blob: a671d0c37f9d830e689989fafe61427aa671ec42 [file] [log] [blame]
Sean Silvad7fb3962012-12-05 00:26:32 +00001==============================================
2Kaleidoscope: Adding JIT and Optimizer Support
3==============================================
4
5.. contents::
6 :local:
7
Sean Silvad7fb3962012-12-05 00:26:32 +00008Chapter 4 Introduction
9======================
10
11Welcome to Chapter 4 of the "`Implementing a language with
12LLVM <index.html>`_" tutorial. Chapters 1-3 described the implementation
13of a simple language and added support for generating LLVM IR. This
14chapter describes two new techniques: adding optimizer support to your
15language, and adding JIT compiler support. These additions will
16demonstrate how to get nice, efficient code for the Kaleidoscope
17language.
18
19Trivial Constant Folding
20========================
21
22Our demonstration for Chapter 3 is elegant and easy to extend.
23Unfortunately, it does not produce wonderful code. The IRBuilder,
24however, does give us obvious optimizations when compiling simple code:
25
26::
27
28 ready> def test(x) 1+2+x;
29 Read function definition:
30 define double @test(double %x) {
31 entry:
32 %addtmp = fadd double 3.000000e+00, %x
33 ret double %addtmp
34 }
35
36This code is not a literal transcription of the AST built by parsing the
37input. That would be:
38
39::
40
41 ready> def test(x) 1+2+x;
42 Read function definition:
43 define double @test(double %x) {
44 entry:
45 %addtmp = fadd double 2.000000e+00, 1.000000e+00
46 %addtmp1 = fadd double %addtmp, %x
47 ret double %addtmp1
48 }
49
50Constant folding, as seen above, in particular, is a very common and
51very important optimization: so much so that many language implementors
52implement constant folding support in their AST representation.
53
54With LLVM, you don't need this support in the AST. Since all calls to
55build LLVM IR go through the LLVM IR builder, the builder itself checked
56to see if there was a constant folding opportunity when you call it. If
57so, it just does the constant fold and return the constant instead of
58creating an instruction.
59
60Well, that was easy :). In practice, we recommend always using
61``IRBuilder`` when generating code like this. It has no "syntactic
62overhead" for its use (you don't have to uglify your compiler with
63constant checks everywhere) and it can dramatically reduce the amount of
64LLVM IR that is generated in some cases (particular for languages with a
65macro preprocessor or that use a lot of constants).
66
67On the other hand, the ``IRBuilder`` is limited by the fact that it does
68all of its analysis inline with the code as it is built. If you take a
69slightly more complex example:
70
71::
72
73 ready> def test(x) (1+2+x)*(x+(1+2));
74 ready> Read function definition:
75 define double @test(double %x) {
76 entry:
77 %addtmp = fadd double 3.000000e+00, %x
78 %addtmp1 = fadd double %x, 3.000000e+00
79 %multmp = fmul double %addtmp, %addtmp1
80 ret double %multmp
81 }
82
83In this case, the LHS and RHS of the multiplication are the same value.
84We'd really like to see this generate "``tmp = x+3; result = tmp*tmp;``"
85instead of computing "``x+3``" twice.
86
87Unfortunately, no amount of local analysis will be able to detect and
88correct this. This requires two transformations: reassociation of
89expressions (to make the add's lexically identical) and Common
90Subexpression Elimination (CSE) to delete the redundant add instruction.
91Fortunately, LLVM provides a broad range of optimizations that you can
92use, in the form of "passes".
93
94LLVM Optimization Passes
95========================
96
97LLVM provides many optimization passes, which do many different sorts of
98things and have different tradeoffs. Unlike other systems, LLVM doesn't
99hold to the mistaken notion that one set of optimizations is right for
100all languages and for all situations. LLVM allows a compiler implementor
101to make complete decisions about what optimizations to use, in which
102order, and in what situation.
103
104As a concrete example, LLVM supports both "whole module" passes, which
105look across as large of body of code as they can (often a whole file,
106but if run at link time, this can be a substantial portion of the whole
107program). It also supports and includes "per-function" passes which just
108operate on a single function at a time, without looking at other
109functions. For more information on passes and how they are run, see the
110`How to Write a Pass <../WritingAnLLVMPass.html>`_ document and the
111`List of LLVM Passes <../Passes.html>`_.
112
113For Kaleidoscope, we are currently generating functions on the fly, one
114at a time, as the user types them in. We aren't shooting for the
115ultimate optimization experience in this setting, but we also want to
116catch the easy and quick stuff where possible. As such, we will choose
117to run a few per-function optimizations as the user types the function
118in. If we wanted to make a "static Kaleidoscope compiler", we would use
119exactly the code we have now, except that we would defer running the
120optimizer until the entire file has been parsed.
121
122In order to get per-function optimizations going, we need to set up a
Alex Denisov596e9792015-12-15 20:50:29 +0000123`FunctionPassManager <../WritingAnLLVMPass.html#what-passmanager-doesr>`_ to hold
Sean Silvad7fb3962012-12-05 00:26:32 +0000124and organize the LLVM optimizations that we want to run. Once we have
Lang Hames2d789c32015-08-26 03:07:41 +0000125that, we can add a set of optimizations to run. We'll need a new
126FunctionPassManager for each module that we want to optimize, so we'll
127write a function to create and initialize both the module and pass manager
128for us:
Sean Silvad7fb3962012-12-05 00:26:32 +0000129
130.. code-block:: c++
131
Lang Hames2d789c32015-08-26 03:07:41 +0000132 void InitializeModuleAndPassManager(void) {
133 // Open a new module.
134 TheModule = llvm::make_unique<Module>("my cool jit", getGlobalContext());
135 TheModule->setDataLayout(TheJIT->getTargetMachine().createDataLayout());
Sean Silvad7fb3962012-12-05 00:26:32 +0000136
Lang Hames2d789c32015-08-26 03:07:41 +0000137 // Create a new pass manager attached to it.
138 TheFPM = llvm::make_unique<FunctionPassManager>(TheModule.get());
139
Sean Silvad7fb3962012-12-05 00:26:32 +0000140 // Provide basic AliasAnalysis support for GVN.
Lang Hames2d789c32015-08-26 03:07:41 +0000141 TheFPM.add(createBasicAliasAnalysisPass());
Sean Silvad7fb3962012-12-05 00:26:32 +0000142 // Do simple "peephole" optimizations and bit-twiddling optzns.
Lang Hames2d789c32015-08-26 03:07:41 +0000143 TheFPM.add(createInstructionCombiningPass());
Sean Silvad7fb3962012-12-05 00:26:32 +0000144 // Reassociate expressions.
Lang Hames2d789c32015-08-26 03:07:41 +0000145 TheFPM.add(createReassociatePass());
Sean Silvad7fb3962012-12-05 00:26:32 +0000146 // Eliminate Common SubExpressions.
Lang Hames2d789c32015-08-26 03:07:41 +0000147 TheFPM.add(createGVNPass());
Sean Silvad7fb3962012-12-05 00:26:32 +0000148 // Simplify the control flow graph (deleting unreachable blocks, etc).
Lang Hames2d789c32015-08-26 03:07:41 +0000149 TheFPM.add(createCFGSimplificationPass());
Sean Silvad7fb3962012-12-05 00:26:32 +0000150
Lang Hames2d789c32015-08-26 03:07:41 +0000151 TheFPM.doInitialization();
152 }
Sean Silvad7fb3962012-12-05 00:26:32 +0000153
Lang Hames2d789c32015-08-26 03:07:41 +0000154This code initializes the global module ``TheModule``, and the function pass
Alex Denisov596e9792015-12-15 20:50:29 +0000155manager ``TheFPM``, which is attached to ``TheModule``. Once the pass manager is
Lang Hames2d789c32015-08-26 03:07:41 +0000156set up, we use a series of "add" calls to add a bunch of LLVM passes.
Sean Silvad7fb3962012-12-05 00:26:32 +0000157
Lang Hames2d789c32015-08-26 03:07:41 +0000158In this case, we choose to add five passes: one analysis pass (alias analysis),
159and four optimization passes. The passes we choose here are a pretty standard set
160of "cleanup" optimizations that are useful for a wide variety of code. I won't
161delve into what they do but, believe me, they are a good starting place :).
Sean Silvad7fb3962012-12-05 00:26:32 +0000162
163Once the PassManager is set up, we need to make use of it. We do this by
164running it after our newly created function is constructed (in
Lang Hames2d789c32015-08-26 03:07:41 +0000165``FunctionAST::codegen()``), but before it is returned to the client:
Sean Silvad7fb3962012-12-05 00:26:32 +0000166
167.. code-block:: c++
168
Lang Hames2d789c32015-08-26 03:07:41 +0000169 if (Value *RetVal = Body->codegen()) {
Sean Silvad7fb3962012-12-05 00:26:32 +0000170 // Finish off the function.
171 Builder.CreateRet(RetVal);
172
173 // Validate the generated code, checking for consistency.
174 verifyFunction(*TheFunction);
175
176 // Optimize the function.
177 TheFPM->run(*TheFunction);
178
179 return TheFunction;
180 }
181
182As you can see, this is pretty straightforward. The
183``FunctionPassManager`` optimizes and updates the LLVM Function\* in
184place, improving (hopefully) its body. With this in place, we can try
185our test above again:
186
187::
188
189 ready> def test(x) (1+2+x)*(x+(1+2));
190 ready> Read function definition:
191 define double @test(double %x) {
192 entry:
193 %addtmp = fadd double %x, 3.000000e+00
194 %multmp = fmul double %addtmp, %addtmp
195 ret double %multmp
196 }
197
198As expected, we now get our nicely optimized code, saving a floating
199point add instruction from every execution of this function.
200
201LLVM provides a wide variety of optimizations that can be used in
202certain circumstances. Some `documentation about the various
203passes <../Passes.html>`_ is available, but it isn't very complete.
204Another good source of ideas can come from looking at the passes that
205``Clang`` runs to get started. The "``opt``" tool allows you to
206experiment with passes from the command line, so you can see if they do
207anything.
208
209Now that we have reasonable code coming out of our front-end, lets talk
210about executing it!
211
212Adding a JIT Compiler
213=====================
214
215Code that is available in LLVM IR can have a wide variety of tools
216applied to it. For example, you can run optimizations on it (as we did
217above), you can dump it out in textual or binary forms, you can compile
218the code to an assembly file (.s) for some target, or you can JIT
219compile it. The nice thing about the LLVM IR representation is that it
220is the "common currency" between many different parts of the compiler.
221
222In this section, we'll add JIT compiler support to our interpreter. The
223basic idea that we want for Kaleidoscope is to have the user enter
224function bodies as they do now, but immediately evaluate the top-level
225expressions they type in. For example, if they type in "1 + 2;", we
226should evaluate and print out 3. If they define a function, they should
227be able to call it from the command line.
228
229In order to do this, we first declare and initialize the JIT. This is
Lang Hames2d789c32015-08-26 03:07:41 +0000230done by adding a global variable ``TheJIT``, and initializing it in
231``main``:
Sean Silvad7fb3962012-12-05 00:26:32 +0000232
233.. code-block:: c++
234
Lang Hames2d789c32015-08-26 03:07:41 +0000235 static std::unique_ptr<KaleidoscopeJIT> TheJIT;
Sean Silvad7fb3962012-12-05 00:26:32 +0000236 ...
237 int main() {
238 ..
Lang Hames2d789c32015-08-26 03:07:41 +0000239 TheJIT = llvm::make_unique<KaleidoscopeJIT>();
240
241 // Run the main "interpreter loop" now.
242 MainLoop();
243
244 return 0;
Sean Silvad7fb3962012-12-05 00:26:32 +0000245 }
246
Lang Hames2d789c32015-08-26 03:07:41 +0000247The KaleidoscopeJIT class is a simple JIT built specifically for these
248tutorials. In later chapters we will look at how it works and extend it with
249new features, but for now we will take it as given. Its API is very simple::
250``addModule`` adds an LLVM IR module to the JIT, making its functions
251available for execution; ``removeModule`` removes a module, freeing any
252memory associated with the code in that module; and ``findSymbol`` allows us
253to look up pointers to the compiled code.
Sean Silvad7fb3962012-12-05 00:26:32 +0000254
Lang Hames2d789c32015-08-26 03:07:41 +0000255We can take this simple API and change our code that parses top-level expressions to
256look like this:
Sean Silvad7fb3962012-12-05 00:26:32 +0000257
258.. code-block:: c++
259
260 static void HandleTopLevelExpression() {
261 // Evaluate a top-level expression into an anonymous function.
Lang Hames09bf4c12015-08-18 18:11:06 +0000262 if (auto FnAST = ParseTopLevelExpr()) {
Lang Hames2d789c32015-08-26 03:07:41 +0000263 if (FnAST->codegen()) {
Sean Silvad7fb3962012-12-05 00:26:32 +0000264
Lang Hames2d789c32015-08-26 03:07:41 +0000265 // JIT the module containing the anonymous expression, keeping a handle so
266 // we can free it later.
267 auto H = TheJIT->addModule(std::move(TheModule));
268 InitializeModuleAndPassManager();
Sean Silvad7fb3962012-12-05 00:26:32 +0000269
Lang Hames2d789c32015-08-26 03:07:41 +0000270 // Search the JIT for the __anon_expr symbol.
271 auto ExprSymbol = TheJIT->findSymbol("__anon_expr");
272 assert(ExprSymbol && "Function not found");
273
274 // Get the symbol's address and cast it to the right type (takes no
275 // arguments, returns a double) so we can call it as a native function.
276 double (*FP)() = (double (*)())(intptr_t)ExprSymbol.getAddress();
Sean Silvad7fb3962012-12-05 00:26:32 +0000277 fprintf(stderr, "Evaluated to %f\n", FP());
Lang Hames2d789c32015-08-26 03:07:41 +0000278
279 // Delete the anonymous expression module from the JIT.
280 TheJIT->removeModule(H);
Sean Silvad7fb3962012-12-05 00:26:32 +0000281 }
282
Lang Hames2d789c32015-08-26 03:07:41 +0000283If parsing and codegen succeeed, the next step is to add the module containing
284the top-level expression to the JIT. We do this by calling addModule, which
285triggers code generation for all the functions in the module, and returns a
286handle that can be used to remove the module from the JIT later. Once the module
287has been added to the JIT it can no longer be modified, so we also open a new
288module to hold subsequent code by calling ``InitializeModuleAndPassManager()``.
289
290Once we've added the module to the JIT we need to get a pointer to the final
291generated code. We do this by calling the JIT's findSymbol method, and passing
292the name of the top-level expression function: ``__anon_expr``. Since we just
293added this function, we assert that findSymbol returned a result.
294
295Next, we get the in-memory address of the ``__anon_expr`` function by calling
296``getAddress()`` on the symbol. Recall that we compile top-level expressions
297into a self-contained LLVM function that takes no arguments and returns the
298computed double. Because the LLVM JIT compiler matches the native platform ABI,
299this means that you can just cast the result pointer to a function pointer of
300that type and call it directly. This means, there is no difference between JIT
301compiled code and native machine code that is statically linked into your
302application.
303
304Finally, since we don't support re-evaluation of top-level expressions, we
305remove the module from the JIT when we're done to free the associated memory.
306Recall, however, that the module we created a few lines earlier (via
307``InitializeModuleAndPassManager``) is still open and waiting for new code to be
308added.
Sean Silvad7fb3962012-12-05 00:26:32 +0000309
310With just these two changes, lets see how Kaleidoscope works now!
311
312::
313
314 ready> 4+5;
315 Read top-level expression:
316 define double @0() {
317 entry:
318 ret double 9.000000e+00
319 }
320
321 Evaluated to 9.000000
322
323Well this looks like it is basically working. The dump of the function
324shows the "no argument function that always returns double" that we
325synthesize for each top-level expression that is typed in. This
326demonstrates very basic functionality, but can we do more?
327
328::
329
330 ready> def testfunc(x y) x + y*2;
331 Read function definition:
332 define double @testfunc(double %x, double %y) {
333 entry:
334 %multmp = fmul double %y, 2.000000e+00
335 %addtmp = fadd double %multmp, %x
336 ret double %addtmp
337 }
338
339 ready> testfunc(4, 10);
340 Read top-level expression:
341 define double @1() {
342 entry:
343 %calltmp = call double @testfunc(double 4.000000e+00, double 1.000000e+01)
344 ret double %calltmp
345 }
346
347 Evaluated to 24.000000
348
Lang Hames2d789c32015-08-26 03:07:41 +0000349 ready> testfunc(5, 10);
350 ready> LLVM ERROR: Program used external function 'testfunc' which could not be resolved!
Sean Silvad7fb3962012-12-05 00:26:32 +0000351
Lang Hames2d789c32015-08-26 03:07:41 +0000352
353Function definitions and calls also work, but something went very wrong on that
354last line. The call looks valid, so what happened? As you may have guessed from
355the the API a Module is a unit of allocation for the JIT, and testfunc was part
356of the same module that contained anonymous expression. When we removed that
357module from the JIT to free the memory for the anonymous expression, we deleted
358the definition of ``testfunc`` along with it. Then, when we tried to call
359testfunc a second time, the JIT could no longer find it.
360
361The easiest way to fix this is to put the anonymous expression in a separate
362module from the rest of the function definitions. The JIT will happily resolve
363function calls across module boundaries, as long as each of the functions called
364has a prototype, and is added to the JIT before it is called. By putting the
365anonymous expression in a different module we can delete it without affecting
366the rest of the functions.
367
368In fact, we're going to go a step further and put every function in its own
369module. Doing so allows us to exploit a useful property of the KaleidoscopeJIT
370that will make our environment more REPL-like: Functions can be added to the
371JIT more than once (unlike a module where every function must have a unique
372definition). When you look up a symbol in KaleidoscopeJIT it will always return
373the most recent definition:
374
375::
376
377 ready> def foo(x) x + 1;
378 Read function definition:
379 define double @foo(double %x) {
380 entry:
381 %addtmp = fadd double %x, 1.000000e+00
382 ret double %addtmp
383 }
384
385 ready> foo(2);
386 Evaluated to 3.000000
387
388 ready> def foo(x) x + 2;
389 define double @foo(double %x) {
390 entry:
391 %addtmp = fadd double %x, 2.000000e+00
392 ret double %addtmp
393 }
394
395 ready> foo(2);
396 Evaluated to 4.000000
397
398
399To allow each function to live in its own module we'll need a way to
400re-generate previous function declarations into each new module we open:
401
402.. code-block:: c++
403
404 static std::unique_ptr<KaleidoscopeJIT> TheJIT;
405
406 ...
407
408 Function *getFunction(std::string Name) {
409 // First, see if the function has already been added to the current module.
410 if (auto *F = TheModule->getFunction(Name))
411 return F;
412
413 // If not, check whether we can codegen the declaration from some existing
414 // prototype.
415 auto FI = FunctionProtos.find(Name);
416 if (FI != FunctionProtos.end())
417 return FI->second->codegen();
418
419 // If no existing prototype exists, return null.
420 return nullptr;
421 }
422
423 ...
424
425 Value *CallExprAST::codegen() {
426 // Look up the name in the global module table.
427 Function *CalleeF = getFunction(Callee);
428
429 ...
430
431 Function *FunctionAST::codegen() {
432 // Transfer ownership of the prototype to the FunctionProtos map, but keep a
433 // reference to it for use below.
434 auto &P = *Proto;
435 FunctionProtos[Proto->getName()] = std::move(Proto);
436 Function *TheFunction = getFunction(P.getName());
437 if (!TheFunction)
438 return nullptr;
439
440
441To enable this, we'll start by adding a new global, ``FunctionProtos``, that
442holds the most recent prototype for each function. We'll also add a convenience
443method, ``getFunction()``, to replace calls to ``TheModule->getFunction()``.
444Our convenience method searches ``TheModule`` for an existing function
445declaration, falling back to generating a new declaration from FunctionProtos if
446it doesn't find one. In ``CallExprAST::codegen()`` we just need to replace the
447call to ``TheModule->getFunction()``. In ``FunctionAST::codegen()`` we need to
448update the FunctionProtos map first, then call ``getFunction()``. With this
449done, we can always obtain a function declaration in the current module for any
450previously declared function.
451
452We also need to update HandleDefinition and HandleExtern:
453
454.. code-block:: c++
455
456 static void HandleDefinition() {
457 if (auto FnAST = ParseDefinition()) {
458 if (auto *FnIR = FnAST->codegen()) {
459 fprintf(stderr, "Read function definition:");
460 FnIR->dump();
461 TheJIT->addModule(std::move(TheModule));
462 InitializeModuleAndPassManager();
463 }
464 } else {
465 // Skip token for error recovery.
466 getNextToken();
467 }
468 }
469
470 static void HandleExtern() {
471 if (auto ProtoAST = ParseExtern()) {
472 if (auto *FnIR = ProtoAST->codegen()) {
473 fprintf(stderr, "Read extern: ");
474 FnIR->dump();
475 FunctionProtos[ProtoAST->getName()] = std::move(ProtoAST);
476 }
477 } else {
478 // Skip token for error recovery.
479 getNextToken();
480 }
481 }
482
483In HandleDefinition, we add two lines to transfer the newly defined function to
484the JIT and open a new module. In HandleExtern, we just need to add one line to
485add the prototype to FunctionProtos.
486
487With these changes made, lets try our REPL again (I removed the dump of the
488anonymous functions this time, you should get the idea by now :) :
489
490::
491
492 ready> def foo(x) x + 1;
493 ready> foo(2);
494 Evaluated to 3.000000
495
496 ready> def foo(x) x + 2;
497 ready> foo(2);
498 Evaluated to 4.000000
499
500It works!
501
502Even with this simple code, we get some surprisingly powerful capabilities -
503check this out:
Sean Silvad7fb3962012-12-05 00:26:32 +0000504
505::
506
507 ready> extern sin(x);
508 Read extern:
509 declare double @sin(double)
510
511 ready> extern cos(x);
512 Read extern:
513 declare double @cos(double)
514
515 ready> sin(1.0);
516 Read top-level expression:
517 define double @2() {
518 entry:
519 ret double 0x3FEAED548F090CEE
520 }
521
522 Evaluated to 0.841471
523
524 ready> def foo(x) sin(x)*sin(x) + cos(x)*cos(x);
525 Read function definition:
526 define double @foo(double %x) {
527 entry:
528 %calltmp = call double @sin(double %x)
529 %multmp = fmul double %calltmp, %calltmp
530 %calltmp2 = call double @cos(double %x)
531 %multmp4 = fmul double %calltmp2, %calltmp2
532 %addtmp = fadd double %multmp, %multmp4
533 ret double %addtmp
534 }
535
536 ready> foo(4.0);
537 Read top-level expression:
538 define double @3() {
539 entry:
540 %calltmp = call double @foo(double 4.000000e+00)
541 ret double %calltmp
542 }
543
544 Evaluated to 1.000000
545
Lang Hames2d789c32015-08-26 03:07:41 +0000546Whoa, how does the JIT know about sin and cos? The answer is surprisingly
547simple: The KaleidoscopeJIT has a straightforward symbol resolution rule that
548it uses to find symbols that aren't available in any given module: First
549it searches all the modules that have already been added to the JIT, from the
550most recent to the oldest, to find the newest definition. If no definition is
551found inside the JIT, it falls back to calling "``dlsym("sin")``" on the
552Kaleidoscope process itself. Since "``sin``" is defined within the JIT's
553address space, it simply patches up calls in the module to call the libm
554version of ``sin`` directly.
Sean Silvad7fb3962012-12-05 00:26:32 +0000555
Lang Hames2d789c32015-08-26 03:07:41 +0000556In the future we'll see how tweaking this symbol resolution rule can be used to
557enable all sorts of useful features, from security (restricting the set of
558symbols available to JIT'd code), to dynamic code generation based on symbol
559names, and even lazy compilation.
Sean Silvad7fb3962012-12-05 00:26:32 +0000560
Lang Hames2d789c32015-08-26 03:07:41 +0000561One immediate benefit of the symbol resolution rule is that we can now extend
562the language by writing arbitrary C++ code to implement operations. For example,
563if we add:
Sean Silvad7fb3962012-12-05 00:26:32 +0000564
565.. code-block:: c++
566
567 /// putchard - putchar that takes a double and returns 0.
Lang Hames59b0da82015-08-19 18:15:58 +0000568 extern "C" double putchard(double X) {
Lang Hamesd76e0672015-08-27 20:31:44 +0000569 fputc((char)X, stderr);
Sean Silvad7fb3962012-12-05 00:26:32 +0000570 return 0;
571 }
572
573Now we can produce simple output to the console by using things like:
574"``extern putchard(x); putchard(120);``", which prints a lowercase 'x'
575on the console (120 is the ASCII code for 'x'). Similar code could be
576used to implement file I/O, console input, and many other capabilities
577in Kaleidoscope.
578
579This completes the JIT and optimizer chapter of the Kaleidoscope
580tutorial. At this point, we can compile a non-Turing-complete
581programming language, optimize and JIT compile it in a user-driven way.
582Next up we'll look into `extending the language with control flow
583constructs <LangImpl5.html>`_, tackling some interesting LLVM IR issues
584along the way.
585
586Full Code Listing
587=================
588
589Here is the complete code listing for our running example, enhanced with
590the LLVM JIT and optimizer. To build this example, use:
591
592.. code-block:: bash
593
594 # Compile
Eric Christophera8c6a0a2015-01-08 19:07:01 +0000595 clang++ -g toy.cpp `llvm-config --cxxflags --ldflags --system-libs --libs core mcjit native` -O3 -o toy
Sean Silvad7fb3962012-12-05 00:26:32 +0000596 # Run
597 ./toy
598
599If you are compiling this on Linux, make sure to add the "-rdynamic"
600option as well. This makes sure that the external functions are resolved
601properly at runtime.
602
603Here is the code:
604
Logan Chien855b17d2013-06-08 09:03:03 +0000605.. literalinclude:: ../../examples/Kaleidoscope/Chapter4/toy.cpp
606 :language: c++
Sean Silvad7fb3962012-12-05 00:26:32 +0000607
608`Next: Extending the language: control flow <LangImpl5.html>`_
609