Gitiles
Code Review Sign In
gerrit-public.fairphone.software / fp2-dev / platform / external / llvm / 118749e02c513f7c4db9b0554f66b635c848e79f / . / docs / tutorial / LangImpl4.html
blob: 5c3f58db868526f643a5b57bb2cb6c1bde0a21ac [file] [log] [blame]
Chris Lattnerc0b42e92007-10-23 06:27:55 +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: Adding JIT and Optimizer Support</title>
7 <meta http-equiv="Content-Type" content="text/html; charset=utf-8">
8 <meta name="author" content="Chris Lattner">
9 <link rel="stylesheet" href="../llvm.css" type="text/css">
10</head>
11
12<body>
13
14<div class="doc_title">Kaleidoscope: Adding JIT and Optimizer Support</div>
15
16<div class="doc_author">
17 <p>Written by <a href="mailto:sabre@nondot.org">Chris Lattner</a></p>
18</div>
19
20<!-- *********************************************************************** -->
21<div class="doc_section"><a name="intro">Part 4 Introduction</a></div>
22<!-- *********************************************************************** -->
23
24<div class="doc_text">
25
26<p>Welcome to part 4 of the "<a href="index.html">Implementing a language with
Chris Lattner118749e2007-10-25 06:23:36 +0000[diff] [blame^]27LLVM</a>" tutorial. Parts 1-3 described the implementation of a simple language
28and included support for generating LLVM IR. This chapter describes two new
29techniques: adding optimizer support to your language, and adding JIT compiler
30support. This shows how to get nice efficient code for your language.</p>
Chris Lattnerc0b42e92007-10-23 06:27:55 +0000[diff] [blame]31
32</div>
33
34<!-- *********************************************************************** -->
Chris Lattner118749e2007-10-25 06:23:36 +0000[diff] [blame^]35<div class="doc_section"><a name="trivialconstfold">Trivial Constant
36Folding</a></div>
Chris Lattnerc0b42e92007-10-23 06:27:55 +0000[diff] [blame]37<!-- *********************************************************************** -->
38
39<div class="doc_text">
40
41<p>
Chris Lattner118749e2007-10-25 06:23:36 +0000[diff] [blame^]42Our demonstration for Chapter 3 is elegant and easy to extend. Unfortunately,
43it does not produce wonderful code. For example, when compiling simple code,
44we don't get obvious optimizations:</p>
Chris Lattnerc0b42e92007-10-23 06:27:55 +0000[diff] [blame]45
46<div class="doc_code">
47<pre>
Chris Lattner118749e2007-10-25 06:23:36 +0000[diff] [blame^]48ready&gt; <b>def test(x) 1+2+x;</b>
49Read function definition:
50define double @test(double %x) {
51entry:
52 %addtmp = add double 1.000000e+00, 2.000000e+00
53 %addtmp1 = add double %addtmp, %x
54 ret double %addtmp1
55}
56</pre>
57</div>
58
59<p>This code is a very very literal transcription of the AST built by parsing
60our code, and as such, lacks optimizations like constant folding (we'd like to
61get "<tt>add x, 3.0</tt>" in the example above) as well as other more important
62optimizations. Constant folding in particular is a very common and very
63important optimization: so much so that many language implementors implement
64constant folding support in their AST representation.</p>
65
66<p>With LLVM, you don't need to. Since all calls to build LLVM IR go through
67the LLVM builder, it would be nice if the builder itself checked to see if there
68was a constant folding opportunity when you call it. If so, it could just do
69the constant fold and return the constant instead of creating an instruction.
70This is exactly what the <tt>LLVMFoldingBuilder</tt> class does. Lets make one
71change:
72
73<div class="doc_code">
74<pre>
75static LLVMFoldingBuilder Builder;
76</pre>
77</div>
78
79<p>All we did was switch from <tt>LLVMBuilder</tt> to
80<tt>LLVMFoldingBuilder</tt>. Though we change no other code, now all of our
81instructions are implicitly constant folded without us having to do anything
82about it. For example, our example above now compiles to:</p>
83
84<div class="doc_code">
85<pre>
86ready&gt; <b>def test(x) 1+2+x;</b>
87Read function definition:
88define double @test(double %x) {
89entry:
90 %addtmp = add double 3.000000e+00, %x
91 ret double %addtmp
92}
93</pre>
94</div>
95
96<p>Well, that was easy. :) In practice, we recommend always using
97<tt>LLVMConstantBuilder</tt> when generating code like this. It has no
98"syntactic overhead" for its use (you don't have to uglify your compiler with
99constant checks everywhere) and it can dramatically reduce the amount of
100LLVM IR that is generated in some cases (particular for languages with a macro
101preprocessor or that use a lot of constants).</p>
102
103<p>On the other hand, the <tt>LLVMFoldingBuilder</tt> is limited by the fact
104that it does all of its analysis inline with the code as it is built. If you
105take a slightly more complex example:</p>
106
107<div class="doc_code">
108<pre>
109ready&gt; <b>def test(x) (1+2+x)*(x+(1+2));</b>
110ready> Read function definition:
111define double @test(double %x) {
112entry:
113 %addtmp = add double 3.000000e+00, %x
114 %addtmp1 = add double %x, 3.000000e+00
115 %multmp = mul double %addtmp, %addtmp1
116 ret double %multmp
117}
118</pre>
119</div>
120
121<p>In this case, the LHS and RHS of the multiplication are the same value. We'd
122really like to see this generate "<tt>tmp = x+3; result = tmp*tmp;</tt>" instead
123of computing "<tt>x*3</tt>" twice.</p>
124
125<p>Unfortunately, no amount of local analysis will be able to detect and correct
126this. This requires two transformations: reassociation of expressions (to
127make the add's lexically identical) and Common Subexpression Elimination (CSE)
128to delete the redundant add instruction. Fortunately, LLVM provides a broad
129range of optimizations that you can use, in the form of "passes".</p>
130
131</div>
132
133<!-- *********************************************************************** -->
134<div class="doc_section"><a name="optimizerpasses">LLVM Optimization
135 Passes</a></div>
136<!-- *********************************************************************** -->
137
138<div class="doc_text">
139
140<p>LLVM provides many optimization passes which do many different sorts of
141things and have different tradeoffs. Unlike other systems, LLVM doesn't hold
142to the mistaken notion that one set of optimizations is right for all languages
143and for all situations. LLVM allows a compiler implementor to make complete
144decisions about what optimizations to use, in which order, and in what
145situation.</p>
146
147<p>As a concrete example, LLVM supports both "whole module" passes, which look
148across as large of body of code as they can (often a whole file, but if run
149at link time, this can be a substantial portion of the whole program). It also
150supports and includes "per-function" passes which just operate on a single
151function at a time, without looking at other functions. For more information
152on passes and how the get run, see the <a href="../WritingAnLLVMPass.html">How
153to Write a Pass</a> document.</p>
154
155<p>For Kaleidoscope, we are currently generating functions on the fly, one at
156a time, as the user types them in. We aren't shooting for the ultimate
157optimization experience in this setting, but we also want to catch the easy and
158quick stuff where possible. As such, we will choose to run a few per-function
159optimizations as the user types the function in. If we wanted to make a "static
160Kaleidoscope compiler", we would use exactly the code we have now, except that
161we would defer running the optimizer until the entire file has been parsed.</p>
162
163<p>In order to get per-function optimizations going, we need to set up a
164<a href="../WritingAnLLVMPass.html#passmanager">FunctionPassManager</a> to hold and
165organize the LLVM optimizations that we want to run. Once we have that, we can
166add a set of optimizations to run. The code looks like this:</p>
167
168<div class="doc_code">
169<pre>
170 ExistingModuleProvider OurModuleProvider(TheModule);
171 FunctionPassManager OurFPM(&amp;OurModuleProvider);
172
173 // Set up the optimizer pipeline. Start with registering info about how the
174 // target lays out data structures.
175 OurFPM.add(new TargetData(*TheExecutionEngine->getTargetData()));
176 // Do simple "peephole" optimizations and bit-twiddling optzns.
177 OurFPM.add(createInstructionCombiningPass());
178 // Reassociate expressions.
179 OurFPM.add(createReassociatePass());
180 // Eliminate Common SubExpressions.
181 OurFPM.add(createGVNPass());
182 // Simplify the control flow graph (deleting unreachable blocks, etc).
183 OurFPM.add(createCFGSimplificationPass());
184
185 // Set the global so the code gen can use this.
186 TheFPM = &amp;OurFPM;
187
188 // Run the main "interpreter loop" now.
189 MainLoop();
190</pre>
191</div>
192
193<p>This code defines two objects, a <tt>ExistingModuleProvider</tt> and a
194<tt>FunctionPassManager</tt>. The former is basically a wrapper around our
195<tt>Module</tt> that the PassManager requires. It provides certain flexibility
196that we're not going to take advantage of here, so I won't dive into what it is
197all about.</p>
198
199<p>The meat of the matter is the definition of the "<tt>OurFPM</tt>". It
200requires a pointer to the <tt>Module</tt> (through the <tt>ModuleProvider</tt>)
201to construct itself. Once it is set up, we use a series of "add" calls to add
202a bunch of LLVM passes. The first pass is basically boilerplate, it adds a pass
203so that later optimizations know how the data structures in the program are
204layed out. The "<tt>TheExecutionEngine</tt>" variable is related to the JIT,
205which we will get to in the next section.</p>
206
207<p>In this case, we choose to add 4 optimization passes. The passes we chose
208here are a pretty standard set of "cleanup" optimizations that are useful for
209a wide variety of code. I won't delve into what they do, but believe that they
210are a good starting place.</p>
211
212<p>Once the passmanager, is set up, we need to make use of it. We do this by
213running it after our newly created function is constructed (in
214<tt>FunctionAST::Codegen</tt>), but before it is returned to the client:</p>
215
216<div class="doc_code">
217<pre>
218 if (Value *RetVal = Body->Codegen()) {
219 // Finish off the function.
220 Builder.CreateRet(RetVal);
221
222 // Validate the generated code, checking for consistency.
223 verifyFunction(*TheFunction);
224
225 // Optimize the function.
226 TheFPM->run(*TheFunction);
227
228 return TheFunction;
229 }
230</pre>
231</div>
232
233<p>As you can see, this is pretty straight-forward. The
234<tt>FunctionPassManager</tt> optimizes and updates the LLVM Function* in place,
235improving (hopefully) its body. With this in place, we can try our test above
236again:</p>
237
238<div class="doc_code">
239<pre>
240ready&gt; <b>def test(x) (1+2+x)*(x+(1+2));</b>
241ready> Read function definition:
242define double @test(double %x) {
243entry:
244 %addtmp = add double %x, 3.000000e+00
245 %multmp = mul double %addtmp, %addtmp
246 ret double %multmp
247}
248</pre>
249</div>
250
251<p>As expected, we now get our nicely optimized code, saving a floating point
252add from the program.</p>
253
254<p>LLVM provides a wide variety of optimizations that can be used in certain
255circumstances. Unfortunately we don't have a good centralized description of
256what every pass does, but you can check out the ones that <tt>llvm-gcc</tt> or
257<tt>llvm-ld</tt> run to get started. The "<tt>opt</tt>" tool allows you to
258experiment with passes from the command line, so you can see if they do
259anything.</p>
260
261<p>Now that we have reasonable code coming out of our front-end, lets talk about
262executing it!</p>
263
264</div>
265
266<!-- *********************************************************************** -->
267<div class="doc_section"><a name="jit">Adding a JIT Compiler</a></div>
268<!-- *********************************************************************** -->
269
270<div class="doc_text">
271
272<p>Once the code is available in LLVM IR form a wide variety of tools can be
273applied to it. For example, you can run optimizations on it (as we did above),
274you can dump it out in textual or binary forms, you can compile the code to an
275assembly file (.s) for some target, or you can JIT compile it. The nice thing
276about the LLVM IR representation is that it is the common currency between many
277different parts of the compiler.
278</p>
279
280<p>In this chapter, we'll add JIT compiler support to our interpreter. The
281basic idea that we want for Kaleidoscope is to have the user enter function
282bodies as they do now, but immediately evaluate the top-level expressions they
283type in. For example, if they type in "1 + 2;", we should evaluate and print
284out 3. If they define a function, they should be able to call it from the
285command line.</p>
286
287<p>In order to do this, we first declare and initialize the JIT. This is done
288by adding a global variable and a call in <tt>main</tt>:</p>
289
290<div class="doc_code">
291<pre>
292static ExecutionEngine *TheExecutionEngine;
293...
294int main() {
295 ..
296 // Create the JIT.
297 TheExecutionEngine = ExecutionEngine::create(TheModule);
298 ..
299}
300</pre>
301</div>
302
303<p>This creates an abstract "Execution Engine" which can be either a JIT
304compiler or the LLVM interpreter. LLVM will automatically pick a JIT compiler
305for you if one is available for your platform, otherwise it will fall back to
306the interpreter.</p>
307
308<p>Once the <tt>ExecutionEngine</tt> is created, the JIT is ready to be used.
309There are a variety of APIs that are useful, but the most simple one is the
310"<tt>getPointerToFunction(F)</tt>" method. This method JIT compiles the
311specified LLVM Function and returns a function pointer to the generated machine
312code. In our case, this means that we can change the code that parses a
313top-level expression to look like this:</p>
314
315<div class="doc_code">
316<pre>
317static void HandleTopLevelExpression() {
318 // Evaluate a top level expression into an anonymous function.
319 if (FunctionAST *F = ParseTopLevelExpr()) {
320 if (Function *LF = F-&gt;Codegen()) {
321 LF->dump(); // Dump the function for exposition purposes.
322
323 // JIT the function, returning a function pointer.
324 void *FPtr = TheExecutionEngine-&gt;getPointerToFunction(LF);
325
326 // Cast it to the right type (takes no arguments, returns a double) so we
327 // can call it as a native function.
328 double (*FP)() = (double (*)())FPtr;
329 fprintf(stderr, "Evaluated to %f\n", FP());
330 }
331</pre>
332</div>
333
334<p>Recall that we compile top-level expressions into a self-contained LLVM
335function that takes no arguments and returns the computed double. Because the
336LLVM JIT compiler matches the native platform ABI, this means that you can just
337cast the result pointer to a function pointer of that type and call it directly.
338As such, there is no difference between JIT compiled code and native machine
339code that is statically linked into your application.</p>
340
341<p>With just these two changes, lets see how Kaleidoscope works now!</p>
342
343<div class="doc_code">
344<pre>
345ready&gt; <b>4+5;</b>
346define double @""() {
347entry:
348 ret double 9.000000e+00
349}
350
351<em>Evaluated to 9.000000</em>
352</pre>
353</div>
354
355<p>Well this looks like it is basically working. The dump of the function
356shows the "no argument function that always returns double" that we synthesize
357for each top level expression that is typed it. This demonstrates very basic
358functionality, but can we do more?</p>
359
360<div class="doc_code">
361<pre>
362ready&gt; def testfunc(x y) x + y*2; </b>
363Read function definition:
364define double @testfunc(double %x, double %y) {
365entry:
366 %multmp = mul double %y, 2.000000e+00
367 %addtmp = add double %multmp, %x
368 ret double %addtmp
369}
370
371ready&gt; <b>testfunc(4, 10);</b>
372define double @""() {
373entry:
374 %calltmp = call double @testfunc( double 4.000000e+00, double 1.000000e+01 )
375 ret double %calltmp
376}
377
378<em>Evaluated to 24.000000</em>
379</pre>
380</div>
381
382<p>This illustrates that we can now call user code, but it is a bit subtle what
383is going on here. Note that we only invoke the JIT on the anonymous functions
384that <em>calls testfunc</em>, but we never invoked it on <em>testfunc
385itself</em>.</p>
386
387<p>What actually happened here is that the anonymous function is
388JIT'd when requested. When the Kaleidoscope app calls through the function
389pointer that is returned, the anonymous function starts executing. It ends up
390making the call for the "testfunc" function, and ends up in a stub that invokes
391the JIT, lazily, on testfunc. Once the JIT finishes lazily compiling testfunc,
392it returns and the code reexecutes the call.</p>
393
394<p>In summary, the JIT will lazily JIT code on the fly as it is needed. The
395JIT provides a number of other more advanced interfaces for things like freeing
396allocated machine code, rejit'ing functions to update them, etc. However, even
397with this simple code, we get some surprisingly powerful capabilities - check
398this out (I removed the dump of the anonymous functions, you should get the idea
399by now :) :</p>
400
401<div class="doc_code">
402<pre>
403ready&gt; <b>extern sin(x);</b>
404Read extern:
405declare double @sin(double)
406
407ready&gt; <b>extern cos(x);</b>
408Read extern:
409declare double @cos(double)
410
411ready&gt; <b>sin(1.0);</b>
412<em>Evaluated to 0.841471</em>
413ready&gt; <b>def foo(x) sin(x)*sin(x) + cos(x)*cos(x);</b>
414Read function definition:
415define double @foo(double %x) {
416entry:
417 %calltmp = call double @sin( double %x )
418 %multmp = mul double %calltmp, %calltmp
419 %calltmp2 = call double @cos( double %x )
420 %multmp4 = mul double %calltmp2, %calltmp2
421 %addtmp = add double %multmp, %multmp4
422 ret double %addtmp
423}
424
425ready&gt; <b>foo(4.0);</b>
426<em>Evaluated to 1.000000</em>
427</pre>
428</div>
429
430<p>Whoa, how does the JIT know about sin and cos? The answer is simple: in this
431example, the JIT started execution of a function and got to a function call. It
432realized that the function was not yet JIT compiled and invoked the standard set
433of routines to resolve the function. In this case, there is no body defined
434for the function, so the JIT ended up calling "<tt>dlsym("sin")</tt>" on itself.
435Since "<tt>sin</tt>" is defined within the JIT's address space, it simply
436patches up calls in the module to call the libm version of <tt>sin</tt>
437directly.</p>
438
439<p>The LLVM JIT provides a number of interfaces (look in the
440<tt>ExecutionEngine.h</tt> file) for controlling how unknown functions get
441resolved. It allows you to establish explicit mappings between IR objects and
442addresses (useful for LLVM global variables that you want to map to static
443tables, for example), allows you to dynamically decide on the fly based on the
444function name, and even allows you to have the JIT abort itself if any lazy
445compilation is attempted.</p>
446
447<p>This completes the JIT and optimizer chapter of the Kaleidoscope tutorial. At
448this point, we can compile a non-Turing-complete programming language, optimize
449and JIT compile it in a user-driven way. Next up we'll look into <a
450href="LangImpl5.html">extending the language with control flow constructs</a>,
451tackling some interesting LLVM IR issues along the way.</p>
452
453</div>
454
455<!-- *********************************************************************** -->
456<div class="doc_section"><a name="code">Full Code Listing</a></div>
457<!-- *********************************************************************** -->
458
459<div class="doc_text">
460
461<p>
462Here is the complete code listing for our running example, enhanced with the
463LLVM JIT and optimizer. To build this example, use:
464</p>
465
466<div class="doc_code">
467<pre>
468 # Compile
469 g++ -g toy.cpp `llvm-config --cppflags --ldflags --libs core jit native` -O3 -o toy
470 # Run
471 ./toy
472</pre>
473</div>
474
475<p>Here is the code:</p>
476
477<div class="doc_code">
478<pre>
479#include "llvm/DerivedTypes.h"
480#include "llvm/ExecutionEngine/ExecutionEngine.h"
481#include "llvm/Module.h"
482#include "llvm/ModuleProvider.h"
483#include "llvm/PassManager.h"
484#include "llvm/Analysis/Verifier.h"
485#include "llvm/Target/TargetData.h"
486#include "llvm/Transforms/Scalar.h"
487#include "llvm/Support/LLVMBuilder.h"
488#include &lt;cstdio&gt;
489#include &lt;string&gt;
490#include &lt;map&gt;
491#include &lt;vector&gt;
492using namespace llvm;
493
494//===----------------------------------------------------------------------===//
495// Lexer
496//===----------------------------------------------------------------------===//
497
498// The lexer returns tokens [0-255] if it is an unknown character, otherwise one
499// of these for known things.
500enum Token {
501 tok_eof = -1,
502
503 // commands
504 tok_def = -2, tok_extern = -3,
505
506 // primary
507 tok_identifier = -4, tok_number = -5,
508};
509
510static std::string IdentifierStr; // Filled in if tok_identifier
511static double NumVal; // Filled in if tok_number
512
513/// gettok - Return the next token from standard input.
514static int gettok() {
515 static int LastChar = ' ';
516
517 // Skip any whitespace.
518 while (isspace(LastChar))
519 LastChar = getchar();
520
521 if (isalpha(LastChar)) { // identifier: [a-zA-Z][a-zA-Z0-9]*
522 IdentifierStr = LastChar;
523 while (isalnum((LastChar = getchar())))
524 IdentifierStr += LastChar;
525
526 if (IdentifierStr == "def") return tok_def;
527 if (IdentifierStr == "extern") return tok_extern;
528 return tok_identifier;
529 }
530
531 if (isdigit(LastChar) || LastChar == '.') { // Number: [0-9.]+
532 std::string NumStr;
533 do {
534 NumStr += LastChar;
535 LastChar = getchar();
536 } while (isdigit(LastChar) || LastChar == '.');
537
538 NumVal = strtod(NumStr.c_str(), 0);
539 return tok_number;
540 }
541
542 if (LastChar == '#') {
543 // Comment until end of line.
544 do LastChar = getchar();
545 while (LastChar != EOF &amp;&amp; LastChar != '\n' &amp; LastChar != '\r');
546
547 if (LastChar != EOF)
548 return gettok();
549 }
550
551 // Check for end of file. Don't eat the EOF.
552 if (LastChar == EOF)
553 return tok_eof;
554
555 // Otherwise, just return the character as its ascii value.
556 int ThisChar = LastChar;
557 LastChar = getchar();
558 return ThisChar;
559}
560
561//===----------------------------------------------------------------------===//
562// Abstract Syntax Tree (aka Parse Tree)
563//===----------------------------------------------------------------------===//
564
Chris Lattnerc0b42e92007-10-23 06:27:55 +0000[diff] [blame]565/// ExprAST - Base class for all expression nodes.
566class ExprAST {
567public:
568 virtual ~ExprAST() {}
569 virtual Value *Codegen() = 0;
570};
571
572/// NumberExprAST - Expression class for numeric literals like "1.0".
573class NumberExprAST : public ExprAST {
574 double Val;
575public:
Chris Lattner118749e2007-10-25 06:23:36 +0000[diff] [blame^]576 NumberExprAST(double val) : Val(val) {}
Chris Lattnerc0b42e92007-10-23 06:27:55 +0000[diff] [blame]577 virtual Value *Codegen();
578};
Chris Lattner118749e2007-10-25 06:23:36 +0000[diff] [blame^]579
580/// VariableExprAST - Expression class for referencing a variable, like "a".
581class VariableExprAST : public ExprAST {
582 std::string Name;
583public:
584 VariableExprAST(const std::string &amp;name) : Name(name) {}
585 virtual Value *Codegen();
586};
587
588/// BinaryExprAST - Expression class for a binary operator.
589class BinaryExprAST : public ExprAST {
590 char Op;
591 ExprAST *LHS, *RHS;
592public:
593 BinaryExprAST(char op, ExprAST *lhs, ExprAST *rhs)
594 : Op(op), LHS(lhs), RHS(rhs) {}
595 virtual Value *Codegen();
596};
597
598/// CallExprAST - Expression class for function calls.
599class CallExprAST : public ExprAST {
600 std::string Callee;
601 std::vector&lt;ExprAST*&gt; Args;
602public:
603 CallExprAST(const std::string &amp;callee, std::vector&lt;ExprAST*&gt; &amp;args)
604 : Callee(callee), Args(args) {}
605 virtual Value *Codegen();
606};
607
608/// PrototypeAST - This class represents the "prototype" for a function,
609/// which captures its argument names as well as if it is an operator.
610class PrototypeAST {
611 std::string Name;
612 std::vector&lt;std::string&gt; Args;
613public:
614 PrototypeAST(const std::string &amp;name, const std::vector&lt;std::string&gt; &amp;args)
615 : Name(name), Args(args) {}
616
617 Function *Codegen();
618};
619
620/// FunctionAST - This class represents a function definition itself.
621class FunctionAST {
622 PrototypeAST *Proto;
623 ExprAST *Body;
624public:
625 FunctionAST(PrototypeAST *proto, ExprAST *body)
626 : Proto(proto), Body(body) {}
627
628 Function *Codegen();
629};
630
631//===----------------------------------------------------------------------===//
632// Parser
633//===----------------------------------------------------------------------===//
634
635/// CurTok/getNextToken - Provide a simple token buffer. CurTok is the current
636/// token the parser it looking at. getNextToken reads another token from the
637/// lexer and updates CurTok with its results.
638static int CurTok;
639static int getNextToken() {
640 return CurTok = gettok();
641}
642
643/// BinopPrecedence - This holds the precedence for each binary operator that is
644/// defined.
645static std::map&lt;char, int&gt; BinopPrecedence;
646
647/// GetTokPrecedence - Get the precedence of the pending binary operator token.
648static int GetTokPrecedence() {
649 if (!isascii(CurTok))
650 return -1;
651
652 // Make sure it's a declared binop.
653 int TokPrec = BinopPrecedence[CurTok];
654 if (TokPrec &lt;= 0) return -1;
655 return TokPrec;
656}
657
658/// Error* - These are little helper functions for error handling.
659ExprAST *Error(const char *Str) { fprintf(stderr, "Error: %s\n", Str);return 0;}
660PrototypeAST *ErrorP(const char *Str) { Error(Str); return 0; }
661FunctionAST *ErrorF(const char *Str) { Error(Str); return 0; }
662
663static ExprAST *ParseExpression();
664
665/// identifierexpr
666/// ::= identifer
667/// ::= identifer '(' expression* ')'
668static ExprAST *ParseIdentifierExpr() {
669 std::string IdName = IdentifierStr;
670
671 getNextToken(); // eat identifer.
672
673 if (CurTok != '(') // Simple variable ref.
674 return new VariableExprAST(IdName);
675
676 // Call.
677 getNextToken(); // eat (
678 std::vector&lt;ExprAST*&gt; Args;
679 while (1) {
680 ExprAST *Arg = ParseExpression();
681 if (!Arg) return 0;
682 Args.push_back(Arg);
683
684 if (CurTok == ')') break;
685
686 if (CurTok != ',')
687 return Error("Expected ')'");
688 getNextToken();
689 }
690
691 // Eat the ')'.
692 getNextToken();
693
694 return new CallExprAST(IdName, Args);
695}
696
697/// numberexpr ::= number
698static ExprAST *ParseNumberExpr() {
699 ExprAST *Result = new NumberExprAST(NumVal);
700 getNextToken(); // consume the number
701 return Result;
702}
703
704/// parenexpr ::= '(' expression ')'
705static ExprAST *ParseParenExpr() {
706 getNextToken(); // eat (.
707 ExprAST *V = ParseExpression();
708 if (!V) return 0;
709
710 if (CurTok != ')')
711 return Error("expected ')'");
712 getNextToken(); // eat ).
713 return V;
714}
715
716/// primary
717/// ::= identifierexpr
718/// ::= numberexpr
719/// ::= parenexpr
720static ExprAST *ParsePrimary() {
721 switch (CurTok) {
722 default: return Error("unknown token when expecting an expression");
723 case tok_identifier: return ParseIdentifierExpr();
724 case tok_number: return ParseNumberExpr();
725 case '(': return ParseParenExpr();
726 }
727}
728
729/// binoprhs
730/// ::= ('+' primary)*
731static ExprAST *ParseBinOpRHS(int ExprPrec, ExprAST *LHS) {
732 // If this is a binop, find its precedence.
733 while (1) {
734 int TokPrec = GetTokPrecedence();
735
736 // If this is a binop that binds at least as tightly as the current binop,
737 // consume it, otherwise we are done.
738 if (TokPrec &lt; ExprPrec)
739 return LHS;
740
741 // Okay, we know this is a binop.
742 int BinOp = CurTok;
743 getNextToken(); // eat binop
744
745 // Parse the primary expression after the binary operator.
746 ExprAST *RHS = ParsePrimary();
747 if (!RHS) return 0;
748
749 // If BinOp binds less tightly with RHS than the operator after RHS, let
750 // the pending operator take RHS as its LHS.
751 int NextPrec = GetTokPrecedence();
752 if (TokPrec &lt; NextPrec) {
753 RHS = ParseBinOpRHS(TokPrec+1, RHS);
754 if (RHS == 0) return 0;
755 }
756
757 // Merge LHS/RHS.
758 LHS = new BinaryExprAST(BinOp, LHS, RHS);
759 }
760}
761
762/// expression
763/// ::= primary binoprhs
764///
765static ExprAST *ParseExpression() {
766 ExprAST *LHS = ParsePrimary();
767 if (!LHS) return 0;
768
769 return ParseBinOpRHS(0, LHS);
770}
771
772/// prototype
773/// ::= id '(' id* ')'
774static PrototypeAST *ParsePrototype() {
775 if (CurTok != tok_identifier)
776 return ErrorP("Expected function name in prototype");
777
778 std::string FnName = IdentifierStr;
779 getNextToken();
780
781 if (CurTok != '(')
782 return ErrorP("Expected '(' in prototype");
783
784 std::vector&lt;std::string&gt; ArgNames;
785 while (getNextToken() == tok_identifier)
786 ArgNames.push_back(IdentifierStr);
787 if (CurTok != ')')
788 return ErrorP("Expected ')' in prototype");
789
790 // success.
791 getNextToken(); // eat ')'.
792
793 return new PrototypeAST(FnName, ArgNames);
794}
795
796/// definition ::= 'def' prototype expression
797static FunctionAST *ParseDefinition() {
798 getNextToken(); // eat def.
799 PrototypeAST *Proto = ParsePrototype();
800 if (Proto == 0) return 0;
801
802 if (ExprAST *E = ParseExpression())
803 return new FunctionAST(Proto, E);
804 return 0;
805}
806
807/// toplevelexpr ::= expression
808static FunctionAST *ParseTopLevelExpr() {
809 if (ExprAST *E = ParseExpression()) {
810 // Make an anonymous proto.
811 PrototypeAST *Proto = new PrototypeAST("", std::vector&lt;std::string&gt;());
812 return new FunctionAST(Proto, E);
813 }
814 return 0;
815}
816
817/// external ::= 'extern' prototype
818static PrototypeAST *ParseExtern() {
819 getNextToken(); // eat extern.
820 return ParsePrototype();
821}
822
823//===----------------------------------------------------------------------===//
824// Code Generation
825//===----------------------------------------------------------------------===//
826
827static Module *TheModule;
828static LLVMFoldingBuilder Builder;
829static std::map&lt;std::string, Value*&gt; NamedValues;
830static FunctionPassManager *TheFPM;
831
832Value *ErrorV(const char *Str) { Error(Str); return 0; }
833
834Value *NumberExprAST::Codegen() {
835 return ConstantFP::get(Type::DoubleTy, APFloat(Val));
836}
837
838Value *VariableExprAST::Codegen() {
839 // Look this variable up in the function.
840 Value *V = NamedValues[Name];
841 return V ? V : ErrorV("Unknown variable name");
842}
843
844Value *BinaryExprAST::Codegen() {
845 Value *L = LHS-&gt;Codegen();
846 Value *R = RHS-&gt;Codegen();
847 if (L == 0 || R == 0) return 0;
848
849 switch (Op) {
850 case '+': return Builder.CreateAdd(L, R, "addtmp");
851 case '-': return Builder.CreateSub(L, R, "subtmp");
852 case '*': return Builder.CreateMul(L, R, "multmp");
853 case '&lt;':
854 L = Builder.CreateFCmpULT(L, R, "multmp");
855 // Convert bool 0/1 to double 0.0 or 1.0
856 return Builder.CreateUIToFP(L, Type::DoubleTy, "booltmp");
857 default: return ErrorV("invalid binary operator");
858 }
859}
860
861Value *CallExprAST::Codegen() {
862 // Look up the name in the global module table.
863 Function *CalleeF = TheModule-&gt;getFunction(Callee);
864 if (CalleeF == 0)
865 return ErrorV("Unknown function referenced");
866
867 // If argument mismatch error.
868 if (CalleeF-&gt;arg_size() != Args.size())
869 return ErrorV("Incorrect # arguments passed");
870
871 std::vector&lt;Value*&gt; ArgsV;
872 for (unsigned i = 0, e = Args.size(); i != e; ++i) {
873 ArgsV.push_back(Args[i]-&gt;Codegen());
874 if (ArgsV.back() == 0) return 0;
875 }
876
877 return Builder.CreateCall(CalleeF, ArgsV.begin(), ArgsV.end(), "calltmp");
878}
879
880Function *PrototypeAST::Codegen() {
881 // Make the function type: double(double,double) etc.
882 std::vector&lt;const Type*&gt; Doubles(Args.size(), Type::DoubleTy);
883 FunctionType *FT = FunctionType::get(Type::DoubleTy, Doubles, false);
884
885 Function *F = new Function(FT, Function::ExternalLinkage, Name, TheModule);
886
887 // If F conflicted, there was already something named 'Name'. If it has a
888 // body, don't allow redefinition or reextern.
889 if (F-&gt;getName() != Name) {
890 // Delete the one we just made and get the existing one.
891 F-&gt;eraseFromParent();
892 F = TheModule-&gt;getFunction(Name);
893
894 // If F already has a body, reject this.
895 if (!F-&gt;empty()) {
896 ErrorF("redefinition of function");
897 return 0;
898 }
899
900 // If F took a different number of args, reject.
901 if (F-&gt;arg_size() != Args.size()) {
902 ErrorF("redefinition of function with different # args");
903 return 0;
904 }
905 }
906
907 // Set names for all arguments.
908 unsigned Idx = 0;
909 for (Function::arg_iterator AI = F-&gt;arg_begin(); Idx != Args.size();
910 ++AI, ++Idx) {
911 AI-&gt;setName(Args[Idx]);
912
913 // Add arguments to variable symbol table.
914 NamedValues[Args[Idx]] = AI;
915 }
916
917 return F;
918}
919
920Function *FunctionAST::Codegen() {
921 NamedValues.clear();
922
923 Function *TheFunction = Proto-&gt;Codegen();
924 if (TheFunction == 0)
925 return 0;
926
927 // Create a new basic block to start insertion into.
928 BasicBlock *BB = new BasicBlock("entry", TheFunction);
929 Builder.SetInsertPoint(BB);
930
931 if (Value *RetVal = Body-&gt;Codegen()) {
932 // Finish off the function.
933 Builder.CreateRet(RetVal);
934
935 // Validate the generated code, checking for consistency.
936 verifyFunction(*TheFunction);
937
938 // Optimize the function.
939 TheFPM-&gt;run(*TheFunction);
940
941 return TheFunction;
942 }
943
944 // Error reading body, remove function.
945 TheFunction-&gt;eraseFromParent();
946 return 0;
947}
948
949//===----------------------------------------------------------------------===//
950// Top-Level parsing and JIT Driver
951//===----------------------------------------------------------------------===//
952
953static ExecutionEngine *TheExecutionEngine;
954
955static void HandleDefinition() {
956 if (FunctionAST *F = ParseDefinition()) {
957 if (Function *LF = F-&gt;Codegen()) {
958 fprintf(stderr, "Read function definition:");
959 LF-&gt;dump();
960 }
961 } else {
962 // Skip token for error recovery.
963 getNextToken();
964 }
965}
966
967static void HandleExtern() {
968 if (PrototypeAST *P = ParseExtern()) {
969 if (Function *F = P-&gt;Codegen()) {
970 fprintf(stderr, "Read extern: ");
971 F-&gt;dump();
972 }
973 } else {
974 // Skip token for error recovery.
975 getNextToken();
976 }
977}
978
979static void HandleTopLevelExpression() {
980 // Evaluate a top level expression into an anonymous function.
981 if (FunctionAST *F = ParseTopLevelExpr()) {
982 if (Function *LF = F-&gt;Codegen()) {
983 // JIT the function, returning a function pointer.
984 void *FPtr = TheExecutionEngine-&gt;getPointerToFunction(LF);
985
986 // Cast it to the right type (takes no arguments, returns a double) so we
987 // can call it as a native function.
988 double (*FP)() = (double (*)())FPtr;
989 fprintf(stderr, "Evaluated to %f\n", FP());
990 }
991 } else {
992 // Skip token for error recovery.
993 getNextToken();
994 }
995}
996
997/// top ::= definition | external | expression | ';'
998static void MainLoop() {
999 while (1) {
1000 fprintf(stderr, "ready&gt; ");
1001 switch (CurTok) {
1002 case tok_eof: return;
1003 case ';': getNextToken(); break; // ignore top level semicolons.
1004 case tok_def: HandleDefinition(); break;
1005 case tok_extern: HandleExtern(); break;
1006 default: HandleTopLevelExpression(); break;
1007 }
1008 }
1009}
1010
1011
1012
1013//===----------------------------------------------------------------------===//
1014// "Library" functions that can be "extern'd" from user code.
1015//===----------------------------------------------------------------------===//
1016
1017/// putchard - putchar that takes a double and returns 0.
1018extern "C"
1019double putchard(double X) {
1020 putchar((char)X);
1021 return 0;
1022}
1023
1024//===----------------------------------------------------------------------===//
1025// Main driver code.
1026//===----------------------------------------------------------------------===//
1027
1028int main() {
1029 // Install standard binary operators.
1030 // 1 is lowest precedence.
1031 BinopPrecedence['&lt;'] = 10;
1032 BinopPrecedence['+'] = 20;
1033 BinopPrecedence['-'] = 20;
1034 BinopPrecedence['*'] = 40; // highest.
1035
1036 // Prime the first token.
1037 fprintf(stderr, "ready&gt; ");
1038 getNextToken();
1039
1040 // Make the module, which holds all the code.
1041 TheModule = new Module("my cool jit");
1042
1043 // Create the JIT.
1044 TheExecutionEngine = ExecutionEngine::create(TheModule);
1045
1046 {
1047 ExistingModuleProvider OurModuleProvider(TheModule);
1048 FunctionPassManager OurFPM(&amp;OurModuleProvider);
1049
1050 // Set up the optimizer pipeline. Start with registering info about how the
1051 // target lays out data structures.
1052 OurFPM.add(new TargetData(*TheExecutionEngine-&gt;getTargetData()));
1053 // Do simple "peephole" optimizations and bit-twiddling optzns.
1054 OurFPM.add(createInstructionCombiningPass());
1055 // Reassociate expressions.
1056 OurFPM.add(createReassociatePass());
1057 // Eliminate Common SubExpressions.
1058 OurFPM.add(createGVNPass());
1059 // Simplify the control flow graph (deleting unreachable blocks, etc).
1060 OurFPM.add(createCFGSimplificationPass());
1061
1062 // Set the global so the code gen can use this.
1063 TheFPM = &amp;OurFPM;
1064
1065 // Run the main "interpreter loop" now.
1066 MainLoop();
1067
1068 TheFPM = 0;
1069 } // Free module provider and pass manager.
1070
1071
1072 // Print out all of the generated code.
1073 TheModule-&gt;dump();
1074 return 0;
1075}
Chris Lattnerc0b42e92007-10-23 06:27:55 +0000[diff] [blame]1076</pre>
1077</div>
1078
Chris Lattnerc0b42e92007-10-23 06:27:55 +0000[diff] [blame]1079</div>
1080
1081<!-- *********************************************************************** -->
1082<hr>
1083<address>
1084 <a href="http://jigsaw.w3.org/css-validator/check/referer"><img
1085 src="http://jigsaw.w3.org/css-validator/images/vcss" alt="Valid CSS!"></a>
1086 <a href="http://validator.w3.org/check/referer"><img
Chris Lattner8eef4b22007-10-23 06:30:50 +0000[diff] [blame]1087 src="http://www.w3.org/Icons/valid-html401" alt="Valid HTML 4.01!"></a>
Chris Lattnerc0b42e92007-10-23 06:27:55 +0000[diff] [blame]1088
1089 <a href="mailto:sabre@nondot.org">Chris Lattner</a><br>
1090 <a href="http://llvm.org">The LLVM Compiler Infrastructure</a><br>
1091 Last modified: $Date: 2007-10-17 11:05:13 -0700 (Wed, 17 Oct 2007) $
1092</address>
1093</body>
1094</html>