				2334	</tr>
				2335	<tr>
				2336	<td>0</td>
				2337	<td>1</td>
				2338	<td>1</td>
				2339	</tr>
				2340	<tr>
				2341	<td>1</td>
				2342	<td>0</td>
				2343	<td>1</td>
				2344	</tr>
				2345	<tr>
				2346	<td>1</td>
				2347	<td>1</td>
				2348	<td>1</td>
				2349	</tr>
				2350	</tbody>
				2351	</table>
				2352	</div>
				2353	<h5>Example:</h5>
				2354	<pre> <result> = or i32 4, %var <i>; yields {i32}:result = 4 \| %var</i>
				2355	<result> = or i32 15, 40 <i>; yields {i32}:result = 47</i>
				2356	<result> = or i32 4, 8 <i>; yields {i32}:result = 12</i>
				2357	</pre>
				2358	</div>
				2359	<!-- _______________________________________________________________________ -->
				2360	<div class="doc_subsubsection"> <a name="i_xor">'<tt>xor</tt>'
				2361	Instruction</a> </div>
				2362	<div class="doc_text">
				2363	<h5>Syntax:</h5>
				2364	<pre> <result> = xor <ty> <var1>, <var2> <i>; yields {ty}:result</i>
				2365	</pre>
				2366	<h5>Overview:</h5>
				2367	<p>The '<tt>xor</tt>' instruction returns the bitwise logical exclusive
				2368	or of its two operands. The <tt>xor</tt> is used to implement the
				2369	"one's complement" operation, which is the "~" operator in C.</p>
				2370	<h5>Arguments:</h5>
				2371	<p>The two arguments to the '<tt>xor</tt>' instruction must be <a
				2372	href="#t_integer">integer</a> values. Both arguments must have
				2373	identical types.</p>
				2374	<h5>Semantics:</h5>
				2375	<p>The truth table used for the '<tt>xor</tt>' instruction is:</p>
				2376	<p> </p>
				2377	<div style="align: center">
				2378	<table border="1" cellspacing="0" cellpadding="4">
				2379	<tbody>
				2380	<tr>
				2381	<td>In0</td>
				2382	<td>In1</td>
				2383	<td>Out</td>
				2384	</tr>
				2385	<tr>
				2386	<td>0</td>
				2387	<td>0</td>
				2388	<td>0</td>
				2389	</tr>
				2390	<tr>
				2391	<td>0</td>
				2392	<td>1</td>
				2393	<td>1</td>
				2394	</tr>
				2395	<tr>
				2396	<td>1</td>
				2397	<td>0</td>
				2398	<td>1</td>
				2399	</tr>
				2400	<tr>
				2401	<td>1</td>
				2402	<td>1</td>
				2403	<td>0</td>
				2404	</tr>
				2405	</tbody>
				2406	</table>
				2407	</div>
				2408	<p> </p>
				2409	<h5>Example:</h5>
				2410	<pre> <result> = xor i32 4, %var <i>; yields {i32}:result = 4 ^ %var</i>
				2411	<result> = xor i32 15, 40 <i>; yields {i32}:result = 39</i>
				2412	<result> = xor i32 4, 8 <i>; yields {i32}:result = 12</i>
				2413	<result> = xor i32 %V, -1 <i>; yields {i32}:result = ~%V</i>
				2414	</pre>
				2415	</div>
				2416
				2417	<!-- ======================================================================= -->
				2418	<div class="doc_subsection">
				2419	<a name="vectorops">Vector Operations</a>
				2420	</div>
				2421
				2422	<div class="doc_text">
				2423
				2424	<p>LLVM supports several instructions to represent vector operations in a
				2425	target-independent manner. These instructions cover the element-access and
				2426	vector-specific operations needed to process vectors effectively. While LLVM
				2427	does directly support these vector operations, many sophisticated algorithms
				2428	will want to use target-specific intrinsics to take full advantage of a specific
				2429	target.</p>
				2430
				2431	</div>
				2432
				2433	<!-- _______________________________________________________________________ -->
				2434	<div class="doc_subsubsection">
				2435	<a name="i_extractelement">'<tt>extractelement</tt>' Instruction</a>
				2436	</div>
				2437
				2438	<div class="doc_text">
				2439
				2440	<h5>Syntax:</h5>
				2441
				2442	<pre>
				2443	<result> = extractelement <n x <ty>> <val>, i32 <idx> <i>; yields <ty></i>
				2444	</pre>
				2445
				2446	<h5>Overview:</h5>
				2447
				2448	<p>
				2449	The '<tt>extractelement</tt>' instruction extracts a single scalar
				2450	element from a vector at a specified index.
				2451	</p>
				2452
				2453
				2454	<h5>Arguments:</h5>
				2455
				2456	<p>
				2457	The first operand of an '<tt>extractelement</tt>' instruction is a
				2458	value of <a href="#t_vector">vector</a> type. The second operand is
				2459	an index indicating the position from which to extract the element.
				2460	The index may be a variable.</p>
				2461
				2462	<h5>Semantics:</h5>
				2463
				2464	<p>
				2465	The result is a scalar of the same type as the element type of
				2466	<tt>val</tt>. Its value is the value at position <tt>idx</tt> of
				2467	<tt>val</tt>. If <tt>idx</tt> exceeds the length of <tt>val</tt>, the
				2468	results are undefined.
				2469	</p>
				2470
				2471	<h5>Example:</h5>
				2472
				2473	<pre>
				2474	%result = extractelement <4 x i32> %vec, i32 0 <i>; yields i32</i>
				2475	</pre>
				2476	</div>
				2477
				2478
				2479	<!-- _______________________________________________________________________ -->
				2480	<div class="doc_subsubsection">
				2481	<a name="i_insertelement">'<tt>insertelement</tt>' Instruction</a>
				2482	</div>
				2483
				2484	<div class="doc_text">
				2485
				2486	<h5>Syntax:</h5>
				2487
				2488	<pre>
				2489	<result> = insertelement <n x <ty>> <val>, <ty> <elt&gt, i32 <idx> <i>; yields <n x <ty>></i>
				2490	</pre>
				2491
				2492	<h5>Overview:</h5>
				2493
				2494	<p>
				2495	The '<tt>insertelement</tt>' instruction inserts a scalar
				2496	element into a vector at a specified index.
				2497	</p>
				2498
				2499
				2500	<h5>Arguments:</h5>
				2501
				2502	<p>
				2503	The first operand of an '<tt>insertelement</tt>' instruction is a
				2504	value of <a href="#t_vector">vector</a> type. The second operand is a
				2505	scalar value whose type must equal the element type of the first
				2506	operand. The third operand is an index indicating the position at
				2507	which to insert the value. The index may be a variable.</p>
				2508
				2509	<h5>Semantics:</h5>
				2510
				2511	<p>
				2512	The result is a vector of the same type as <tt>val</tt>. Its
				2513	element values are those of <tt>val</tt> except at position
				2514	<tt>idx</tt>, where it gets the value <tt>elt</tt>. If <tt>idx</tt>
				2515	exceeds the length of <tt>val</tt>, the results are undefined.
				2516	</p>
				2517
				2518	<h5>Example:</h5>
				2519
				2520	<pre>
				2521	%result = insertelement <4 x i32> %vec, i32 1, i32 0 <i>; yields <4 x i32></i>
				2522	</pre>
				2523	</div>
				2524
				2525	<!-- _______________________________________________________________________ -->
				2526	<div class="doc_subsubsection">
				2527	<a name="i_shufflevector">'<tt>shufflevector</tt>' Instruction</a>
				2528	</div>
				2529
				2530	<div class="doc_text">
				2531
				2532	<h5>Syntax:</h5>
				2533
				2534	<pre>
				2535	<result> = shufflevector <n x <ty>> <v1>, <n x <ty>> <v2>, <n x i32> <mask> <i>; yields <n x <ty>></i>
				2536	</pre>
				2537
				2538	<h5>Overview:</h5>
				2539
				2540	<p>
				2541	The '<tt>shufflevector</tt>' instruction constructs a permutation of elements
				2542	from two input vectors, returning a vector of the same type.
				2543	</p>
				2544
				2545	<h5>Arguments:</h5>
				2546
				2547	<p>
				2548	The first two operands of a '<tt>shufflevector</tt>' instruction are vectors
				2549	with types that match each other and types that match the result of the
				2550	instruction. The third argument is a shuffle mask, which has the same number
				2551	of elements as the other vector type, but whose element type is always 'i32'.
				2552	</p>
				2553
				2554	<p>
				2555	The shuffle mask operand is required to be a constant vector with either
				2556	constant integer or undef values.
				2557	</p>
				2558
				2559	<h5>Semantics:</h5>
				2560
				2561	<p>
				2562	The elements of the two input vectors are numbered from left to right across
				2563	both of the vectors. The shuffle mask operand specifies, for each element of
				2564	the result vector, which element of the two input registers the result element
				2565	gets. The element selector may be undef (meaning "don't care") and the second
				2566	operand may be undef if performing a shuffle from only one vector.
				2567	</p>
				2568
				2569	<h5>Example:</h5>
				2570
				2571	<pre>
				2572	%result = shufflevector <4 x i32> %v1, <4 x i32> %v2,
				2573	<4 x i32> <i32 0, i32 4, i32 1, i32 5> <i>; yields <4 x i32></i>
				2574	%result = shufflevector <4 x i32> %v1, <4 x i32> undef,
				2575	<4 x i32> <i32 0, i32 1, i32 2, i32 3> <i>; yields <4 x i32></i> - Identity shuffle.
				2576	</pre>
				2577	</div>
				2578
				2579
				2580	<!-- ======================================================================= -->
				2581	<div class="doc_subsection">
				2582	<a name="memoryops">Memory Access and Addressing Operations</a>
				2583	</div>
				2584
				2585	<div class="doc_text">
				2586
				2587	<p>A key design point of an SSA-based representation is how it
				2588	represents memory. In LLVM, no memory locations are in SSA form, which
				2589	makes things very simple. This section describes how to read, write,
				2590	allocate, and free memory in LLVM.</p>
				2591
				2592	</div>
				2593
				2594	<!-- _______________________________________________________________________ -->
				2595	<div class="doc_subsubsection">
				2596	<a name="i_malloc">'<tt>malloc</tt>' Instruction</a>
				2597	</div>
				2598
				2599	<div class="doc_text">
				2600
				2601	<h5>Syntax:</h5>
				2602
				2603	<pre>
				2604	<result> = malloc <type>[, i32 <NumElements>][, align <alignment>] <i>; yields {type*}:result</i>
				2605	</pre>
				2606
				2607	<h5>Overview:</h5>
				2608
				2609	<p>The '<tt>malloc</tt>' instruction allocates memory from the system
				2610	heap and returns a pointer to it.</p>
				2611
				2612	<h5>Arguments:</h5>
				2613
				2614	<p>The '<tt>malloc</tt>' instruction allocates
				2615	<tt>sizeof(<type>)*NumElements</tt>
				2616	bytes of memory from the operating system and returns a pointer of the
				2617	appropriate type to the program. If "NumElements" is specified, it is the
				2618	number of elements allocated. If an alignment is specified, the value result
				2619	of the allocation is guaranteed to be aligned to at least that boundary. If
				2620	not specified, or if zero, the target can choose to align the allocation on any
				2621	convenient boundary.</p>
				2622
				2623	<p>'<tt>type</tt>' must be a sized type.</p>
				2624
				2625	<h5>Semantics:</h5>
				2626
				2627	<p>Memory is allocated using the system "<tt>malloc</tt>" function, and
				2628	a pointer is returned.</p>
				2629
				2630	<h5>Example:</h5>
				2631
				2632	<pre>
				2633	%array = malloc [4 x i8 ] <i>; yields {[%4 x i8]*}:array</i>
				2634
				2635	%size = <a href="#i_add">add</a> i32 2, 2 <i>; yields {i32}:size = i32 4</i>
				2636	%array1 = malloc i8, i32 4 <i>; yields {i8*}:array1</i>
				2637	%array2 = malloc [12 x i8], i32 %size <i>; yields {[12 x i8]*}:array2</i>
				2638	%array3 = malloc i32, i32 4, align 1024 <i>; yields {i32*}:array3</i>
				2639	%array4 = malloc i32, align 1024 <i>; yields {i32*}:array4</i>
				2640	</pre>
				2641	</div>
				2642
				2643	<!-- _______________________________________________________________________ -->
				2644	<div class="doc_subsubsection">
				2645	<a name="i_free">'<tt>free</tt>' Instruction</a>
				2646	</div>
				2647
				2648	<div class="doc_text">
				2649
				2650	<h5>Syntax:</h5>
				2651
				2652	<pre>
				2653	free <type> <value> <i>; yields {void}</i>
				2654	</pre>
				2655
				2656	<h5>Overview:</h5>
				2657
				2658	<p>The '<tt>free</tt>' instruction returns memory back to the unused
				2659	memory heap to be reallocated in the future.</p>
				2660
				2661	<h5>Arguments:</h5>
				2662
				2663	<p>'<tt>value</tt>' shall be a pointer value that points to a value
				2664	that was allocated with the '<tt><a href="#i_malloc">malloc</a></tt>'
				2665	instruction.</p>
				2666
				2667	<h5>Semantics:</h5>
				2668
				2669	<p>Access to the memory pointed to by the pointer is no longer defined
				2670	after this instruction executes.</p>
				2671
				2672	<h5>Example:</h5>
				2673
				2674	<pre>
				2675	%array = <a href="#i_malloc">malloc</a> [4 x i8] <i>; yields {[4 x i8]*}:array</i>
				2676	free [4 x i8]* %array
				2677	</pre>
				2678	</div>
				2679
				2680	<!-- _______________________________________________________________________ -->
				2681	<div class="doc_subsubsection">
				2682	<a name="i_alloca">'<tt>alloca</tt>' Instruction</a>
				2683	</div>
				2684
				2685	<div class="doc_text">
				2686
				2687	<h5>Syntax:</h5>
				2688
				2689	<pre>
				2690	<result> = alloca <type>[, i32 <NumElements>][, align <alignment>] <i>; yields {type*}:result</i>
				2691	</pre>
				2692
				2693	<h5>Overview:</h5>
				2694
				2695	<p>The '<tt>alloca</tt>' instruction allocates memory on the stack frame of the
				2696	currently executing function, to be automatically released when this function
				2697	returns to its caller.</p>
				2698
				2699	<h5>Arguments:</h5>
				2700
				2701	<p>The '<tt>alloca</tt>' instruction allocates <tt>sizeof(<type>)*NumElements</tt>
				2702	bytes of memory on the runtime stack, returning a pointer of the
				2703	appropriate type to the program. If "NumElements" is specified, it is the
				2704	number of elements allocated. If an alignment is specified, the value result
				2705	of the allocation is guaranteed to be aligned to at least that boundary. If
				2706	not specified, or if zero, the target can choose to align the allocation on any
				2707	convenient boundary.</p>
				2708
				2709	<p>'<tt>type</tt>' may be any sized type.</p>
				2710
				2711	<h5>Semantics:</h5>
				2712
				2713	<p>Memory is allocated; a pointer is returned. '<tt>alloca</tt>'d
				2714	memory is automatically released when the function returns. The '<tt>alloca</tt>'
				2715	instruction is commonly used to represent automatic variables that must
				2716	have an address available. When the function returns (either with the <tt><a
				2717	href="#i_ret">ret</a></tt> or <tt><a href="#i_unwind">unwind</a></tt>
				2718	instructions), the memory is reclaimed.</p>
				2719
				2720	<h5>Example:</h5>
				2721
				2722	<pre>
				2723	%ptr = alloca i32 <i>; yields {i32*}:ptr</i>
				2724	%ptr = alloca i32, i32 4 <i>; yields {i32*}:ptr</i>
				2725	%ptr = alloca i32, i32 4, align 1024 <i>; yields {i32*}:ptr</i>
				2726	%ptr = alloca i32, align 1024 <i>; yields {i32*}:ptr</i>
				2727	</pre>
				2728	</div>
				2729
				2730	<!-- _______________________________________________________________________ -->
				2731	<div class="doc_subsubsection"> <a name="i_load">'<tt>load</tt>'
				2732	Instruction</a> </div>
				2733	<div class="doc_text">
				2734	<h5>Syntax:</h5>
				2735	<pre> <result> = load <ty>* <pointer>[, align <alignment>]<br> <result> = volatile load <ty>* <pointer>[, align <alignment>]<br></pre>
				2736	<h5>Overview:</h5>
				2737	<p>The '<tt>load</tt>' instruction is used to read from memory.</p>
				2738	<h5>Arguments:</h5>
				2739	<p>The argument to the '<tt>load</tt>' instruction specifies the memory
				2740	address from which to load. The pointer must point to a <a
				2741	href="#t_firstclass">first class</a> type. If the <tt>load</tt> is
				2742	marked as <tt>volatile</tt>, then the optimizer is not allowed to modify
				2743	the number or order of execution of this <tt>load</tt> with other
				2744	volatile <tt>load</tt> and <tt><a href="#i_store">store</a></tt>
				2745	instructions. </p>
				2746	<h5>Semantics:</h5>
				2747	<p>The location of memory pointed to is loaded.</p>
				2748	<h5>Examples:</h5>
				2749	<pre> %ptr = <a href="#i_alloca">alloca</a> i32 <i>; yields {i32*}:ptr</i>
				2750	<a
				2751	href="#i_store">store</a> i32 3, i32* %ptr <i>; yields {void}</i>
				2752	%val = load i32* %ptr <i>; yields {i32}:val = i32 3</i>
				2753	</pre>
				2754	</div>
				2755	<!-- _______________________________________________________________________ -->
				2756	<div class="doc_subsubsection"> <a name="i_store">'<tt>store</tt>'
				2757	Instruction</a> </div>
				2758	<div class="doc_text">
				2759	<h5>Syntax:</h5>
				2760	<pre> store <ty> <value>, <ty>* <pointer>[, align <alignment>] <i>; yields {void}</i>
				2761	volatile store <ty> <value>, <ty>* <pointer>[, align <alignment>] <i>; yields {void}</i>
				2762	</pre>
				2763	<h5>Overview:</h5>
				2764	<p>The '<tt>store</tt>' instruction is used to write to memory.</p>
				2765	<h5>Arguments:</h5>
				2766	<p>There are two arguments to the '<tt>store</tt>' instruction: a value
				2767	to store and an address at which to store it. The type of the '<tt><pointer></tt>'
				2768	operand must be a pointer to the type of the '<tt><value></tt>'
				2769	operand. If the <tt>store</tt> is marked as <tt>volatile</tt>, then the
				2770	optimizer is not allowed to modify the number or order of execution of
				2771	this <tt>store</tt> with other volatile <tt>load</tt> and <tt><a
				2772	href="#i_store">store</a></tt> instructions.</p>
				2773	<h5>Semantics:</h5>
				2774	<p>The contents of memory are updated to contain '<tt><value></tt>'
				2775	at the location specified by the '<tt><pointer></tt>' operand.</p>
				2776	<h5>Example:</h5>
				2777	<pre> %ptr = <a href="#i_alloca">alloca</a> i32 <i>; yields {i32*}:ptr</i>
				2778	<a
				2779	href="#i_store">store</a> i32 3, i32* %ptr <i>; yields {void}</i>
				2780	%val = load i32* %ptr <i>; yields {i32}:val = i32 3</i>
				2781	</pre>
				2782	</div>
				2783
				2784	<!-- _______________________________________________________________________ -->
				2785	<div class="doc_subsubsection">
				2786	<a name="i_getelementptr">'<tt>getelementptr</tt>' Instruction</a>
				2787	</div>
				2788
				2789	<div class="doc_text">
				2790	<h5>Syntax:</h5>
				2791	<pre>
				2792	<result> = getelementptr <ty>* <ptrval>{, <ty> <idx>}*
				2793	</pre>
				2794
				2795	<h5>Overview:</h5>
				2796
				2797	<p>
				2798	The '<tt>getelementptr</tt>' instruction is used to get the address of a
				2799	subelement of an aggregate data structure.</p>
				2800
				2801	<h5>Arguments:</h5>
				2802
				2803	<p>This instruction takes a list of integer operands that indicate what
				2804	elements of the aggregate object to index to. The actual types of the arguments
				2805	provided depend on the type of the first pointer argument. The
				2806	'<tt>getelementptr</tt>' instruction is used to index down through the type
				2807	levels of a structure or to a specific index in an array. When indexing into a
				2808	structure, only <tt>i32</tt> integer constants are allowed. When indexing
				2809	into an array or pointer, only integers of 32 or 64 bits are allowed, and will
				2810	be sign extended to 64-bit values.</p>
				2811
				2812	<p>For example, let's consider a C code fragment and how it gets
				2813	compiled to LLVM:</p>
				2814
				2815	<div class="doc_code">
				2816	<pre>
				2817	struct RT {
				2818	char A;
				2819	int B[10][20];
				2820	char C;
				2821	};
				2822	struct ST {
				2823	int X;
				2824	double Y;
				2825	struct RT Z;
				2826	};
				2827
				2828	int foo(struct ST s) {
				2829	return &s[1].Z.B[5][13];
				2830	}
				2831	</pre>
				2832	</div>
				2833
				2834	<p>The LLVM code generated by the GCC frontend is:</p>
				2835
				2836	<div class="doc_code">
				2837	<pre>
				2838	%RT = type { i8 , [10 x [20 x i32]], i8 }
				2839	%ST = type { i32, double, %RT }
				2840
				2841	define i32* %foo(%ST* %s) {
				2842	entry:
				2843	%reg = getelementptr %ST* %s, i32 1, i32 2, i32 1, i32 5, i32 13
				2844	ret i32* %reg
				2845	}
				2846	</pre>
				2847	</div>
				2848
				2849	<h5>Semantics:</h5>
				2850
				2851	<p>The index types specified for the '<tt>getelementptr</tt>' instruction depend
				2852	on the pointer type that is being indexed into. <a href="#t_pointer">Pointer</a>
				2853	and <a href="#t_array">array</a> types can use a 32-bit or 64-bit
				2854	<a href="#t_integer">integer</a> type but the value will always be sign extended
				2855	to 64-bits. <a href="#t_struct">Structure</a> types require <tt>i32</tt>
				2856	<b>constants</b>.</p>
				2857
				2858	<p>In the example above, the first index is indexing into the '<tt>%ST*</tt>'
				2859	type, which is a pointer, yielding a '<tt>%ST</tt>' = '<tt>{ i32, double, %RT
				2860	}</tt>' type, a structure. The second index indexes into the third element of
				2861	the structure, yielding a '<tt>%RT</tt>' = '<tt>{ i8 , [10 x [20 x i32]],
				2862	i8 }</tt>' type, another structure. The third index indexes into the second
				2863	element of the structure, yielding a '<tt>[10 x [20 x i32]]</tt>' type, an
				2864	array. The two dimensions of the array are subscripted into, yielding an
				2865	'<tt>i32</tt>' type. The '<tt>getelementptr</tt>' instruction returns a pointer
				2866	to this element, thus computing a value of '<tt>i32*</tt>' type.</p>
				2867
				2868	<p>Note that it is perfectly legal to index partially through a
				2869	structure, returning a pointer to an inner element. Because of this,
				2870	the LLVM code for the given testcase is equivalent to:</p>
				2871
				2872	<pre>
				2873	define i32* %foo(%ST* %s) {
				2874	%t1 = getelementptr %ST* %s, i32 1 <i>; yields %ST*:%t1</i>
				2875	%t2 = getelementptr %ST* %t1, i32 0, i32 2 <i>; yields %RT*:%t2</i>
				2876	%t3 = getelementptr %RT* %t2, i32 0, i32 1 <i>; yields [10 x [20 x i32]]*:%t3</i>
				2877	%t4 = getelementptr [10 x [20 x i32]]* %t3, i32 0, i32 5 <i>; yields [20 x i32]*:%t4</i>
				2878	%t5 = getelementptr [20 x i32]* %t4, i32 0, i32 13 <i>; yields i32*:%t5</i>
				2879	ret i32* %t5
				2880	}
				2881	</pre>
				2882
				2883	<p>Note that it is undefined to access an array out of bounds: array and
				2884	pointer indexes must always be within the defined bounds of the array type.
				2885	The one exception for this rules is zero length arrays. These arrays are
				2886	defined to be accessible as variable length arrays, which requires access
				2887	beyond the zero'th element.</p>
				2888
				2889	<p>The getelementptr instruction is often confusing. For some more insight
				2890	into how it works, see <a href="GetElementPtr.html">the getelementptr
				2891	FAQ</a>.</p>
				2892
				2893	<h5>Example:</h5>
				2894
				2895	<pre>
				2896	<i>; yields [12 x i8]*:aptr</i>
				2897	%aptr = getelementptr {i32, [12 x i8]}* %sptr, i64 0, i32 1
				2898	</pre>
				2899	</div>
				2900
				2901	<!-- ======================================================================= -->
				2902	<div class="doc_subsection"> <a name="convertops">Conversion Operations</a>
				2903	</div>
				2904	<div class="doc_text">
				2905	<p>The instructions in this category are the conversion instructions (casting)
				2906	which all take a single operand and a type. They perform various bit conversions
				2907	on the operand.</p>
				2908	</div>
				2909
				2910	<!-- _______________________________________________________________________ -->
				2911	<div class="doc_subsubsection">
				2912	<a name="i_trunc">'<tt>trunc .. to</tt>' Instruction</a>
				2913	</div>
				2914	<div class="doc_text">
				2915
				2916	<h5>Syntax:</h5>
				2917	<pre>
				2918	<result> = trunc <ty> <value> to <ty2> <i>; yields ty2</i>
				2919	</pre>
				2920
				2921	<h5>Overview:</h5>
				2922	<p>
				2923	The '<tt>trunc</tt>' instruction truncates its operand to the type <tt>ty2</tt>.
				2924	</p>
				2925
				2926	<h5>Arguments:</h5>
				2927	<p>
				2928	The '<tt>trunc</tt>' instruction takes a <tt>value</tt> to trunc, which must
				2929	be an <a href="#t_integer">integer</a> type, and a type that specifies the size
				2930	and type of the result, which must be an <a href="#t_integer">integer</a>
				2931	type. The bit size of <tt>value</tt> must be larger than the bit size of
				2932	<tt>ty2</tt>. Equal sized types are not allowed.</p>
				2933
				2934	<h5>Semantics:</h5>
				2935	<p>
				2936	The '<tt>trunc</tt>' instruction truncates the high order bits in <tt>value</tt>
				2937	and converts the remaining bits to <tt>ty2</tt>. Since the source size must be
				2938	larger than the destination size, <tt>trunc</tt> cannot be a <i>no-op cast</i>.
				2939	It will always truncate bits.</p>
				2940
				2941	<h5>Example:</h5>
				2942	<pre>
				2943	%X = trunc i32 257 to i8 <i>; yields i8:1</i>
				2944	%Y = trunc i32 123 to i1 <i>; yields i1:true</i>
				2945	%Y = trunc i32 122 to i1 <i>; yields i1:false</i>
				2946	</pre>
				2947	</div>
				2948
				2949	<!-- _______________________________________________________________________ -->
				2950	<div class="doc_subsubsection">
				2951	<a name="i_zext">'<tt>zext .. to</tt>' Instruction</a>
				2952	</div>
				2953	<div class="doc_text">
				2954
				2955	<h5>Syntax:</h5>
				2956	<pre>
				2957	<result> = zext <ty> <value> to <ty2> <i>; yields ty2</i>
				2958	</pre>
				2959
				2960	<h5>Overview:</h5>
				2961	<p>The '<tt>zext</tt>' instruction zero extends its operand to type
				2962	<tt>ty2</tt>.</p>
				2963
				2964
				2965	<h5>Arguments:</h5>
				2966	<p>The '<tt>zext</tt>' instruction takes a value to cast, which must be of
				2967	<a href="#t_integer">integer</a> type, and a type to cast it to, which must
				2968	also be of <a href="#t_integer">integer</a> type. The bit size of the
				2969	<tt>value</tt> must be smaller than the bit size of the destination type,
				2970	<tt>ty2</tt>.</p>
				2971
				2972	<h5>Semantics:</h5>
				2973	<p>The <tt>zext</tt> fills the high order bits of the <tt>value</tt> with zero
				2974	bits until it reaches the size of the destination type, <tt>ty2</tt>.</p>
				2975
				2976	<p>When zero extending from i1, the result will always be either 0 or 1.</p>
				2977
				2978	<h5>Example:</h5>
				2979	<pre>
				2980	%X = zext i32 257 to i64 <i>; yields i64:257</i>
				2981	%Y = zext i1 true to i32 <i>; yields i32:1</i>
				2982	</pre>
				2983	</div>
				2984
				2985	<!-- _______________________________________________________________________ -->
				2986	<div class="doc_subsubsection">
				2987	<a name="i_sext">'<tt>sext .. to</tt>' Instruction</a>
				2988	</div>
				2989	<div class="doc_text">
				2990
				2991	<h5>Syntax:</h5>
				2992	<pre>
				2993	<result> = sext <ty> <value> to <ty2> <i>; yields ty2</i>
				2994	</pre>
				2995
				2996	<h5>Overview:</h5>
				2997	<p>The '<tt>sext</tt>' sign extends <tt>value</tt> to the type <tt>ty2</tt>.</p>
				2998
				2999	<h5>Arguments:</h5>
				3000	<p>
				3001	The '<tt>sext</tt>' instruction takes a value to cast, which must be of
				3002	<a href="#t_integer">integer</a> type, and a type to cast it to, which must
				3003	also be of <a href="#t_integer">integer</a> type. The bit size of the
				3004	<tt>value</tt> must be smaller than the bit size of the destination type,
				3005	<tt>ty2</tt>.</p>
				3006
				3007	<h5>Semantics:</h5>
				3008	<p>
				3009	The '<tt>sext</tt>' instruction performs a sign extension by copying the sign
				3010	bit (highest order bit) of the <tt>value</tt> until it reaches the bit size of
				3011	the type <tt>ty2</tt>.</p>
				3012
				3013	<p>When sign extending from i1, the extension always results in -1 or 0.</p>
				3014
				3015	<h5>Example:</h5>
				3016	<pre>
				3017	%X = sext i8 -1 to i16 <i>; yields i16 :65535</i>
				3018	%Y = sext i1 true to i32 <i>; yields i32:-1</i>
				3019	</pre>
				3020	</div>
				3021
				3022	<!-- _______________________________________________________________________ -->
				3023	<div class="doc_subsubsection">
				3024	<a name="i_fptrunc">'<tt>fptrunc .. to</tt>' Instruction</a>
				3025	</div>
				3026
				3027	<div class="doc_text">
				3028
				3029	<h5>Syntax:</h5>
				3030
				3031	<pre>
				3032	<result> = fptrunc <ty> <value> to <ty2> <i>; yields ty2</i>
				3033	</pre>
				3034
				3035	<h5>Overview:</h5>
				3036	<p>The '<tt>fptrunc</tt>' instruction truncates <tt>value</tt> to type
				3037	<tt>ty2</tt>.</p>
				3038
				3039
				3040	<h5>Arguments:</h5>
				3041	<p>The '<tt>fptrunc</tt>' instruction takes a <a href="#t_floating">floating
				3042	point</a> value to cast and a <a href="#t_floating">floating point</a> type to
				3043	cast it to. The size of <tt>value</tt> must be larger than the size of
				3044	<tt>ty2</tt>. This implies that <tt>fptrunc</tt> cannot be used to make a
				3045	<i>no-op cast</i>.</p>
				3046
				3047	<h5>Semantics:</h5>
				3048	<p> The '<tt>fptrunc</tt>' instruction truncates a <tt>value</tt> from a larger
				3049	<a href="#t_floating">floating point</a> type to a smaller
				3050	<a href="#t_floating">floating point</a> type. If the value cannot fit within
				3051	the destination type, <tt>ty2</tt>, then the results are undefined.</p>
				3052
				3053	<h5>Example:</h5>
				3054	<pre>
				3055	%X = fptrunc double 123.0 to float <i>; yields float:123.0</i>
				3056	%Y = fptrunc double 1.0E+300 to float <i>; yields undefined</i>
				3057	</pre>
				3058	</div>
				3059
				3060	<!-- _______________________________________________________________________ -->
				3061	<div class="doc_subsubsection">
				3062	<a name="i_fpext">'<tt>fpext .. to</tt>' Instruction</a>
				3063	</div>
				3064	<div class="doc_text">
				3065
				3066	<h5>Syntax:</h5>
				3067	<pre>
				3068	<result> = fpext <ty> <value> to <ty2> <i>; yields ty2</i>
				3069	</pre>
				3070
				3071	<h5>Overview:</h5>
				3072	<p>The '<tt>fpext</tt>' extends a floating point <tt>value</tt> to a larger
				3073	floating point value.</p>
				3074
				3075	<h5>Arguments:</h5>
				3076	<p>The '<tt>fpext</tt>' instruction takes a
				3077	<a href="#t_floating">floating point</a> <tt>value</tt> to cast,
				3078	and a <a href="#t_floating">floating point</a> type to cast it to. The source
				3079	type must be smaller than the destination type.</p>
				3080
				3081	<h5>Semantics:</h5>
				3082	<p>The '<tt>fpext</tt>' instruction extends the <tt>value</tt> from a smaller
				3083	<a href="#t_floating">floating point</a> type to a larger
				3084	<a href="#t_floating">floating point</a> type. The <tt>fpext</tt> cannot be
				3085	used to make a <i>no-op cast</i> because it always changes bits. Use
				3086	<tt>bitcast</tt> to make a <i>no-op cast</i> for a floating point cast.</p>
				3087
				3088	<h5>Example:</h5>
				3089	<pre>
				3090	%X = fpext float 3.1415 to double <i>; yields double:3.1415</i>
				3091	%Y = fpext float 1.0 to float <i>; yields float:1.0 (no-op)</i>
				3092	</pre>
				3093	</div>
				3094
				3095	<!-- _______________________________________________________________________ -->
				3096	<div class="doc_subsubsection">
				3097	<a name="i_fptoui">'<tt>fptoui .. to</tt>' Instruction</a>
				3098	</div>
				3099	<div class="doc_text">
				3100
				3101	<h5>Syntax:</h5>
				3102	<pre>
Reid Spencer	e6adee8	2007-07-31 14:40:14 +0000	[diff] [blame]	3103	<result> = fptoui <ty> <value> to <ty2> <i>; yields ty2</i>
Dan Gohman	f17a25c	2007-07-18 16:29:46 +0000	[diff] [blame]	3104	</pre>
				3105
				3106	<h5>Overview:</h5>
Reid Spencer	e6adee8	2007-07-31 14:40:14 +0000	[diff] [blame]	3107	<p>The '<tt>fptoui</tt>' converts a floating point <tt>value</tt> to its
Dan Gohman	f17a25c	2007-07-18 16:29:46 +0000	[diff] [blame]	3108	unsigned integer equivalent of type <tt>ty2</tt>.
				3109	</p>
				3110
				3111	<h5>Arguments:</h5>
Reid Spencer	e6adee8	2007-07-31 14:40:14 +0000	[diff] [blame]	3112	<p>The '<tt>fptoui</tt>' instruction takes a value to cast, which must be a
Dan Gohman	f17a25c	2007-07-18 16:29:46 +0000	[diff] [blame]	3113	<a href="#t_floating">floating point</a> value, and a type to cast it to, which
				3114	must be an <a href="#t_integer">integer</a> type.</p>
				3115
				3116	<h5>Semantics:</h5>
Reid Spencer	e6adee8	2007-07-31 14:40:14 +0000	[diff] [blame]	3117	<p> The '<tt>fptoui</tt>' instruction converts its
Dan Gohman	f17a25c	2007-07-18 16:29:46 +0000	[diff] [blame]	3118	<a href="#t_floating">floating point</a> operand into the nearest (rounding
				3119	towards zero) unsigned integer value. If the value cannot fit in <tt>ty2</tt>,
				3120	the results are undefined.</p>
				3121
				3122	<p>When converting to i1, the conversion is done as a comparison against
				3123	zero. If the <tt>value</tt> was zero, the i1 result will be <tt>false</tt>.
				3124	If the <tt>value</tt> was non-zero, the i1 result will be <tt>true</tt>.</p>
				3125
				3126	<h5>Example:</h5>
				3127	<pre>
Reid Spencer	e6adee8	2007-07-31 14:40:14 +0000	[diff] [blame]	3128	%X = fptoui double 123.0 to i32 <i>; yields i32:123</i>
				3129	%Y = fptoui float 1.0E+300 to i1 <i>; yields i1:true</i>
				3130	%X = fptoui float 1.04E+17 to i8 <i>; yields undefined:1</i>
Dan Gohman	f17a25c	2007-07-18 16:29:46 +0000	[diff] [blame]	3131	</pre>
				3132	</div>
				3133
				3134	<!-- _______________________________________________________________________ -->
				3135	<div class="doc_subsubsection">
				3136	<a name="i_fptosi">'<tt>fptosi .. to</tt>' Instruction</a>
				3137	</div>
				3138	<div class="doc_text">
				3139
				3140	<h5>Syntax:</h5>
				3141	<pre>
				3142	<result> = fptosi <ty> <value> to <ty2> <i>; yields ty2</i>
				3143	</pre>
				3144
				3145	<h5>Overview:</h5>
				3146	<p>The '<tt>fptosi</tt>' instruction converts
				3147	<a href="#t_floating">floating point</a> <tt>value</tt> to type <tt>ty2</tt>.
				3148	</p>
				3149
				3150
				3151	<h5>Arguments:</h5>
				3152	<p> The '<tt>fptosi</tt>' instruction takes a value to cast, which must be a
				3153	<a href="#t_floating">floating point</a> value, and a type to cast it to, which
				3154	must also be an <a href="#t_integer">integer</a> type.</p>
				3155
				3156	<h5>Semantics:</h5>
				3157	<p>The '<tt>fptosi</tt>' instruction converts its
				3158	<a href="#t_floating">floating point</a> operand into the nearest (rounding
				3159	towards zero) signed integer value. If the value cannot fit in <tt>ty2</tt>,
				3160	the results are undefined.</p>
				3161
				3162	<p>When converting to i1, the conversion is done as a comparison against
				3163	zero. If the <tt>value</tt> was zero, the i1 result will be <tt>false</tt>.
				3164	If the <tt>value</tt> was non-zero, the i1 result will be <tt>true</tt>.</p>
				3165
				3166	<h5>Example:</h5>
				3167	<pre>
				3168	%X = fptosi double -123.0 to i32 <i>; yields i32:-123</i>
				3169	%Y = fptosi float 1.0E-247 to i1 <i>; yields i1:true</i>
				3170	%X = fptosi float 1.04E+17 to i8 <i>; yields undefined:1</i>
				3171	</pre>
				3172	</div>
				3173
				3174	<!-- _______________________________________________________________________ -->
				3175	<div class="doc_subsubsection">
				3176	<a name="i_uitofp">'<tt>uitofp .. to</tt>' Instruction</a>
				3177	</div>
				3178	<div class="doc_text">
				3179
				3180	<h5>Syntax:</h5>
				3181	<pre>
				3182	<result> = uitofp <ty> <value> to <ty2> <i>; yields ty2</i>
				3183	</pre>
				3184
				3185	<h5>Overview:</h5>
				3186	<p>The '<tt>uitofp</tt>' instruction regards <tt>value</tt> as an unsigned
				3187	integer and converts that value to the <tt>ty2</tt> type.</p>
				3188
				3189
				3190	<h5>Arguments:</h5>
				3191	<p>The '<tt>uitofp</tt>' instruction takes a value to cast, which must be an
				3192	<a href="#t_integer">integer</a> value, and a type to cast it to, which must
				3193	be a <a href="#t_floating">floating point</a> type.</p>
				3194
				3195	<h5>Semantics:</h5>
				3196	<p>The '<tt>uitofp</tt>' instruction interprets its operand as an unsigned
				3197	integer quantity and converts it to the corresponding floating point value. If
				3198	the value cannot fit in the floating point value, the results are undefined.</p>
				3199
				3200
				3201	<h5>Example:</h5>
				3202	<pre>
				3203	%X = uitofp i32 257 to float <i>; yields float:257.0</i>
				3204	%Y = uitofp i8 -1 to double <i>; yields double:255.0</i>
				3205	</pre>
				3206	</div>
				3207
				3208	<!-- _______________________________________________________________________ -->
				3209	<div class="doc_subsubsection">
				3210	<a name="i_sitofp">'<tt>sitofp .. to</tt>' Instruction</a>
				3211	</div>
				3212	<div class="doc_text">
				3213
				3214	<h5>Syntax:</h5>
				3215	<pre>
				3216	<result> = sitofp <ty> <value> to <ty2> <i>; yields ty2</i>
				3217	</pre>
				3218
				3219	<h5>Overview:</h5>
				3220	<p>The '<tt>sitofp</tt>' instruction regards <tt>value</tt> as a signed
				3221	integer and converts that value to the <tt>ty2</tt> type.</p>
				3222
				3223	<h5>Arguments:</h5>
				3224	<p>The '<tt>sitofp</tt>' instruction takes a value to cast, which must be an
				3225	<a href="#t_integer">integer</a> value, and a type to cast it to, which must be
				3226	a <a href="#t_floating">floating point</a> type.</p>
				3227
				3228	<h5>Semantics:</h5>
				3229	<p>The '<tt>sitofp</tt>' instruction interprets its operand as a signed
				3230	integer quantity and converts it to the corresponding floating point value. If
				3231	the value cannot fit in the floating point value, the results are undefined.</p>
				3232
				3233	<h5>Example:</h5>
				3234	<pre>
				3235	%X = sitofp i32 257 to float <i>; yields float:257.0</i>
				3236	%Y = sitofp i8 -1 to double <i>; yields double:-1.0</i>
				3237	</pre>
				3238	</div>
				3239
				3240	<!-- _______________________________________________________________________ -->
				3241	<div class="doc_subsubsection">
				3242	<a name="i_ptrtoint">'<tt>ptrtoint .. to</tt>' Instruction</a>
				3243	</div>
				3244	<div class="doc_text">
				3245
				3246	<h5>Syntax:</h5>
				3247	<pre>
				3248	<result> = ptrtoint <ty> <value> to <ty2> <i>; yields ty2</i>
				3249	</pre>
				3250
				3251	<h5>Overview:</h5>
				3252	<p>The '<tt>ptrtoint</tt>' instruction converts the pointer <tt>value</tt> to
				3253	the integer type <tt>ty2</tt>.</p>
				3254
				3255	<h5>Arguments:</h5>
				3256	<p>The '<tt>ptrtoint</tt>' instruction takes a <tt>value</tt> to cast, which
				3257	must be a <a href="#t_pointer">pointer</a> value, and a type to cast it to
				3258	<tt>ty2</tt>, which must be an <a href="#t_integer">integer</a> type.
				3259
				3260	<h5>Semantics:</h5>
				3261	<p>The '<tt>ptrtoint</tt>' instruction converts <tt>value</tt> to integer type
				3262	<tt>ty2</tt> by interpreting the pointer value as an integer and either
				3263	truncating or zero extending that value to the size of the integer type. If
				3264	<tt>value</tt> is smaller than <tt>ty2</tt> then a zero extension is done. If
				3265	<tt>value</tt> is larger than <tt>ty2</tt> then a truncation is done. If they
				3266	are the same size, then nothing is done (<i>no-op cast</i>) other than a type
				3267	change.</p>
				3268
				3269	<h5>Example:</h5>
				3270	<pre>
				3271	%X = ptrtoint i32* %X to i8 <i>; yields truncation on 32-bit architecture</i>
				3272	%Y = ptrtoint i32* %x to i64 <i>; yields zero extension on 32-bit architecture</i>
				3273	</pre>
				3274	</div>
				3275
				3276	<!-- _______________________________________________________________________ -->
				3277	<div class="doc_subsubsection">
				3278	<a name="i_inttoptr">'<tt>inttoptr .. to</tt>' Instruction</a>
				3279	</div>
				3280	<div class="doc_text">
				3281
				3282	<h5>Syntax:</h5>
				3283	<pre>
				3284	<result> = inttoptr <ty> <value> to <ty2> <i>; yields ty2</i>
				3285	</pre>
				3286
				3287	<h5>Overview:</h5>
				3288	<p>The '<tt>inttoptr</tt>' instruction converts an integer <tt>value</tt> to
				3289	a pointer type, <tt>ty2</tt>.</p>
				3290
				3291	<h5>Arguments:</h5>
				3292	<p>The '<tt>inttoptr</tt>' instruction takes an <a href="#t_integer">integer</a>
				3293	value to cast, and a type to cast it to, which must be a
				3294	<a href="#t_pointer">pointer</a> type.
				3295
				3296	<h5>Semantics:</h5>
				3297	<p>The '<tt>inttoptr</tt>' instruction converts <tt>value</tt> to type
				3298	<tt>ty2</tt> by applying either a zero extension or a truncation depending on
				3299	the size of the integer <tt>value</tt>. If <tt>value</tt> is larger than the
				3300	size of a pointer then a truncation is done. If <tt>value</tt> is smaller than
				3301	the size of a pointer then a zero extension is done. If they are the same size,
				3302	nothing is done (<i>no-op cast</i>).</p>
				3303
				3304	<h5>Example:</h5>
				3305	<pre>
				3306	%X = inttoptr i32 255 to i32* <i>; yields zero extension on 64-bit architecture</i>
				3307	%X = inttoptr i32 255 to i32* <i>; yields no-op on 32-bit architecture</i>
				3308	%Y = inttoptr i64 0 to i32* <i>; yields truncation on 32-bit architecture</i>
				3309	</pre>
				3310	</div>
				3311
				3312	<!-- _______________________________________________________________________ -->
				3313	<div class="doc_subsubsection">
				3314	<a name="i_bitcast">'<tt>bitcast .. to</tt>' Instruction</a>
				3315	</div>
				3316	<div class="doc_text">
				3317
				3318	<h5>Syntax:</h5>
				3319	<pre>
				3320	<result> = bitcast <ty> <value> to <ty2> <i>; yields ty2</i>
				3321	</pre>
				3322
				3323	<h5>Overview:</h5>
				3324	<p>The '<tt>bitcast</tt>' instruction converts <tt>value</tt> to type
				3325	<tt>ty2</tt> without changing any bits.</p>
				3326
				3327	<h5>Arguments:</h5>
				3328	<p>The '<tt>bitcast</tt>' instruction takes a value to cast, which must be
				3329	a first class value, and a type to cast it to, which must also be a <a
				3330	href="#t_firstclass">first class</a> type. The bit sizes of <tt>value</tt>
				3331	and the destination type, <tt>ty2</tt>, must be identical. If the source
				3332	type is a pointer, the destination type must also be a pointer.</p>
				3333
				3334	<h5>Semantics:</h5>
				3335	<p>The '<tt>bitcast</tt>' instruction converts <tt>value</tt> to type
				3336	<tt>ty2</tt>. It is always a <i>no-op cast</i> because no bits change with
				3337	this conversion. The conversion is done as if the <tt>value</tt> had been
				3338	stored to memory and read back as type <tt>ty2</tt>. Pointer types may only be
				3339	converted to other pointer types with this instruction. To convert pointers to
				3340	other types, use the <a href="#i_inttoptr">inttoptr</a> or
				3341	<a href="#i_ptrtoint">ptrtoint</a> instructions first.</p>
				3342
				3343	<h5>Example:</h5>
				3344	<pre>
				3345	%X = bitcast i8 255 to i8 <i>; yields i8 :-1</i>
				3346	%Y = bitcast i32* %x to sint* <i>; yields sint*:%x</i>
				3347	%Z = bitcast <2xint> %V to i64; <i>; yields i64: %V</i>
				3348	</pre>
				3349	</div>
				3350
				3351	<!-- ======================================================================= -->
				3352	<div class="doc_subsection"> <a name="otherops">Other Operations</a> </div>
				3353	<div class="doc_text">
				3354	<p>The instructions in this category are the "miscellaneous"
				3355	instructions, which defy better classification.</p>
				3356	</div>
				3357
				3358	<!-- _______________________________________________________________________ -->
				3359	<div class="doc_subsubsection"><a name="i_icmp">'<tt>icmp</tt>' Instruction</a>
				3360	</div>
				3361	<div class="doc_text">
				3362	<h5>Syntax:</h5>
				3363	<pre> <result> = icmp <cond> <ty> <var1>, <var2> <i>; yields {i1}:result</i>
				3364	</pre>
				3365	<h5>Overview:</h5>
				3366	<p>The '<tt>icmp</tt>' instruction returns a boolean value based on comparison
				3367	of its two integer operands.</p>
				3368	<h5>Arguments:</h5>
				3369	<p>The '<tt>icmp</tt>' instruction takes three operands. The first operand is
				3370	the condition code indicating the kind of comparison to perform. It is not
				3371	a value, just a keyword. The possible condition code are:
				3372	<ol>
				3373	<li><tt>eq</tt>: equal</li>
				3374	<li><tt>ne</tt>: not equal </li>
				3375	<li><tt>ugt</tt>: unsigned greater than</li>
				3376	<li><tt>uge</tt>: unsigned greater or equal</li>
				3377	<li><tt>ult</tt>: unsigned less than</li>
				3378	<li><tt>ule</tt>: unsigned less or equal</li>
				3379	<li><tt>sgt</tt>: signed greater than</li>
				3380	<li><tt>sge</tt>: signed greater or equal</li>
				3381	<li><tt>slt</tt>: signed less than</li>
				3382	<li><tt>sle</tt>: signed less or equal</li>
				3383	</ol>
				3384	<p>The remaining two arguments must be <a href="#t_integer">integer</a> or
				3385	<a href="#t_pointer">pointer</a> typed. They must also be identical types.</p>
				3386	<h5>Semantics:</h5>
				3387	<p>The '<tt>icmp</tt>' compares <tt>var1</tt> and <tt>var2</tt> according to
				3388	the condition code given as <tt>cond</tt>. The comparison performed always
				3389	yields a <a href="#t_primitive">i1</a> result, as follows:
				3390	<ol>
				3391	<li><tt>eq</tt>: yields <tt>true</tt> if the operands are equal,
				3392	<tt>false</tt> otherwise. No sign interpretation is necessary or performed.
				3393	</li>
				3394	<li><tt>ne</tt>: yields <tt>true</tt> if the operands are unequal,
				3395	<tt>false</tt> otherwise. No sign interpretation is necessary or performed.
				3396	<li><tt>ugt</tt>: interprets the operands as unsigned values and yields
				3397	<tt>true</tt> if <tt>var1</tt> is greater than <tt>var2</tt>.</li>
				3398	<li><tt>uge</tt>: interprets the operands as unsigned values and yields
				3399	<tt>true</tt> if <tt>var1</tt> is greater than or equal to <tt>var2</tt>.</li>
				3400	<li><tt>ult</tt>: interprets the operands as unsigned values and yields
				3401	<tt>true</tt> if <tt>var1</tt> is less than <tt>var2</tt>.</li>
				3402	<li><tt>ule</tt>: interprets the operands as unsigned values and yields
				3403	<tt>true</tt> if <tt>var1</tt> is less than or equal to <tt>var2</tt>.</li>
				3404	<li><tt>sgt</tt>: interprets the operands as signed values and yields
				3405	<tt>true</tt> if <tt>var1</tt> is greater than <tt>var2</tt>.</li>
				3406	<li><tt>sge</tt>: interprets the operands as signed values and yields
				3407	<tt>true</tt> if <tt>var1</tt> is greater than or equal to <tt>var2</tt>.</li>
				3408	<li><tt>slt</tt>: interprets the operands as signed values and yields
				3409	<tt>true</tt> if <tt>var1</tt> is less than <tt>var2</tt>.</li>
				3410	<li><tt>sle</tt>: interprets the operands as signed values and yields
				3411	<tt>true</tt> if <tt>var1</tt> is less than or equal to <tt>var2</tt>.</li>
				3412	</ol>
				3413	<p>If the operands are <a href="#t_pointer">pointer</a> typed, the pointer
				3414	values are compared as if they were integers.</p>
				3415
				3416	<h5>Example:</h5>
				3417	<pre> <result> = icmp eq i32 4, 5 <i>; yields: result=false</i>
				3418	<result> = icmp ne float* %X, %X <i>; yields: result=false</i>
				3419	<result> = icmp ult i16 4, 5 <i>; yields: result=true</i>
				3420	<result> = icmp sgt i16 4, 5 <i>; yields: result=false</i>
				3421	<result> = icmp ule i16 -4, 5 <i>; yields: result=false</i>
				3422	<result> = icmp sge i16 4, 5 <i>; yields: result=false</i>
				3423	</pre>
				3424	</div>
				3425
				3426	<!-- _______________________________________________________________________ -->
				3427	<div class="doc_subsubsection"><a name="i_fcmp">'<tt>fcmp</tt>' Instruction</a>
				3428	</div>
				3429	<div class="doc_text">
				3430	<h5>Syntax:</h5>
				3431	<pre> <result> = fcmp <cond> <ty> <var1>, <var2> <i>; yields {i1}:result</i>
				3432	</pre>
				3433	<h5>Overview:</h5>
				3434	<p>The '<tt>fcmp</tt>' instruction returns a boolean value based on comparison
				3435	of its floating point operands.</p>
				3436	<h5>Arguments:</h5>
				3437	<p>The '<tt>fcmp</tt>' instruction takes three operands. The first operand is
				3438	the condition code indicating the kind of comparison to perform. It is not
				3439	a value, just a keyword. The possible condition code are:
				3440	<ol>
				3441	<li><tt>false</tt>: no comparison, always returns false</li>
				3442	<li><tt>oeq</tt>: ordered and equal</li>
				3443	<li><tt>ogt</tt>: ordered and greater than </li>
				3444	<li><tt>oge</tt>: ordered and greater than or equal</li>
				3445	<li><tt>olt</tt>: ordered and less than </li>
				3446	<li><tt>ole</tt>: ordered and less than or equal</li>
				3447	<li><tt>one</tt>: ordered and not equal</li>
				3448	<li><tt>ord</tt>: ordered (no nans)</li>
				3449	<li><tt>ueq</tt>: unordered or equal</li>
				3450	<li><tt>ugt</tt>: unordered or greater than </li>
				3451	<li><tt>uge</tt>: unordered or greater than or equal</li>
				3452	<li><tt>ult</tt>: unordered or less than </li>
				3453	<li><tt>ule</tt>: unordered or less than or equal</li>
				3454	<li><tt>une</tt>: unordered or not equal</li>
				3455	<li><tt>uno</tt>: unordered (either nans)</li>
				3456	<li><tt>true</tt>: no comparison, always returns true</li>
				3457	</ol>
				3458	<p><i>Ordered</i> means that neither operand is a QNAN while
				3459	<i>unordered</i> means that either operand may be a QNAN.</p>
				3460	<p>The <tt>val1</tt> and <tt>val2</tt> arguments must be
				3461	<a href="#t_floating">floating point</a> typed. They must have identical
				3462	types.</p>
				3463	<h5>Semantics:</h5>
				3464	<p>The '<tt>fcmp</tt>' compares <tt>var1</tt> and <tt>var2</tt> according to
				3465	the condition code given as <tt>cond</tt>. The comparison performed always
				3466	yields a <a href="#t_primitive">i1</a> result, as follows:
				3467	<ol>
				3468	<li><tt>false</tt>: always yields <tt>false</tt>, regardless of operands.</li>
				3469	<li><tt>oeq</tt>: yields <tt>true</tt> if both operands are not a QNAN and
				3470	<tt>var1</tt> is equal to <tt>var2</tt>.</li>
				3471	<li><tt>ogt</tt>: yields <tt>true</tt> if both operands are not a QNAN and
				3472	<tt>var1</tt> is greather than <tt>var2</tt>.</li>
				3473	<li><tt>oge</tt>: yields <tt>true</tt> if both operands are not a QNAN and
				3474	<tt>var1</tt> is greater than or equal to <tt>var2</tt>.</li>
				3475	<li><tt>olt</tt>: yields <tt>true</tt> if both operands are not a QNAN and
				3476	<tt>var1</tt> is less than <tt>var2</tt>.</li>
				3477	<li><tt>ole</tt>: yields <tt>true</tt> if both operands are not a QNAN and
				3478	<tt>var1</tt> is less than or equal to <tt>var2</tt>.</li>
				3479	<li><tt>one</tt>: yields <tt>true</tt> if both operands are not a QNAN and
				3480	<tt>var1</tt> is not equal to <tt>var2</tt>.</li>
				3481	<li><tt>ord</tt>: yields <tt>true</tt> if both operands are not a QNAN.</li>
				3482	<li><tt>ueq</tt>: yields <tt>true</tt> if either operand is a QNAN or
				3483	<tt>var1</tt> is equal to <tt>var2</tt>.</li>
				3484	<li><tt>ugt</tt>: yields <tt>true</tt> if either operand is a QNAN or
				3485	<tt>var1</tt> is greater than <tt>var2</tt>.</li>
				3486	<li><tt>uge</tt>: yields <tt>true</tt> if either operand is a QNAN or
				3487	<tt>var1</tt> is greater than or equal to <tt>var2</tt>.</li>
				3488	<li><tt>ult</tt>: yields <tt>true</tt> if either operand is a QNAN or
				3489	<tt>var1</tt> is less than <tt>var2</tt>.</li>
				3490	<li><tt>ule</tt>: yields <tt>true</tt> if either operand is a QNAN or
				3491	<tt>var1</tt> is less than or equal to <tt>var2</tt>.</li>
				3492	<li><tt>une</tt>: yields <tt>true</tt> if either operand is a QNAN or
				3493	<tt>var1</tt> is not equal to <tt>var2</tt>.</li>
				3494	<li><tt>uno</tt>: yields <tt>true</tt> if either operand is a QNAN.</li>
				3495	<li><tt>true</tt>: always yields <tt>true</tt>, regardless of operands.</li>
				3496	</ol>
				3497
				3498	<h5>Example:</h5>
				3499	<pre> <result> = fcmp oeq float 4.0, 5.0 <i>; yields: result=false</i>
				3500	<result> = icmp one float 4.0, 5.0 <i>; yields: result=true</i>
				3501	<result> = icmp olt float 4.0, 5.0 <i>; yields: result=true</i>
				3502	<result> = icmp ueq double 1.0, 2.0 <i>; yields: result=false</i>
				3503	</pre>
				3504	</div>
				3505
				3506	<!-- _______________________________________________________________________ -->
				3507	<div class="doc_subsubsection"> <a name="i_phi">'<tt>phi</tt>'
				3508	Instruction</a> </div>
				3509	<div class="doc_text">
				3510	<h5>Syntax:</h5>
				3511	<pre> <result> = phi <ty> [ <val0>, <label0>], ...<br></pre>
				3512	<h5>Overview:</h5>
				3513	<p>The '<tt>phi</tt>' instruction is used to implement the φ node in
				3514	the SSA graph representing the function.</p>
				3515	<h5>Arguments:</h5>
				3516	<p>The type of the incoming values is specified with the first type
				3517	field. After this, the '<tt>phi</tt>' instruction takes a list of pairs
				3518	as arguments, with one pair for each predecessor basic block of the
				3519	current block. Only values of <a href="#t_firstclass">first class</a>
				3520	type may be used as the value arguments to the PHI node. Only labels
				3521	may be used as the label arguments.</p>
				3522	<p>There must be no non-phi instructions between the start of a basic
				3523	block and the PHI instructions: i.e. PHI instructions must be first in
				3524	a basic block.</p>
				3525	<h5>Semantics:</h5>
				3526	<p>At runtime, the '<tt>phi</tt>' instruction logically takes on the value
				3527	specified by the pair corresponding to the predecessor basic block that executed
				3528	just prior to the current block.</p>
				3529	<h5>Example:</h5>
				3530	<pre>Loop: ; Infinite loop that counts from 0 on up...<br> %indvar = phi i32 [ 0, %LoopHeader ], [ %nextindvar, %Loop ]<br> %nextindvar = add i32 %indvar, 1<br> br label %Loop<br></pre>
				3531	</div>
				3532
				3533	<!-- _______________________________________________________________________ -->
				3534	<div class="doc_subsubsection">
				3535	<a name="i_select">'<tt>select</tt>' Instruction</a>
				3536	</div>
				3537
				3538	<div class="doc_text">
				3539
				3540	<h5>Syntax:</h5>
				3541
				3542	<pre>
				3543	<result> = select i1 <cond>, <ty> <val1>, <ty> <val2> <i>; yields ty</i>
				3544	</pre>
				3545
				3546	<h5>Overview:</h5>
				3547
				3548	<p>
				3549	The '<tt>select</tt>' instruction is used to choose one value based on a
				3550	condition, without branching.
				3551	</p>
				3552
				3553
				3554	<h5>Arguments:</h5>
				3555
				3556	<p>
				3557	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.
				3558	</p>
				3559
				3560	<h5>Semantics:</h5>
				3561
				3562	<p>
				3563	If the boolean condition evaluates to true, the instruction returns the first
				3564	value argument; otherwise, it returns the second value argument.
				3565	</p>
				3566
				3567	<h5>Example:</h5>
				3568
				3569	<pre>
				3570	%X = select i1 true, i8 17, i8 42 <i>; yields i8:17</i>
				3571	</pre>
				3572	</div>
				3573
				3574
				3575	<!-- _______________________________________________________________________ -->
				3576	<div class="doc_subsubsection">
				3577	<a name="i_call">'<tt>call</tt>' Instruction</a>
				3578	</div>
				3579
				3580	<div class="doc_text">
				3581
				3582	<h5>Syntax:</h5>
				3583	<pre>
				3584	<result> = [tail] call [<a href="#callingconv">cconv</a>] <ty>* <fnptrval>(<param list>)
				3585	</pre>
				3586
				3587	<h5>Overview:</h5>
				3588
				3589	<p>The '<tt>call</tt>' instruction represents a simple function call.</p>
				3590
				3591	<h5>Arguments:</h5>
				3592
				3593	<p>This instruction requires several arguments:</p>
				3594
				3595	<ol>
				3596	<li>
				3597	<p>The optional "tail" marker indicates whether the callee function accesses
				3598	any allocas or varargs in the caller. If the "tail" marker is present, the
				3599	function call is eligible for tail call optimization. Note that calls may
				3600	be marked "tail" even if they do not occur before a <a
				3601	href="#i_ret"><tt>ret</tt></a> instruction.
				3602	</li>
				3603	<li>
				3604	<p>The optional "cconv" marker indicates which <a href="#callingconv">calling
				3605	convention</a> the call should use. If none is specified, the call defaults
				3606	to using C calling conventions.
				3607	</li>
				3608	<li>
				3609	<p>'<tt>ty</tt>': shall be the signature of the pointer to function value
				3610	being invoked. The argument types must match the types implied by this
				3611	signature. This type can be omitted if the function is not varargs and
				3612	if the function type does not return a pointer to a function.</p>
				3613	</li>
				3614	<li>
				3615	<p>'<tt>fnptrval</tt>': An LLVM value containing a pointer to a function to
				3616	be invoked. In most cases, this is a direct function invocation, but
				3617	indirect <tt>call</tt>s are just as possible, calling an arbitrary pointer
				3618	to function value.</p>
				3619	</li>
				3620	<li>
				3621	<p>'<tt>function args</tt>': argument list whose types match the
				3622	function signature argument types. All arguments must be of
				3623	<a href="#t_firstclass">first class</a> type. If the function signature
				3624	indicates the function accepts a variable number of arguments, the extra
				3625	arguments can be specified.</p>
				3626	</li>
				3627	</ol>
				3628
				3629	<h5>Semantics:</h5>
				3630
				3631	<p>The '<tt>call</tt>' instruction is used to cause control flow to
				3632	transfer to a specified function, with its incoming arguments bound to
				3633	the specified values. Upon a '<tt><a href="#i_ret">ret</a></tt>'
				3634	instruction in the called function, control flow continues with the
				3635	instruction after the function call, and the return value of the
				3636	function is bound to the result argument. This is a simpler case of
				3637	the <a href="#i_invoke">invoke</a> instruction.</p>
				3638
				3639	<h5>Example:</h5>
				3640
				3641	<pre>
				3642	%retval = call i32 %test(i32 %argc)
				3643	call i32(i8 , ...) %printf(i8 * %msg, i32 12, i8 42);
				3644	%X = tail call i32 %foo()
				3645	%Y = tail call <a href="#callingconv">fastcc</a> i32 %foo()
				3646	</pre>
				3647
				3648	</div>
				3649
				3650	<!-- _______________________________________________________________________ -->
				3651	<div class="doc_subsubsection">
				3652	<a name="i_va_arg">'<tt>va_arg</tt>' Instruction</a>
				3653	</div>
				3654
				3655	<div class="doc_text">
				3656
				3657	<h5>Syntax:</h5>
				3658
				3659	<pre>
				3660	<resultval> = va_arg <va_list*> <arglist>, <argty>
				3661	</pre>
				3662
				3663	<h5>Overview:</h5>
				3664
				3665	<p>The '<tt>va_arg</tt>' instruction is used to access arguments passed through
				3666	the "variable argument" area of a function call. It is used to implement the
				3667	<tt>va_arg</tt> macro in C.</p>
				3668
				3669	<h5>Arguments:</h5>
				3670
				3671	<p>This instruction takes a <tt>va_list*</tt> value and the type of
				3672	the argument. It returns a value of the specified argument type and
				3673	increments the <tt>va_list</tt> to point to the next argument. The
				3674	actual type of <tt>va_list</tt> is target specific.</p>
				3675
				3676	<h5>Semantics:</h5>
				3677
				3678	<p>The '<tt>va_arg</tt>' instruction loads an argument of the specified
				3679	type from the specified <tt>va_list</tt> and causes the
				3680	<tt>va_list</tt> to point to the next argument. For more information,
				3681	see the variable argument handling <a href="#int_varargs">Intrinsic
				3682	Functions</a>.</p>
				3683
				3684	<p>It is legal for this instruction to be called in a function which does not
				3685	take a variable number of arguments, for example, the <tt>vfprintf</tt>
				3686	function.</p>
				3687
				3688	<p><tt>va_arg</tt> is an LLVM instruction instead of an <a
				3689	href="#intrinsics">intrinsic function</a> because it takes a type as an
				3690	argument.</p>
				3691
				3692	<h5>Example:</h5>
				3693
				3694	<p>See the <a href="#int_varargs">variable argument processing</a> section.</p>
				3695
				3696	</div>
				3697
				3698	<!-- *********************************************************************** -->
				3699	<div class="doc_section"> <a name="intrinsics">Intrinsic Functions</a> </div>
				3700	<!-- *********************************************************************** -->
				3701
				3702	<div class="doc_text">
				3703
				3704	<p>LLVM supports the notion of an "intrinsic function". These functions have
				3705	well known names and semantics and are required to follow certain restrictions.
				3706	Overall, these intrinsics represent an extension mechanism for the LLVM
				3707	language that does not require changing all of the transformations in LLVM when
				3708	adding to the language (or the bitcode reader/writer, the parser, etc...).</p>
				3709
				3710	<p>Intrinsic function names must all start with an "<tt>llvm.</tt>" prefix. This
				3711	prefix is reserved in LLVM for intrinsic names; thus, function names may not
				3712	begin with this prefix. Intrinsic functions must always be external functions:
				3713	you cannot define the body of intrinsic functions. Intrinsic functions may
				3714	only be used in call or invoke instructions: it is illegal to take the address
				3715	of an intrinsic function. Additionally, because intrinsic functions are part
				3716	of the LLVM language, it is required if any are added that they be documented
				3717	here.</p>
				3718
Chandler Carruth	a228e39	2007-08-04 01:51:18 +0000	[diff] [blame]	3719	<p>Some intrinsic functions can be overloaded, i.e., the intrinsic represents
				3720	a family of functions that perform the same operation but on different data
				3721	types. Because LLVM can represent over 8 million different integer types,
				3722	overloading is used commonly to allow an intrinsic function to operate on any
				3723	integer type. One or more of the argument types or the result type can be
				3724	overloaded to accept any integer type. Argument types may also be defined as
				3725	exactly matching a previous argument's type or the result type. This allows an
				3726	intrinsic function which accepts multiple arguments, but needs all of them to
				3727	be of the same type, to only be overloaded with respect to a single argument or
				3728	the result.</p>
Dan Gohman	f17a25c	2007-07-18 16:29:46 +0000	[diff] [blame]	3729
Chandler Carruth	a228e39	2007-08-04 01:51:18 +0000	[diff] [blame]	3730	<p>Overloaded intrinsics will have the names of its overloaded argument types
				3731	encoded into its function name, each preceded by a period. Only those types
				3732	which are overloaded result in a name suffix. Arguments whose type is matched
				3733	against another type do not. For example, the <tt>llvm.ctpop</tt> function can
				3734	take an integer of any width and returns an integer of exactly the same integer
				3735	width. This leads to a family of functions such as
				3736	<tt>i8 @llvm.ctpop.i8(i8 %val)</tt> and <tt>i29 @llvm.ctpop.i29(i29 %val)</tt>.
				3737	Only one type, the return type, is overloaded, and only one type suffix is
				3738	required. Because the argument's type is matched against the return type, it
				3739	does not require its own name suffix.</p>
Dan Gohman	f17a25c	2007-07-18 16:29:46 +0000	[diff] [blame]	3740
				3741	<p>To learn how to add an intrinsic function, please see the
				3742	<a href="ExtendingLLVM.html">Extending LLVM Guide</a>.
				3743	</p>
				3744
				3745	</div>
				3746
				3747	<!-- ======================================================================= -->
				3748	<div class="doc_subsection">
				3749	<a name="int_varargs">Variable Argument Handling Intrinsics</a>
				3750	</div>
				3751
				3752	<div class="doc_text">
				3753
				3754	<p>Variable argument support is defined in LLVM with the <a
				3755	href="#i_va_arg"><tt>va_arg</tt></a> instruction and these three
				3756	intrinsic functions. These functions are related to the similarly
				3757	named macros defined in the <tt><stdarg.h></tt> header file.</p>
				3758
				3759	<p>All of these functions operate on arguments that use a
				3760	target-specific value type "<tt>va_list</tt>". The LLVM assembly
				3761	language reference manual does not define what this type is, so all
				3762	transformations should be prepared to handle these functions regardless of
				3763	the type used.</p>
				3764
				3765	<p>This example shows how the <a href="#i_va_arg"><tt>va_arg</tt></a>
				3766	instruction and the variable argument handling intrinsic functions are
				3767	used.</p>
				3768
				3769	<div class="doc_code">
				3770	<pre>
				3771	define i32 @test(i32 %X, ...) {
				3772	; Initialize variable argument processing
				3773	%ap = alloca i8*
				3774	%ap2 = bitcast i8** %ap to i8*
				3775	call void @llvm.va_start(i8* %ap2)
				3776
				3777	; Read a single integer argument
				3778	%tmp = va_arg i8** %ap, i32
				3779
				3780	; Demonstrate usage of llvm.va_copy and llvm.va_end
				3781	%aq = alloca i8*
				3782	%aq2 = bitcast i8** %aq to i8*
				3783	call void @llvm.va_copy(i8* %aq2, i8* %ap2)
				3784	call void @llvm.va_end(i8* %aq2)
				3785
				3786	; Stop processing of arguments.
				3787	call void @llvm.va_end(i8* %ap2)
				3788	ret i32 %tmp
				3789	}
				3790
				3791	declare void @llvm.va_start(i8*)
				3792	declare void @llvm.va_copy(i8, i8)
				3793	declare void @llvm.va_end(i8*)
				3794	</pre>
				3795	</div>
				3796
				3797	</div>
				3798
				3799	<!-- _______________________________________________________________________ -->
				3800	<div class="doc_subsubsection">
				3801	<a name="int_va_start">'<tt>llvm.va_start</tt>' Intrinsic</a>
				3802	</div>
				3803
				3804
				3805	<div class="doc_text">
				3806	<h5>Syntax:</h5>
				3807	<pre> declare void %llvm.va_start(i8* <arglist>)<br></pre>
				3808	<h5>Overview:</h5>
				3809	<P>The '<tt>llvm.va_start</tt>' intrinsic initializes
				3810	<tt>*<arglist></tt> for subsequent use by <tt><a
				3811	href="#i_va_arg">va_arg</a></tt>.</p>
				3812
				3813	<h5>Arguments:</h5>
				3814
				3815	<P>The argument is a pointer to a <tt>va_list</tt> element to initialize.</p>
				3816
				3817	<h5>Semantics:</h5>
				3818
				3819	<P>The '<tt>llvm.va_start</tt>' intrinsic works just like the <tt>va_start</tt>
				3820	macro available in C. In a target-dependent way, it initializes the
				3821	<tt>va_list</tt> element to which the argument points, so that the next call to
				3822	<tt>va_arg</tt> will produce the first variable argument passed to the function.
				3823	Unlike the C <tt>va_start</tt> macro, this intrinsic does not need to know the
				3824	last argument of the function as the compiler can figure that out.</p>
				3825
				3826	</div>
				3827
				3828	<!-- _______________________________________________________________________ -->
				3829	<div class="doc_subsubsection">
				3830	<a name="int_va_end">'<tt>llvm.va_end</tt>' Intrinsic</a>
				3831	</div>
				3832
				3833	<div class="doc_text">
				3834	<h5>Syntax:</h5>
				3835	<pre> declare void @llvm.va_end(i8* <arglist>)<br></pre>
				3836	<h5>Overview:</h5>
				3837
				3838	<p>The '<tt>llvm.va_end</tt>' intrinsic destroys <tt>*<arglist></tt>,
				3839	which has been initialized previously with <tt><a href="#int_va_start">llvm.va_start</a></tt>
				3840	or <tt><a href="#i_va_copy">llvm.va_copy</a></tt>.</p>
				3841
				3842	<h5>Arguments:</h5>
				3843
				3844	<p>The argument is a pointer to a <tt>va_list</tt> to destroy.</p>
				3845
				3846	<h5>Semantics:</h5>
				3847
				3848	<p>The '<tt>llvm.va_end</tt>' intrinsic works just like the <tt>va_end</tt>
				3849	macro available in C. In a target-dependent way, it destroys the
				3850	<tt>va_list</tt> element to which the argument points. Calls to <a
				3851	href="#int_va_start"><tt>llvm.va_start</tt></a> and <a href="#int_va_copy">
				3852	<tt>llvm.va_copy</tt></a> must be matched exactly with calls to
				3853	<tt>llvm.va_end</tt>.</p>
				3854
				3855	</div>
				3856
				3857	<!-- _______________________________________________________________________ -->
				3858	<div class="doc_subsubsection">
				3859	<a name="int_va_copy">'<tt>llvm.va_copy</tt>' Intrinsic</a>
				3860	</div>
				3861
				3862	<div class="doc_text">
				3863
				3864	<h5>Syntax:</h5>
				3865
				3866	<pre>
				3867	declare void @llvm.va_copy(i8* <destarglist>, i8* <srcarglist>)
				3868	</pre>
				3869
				3870	<h5>Overview:</h5>
				3871
				3872	<p>The '<tt>llvm.va_copy</tt>' intrinsic copies the current argument position
				3873	from the source argument list to the destination argument list.</p>
				3874
				3875	<h5>Arguments:</h5>
				3876
				3877	<p>The first argument is a pointer to a <tt>va_list</tt> element to initialize.
				3878	The second argument is a pointer to a <tt>va_list</tt> element to copy from.</p>
				3879
				3880
				3881	<h5>Semantics:</h5>
				3882
				3883	<p>The '<tt>llvm.va_copy</tt>' intrinsic works just like the <tt>va_copy</tt>
				3884	macro available in C. In a target-dependent way, it copies the source
				3885	<tt>va_list</tt> element into the destination <tt>va_list</tt> element. This
				3886	intrinsic is necessary because the <tt><a href="#int_va_start">
				3887	llvm.va_start</a></tt> intrinsic may be arbitrarily complex and require, for
				3888	example, memory allocation.</p>
				3889
				3890	</div>
				3891
				3892	<!-- ======================================================================= -->
				3893	<div class="doc_subsection">
				3894	<a name="int_gc">Accurate Garbage Collection Intrinsics</a>
				3895	</div>
				3896
				3897	<div class="doc_text">
				3898
				3899	<p>
				3900	LLVM support for <a href="GarbageCollection.html">Accurate Garbage
				3901	Collection</a> requires the implementation and generation of these intrinsics.
				3902	These intrinsics allow identification of <a href="#int_gcroot">GC roots on the
				3903	stack</a>, as well as garbage collector implementations that require <a
				3904	href="#int_gcread">read</a> and <a href="#int_gcwrite">write</a> barriers.
				3905	Front-ends for type-safe garbage collected languages should generate these
				3906	intrinsics to make use of the LLVM garbage collectors. For more details, see <a
				3907	href="GarbageCollection.html">Accurate Garbage Collection with LLVM</a>.
				3908	</p>
				3909	</div>
				3910
				3911	<!-- _______________________________________________________________________ -->
				3912	<div class="doc_subsubsection">
				3913	<a name="int_gcroot">'<tt>llvm.gcroot</tt>' Intrinsic</a>
				3914	</div>
				3915
				3916	<div class="doc_text">
				3917
				3918	<h5>Syntax:</h5>
				3919
				3920	<pre>
				3921	declare void @llvm.gcroot(<ty>** %ptrloc, <ty2>* %metadata)
				3922	</pre>
				3923
				3924	<h5>Overview:</h5>
				3925
				3926	<p>The '<tt>llvm.gcroot</tt>' intrinsic declares the existence of a GC root to
				3927	the code generator, and allows some metadata to be associated with it.</p>
				3928
				3929	<h5>Arguments:</h5>
				3930
				3931	<p>The first argument specifies the address of a stack object that contains the
				3932	root pointer. The second pointer (which must be either a constant or a global
				3933	value address) contains the meta-data to be associated with the root.</p>
				3934
				3935	<h5>Semantics:</h5>
				3936
				3937	<p>At runtime, a call to this intrinsics stores a null pointer into the "ptrloc"
				3938	location. At compile-time, the code generator generates information to allow
				3939	the runtime to find the pointer at GC safe points.
				3940	</p>
				3941
				3942	</div>
				3943
				3944
				3945	<!-- _______________________________________________________________________ -->
				3946	<div class="doc_subsubsection">
				3947	<a name="int_gcread">'<tt>llvm.gcread</tt>' Intrinsic</a>
				3948	</div>
				3949
				3950	<div class="doc_text">
				3951
				3952	<h5>Syntax:</h5>
				3953
				3954	<pre>
				3955	declare i8 * @llvm.gcread(i8 * %ObjPtr, i8 ** %Ptr)
				3956	</pre>
				3957
				3958	<h5>Overview:</h5>
				3959
				3960	<p>The '<tt>llvm.gcread</tt>' intrinsic identifies reads of references from heap
				3961	locations, allowing garbage collector implementations that require read
				3962	barriers.</p>
				3963
				3964	<h5>Arguments:</h5>
				3965
				3966	<p>The second argument is the address to read from, which should be an address
				3967	allocated from the garbage collector. The first object is a pointer to the
				3968	start of the referenced object, if needed by the language runtime (otherwise
				3969	null).</p>
				3970
				3971	<h5>Semantics:</h5>
				3972
				3973	<p>The '<tt>llvm.gcread</tt>' intrinsic has the same semantics as a load
				3974	instruction, but may be replaced with substantially more complex code by the
				3975	garbage collector runtime, as needed.</p>
				3976
				3977	</div>
				3978
				3979
				3980	<!-- _______________________________________________________________________ -->
				3981	<div class="doc_subsubsection">
				3982	<a name="int_gcwrite">'<tt>llvm.gcwrite</tt>' Intrinsic</a>
				3983	</div>
				3984
				3985	<div class="doc_text">
				3986
				3987	<h5>Syntax:</h5>
				3988
				3989	<pre>
				3990	declare void @llvm.gcwrite(i8 * %P1, i8 * %Obj, i8 ** %P2)
				3991	</pre>
				3992
				3993	<h5>Overview:</h5>
				3994
				3995	<p>The '<tt>llvm.gcwrite</tt>' intrinsic identifies writes of references to heap
				3996	locations, allowing garbage collector implementations that require write
				3997	barriers (such as generational or reference counting collectors).</p>
				3998
				3999	<h5>Arguments:</h5>
				4000
				4001	<p>The first argument is the reference to store, the second is the start of the
				4002	object to store it to, and the third is the address of the field of Obj to
				4003	store to. If the runtime does not require a pointer to the object, Obj may be
				4004	null.</p>
				4005
				4006	<h5>Semantics:</h5>
				4007
				4008	<p>The '<tt>llvm.gcwrite</tt>' intrinsic has the same semantics as a store
				4009	instruction, but may be replaced with substantially more complex code by the
				4010	garbage collector runtime, as needed.</p>
				4011
				4012	</div>
				4013
				4014
				4015
				4016	<!-- ======================================================================= -->
				4017	<div class="doc_subsection">
				4018	<a name="int_codegen">Code Generator Intrinsics</a>
				4019	</div>
				4020
				4021	<div class="doc_text">
				4022	<p>
				4023	These intrinsics are provided by LLVM to expose special features that may only
				4024	be implemented with code generator support.
				4025	</p>
				4026
				4027	</div>
				4028
				4029	<!-- _______________________________________________________________________ -->
				4030	<div class="doc_subsubsection">
				4031	<a name="int_returnaddress">'<tt>llvm.returnaddress</tt>' Intrinsic</a>
				4032	</div>
				4033
				4034	<div class="doc_text">
				4035
				4036	<h5>Syntax:</h5>
				4037	<pre>
				4038	declare i8 *@llvm.returnaddress(i32 <level>)
				4039	</pre>
				4040
				4041	<h5>Overview:</h5>
				4042
				4043	<p>
				4044	The '<tt>llvm.returnaddress</tt>' intrinsic attempts to compute a
				4045	target-specific value indicating the return address of the current function
				4046	or one of its callers.
				4047	</p>
				4048
				4049	<h5>Arguments:</h5>
				4050
				4051	<p>
				4052	The argument to this intrinsic indicates which function to return the address
				4053	for. Zero indicates the calling function, one indicates its caller, etc. The
				4054	argument is <b>required</b> to be a constant integer value.
				4055	</p>
				4056
				4057	<h5>Semantics:</h5>
				4058
				4059	<p>
				4060	The '<tt>llvm.returnaddress</tt>' intrinsic either returns a pointer indicating
				4061	the return address of the specified call frame, or zero if it cannot be
				4062	identified. The value returned by this intrinsic is likely to be incorrect or 0
				4063	for arguments other than zero, so it should only be used for debugging purposes.
				4064	</p>
				4065
				4066	<p>
				4067	Note that calling this intrinsic does not prevent function inlining or other
				4068	aggressive transformations, so the value returned may not be that of the obvious
				4069	source-language caller.
				4070	</p>
				4071	</div>
				4072
				4073
				4074	<!-- _______________________________________________________________________ -->
				4075	<div class="doc_subsubsection">
				4076	<a name="int_frameaddress">'<tt>llvm.frameaddress</tt>' Intrinsic</a>
				4077	</div>
				4078
				4079	<div class="doc_text">
				4080
				4081	<h5>Syntax:</h5>
				4082	<pre>
				4083	declare i8 *@llvm.frameaddress(i32 <level>)
				4084	</pre>
				4085
				4086	<h5>Overview:</h5>
				4087
				4088	<p>
				4089	The '<tt>llvm.frameaddress</tt>' intrinsic attempts to return the
				4090	target-specific frame pointer value for the specified stack frame.
				4091	</p>
				4092
				4093	<h5>Arguments:</h5>
				4094
				4095	<p>
				4096	The argument to this intrinsic indicates which function to return the frame
				4097	pointer for. Zero indicates the calling function, one indicates its caller,
				4098	etc. The argument is <b>required</b> to be a constant integer value.
				4099	</p>
				4100
				4101	<h5>Semantics:</h5>
				4102
				4103	<p>
				4104	The '<tt>llvm.frameaddress</tt>' intrinsic either returns a pointer indicating
				4105	the frame address of the specified call frame, or zero if it cannot be
				4106	identified. The value returned by this intrinsic is likely to be incorrect or 0
				4107	for arguments other than zero, so it should only be used for debugging purposes.
				4108	</p>
				4109
				4110	<p>
				4111	Note that calling this intrinsic does not prevent function inlining or other
				4112	aggressive transformations, so the value returned may not be that of the obvious
				4113	source-language caller.
				4114	</p>
				4115	</div>
				4116
				4117	<!-- _______________________________________________________________________ -->
				4118	<div class="doc_subsubsection">
				4119	<a name="int_stacksave">'<tt>llvm.stacksave</tt>' Intrinsic</a>
				4120	</div>
				4121
				4122	<div class="doc_text">
				4123
				4124	<h5>Syntax:</h5>
				4125	<pre>
				4126	declare i8 *@llvm.stacksave()
				4127	</pre>
				4128
				4129	<h5>Overview:</h5>
				4130
				4131	<p>
				4132	The '<tt>llvm.stacksave</tt>' intrinsic is used to remember the current state of
				4133	the function stack, for use with <a href="#int_stackrestore">
				4134	<tt>llvm.stackrestore</tt></a>. This is useful for implementing language
				4135	features like scoped automatic variable sized arrays in C99.
				4136	</p>
				4137
				4138	<h5>Semantics:</h5>
				4139
				4140	<p>
				4141	This intrinsic returns a opaque pointer value that can be passed to <a
				4142	href="#int_stackrestore"><tt>llvm.stackrestore</tt></a>. When an
				4143	<tt>llvm.stackrestore</tt> intrinsic is executed with a value saved from
				4144	<tt>llvm.stacksave</tt>, it effectively restores the state of the stack to the
				4145	state it was in when the <tt>llvm.stacksave</tt> intrinsic executed. In
				4146	practice, this pops any <a href="#i_alloca">alloca</a> blocks from the stack
				4147	that were allocated after the <tt>llvm.stacksave</tt> was executed.
				4148	</p>
				4149
				4150	</div>
				4151
				4152	<!-- _______________________________________________________________________ -->
				4153	<div class="doc_subsubsection">
				4154	<a name="int_stackrestore">'<tt>llvm.stackrestore</tt>' Intrinsic</a>
				4155	</div>
				4156
				4157	<div class="doc_text">
				4158
				4159	<h5>Syntax:</h5>
				4160	<pre>
				4161	declare void @llvm.stackrestore(i8 * %ptr)
				4162	</pre>
				4163
				4164	<h5>Overview:</h5>
				4165
				4166	<p>
				4167	The '<tt>llvm.stackrestore</tt>' intrinsic is used to restore the state of
				4168	the function stack to the state it was in when the corresponding <a
				4169	href="#int_stacksave"><tt>llvm.stacksave</tt></a> intrinsic executed. This is
				4170	useful for implementing language features like scoped automatic variable sized
				4171	arrays in C99.
				4172	</p>
				4173
				4174	<h5>Semantics:</h5>
				4175
				4176	<p>
				4177	See the description for <a href="#int_stacksave"><tt>llvm.stacksave</tt></a>.
				4178	</p>
				4179
				4180	</div>
				4181
				4182
				4183	<!-- _______________________________________________________________________ -->
				4184	<div class="doc_subsubsection">
				4185	<a name="int_prefetch">'<tt>llvm.prefetch</tt>' Intrinsic</a>
				4186	</div>
				4187
				4188	<div class="doc_text">
				4189
				4190	<h5>Syntax:</h5>
				4191	<pre>
				4192	declare void @llvm.prefetch(i8 * <address>,
				4193	i32 <rw>, i32 <locality>)
				4194	</pre>
				4195
				4196	<h5>Overview:</h5>
				4197
				4198
				4199	<p>
				4200	The '<tt>llvm.prefetch</tt>' intrinsic is a hint to the code generator to insert
				4201	a prefetch instruction if supported; otherwise, it is a noop. Prefetches have
				4202	no
				4203	effect on the behavior of the program but can change its performance
				4204	characteristics.
				4205	</p>
				4206
				4207	<h5>Arguments:</h5>
				4208
				4209	<p>
				4210	<tt>address</tt> is the address to be prefetched, <tt>rw</tt> is the specifier
				4211	determining if the fetch should be for a read (0) or write (1), and
				4212	<tt>locality</tt> is a temporal locality specifier ranging from (0) - no
				4213	locality, to (3) - extremely local keep in cache. The <tt>rw</tt> and
				4214	<tt>locality</tt> arguments must be constant integers.
				4215	</p>
				4216
				4217	<h5>Semantics:</h5>
				4218
				4219	<p>
				4220	This intrinsic does not modify the behavior of the program. In particular,
				4221	prefetches cannot trap and do not produce a value. On targets that support this
				4222	intrinsic, the prefetch can provide hints to the processor cache for better
				4223	performance.
				4224	</p>
				4225
				4226	</div>
				4227
				4228	<!-- _______________________________________________________________________ -->
				4229	<div class="doc_subsubsection">
				4230	<a name="int_pcmarker">'<tt>llvm.pcmarker</tt>' Intrinsic</a>
				4231	</div>
				4232
				4233	<div class="doc_text">
				4234
				4235	<h5>Syntax:</h5>
				4236	<pre>
				4237	declare void @llvm.pcmarker( i32 <id> )
				4238	</pre>
				4239
				4240	<h5>Overview:</h5>
				4241
				4242
				4243	<p>
				4244	The '<tt>llvm.pcmarker</tt>' intrinsic is a method to export a Program Counter
				4245	(PC) in a region of
				4246	code to simulators and other tools. The method is target specific, but it is
				4247	expected that the marker will use exported symbols to transmit the PC of the marker.
				4248	The marker makes no guarantees that it will remain with any specific instruction
				4249	after optimizations. It is possible that the presence of a marker will inhibit
				4250	optimizations. The intended use is to be inserted after optimizations to allow
				4251	correlations of simulation runs.
				4252	</p>
				4253
				4254	<h5>Arguments:</h5>
				4255
				4256	<p>
				4257	<tt>id</tt> is a numerical id identifying the marker.
				4258	</p>
				4259
				4260	<h5>Semantics:</h5>
				4261
				4262	<p>
				4263	This intrinsic does not modify the behavior of the program. Backends that do not
				4264	support this intrinisic may ignore it.
				4265	</p>
				4266
				4267	</div>
				4268
				4269	<!-- _______________________________________________________________________ -->
				4270	<div class="doc_subsubsection">
				4271	<a name="int_readcyclecounter">'<tt>llvm.readcyclecounter</tt>' Intrinsic</a>
				4272	</div>
				4273
				4274	<div class="doc_text">
				4275
				4276	<h5>Syntax:</h5>
				4277	<pre>
				4278	declare i64 @llvm.readcyclecounter( )
				4279	</pre>
				4280
				4281	<h5>Overview:</h5>
				4282
				4283
				4284	<p>
				4285	The '<tt>llvm.readcyclecounter</tt>' intrinsic provides access to the cycle
				4286	counter register (or similar low latency, high accuracy clocks) on those targets
				4287	that support it. On X86, it should map to RDTSC. On Alpha, it should map to RPCC.
				4288	As the backing counters overflow quickly (on the order of 9 seconds on alpha), this
				4289	should only be used for small timings.
				4290	</p>
				4291
				4292	<h5>Semantics:</h5>
				4293
				4294	<p>
				4295	When directly supported, reading the cycle counter should not modify any memory.
				4296	Implementations are allowed to either return a application specific value or a
				4297	system wide value. On backends without support, this is lowered to a constant 0.
				4298	</p>
				4299
				4300	</div>
				4301
				4302	<!-- ======================================================================= -->
				4303	<div class="doc_subsection">
				4304	<a name="int_libc">Standard C Library Intrinsics</a>
				4305	</div>
				4306
				4307	<div class="doc_text">
				4308	<p>
				4309	LLVM provides intrinsics for a few important standard C library functions.
				4310	These intrinsics allow source-language front-ends to pass information about the
				4311	alignment of the pointer arguments to the code generator, providing opportunity
				4312	for more efficient code generation.
				4313	</p>
				4314
				4315	</div>
				4316
				4317	<!-- _______________________________________________________________________ -->
				4318	<div class="doc_subsubsection">
				4319	<a name="int_memcpy">'<tt>llvm.memcpy</tt>' Intrinsic</a>
				4320	</div>
				4321
				4322	<div class="doc_text">
				4323
				4324	<h5>Syntax:</h5>
				4325	<pre>
				4326	declare void @llvm.memcpy.i32(i8 * <dest>, i8 * <src>,
				4327	i32 <len>, i32 <align>)
				4328	declare void @llvm.memcpy.i64(i8 * <dest>, i8 * <src>,
				4329	i64 <len>, i32 <align>)
				4330	</pre>
				4331
				4332	<h5>Overview:</h5>
				4333
				4334	<p>
				4335	The '<tt>llvm.memcpy.*</tt>' intrinsics copy a block of memory from the source
				4336	location to the destination location.
				4337	</p>
				4338
				4339	<p>
				4340	Note that, unlike the standard libc function, the <tt>llvm.memcpy.*</tt>
				4341	intrinsics do not return a value, and takes an extra alignment argument.
				4342	</p>
				4343
				4344	<h5>Arguments:</h5>
				4345
				4346	<p>
				4347	The first argument is a pointer to the destination, the second is a pointer to
				4348	the source. The third argument is an integer argument
				4349	specifying the number of bytes to copy, and the fourth argument is the alignment
				4350	of the source and destination locations.
				4351	</p>
				4352
				4353	<p>
				4354	If the call to this intrinisic has an alignment value that is not 0 or 1, then
				4355	the caller guarantees that both the source and destination pointers are aligned
				4356	to that boundary.
				4357	</p>
				4358
				4359	<h5>Semantics:</h5>
				4360
				4361	<p>
				4362	The '<tt>llvm.memcpy.*</tt>' intrinsics copy a block of memory from the source
				4363	location to the destination location, which are not allowed to overlap. It
				4364	copies "len" bytes of memory over. If the argument is known to be aligned to
				4365	some boundary, this can be specified as the fourth argument, otherwise it should
				4366	be set to 0 or 1.
				4367	</p>
				4368	</div>
				4369
				4370
				4371	<!-- _______________________________________________________________________ -->
				4372	<div class="doc_subsubsection">
				4373	<a name="int_memmove">'<tt>llvm.memmove</tt>' Intrinsic</a>
				4374	</div>
				4375
				4376	<div class="doc_text">
				4377
				4378	<h5>Syntax:</h5>
				4379	<pre>
				4380	declare void @llvm.memmove.i32(i8 * <dest>, i8 * <src>,
				4381	i32 <len>, i32 <align>)
				4382	declare void @llvm.memmove.i64(i8 * <dest>, i8 * <src>,
				4383	i64 <len>, i32 <align>)
				4384	</pre>
				4385
				4386	<h5>Overview:</h5>
				4387
				4388	<p>
				4389	The '<tt>llvm.memmove.*</tt>' intrinsics move a block of memory from the source
				4390	location to the destination location. It is similar to the
				4391	'<tt>llvm.memcmp</tt>' intrinsic but allows the two memory locations to overlap.
				4392	</p>
				4393
				4394	<p>
				4395	Note that, unlike the standard libc function, the <tt>llvm.memmove.*</tt>
				4396	intrinsics do not return a value, and takes an extra alignment argument.
				4397	</p>
				4398
				4399	<h5>Arguments:</h5>
				4400
				4401	<p>
				4402	The first argument is a pointer to the destination, the second is a pointer to
				4403	the source. The third argument is an integer argument
				4404	specifying the number of bytes to copy, and the fourth argument is the alignment
				4405	of the source and destination locations.
				4406	</p>
				4407
				4408	<p>
				4409	If the call to this intrinisic has an alignment value that is not 0 or 1, then
				4410	the caller guarantees that the source and destination pointers are aligned to
				4411	that boundary.
				4412	</p>
				4413
				4414	<h5>Semantics:</h5>
				4415
				4416	<p>
				4417	The '<tt>llvm.memmove.*</tt>' intrinsics copy a block of memory from the source
				4418	location to the destination location, which may overlap. It
				4419	copies "len" bytes of memory over. If the argument is known to be aligned to
				4420	some boundary, this can be specified as the fourth argument, otherwise it should
				4421	be set to 0 or 1.
				4422	</p>
				4423	</div>
				4424
				4425
				4426	<!-- _______________________________________________________________________ -->
				4427	<div class="doc_subsubsection">
				4428	<a name="int_memset">'<tt>llvm.memset.*</tt>' Intrinsics</a>
				4429	</div>
				4430
				4431	<div class="doc_text">
				4432
				4433	<h5>Syntax:</h5>
				4434	<pre>
				4435	declare void @llvm.memset.i32(i8 * <dest>, i8 <val>,
				4436	i32 <len>, i32 <align>)
				4437	declare void @llvm.memset.i64(i8 * <dest>, i8 <val>,
				4438	i64 <len>, i32 <align>)
				4439	</pre>
				4440
				4441	<h5>Overview:</h5>
				4442
				4443	<p>
				4444	The '<tt>llvm.memset.*</tt>' intrinsics fill a block of memory with a particular
				4445	byte value.
				4446	</p>
				4447
				4448	<p>
				4449	Note that, unlike the standard libc function, the <tt>llvm.memset</tt> intrinsic
				4450	does not return a value, and takes an extra alignment argument.
				4451	</p>
				4452
				4453	<h5>Arguments:</h5>
				4454
				4455	<p>
				4456	The first argument is a pointer to the destination to fill, the second is the
				4457	byte value to fill it with, the third argument is an integer
				4458	argument specifying the number of bytes to fill, and the fourth argument is the
				4459	known alignment of destination location.
				4460	</p>
				4461
				4462	<p>
				4463	If the call to this intrinisic has an alignment value that is not 0 or 1, then
				4464	the caller guarantees that the destination pointer is aligned to that boundary.
				4465	</p>
				4466
				4467	<h5>Semantics:</h5>
				4468
				4469	<p>
				4470	The '<tt>llvm.memset.*</tt>' intrinsics fill "len" bytes of memory starting at
				4471	the
				4472	destination location. If the argument is known to be aligned to some boundary,
				4473	this can be specified as the fourth argument, otherwise it should be set to 0 or
				4474	1.
				4475	</p>
				4476	</div>
				4477
				4478
				4479	<!-- _______________________________________________________________________ -->
				4480	<div class="doc_subsubsection">
				4481	<a name="int_sqrt">'<tt>llvm.sqrt.*</tt>' Intrinsic</a>
				4482	</div>
				4483
				4484	<div class="doc_text">
				4485
				4486	<h5>Syntax:</h5>
				4487	<pre>
				4488	declare float @llvm.sqrt.f32(float %Val)
				4489	declare double @llvm.sqrt.f64(double %Val)
				4490	</pre>
				4491
				4492	<h5>Overview:</h5>
				4493
				4494	<p>
				4495	The '<tt>llvm.sqrt</tt>' intrinsics return the sqrt of the specified operand,
				4496	returning the same value as the libm '<tt>sqrt</tt>' function would. Unlike
				4497	<tt>sqrt</tt> in libm, however, <tt>llvm.sqrt</tt> has undefined behavior for
				4498	negative numbers (which allows for better optimization).
				4499	</p>
				4500
				4501	<h5>Arguments:</h5>
				4502
				4503	<p>
				4504	The argument and return value are floating point numbers of the same type.
				4505	</p>
				4506
				4507	<h5>Semantics:</h5>
				4508
				4509	<p>
				4510	This function returns the sqrt of the specified operand if it is a nonnegative
				4511	floating point number.
				4512	</p>
				4513	</div>
				4514
				4515	<!-- _______________________________________________________________________ -->
				4516	<div class="doc_subsubsection">
				4517	<a name="int_powi">'<tt>llvm.powi.*</tt>' Intrinsic</a>
				4518	</div>
				4519
				4520	<div class="doc_text">
				4521
				4522	<h5>Syntax:</h5>
				4523	<pre>
				4524	declare float @llvm.powi.f32(float %Val, i32 %power)
				4525	declare double @llvm.powi.f64(double %Val, i32 %power)
				4526	</pre>
				4527
				4528	<h5>Overview:</h5>
				4529
				4530	<p>
				4531	The '<tt>llvm.powi.*</tt>' intrinsics return the first operand raised to the
				4532	specified (positive or negative) power. The order of evaluation of
				4533	multiplications is not defined.
				4534	</p>
				4535
				4536	<h5>Arguments:</h5>
				4537
				4538	<p>
				4539	The second argument is an integer power, and the first is a value to raise to
				4540	that power.
				4541	</p>
				4542
				4543	<h5>Semantics:</h5>
				4544
				4545	<p>
				4546	This function returns the first value raised to the second power with an
				4547	unspecified sequence of rounding operations.</p>
				4548	</div>
				4549
				4550
				4551	<!-- ======================================================================= -->
				4552	<div class="doc_subsection">
				4553	<a name="int_manip">Bit Manipulation Intrinsics</a>
				4554	</div>
				4555
				4556	<div class="doc_text">
				4557	<p>
				4558	LLVM provides intrinsics for a few important bit manipulation operations.
				4559	These allow efficient code generation for some algorithms.
				4560	</p>
				4561
				4562	</div>
				4563
				4564	<!-- _______________________________________________________________________ -->
				4565	<div class="doc_subsubsection">
				4566	<a name="int_bswap">'<tt>llvm.bswap.*</tt>' Intrinsics</a>
				4567	</div>
				4568
				4569	<div class="doc_text">
				4570
				4571	<h5>Syntax:</h5>
				4572	<p>This is an overloaded intrinsic function. You can use bswap on any integer
Chandler Carruth	a228e39	2007-08-04 01:51:18 +0000	[diff] [blame]	4573	type that is an even number of bytes (i.e. BitWidth % 16 == 0).
Dan Gohman	f17a25c	2007-07-18 16:29:46 +0000	[diff] [blame]	4574	<pre>
Chandler Carruth	a228e39	2007-08-04 01:51:18 +0000	[diff] [blame]	4575	declare i16 @llvm.bswap.i16(i16 <id>)
				4576	declare i32 @llvm.bswap.i32(i32 <id>)
				4577	declare i64 @llvm.bswap.i64(i64 <id>)
Dan Gohman	f17a25c	2007-07-18 16:29:46 +0000	[diff] [blame]	4578	</pre>
				4579
				4580	<h5>Overview:</h5>
				4581
				4582	<p>
				4583	The '<tt>llvm.bswap</tt>' family of intrinsics is used to byte swap integer
				4584	values with an even number of bytes (positive multiple of 16 bits). These are
				4585	useful for performing operations on data that is not in the target's native
				4586	byte order.
				4587	</p>
				4588
				4589	<h5>Semantics:</h5>
				4590
				4591	<p>
Chandler Carruth	a228e39	2007-08-04 01:51:18 +0000	[diff] [blame]	4592	The <tt>llvm.bswap.i16</tt> intrinsic returns an i16 value that has the high
Dan Gohman	f17a25c	2007-07-18 16:29:46 +0000	[diff] [blame]	4593	and low byte of the input i16 swapped. Similarly, the <tt>llvm.bswap.i32</tt>
				4594	intrinsic returns an i32 value that has the four bytes of the input i32
				4595	swapped, so that if the input bytes are numbered 0, 1, 2, 3 then the returned
Chandler Carruth	a228e39	2007-08-04 01:51:18 +0000	[diff] [blame]	4596	i32 will have its bytes in 3, 2, 1, 0 order. The <tt>llvm.bswap.i48</tt>,
				4597	<tt>llvm.bswap.i64</tt> and other intrinsics extend this concept to
Dan Gohman	f17a25c	2007-07-18 16:29:46 +0000	[diff] [blame]	4598	additional even-byte lengths (6 bytes, 8 bytes and more, respectively).
				4599	</p>
				4600
				4601	</div>
				4602
				4603	<!-- _______________________________________________________________________ -->
				4604	<div class="doc_subsubsection">
				4605	<a name="int_ctpop">'<tt>llvm.ctpop.*</tt>' Intrinsic</a>
				4606	</div>
				4607
				4608	<div class="doc_text">
				4609
				4610	<h5>Syntax:</h5>
				4611	<p>This is an overloaded intrinsic. You can use llvm.ctpop on any integer bit
				4612	width. Not all targets support all bit widths however.
				4613	<pre>
Chandler Carruth	a228e39	2007-08-04 01:51:18 +0000	[diff] [blame]	4614	declare i8 @llvm.ctpop.i8 (i8 <src>)
				4615	declare i16 @llvm.ctpop.i16(i16 <src>)
Dan Gohman	f17a25c	2007-07-18 16:29:46 +0000	[diff] [blame]	4616	declare i32 @llvm.ctpop.i32(i32 <src>)
Chandler Carruth	a228e39	2007-08-04 01:51:18 +0000	[diff] [blame]	4617	declare i64 @llvm.ctpop.i64(i64 <src>)
				4618	declare i256 @llvm.ctpop.i256(i256 <src>)
Dan Gohman	f17a25c	2007-07-18 16:29:46 +0000	[diff] [blame]	4619	</pre>
				4620
				4621	<h5>Overview:</h5>
				4622
				4623	<p>
				4624	The '<tt>llvm.ctpop</tt>' family of intrinsics counts the number of bits set in a
				4625	value.
				4626	</p>
				4627
				4628	<h5>Arguments:</h5>
				4629
				4630	<p>
				4631	The only argument is the value to be counted. The argument may be of any
				4632	integer type. The return type must match the argument type.
				4633	</p>
				4634
				4635	<h5>Semantics:</h5>
				4636
				4637	<p>
				4638	The '<tt>llvm.ctpop</tt>' intrinsic counts the 1's in a variable.
				4639	</p>
				4640	</div>
				4641
				4642	<!-- _______________________________________________________________________ -->
				4643	<div class="doc_subsubsection">
				4644	<a name="int_ctlz">'<tt>llvm.ctlz.*</tt>' Intrinsic</a>
				4645	</div>
				4646
				4647	<div class="doc_text">
				4648
				4649	<h5>Syntax:</h5>
				4650	<p>This is an overloaded intrinsic. You can use <tt>llvm.ctlz</tt> on any
				4651	integer bit width. Not all targets support all bit widths however.
				4652	<pre>
Chandler Carruth	a228e39	2007-08-04 01:51:18 +0000	[diff] [blame]	4653	declare i8 @llvm.ctlz.i8 (i8 <src>)
				4654	declare i16 @llvm.ctlz.i16(i16 <src>)
Dan Gohman	f17a25c	2007-07-18 16:29:46 +0000	[diff] [blame]	4655	declare i32 @llvm.ctlz.i32(i32 <src>)
Chandler Carruth	a228e39	2007-08-04 01:51:18 +0000	[diff] [blame]	4656	declare i64 @llvm.ctlz.i64(i64 <src>)
				4657	declare i256 @llvm.ctlz.i256(i256 <src>)
Dan Gohman	f17a25c	2007-07-18 16:29:46 +0000	[diff] [blame]	4658	</pre>
				4659
				4660	<h5>Overview:</h5>
				4661
				4662	<p>
				4663	The '<tt>llvm.ctlz</tt>' family of intrinsic functions counts the number of
				4664	leading zeros in a variable.
				4665	</p>
				4666
				4667	<h5>Arguments:</h5>
				4668
				4669	<p>
				4670	The only argument is the value to be counted. The argument may be of any
				4671	integer type. The return type must match the argument type.
				4672	</p>
				4673
				4674	<h5>Semantics:</h5>
				4675
				4676	<p>
				4677	The '<tt>llvm.ctlz</tt>' intrinsic counts the leading (most significant) zeros
				4678	in a variable. If the src == 0 then the result is the size in bits of the type
				4679	of src. For example, <tt>llvm.ctlz(i32 2) = 30</tt>.
				4680	</p>
				4681	</div>
				4682
				4683
				4684
				4685	<!-- _______________________________________________________________________ -->
				4686	<div class="doc_subsubsection">
				4687	<a name="int_cttz">'<tt>llvm.cttz.*</tt>' Intrinsic</a>
				4688	</div>
				4689
				4690	<div class="doc_text">
				4691
				4692	<h5>Syntax:</h5>
				4693	<p>This is an overloaded intrinsic. You can use <tt>llvm.cttz</tt> on any
				4694	integer bit width. Not all targets support all bit widths however.
				4695	<pre>
Chandler Carruth	a228e39	2007-08-04 01:51:18 +0000	[diff] [blame]	4696	declare i8 @llvm.cttz.i8 (i8 <src>)
				4697	declare i16 @llvm.cttz.i16(i16 <src>)
Dan Gohman	f17a25c	2007-07-18 16:29:46 +0000	[diff] [blame]	4698	declare i32 @llvm.cttz.i32(i32 <src>)
Chandler Carruth	a228e39	2007-08-04 01:51:18 +0000	[diff] [blame]	4699	declare i64 @llvm.cttz.i64(i64 <src>)
				4700	declare i256 @llvm.cttz.i256(i256 <src>)
Dan Gohman	f17a25c	2007-07-18 16:29:46 +0000	[diff] [blame]	4701	</pre>
				4702
				4703	<h5>Overview:</h5>
				4704
				4705	<p>
				4706	The '<tt>llvm.cttz</tt>' family of intrinsic functions counts the number of
				4707	trailing zeros.
				4708	</p>
				4709
				4710	<h5>Arguments:</h5>
				4711
				4712	<p>
				4713	The only argument is the value to be counted. The argument may be of any
				4714	integer type. The return type must match the argument type.
				4715	</p>
				4716
				4717	<h5>Semantics:</h5>
				4718
				4719	<p>
				4720	The '<tt>llvm.cttz</tt>' intrinsic counts the trailing (least significant) zeros
				4721	in a variable. If the src == 0 then the result is the size in bits of the type
				4722	of src. For example, <tt>llvm.cttz(2) = 1</tt>.
				4723	</p>
				4724	</div>
				4725
				4726	<!-- _______________________________________________________________________ -->
				4727	<div class="doc_subsubsection">
				4728	<a name="int_part_select">'<tt>llvm.part.select.*</tt>' Intrinsic</a>
				4729	</div>
				4730
				4731	<div class="doc_text">
				4732
				4733	<h5>Syntax:</h5>
				4734	<p>This is an overloaded intrinsic. You can use <tt>llvm.part.select</tt>
				4735	on any integer bit width.
				4736	<pre>
Chandler Carruth	a228e39	2007-08-04 01:51:18 +0000	[diff] [blame]	4737	declare i17 @llvm.part.select.i17 (i17 %val, i32 %loBit, i32 %hiBit)
				4738	declare i29 @llvm.part.select.i29 (i29 %val, i32 %loBit, i32 %hiBit)
Dan Gohman	f17a25c	2007-07-18 16:29:46 +0000	[diff] [blame]	4739	</pre>
				4740
				4741	<h5>Overview:</h5>
				4742	<p>The '<tt>llvm.part.select</tt>' family of intrinsic functions selects a
				4743	range of bits from an integer value and returns them in the same bit width as
				4744	the original value.</p>
				4745
				4746	<h5>Arguments:</h5>
				4747	<p>The first argument, <tt>%val</tt> and the result may be integer types of
				4748	any bit width but they must have the same bit width. The second and third
				4749	arguments must be <tt>i32</tt> type since they specify only a bit index.</p>
				4750
				4751	<h5>Semantics:</h5>
				4752	<p>The operation of the '<tt>llvm.part.select</tt>' intrinsic has two modes
				4753	of operation: forwards and reverse. If <tt>%loBit</tt> is greater than
				4754	<tt>%hiBits</tt> then the intrinsic operates in reverse mode. Otherwise it
				4755	operates in forward mode.</p>
				4756	<p>In forward mode, this intrinsic is the equivalent of shifting <tt>%val</tt>
				4757	right by <tt>%loBit</tt> bits and then ANDing it with a mask with
				4758	only the <tt>%hiBit - %loBit</tt> bits set, as follows:</p>
				4759	<ol>
				4760	<li>The <tt>%val</tt> is shifted right (LSHR) by the number of bits specified
				4761	by <tt>%loBits</tt>. This normalizes the value to the low order bits.</li>
				4762	<li>The <tt>%loBits</tt> value is subtracted from the <tt>%hiBits</tt> value
				4763	to determine the number of bits to retain.</li>
				4764	<li>A mask of the retained bits is created by shifting a -1 value.</li>
				4765	<li>The mask is ANDed with <tt>%val</tt> to produce the result.
				4766	</ol>
				4767	<p>In reverse mode, a similar computation is made except that the bits are
				4768	returned in the reverse order. So, for example, if <tt>X</tt> has the value
				4769	<tt>i16 0x0ACF (101011001111)</tt> and we apply
				4770	<tt>part.select(i16 X, 8, 3)</tt> to it, we get back the value
				4771	<tt>i16 0x0026 (000000100110)</tt>.</p>
				4772	</div>
				4773
				4774	<div class="doc_subsubsection">
				4775	<a name="int_part_set">'<tt>llvm.part.set.*</tt>' Intrinsic</a>
				4776	</div>
				4777
				4778	<div class="doc_text">
				4779
				4780	<h5>Syntax:</h5>
				4781	<p>This is an overloaded intrinsic. You can use <tt>llvm.part.set</tt>
				4782	on any integer bit width.
				4783	<pre>
Chandler Carruth	a228e39	2007-08-04 01:51:18 +0000	[diff] [blame]	4784	declare i17 @llvm.part.set.i17.i9 (i17 %val, i9 %repl, i32 %lo, i32 %hi)
				4785	declare i29 @llvm.part.set.i29.i9 (i29 %val, i9 %repl, i32 %lo, i32 %hi)
Dan Gohman	f17a25c	2007-07-18 16:29:46 +0000	[diff] [blame]	4786	</pre>
				4787
				4788	<h5>Overview:</h5>
				4789	<p>The '<tt>llvm.part.set</tt>' family of intrinsic functions replaces a range
				4790	of bits in an integer value with another integer value. It returns the integer
				4791	with the replaced bits.</p>
				4792
				4793	<h5>Arguments:</h5>
				4794	<p>The first argument, <tt>%val</tt> and the result may be integer types of
				4795	any bit width but they must have the same bit width. <tt>%val</tt> is the value
				4796	whose bits will be replaced. The second argument, <tt>%repl</tt> may be an
				4797	integer of any bit width. The third and fourth arguments must be <tt>i32</tt>
				4798	type since they specify only a bit index.</p>
				4799
				4800	<h5>Semantics:</h5>
				4801	<p>The operation of the '<tt>llvm.part.set</tt>' intrinsic has two modes
				4802	of operation: forwards and reverse. If <tt>%lo</tt> is greater than
				4803	<tt>%hi</tt> then the intrinsic operates in reverse mode. Otherwise it
				4804	operates in forward mode.</p>
				4805	<p>For both modes, the <tt>%repl</tt> value is prepared for use by either
				4806	truncating it down to the size of the replacement area or zero extending it
				4807	up to that size.</p>
				4808	<p>In forward mode, the bits between <tt>%lo</tt> and <tt>%hi</tt> (inclusive)
				4809	are replaced with corresponding bits from <tt>%repl</tt>. That is the 0th bit
				4810	in <tt>%repl</tt> replaces the <tt>%lo</tt>th bit in <tt>%val</tt> and etc. up
				4811	to the <tt>%hi</tt>th bit.
				4812	<p>In reverse mode, a similar computation is made except that the bits are
				4813	reversed. That is, the <tt>0</tt>th bit in <tt>%repl</tt> replaces the
				4814	<tt>%hi</tt> bit in <tt>%val</tt> and etc. down to the <tt>%lo</tt>th bit.
				4815	<h5>Examples:</h5>
				4816	<pre>
				4817	llvm.part.set(0xFFFF, 0, 4, 7) -> 0xFF0F
				4818	llvm.part.set(0xFFFF, 0, 7, 4) -> 0xFF0F
				4819	llvm.part.set(0xFFFF, 1, 7, 4) -> 0xFF8F
				4820	llvm.part.set(0xFFFF, F, 8, 3) -> 0xFFE7
				4821	llvm.part.set(0xFFFF, 0, 3, 8) -> 0xFE07
				4822	</pre>
				4823	</div>
				4824
				4825	<!-- ======================================================================= -->
				4826	<div class="doc_subsection">
				4827	<a name="int_debugger">Debugger Intrinsics</a>
				4828	</div>
				4829
				4830	<div class="doc_text">
				4831	<p>
				4832	The LLVM debugger intrinsics (which all start with <tt>llvm.dbg.</tt> prefix),
				4833	are described in the <a
				4834	href="SourceLevelDebugging.html#format_common_intrinsics">LLVM Source Level
				4835	Debugging</a> document.
				4836	</p>
				4837	</div>
				4838
				4839
				4840	<!-- ======================================================================= -->
				4841	<div class="doc_subsection">
				4842	<a name="int_eh">Exception Handling Intrinsics</a>
				4843	</div>
				4844
				4845	<div class="doc_text">
				4846	<p> The LLVM exception handling intrinsics (which all start with
				4847	<tt>llvm.eh.</tt> prefix), are described in the <a
				4848	href="ExceptionHandling.html#format_common_intrinsics">LLVM Exception
				4849	Handling</a> document. </p>
				4850	</div>
				4851
				4852	<!-- ======================================================================= -->
				4853	<div class="doc_subsection">
Chandler Carruth	c0470a8	2007-07-20 19:34:37 +0000	[diff] [blame]	4854	<a name="int_atomics">Atomic Operations and Synchronization Intrinsics</a>
				4855	</div>
				4856
				4857	<div class="doc_text">
				4858	<p>
				4859	These intrinsic functions expand the "universal IR" of LLVM to represent
				4860	hardware constructs for atomic operations and memory synchronization. This
				4861	provides an interface to the hardware, not an interface to the programmer. It
				4862	is aimed at a low enough level to allow any programming models or APIs which
				4863	need atomic behaviors to map cleanly onto it. It is also modeled primarily on
Chandler Carruth	3060d9a	2007-07-20 20:14:52 +0000	[diff] [blame]	4864	hardware behavior. Just as hardware provides a "universal IR" for source
Chandler Carruth	c0470a8	2007-07-20 19:34:37 +0000	[diff] [blame]	4865	languages, it also provides a starting point for developing a "universal"
				4866	atomic operation and synchronization IR.
				4867	</p>
				4868	<p>
				4869	These do <em>not</em> form an API such as high-level threading libraries,
				4870	software transaction memory systems, atomic primitives, and intrinsic
Reid Spencer	b043f67	2007-07-20 19:59:11 +0000	[diff] [blame]	4871	functions as found in BSD, GNU libc, atomic_ops, APR, and other system and
Chandler Carruth	c0470a8	2007-07-20 19:34:37 +0000	[diff] [blame]	4872	application libraries. The hardware interface provided by LLVM should allow
				4873	a clean implementation of all of these APIs and parallel programming models.
				4874	No one model or paradigm should be selected above others unless the hardware
				4875	itself ubiquitously does so.
				4876	</p>
				4877	</div>
				4878
				4879	<!-- _______________________________________________________________________ -->
				4880	<div class="doc_subsubsection">
				4881	<a name="int_lcs">'<tt>llvm.atomic.lcs.*</tt>' Intrinsic</a>
				4882	</div>
				4883	<div class="doc_text">
				4884	<h5>Syntax:</h5>
				4885	<p>
				4886	This is an overloaded intrinsic. You can use <tt>llvm.atomic.lcs</tt> on any
Reid Spencer	b043f67	2007-07-20 19:59:11 +0000	[diff] [blame]	4887	integer bit width. Not all targets support all bit widths however.</p>
Chandler Carruth	c0470a8	2007-07-20 19:34:37 +0000	[diff] [blame]	4888	<pre>
				4889	declare i8 @llvm.atomic.lcs.i8.i8p.i8.i8( i8* <ptr>, i8 <cmp>, i8 <val> )
				4890	declare i16 @llvm.atomic.lcs.i16.i16p.i16.i16( i16* <ptr>, i16 <cmp>, i16 <val> )
				4891	declare i32 @llvm.atomic.lcs.i32.i32p.i32.i32( i32* <ptr>, i32 <cmp>, i32 <val> )
				4892	declare i64 @llvm.atomic.lcs.i64.i64p.i64.i64( i64* <ptr>, i64 <cmp>, i64 <val> )
				4893	</pre>
Chandler Carruth	c0470a8	2007-07-20 19:34:37 +0000	[diff] [blame]	4894	<h5>Overview:</h5>
				4895	<p>
Chandler Carruth	3060d9a	2007-07-20 20:14:52 +0000	[diff] [blame]	4896	This loads a value in memory and compares it to a given value. If they are
				4897	equal, it stores a new value into the memory.
Chandler Carruth	c0470a8	2007-07-20 19:34:37 +0000	[diff] [blame]	4898	</p>
				4899	<h5>Arguments:</h5>
				4900	<p>
				4901	The <tt>llvm.atomic.lcs</tt> intrinsic takes three arguments. The result as
				4902	well as both <tt>cmp</tt> and <tt>val</tt> must be integer values with the
				4903	same bit width. The <tt>ptr</tt> argument must be a pointer to a value of
				4904	this integer type. While any bit width integer may be used, targets may only
				4905	lower representations they support in hardware.
				4906	</p>
				4907	<h5>Semantics:</h5>
				4908	<p>
				4909	This entire intrinsic must be executed atomically. It first loads the value
Chandler Carruth	3060d9a	2007-07-20 20:14:52 +0000	[diff] [blame]	4910	in memory pointed to by <tt>ptr</tt> and compares it with the value
				4911	<tt>cmp</tt>. If they are equal, <tt>val</tt> is stored into the memory. The
				4912	loaded value is yielded in all cases. This provides the equivalent of an
				4913	atomic compare-and-swap operation within the SSA framework.
Chandler Carruth	c0470a8	2007-07-20 19:34:37 +0000	[diff] [blame]	4914	</p>
				4915	<h5>Examples:</h5>
				4916	<pre>
				4917	%ptr = malloc i32
				4918	store i32 4, %ptr
				4919
				4920	%val1 = add i32 4, 4
				4921	%result1 = call i32 @llvm.atomic.lcs( i32* %ptr, i32 4, %val1 )
				4922	<i>; yields {i32}:result1 = 4</i>
				4923	%stored1 = icmp eq i32 %result1, 4 <i>; yields {i1}:stored1 = true</i>
				4924	%memval1 = load i32* %ptr <i>; yields {i32}:memval1 = 8</i>
				4925
				4926	%val2 = add i32 1, 1
				4927	%result2 = call i32 @llvm.atomic.lcs( i32* %ptr, i32 5, %val2 )
				4928	<i>; yields {i32}:result2 = 8</i>
				4929	%stored2 = icmp eq i32 %result2, 5 <i>; yields {i1}:stored2 = false</i>
				4930	%memval2 = load i32* %ptr <i>; yields {i32}:memval2 = 8</i>
				4931	</pre>
				4932	</div>
				4933
				4934	<!-- _______________________________________________________________________ -->
				4935	<div class="doc_subsubsection">
				4936	<a name="int_ls">'<tt>llvm.atomic.ls.*</tt>' Intrinsic</a>
				4937	</div>
				4938	<div class="doc_text">
				4939	<h5>Syntax:</h5>
				4940	<p>
				4941	This is an overloaded intrinsic. You can use <tt>llvm.atomic.ls</tt> on any
Reid Spencer	b043f67	2007-07-20 19:59:11 +0000	[diff] [blame]	4942	integer bit width. Not all targets support all bit widths however.</p>
Chandler Carruth	c0470a8	2007-07-20 19:34:37 +0000	[diff] [blame]	4943	<pre>
				4944	declare i8 @llvm.atomic.ls.i8.i8p.i8( i8* <ptr>, i8 <val> )
				4945	declare i16 @llvm.atomic.ls.i16.i16p.i16( i16* <ptr>, i16 <val> )
				4946	declare i32 @llvm.atomic.ls.i32.i32p.i32( i32* <ptr>, i32 <val> )
				4947	declare i64 @llvm.atomic.ls.i64.i64p.i64( i64* <ptr>, i64 <val> )
				4948	</pre>
Chandler Carruth	c0470a8	2007-07-20 19:34:37 +0000	[diff] [blame]	4949	<h5>Overview:</h5>
				4950	<p>
Chandler Carruth	3060d9a	2007-07-20 20:14:52 +0000	[diff] [blame]	4951	This intrinsic loads the value stored in memory at <tt>ptr</tt> and yields
				4952	the value from memory. It then stores the value in <tt>val</tt> in the memory
				4953	at <tt>ptr</tt>.
Chandler Carruth	c0470a8	2007-07-20 19:34:37 +0000	[diff] [blame]	4954	</p>
				4955	<h5>Arguments:</h5>
				4956	<p>
				4957	The <tt>llvm.atomic.ls</tt> intrinsic takes two arguments. Both the
				4958	<tt>val</tt> argument and the result must be integers of the same bit width.
				4959	The first argument, <tt>ptr</tt>, must be a pointer to a value of this
				4960	integer type. The targets may only lower integer representations they
				4961	support.
				4962	</p>
				4963	<h5>Semantics:</h5>
				4964	<p>
				4965	This intrinsic loads the value pointed to by <tt>ptr</tt>, yields it, and
				4966	stores <tt>val</tt> back into <tt>ptr</tt> atomically. This provides the
				4967	equivalent of an atomic swap operation within the SSA framework.
				4968	</p>
				4969	<h5>Examples:</h5>
				4970	<pre>
				4971	%ptr = malloc i32
				4972	store i32 4, %ptr
				4973
				4974	%val1 = add i32 4, 4
				4975	%result1 = call i32 @llvm.atomic.ls( i32* %ptr, i32 %val1 )
				4976	<i>; yields {i32}:result1 = 4</i>
				4977	%stored1 = icmp eq i32 %result1, 4 <i>; yields {i1}:stored1 = true</i>
				4978	%memval1 = load i32* %ptr <i>; yields {i32}:memval1 = 8</i>
				4979
				4980	%val2 = add i32 1, 1
				4981	%result2 = call i32 @llvm.atomic.ls( i32* %ptr, i32 %val2 )
				4982	<i>; yields {i32}:result2 = 8</i>
				4983	%stored2 = icmp eq i32 %result2, 8 <i>; yields {i1}:stored2 = true</i>
				4984	%memval2 = load i32* %ptr <i>; yields {i32}:memval2 = 2</i>
				4985	</pre>
				4986	</div>
				4987
				4988	<!-- _______________________________________________________________________ -->
				4989	<div class="doc_subsubsection">
				4990	<a name="int_las">'<tt>llvm.atomic.las.*</tt>' Intrinsic</a>
				4991	</div>
				4992	<div class="doc_text">
				4993	<h5>Syntax:</h5>
				4994	<p>
				4995	This is an overloaded intrinsic. You can use <tt>llvm.atomic.las</tt> on any
Reid Spencer	b043f67	2007-07-20 19:59:11 +0000	[diff] [blame]	4996	integer bit width. Not all targets support all bit widths however.</p>
Chandler Carruth	c0470a8	2007-07-20 19:34:37 +0000	[diff] [blame]	4997	<pre>
				4998	declare i8 @llvm.atomic.las.i8.i8p.i8( i8* <ptr>, i8 <delta> )
				4999	declare i16 @llvm.atomic.las.i16.i16p.i16( i16* <ptr>, i16 <delta> )
				5000	declare i32 @llvm.atomic.las.i32.i32p.i32( i32* <ptr>, i32 <delta> )
				5001	declare i64 @llvm.atomic.las.i64.i64p.i64( i64* <ptr>, i64 <delta> )
				5002	</pre>
Chandler Carruth	c0470a8	2007-07-20 19:34:37 +0000	[diff] [blame]	5003	<h5>Overview:</h5>
				5004	<p>
Chandler Carruth	3060d9a	2007-07-20 20:14:52 +0000	[diff] [blame]	5005	This intrinsic adds <tt>delta</tt> to the value stored in memory at
Chandler Carruth	c0470a8	2007-07-20 19:34:37 +0000	[diff] [blame]	5006	<tt>ptr</tt>. It yields the original value at <tt>ptr</tt>.
				5007	</p>
				5008	<h5>Arguments:</h5>
				5009	<p>
				5010	The intrinsic takes two arguments, the first a pointer to an integer value
				5011	and the second an integer value. The result is also an integer value. These
				5012	integer types can have any bit width, but they must all have the same bit
				5013	width. The targets may only lower integer representations they support.
				5014	</p>
				5015	<h5>Semantics:</h5>
				5016	<p>
				5017	This intrinsic does a series of operations atomically. It first loads the
				5018	value stored at <tt>ptr</tt>. It then adds <tt>delta</tt>, stores the result
				5019	to <tt>ptr</tt>. It yields the original value stored at <tt>ptr</tt>.
				5020	</p>
				5021	<h5>Examples:</h5>
				5022	<pre>
				5023	%ptr = malloc i32
				5024	store i32 4, %ptr
				5025	%result1 = call i32 @llvm.atomic.las( i32* %ptr, i32 4 )
				5026	<i>; yields {i32}:result1 = 4</i>
				5027	%result2 = call i32 @llvm.atomic.las( i32* %ptr, i32 2 )
				5028	<i>; yields {i32}:result2 = 8</i>
				5029	%result3 = call i32 @llvm.atomic.las( i32* %ptr, i32 5 )
				5030	<i>; yields {i32}:result3 = 10</i>
				5031	%memval = load i32* %ptr <i>; yields {i32}:memval1 = 15</i>
				5032	</pre>
				5033	</div>
				5034
				5035	<!-- _______________________________________________________________________ -->
				5036	<div class="doc_subsubsection">
				5037	<a name="int_lss">'<tt>llvm.atomic.lss.*</tt>' Intrinsic</a>
				5038	</div>
				5039	<div class="doc_text">
				5040	<h5>Syntax:</h5>
				5041	<p>
				5042	This is an overloaded intrinsic. You can use <tt>llvm.atomic.lss</tt> on any
Reid Spencer	b043f67	2007-07-20 19:59:11 +0000	[diff] [blame]	5043	integer bit width. Not all targets support all bit widths however.</p>
Chandler Carruth	c0470a8	2007-07-20 19:34:37 +0000	[diff] [blame]	5044	<pre>
				5045	declare i8 @llvm.atomic.lss.i8.i8.i8( i8* <ptr>, i8 <delta> )
				5046	declare i16 @llvm.atomic.lss.i16.i16.i16( i16* <ptr>, i16 <delta> )
				5047	declare i32 @llvm.atomic.lss.i32.i32.i32( i32* <ptr>, i32 <delta> )
				5048	declare i64 @llvm.atomic.lss.i64.i64.i64( i64* <ptr>, i64 <delta> )
				5049	</pre>
Chandler Carruth	c0470a8	2007-07-20 19:34:37 +0000	[diff] [blame]	5050	<h5>Overview:</h5>
				5051	<p>
Chandler Carruth	3060d9a	2007-07-20 20:14:52 +0000	[diff] [blame]	5052	This intrinsic subtracts <tt>delta</tt> from the value stored in memory at
				5053	<tt>ptr</tt>. It yields the original value at <tt>ptr</tt>.
Chandler Carruth	c0470a8	2007-07-20 19:34:37 +0000	[diff] [blame]	5054	</p>
				5055	<h5>Arguments:</h5>
				5056	<p>
				5057	The intrinsic takes two arguments, the first a pointer to an integer value
				5058	and the second an integer value. The result is also an integer value. These
				5059	integer types can have any bit width, but they must all have the same bit
				5060	width. The targets may only lower integer representations they support.
				5061	</p>
				5062	<h5>Semantics:</h5>
				5063	<p>
				5064	This intrinsic does a series of operations atomically. It first loads the
				5065	value stored at <tt>ptr</tt>. It then subtracts <tt>delta</tt>,
				5066	stores the result to <tt>ptr</tt>. It yields the original value stored
				5067	at <tt>ptr</tt>.
				5068	</p>
				5069	<h5>Examples:</h5>
				5070	<pre>
				5071	%ptr = malloc i32
				5072	store i32 32, %ptr
				5073	%result1 = call i32 @llvm.atomic.lss( i32* %ptr, i32 4 )
				5074	<i>; yields {i32}:result1 = 32</i>
				5075	%result2 = call i32 @llvm.atomic.lss( i32* %ptr, i32 2 )
				5076	<i>; yields {i32}:result2 = 28</i>
				5077	%result3 = call i32 @llvm.atomic.lss( i32* %ptr, i32 5 )
				5078	<i>; yields {i32}:result3 = 26</i>
				5079	%memval = load i32* %ptr <i>; yields {i32}:memval1 = 21</i>
				5080	</pre>
				5081	</div>
				5082
				5083	<!-- _______________________________________________________________________ -->
				5084	<div class="doc_subsubsection">
				5085	<a name="int_memory_barrier">'<tt>llvm.memory.barrier</tt>' Intrinsic</a>
				5086	</div>
				5087	<div class="doc_text">
				5088	<h5>Syntax:</h5>
Chandler Carruth	c0470a8	2007-07-20 19:34:37 +0000	[diff] [blame]	5089	<pre>
				5090	declare void @llvm.memory.barrier( i1 <ll>, i1 <ls>, i1 <sl>, i1 <ss> )
				5091	</pre>
Chandler Carruth	c0470a8	2007-07-20 19:34:37 +0000	[diff] [blame]	5092	<h5>Overview:</h5>
				5093	<p>
				5094	The <tt>llvm.memory.barrier</tt> intrinsic guarantees ordering between
				5095	specific pairs of memory access types.
				5096	</p>
				5097	<h5>Arguments:</h5>
				5098	<p>
				5099	The <tt>llvm.memory.barrier</tt> intrinsic requires four boolean arguments.
				5100	Each argument enables a specific barrier as listed below.
Reid Spencer	c64165c	2007-07-20 20:03:33 +0000	[diff] [blame]	5101	</p>
Chandler Carruth	c0470a8	2007-07-20 19:34:37 +0000	[diff] [blame]	5102	<ul>
				5103	<li><tt>ll</tt>: load-load barrier</li>
				5104	<li><tt>ls</tt>: load-store barrier</li>
				5105	<li><tt>sl</tt>: store-load barrier</li>
				5106	<li><tt>ss</tt>: store-store barrier</li>
				5107	</ul>
Chandler Carruth	c0470a8	2007-07-20 19:34:37 +0000	[diff] [blame]	5108	<h5>Semantics:</h5>
				5109	<p>
				5110	This intrinsic causes the system to enforce some ordering constraints upon
				5111	the loads and stores of the program. This barrier does not indicate
				5112	<em>when</em> any events will occur, it only enforces an <em>order</em> in
				5113	which they occur. For any of the specified pairs of load and store operations
				5114	(f.ex. load-load, or store-load), all of the first operations preceding the
				5115	barrier will complete before any of the second operations succeeding the
				5116	barrier begin. Specifically the semantics for each pairing is as follows:
Reid Spencer	c64165c	2007-07-20 20:03:33 +0000	[diff] [blame]	5117	</p>
Chandler Carruth	c0470a8	2007-07-20 19:34:37 +0000	[diff] [blame]	5118	<ul>
				5119	<li><tt>ll</tt>: All loads before the barrier must complete before any load
				5120	after the barrier begins.</li>
				5121	<li><tt>ls</tt>: All loads before the barrier must complete before any
				5122	store after the barrier begins.</li>
				5123	<li><tt>ss</tt>: All stores before the barrier must complete before any
				5124	store after the barrier begins.</li>
				5125	<li><tt>sl</tt>: All stores before the barrier must complete before any
				5126	load after the barrier begins.</li>
				5127	</ul>
Reid Spencer	c64165c	2007-07-20 20:03:33 +0000	[diff] [blame]	5128	<p>
Chandler Carruth	c0470a8	2007-07-20 19:34:37 +0000	[diff] [blame]	5129	These semantics are applied with a logical "and" behavior when more than one
				5130	is enabled in a single memory barrier intrinsic.
				5131	</p>
				5132	<h5>Example:</h5>
				5133	<pre>
				5134	%ptr = malloc i32
				5135	store i32 4, %ptr
				5136
				5137	%result1 = load i32* %ptr <i>; yields {i32}:result1 = 4</i>
				5138	call void @llvm.memory.barrier( i1 false, i1 true, i1 false, i1 false )
				5139	<i>; guarantee the above finishes</i>
				5140	store i32 8, %ptr <i>; before this begins</i>
				5141	</pre>
				5142	</div>
				5143
				5144	<!-- ======================================================================= -->
				5145	<div class="doc_subsection">
Duncan Sands	38947cd	2007-07-27 12:58:54 +0000	[diff] [blame]	5146	<a name="int_trampoline">Trampoline Intrinsics</a>
				5147	</div>
				5148
				5149	<div class="doc_text">
				5150	<p>
				5151	These intrinsics make it possible to excise one parameter, marked with
				5152	the <tt>nest</tt> attribute, from a function. The result is a callable
				5153	function pointer lacking the nest parameter - the caller does not need
				5154	to provide a value for it. Instead, the value to use is stored in
				5155	advance in a "trampoline", a block of memory usually allocated
				5156	on the stack, which also contains code to splice the nest value into the
				5157	argument list. This is used to implement the GCC nested function address
				5158	extension.
				5159	</p>
				5160	<p>
				5161	For example, if the function is
				5162	<tt>i32 f(i8* nest %c, i32 %x, i32 %y)</tt> then the resulting function
				5163	pointer has signature <tt>i32 (i32, i32)*</tt>. It can be created as follows:
				5164	<pre>
				5165	%tramp1 = alloca [10 x i8], align 4 ; size and alignment only correct for X86
				5166	%tramp = getelementptr [10 x i8]* %tramp1, i32 0, i32 0
				5167	call void @llvm.init.trampoline( i8* %tramp, i8* bitcast (i32 (i8* nest , i32, i32)* @f to i8), i8 %nval )
				5168	%adj = call i8* @llvm.adjust.trampoline( i8* %tramp )
				5169	%fp = bitcast i8* %adj to i32 (i32, i32)*
				5170	</pre>
				5171	The call <tt>%val = call i32 %fp( i32 %x, i32 %y )</tt> is then equivalent to
				5172	<tt>%val = call i32 %f( i8* %nval, i32 %x, i32 %y )</tt>.
				5173	</p>
Duncan Sands	38947cd	2007-07-27 12:58:54 +0000	[diff] [blame]	5174	</div>
				5175
				5176	<!-- _______________________________________________________________________ -->
				5177	<div class="doc_subsubsection">
				5178	<a name="int_it">'<tt>llvm.init.trampoline</tt>' Intrinsic</a>
				5179	</div>
				5180	<div class="doc_text">
				5181	<h5>Syntax:</h5>
				5182	<pre>
				5183	declare void @llvm.init.trampoline(i8* <tramp>, i8* <func>, i8* <nval>)
				5184	</pre>
				5185	<h5>Overview:</h5>
				5186	<p>
				5187	This initializes the memory pointed to by <tt>tramp</tt> as a trampoline.
				5188	</p>
				5189	<h5>Arguments:</h5>
				5190	<p>
				5191	The <tt>llvm.init.trampoline</tt> intrinsic takes three arguments, all
				5192	pointers. The <tt>tramp</tt> argument must point to a sufficiently large
				5193	and sufficiently aligned block of memory; this memory is written to by the
Duncan Sands	3501221	2007-08-22 23:39:54 +0000	[diff] [blame]	5194	intrinsic. Note that the size and the alignment are target-specific - LLVM
				5195	currently provides no portable way of determining them, so a front-end that
				5196	generates this intrinsic needs to have some target-specific knowledge.
				5197	The <tt>func</tt> argument must hold a function bitcast to an <tt>i8*</tt>.
Duncan Sands	38947cd	2007-07-27 12:58:54 +0000	[diff] [blame]	5198	</p>
				5199	<h5>Semantics:</h5>
				5200	<p>
				5201	The block of memory pointed to by <tt>tramp</tt> is filled with target
				5202	dependent code, turning it into a function.
				5203	The new function's signature is the same as that of <tt>func</tt> with
				5204	any arguments marked with the <tt>nest</tt> attribute removed. At most
				5205	one such <tt>nest</tt> argument is allowed, and it must be of pointer
				5206	type. Calling the new function is equivalent to calling <tt>func</tt>
				5207	with the same argument list, but with <tt>nval</tt> used for the missing
				5208	<tt>nest</tt> argument.
				5209	</p>
				5210	</div>
				5211
				5212	<!-- _______________________________________________________________________ -->
				5213	<div class="doc_subsubsection">
				5214	<a name="int_at">'<tt>llvm.adjust.trampoline</tt>' Intrinsic</a>
				5215	</div>
				5216	<div class="doc_text">
				5217	<h5>Syntax:</h5>
				5218	<pre>
				5219	declare i8* @llvm.adjust.trampoline(i8* <tramp>)
				5220	</pre>
				5221	<h5>Overview:</h5>
				5222	<p>
				5223	This intrinsic returns a function pointer suitable for executing
				5224	the trampoline code pointed to by <tt>tramp</tt>.
				5225	</p>
				5226	<h5>Arguments:</h5>
				5227	<p>
				5228	The <tt>llvm.adjust.trampoline</tt> takes one argument, a pointer to a
				5229	trampoline initialized by the
				5230	<a href="#int_it">'<tt>llvm.init.trampoline</tt>' intrinsic</a>.
				5231	</p>
				5232	<h5>Semantics:</h5>
				5233	<p>
				5234	A function pointer that can be used to execute the trampoline code in
				5235	<tt>tramp</tt> is returned. The returned value should be bitcast to an
				5236	<a href="#int_trampoline">appropriate function pointer type</a>
				5237	before being called.
				5238	</p>
				5239	</div>
				5240
				5241	<!-- ======================================================================= -->
				5242	<div class="doc_subsection">
Dan Gohman	f17a25c	2007-07-18 16:29:46 +0000	[diff] [blame]	5243	<a name="int_general">General Intrinsics</a>
				5244	</div>
				5245
				5246	<div class="doc_text">
				5247	<p> This class of intrinsics is designed to be generic and has
				5248	no specific purpose. </p>
				5249	</div>
				5250
				5251	<!-- _______________________________________________________________________ -->
				5252	<div class="doc_subsubsection">
				5253	<a name="int_var_annotation">'<tt>llvm.var.annotation</tt>' Intrinsic</a>
				5254	</div>
				5255
				5256	<div class="doc_text">
				5257
				5258	<h5>Syntax:</h5>
				5259	<pre>
				5260	declare void @llvm.var.annotation(i8* <val>, i8* <str>, i8* <str>, i32 <int> )
				5261	</pre>
				5262
				5263	<h5>Overview:</h5>
				5264
				5265	<p>
				5266	The '<tt>llvm.var.annotation</tt>' intrinsic
				5267	</p>
				5268
				5269	<h5>Arguments:</h5>
				5270
				5271	<p>
				5272	The first argument is a pointer to a value, the second is a pointer to a
				5273	global string, the third is a pointer to a global string which is the source
				5274	file name, and the last argument is the line number.
				5275	</p>
				5276
				5277	<h5>Semantics:</h5>
				5278
				5279	<p>
				5280	This intrinsic allows annotation of local variables with arbitrary strings.
				5281	This can be useful for special purpose optimizations that want to look for these
				5282	annotations. These have no other defined use, they are ignored by code
				5283	generation and optimization.
				5284	</div>
				5285
				5286
				5287	<!-- *********************************************************************** -->
				5288	<hr>
				5289	<address>
				5290	<a href="http://jigsaw.w3.org/css-validator/check/referer"><img
				5291	src="http://jigsaw.w3.org/css-validator/images/vcss" alt="Valid CSS!"></a>
				5292	<a href="http://validator.w3.org/check/referer"><img
				5293	src="http://www.w3.org/Icons/valid-html401" alt="Valid HTML 4.01!" /></a>
				5294
				5295	<a href="mailto:sabre@nondot.org">Chris Lattner</a><br>
				5296	<a href="http://llvm.org">The LLVM Compiler Infrastructure</a><br>
				5297	Last modified: $Date$
				5298	</address>
				5299	</body>
				5300	</html>