Blame - docs/tutorial/OCamlLangImpl3.html - fp2-dev/platform/external/llvm

blob: 0edc726480ce474e09617c0a34d99b90efc316c5 [file] [log] [blame]

Erick Tryzelaar	37c076b	2008-03-30 09:57:12 +0000	[diff] [blame^]	1	<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01//EN"
				2	"http://www.w3.org/TR/html4/strict.dtd">
				3
				4	<html>
				5	<head>
				6	<title>Kaleidoscope: Implementing code generation to LLVM IR</title>
				7	<meta http-equiv="Content-Type" content="text/html; charset=utf-8">
				8	<meta name="author" content="Chris Lattner">
				9	<meta name="author" content="Erick Tryzelaar">
				10	<link rel="stylesheet" href="../llvm.css" type="text/css">
				11	</head>
				12
				13	<body>
				14
				15	<div class="doc_title">Kaleidoscope: Code generation to LLVM IR</div>
				16
				17	<ul>
				18	<li><a href="index.html">Up to Tutorial Index</a></li>
				19	<li>Chapter 3
				20	<ol>
				21	<li><a href="#intro">Chapter 3 Introduction</a></li>
				22	<li><a href="#basics">Code Generation Setup</a></li>
				23	<li><a href="#exprs">Expression Code Generation</a></li>
				24	<li><a href="#funcs">Function Code Generation</a></li>
				25	<li><a href="#driver">Driver Changes and Closing Thoughts</a></li>
				26	<li><a href="#code">Full Code Listing</a></li>
				27	</ol>
				28	</li>
				29	<li><a href="LangImpl4.html">Chapter 4</a>: Adding JIT and Optimizer
				30	Support</li>
				31	</ul>
				32
				33	<div class="doc_author">
				34	<p>
				35	Written by <a href="mailto:sabre@nondot.org">Chris Lattner</a>
				36	and <a href="mailto:idadesub@users.sourceforge.net">Erick Tryzelaar</a>
				37	</p>
				38	</div>
				39
				40	<!-- *********************************************************************** -->
				41	<div class="doc_section"><a name="intro">Chapter 3 Introduction</a></div>
				42	<!-- *********************************************************************** -->
				43
				44	<div class="doc_text">
				45
				46	<p>Welcome to Chapter 3 of the "<a href="index.html">Implementing a language
				47	with LLVM</a>" tutorial. This chapter shows you how to transform the <a
				48	href="OCamlLangImpl2.html">Abstract Syntax Tree</a>, built in Chapter 2, into
				49	LLVM IR. This will teach you a little bit about how LLVM does things, as well
				50	as demonstrate how easy it is to use. It's much more work to build a lexer and
				51	parser than it is to generate LLVM IR code. :)
				52	</p>
				53
				54	<p><b>Please note</b>: the code in this chapter and later require LLVM 2.3 or
				55	LLVM SVN to work. LLVM 2.2 and before will not work with it.</p>
				56
				57	</div>
				58
				59	<!-- *********************************************************************** -->
				60	<div class="doc_section"><a name="basics">Code Generation Setup</a></div>
				61	<!-- *********************************************************************** -->
				62
				63	<div class="doc_text">
				64
				65	<p>
				66	In order to generate LLVM IR, we want some simple setup to get started. First
				67	we define virtual code generation (codegen) methods in each AST class:</p>
				68
				69	<div class="doc_code">
				70	<pre>
				71	let rec codegen_expr = function
				72	\| Ast.Number n -> ...
				73	\| Ast.Variable name -> ...
				74	</pre>
				75	</div>
				76
				77	<p>The <tt>Codegen.codegen_expr</tt> function says to emit IR for that AST node
				78	along with all the things it depends on, and they all return an LLVM Value
				79	object. "Value" is the class used to represent a "<a
				80	href="http://en.wikipedia.org/wiki/Static_single_assignment_form">Static Single
				81	Assignment (SSA)</a> register" or "SSA value" in LLVM. The most distinct aspect
				82	of SSA values is that their value is computed as the related instruction
				83	executes, and it does not get a new value until (and if) the instruction
				84	re-executes. In other words, there is no way to "change" an SSA value. For
				85	more information, please read up on <a
				86	href="http://en.wikipedia.org/wiki/Static_single_assignment_form">Static Single
				87	Assignment</a> - the concepts are really quite natural once you grok them.</p>
				88
				89	<p>The
				90	second thing we want is an "Error" exception like we used for the parser, which
				91	will be used to report errors found during code generation (for example, use of
				92	an undeclared parameter):</p>
				93
				94	<div class="doc_code">
				95	<pre>
				96	exception Error of string
				97
				98	let the_module = create_module "my cool jit"
				99	let builder = builder ()
				100	let named_values:(string, llvalue) Hashtbl.t = Hashtbl.create 10
				101	</pre>
				102	</div>
				103
				104	<p>The static variables will be used during code generation.
				105	<tt>Codgen.the_module</tt> is the LLVM construct that contains all of the
				106	functions and global variables in a chunk of code. In many ways, it is the
				107	top-level structure that the LLVM IR uses to contain code.</p>
				108
				109	<p>The <tt>Codegen.builder</tt> object is a helper object that makes it easy to
				110	generate LLVM instructions. Instances of the <a
				111	href="http://llvm.org/doxygen/LLVMBuilder_8h-source.html"><tt>LLVMBuilder</tt></a>
				112	class keep track of the current place to insert instructions and has methods to
				113	create new instructions.</p>
				114
				115	<p>The <tt>Codegen.named_values</tt> map keeps track of which values are defined
				116	in the current scope and what their LLVM representation is. (In other words, it
				117	is a symbol table for the code). In this form of Kaleidoscope, the only things
				118	that can be referenced are function parameters. As such, function parameters
				119	will be in this map when generating code for their function body.</p>
				120
				121	<p>
				122	With these basics in place, we can start talking about how to generate code for
				123	each expression. Note that this assumes that the <tt>Codgen.builder</tt> has
				124	been set up to generate code <em>into</em> something. For now, we'll assume
				125	that this has already been done, and we'll just use it to emit code.</p>
				126
				127	</div>
				128
				129	<!-- *********************************************************************** -->
				130	<div class="doc_section"><a name="exprs">Expression Code Generation</a></div>
				131	<!-- *********************************************************************** -->
				132
				133	<div class="doc_text">
				134
				135	<p>Generating LLVM code for expression nodes is very straightforward: less
				136	than 30 lines of commented code for all four of our expression nodes. First
				137	we'll do numeric literals:</p>
				138
				139	<div class="doc_code">
				140	<pre>
				141	\| Ast.Number n -> const_float double_type n
				142	</pre>
				143	</div>
				144
				145	<p>In the LLVM IR, numeric constants are represented with the
				146	<tt>ConstantFP</tt> class, which holds the numeric value in an <tt>APFloat</tt>
				147	internally (<tt>APFloat</tt> has the capability of holding floating point
				148	constants of <em>A</em>rbitrary <em>P</em>recision). This code basically just
				149	creates and returns a <tt>ConstantFP</tt>. Note that in the LLVM IR
				150	that constants are all uniqued together and shared. For this reason, the API
				151	uses "the foo::get(..)" idiom instead of "new foo(..)" or "foo::create(..)".</p>
				152
				153	<div class="doc_code">
				154	<pre>
				155	\| Ast.Variable name ->
				156	(try Hashtbl.find named_values name with
				157	\| Not_found -> raise (Error "unknown variable name"))
				158	</pre>
				159	</div>
				160
				161	<p>References to variables are also quite simple using LLVM. In the simple
				162	version of Kaleidoscope, we assume that the variable has already been emited
				163	somewhere and its value is available. In practice, the only values that can be
				164	in the <tt>Codegen.named_values</tt> map are function arguments. This code
				165	simply checks to see that the specified name is in the map (if not, an unknown
				166	variable is being referenced) and returns the value for it. In future chapters,
				167	we'll add support for <a href="LangImpl5.html#for">loop induction variables</a>
				168	in the symbol table, and for <a href="LangImpl7.html#localvars">local
				169	variables</a>.</p>
				170
				171	<div class="doc_code">
				172	<pre>
				173	\| Ast.Binary (op, lhs, rhs) ->
				174	let lhs_val = codegen_expr lhs in
				175	let rhs_val = codegen_expr rhs in
				176	begin
				177	match op with
				178	\| '+' -> build_add lhs_val rhs_val "addtmp" builder
				179	\| '-' -> build_sub lhs_val rhs_val "subtmp" builder
				180	\| '*' -> build_mul lhs_val rhs_val "multmp" builder
				181	\| '<' ->
				182	(* Convert bool 0/1 to double 0.0 or 1.0 *)
				183	let i = build_fcmp Fcmp.Ult lhs_val rhs_val "cmptmp" builder in
				184	build_uitofp i double_type "booltmp" builder
				185	\| _ -> raise (Error "invalid binary operator")
				186	end
				187	</pre>
				188	</div>
				189
				190	<p>Binary operators start to get more interesting. The basic idea here is that
				191	we recursively emit code for the left-hand side of the expression, then the
				192	right-hand side, then we compute the result of the binary expression. In this
				193	code, we do a simple switch on the opcode to create the right LLVM instruction.
				194	</p>
				195
				196	<p>In the example above, the LLVM builder class is starting to show its value.
				197	LLVMBuilder knows where to insert the newly created instruction, all you have to
				198	do is specify what instruction to create (e.g. with <tt>Llvm.create_add</tt>),
				199	which operands to use (<tt>lhs</tt> and <tt>rhs</tt> here) and optionally
				200	provide a name for the generated instruction.</p>
				201
				202	<p>One nice thing about LLVM is that the name is just a hint. For instance, if
				203	the code above emits multiple "addtmp" variables, LLVM will automatically
				204	provide each one with an increasing, unique numeric suffix. Local value names
				205	for instructions are purely optional, but it makes it much easier to read the
				206	IR dumps.</p>
				207
				208	<p><a href="../LangRef.html#instref">LLVM instructions</a> are constrained by
				209	strict rules: for example, the Left and Right operators of
				210	an <a href="../LangRef.html#i_add">add instruction</a> must have the same
				211	type, and the result type of the add must match the operand types. Because
				212	all values in Kaleidoscope are doubles, this makes for very simple code for add,
				213	sub and mul.</p>
				214
				215	<p>On the other hand, LLVM specifies that the <a
				216	href="../LangRef.html#i_fcmp">fcmp instruction</a> always returns an 'i1' value
				217	(a one bit integer). The problem with this is that Kaleidoscope wants the value to be a 0.0 or 1.0 value. In order to get these semantics, we combine the fcmp instruction with
				218	a <a href="../LangRef.html#i_uitofp">uitofp instruction</a>. This instruction
				219	converts its input integer into a floating point value by treating the input
				220	as an unsigned value. In contrast, if we used the <a
				221	href="../LangRef.html#i_sitofp">sitofp instruction</a>, the Kaleidoscope '<'
				222	operator would return 0.0 and -1.0, depending on the input value.</p>
				223
				224	<div class="doc_code">
				225	<pre>
				226	\| Ast.Call (callee, args) ->
				227	(* Look up the name in the module table. *)
				228	let callee =
				229	match lookup_function callee the_module with
				230	\| Some callee -> callee
				231	\| None -> raise (Error "unknown function referenced")
				232	in
				233	let params = params callee in
				234
				235	(* If argument mismatch error. *)
				236	if Array.length params == Array.length args then () else
				237	raise (Error "incorrect # arguments passed");
				238	let args = Array.map codegen_expr args in
				239	build_call callee args "calltmp" builder
				240	</pre>
				241	</div>
				242
				243	<p>Code generation for function calls is quite straightforward with LLVM. The
				244	code above initially does a function name lookup in the LLVM Module's symbol
				245	table. Recall that the LLVM Module is the container that holds all of the
				246	functions we are JIT'ing. By giving each function the same name as what the
				247	user specifies, we can use the LLVM symbol table to resolve function names for
				248	us.</p>
				249
				250	<p>Once we have the function to call, we recursively codegen each argument that
				251	is to be passed in, and create an LLVM <a href="../LangRef.html#i_call">call
				252	instruction</a>. Note that LLVM uses the native C calling conventions by
				253	default, allowing these calls to also call into standard library functions like
				254	"sin" and "cos", with no additional effort.</p>
				255
				256	<p>This wraps up our handling of the four basic expressions that we have so far
				257	in Kaleidoscope. Feel free to go in and add some more. For example, by
				258	browsing the <a href="../LangRef.html">LLVM language reference</a> you'll find
				259	several other interesting instructions that are really easy to plug into our
				260	basic framework.</p>
				261
				262	</div>
				263
				264	<!-- *********************************************************************** -->
				265	<div class="doc_section"><a name="funcs">Function Code Generation</a></div>
				266	<!-- *********************************************************************** -->
				267
				268	<div class="doc_text">
				269
				270	<p>Code generation for prototypes and functions must handle a number of
				271	details, which make their code less beautiful than expression code
				272	generation, but allows us to illustrate some important points. First, lets
				273	talk about code generation for prototypes: they are used both for function
				274	bodies and external function declarations. The code starts with:</p>
				275
				276	<div class="doc_code">
				277	<pre>
				278	let codegen_proto = function
				279	\| Ast.Prototype (name, args) ->
				280	(* Make the function type: double(double,double) etc. *)
				281	let doubles = Array.make (Array.length args) double_type in
				282	let ft = function_type double_type doubles in
				283	let f =
				284	match lookup_function name the_module with
				285	</pre>
				286	</div>
				287
				288	<p>This code packs a lot of power into a few lines. Note first that this
				289	function returns a "Function" instead of a "Value" (although at the moment
				290	they both are modeled by <tt>llvalue</tt> in ocaml). Because a "prototype"
				291	really talks about the external interface for a function (not the value computed
				292	by an expression), it makes sense for it to return the LLVM Function it
				293	corresponds to when codegen'd.</p>
				294
				295	<p>The call to <tt>Llvm.function_type</tt> creates the <tt>Llvm.llvalue</tt>
				296	that should be used for a given Prototype. Since all function arguments in
				297	Kaleidoscope are of type double, the first line creates a vector of "N" LLVM
				298	double types. It then uses the <tt>Llvm.function_type</tt> method to create a
				299	function type that takes "N" doubles as arguments, returns one double as a
				300	result, and that is not vararg (that uses the function
				301	<tt>Llvm.var_arg_function_type</tt>). Note that Types in LLVM are uniqued just
				302	like <tt>Constant</tt>s are, so you don't "new" a type, you "get" it.</p>
				303
				304	<p>The final line above checks if the function has already been defined in
				305	<tt>Codegen.the_module</tt>. If not, we will create it.</p>
				306
				307	<div class="doc_code">
				308	<pre>
				309	\| None -> declare_function name ft the_module
				310	</pre>
				311	</div>
				312
				313	<p>This indicates the type and name to use, as well as which module to insert
				314	into. By default we assume a function has
				315	<tt>Llvm.Linkage.ExternalLinkage</tt>. "<a href="LangRef.html#linkage">external
				316	linkage</a>" means that the function may be defined outside the current module
				317	and/or that it is callable by functions outside the module. The "<tt>name</tt>"
				318	passed in is the name the user specified: this name is registered in
				319	"<tt>Codegen.the_module</tt>"s symbol table, which is used by the function call
				320	code above.</p>
				321
				322	<p>In Kaleidoscope, I choose to allow redefinitions of functions in two cases:
				323	first, we want to allow 'extern'ing a function more than once, as long as the
				324	prototypes for the externs match (since all arguments have the same type, we
				325	just have to check that the number of arguments match). Second, we want to
				326	allow 'extern'ing a function and then definining a body for it. This is useful
				327	when defining mutually recursive functions.</p>
				328
				329	<div class="doc_code">
				330	<pre>
				331	(* If 'f' conflicted, there was already something named 'name'. If it
				332	* has a body, don't allow redefinition or reextern. *)
				333	\| Some f ->
				334	(* If 'f' already has a body, reject this. *)
				335	if Array.length (basic_blocks f) == 0 then () else
				336	raise (Error "redefinition of function");
				337
				338	(* If 'f' took a different number of arguments, reject. *)
				339	if Array.length (params f) == Array.length args then () else
				340	raise (Error "redefinition of function with different # args");
				341	f
				342	in
				343	</pre>
				344	</div>
				345
				346	<p>In order to verify the logic above, we first check to see if the pre-existing
				347	function is "empty". In this case, empty means that it has no basic blocks in
				348	it, which means it has no body. If it has no body, it is a forward
				349	declaration. Since we don't allow anything after a full definition of the
				350	function, the code rejects this case. If the previous reference to a function
				351	was an 'extern', we simply verify that the number of arguments for that
				352	definition and this one match up. If not, we emit an error.</p>
				353
				354	<div class="doc_code">
				355	<pre>
				356	(* Set names for all arguments. *)
				357	Array.iteri (fun i a ->
				358	let n = args.(i) in
				359	set_value_name n a;
				360	Hashtbl.add named_values n a;
				361	) (params f);
				362	f
				363	</pre>
				364	</div>
				365
				366	<p>The last bit of code for prototypes loops over all of the arguments in the
				367	function, setting the name of the LLVM Argument objects to match, and registering
				368	the arguments in the <tt>Codegen.named_values</tt> map for future use by the
				369	<tt>Ast.Variable</tt> variant. Once this is set up, it returns the Function
				370	object to the caller. Note that we don't check for conflicting
				371	argument names here (e.g. "extern foo(a b a)"). Doing so would be very
				372	straight-forward with the mechanics we have already used above.</p>
				373
				374	<div class="doc_code">
				375	<pre>
				376	let codegen_func = function
				377	\| Ast.Function (proto, body) ->
				378	Hashtbl.clear named_values;
				379	let the_function = codegen_proto proto in
				380	</pre>
				381	</div>
				382
				383	<p>Code generation for function definitions starts out simply enough: we just
				384	codegen the prototype (Proto) and verify that it is ok. We then clear out the
				385	<tt>Codegen.named_values</tt> map to make sure that there isn't anything in it
				386	from the last function we compiled. Code generation of the prototype ensures
				387	that there is an LLVM Function object that is ready to go for us.</p>
				388
				389	<div class="doc_code">
				390	<pre>
				391	(* Create a new basic block to start insertion into. *)
				392	let bb = append_block "entry" the_function in
				393	position_at_end bb builder;
				394
				395	try
				396	let ret_val = codegen_expr body in
				397	</pre>
				398	</div>
				399
				400	<p>Now we get to the point where the <tt>Codegen.builder</tt> is set up. The
				401	first line creates a new
				402	<a href="http://en.wikipedia.org/wiki/Basic_block">basic block</a> (named
				403	"entry"), which is inserted into <tt>the_function</tt>. The second line then
				404	tells the builder that new instructions should be inserted into the end of the
				405	new basic block. Basic blocks in LLVM are an important part of functions that
				406	define the <a
				407	href="http://en.wikipedia.org/wiki/Control_flow_graph">Control Flow Graph</a>.
				408	Since we don't have any control flow, our functions will only contain one
				409	block at this point. We'll fix this in <a href="OCamlLangImpl5.html">Chapter
				410	5</a> :).</p>
				411
				412	<div class="doc_code">
				413	<pre>
				414	let ret_val = codegen_expr body in
				415
				416	(* Finish off the function. *)
				417	let _ = build_ret ret_val builder in
				418
				419	(* Validate the generated code, checking for consistency. *)
				420	Llvm_analysis.assert_valid_function the_function;
				421
				422	the_function
				423	</pre>
				424	</div>
				425
				426	<p>Once the insertion point is set up, we call the <tt>Codegen.codegen_func</tt>
				427	method for the root expression of the function. If no error happens, this emits
				428	code to compute the expression into the entry block and returns the value that
				429	was computed. Assuming no error, we then create an LLVM <a
				430	href="../LangRef.html#i_ret">ret instruction</a>, which completes the function.
				431	Once the function is built, we call
				432	<tt>Llvm_analysis.assert_valid_function</tt>, which is provided by LLVM. This
				433	function does a variety of consistency checks on the generated code, to
				434	determine if our compiler is doing everything right. Using this is important:
				435	it can catch a lot of bugs. Once the function is finished and validated, we
				436	return it.</p>
				437
				438	<div class="doc_code">
				439	<pre>
				440	with e ->
				441	delete_function the_function;
				442	raise e
				443	</pre>
				444	</div>
				445
				446	<p>The only piece left here is handling of the error case. For simplicity, we
				447	handle this by merely deleting the function we produced with the
				448	<tt>Llvm.delete_function</tt> method. This allows the user to redefine a
				449	function that they incorrectly typed in before: if we didn't delete it, it
				450	would live in the symbol table, with a body, preventing future redefinition.</p>
				451
				452	<p>This code does have a bug, though. Since the <tt>Codegen.codegen_proto</tt>
				453	can return a previously defined forward declaration, our code can actually delete
				454	a forward declaration. There are a number of ways to fix this bug, see what you
				455	can come up with! Here is a testcase:</p>
				456
				457	<div class="doc_code">
				458	<pre>
				459	extern foo(a b); # ok, defines foo.
				460	def foo(a b) c; # error, 'c' is invalid.
				461	def bar() foo(1, 2); # error, unknown function "foo"
				462	</pre>
				463	</div>
				464
				465	</div>
				466
				467	<!-- *********************************************************************** -->
				468	<div class="doc_section"><a name="driver">Driver Changes and
				469	Closing Thoughts</a></div>
				470	<!-- *********************************************************************** -->
				471
				472	<div class="doc_text">
				473
				474	<p>
				475	For now, code generation to LLVM doesn't really get us much, except that we can
				476	look at the pretty IR calls. The sample code inserts calls to Codegen into the
				477	"<tt>Toplevel.main_loop</tt>", and then dumps out the LLVM IR. This gives a
				478	nice way to look at the LLVM IR for simple functions. For example:
				479	</p>
				480
				481	<div class="doc_code">
				482	<pre>
				483	ready> <b>4+5</b>;
				484	Read top-level expression:
				485	define double @""() {
				486	entry:
				487	%addtmp = add double 4.000000e+00, 5.000000e+00
				488	ret double %addtmp
				489	}
				490	</pre>
				491	</div>
				492
				493	<p>Note how the parser turns the top-level expression into anonymous functions
				494	for us. This will be handy when we add <a href="LangImpl4.html#jit">JIT
				495	support</a> in the next chapter. Also note that the code is very literally
				496	transcribed, no optimizations are being performed. We will
				497	<a href="OCamlLangImpl4.html#trivialconstfold">add optimizations</a> explicitly
				498	in the next chapter.</p>
				499
				500	<div class="doc_code">
				501	<pre>
				502	ready> <b>def foo(a b) aa + 2ab + bb;</b>
				503	Read function definition:
				504	define double @foo(double %a, double %b) {
				505	entry:
				506	%multmp = mul double %a, %a
				507	%multmp1 = mul double 2.000000e+00, %a
				508	%multmp2 = mul double %multmp1, %b
				509	%addtmp = add double %multmp, %multmp2
				510	%multmp3 = mul double %b, %b
				511	%addtmp4 = add double %addtmp, %multmp3
				512	ret double %addtmp4
				513	}
				514	</pre>
				515	</div>
				516
				517	<p>This shows some simple arithmetic. Notice the striking similarity to the
				518	LLVM builder calls that we use to create the instructions.</p>
				519
				520	<div class="doc_code">
				521	<pre>
				522	ready> <b>def bar(a) foo(a, 4.0) + bar(31337);</b>
				523	Read function definition:
				524	define double @bar(double %a) {
				525	entry:
				526	%calltmp = call double @foo( double %a, double 4.000000e+00 )
				527	%calltmp1 = call double @bar( double 3.133700e+04 )
				528	%addtmp = add double %calltmp, %calltmp1
				529	ret double %addtmp
				530	}
				531	</pre>
				532	</div>
				533
				534	<p>This shows some function calls. Note that this function will take a long
				535	time to execute if you call it. In the future we'll add conditional control
				536	flow to actually make recursion useful :).</p>
				537
				538	<div class="doc_code">
				539	<pre>
				540	ready> <b>extern cos(x);</b>
				541	Read extern:
				542	declare double @cos(double)
				543
				544	ready> <b>cos(1.234);</b>
				545	Read top-level expression:
				546	define double @""() {
				547	entry:
				548	%calltmp = call double @cos( double 1.234000e+00 )
				549	ret double %calltmp
				550	}
				551	</pre>
				552	</div>
				553
				554	<p>This shows an extern for the libm "cos" function, and a call to it.</p>
				555
				556
				557	<div class="doc_code">
				558	<pre>
				559	ready> <b>^D</b>
				560	; ModuleID = 'my cool jit'
				561
				562	define double @""() {
				563	entry:
				564	%addtmp = add double 4.000000e+00, 5.000000e+00
				565	ret double %addtmp
				566	}
				567
				568	define double @foo(double %a, double %b) {
				569	entry:
				570	%multmp = mul double %a, %a
				571	%multmp1 = mul double 2.000000e+00, %a
				572	%multmp2 = mul double %multmp1, %b
				573	%addtmp = add double %multmp, %multmp2
				574	%multmp3 = mul double %b, %b
				575	%addtmp4 = add double %addtmp, %multmp3
				576	ret double %addtmp4
				577	}
				578
				579	define double @bar(double %a) {
				580	entry:
				581	%calltmp = call double @foo( double %a, double 4.000000e+00 )
				582	%calltmp1 = call double @bar( double 3.133700e+04 )
				583	%addtmp = add double %calltmp, %calltmp1
				584	ret double %addtmp
				585	}
				586
				587	declare double @cos(double)
				588
				589	define double @""() {
				590	entry:
				591	%calltmp = call double @cos( double 1.234000e+00 )
				592	ret double %calltmp
				593	}
				594	</pre>
				595	</div>
				596
				597	<p>When you quit the current demo, it dumps out the IR for the entire module
				598	generated. Here you can see the big picture with all the functions referencing
				599	each other.</p>
				600
				601	<p>This wraps up the third chapter of the Kaleidoscope tutorial. Up next, we'll
				602	describe how to <a href="LangImpl4.html">add JIT codegen and optimizer
				603	support</a> to this so we can actually start running code!</p>
				604
				605	</div>
				606
				607
				608	<!-- *********************************************************************** -->
				609	<div class="doc_section"><a name="code">Full Code Listing</a></div>
				610	<!-- *********************************************************************** -->
				611
				612	<div class="doc_text">
				613
				614	<p>
				615	Here is the complete code listing for our running example, enhanced with the
				616	LLVM code generator. Because this uses the LLVM libraries, we need to link
				617	them in. To do this, we use the <a
				618	href="http://llvm.org/cmds/llvm-config.html">llvm-config</a> tool to inform
				619	our makefile/command line about which options to use:</p>
				620
				621	<div class="doc_code">
				622	<pre>
				623	# Compile
				624	ocamlbuild toy.byte
				625	# Run
				626	./toy.byte
				627	</pre>
				628	</div>
				629
				630	<p>Here is the code:</p>
				631
				632	<dl>
				633	<dt>_tags:</dt>
				634	<dd class="doc_code">
				635	<pre>
				636	<{lexer,parser}.ml>: use_camlp4, pp(camlp4of)
				637	<*.{byte,native}>: g++, use_llvm, use_llvm_analysis
				638	</pre>
				639	</dd>
				640
				641	<dt>myocamlbuild.ml:</dt>
				642	<dd class="doc_code">
				643	<pre>
				644	open Ocamlbuild_plugin;;
				645
				646	ocaml_lib ~extern:true "llvm";;
				647	ocaml_lib ~extern:true "llvm_analysis";;
				648
				649	flag ["link"; "ocaml"; "g++"] (S[A"-cc"; A"g++"]);;
				650	</pre>
				651	</dd>
				652
				653	<dt>token.ml:</dt>
				654	<dd class="doc_code">
				655	<pre>
				656	(*===----------------------------------------------------------------------===
				657	* Lexer Tokens
				658	===----------------------------------------------------------------------===)
				659
				660	(* The lexer returns these 'Kwd' if it is an unknown character, otherwise one of
				661	* these others for known things. *)
				662	type token =
				663	(* commands *)
				664	\| Def \| Extern
				665
				666	(* primary *)
				667	\| Ident of string \| Number of float
				668
				669	(* unknown *)
				670	\| Kwd of char
				671	</pre>
				672	</dd>
				673
				674	<dt>lexer.ml:</dt>
				675	<dd class="doc_code">
				676	<pre>
				677	(*===----------------------------------------------------------------------===
				678	* Lexer
				679	===----------------------------------------------------------------------===)
				680
				681	let rec lex = parser
				682	(* Skip any whitespace. *)
				683	\| [< ' (' ' \| '\n' \| '\r' \| '\t'); stream >] -> lex stream
				684
				685	(* identifier: [a-zA-Z][a-zA-Z0-9] *)
				686	\| [< ' ('A' .. 'Z' \| 'a' .. 'z' as c); stream >] ->
				687	let buffer = Buffer.create 1 in
				688	Buffer.add_char buffer c;
				689	lex_ident buffer stream
				690
				691	(* number: [0-9.]+ *)
				692	\| [< ' ('0' .. '9' as c); stream >] ->
				693	let buffer = Buffer.create 1 in
				694	Buffer.add_char buffer c;
				695	lex_number buffer stream
				696
				697	(* Comment until end of line. *)
				698	\| [< ' ('#'); stream >] ->
				699	lex_comment stream
				700
				701	(* Otherwise, just return the character as its ascii value. *)
				702	\| [< 'c; stream >] ->
				703	[< 'Token.Kwd c; lex stream >]
				704
				705	(* end of stream. *)
				706	\| [< >] -> [< >]
				707
				708	and lex_number buffer = parser
				709	\| [< ' ('0' .. '9' \| '.' as c); stream >] ->
				710	Buffer.add_char buffer c;
				711	lex_number buffer stream
				712	\| [< stream=lex >] ->
				713	[< 'Token.Number (float_of_string (Buffer.contents buffer)); stream >]
				714
				715	and lex_ident buffer = parser
				716	\| [< ' ('A' .. 'Z' \| 'a' .. 'z' \| '0' .. '9' as c); stream >] ->
				717	Buffer.add_char buffer c;
				718	lex_ident buffer stream
				719	\| [< stream=lex >] ->
				720	match Buffer.contents buffer with
				721	\| "def" -> [< 'Token.Def; stream >]
				722	\| "extern" -> [< 'Token.Extern; stream >]
				723	\| id -> [< 'Token.Ident id; stream >]
				724
				725	and lex_comment = parser
				726	\| [< ' ('\n'); stream=lex >] -> stream
				727	\| [< 'c; e=lex_comment >] -> e
				728	\| [< >] -> [< >]
				729	</pre>
				730	</dd>
				731
				732	<dt>ast.ml:</dt>
				733	<dd class="doc_code">
				734	<pre>
				735	(*===----------------------------------------------------------------------===
				736	* Abstract Syntax Tree (aka Parse Tree)
				737	===----------------------------------------------------------------------===)
				738
				739	(* expr - Base type for all expression nodes. *)
				740	type expr =
				741	(* variant for numeric literals like "1.0". *)
				742	\| Number of float
				743
				744	(* variant for referencing a variable, like "a". *)
				745	\| Variable of string
				746
				747	(* variant for a binary operator. *)
				748	\| Binary of char * expr * expr
				749
				750	(* variant for function calls. *)
				751	\| Call of string * expr array
				752
				753	(* proto - This type represents the "prototype" for a function, which captures
				754	* its name, and its argument names (thus implicitly the number of arguments the
				755	* function takes). *)
				756	type proto = Prototype of string * string array
				757
				758	(* func - This type represents a function definition itself. *)
				759	type func = Function of proto * expr
				760	</pre>
				761	</dd>
				762
				763	<dt>parser.ml:</dt>
				764	<dd class="doc_code">
				765	<pre>
				766	(*===---------------------------------------------------------------------===
				767	* Parser
				768	===---------------------------------------------------------------------===)
				769
				770	(* binop_precedence - This holds the precedence for each binary operator that is
				771	* defined *)
				772	let binop_precedence:(char, int) Hashtbl.t = Hashtbl.create 10
				773
				774	(* precedence - Get the precedence of the pending binary operator token. *)
				775	let precedence c = try Hashtbl.find binop_precedence c with Not_found -> -1
				776
				777	(* primary
				778	* ::= identifier
				779	* ::= numberexpr
				780	* ::= parenexpr *)
				781	let rec parse_primary = parser
				782	(* numberexpr ::= number *)
				783	\| [< 'Token.Number n >] -> Ast.Number n
				784
				785	(* parenexpr ::= '(' expression ')' *)
				786	\| [< 'Token.Kwd '('; e=parse_expr; 'Token.Kwd ')' ?? "expected ')'" >] -> e
				787
				788	(* identifierexpr
				789	* ::= identifier
				790	* ::= identifier '(' argumentexpr ')' *)
				791	\| [< 'Token.Ident id; stream >] ->
				792	let rec parse_args accumulator = parser
				793	\| [< e=parse_expr; stream >] ->
				794	begin parser
				795	\| [< 'Token.Kwd ','; e=parse_args (e :: accumulator) >] -> e
				796	\| [< >] -> e :: accumulator
				797	end stream
				798	\| [< >] -> accumulator
				799	in
				800	let rec parse_ident id = parser
				801	(* Call. *)
				802	\| [< 'Token.Kwd '(';
				803	args=parse_args [];
				804	'Token.Kwd ')' ?? "expected ')'">] ->
				805	Ast.Call (id, Array.of_list (List.rev args))
				806
				807	(* Simple variable ref. *)
				808	\| [< >] -> Ast.Variable id
				809	in
				810	parse_ident id stream
				811
				812	\| [< >] -> raise (Stream.Error "unknown token when expecting an expression.")
				813
				814	(* binoprhs
				815	* ::= ('+' primary)* *)
				816	and parse_bin_rhs expr_prec lhs stream =
				817	match Stream.peek stream with
				818	(* If this is a binop, find its precedence. *)
				819	\| Some (Token.Kwd c) when Hashtbl.mem binop_precedence c ->
				820	let token_prec = precedence c in
				821
				822	(* If this is a binop that binds at least as tightly as the current binop,
				823	* consume it, otherwise we are done. *)
				824	if token_prec < expr_prec then lhs else begin
				825	(* Eat the binop. *)
				826	Stream.junk stream;
				827
				828	(* Parse the primary expression after the binary operator. *)
				829	let rhs = parse_primary stream in
				830
				831	(* Okay, we know this is a binop. *)
				832	let rhs =
				833	match Stream.peek stream with
				834	\| Some (Token.Kwd c2) ->
				835	(* If BinOp binds less tightly with rhs than the operator after
				836	* rhs, let the pending operator take rhs as its lhs. *)
				837	let next_prec = precedence c2 in
				838	if token_prec < next_prec
				839	then parse_bin_rhs (token_prec + 1) rhs stream
				840	else rhs
				841	\| _ -> rhs
				842	in
				843
				844	(* Merge lhs/rhs. *)
				845	let lhs = Ast.Binary (c, lhs, rhs) in
				846	parse_bin_rhs expr_prec lhs stream
				847	end
				848	\| _ -> lhs
				849
				850	(* expression
				851	* ::= primary binoprhs *)
				852	and parse_expr = parser
				853	\| [< lhs=parse_primary; stream >] -> parse_bin_rhs 0 lhs stream
				854
				855	(* prototype
				856	* ::= id '(' id* ')' *)
				857	let parse_prototype =
				858	let rec parse_args accumulator = parser
				859	\| [< 'Token.Ident id; e=parse_args (id::accumulator) >] -> e
				860	\| [< >] -> accumulator
				861	in
				862
				863	parser
				864	\| [< 'Token.Ident id;
				865	'Token.Kwd '(' ?? "expected '(' in prototype";
				866	args=parse_args [];
				867	'Token.Kwd ')' ?? "expected ')' in prototype" >] ->
				868	(* success. *)
				869	Ast.Prototype (id, Array.of_list (List.rev args))
				870
				871	\| [< >] ->
				872	raise (Stream.Error "expected function name in prototype")
				873
				874	(* definition ::= 'def' prototype expression *)
				875	let parse_definition = parser
				876	\| [< 'Token.Def; p=parse_prototype; e=parse_expr >] ->
				877	Ast.Function (p, e)
				878
				879	(* toplevelexpr ::= expression *)
				880	let parse_toplevel = parser
				881	\| [< e=parse_expr >] ->
				882	(* Make an anonymous proto. *)
				883	Ast.Function (Ast.Prototype ("", [\|\|]), e)
				884
				885	(* external ::= 'extern' prototype *)
				886	let parse_extern = parser
				887	\| [< 'Token.Extern; e=parse_prototype >] -> e
				888	</pre>
				889	</dd>
				890
				891	<dt>codegen.ml:</dt>
				892	<dd class="doc_code">
				893	<pre>
				894	(*===----------------------------------------------------------------------===
				895	* Code Generation
				896	===----------------------------------------------------------------------===)
				897
				898	open Llvm
				899
				900	exception Error of string
				901
				902	let the_module = create_module "my cool jit"
				903	let builder = builder ()
				904	let named_values:(string, llvalue) Hashtbl.t = Hashtbl.create 10
				905
				906	let rec codegen_expr = function
				907	\| Ast.Number n -> const_float double_type n
				908	\| Ast.Variable name ->
				909	(try Hashtbl.find named_values name with
				910	\| Not_found -> raise (Error "unknown variable name"))
				911	\| Ast.Binary (op, lhs, rhs) ->
				912	let lhs_val = codegen_expr lhs in
				913	let rhs_val = codegen_expr rhs in
				914	begin
				915	match op with
				916	\| '+' -> build_add lhs_val rhs_val "addtmp" builder
				917	\| '-' -> build_sub lhs_val rhs_val "subtmp" builder
				918	\| '*' -> build_mul lhs_val rhs_val "multmp" builder
				919	\| '<' ->
				920	(* Convert bool 0/1 to double 0.0 or 1.0 *)
				921	let i = build_fcmp Fcmp.Ult lhs_val rhs_val "cmptmp" builder in
				922	build_uitofp i double_type "booltmp" builder
				923	\| _ -> raise (Error "invalid binary operator")
				924	end
				925	\| Ast.Call (callee, args) ->
				926	(* Look up the name in the module table. *)
				927	let callee =
				928	match lookup_function callee the_module with
				929	\| Some callee -> callee
				930	\| None -> raise (Error "unknown function referenced")
				931	in
				932	let params = params callee in
				933
				934	(* If argument mismatch error. *)
				935	if Array.length params == Array.length args then () else
				936	raise (Error "incorrect # arguments passed");
				937	let args = Array.map codegen_expr args in
				938	build_call callee args "calltmp" builder
				939
				940	let codegen_proto = function
				941	\| Ast.Prototype (name, args) ->
				942	(* Make the function type: double(double,double) etc. *)
				943	let doubles = Array.make (Array.length args) double_type in
				944	let ft = function_type double_type doubles in
				945	let f =
				946	match lookup_function name the_module with
				947	\| None -> declare_function name ft the_module
				948
				949	(* If 'f' conflicted, there was already something named 'name'. If it
				950	* has a body, don't allow redefinition or reextern. *)
				951	\| Some f ->
				952	(* If 'f' already has a body, reject this. *)
				953	if block_begin f <> At_end f then
				954	raise (Error "redefinition of function");
				955
				956	(* If 'f' took a different number of arguments, reject. *)
				957	if element_type (type_of f) <> ft then
				958	raise (Error "redefinition of function with different # args");
				959	f
				960	in
				961
				962	(* Set names for all arguments. *)
				963	Array.iteri (fun i a ->
				964	let n = args.(i) in
				965	set_value_name n a;
				966	Hashtbl.add named_values n a;
				967	) (params f);
				968	f
				969
				970	let codegen_func = function
				971	\| Ast.Function (proto, body) ->
				972	Hashtbl.clear named_values;
				973	let the_function = codegen_proto proto in
				974
				975	(* Create a new basic block to start insertion into. *)
				976	let bb = append_block "entry" the_function in
				977	position_at_end bb builder;
				978
				979	try
				980	let ret_val = codegen_expr body in
				981
				982	(* Finish off the function. *)
				983	let _ = build_ret ret_val builder in
				984
				985	(* Validate the generated code, checking for consistency. *)
				986	Llvm_analysis.assert_valid_function the_function;
				987
				988	the_function
				989	with e ->
				990	delete_function the_function;
				991	raise e
				992	</pre>
				993	</dd>
				994
				995	<dt>toplevel.ml:</dt>
				996	<dd class="doc_code">
				997	<pre>
				998	(*===----------------------------------------------------------------------===
				999	* Top-Level parsing and JIT Driver
				1000	===----------------------------------------------------------------------===)
				1001
				1002	open Llvm
				1003
				1004	(* top ::= definition \| external \| expression \| ';' *)
				1005	let rec main_loop stream =
				1006	match Stream.peek stream with
				1007	\| None -> ()
				1008
				1009	(* ignore top-level semicolons. *)
				1010	\| Some (Token.Kwd ';') ->
				1011	Stream.junk stream;
				1012	main_loop stream
				1013
				1014	\| Some token ->
				1015	begin
				1016	try match token with
				1017	\| Token.Def ->
				1018	let e = Parser.parse_definition stream in
				1019	print_endline "parsed a function definition.";
				1020	dump_value (Codegen.codegen_func e);
				1021	\| Token.Extern ->
				1022	let e = Parser.parse_extern stream in
				1023	print_endline "parsed an extern.";
				1024	dump_value (Codegen.codegen_proto e);
				1025	\| _ ->
				1026	(* Evaluate a top-level expression into an anonymous function. *)
				1027	let e = Parser.parse_toplevel stream in
				1028	print_endline "parsed a top-level expr";
				1029	dump_value (Codegen.codegen_func e);
				1030	with Stream.Error s \| Codegen.Error s ->
				1031	(* Skip token for error recovery. *)
				1032	Stream.junk stream;
				1033	print_endline s;
				1034	end;
				1035	print_string "ready> "; flush stdout;
				1036	main_loop stream
				1037	</pre>
				1038	</dd>
				1039
				1040	<dt>toy.ml:</dt>
				1041	<dd class="doc_code">
				1042	<pre>
				1043	(*===----------------------------------------------------------------------===
				1044	* Main driver code.
				1045	===----------------------------------------------------------------------===)
				1046
				1047	open Llvm
				1048
				1049	let main () =
				1050	(* Install standard binary operators.
				1051	* 1 is the lowest precedence. *)
				1052	Hashtbl.add Parser.binop_precedence '<' 10;
				1053	Hashtbl.add Parser.binop_precedence '+' 20;
				1054	Hashtbl.add Parser.binop_precedence '-' 20;
				1055	Hashtbl.add Parser.binop_precedence '' 40; ( highest. *)
				1056
				1057	(* Prime the first token. *)
				1058	print_string "ready> "; flush stdout;
				1059	let stream = Lexer.lex (Stream.of_channel stdin) in
				1060
				1061	(* Run the main "interpreter loop" now. *)
				1062	Toplevel.main_loop stream;
				1063
				1064	(* Print out all the generated code. *)
				1065	dump_module Codegen.the_module
				1066	;;
				1067
				1068	main ()
				1069	</pre>
				1070	</dd>
				1071	</dl>
				1072
				1073	<a href="OCamlLangImpl4.html">Next: Adding JIT and Optimizer Support</a>
				1074	</div>
				1075
				1076	<!-- *********************************************************************** -->
				1077	<hr>
				1078	<address>
				1079	<a href="http://jigsaw.w3.org/css-validator/check/referer"><img
				1080	src="http://jigsaw.w3.org/css-validator/images/vcss" alt="Valid CSS!"></a>
				1081	<a href="http://validator.w3.org/check/referer"><img
				1082	src="http://www.w3.org/Icons/valid-html401" alt="Valid HTML 4.01!"></a>
				1083
				1084	<a href="mailto:sabre@nondot.org">Chris Lattner</a><br>
				1085	<a href="mailto:idadesub@users.sourceforge.net">Erick Tryzelaar</a><br>
				1086	<a href="http://llvm.org">The LLVM Compiler Infrastructure</a><br>
				1087	Last modified: $Date: 2007-10-17 11:05:13 -0700 (Wed, 17 Oct 2007) $
				1088	</address>
				1089	</body>
				1090	</html>