docs/tutorial/OCamlLangImpl4.html

   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   <meta name="author" content="Erick Tryzelaar">
  10   <link rel="stylesheet" href="../llvm.css" type="text/css">
  11 </head>
  12
  13 <body>
  14
  15 <h1>Kaleidoscope: Adding JIT and Optimizer Support</h1>
  16
  17 <ul>
  18 <li><a href="index.html">Up to Tutorial Index</a></li>
  19 <li>Chapter 4
  20   <ol>
  21     <li><a href="#intro">Chapter 4 Introduction</a></li>
  22     <li><a href="#trivialconstfold">Trivial Constant Folding</a></li>
  23     <li><a href="#optimizerpasses">LLVM Optimization Passes</a></li>
  24     <li><a href="#jit">Adding a JIT Compiler</a></li>
  25     <li><a href="#code">Full Code Listing</a></li>
  26   </ol>
  27 </li>
  28 <li><a href="OCamlLangImpl5.html">Chapter 5</a>: Extending the Language: Control
  29 Flow</li>
  30 </ul>
  31
  32 <div class="doc_author">
  33         <p>
  34                 Written by <a href="mailto:sabre@nondot.org">Chris Lattner</a>
  35                 and <a href="mailto:idadesub@users.sourceforge.net">Erick Tryzelaar</a>
  36         </p>
  37 </div>
  38
  39 <!-- *********************************************************************** -->
  40 <h2><a name="intro">Chapter 4 Introduction</a></h2>
  41 <!-- *********************************************************************** -->
  42
  43 <div>
  44
  45 <p>Welcome to Chapter 4 of the "<a href="index.html">Implementing a language
  46 with LLVM</a>" tutorial.  Chapters 1-3 described the implementation of a simple
  47 language and added support for generating LLVM IR.  This chapter describes
  48 two new techniques: adding optimizer support to your language, and adding JIT
  49 compiler support.  These additions will demonstrate how to get nice, efficient code
  50 for the Kaleidoscope language.</p>
  51
  52 </div>
  53
  54 <!-- *********************************************************************** -->
  55 <h2><a name="trivialconstfold">Trivial Constant Folding</a></h2>
  56 <!-- *********************************************************************** -->
  57
  58 <div>
  59
  60 <p><b>Note:</b> the default <tt>IRBuilder</tt> now always includes the constant
  61 folding optimisations below.<p>
  62
  63 <p>
  64 Our demonstration for Chapter 3 is elegant and easy to extend.  Unfortunately,
  65 it does not produce wonderful code.  For example, when compiling simple code,
  66 we don't get obvious optimizations:</p>
  67
  68 <div class="doc_code">
  69 <pre>
  70 ready&gt; <b>def test(x) 1+2+x;</b>
  71 Read function definition:
  72 define double @test(double %x) {
  73 entry:
  74         %addtmp = fadd double 1.000000e+00, 2.000000e+00
  75         %addtmp1 = fadd double %addtmp, %x
  76         ret double %addtmp1
  77 }
  78 </pre>
  79 </div>
  80
  81 <p>This code is a very, very literal transcription of the AST built by parsing
  82 the input. As such, this transcription lacks optimizations like constant folding
  83 (we'd like to get "<tt>add x, 3.0</tt>" in the example above) as well as other
  84 more important optimizations.  Constant folding, in particular, is a very common
  85 and very important optimization: so much so that many language implementors
  86 implement constant folding support in their AST representation.</p>
  87
  88 <p>With LLVM, you don't need this support in the AST.  Since all calls to build
  89 LLVM IR go through the LLVM builder, it would be nice if the builder itself
  90 checked to see if there was a constant folding opportunity when you call it.
  91 If so, it could just do the constant fold and return the constant instead of
  92 creating an instruction.  This is exactly what the <tt>LLVMFoldingBuilder</tt>
  93 class does.
  94
  95 <p>All we did was switch from <tt>LLVMBuilder</tt> to
  96 <tt>LLVMFoldingBuilder</tt>.  Though we change no other code, we now have all of our
  97 instructions implicitly constant folded without us having to do anything
  98 about it.  For example, the input above now compiles to:</p>
  99
 100 <div class="doc_code">
 101 <pre>
 102 ready&gt; <b>def test(x) 1+2+x;</b>
 103 Read function definition:
 104 define double @test(double %x) {
 105 entry:
 106         %addtmp = fadd double 3.000000e+00, %x
 107         ret double %addtmp
 108 }
 109 </pre>
 110 </div>
 111
 112 <p>Well, that was easy :).  In practice, we recommend always using
 113 <tt>LLVMFoldingBuilder</tt> when generating code like this.  It has no
 114 "syntactic overhead" for its use (you don't have to uglify your compiler with
 115 constant checks everywhere) and it can dramatically reduce the amount of
 116 LLVM IR that is generated in some cases (particular for languages with a macro
 117 preprocessor or that use a lot of constants).</p>
 118
 119 <p>On the other hand, the <tt>LLVMFoldingBuilder</tt> is limited by the fact
 120 that it does all of its analysis inline with the code as it is built.  If you
 121 take a slightly more complex example:</p>
 122
 123 <div class="doc_code">
 124 <pre>
 125 ready&gt; <b>def test(x) (1+2+x)*(x+(1+2));</b>
 126 ready&gt; Read function definition:
 127 define double @test(double %x) {
 128 entry:
 129         %addtmp = fadd double 3.000000e+00, %x
 130         %addtmp1 = fadd double %x, 3.000000e+00
 131         %multmp = fmul double %addtmp, %addtmp1
 132         ret double %multmp
 133 }
 134 </pre>
 135 </div>
 136
 137 <p>In this case, the LHS and RHS of the multiplication are the same value.  We'd
 138 really like to see this generate "<tt>tmp = x+3; result = tmp*tmp;</tt>" instead
 139 of computing "<tt>x*3</tt>" twice.</p>
 140
 141 <p>Unfortunately, no amount of local analysis will be able to detect and correct
 142 this.  This requires two transformations: reassociation of expressions (to
 143 make the add's lexically identical) and Common Subexpression Elimination (CSE)
 144 to  delete the redundant add instruction.  Fortunately, LLVM provides a broad
 145 range of optimizations that you can use, in the form of "passes".</p>
 146
 147 </div>
 148
 149 <!-- *********************************************************************** -->
 150 <h2><a name="optimizerpasses">LLVM Optimization Passes</a></h2>
 151 <!-- *********************************************************************** -->
 152
 153 <div>
 154
 155 <p>LLVM provides many optimization passes, which do many different sorts of
 156 things and have different tradeoffs.  Unlike other systems, LLVM doesn't hold
 157 to the mistaken notion that one set of optimizations is right for all languages
 158 and for all situations.  LLVM allows a compiler implementor to make complete
 159 decisions about what optimizations to use, in which order, and in what
 160 situation.</p>
 161
 162 <p>As a concrete example, LLVM supports both "whole module" passes, which look
 163 across as large of body of code as they can (often a whole file, but if run
 164 at link time, this can be a substantial portion of the whole program).  It also
 165 supports and includes "per-function" passes which just operate on a single
 166 function at a time, without looking at other functions.  For more information
 167 on passes and how they are run, see the <a href="../WritingAnLLVMPass.html">How
 168 to Write a Pass</a> document and the <a href="../Passes.html">List of LLVM
 169 Passes</a>.</p>
 170
 171 <p>For Kaleidoscope, we are currently generating functions on the fly, one at
 172 a time, as the user types them in.  We aren't shooting for the ultimate
 173 optimization experience in this setting, but we also want to catch the easy and
 174 quick stuff where possible.  As such, we will choose to run a few per-function
 175 optimizations as the user types the function in.  If we wanted to make a "static
 176 Kaleidoscope compiler", we would use exactly the code we have now, except that
 177 we would defer running the optimizer until the entire file has been parsed.</p>
 178
 179 <p>In order to get per-function optimizations going, we need to set up a
 180 <a href="../WritingAnLLVMPass.html#passmanager">Llvm.PassManager</a> to hold and
 181 organize the LLVM optimizations that we want to run.  Once we have that, we can
 182 add a set of optimizations to run.  The code looks like this:</p>
 183
 184 <div class="doc_code">
 185 <pre>
 186   (* Create the JIT. *)
 187   let the_execution_engine = ExecutionEngine.create Codegen.the_module in
 188   let the_fpm = PassManager.create_function Codegen.the_module in
 189
 190   (* Set up the optimizer pipeline.  Start with registering info about how the
 191    * target lays out data structures. *)
 192   TargetData.add (ExecutionEngine.target_data the_execution_engine) the_fpm;
 193
 194   (* Do simple "peephole" optimizations and bit-twiddling optzn. *)
 195   add_instruction_combining the_fpm;
 196
 197   (* reassociate expressions. *)
 198   add_reassociation the_fpm;
 199
 200   (* Eliminate Common SubExpressions. *)
 201   add_gvn the_fpm;
 202
 203   (* Simplify the control flow graph (deleting unreachable blocks, etc). *)
 204   add_cfg_simplification the_fpm;
 205
 206   ignore (PassManager.initialize the_fpm);
 207
 208   (* Run the main "interpreter loop" now. *)
 209   Toplevel.main_loop the_fpm the_execution_engine stream;
 210 </pre>
 211 </div>
 212
 213 <p>The meat of the matter here, is the definition of "<tt>the_fpm</tt>".  It
 214 requires a pointer to the <tt>the_module</tt> to construct itself.  Once it is
 215 set up, we use a series of "add" calls to add a bunch of LLVM passes.  The
 216 first pass is basically boilerplate, it adds a pass so that later optimizations
 217 know how the data structures in the program are laid out.  The
 218 "<tt>the_execution_engine</tt>" variable is related to the JIT, which we will
 219 get to in the next section.</p>
 220
 221 <p>In this case, we choose to add 4 optimization passes.  The passes we chose
 222 here are a pretty standard set of "cleanup" optimizations that are useful for
 223 a wide variety of code.  I won't delve into what they do but, believe me,
 224 they are a good starting place :).</p>
 225
 226 <p>Once the <tt>Llvm.PassManager.</tt> is set up, we need to make use of it.
 227 We do this by running it after our newly created function is constructed (in
 228 <tt>Codegen.codegen_func</tt>), but before it is returned to the client:</p>
 229
 230 <div class="doc_code">
 231 <pre>
 232 let codegen_func the_fpm = function
 233       ...
 234       try
 235         let ret_val = codegen_expr body in
 236
 237         (* Finish off the function. *)
 238         let _ = build_ret ret_val builder in
 239
 240         (* Validate the generated code, checking for consistency. *)
 241         Llvm_analysis.assert_valid_function the_function;
 242
 243         (* Optimize the function. *)
 244         let _ = PassManager.run_function the_function the_fpm in
 245
 246         the_function
 247 </pre>
 248 </div>
 249
 250 <p>As you can see, this is pretty straightforward.  The <tt>the_fpm</tt>
 251 optimizes and updates the LLVM Function* in place, improving (hopefully) its
 252 body.  With this in place, we can try our test above again:</p>
 253
 254 <div class="doc_code">
 255 <pre>
 256 ready&gt; <b>def test(x) (1+2+x)*(x+(1+2));</b>
 257 ready&gt; Read function definition:
 258 define double @test(double %x) {
 259 entry:
 260         %addtmp = fadd double %x, 3.000000e+00
 261         %multmp = fmul double %addtmp, %addtmp
 262         ret double %multmp
 263 }
 264 </pre>
 265 </div>
 266
 267 <p>As expected, we now get our nicely optimized code, saving a floating point
 268 add instruction from every execution of this function.</p>
 269
 270 <p>LLVM provides a wide variety of optimizations that can be used in certain
 271 circumstances.  Some <a href="../Passes.html">documentation about the various
 272 passes</a> is available, but it isn't very complete.  Another good source of
 273 ideas can come from looking at the passes that <tt>llvm-gcc</tt> or
 274 <tt>llvm-ld</tt> run to get started.  The "<tt>opt</tt>" tool allows you to
 275 experiment with passes from the command line, so you can see if they do
 276 anything.</p>
 277
 278 <p>Now that we have reasonable code coming out of our front-end, lets talk about
 279 executing it!</p>
 280
 281 </div>
 282
 283 <!-- *********************************************************************** -->
 284 <h2><a name="jit">Adding a JIT Compiler</a></h2>
 285 <!-- *********************************************************************** -->
 286
 287 <div>
 288
 289 <p>Code that is available in LLVM IR can have a wide variety of tools
 290 applied to it.  For example, you can run optimizations on it (as we did above),
 291 you can dump it out in textual or binary forms, you can compile the code to an
 292 assembly file (.s) for some target, or you can JIT compile it.  The nice thing
 293 about the LLVM IR representation is that it is the "common currency" between
 294 many different parts of the compiler.
 295 </p>
 296
 297 <p>In this section, we'll add JIT compiler support to our interpreter.  The
 298 basic idea that we want for Kaleidoscope is to have the user enter function
 299 bodies as they do now, but immediately evaluate the top-level expressions they
 300 type in.  For example, if they type in "1 + 2;", we should evaluate and print
 301 out 3.  If they define a function, they should be able to call it from the
 302 command line.</p>
 303
 304 <p>In order to do this, we first declare and initialize the JIT.  This is done
 305 by adding a global variable and a call in <tt>main</tt>:</p>
 306
 307 <div class="doc_code">
 308 <pre>
 309 ...
 310 let main () =
 311   ...
 312   <b>(* Create the JIT. *)
 313   let the_execution_engine = ExecutionEngine.create Codegen.the_module in</b>
 314   ...
 315 </pre>
 316 </div>
 317
 318 <p>This creates an abstract "Execution Engine" which can be either a JIT
 319 compiler or the LLVM interpreter.  LLVM will automatically pick a JIT compiler
 320 for you if one is available for your platform, otherwise it will fall back to
 321 the interpreter.</p>
 322
 323 <p>Once the <tt>Llvm_executionengine.ExecutionEngine.t</tt> is created, the JIT
 324 is ready to be used.  There are a variety of APIs that are useful, but the
 325 simplest one is the "<tt>Llvm_executionengine.ExecutionEngine.run_function</tt>"
 326 function.  This method JIT compiles the specified LLVM Function and returns a
 327 function pointer to the generated machine code.  In our case, this means that we
 328 can change the code that parses a top-level expression to look like this:</p>
 329
 330 <div class="doc_code">
 331 <pre>
 332             (* Evaluate a top-level expression into an anonymous function. *)
 333             let e = Parser.parse_toplevel stream in
 334             print_endline "parsed a top-level expr";
 335             let the_function = Codegen.codegen_func the_fpm e in
 336             dump_value the_function;
 337
 338             (* JIT the function, returning a function pointer. *)
 339             let result = ExecutionEngine.run_function the_function [||]
 340               the_execution_engine in
 341
 342             print_string "Evaluated to ";
 343             print_float (GenericValue.as_float Codegen.double_type result);
 344             print_newline ();
 345 </pre>
 346 </div>
 347
 348 <p>Recall that we compile top-level expressions into a self-contained LLVM
 349 function that takes no arguments and returns the computed double.  Because the
 350 LLVM JIT compiler matches the native platform ABI, this means that you can just
 351 cast the result pointer to a function pointer of that type and call it directly.
 352 This means, there is no difference between JIT compiled code and native machine
 353 code that is statically linked into your application.</p>
 354
 355 <p>With just these two changes, lets see how Kaleidoscope works now!</p>
 356
 357 <div class="doc_code">
 358 <pre>
 359 ready&gt; <b>4+5;</b>
 360 define double @""() {
 361 entry:
 362         ret double 9.000000e+00
 363 }
 364
 365 <em>Evaluated to 9.000000</em>
 366 </pre>
 367 </div>
 368
 369 <p>Well this looks like it is basically working.  The dump of the function
 370 shows the "no argument function that always returns double" that we synthesize
 371 for each top level expression that is typed in.  This demonstrates very basic
 372 functionality, but can we do more?</p>
 373
 374 <div class="doc_code">
 375 <pre>
 376 ready&gt; <b>def testfunc(x y) x + y*2; </b>
 377 Read function definition:
 378 define double @testfunc(double %x, double %y) {
 379 entry:
 380         %multmp = fmul double %y, 2.000000e+00
 381         %addtmp = fadd double %multmp, %x
 382         ret double %addtmp
 383 }
 384
 385 ready&gt; <b>testfunc(4, 10);</b>
 386 define double @""() {
 387 entry:
 388         %calltmp = call double @testfunc(double 4.000000e+00, double 1.000000e+01)
 389         ret double %calltmp
 390 }
 391
 392 <em>Evaluated to 24.000000</em>
 393 </pre>
 394 </div>
 395
 396 <p>This illustrates that we can now call user code, but there is something a bit
 397 subtle going on here.  Note that we only invoke the JIT on the anonymous
 398 functions that <em>call testfunc</em>, but we never invoked it
 399 on <em>testfunc</em> itself.  What actually happened here is that the JIT
 400 scanned for all non-JIT'd functions transitively called from the anonymous
 401 function and compiled all of them before returning
 402 from <tt>run_function</tt>.</p>
 403
 404 <p>The JIT provides a number of other more advanced interfaces for things like
 405 freeing allocated machine code, rejit'ing functions to update them, etc.
 406 However, even with this simple code, we get some surprisingly powerful
 407 capabilities - check this out (I removed the dump of the anonymous functions,
 408 you should get the idea by now :) :</p>
 409
 410 <div class="doc_code">
 411 <pre>
 412 ready&gt; <b>extern sin(x);</b>
 413 Read extern:
 414 declare double @sin(double)
 415
 416 ready&gt; <b>extern cos(x);</b>
 417 Read extern:
 418 declare double @cos(double)
 419
 420 ready&gt; <b>sin(1.0);</b>
 421 <em>Evaluated to 0.841471</em>
 422
 423 ready&gt; <b>def foo(x) sin(x)*sin(x) + cos(x)*cos(x);</b>
 424 Read function definition:
 425 define double @foo(double %x) {
 426 entry:
 427         %calltmp = call double @sin(double %x)
 428         %multmp = fmul double %calltmp, %calltmp
 429         %calltmp2 = call double @cos(double %x)
 430         %multmp4 = fmul double %calltmp2, %calltmp2
 431         %addtmp = fadd double %multmp, %multmp4
 432         ret double %addtmp
 433 }
 434
 435 ready&gt; <b>foo(4.0);</b>
 436 <em>Evaluated to 1.000000</em>
 437 </pre>
 438 </div>
 439
 440 <p>Whoa, how does the JIT know about sin and cos?  The answer is surprisingly
 441 simple: in this example, the JIT started execution of a function and got to a
 442 function call.  It realized that the function was not yet JIT compiled and
 443 invoked the standard set of routines to resolve the function.  In this case,
 444 there is no body defined for the function, so the JIT ended up calling
 445 "<tt>dlsym("sin")</tt>" on the Kaleidoscope process itself.  Since
 446 "<tt>sin</tt>" is defined within the JIT's address space, it simply patches up
 447 calls in the module to call the libm version of <tt>sin</tt> directly.</p>
 448
 449 <p>The LLVM JIT provides a number of interfaces (look in the
 450 <tt>llvm_executionengine.mli</tt> file) for controlling how unknown functions
 451 get resolved.  It allows you to establish explicit mappings between IR objects
 452 and addresses (useful for LLVM global variables that you want to map to static
 453 tables, for example), allows you to dynamically decide on the fly based on the
 454 function name, and even allows you to have the JIT compile functions lazily the
 455 first time they're called.</p>
 456
 457 <p>One interesting application of this is that we can now extend the language
 458 by writing arbitrary C code to implement operations.  For example, if we add:
 459 </p>
 460
 461 <div class="doc_code">
 462 <pre>
 463 /* putchard - putchar that takes a double and returns 0. */
 464 extern "C"
 465 double putchard(double X) {
 466   putchar((char)X);
 467   return 0;
 468 }
 469 </pre>
 470 </div>
 471
 472 <p>Now we can produce simple output to the console by using things like:
 473 "<tt>extern putchard(x); putchard(120);</tt>", which prints a lowercase 'x' on
 474 the console (120 is the ASCII code for 'x').  Similar code could be used to
 475 implement file I/O, console input, and many other capabilities in
 476 Kaleidoscope.</p>
 477
 478 <p>This completes the JIT and optimizer chapter of the Kaleidoscope tutorial. At
 479 this point, we can compile a non-Turing-complete programming language, optimize
 480 and JIT compile it in a user-driven way.  Next up we'll look into <a
 481 href="OCamlLangImpl5.html">extending the language with control flow
 482 constructs</a>, tackling some interesting LLVM IR issues along the way.</p>
 483
 484 </div>
 485
 486 <!-- *********************************************************************** -->
 487 <h2><a name="code">Full Code Listing</a></h2>
 488 <!-- *********************************************************************** -->
 489
 490 <div>
 491
 492 <p>
 493 Here is the complete code listing for our running example, enhanced with the
 494 LLVM JIT and optimizer.  To build this example, use:
 495 </p>
 496
 497 <div class="doc_code">
 498 <pre>
 499 # Compile
 500 ocamlbuild toy.byte
 501 # Run
 502 ./toy.byte
 503 </pre>
 504 </div>
 505
 506 <p>Here is the code:</p>
 507
 508 <dl>
 509 <dt>_tags:</dt>
 510 <dd class="doc_code">
 511 <pre>
 512 &lt;{lexer,parser}.ml&gt;: use_camlp4, pp(camlp4of)
 513 &lt;*.{byte,native}&gt;: g++, use_llvm, use_llvm_analysis
 514 &lt;*.{byte,native}&gt;: use_llvm_executionengine, use_llvm_target
 515 &lt;*.{byte,native}&gt;: use_llvm_scalar_opts, use_bindings
 516 </pre>
 517 </dd>
 518
 519 <dt>myocamlbuild.ml:</dt>
 520 <dd class="doc_code">
 521 <pre>
 522 open Ocamlbuild_plugin;;
 523
 524 ocaml_lib ~extern:true "llvm";;
 525 ocaml_lib ~extern:true "llvm_analysis";;
 526 ocaml_lib ~extern:true "llvm_executionengine";;
 527 ocaml_lib ~extern:true "llvm_target";;
 528 ocaml_lib ~extern:true "llvm_scalar_opts";;
 529
 530 flag ["link"; "ocaml"; "g++"] (S[A"-cc"; A"g++"]);;
 531 dep ["link"; "ocaml"; "use_bindings"] ["bindings.o"];;
 532 </pre>
 533 </dd>
 534
 535 <dt>token.ml:</dt>
 536 <dd class="doc_code">
 537 <pre>
 538 (*===----------------------------------------------------------------------===
 539  * Lexer Tokens
 540  *===----------------------------------------------------------------------===*)
 541
 542 (* The lexer returns these 'Kwd' if it is an unknown character, otherwise one of
 543  * these others for known things. *)
 544 type token =
 545   (* commands *)
 546   | Def | Extern
 547
 548   (* primary *)
 549   | Ident of string | Number of float
 550
 551   (* unknown *)
 552   | Kwd of char
 553 </pre>
 554 </dd>
 555
 556 <dt>lexer.ml:</dt>
 557 <dd class="doc_code">
 558 <pre>
 559 (*===----------------------------------------------------------------------===
 560  * Lexer
 561  *===----------------------------------------------------------------------===*)
 562
 563 let rec lex = parser
 564   (* Skip any whitespace. *)
 565   | [&lt; ' (' ' | '\n' | '\r' | '\t'); stream &gt;] -&gt; lex stream
 566
 567   (* identifier: [a-zA-Z][a-zA-Z0-9] *)
 568   | [&lt; ' ('A' .. 'Z' | 'a' .. 'z' as c); stream &gt;] -&gt;
 569       let buffer = Buffer.create 1 in
 570       Buffer.add_char buffer c;
 571       lex_ident buffer stream
 572
 573   (* number: [0-9.]+ *)
 574   | [&lt; ' ('0' .. '9' as c); stream &gt;] -&gt;
 575       let buffer = Buffer.create 1 in
 576       Buffer.add_char buffer c;
 577       lex_number buffer stream
 578
 579   (* Comment until end of line. *)
 580   | [&lt; ' ('#'); stream &gt;] -&gt;
 581       lex_comment stream
 582
 583   (* Otherwise, just return the character as its ascii value. *)
 584   | [&lt; 'c; stream &gt;] -&gt;
 585       [&lt; 'Token.Kwd c; lex stream &gt;]
 586
 587   (* end of stream. *)
 588   | [&lt; &gt;] -&gt; [&lt; &gt;]
 589
 590 and lex_number buffer = parser
 591   | [&lt; ' ('0' .. '9' | '.' as c); stream &gt;] -&gt;
 592       Buffer.add_char buffer c;
 593       lex_number buffer stream
 594   | [&lt; stream=lex &gt;] -&gt;
 595       [&lt; 'Token.Number (float_of_string (Buffer.contents buffer)); stream &gt;]
 596
 597 and lex_ident buffer = parser
 598   | [&lt; ' ('A' .. 'Z' | 'a' .. 'z' | '0' .. '9' as c); stream &gt;] -&gt;
 599       Buffer.add_char buffer c;
 600       lex_ident buffer stream
 601   | [&lt; stream=lex &gt;] -&gt;
 602       match Buffer.contents buffer with
 603       | "def" -&gt; [&lt; 'Token.Def; stream &gt;]
 604       | "extern" -&gt; [&lt; 'Token.Extern; stream &gt;]
 605       | id -&gt; [&lt; 'Token.Ident id; stream &gt;]
 606
 607 and lex_comment = parser
 608   | [&lt; ' ('\n'); stream=lex &gt;] -&gt; stream
 609   | [&lt; 'c; e=lex_comment &gt;] -&gt; e
 610   | [&lt; &gt;] -&gt; [&lt; &gt;]
 611 </pre>
 612 </dd>
 613
 614 <dt>ast.ml:</dt>
 615 <dd class="doc_code">
 616 <pre>
 617 (*===----------------------------------------------------------------------===
 618  * Abstract Syntax Tree (aka Parse Tree)
 619  *===----------------------------------------------------------------------===*)
 620
 621 (* expr - Base type for all expression nodes. *)
 622 type expr =
 623   (* variant for numeric literals like "1.0". *)
 624   | Number of float
 625
 626   (* variant for referencing a variable, like "a". *)
 627   | Variable of string
 628
 629   (* variant for a binary operator. *)
 630   | Binary of char * expr * expr
 631
 632   (* variant for function calls. *)
 633   | Call of string * expr array
 634
 635 (* proto - This type represents the "prototype" for a function, which captures
 636  * its name, and its argument names (thus implicitly the number of arguments the
 637  * function takes). *)
 638 type proto = Prototype of string * string array
 639
 640 (* func - This type represents a function definition itself. *)
 641 type func = Function of proto * expr
 642 </pre>
 643 </dd>
 644
 645 <dt>parser.ml:</dt>
 646 <dd class="doc_code">
 647 <pre>
 648 (*===---------------------------------------------------------------------===
 649  * Parser
 650  *===---------------------------------------------------------------------===*)
 651
 652 (* binop_precedence - This holds the precedence for each binary operator that is
 653  * defined *)
 654 let binop_precedence:(char, int) Hashtbl.t = Hashtbl.create 10
 655
 656 (* precedence - Get the precedence of the pending binary operator token. *)
 657 let precedence c = try Hashtbl.find binop_precedence c with Not_found -&gt; -1
 658
 659 (* primary
 660  *   ::= identifier
 661  *   ::= numberexpr
 662  *   ::= parenexpr *)
 663 let rec parse_primary = parser
 664   (* numberexpr ::= number *)
 665   | [&lt; 'Token.Number n &gt;] -&gt; Ast.Number n
 666
 667   (* parenexpr ::= '(' expression ')' *)
 668   | [&lt; 'Token.Kwd '('; e=parse_expr; 'Token.Kwd ')' ?? "expected ')'" &gt;] -&gt; e
 669
 670   (* identifierexpr
 671    *   ::= identifier
 672    *   ::= identifier '(' argumentexpr ')' *)
 673   | [&lt; 'Token.Ident id; stream &gt;] -&gt;
 674       let rec parse_args accumulator = parser
 675         | [&lt; e=parse_expr; stream &gt;] -&gt;
 676             begin parser
 677               | [&lt; 'Token.Kwd ','; e=parse_args (e :: accumulator) &gt;] -&gt; e
 678               | [&lt; &gt;] -&gt; e :: accumulator
 679             end stream
 680         | [&lt; &gt;] -&gt; accumulator
 681       in
 682       let rec parse_ident id = parser
 683         (* Call. *)
 684         | [&lt; 'Token.Kwd '(';
 685              args=parse_args [];
 686              'Token.Kwd ')' ?? "expected ')'"&gt;] -&gt;
 687             Ast.Call (id, Array.of_list (List.rev args))
 688
 689         (* Simple variable ref. *)
 690         | [&lt; &gt;] -&gt; Ast.Variable id
 691       in
 692       parse_ident id stream
 693
 694   | [&lt; &gt;] -&gt; raise (Stream.Error "unknown token when expecting an expression.")
 695
 696 (* binoprhs
 697  *   ::= ('+' primary)* *)
 698 and parse_bin_rhs expr_prec lhs stream =
 699   match Stream.peek stream with
 700   (* If this is a binop, find its precedence. *)
 701   | Some (Token.Kwd c) when Hashtbl.mem binop_precedence c -&gt;
 702       let token_prec = precedence c in
 703
 704       (* If this is a binop that binds at least as tightly as the current binop,
 705        * consume it, otherwise we are done. *)
 706       if token_prec &lt; expr_prec then lhs else begin
 707         (* Eat the binop. *)
 708         Stream.junk stream;
 709
 710         (* Parse the primary expression after the binary operator. *)
 711         let rhs = parse_primary stream in
 712
 713         (* Okay, we know this is a binop. *)
 714         let rhs =
 715           match Stream.peek stream with
 716           | Some (Token.Kwd c2) -&gt;
 717               (* If BinOp binds less tightly with rhs than the operator after
 718                * rhs, let the pending operator take rhs as its lhs. *)
 719               let next_prec = precedence c2 in
 720               if token_prec &lt; next_prec
 721               then parse_bin_rhs (token_prec + 1) rhs stream
 722               else rhs
 723           | _ -&gt; rhs
 724         in
 725
 726         (* Merge lhs/rhs. *)
 727         let lhs = Ast.Binary (c, lhs, rhs) in
 728         parse_bin_rhs expr_prec lhs stream
 729       end
 730   | _ -&gt; lhs
 731
 732 (* expression
 733  *   ::= primary binoprhs *)
 734 and parse_expr = parser
 735   | [&lt; lhs=parse_primary; stream &gt;] -&gt; parse_bin_rhs 0 lhs stream
 736
 737 (* prototype
 738  *   ::= id '(' id* ')' *)
 739 let parse_prototype =
 740   let rec parse_args accumulator = parser
 741     | [&lt; 'Token.Ident id; e=parse_args (id::accumulator) &gt;] -&gt; e
 742     | [&lt; &gt;] -&gt; accumulator
 743   in
 744
 745   parser
 746   | [&lt; 'Token.Ident id;
 747        'Token.Kwd '(' ?? "expected '(' in prototype";
 748        args=parse_args [];
 749        'Token.Kwd ')' ?? "expected ')' in prototype" &gt;] -&gt;
 750       (* success. *)
 751       Ast.Prototype (id, Array.of_list (List.rev args))
 752
 753   | [&lt; &gt;] -&gt;
 754       raise (Stream.Error "expected function name in prototype")
 755
 756 (* definition ::= 'def' prototype expression *)
 757 let parse_definition = parser
 758   | [&lt; 'Token.Def; p=parse_prototype; e=parse_expr &gt;] -&gt;
 759       Ast.Function (p, e)
 760
 761 (* toplevelexpr ::= expression *)
 762 let parse_toplevel = parser
 763   | [&lt; e=parse_expr &gt;] -&gt;
 764       (* Make an anonymous proto. *)
 765       Ast.Function (Ast.Prototype ("", [||]), e)
 766
 767 (*  external ::= 'extern' prototype *)
 768 let parse_extern = parser
 769   | [&lt; 'Token.Extern; e=parse_prototype &gt;] -&gt; e
 770 </pre>
 771 </dd>
 772
 773 <dt>codegen.ml:</dt>
 774 <dd class="doc_code">
 775 <pre>
 776 (*===----------------------------------------------------------------------===
 777  * Code Generation
 778  *===----------------------------------------------------------------------===*)
 779
 780 open Llvm
 781
 782 exception Error of string
 783
 784 let context = global_context ()
 785 let the_module = create_module context "my cool jit"
 786 let builder = builder context
 787 let named_values:(string, llvalue) Hashtbl.t = Hashtbl.create 10
 788 let double_type = double_type context
 789
 790 let rec codegen_expr = function
 791   | Ast.Number n -&gt; const_float double_type n
 792   | Ast.Variable name -&gt;
 793       (try Hashtbl.find named_values name with
 794         | Not_found -&gt; raise (Error "unknown variable name"))
 795   | Ast.Binary (op, lhs, rhs) -&gt;
 796       let lhs_val = codegen_expr lhs in
 797       let rhs_val = codegen_expr rhs in
 798       begin
 799         match op with
 800         | '+' -&gt; build_add lhs_val rhs_val "addtmp" builder
 801         | '-' -&gt; build_sub lhs_val rhs_val "subtmp" builder
 802         | '*' -&gt; build_mul lhs_val rhs_val "multmp" builder
 803         | '&lt;' -&gt;
 804             (* Convert bool 0/1 to double 0.0 or 1.0 *)
 805             let i = build_fcmp Fcmp.Ult lhs_val rhs_val "cmptmp" builder in
 806             build_uitofp i double_type "booltmp" builder
 807         | _ -&gt; raise (Error "invalid binary operator")
 808       end
 809   | Ast.Call (callee, args) -&gt;
 810       (* Look up the name in the module table. *)
 811       let callee =
 812         match lookup_function callee the_module with
 813         | Some callee -&gt; callee
 814         | None -&gt; raise (Error "unknown function referenced")
 815       in
 816       let params = params callee in
 817
 818       (* If argument mismatch error. *)
 819       if Array.length params == Array.length args then () else
 820         raise (Error "incorrect # arguments passed");
 821       let args = Array.map codegen_expr args in
 822       build_call callee args "calltmp" builder
 823
 824 let codegen_proto = function
 825   | Ast.Prototype (name, args) -&gt;
 826       (* Make the function type: double(double,double) etc. *)
 827       let doubles = Array.make (Array.length args) double_type in
 828       let ft = function_type double_type doubles in
 829       let f =
 830         match lookup_function name the_module with
 831         | None -&gt; declare_function name ft the_module
 832
 833         (* If 'f' conflicted, there was already something named 'name'. If it
 834          * has a body, don't allow redefinition or reextern. *)
 835         | Some f -&gt;
 836             (* If 'f' already has a body, reject this. *)
 837             if block_begin f &lt;&gt; At_end f then
 838               raise (Error "redefinition of function");
 839
 840             (* If 'f' took a different number of arguments, reject. *)
 841             if element_type (type_of f) &lt;&gt; ft then
 842               raise (Error "redefinition of function with different # args");
 843             f
 844       in
 845
 846       (* Set names for all arguments. *)
 847       Array.iteri (fun i a -&gt;
 848         let n = args.(i) in
 849         set_value_name n a;
 850         Hashtbl.add named_values n a;
 851       ) (params f);
 852       f
 853
 854 let codegen_func the_fpm = function
 855   | Ast.Function (proto, body) -&gt;
 856       Hashtbl.clear named_values;
 857       let the_function = codegen_proto proto in
 858
 859       (* Create a new basic block to start insertion into. *)
 860       let bb = append_block context "entry" the_function in
 861       position_at_end bb builder;
 862
 863       try
 864         let ret_val = codegen_expr body in
 865
 866         (* Finish off the function. *)
 867         let _ = build_ret ret_val builder in
 868
 869         (* Validate the generated code, checking for consistency. *)
 870         Llvm_analysis.assert_valid_function the_function;
 871
 872         (* Optimize the function. *)
 873         let _ = PassManager.run_function the_function the_fpm in
 874
 875         the_function
 876       with e -&gt;
 877         delete_function the_function;
 878         raise e
 879 </pre>
 880 </dd>
 881
 882 <dt>toplevel.ml:</dt>
 883 <dd class="doc_code">
 884 <pre>
 885 (*===----------------------------------------------------------------------===
 886  * Top-Level parsing and JIT Driver
 887  *===----------------------------------------------------------------------===*)
 888
 889 open Llvm
 890 open Llvm_executionengine
 891
 892 (* top ::= definition | external | expression | ';' *)
 893 let rec main_loop the_fpm the_execution_engine stream =
 894   match Stream.peek stream with
 895   | None -&gt; ()
 896
 897   (* ignore top-level semicolons. *)
 898   | Some (Token.Kwd ';') -&gt;
 899       Stream.junk stream;
 900       main_loop the_fpm the_execution_engine stream
 901
 902   | Some token -&gt;
 903       begin
 904         try match token with
 905         | Token.Def -&gt;
 906             let e = Parser.parse_definition stream in
 907             print_endline "parsed a function definition.";
 908             dump_value (Codegen.codegen_func the_fpm e);
 909         | Token.Extern -&gt;
 910             let e = Parser.parse_extern stream in
 911             print_endline "parsed an extern.";
 912             dump_value (Codegen.codegen_proto e);
 913         | _ -&gt;
 914             (* Evaluate a top-level expression into an anonymous function. *)
 915             let e = Parser.parse_toplevel stream in
 916             print_endline "parsed a top-level expr";
 917             let the_function = Codegen.codegen_func the_fpm e in
 918             dump_value the_function;
 919
 920             (* JIT the function, returning a function pointer. *)
 921             let result = ExecutionEngine.run_function the_function [||]
 922               the_execution_engine in
 923
 924             print_string "Evaluated to ";
 925             print_float (GenericValue.as_float Codegen.double_type result);
 926             print_newline ();
 927         with Stream.Error s | Codegen.Error s -&gt;
 928           (* Skip token for error recovery. *)
 929           Stream.junk stream;
 930           print_endline s;
 931       end;
 932       print_string "ready&gt; "; flush stdout;
 933       main_loop the_fpm the_execution_engine stream
 934 </pre>
 935 </dd>
 936
 937 <dt>toy.ml:</dt>
 938 <dd class="doc_code">
 939 <pre>
 940 (*===----------------------------------------------------------------------===
 941  * Main driver code.
 942  *===----------------------------------------------------------------------===*)
 943
 944 open Llvm
 945 open Llvm_executionengine
 946 open Llvm_target
 947 open Llvm_scalar_opts
 948
 949 let main () =
 950   ignore (initialize_native_target ());
 951
 952   (* Install standard binary operators.
 953    * 1 is the lowest precedence. *)
 954   Hashtbl.add Parser.binop_precedence '&lt;' 10;
 955   Hashtbl.add Parser.binop_precedence '+' 20;
 956   Hashtbl.add Parser.binop_precedence '-' 20;
 957   Hashtbl.add Parser.binop_precedence '*' 40;    (* highest. *)
 958
 959   (* Prime the first token. *)
 960   print_string "ready&gt; "; flush stdout;
 961   let stream = Lexer.lex (Stream.of_channel stdin) in
 962
 963   (* Create the JIT. *)
 964   let the_execution_engine = ExecutionEngine.create Codegen.the_module in
 965   let the_fpm = PassManager.create_function Codegen.the_module in
 966
 967   (* Set up the optimizer pipeline.  Start with registering info about how the
 968    * target lays out data structures. *)
 969   TargetData.add (ExecutionEngine.target_data the_execution_engine) the_fpm;
 970
 971   (* Do simple "peephole" optimizations and bit-twiddling optzn. *)
 972   add_instruction_combination the_fpm;
 973
 974   (* reassociate expressions. *)
 975   add_reassociation the_fpm;
 976
 977   (* Eliminate Common SubExpressions. *)
 978   add_gvn the_fpm;
 979
 980   (* Simplify the control flow graph (deleting unreachable blocks, etc). *)
 981   add_cfg_simplification the_fpm;
 982
 983   ignore (PassManager.initialize the_fpm);
 984
 985   (* Run the main "interpreter loop" now. *)
 986   Toplevel.main_loop the_fpm the_execution_engine stream;
 987
 988   (* Print out all the generated code. *)
 989   dump_module Codegen.the_module
 990 ;;
 991
 992 main ()
 993 </pre>
 994 </dd>
 995
 996 <dt>bindings.c</dt>
 997 <dd class="doc_code">
 998 <pre>
 999 #include &lt;stdio.h&gt;
1000
1001 /* putchard - putchar that takes a double and returns 0. */
1002 extern double putchard(double X) {
1003   putchar((char)X);
1004   return 0;
1005 }
1006 </pre>
1007 </dd>
1008 </dl>
1009
1010 <a href="OCamlLangImpl5.html">Next: Extending the language: control flow</a>
1011 </div>
1012
1013 <!-- *********************************************************************** -->
1014 <hr>
1015 <address>
1016   <a href="http://jigsaw.w3.org/css-validator/check/referer"><img
1017   src="http://jigsaw.w3.org/css-validator/images/vcss" alt="Valid CSS!"></a>
1018   <a href="http://validator.w3.org/check/referer"><img
1019   src="http://www.w3.org/Icons/valid-html401" alt="Valid HTML 4.01!"></a>
1020
1021   <a href="mailto:sabre@nondot.org">Chris Lattner</a><br>
1022   <a href="mailto:idadesub@users.sourceforge.net">Erick Tryzelaar</a><br>
1023   <a href="http://llvm.org/">The LLVM Compiler Infrastructure</a><br>
1024   Last modified: $Date$
1025 </address>
1026 </body>
1027 </html>