| Chris Lattner | c9d5d2c | 2007-11-01 06:49:54 +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> | 
| Chris Lattner | 58f2c87 | 2007-11-02 05:42:52 +0000 | [diff] [blame] | 6 |   <title>Kaleidoscope: Extending the Language: User-defined Operators</title> | 
| Chris Lattner | c9d5d2c | 2007-11-01 06:49:54 +0000 | [diff] [blame] | 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 |  | 
| Chris Lattner | 58f2c87 | 2007-11-02 05:42:52 +0000 | [diff] [blame] | 14 | <div class="doc_title">Kaleidoscope: Extending the Language: User-defined Operators</div> | 
| Chris Lattner | c9d5d2c | 2007-11-01 06:49:54 +0000 | [diff] [blame] | 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 6 Introduction</a></div> | 
 | 22 | <!-- *********************************************************************** --> | 
 | 23 |  | 
 | 24 | <div class="doc_text"> | 
 | 25 |  | 
 | 26 | <p>Welcome to Part 6 of the "<a href="index.html">Implementing a language with | 
 | 27 | LLVM</a>" tutorial.  At this point in our tutorial, we now have a fully | 
 | 28 | functional language that is fairly minimal, but also useful.  One big problem | 
 | 29 | with it though is that it doesn't have many useful operators (like division, | 
 | 30 | logical negation, or even any comparisons other than less-than.</p> | 
 | 31 |  | 
| Chris Lattner | 58f2c87 | 2007-11-02 05:42:52 +0000 | [diff] [blame] | 32 | <p>This chapter of the tutorial takes a wild digression into adding user-defined | 
 | 33 | operators to the simple and beautiful Kaleidoscope language, giving us a  | 
| Chris Lattner | c9d5d2c | 2007-11-01 06:49:54 +0000 | [diff] [blame] | 34 | simple and ugly language in some ways, but also a powerful one at the same time. | 
 | 35 | One of the great things about creating your own language is that you get to | 
 | 36 | decide what is good or bad.  In this tutorial we'll assume that it is okay and | 
 | 37 | use this as a way to show some interesting parsing techniques.</p> | 
 | 38 |  | 
| Chris Lattner | 58f2c87 | 2007-11-02 05:42:52 +0000 | [diff] [blame] | 39 | <p>At the end of this tutorial, we'll <a href="#example">run through a nice | 
 | 40 | little example</a> that shows an example application that you can build with | 
 | 41 | Kaleidoscope and the feature set it now has.</p> | 
 | 42 |  | 
| Chris Lattner | c9d5d2c | 2007-11-01 06:49:54 +0000 | [diff] [blame] | 43 | </div> | 
 | 44 |  | 
 | 45 | <!-- *********************************************************************** --> | 
| Chris Lattner | 58f2c87 | 2007-11-02 05:42:52 +0000 | [diff] [blame] | 46 | <div class="doc_section"><a name="idea">User-defined Operators: the Idea</a></div> | 
| Chris Lattner | c9d5d2c | 2007-11-01 06:49:54 +0000 | [diff] [blame] | 47 | <!-- *********************************************************************** --> | 
 | 48 |  | 
 | 49 | <div class="doc_text"> | 
 | 50 |  | 
 | 51 | <p> | 
| Chris Lattner | 58f2c87 | 2007-11-02 05:42:52 +0000 | [diff] [blame] | 52 | The "operator overloading" that we will add to Kaleidoscope is more general than | 
| Chris Lattner | c9d5d2c | 2007-11-01 06:49:54 +0000 | [diff] [blame] | 53 | languages like C++.  In C++, you are only allowed to redefine existing | 
 | 54 | operators: you can't programatically change the grammar, introduce new | 
 | 55 | operators, change precedence levels, etc.  In this chapter, we will add this | 
 | 56 | capability to Kaleidoscope, which will allow us to round out the set of | 
 | 57 | operators that are supported, culminating in a more interesting example app.</p> | 
 | 58 |  | 
| Chris Lattner | 58f2c87 | 2007-11-02 05:42:52 +0000 | [diff] [blame] | 59 | <p>The point of going into user-defined operators in a tutorial like this is to | 
| Chris Lattner | c9d5d2c | 2007-11-01 06:49:54 +0000 | [diff] [blame] | 60 | show the power and flexibility of using a hand-written parser.  The parser we | 
 | 61 | are using so far is using recursive descent for most parts of the grammar, and  | 
 | 62 | operator precedence parsing for the expressions.  See <a  | 
 | 63 | href="LangImpl2.html">Chapter 2</a> for details.  Without using operator | 
 | 64 | precedence parsing, it would be very difficult to allow the programmer to | 
 | 65 | introduce new operators into the grammar: the grammar is dynamically extensible | 
 | 66 | as the JIT runs.</p> | 
 | 67 |  | 
 | 68 | <p>The two specific features we'll add are programmable unary operators (right | 
 | 69 | now, Kaleidoscope has no unary operators at all) as well as binary operators. | 
 | 70 | An example of this is:</p> | 
 | 71 |  | 
 | 72 | <div class="doc_code"> | 
 | 73 | <pre> | 
 | 74 | # Logical unary not. | 
 | 75 | def unary!(v) | 
 | 76 |   if v then | 
 | 77 |     0 | 
 | 78 |   else | 
 | 79 |     1; | 
 | 80 |  | 
 | 81 | # Define > with the same precedence as <. | 
 | 82 | def binary> 10 (LHS RHS) | 
 | 83 |   !(LHS < RHS);     # alternatively, could just use "RHS < LHS" | 
 | 84 |  | 
 | 85 | # Binary "logical or", (note that it does not "short circuit") | 
 | 86 | def binary| 5 (LHS RHS) | 
 | 87 |   if LHS then | 
 | 88 |     1 | 
 | 89 |   else if RHS then | 
 | 90 |     1 | 
 | 91 |   else | 
 | 92 |     0; | 
 | 93 |  | 
 | 94 | # Define = with slightly lower precedence than relationals. | 
 | 95 | def binary= 9 (LHS RHS) | 
 | 96 |   !(LHS < RHS | LHS > RHS); | 
 | 97 | </pre> | 
 | 98 | </div> | 
 | 99 |  | 
 | 100 | <p>Many languages aspire to being able to implement their standard runtime | 
 | 101 | library in the language itself.  In Kaleidoscope, we can implement significant | 
 | 102 | parts of the language in the library!</p> | 
 | 103 |  | 
 | 104 | <p>We will break down implementation of these features into two parts: | 
| Chris Lattner | 58f2c87 | 2007-11-02 05:42:52 +0000 | [diff] [blame] | 105 | implementing support for user-defined binary operators and adding unary | 
| Chris Lattner | c9d5d2c | 2007-11-01 06:49:54 +0000 | [diff] [blame] | 106 | operators.</p> | 
 | 107 |  | 
 | 108 | </div> | 
 | 109 |  | 
 | 110 | <!-- *********************************************************************** --> | 
| Chris Lattner | 58f2c87 | 2007-11-02 05:42:52 +0000 | [diff] [blame] | 111 | <div class="doc_section"><a name="binary">User-defined Binary Operators</a></div> | 
| Chris Lattner | c9d5d2c | 2007-11-01 06:49:54 +0000 | [diff] [blame] | 112 | <!-- *********************************************************************** --> | 
 | 113 |  | 
 | 114 | <div class="doc_text"> | 
 | 115 |  | 
| Chris Lattner | 58f2c87 | 2007-11-02 05:42:52 +0000 | [diff] [blame] | 116 | <p>Adding support for user-defined binary operators is pretty simple with our | 
| Chris Lattner | c9d5d2c | 2007-11-01 06:49:54 +0000 | [diff] [blame] | 117 | current framework.  We'll first add support for the unary/binary keywords:</p> | 
 | 118 |  | 
 | 119 | <div class="doc_code"> | 
 | 120 | <pre> | 
 | 121 | enum Token { | 
 | 122 |   ... | 
 | 123 |   <b>// operators | 
 | 124 |   tok_binary = -11, tok_unary = -12</b> | 
 | 125 | }; | 
 | 126 | ... | 
 | 127 | static int gettok() { | 
 | 128 | ... | 
 | 129 |     if (IdentifierStr == "for") return tok_for; | 
 | 130 |     if (IdentifierStr == "in") return tok_in; | 
 | 131 |     <b>if (IdentifierStr == "binary") return tok_binary; | 
 | 132 |     if (IdentifierStr == "unary") return tok_unary;</b> | 
 | 133 |     return tok_identifier; | 
 | 134 | </pre> | 
 | 135 | </div> | 
 | 136 |  | 
 | 137 | <p>This just adds lexer support for the unary and binary keywords, like we | 
 | 138 | did in <a href="LangImpl5.html#iflexer">previous chapters</a>.  One nice thing | 
 | 139 | about our current AST is that we represent binary operators fully generally | 
 | 140 | with their ASCII code as the opcode.  For our extended operators, we'll use the | 
 | 141 | same representation, so we don't need any new AST or parser support.</p> | 
 | 142 |  | 
 | 143 | <p>On the other hand, we have to be able to represent the definitions of these | 
 | 144 | new operators, in the "def binary| 5" part of the function definition.  In the | 
 | 145 | grammar so far, the "name" for the function definition is parsed as the | 
 | 146 | "prototype" production and into the <tt>PrototypeAST</tt> AST node.  To | 
 | 147 | represent our new user-defined operators as prototypes, we have to extend | 
 | 148 | the  <tt>PrototypeAST</tt> AST node like this:</p> | 
 | 149 |  | 
 | 150 | <div class="doc_code"> | 
 | 151 | <pre> | 
 | 152 | /// PrototypeAST - This class represents the "prototype" for a function, | 
 | 153 | /// which captures its argument names as well as if it is an operator. | 
 | 154 | class PrototypeAST { | 
 | 155 |   std::string Name; | 
 | 156 |   std::vector<std::string> Args; | 
 | 157 |   <b>bool isOperator; | 
 | 158 |   unsigned Precedence;  // Precedence if a binary op.</b> | 
 | 159 | public: | 
 | 160 |   PrototypeAST(const std::string &name, const std::vector<std::string> &args, | 
 | 161 |                <b>bool isoperator = false, unsigned prec = 0</b>) | 
 | 162 |   : Name(name), Args(args), <b>isOperator(isoperator), Precedence(prec)</b> {} | 
 | 163 |    | 
 | 164 |   <b>bool isUnaryOp() const { return isOperator && Args.size() == 1; } | 
 | 165 |   bool isBinaryOp() const { return isOperator && Args.size() == 2; } | 
 | 166 |    | 
 | 167 |   char getOperatorName() const { | 
 | 168 |     assert(isUnaryOp() || isBinaryOp()); | 
 | 169 |     return Name[Name.size()-1]; | 
 | 170 |   } | 
 | 171 |    | 
 | 172 |   unsigned getBinaryPrecedence() const { return Precedence; }</b> | 
 | 173 |    | 
 | 174 |   Function *Codegen(); | 
 | 175 | }; | 
 | 176 | </pre> | 
 | 177 | </div> | 
 | 178 |  | 
 | 179 | <p>Basically, in addition to knowing a name for the prototype, we now keep track | 
 | 180 | of whether it was an operator, and if it was, what precedence level the operator | 
| Chris Lattner | 58f2c87 | 2007-11-02 05:42:52 +0000 | [diff] [blame] | 181 | is at.  The precedence is only used for binary operators (as you'll see below, | 
 | 182 | it just doesn't apply for unary operators).  Now that we have a way to represent | 
 | 183 | the prototype for a user-defined operator, we need to parse it:</p> | 
 | 184 |  | 
 | 185 | <div class="doc_code"> | 
 | 186 | <pre> | 
 | 187 | /// prototype | 
 | 188 | ///   ::= id '(' id* ')' | 
 | 189 | <b>///   ::= binary LETTER number? (id, id)</b> | 
 | 190 | static PrototypeAST *ParsePrototype() { | 
 | 191 |   std::string FnName; | 
 | 192 |    | 
 | 193 |   <b>int Kind = 0;  // 0 = identifier, 1 = unary, 2 = binary. | 
 | 194 |   unsigned BinaryPrecedence = 30;</b> | 
 | 195 |    | 
 | 196 |   switch (CurTok) { | 
 | 197 |   default: | 
 | 198 |     return ErrorP("Expected function name in prototype"); | 
 | 199 |   case tok_identifier: | 
 | 200 |     FnName = IdentifierStr; | 
 | 201 |     Kind = 0; | 
 | 202 |     getNextToken(); | 
 | 203 |     break; | 
 | 204 |   <b>case tok_binary: | 
 | 205 |     getNextToken(); | 
 | 206 |     if (!isascii(CurTok)) | 
 | 207 |       return ErrorP("Expected binary operator"); | 
 | 208 |     FnName = "binary"; | 
 | 209 |     FnName += (char)CurTok; | 
 | 210 |     Kind = 2; | 
 | 211 |     getNextToken(); | 
 | 212 |      | 
 | 213 |     // Read the precedence if present. | 
 | 214 |     if (CurTok == tok_number) { | 
 | 215 |       if (NumVal < 1 || NumVal > 100) | 
 | 216 |         return ErrorP("Invalid precedecnce: must be 1..100"); | 
 | 217 |       BinaryPrecedence = (unsigned)NumVal; | 
 | 218 |       getNextToken(); | 
 | 219 |     } | 
 | 220 |     break;</b> | 
 | 221 |   } | 
 | 222 |    | 
 | 223 |   if (CurTok != '(') | 
 | 224 |     return ErrorP("Expected '(' in prototype"); | 
 | 225 |    | 
 | 226 |   std::vector<std::string> ArgNames; | 
 | 227 |   while (getNextToken() == tok_identifier) | 
 | 228 |     ArgNames.push_back(IdentifierStr); | 
 | 229 |   if (CurTok != ')') | 
 | 230 |     return ErrorP("Expected ')' in prototype"); | 
 | 231 |    | 
 | 232 |   // success. | 
 | 233 |   getNextToken();  // eat ')'. | 
 | 234 |    | 
 | 235 |   <b>// Verify right number of names for operator. | 
 | 236 |   if (Kind && ArgNames.size() != Kind) | 
 | 237 |     return ErrorP("Invalid number of operands for operator"); | 
 | 238 |    | 
 | 239 |   return new PrototypeAST(FnName, ArgNames, Kind != 0, BinaryPrecedence);</b> | 
 | 240 | } | 
 | 241 | </pre> | 
 | 242 | </div> | 
 | 243 |  | 
 | 244 | <p>This is all fairly straight-forward parsing code, and we have already seen | 
 | 245 | a lot of similar code in the past.  One interesting piece of this is the part | 
 | 246 | that sets up <tt>FnName</tt> for binary operators.  This builds names like | 
 | 247 | "binary@" for a newly defined "@" operator.  This takes advantage of the fact | 
 | 248 | that symbol names in the LLVM symbol table are allowed to have any character in | 
 | 249 | them, inluding embedded nul characters.</p> | 
 | 250 |  | 
 | 251 | <p>The next interesting piece is codegen support for these binary operators. | 
 | 252 | Given our current structure, this is a simple addition of a default case for our | 
 | 253 | existing binary operator node:</p> | 
 | 254 |  | 
 | 255 | <div class="doc_code"> | 
 | 256 | <pre> | 
 | 257 | Value *BinaryExprAST::Codegen() { | 
 | 258 |   Value *L = LHS->Codegen(); | 
 | 259 |   Value *R = RHS->Codegen(); | 
 | 260 |   if (L == 0 || R == 0) return 0; | 
 | 261 |    | 
 | 262 |   switch (Op) { | 
 | 263 |   case '+': return Builder.CreateAdd(L, R, "addtmp"); | 
 | 264 |   case '-': return Builder.CreateSub(L, R, "subtmp"); | 
 | 265 |   case '*': return Builder.CreateMul(L, R, "multmp"); | 
 | 266 |   case '<': | 
 | 267 |     L = Builder.CreateFCmpULT(L, R, "multmp"); | 
 | 268 |     // Convert bool 0/1 to double 0.0 or 1.0 | 
 | 269 |     return Builder.CreateUIToFP(L, Type::DoubleTy, "booltmp"); | 
 | 270 |   <b>default: break;</b> | 
 | 271 |   } | 
 | 272 |    | 
 | 273 |   <b>// If it wasn't a builtin binary operator, it must be a user defined one. Emit | 
 | 274 |   // a call to it. | 
 | 275 |   Function *F = TheModule->getFunction(std::string("binary")+Op); | 
 | 276 |   assert(F && "binary operator not found!"); | 
 | 277 |    | 
 | 278 |   Value *Ops[] = { L, R }; | 
 | 279 |   return Builder.CreateCall(F, Ops, Ops+2, "binop");</b> | 
 | 280 | } | 
 | 281 |  | 
 | 282 | </pre> | 
 | 283 | </div> | 
 | 284 |  | 
 | 285 | <p>As you can see above, the new code is actually really simple.  It just does | 
 | 286 | a lookup for the appropriate operator in the symbol table and generates a  | 
 | 287 | function call to it.  Since user-defined operators are just built as normal | 
 | 288 | functions (because the "prototype" boils down into a function with the right | 
 | 289 | name) everything falls into place.</p> | 
 | 290 |  | 
 | 291 | <p>The final missing piece is a bit of top level magic, here:</p> | 
 | 292 |  | 
 | 293 | <div class="doc_code"> | 
 | 294 | <pre> | 
 | 295 | Function *FunctionAST::Codegen() { | 
 | 296 |   NamedValues.clear(); | 
 | 297 |    | 
 | 298 |   Function *TheFunction = Proto->Codegen(); | 
 | 299 |   if (TheFunction == 0) | 
 | 300 |     return 0; | 
 | 301 |    | 
 | 302 |   <b>// If this is an operator, install it. | 
 | 303 |   if (Proto->isBinaryOp()) | 
 | 304 |     BinopPrecedence[Proto->getOperatorName()] = Proto->getBinaryPrecedence();</b> | 
 | 305 |    | 
 | 306 |   // Create a new basic block to start insertion into. | 
 | 307 |   BasicBlock *BB = new BasicBlock("entry", TheFunction); | 
 | 308 |   Builder.SetInsertPoint(BB); | 
 | 309 |    | 
 | 310 |   if (Value *RetVal = Body->Codegen()) { | 
 | 311 |     ... | 
 | 312 | </pre> | 
 | 313 | </div> | 
 | 314 |  | 
 | 315 | <p>Basically, before codegening a function, if it is a user-defined operator, we | 
 | 316 | register it in the precedence table.  This allows the binary operator parsing | 
 | 317 | logic we already have to handle it.  Since it is a fully-general operator | 
 | 318 | precedence parser, this is all we need to do to "extend the grammar".</p> | 
 | 319 |  | 
 | 320 | <p>With that, we have useful user-defined binary operators.  This builds a lot | 
 | 321 | on the previous framework we built for other operators.  Adding unary operators | 
 | 322 | is a bit more challenging, because we don't have any framework for it yet, lets | 
 | 323 | see what it takes.</p> | 
 | 324 |  | 
 | 325 | </div> | 
 | 326 |  | 
 | 327 | <!-- *********************************************************************** --> | 
 | 328 | <div class="doc_section"><a name="unary">User-defined Unary Operators</a></div> | 
 | 329 | <!-- *********************************************************************** --> | 
 | 330 |  | 
 | 331 | <div class="doc_text"> | 
 | 332 |  | 
 | 333 | <p>Since we don't currently support unary operators in the Kaleidoscope | 
 | 334 | langugage, we'll need to add everything for them.  Above, we added simple | 
 | 335 | support for the 'unary' keyword to the lexer.  In addition to that, we need an | 
 | 336 | AST node:</p> | 
 | 337 |  | 
 | 338 | <div class="doc_code"> | 
 | 339 | <pre> | 
 | 340 | /// UnaryExprAST - Expression class for a unary operator. | 
 | 341 | class UnaryExprAST : public ExprAST { | 
 | 342 |   char Opcode; | 
 | 343 |   ExprAST *Operand; | 
 | 344 | public: | 
 | 345 |   UnaryExprAST(char opcode, ExprAST *operand)  | 
 | 346 |     : Opcode(opcode), Operand(operand) {} | 
 | 347 |   virtual Value *Codegen(); | 
 | 348 | }; | 
 | 349 | </pre> | 
 | 350 | </div> | 
 | 351 |  | 
 | 352 | <p>This AST node is very simple and obvious by now.  It directly mirrors the | 
 | 353 | binary operator AST node, except that it only has one child.  With this, we | 
 | 354 | need to add the parsing logic.  Parsing a unary operator is pretty simple: we'll | 
 | 355 | add a new function to do it:</p> | 
 | 356 |  | 
 | 357 | <div class="doc_code"> | 
 | 358 | <pre> | 
 | 359 | /// unary | 
 | 360 | ///   ::= primary | 
 | 361 | ///   ::= '!' unary | 
 | 362 | static ExprAST *ParseUnary() { | 
 | 363 |   // If the current token is not an operator, it must be a primary expr. | 
 | 364 |   if (!isascii(CurTok) || CurTok == '(' || CurTok == ',') | 
 | 365 |     return ParsePrimary(); | 
 | 366 |    | 
 | 367 |   // If this is a unary operator, read it. | 
 | 368 |   int Opc = CurTok; | 
 | 369 |   getNextToken(); | 
 | 370 |   if (ExprAST *Operand = ParseUnary()) | 
 | 371 |     return new UnaryExprAST(Opc, Operand); | 
 | 372 |   return 0; | 
 | 373 | } | 
 | 374 | </pre> | 
 | 375 | </div> | 
 | 376 |  | 
 | 377 | <p>The grammar we add is pretty straight-forward here.  If we see a unary | 
 | 378 | operator when parsing a primary operator, we eat the operator as a prefix and | 
 | 379 | parse the remaining piece as another unary operator.  This allows us to handle | 
 | 380 | multiple unary operators (e.g. "!!x").  Note that unary operators can't have  | 
 | 381 | ambiguous parses like binary operators can, so there is no need for precedence | 
 | 382 | information.</p> | 
 | 383 |  | 
 | 384 | <p>The problem with the above is that we need to call ParseUnary from somewhere. | 
 | 385 | To do this, we change previous callers of ParsePrimary to call ParseUnary | 
 | 386 | instead:</p> | 
 | 387 |  | 
 | 388 | <div class="doc_code"> | 
 | 389 | <pre> | 
 | 390 | /// binoprhs | 
 | 391 | ///   ::= ('+' unary)* | 
 | 392 | static ExprAST *ParseBinOpRHS(int ExprPrec, ExprAST *LHS) { | 
 | 393 |   ... | 
 | 394 |     <b>// Parse the unary expression after the binary operator. | 
 | 395 |     ExprAST *RHS = ParseUnary(); | 
 | 396 |     if (!RHS) return 0;</b> | 
 | 397 |   ... | 
 | 398 | } | 
 | 399 | /// expression | 
 | 400 | ///   ::= unary binoprhs | 
 | 401 | /// | 
 | 402 | static ExprAST *ParseExpression() { | 
 | 403 |   <b>ExprAST *LHS = ParseUnary();</b> | 
 | 404 |   if (!LHS) return 0; | 
 | 405 |    | 
 | 406 |   return ParseBinOpRHS(0, LHS); | 
 | 407 | } | 
 | 408 | </pre> | 
 | 409 | </div> | 
 | 410 |  | 
 | 411 | <p>With these two simple changes, we now parse unary operators and build the | 
 | 412 | AST for them.  Next up, we need to add parser support for prototypes, to parse | 
 | 413 | the unary operator prototype.  We extend the binary operator code above  | 
 | 414 | with:</p> | 
 | 415 |  | 
 | 416 | <div class="doc_code"> | 
 | 417 | <pre> | 
 | 418 | /// prototype | 
 | 419 | ///   ::= id '(' id* ')' | 
 | 420 | ///   ::= binary LETTER number? (id, id) | 
 | 421 | <b>///   ::= unary LETTER (id)</b> | 
 | 422 | static PrototypeAST *ParsePrototype() { | 
 | 423 |   std::string FnName; | 
 | 424 |    | 
 | 425 |   int Kind = 0;  // 0 = identifier, 1 = unary, 2 = binary. | 
 | 426 |   unsigned BinaryPrecedence = 30; | 
 | 427 |    | 
 | 428 |   switch (CurTok) { | 
 | 429 |   default: | 
 | 430 |     return ErrorP("Expected function name in prototype"); | 
 | 431 |   case tok_identifier: | 
 | 432 |     FnName = IdentifierStr; | 
 | 433 |     Kind = 0; | 
 | 434 |     getNextToken(); | 
 | 435 |     break; | 
 | 436 |   <b>case tok_unary: | 
 | 437 |     getNextToken(); | 
 | 438 |     if (!isascii(CurTok)) | 
 | 439 |       return ErrorP("Expected unary operator"); | 
 | 440 |     FnName = "unary"; | 
 | 441 |     FnName += (char)CurTok; | 
 | 442 |     Kind = 1; | 
 | 443 |     getNextToken(); | 
 | 444 |     break;</b> | 
 | 445 |   case tok_binary: | 
 | 446 |     ... | 
 | 447 | </pre> | 
 | 448 | </div> | 
 | 449 |  | 
 | 450 | <p>As with binary operators, we name unary operators with a name that includes | 
 | 451 | the operator character.  This assists us at code generation time.  Speaking of, | 
 | 452 | the final piece we need to add is codegen support for unary operators.  It looks | 
 | 453 | like this:</p> | 
 | 454 |  | 
 | 455 | <div class="doc_code"> | 
 | 456 | <pre> | 
 | 457 | Value *UnaryExprAST::Codegen() { | 
 | 458 |   Value *OperandV = Operand->Codegen(); | 
 | 459 |   if (OperandV == 0) return 0; | 
 | 460 |    | 
 | 461 |   Function *F = TheModule->getFunction(std::string("unary")+Opcode); | 
 | 462 |   if (F == 0) | 
 | 463 |     return ErrorV("Unknown unary operator"); | 
 | 464 |    | 
 | 465 |   return Builder.CreateCall(F, OperandV, "unop"); | 
 | 466 | } | 
 | 467 | </pre> | 
 | 468 | </div> | 
 | 469 |  | 
 | 470 | <p>This code is similar to, but simpler than, the code for binary operators.  It | 
 | 471 | is simpler primarily because it doesn't need to handle any predefined operators. | 
 | 472 | </p> | 
 | 473 |  | 
 | 474 | </div> | 
 | 475 |  | 
 | 476 | <!-- *********************************************************************** --> | 
 | 477 | <div class="doc_section"><a name="example">Kicking the Tires</a></div> | 
 | 478 | <!-- *********************************************************************** --> | 
 | 479 |  | 
 | 480 | <div class="doc_text"> | 
 | 481 |  | 
 | 482 | <p>It is somewhat hard to believe, but with a few simple extensions we've | 
| Chris Lattner | b5d81b3 | 2007-11-02 05:54:25 +0000 | [diff] [blame] | 483 | covered in the last chapters, we have grown a real-ish language.  With this, we  | 
| Chris Lattner | 58f2c87 | 2007-11-02 05:42:52 +0000 | [diff] [blame] | 484 | can do a lot of interesting things, including I/O, math, and a bunch of other | 
 | 485 | things.  For example, we can now add a nice sequencing operator (printd is | 
 | 486 | defined to print out the specified value and a newline):</p> | 
 | 487 |  | 
 | 488 | <div class="doc_code"> | 
 | 489 | <pre> | 
 | 490 | ready> <b>extern printd(x);</b> | 
 | 491 | Read extern: declare double @printd(double) | 
 | 492 | ready> <b>def binary : 1 (x y) 0;  # Low-precedence operator that ignores operands.</b> | 
 | 493 | .. | 
 | 494 | ready> <b>printd(123) : printd(456) : printd(789);</b> | 
 | 495 | 123.000000 | 
 | 496 | 456.000000 | 
 | 497 | 789.000000 | 
 | 498 | Evaluated to 0.000000 | 
 | 499 | </pre> | 
 | 500 | </div> | 
 | 501 |  | 
| Chris Lattner | b5d81b3 | 2007-11-02 05:54:25 +0000 | [diff] [blame] | 502 | <p>We can also define a bunch of other "primitive" operations, such as:</p> | 
| Chris Lattner | 58f2c87 | 2007-11-02 05:42:52 +0000 | [diff] [blame] | 503 |  | 
 | 504 | <div class="doc_code"> | 
 | 505 | <pre> | 
 | 506 | # Logical unary not. | 
 | 507 | def unary!(v) | 
 | 508 |   if v then | 
 | 509 |     0 | 
 | 510 |   else | 
 | 511 |     1; | 
 | 512 |      | 
 | 513 | # Unary negate. | 
 | 514 | def unary-(v) | 
 | 515 |   0-v; | 
 | 516 |  | 
 | 517 | # Define > with the same precedence as >.  We could also easily define | 
 | 518 | # <= etc. | 
 | 519 | def binary> 10 (LHS RHS) | 
 | 520 |   !(LHS < RHS); | 
 | 521 |  | 
 | 522 | # Binary logical or, which does not short circuit.  | 
 | 523 | def binary| 5 (LHS RHS) | 
 | 524 |   if LHS then | 
 | 525 |     1 | 
 | 526 |   else if RHS then | 
 | 527 |     1 | 
 | 528 |   else | 
 | 529 |     0; | 
 | 530 |  | 
 | 531 | # Binary logical and, which does not short circuit.  | 
 | 532 | def binary& 6 (LHS RHS) | 
 | 533 |   if !LHS then | 
 | 534 |     0 | 
 | 535 |   else | 
 | 536 |     !!RHS; | 
 | 537 |  | 
 | 538 | # Define = with slightly lower precedence than relationals. | 
 | 539 | def binary = 9 (LHS RHS) | 
 | 540 |   !(LHS < RHS | LHS > RHS); | 
 | 541 |  | 
 | 542 | </pre> | 
 | 543 | </div> | 
| Chris Lattner | c9d5d2c | 2007-11-01 06:49:54 +0000 | [diff] [blame] | 544 |  | 
 | 545 |  | 
| Chris Lattner | 58f2c87 | 2007-11-02 05:42:52 +0000 | [diff] [blame] | 546 | <p>Given the previous if/then/else support, we can also define interesting | 
 | 547 | functions for I/O.  For example, the following prints out a character whose | 
 | 548 | "density" reflects the value passed in: the lower the value, the denser the | 
 | 549 | character:</p> | 
 | 550 |  | 
 | 551 | <div class="doc_code"> | 
 | 552 | <pre> | 
 | 553 | ready> | 
 | 554 | <b> | 
 | 555 | extern putchard(char) | 
 | 556 | def printdensity(d) | 
 | 557 |   if d > 8 then | 
 | 558 |     putchard(32)  # ' ' | 
 | 559 |   else if d > 4 then | 
 | 560 |     putchard(46)  # '.' | 
 | 561 |   else if d > 2 then | 
 | 562 |     putchard(43)  # '+' | 
 | 563 |   else | 
 | 564 |     putchard(42); # '*'</b> | 
 | 565 | ... | 
 | 566 | ready> <b>printdensity(1): printdensity(2): printdensity(3) :  | 
 | 567 |           printdensity(4): printdensity(5): printdensity(9): putchard(10);</b> | 
 | 568 | *++..  | 
 | 569 | Evaluated to 0.000000 | 
 | 570 | </pre> | 
 | 571 | </div> | 
 | 572 |  | 
 | 573 | <p>Based on these simple primitive operations, we can start to define more | 
 | 574 | interesting things.  For example, here's a little function that solves for the | 
 | 575 | number of iterations it takes for a function in the complex plane to | 
 | 576 | converge:</p> | 
 | 577 |  | 
 | 578 | <div class="doc_code"> | 
 | 579 | <pre> | 
 | 580 | # determine whether the specific location diverges. | 
 | 581 | # Solve for z = z^2 + c in the complex plane. | 
 | 582 | def mandleconverger(real imag iters creal cimag) | 
 | 583 |   if iters > 255 | (real*real + imag*imag > 4) then | 
 | 584 |     iters | 
 | 585 |   else | 
 | 586 |     mandleconverger(real*real - imag*imag + creal, | 
 | 587 |                     2*real*imag + cimag, | 
 | 588 |                     iters+1, creal, cimag); | 
 | 589 |  | 
 | 590 | # return the number of iterations required for the iteration to escape | 
 | 591 | def mandleconverge(real imag) | 
 | 592 |   mandleconverger(real, imag, 0, real, imag); | 
 | 593 | </pre> | 
 | 594 | </div> | 
 | 595 |  | 
| Chris Lattner | b5d81b3 | 2007-11-02 05:54:25 +0000 | [diff] [blame] | 596 | <p>This "z = z<sup>2</sup> + c" function is a beautiful little creature that is the basis | 
| Chris Lattner | 58f2c87 | 2007-11-02 05:42:52 +0000 | [diff] [blame] | 597 | for computation of the <a  | 
 | 598 | href="http://en.wikipedia.org/wiki/Mandelbrot_set">Mandelbrot Set</a>.  Our | 
 | 599 | <tt>mandelconverge</tt> function returns the number of iterations that it takes | 
 | 600 | for a complex orbit to escape, saturating to 255.  This is not a very useful | 
 | 601 | function by itself, but if you plot its value over a two-dimensional plane, | 
 | 602 | you can see the mandelbrot set.  Given that we are limited to using putchard | 
 | 603 | here, our amazing graphical output is limited, but we can whip together | 
 | 604 | something using the density plotter above:</p> | 
 | 605 |  | 
 | 606 | <div class="doc_code"> | 
 | 607 | <pre> | 
 | 608 | # compute and plot the mandlebrot set with the specified 2 dimentional range | 
 | 609 | # info. | 
 | 610 | def mandelhelp(xmin xmax xstep   ymin ymax ystep) | 
 | 611 |   for y = ymin, y < ymax, ystep in ( | 
 | 612 |     (for x = xmin, x < xmax, xstep in | 
 | 613 |        printdensity(mandleconverge(x,y))) | 
 | 614 |     : putchard(10) | 
 | 615 |   ) | 
 | 616 |   | 
 | 617 | # mandel - This is a convenient helper function for ploting the mandelbrot set | 
 | 618 | # from the specified position with the specified magnification. | 
 | 619 | def mandel(realstart imagstart realmag imagmag)  | 
 | 620 |   mandelhelp(realstart, realstart+realmag*78, realmag, | 
 | 621 |              imagstart, imagstart+imagmag*40, imagmag); | 
 | 622 | </pre> | 
 | 623 | </div> | 
 | 624 |  | 
 | 625 | <p>Given this, we can try plotting out the mandlebrot set!  Lets try it out:</p> | 
 | 626 |  | 
 | 627 | <div class="doc_code"> | 
 | 628 | <pre> | 
 | 629 | ready> <b>mandel(-2.3, -1.3, 0.05, 0.07);</b> | 
 | 630 | *******************************+++++++++++************************************* | 
 | 631 | *************************+++++++++++++++++++++++******************************* | 
 | 632 | **********************+++++++++++++++++++++++++++++**************************** | 
 | 633 | *******************+++++++++++++++++++++.. ...++++++++************************* | 
 | 634 | *****************++++++++++++++++++++++.... ...+++++++++*********************** | 
 | 635 | ***************+++++++++++++++++++++++.....   ...+++++++++********************* | 
 | 636 | **************+++++++++++++++++++++++....     ....+++++++++******************** | 
 | 637 | *************++++++++++++++++++++++......      .....++++++++******************* | 
 | 638 | ************+++++++++++++++++++++.......       .......+++++++****************** | 
 | 639 | ***********+++++++++++++++++++....                ... .+++++++***************** | 
 | 640 | **********+++++++++++++++++.......                     .+++++++**************** | 
 | 641 | *********++++++++++++++...........                    ...+++++++*************** | 
 | 642 | ********++++++++++++............                      ...++++++++************** | 
 | 643 | ********++++++++++... ..........                        .++++++++************** | 
 | 644 | *******+++++++++.....                                   .+++++++++************* | 
 | 645 | *******++++++++......                                  ..+++++++++************* | 
 | 646 | *******++++++.......                                   ..+++++++++************* | 
 | 647 | *******+++++......                                     ..+++++++++************* | 
 | 648 | *******.... ....                                      ...+++++++++************* | 
 | 649 | *******.... .                                         ...+++++++++************* | 
 | 650 | *******+++++......                                    ...+++++++++************* | 
 | 651 | *******++++++.......                                   ..+++++++++************* | 
 | 652 | *******++++++++......                                   .+++++++++************* | 
 | 653 | *******+++++++++.....                                  ..+++++++++************* | 
 | 654 | ********++++++++++... ..........                        .++++++++************** | 
 | 655 | ********++++++++++++............                      ...++++++++************** | 
 | 656 | *********++++++++++++++..........                     ...+++++++*************** | 
 | 657 | **********++++++++++++++++........                     .+++++++**************** | 
 | 658 | **********++++++++++++++++++++....                ... ..+++++++**************** | 
 | 659 | ***********++++++++++++++++++++++.......       .......++++++++***************** | 
 | 660 | ************+++++++++++++++++++++++......      ......++++++++****************** | 
 | 661 | **************+++++++++++++++++++++++....      ....++++++++******************** | 
 | 662 | ***************+++++++++++++++++++++++.....   ...+++++++++********************* | 
 | 663 | *****************++++++++++++++++++++++....  ...++++++++*********************** | 
 | 664 | *******************+++++++++++++++++++++......++++++++************************* | 
 | 665 | *********************++++++++++++++++++++++.++++++++*************************** | 
 | 666 | *************************+++++++++++++++++++++++******************************* | 
 | 667 | ******************************+++++++++++++************************************ | 
 | 668 | ******************************************************************************* | 
 | 669 | ******************************************************************************* | 
 | 670 | ******************************************************************************* | 
 | 671 | Evaluated to 0.000000 | 
 | 672 | ready> <b>mandel(-2, -1, 0.02, 0.04);</b> | 
 | 673 | **************************+++++++++++++++++++++++++++++++++++++++++++++++++++++ | 
 | 674 | ***********************++++++++++++++++++++++++++++++++++++++++++++++++++++++++ | 
 | 675 | *********************+++++++++++++++++++++++++++++++++++++++++++++++++++++++++. | 
 | 676 | *******************+++++++++++++++++++++++++++++++++++++++++++++++++++++++++... | 
 | 677 | *****************+++++++++++++++++++++++++++++++++++++++++++++++++++++++++..... | 
 | 678 | ***************++++++++++++++++++++++++++++++++++++++++++++++++++++++++........ | 
 | 679 | **************++++++++++++++++++++++++++++++++++++++++++++++++++++++........... | 
 | 680 | ************+++++++++++++++++++++++++++++++++++++++++++++++++++++.............. | 
 | 681 | ***********++++++++++++++++++++++++++++++++++++++++++++++++++........        .  | 
 | 682 | **********++++++++++++++++++++++++++++++++++++++++++++++.............           | 
 | 683 | ********+++++++++++++++++++++++++++++++++++++++++++..................           | 
 | 684 | *******+++++++++++++++++++++++++++++++++++++++.......................           | 
 | 685 | ******+++++++++++++++++++++++++++++++++++...........................            | 
 | 686 | *****++++++++++++++++++++++++++++++++............................               | 
 | 687 | *****++++++++++++++++++++++++++++...............................                | 
 | 688 | ****++++++++++++++++++++++++++......   .........................                | 
 | 689 | ***++++++++++++++++++++++++.........     ......    ...........                  | 
 | 690 | ***++++++++++++++++++++++............                                           | 
 | 691 | **+++++++++++++++++++++..............                                           | 
 | 692 | **+++++++++++++++++++................                                           | 
 | 693 | *++++++++++++++++++.................                                            | 
 | 694 | *++++++++++++++++............ ...                                               | 
 | 695 | *++++++++++++++..............                                                   | 
 | 696 | *+++....++++................                                                    | 
 | 697 | *..........  ...........                                                        | 
 | 698 | *                                                                               | 
 | 699 | *..........  ...........                                                        | 
 | 700 | *+++....++++................                                                    | 
 | 701 | *++++++++++++++..............                                                   | 
 | 702 | *++++++++++++++++............ ...                                               | 
 | 703 | *++++++++++++++++++.................                                            | 
 | 704 | **+++++++++++++++++++................                                           | 
 | 705 | **+++++++++++++++++++++..............                                           | 
 | 706 | ***++++++++++++++++++++++............                                           | 
 | 707 | ***++++++++++++++++++++++++.........     ......    ...........                  | 
 | 708 | ****++++++++++++++++++++++++++......   .........................                | 
 | 709 | *****++++++++++++++++++++++++++++...............................                | 
 | 710 | *****++++++++++++++++++++++++++++++++............................               | 
 | 711 | ******+++++++++++++++++++++++++++++++++++...........................            | 
 | 712 | *******+++++++++++++++++++++++++++++++++++++++.......................           | 
 | 713 | ********+++++++++++++++++++++++++++++++++++++++++++..................           | 
 | 714 | Evaluated to 0.000000 | 
 | 715 | ready> <b>mandel(-0.9, -1.4, 0.02, 0.03);</b> | 
 | 716 | ******************************************************************************* | 
 | 717 | ******************************************************************************* | 
 | 718 | ******************************************************************************* | 
 | 719 | **********+++++++++++++++++++++************************************************ | 
 | 720 | *+++++++++++++++++++++++++++++++++++++++*************************************** | 
 | 721 | +++++++++++++++++++++++++++++++++++++++++++++********************************** | 
 | 722 | ++++++++++++++++++++++++++++++++++++++++++++++++++***************************** | 
 | 723 | ++++++++++++++++++++++++++++++++++++++++++++++++++++++************************* | 
 | 724 | +++++++++++++++++++++++++++++++++++++++++++++++++++++++++********************** | 
 | 725 | +++++++++++++++++++++++++++++++++.........++++++++++++++++++******************* | 
 | 726 | +++++++++++++++++++++++++++++++....   ......+++++++++++++++++++**************** | 
 | 727 | +++++++++++++++++++++++++++++.......  ........+++++++++++++++++++************** | 
 | 728 | ++++++++++++++++++++++++++++........   ........++++++++++++++++++++************ | 
 | 729 | +++++++++++++++++++++++++++.........     ..  ...+++++++++++++++++++++********** | 
 | 730 | ++++++++++++++++++++++++++...........        ....++++++++++++++++++++++******** | 
 | 731 | ++++++++++++++++++++++++.............       .......++++++++++++++++++++++****** | 
 | 732 | +++++++++++++++++++++++.............        ........+++++++++++++++++++++++**** | 
 | 733 | ++++++++++++++++++++++...........           ..........++++++++++++++++++++++*** | 
 | 734 | ++++++++++++++++++++...........                .........++++++++++++++++++++++* | 
 | 735 | ++++++++++++++++++............                  ...........++++++++++++++++++++ | 
 | 736 | ++++++++++++++++...............                 .............++++++++++++++++++ | 
 | 737 | ++++++++++++++.................                 ...............++++++++++++++++ | 
 | 738 | ++++++++++++..................                  .................++++++++++++++ | 
 | 739 | +++++++++..................                      .................+++++++++++++ | 
 | 740 | ++++++........        .                               .........  ..++++++++++++ | 
 | 741 | ++............                                         ......    ....++++++++++ | 
 | 742 | ..............                                                    ...++++++++++ | 
 | 743 | ..............                                                    ....+++++++++ | 
 | 744 | ..............                                                    .....++++++++ | 
 | 745 | .............                                                    ......++++++++ | 
 | 746 | ...........                                                     .......++++++++ | 
 | 747 | .........                                                       ........+++++++ | 
 | 748 | .........                                                       ........+++++++ | 
 | 749 | .........                                                           ....+++++++ | 
 | 750 | ........                                                             ...+++++++ | 
 | 751 | .......                                                              ...+++++++ | 
 | 752 |                                                                     ....+++++++ | 
 | 753 |                                                                    .....+++++++ | 
 | 754 |                                                                     ....+++++++ | 
 | 755 |                                                                     ....+++++++ | 
 | 756 |                                                                     ....+++++++ | 
 | 757 | Evaluated to 0.000000 | 
 | 758 | ready> <b>^D</b> | 
 | 759 | </pre> | 
 | 760 | </div> | 
 | 761 |  | 
 | 762 | <p>At this point, you may be starting to realize that Kaleidoscope is a real | 
 | 763 | and powerful language.  It may not be self-similar :), but it can be used to | 
 | 764 | plot things that are!</p> | 
 | 765 |  | 
 | 766 | <p>With this, we conclude the "adding user-defined operators" chapter of the | 
 | 767 | tutorial.  We successfully extended our language with the ability to extend the | 
 | 768 | language in the library, and showed how this can be used to build a simple but | 
 | 769 | interesting end user application in Kaleidoscope.  At this point, Kaleidoscope | 
 | 770 | can build a variety of applications that are functional and can call functions | 
| Chris Lattner | b5d81b3 | 2007-11-02 05:54:25 +0000 | [diff] [blame] | 771 | with side-effects, but it can't actually define and mutate a variable itself. | 
| Chris Lattner | 58f2c87 | 2007-11-02 05:42:52 +0000 | [diff] [blame] | 772 | </p> | 
 | 773 |  | 
 | 774 | <p>Strikingly, lack of this feature is an important limitation for some | 
 | 775 | languages, and it is not at all obvious how to <a href="LangImpl7.html">add | 
 | 776 | support for mutable variables</a> without having to add an "SSA construction" | 
 | 777 | phase to your front-end.  In the next chapter, we will describe how you can | 
 | 778 | add this without building SSA in your front-end.</p> | 
| Chris Lattner | c9d5d2c | 2007-11-01 06:49:54 +0000 | [diff] [blame] | 779 |  | 
 | 780 | </div> | 
 | 781 |  | 
 | 782 |  | 
 | 783 | <!-- *********************************************************************** --> | 
 | 784 | <div class="doc_section"><a name="code">Full Code Listing</a></div> | 
 | 785 | <!-- *********************************************************************** --> | 
 | 786 |  | 
 | 787 | <div class="doc_text"> | 
 | 788 |  | 
 | 789 | <p> | 
 | 790 | Here is the complete code listing for our running example, enhanced with the | 
 | 791 | if/then/else and for expressions..  To build this example, use: | 
 | 792 | </p> | 
 | 793 |  | 
 | 794 | <div class="doc_code"> | 
 | 795 | <pre> | 
 | 796 |    # Compile | 
 | 797 |    g++ -g toy.cpp `llvm-config --cppflags --ldflags --libs core jit native` -O3 -o toy | 
 | 798 |    # Run | 
 | 799 |    ./toy | 
 | 800 | </pre> | 
 | 801 | </div> | 
 | 802 |  | 
 | 803 | <p>Here is the code:</p> | 
 | 804 |  | 
 | 805 | <div class="doc_code"> | 
 | 806 | <pre> | 
| Chris Lattner | 58f2c87 | 2007-11-02 05:42:52 +0000 | [diff] [blame] | 807 | #include "llvm/DerivedTypes.h" | 
 | 808 | #include "llvm/ExecutionEngine/ExecutionEngine.h" | 
 | 809 | #include "llvm/Module.h" | 
 | 810 | #include "llvm/ModuleProvider.h" | 
 | 811 | #include "llvm/PassManager.h" | 
 | 812 | #include "llvm/Analysis/Verifier.h" | 
 | 813 | #include "llvm/Target/TargetData.h" | 
 | 814 | #include "llvm/Transforms/Scalar.h" | 
 | 815 | #include "llvm/Support/LLVMBuilder.h" | 
 | 816 | #include <cstdio> | 
 | 817 | #include <string> | 
 | 818 | #include <map> | 
 | 819 | #include <vector> | 
 | 820 | using namespace llvm; | 
 | 821 |  | 
 | 822 | //===----------------------------------------------------------------------===// | 
 | 823 | // Lexer | 
 | 824 | //===----------------------------------------------------------------------===// | 
 | 825 |  | 
 | 826 | // The lexer returns tokens [0-255] if it is an unknown character, otherwise one | 
 | 827 | // of these for known things. | 
 | 828 | enum Token { | 
 | 829 |   tok_eof = -1, | 
 | 830 |  | 
 | 831 |   // commands | 
 | 832 |   tok_def = -2, tok_extern = -3, | 
 | 833 |  | 
 | 834 |   // primary | 
 | 835 |   tok_identifier = -4, tok_number = -5, | 
 | 836 |    | 
 | 837 |   // control | 
 | 838 |   tok_if = -6, tok_then = -7, tok_else = -8, | 
 | 839 |   tok_for = -9, tok_in = -10, | 
 | 840 |    | 
 | 841 |   // operators | 
 | 842 |   tok_binary = -11, tok_unary = -12 | 
 | 843 | }; | 
 | 844 |  | 
 | 845 | static std::string IdentifierStr;  // Filled in if tok_identifier | 
 | 846 | static double NumVal;              // Filled in if tok_number | 
 | 847 |  | 
 | 848 | /// gettok - Return the next token from standard input. | 
 | 849 | static int gettok() { | 
 | 850 |   static int LastChar = ' '; | 
 | 851 |  | 
 | 852 |   // Skip any whitespace. | 
 | 853 |   while (isspace(LastChar)) | 
 | 854 |     LastChar = getchar(); | 
 | 855 |  | 
 | 856 |   if (isalpha(LastChar)) { // identifier: [a-zA-Z][a-zA-Z0-9]* | 
 | 857 |     IdentifierStr = LastChar; | 
 | 858 |     while (isalnum((LastChar = getchar()))) | 
 | 859 |       IdentifierStr += LastChar; | 
 | 860 |  | 
 | 861 |     if (IdentifierStr == "def") return tok_def; | 
 | 862 |     if (IdentifierStr == "extern") return tok_extern; | 
 | 863 |     if (IdentifierStr == "if") return tok_if; | 
 | 864 |     if (IdentifierStr == "then") return tok_then; | 
 | 865 |     if (IdentifierStr == "else") return tok_else; | 
 | 866 |     if (IdentifierStr == "for") return tok_for; | 
 | 867 |     if (IdentifierStr == "in") return tok_in; | 
 | 868 |     if (IdentifierStr == "binary") return tok_binary; | 
 | 869 |     if (IdentifierStr == "unary") return tok_unary; | 
 | 870 |     return tok_identifier; | 
 | 871 |   } | 
 | 872 |  | 
 | 873 |   if (isdigit(LastChar) || LastChar == '.') {   // Number: [0-9.]+ | 
 | 874 |     std::string NumStr; | 
 | 875 |     do { | 
 | 876 |       NumStr += LastChar; | 
 | 877 |       LastChar = getchar(); | 
 | 878 |     } while (isdigit(LastChar) || LastChar == '.'); | 
 | 879 |  | 
 | 880 |     NumVal = strtod(NumStr.c_str(), 0); | 
 | 881 |     return tok_number; | 
 | 882 |   } | 
 | 883 |  | 
 | 884 |   if (LastChar == '#') { | 
 | 885 |     // Comment until end of line. | 
 | 886 |     do LastChar = getchar(); | 
 | 887 |     while (LastChar != EOF && LastChar != '\n' & LastChar != '\r'); | 
 | 888 |      | 
 | 889 |     if (LastChar != EOF) | 
 | 890 |       return gettok(); | 
 | 891 |   } | 
 | 892 |    | 
 | 893 |   // Check for end of file.  Don't eat the EOF. | 
 | 894 |   if (LastChar == EOF) | 
 | 895 |     return tok_eof; | 
 | 896 |  | 
 | 897 |   // Otherwise, just return the character as its ascii value. | 
 | 898 |   int ThisChar = LastChar; | 
 | 899 |   LastChar = getchar(); | 
 | 900 |   return ThisChar; | 
 | 901 | } | 
 | 902 |  | 
 | 903 | //===----------------------------------------------------------------------===// | 
 | 904 | // Abstract Syntax Tree (aka Parse Tree) | 
 | 905 | //===----------------------------------------------------------------------===// | 
 | 906 |  | 
 | 907 | /// ExprAST - Base class for all expression nodes. | 
 | 908 | class ExprAST { | 
 | 909 | public: | 
 | 910 |   virtual ~ExprAST() {} | 
 | 911 |   virtual Value *Codegen() = 0; | 
 | 912 | }; | 
 | 913 |  | 
 | 914 | /// NumberExprAST - Expression class for numeric literals like "1.0". | 
 | 915 | class NumberExprAST : public ExprAST { | 
 | 916 |   double Val; | 
 | 917 | public: | 
 | 918 |   NumberExprAST(double val) : Val(val) {} | 
 | 919 |   virtual Value *Codegen(); | 
 | 920 | }; | 
 | 921 |  | 
 | 922 | /// VariableExprAST - Expression class for referencing a variable, like "a". | 
 | 923 | class VariableExprAST : public ExprAST { | 
 | 924 |   std::string Name; | 
 | 925 | public: | 
 | 926 |   VariableExprAST(const std::string &name) : Name(name) {} | 
 | 927 |   virtual Value *Codegen(); | 
 | 928 | }; | 
 | 929 |  | 
 | 930 | /// UnaryExprAST - Expression class for a unary operator. | 
 | 931 | class UnaryExprAST : public ExprAST { | 
 | 932 |   char Opcode; | 
 | 933 |   ExprAST *Operand; | 
 | 934 | public: | 
 | 935 |   UnaryExprAST(char opcode, ExprAST *operand)  | 
 | 936 |     : Opcode(opcode), Operand(operand) {} | 
 | 937 |   virtual Value *Codegen(); | 
 | 938 | }; | 
 | 939 |  | 
 | 940 | /// BinaryExprAST - Expression class for a binary operator. | 
 | 941 | class BinaryExprAST : public ExprAST { | 
 | 942 |   char Op; | 
 | 943 |   ExprAST *LHS, *RHS; | 
 | 944 | public: | 
 | 945 |   BinaryExprAST(char op, ExprAST *lhs, ExprAST *rhs)  | 
 | 946 |     : Op(op), LHS(lhs), RHS(rhs) {} | 
 | 947 |   virtual Value *Codegen(); | 
 | 948 | }; | 
 | 949 |  | 
 | 950 | /// CallExprAST - Expression class for function calls. | 
 | 951 | class CallExprAST : public ExprAST { | 
 | 952 |   std::string Callee; | 
 | 953 |   std::vector<ExprAST*> Args; | 
 | 954 | public: | 
 | 955 |   CallExprAST(const std::string &callee, std::vector<ExprAST*> &args) | 
 | 956 |     : Callee(callee), Args(args) {} | 
 | 957 |   virtual Value *Codegen(); | 
 | 958 | }; | 
 | 959 |  | 
 | 960 | /// IfExprAST - Expression class for if/then/else. | 
 | 961 | class IfExprAST : public ExprAST { | 
 | 962 |   ExprAST *Cond, *Then, *Else; | 
 | 963 | public: | 
 | 964 |   IfExprAST(ExprAST *cond, ExprAST *then, ExprAST *_else) | 
 | 965 |   : Cond(cond), Then(then), Else(_else) {} | 
 | 966 |   virtual Value *Codegen(); | 
 | 967 | }; | 
 | 968 |  | 
 | 969 | /// ForExprAST - Expression class for for/in. | 
 | 970 | class ForExprAST : public ExprAST { | 
 | 971 |   std::string VarName; | 
 | 972 |   ExprAST *Start, *End, *Step, *Body; | 
 | 973 | public: | 
 | 974 |   ForExprAST(const std::string &varname, ExprAST *start, ExprAST *end, | 
 | 975 |              ExprAST *step, ExprAST *body) | 
 | 976 |     : VarName(varname), Start(start), End(end), Step(step), Body(body) {} | 
 | 977 |   virtual Value *Codegen(); | 
 | 978 | }; | 
 | 979 |  | 
 | 980 | /// PrototypeAST - This class represents the "prototype" for a function, | 
 | 981 | /// which captures its argument names as well as if it is an operator. | 
 | 982 | class PrototypeAST { | 
 | 983 |   std::string Name; | 
 | 984 |   std::vector<std::string> Args; | 
 | 985 |   bool isOperator; | 
 | 986 |   unsigned Precedence;  // Precedence if a binary op. | 
 | 987 | public: | 
 | 988 |   PrototypeAST(const std::string &name, const std::vector<std::string> &args, | 
 | 989 |                bool isoperator = false, unsigned prec = 0) | 
 | 990 |   : Name(name), Args(args), isOperator(isoperator), Precedence(prec) {} | 
 | 991 |    | 
 | 992 |   bool isUnaryOp() const { return isOperator && Args.size() == 1; } | 
 | 993 |   bool isBinaryOp() const { return isOperator && Args.size() == 2; } | 
 | 994 |    | 
 | 995 |   char getOperatorName() const { | 
 | 996 |     assert(isUnaryOp() || isBinaryOp()); | 
 | 997 |     return Name[Name.size()-1]; | 
 | 998 |   } | 
 | 999 |    | 
 | 1000 |   unsigned getBinaryPrecedence() const { return Precedence; } | 
 | 1001 |    | 
 | 1002 |   Function *Codegen(); | 
 | 1003 | }; | 
 | 1004 |  | 
 | 1005 | /// FunctionAST - This class represents a function definition itself. | 
 | 1006 | class FunctionAST { | 
 | 1007 |   PrototypeAST *Proto; | 
 | 1008 |   ExprAST *Body; | 
 | 1009 | public: | 
 | 1010 |   FunctionAST(PrototypeAST *proto, ExprAST *body) | 
 | 1011 |     : Proto(proto), Body(body) {} | 
 | 1012 |    | 
 | 1013 |   Function *Codegen(); | 
 | 1014 | }; | 
 | 1015 |  | 
 | 1016 | //===----------------------------------------------------------------------===// | 
 | 1017 | // Parser | 
 | 1018 | //===----------------------------------------------------------------------===// | 
 | 1019 |  | 
 | 1020 | /// CurTok/getNextToken - Provide a simple token buffer.  CurTok is the current | 
 | 1021 | /// token the parser it looking at.  getNextToken reads another token from the | 
 | 1022 | /// lexer and updates CurTok with its results. | 
 | 1023 | static int CurTok; | 
 | 1024 | static int getNextToken() { | 
 | 1025 |   return CurTok = gettok(); | 
 | 1026 | } | 
 | 1027 |  | 
 | 1028 | /// BinopPrecedence - This holds the precedence for each binary operator that is | 
 | 1029 | /// defined. | 
 | 1030 | static std::map<char, int> BinopPrecedence; | 
 | 1031 |  | 
 | 1032 | /// GetTokPrecedence - Get the precedence of the pending binary operator token. | 
 | 1033 | static int GetTokPrecedence() { | 
 | 1034 |   if (!isascii(CurTok)) | 
 | 1035 |     return -1; | 
 | 1036 |    | 
 | 1037 |   // Make sure it's a declared binop. | 
 | 1038 |   int TokPrec = BinopPrecedence[CurTok]; | 
 | 1039 |   if (TokPrec <= 0) return -1; | 
 | 1040 |   return TokPrec; | 
 | 1041 | } | 
 | 1042 |  | 
 | 1043 | /// Error* - These are little helper functions for error handling. | 
 | 1044 | ExprAST *Error(const char *Str) { fprintf(stderr, "Error: %s\n", Str);return 0;} | 
 | 1045 | PrototypeAST *ErrorP(const char *Str) { Error(Str); return 0; } | 
 | 1046 | FunctionAST *ErrorF(const char *Str) { Error(Str); return 0; } | 
 | 1047 |  | 
 | 1048 | static ExprAST *ParseExpression(); | 
 | 1049 |  | 
 | 1050 | /// identifierexpr | 
 | 1051 | ///   ::= identifer | 
 | 1052 | ///   ::= identifer '(' expression* ')' | 
 | 1053 | static ExprAST *ParseIdentifierExpr() { | 
 | 1054 |   std::string IdName = IdentifierStr; | 
 | 1055 |    | 
 | 1056 |   getNextToken();  // eat identifer. | 
 | 1057 |    | 
 | 1058 |   if (CurTok != '(') // Simple variable ref. | 
 | 1059 |     return new VariableExprAST(IdName); | 
 | 1060 |    | 
 | 1061 |   // Call. | 
 | 1062 |   getNextToken();  // eat ( | 
 | 1063 |   std::vector<ExprAST*> Args; | 
 | 1064 |   if (CurTok != ')') { | 
 | 1065 |     while (1) { | 
 | 1066 |       ExprAST *Arg = ParseExpression(); | 
 | 1067 |       if (!Arg) return 0; | 
 | 1068 |       Args.push_back(Arg); | 
 | 1069 |        | 
 | 1070 |       if (CurTok == ')') break; | 
 | 1071 |        | 
 | 1072 |       if (CurTok != ',') | 
 | 1073 |         return Error("Expected ')'"); | 
 | 1074 |       getNextToken(); | 
 | 1075 |     } | 
 | 1076 |   } | 
 | 1077 |  | 
 | 1078 |   // Eat the ')'. | 
 | 1079 |   getNextToken(); | 
 | 1080 |    | 
 | 1081 |   return new CallExprAST(IdName, Args); | 
 | 1082 | } | 
 | 1083 |  | 
 | 1084 | /// numberexpr ::= number | 
 | 1085 | static ExprAST *ParseNumberExpr() { | 
 | 1086 |   ExprAST *Result = new NumberExprAST(NumVal); | 
 | 1087 |   getNextToken(); // consume the number | 
 | 1088 |   return Result; | 
 | 1089 | } | 
 | 1090 |  | 
 | 1091 | /// parenexpr ::= '(' expression ')' | 
 | 1092 | static ExprAST *ParseParenExpr() { | 
 | 1093 |   getNextToken();  // eat (. | 
 | 1094 |   ExprAST *V = ParseExpression(); | 
 | 1095 |   if (!V) return 0; | 
 | 1096 |    | 
 | 1097 |   if (CurTok != ')') | 
 | 1098 |     return Error("expected ')'"); | 
 | 1099 |   getNextToken();  // eat ). | 
 | 1100 |   return V; | 
 | 1101 | } | 
 | 1102 |  | 
 | 1103 | /// ifexpr ::= 'if' expression 'then' expression 'else' expression | 
 | 1104 | static ExprAST *ParseIfExpr() { | 
 | 1105 |   getNextToken();  // eat the if. | 
 | 1106 |    | 
 | 1107 |   // condition. | 
 | 1108 |   ExprAST *Cond = ParseExpression(); | 
 | 1109 |   if (!Cond) return 0; | 
 | 1110 |    | 
 | 1111 |   if (CurTok != tok_then) | 
 | 1112 |     return Error("expected then"); | 
 | 1113 |   getNextToken();  // eat the then | 
 | 1114 |    | 
 | 1115 |   ExprAST *Then = ParseExpression(); | 
 | 1116 |   if (Then == 0) return 0; | 
 | 1117 |    | 
 | 1118 |   if (CurTok != tok_else) | 
 | 1119 |     return Error("expected else"); | 
 | 1120 |    | 
 | 1121 |   getNextToken(); | 
 | 1122 |    | 
 | 1123 |   ExprAST *Else = ParseExpression(); | 
 | 1124 |   if (!Else) return 0; | 
 | 1125 |    | 
 | 1126 |   return new IfExprAST(Cond, Then, Else); | 
 | 1127 | } | 
 | 1128 |  | 
 | 1129 | /// forexpr ::= 'for' identifer '=' expr ',' expr (',' expr)? 'in' expression | 
 | 1130 | static ExprAST *ParseForExpr() { | 
 | 1131 |   getNextToken();  // eat the for. | 
 | 1132 |  | 
 | 1133 |   if (CurTok != tok_identifier) | 
 | 1134 |     return Error("expected identifier after for"); | 
 | 1135 |    | 
 | 1136 |   std::string IdName = IdentifierStr; | 
 | 1137 |   getNextToken();  // eat identifer. | 
 | 1138 |    | 
 | 1139 |   if (CurTok != '=') | 
 | 1140 |     return Error("expected '=' after for"); | 
 | 1141 |   getNextToken();  // eat '='. | 
 | 1142 |    | 
 | 1143 |    | 
 | 1144 |   ExprAST *Start = ParseExpression(); | 
 | 1145 |   if (Start == 0) return 0; | 
 | 1146 |   if (CurTok != ',') | 
 | 1147 |     return Error("expected ',' after for start value"); | 
 | 1148 |   getNextToken(); | 
 | 1149 |    | 
 | 1150 |   ExprAST *End = ParseExpression(); | 
 | 1151 |   if (End == 0) return 0; | 
 | 1152 |    | 
 | 1153 |   // The step value is optional. | 
 | 1154 |   ExprAST *Step = 0; | 
 | 1155 |   if (CurTok == ',') { | 
 | 1156 |     getNextToken(); | 
 | 1157 |     Step = ParseExpression(); | 
 | 1158 |     if (Step == 0) return 0; | 
 | 1159 |   } | 
 | 1160 |    | 
 | 1161 |   if (CurTok != tok_in) | 
 | 1162 |     return Error("expected 'in' after for"); | 
 | 1163 |   getNextToken();  // eat 'in'. | 
 | 1164 |    | 
 | 1165 |   ExprAST *Body = ParseExpression(); | 
 | 1166 |   if (Body == 0) return 0; | 
 | 1167 |  | 
 | 1168 |   return new ForExprAST(IdName, Start, End, Step, Body); | 
 | 1169 | } | 
 | 1170 |  | 
 | 1171 |  | 
 | 1172 | /// primary | 
 | 1173 | ///   ::= identifierexpr | 
 | 1174 | ///   ::= numberexpr | 
 | 1175 | ///   ::= parenexpr | 
 | 1176 | ///   ::= ifexpr | 
 | 1177 | ///   ::= forexpr | 
 | 1178 | static ExprAST *ParsePrimary() { | 
 | 1179 |   switch (CurTok) { | 
 | 1180 |   default: return Error("unknown token when expecting an expression"); | 
 | 1181 |   case tok_identifier: return ParseIdentifierExpr(); | 
 | 1182 |   case tok_number:     return ParseNumberExpr(); | 
 | 1183 |   case '(':            return ParseParenExpr(); | 
 | 1184 |   case tok_if:         return ParseIfExpr(); | 
 | 1185 |   case tok_for:        return ParseForExpr(); | 
 | 1186 |   } | 
 | 1187 | } | 
 | 1188 |  | 
 | 1189 | /// unary | 
 | 1190 | ///   ::= primary | 
 | 1191 | ///   ::= '!' unary | 
 | 1192 | static ExprAST *ParseUnary() { | 
 | 1193 |   // If the current token is not an operator, it must be a primary expr. | 
 | 1194 |   if (!isascii(CurTok) || CurTok == '(' || CurTok == ',') | 
 | 1195 |     return ParsePrimary(); | 
 | 1196 |    | 
 | 1197 |   // If this is a unary operator, read it. | 
 | 1198 |   int Opc = CurTok; | 
 | 1199 |   getNextToken(); | 
 | 1200 |   if (ExprAST *Operand = ParseUnary()) | 
 | 1201 |     return new UnaryExprAST(Opc, Operand); | 
 | 1202 |   return 0; | 
 | 1203 | } | 
 | 1204 |  | 
 | 1205 | /// binoprhs | 
 | 1206 | ///   ::= ('+' unary)* | 
 | 1207 | static ExprAST *ParseBinOpRHS(int ExprPrec, ExprAST *LHS) { | 
 | 1208 |   // If this is a binop, find its precedence. | 
 | 1209 |   while (1) { | 
 | 1210 |     int TokPrec = GetTokPrecedence(); | 
 | 1211 |      | 
 | 1212 |     // If this is a binop that binds at least as tightly as the current binop, | 
 | 1213 |     // consume it, otherwise we are done. | 
 | 1214 |     if (TokPrec < ExprPrec) | 
 | 1215 |       return LHS; | 
 | 1216 |      | 
 | 1217 |     // Okay, we know this is a binop. | 
 | 1218 |     int BinOp = CurTok; | 
 | 1219 |     getNextToken();  // eat binop | 
 | 1220 |      | 
 | 1221 |     // Parse the unary expression after the binary operator. | 
 | 1222 |     ExprAST *RHS = ParseUnary(); | 
 | 1223 |     if (!RHS) return 0; | 
 | 1224 |      | 
 | 1225 |     // If BinOp binds less tightly with RHS than the operator after RHS, let | 
 | 1226 |     // the pending operator take RHS as its LHS. | 
 | 1227 |     int NextPrec = GetTokPrecedence(); | 
 | 1228 |     if (TokPrec < NextPrec) { | 
 | 1229 |       RHS = ParseBinOpRHS(TokPrec+1, RHS); | 
 | 1230 |       if (RHS == 0) return 0; | 
 | 1231 |     } | 
 | 1232 |      | 
 | 1233 |     // Merge LHS/RHS. | 
 | 1234 |     LHS = new BinaryExprAST(BinOp, LHS, RHS); | 
 | 1235 |   } | 
 | 1236 | } | 
 | 1237 |  | 
 | 1238 | /// expression | 
 | 1239 | ///   ::= unary binoprhs | 
 | 1240 | /// | 
 | 1241 | static ExprAST *ParseExpression() { | 
 | 1242 |   ExprAST *LHS = ParseUnary(); | 
 | 1243 |   if (!LHS) return 0; | 
 | 1244 |    | 
 | 1245 |   return ParseBinOpRHS(0, LHS); | 
 | 1246 | } | 
 | 1247 |  | 
 | 1248 | /// prototype | 
 | 1249 | ///   ::= id '(' id* ')' | 
 | 1250 | ///   ::= binary LETTER number? (id, id) | 
 | 1251 | ///   ::= unary LETTER (id) | 
 | 1252 | static PrototypeAST *ParsePrototype() { | 
 | 1253 |   std::string FnName; | 
 | 1254 |    | 
 | 1255 |   int Kind = 0;  // 0 = identifier, 1 = unary, 2 = binary. | 
 | 1256 |   unsigned BinaryPrecedence = 30; | 
 | 1257 |    | 
 | 1258 |   switch (CurTok) { | 
 | 1259 |   default: | 
 | 1260 |     return ErrorP("Expected function name in prototype"); | 
 | 1261 |   case tok_identifier: | 
 | 1262 |     FnName = IdentifierStr; | 
 | 1263 |     Kind = 0; | 
 | 1264 |     getNextToken(); | 
 | 1265 |     break; | 
 | 1266 |   case tok_unary: | 
 | 1267 |     getNextToken(); | 
 | 1268 |     if (!isascii(CurTok)) | 
 | 1269 |       return ErrorP("Expected unary operator"); | 
 | 1270 |     FnName = "unary"; | 
 | 1271 |     FnName += (char)CurTok; | 
 | 1272 |     Kind = 1; | 
 | 1273 |     getNextToken(); | 
 | 1274 |     break; | 
 | 1275 |   case tok_binary: | 
 | 1276 |     getNextToken(); | 
 | 1277 |     if (!isascii(CurTok)) | 
 | 1278 |       return ErrorP("Expected binary operator"); | 
 | 1279 |     FnName = "binary"; | 
 | 1280 |     FnName += (char)CurTok; | 
 | 1281 |     Kind = 2; | 
 | 1282 |     getNextToken(); | 
 | 1283 |      | 
 | 1284 |     // Read the precedence if present. | 
 | 1285 |     if (CurTok == tok_number) { | 
 | 1286 |       if (NumVal < 1 || NumVal > 100) | 
 | 1287 |         return ErrorP("Invalid precedecnce: must be 1..100"); | 
 | 1288 |       BinaryPrecedence = (unsigned)NumVal; | 
 | 1289 |       getNextToken(); | 
 | 1290 |     } | 
 | 1291 |     break; | 
 | 1292 |   } | 
 | 1293 |    | 
 | 1294 |   if (CurTok != '(') | 
 | 1295 |     return ErrorP("Expected '(' in prototype"); | 
 | 1296 |    | 
 | 1297 |   std::vector<std::string> ArgNames; | 
 | 1298 |   while (getNextToken() == tok_identifier) | 
 | 1299 |     ArgNames.push_back(IdentifierStr); | 
 | 1300 |   if (CurTok != ')') | 
 | 1301 |     return ErrorP("Expected ')' in prototype"); | 
 | 1302 |    | 
 | 1303 |   // success. | 
 | 1304 |   getNextToken();  // eat ')'. | 
 | 1305 |    | 
 | 1306 |   // Verify right number of names for operator. | 
 | 1307 |   if (Kind && ArgNames.size() != Kind) | 
 | 1308 |     return ErrorP("Invalid number of operands for operator"); | 
 | 1309 |    | 
 | 1310 |   return new PrototypeAST(FnName, ArgNames, Kind != 0, BinaryPrecedence); | 
 | 1311 | } | 
 | 1312 |  | 
 | 1313 | /// definition ::= 'def' prototype expression | 
 | 1314 | static FunctionAST *ParseDefinition() { | 
 | 1315 |   getNextToken();  // eat def. | 
 | 1316 |   PrototypeAST *Proto = ParsePrototype(); | 
 | 1317 |   if (Proto == 0) return 0; | 
 | 1318 |  | 
 | 1319 |   if (ExprAST *E = ParseExpression()) | 
 | 1320 |     return new FunctionAST(Proto, E); | 
 | 1321 |   return 0; | 
 | 1322 | } | 
 | 1323 |  | 
 | 1324 | /// toplevelexpr ::= expression | 
 | 1325 | static FunctionAST *ParseTopLevelExpr() { | 
 | 1326 |   if (ExprAST *E = ParseExpression()) { | 
 | 1327 |     // Make an anonymous proto. | 
 | 1328 |     PrototypeAST *Proto = new PrototypeAST("", std::vector<std::string>()); | 
 | 1329 |     return new FunctionAST(Proto, E); | 
 | 1330 |   } | 
 | 1331 |   return 0; | 
 | 1332 | } | 
 | 1333 |  | 
 | 1334 | /// external ::= 'extern' prototype | 
 | 1335 | static PrototypeAST *ParseExtern() { | 
 | 1336 |   getNextToken();  // eat extern. | 
 | 1337 |   return ParsePrototype(); | 
 | 1338 | } | 
 | 1339 |  | 
 | 1340 | //===----------------------------------------------------------------------===// | 
 | 1341 | // Code Generation | 
 | 1342 | //===----------------------------------------------------------------------===// | 
 | 1343 |  | 
 | 1344 | static Module *TheModule; | 
 | 1345 | static LLVMFoldingBuilder Builder; | 
 | 1346 | static std::map<std::string, Value*> NamedValues; | 
 | 1347 | static FunctionPassManager *TheFPM; | 
 | 1348 |  | 
 | 1349 | Value *ErrorV(const char *Str) { Error(Str); return 0; } | 
 | 1350 |  | 
 | 1351 | Value *NumberExprAST::Codegen() { | 
 | 1352 |   return ConstantFP::get(Type::DoubleTy, APFloat(Val)); | 
 | 1353 | } | 
 | 1354 |  | 
 | 1355 | Value *VariableExprAST::Codegen() { | 
 | 1356 |   // Look this variable up in the function. | 
 | 1357 |   Value *V = NamedValues[Name]; | 
 | 1358 |   return V ? V : ErrorV("Unknown variable name"); | 
 | 1359 | } | 
 | 1360 |  | 
 | 1361 | Value *UnaryExprAST::Codegen() { | 
 | 1362 |   Value *OperandV = Operand->Codegen(); | 
 | 1363 |   if (OperandV == 0) return 0; | 
 | 1364 |    | 
 | 1365 |   Function *F = TheModule->getFunction(std::string("unary")+Opcode); | 
 | 1366 |   if (F == 0) | 
 | 1367 |     return ErrorV("Unknown unary operator"); | 
 | 1368 |    | 
 | 1369 |   return Builder.CreateCall(F, OperandV, "unop"); | 
 | 1370 | } | 
 | 1371 |  | 
 | 1372 |  | 
 | 1373 | Value *BinaryExprAST::Codegen() { | 
 | 1374 |   Value *L = LHS->Codegen(); | 
 | 1375 |   Value *R = RHS->Codegen(); | 
 | 1376 |   if (L == 0 || R == 0) return 0; | 
 | 1377 |    | 
 | 1378 |   switch (Op) { | 
 | 1379 |   case '+': return Builder.CreateAdd(L, R, "addtmp"); | 
 | 1380 |   case '-': return Builder.CreateSub(L, R, "subtmp"); | 
 | 1381 |   case '*': return Builder.CreateMul(L, R, "multmp"); | 
 | 1382 |   case '<': | 
 | 1383 |     L = Builder.CreateFCmpULT(L, R, "multmp"); | 
 | 1384 |     // Convert bool 0/1 to double 0.0 or 1.0 | 
 | 1385 |     return Builder.CreateUIToFP(L, Type::DoubleTy, "booltmp"); | 
 | 1386 |   default: break; | 
 | 1387 |   } | 
 | 1388 |    | 
 | 1389 |   // If it wasn't a builtin binary operator, it must be a user defined one. Emit | 
 | 1390 |   // a call to it. | 
 | 1391 |   Function *F = TheModule->getFunction(std::string("binary")+Op); | 
 | 1392 |   assert(F && "binary operator not found!"); | 
 | 1393 |    | 
 | 1394 |   Value *Ops[] = { L, R }; | 
 | 1395 |   return Builder.CreateCall(F, Ops, Ops+2, "binop"); | 
 | 1396 | } | 
 | 1397 |  | 
 | 1398 | Value *CallExprAST::Codegen() { | 
 | 1399 |   // Look up the name in the global module table. | 
 | 1400 |   Function *CalleeF = TheModule->getFunction(Callee); | 
 | 1401 |   if (CalleeF == 0) | 
 | 1402 |     return ErrorV("Unknown function referenced"); | 
 | 1403 |    | 
 | 1404 |   // If argument mismatch error. | 
 | 1405 |   if (CalleeF->arg_size() != Args.size()) | 
 | 1406 |     return ErrorV("Incorrect # arguments passed"); | 
 | 1407 |  | 
 | 1408 |   std::vector<Value*> ArgsV; | 
 | 1409 |   for (unsigned i = 0, e = Args.size(); i != e; ++i) { | 
 | 1410 |     ArgsV.push_back(Args[i]->Codegen()); | 
 | 1411 |     if (ArgsV.back() == 0) return 0; | 
 | 1412 |   } | 
 | 1413 |    | 
 | 1414 |   return Builder.CreateCall(CalleeF, ArgsV.begin(), ArgsV.end(), "calltmp"); | 
 | 1415 | } | 
 | 1416 |  | 
 | 1417 | Value *IfExprAST::Codegen() { | 
 | 1418 |   Value *CondV = Cond->Codegen(); | 
 | 1419 |   if (CondV == 0) return 0; | 
 | 1420 |    | 
 | 1421 |   // Convert condition to a bool by comparing equal to 0.0. | 
 | 1422 |   CondV = Builder.CreateFCmpONE(CondV,  | 
 | 1423 |                                 ConstantFP::get(Type::DoubleTy, APFloat(0.0)), | 
 | 1424 |                                 "ifcond"); | 
 | 1425 |    | 
 | 1426 |   Function *TheFunction = Builder.GetInsertBlock()->getParent(); | 
 | 1427 |    | 
 | 1428 |   // Create blocks for the then and else cases.  Insert the 'then' block at the | 
 | 1429 |   // end of the function. | 
 | 1430 |   BasicBlock *ThenBB = new BasicBlock("then", TheFunction); | 
 | 1431 |   BasicBlock *ElseBB = new BasicBlock("else"); | 
 | 1432 |   BasicBlock *MergeBB = new BasicBlock("ifcont"); | 
 | 1433 |    | 
 | 1434 |   Builder.CreateCondBr(CondV, ThenBB, ElseBB); | 
 | 1435 |    | 
 | 1436 |   // Emit then value. | 
 | 1437 |   Builder.SetInsertPoint(ThenBB); | 
 | 1438 |    | 
 | 1439 |   Value *ThenV = Then->Codegen(); | 
 | 1440 |   if (ThenV == 0) return 0; | 
 | 1441 |    | 
 | 1442 |   Builder.CreateBr(MergeBB); | 
 | 1443 |   // Codegen of 'Then' can change the current block, update ThenBB for the PHI. | 
 | 1444 |   ThenBB = Builder.GetInsertBlock(); | 
 | 1445 |    | 
 | 1446 |   // Emit else block. | 
 | 1447 |   TheFunction->getBasicBlockList().push_back(ElseBB); | 
 | 1448 |   Builder.SetInsertPoint(ElseBB); | 
 | 1449 |    | 
 | 1450 |   Value *ElseV = Else->Codegen(); | 
 | 1451 |   if (ElseV == 0) return 0; | 
 | 1452 |    | 
 | 1453 |   Builder.CreateBr(MergeBB); | 
 | 1454 |   // Codegen of 'Else' can change the current block, update ElseBB for the PHI. | 
 | 1455 |   ElseBB = Builder.GetInsertBlock(); | 
 | 1456 |    | 
 | 1457 |   // Emit merge block. | 
 | 1458 |   TheFunction->getBasicBlockList().push_back(MergeBB); | 
 | 1459 |   Builder.SetInsertPoint(MergeBB); | 
 | 1460 |   PHINode *PN = Builder.CreatePHI(Type::DoubleTy, "iftmp"); | 
 | 1461 |    | 
 | 1462 |   PN->addIncoming(ThenV, ThenBB); | 
 | 1463 |   PN->addIncoming(ElseV, ElseBB); | 
 | 1464 |   return PN; | 
 | 1465 | } | 
 | 1466 |  | 
 | 1467 | Value *ForExprAST::Codegen() { | 
 | 1468 |   // Output this as: | 
 | 1469 |   //   ... | 
 | 1470 |   //   start = startexpr | 
 | 1471 |   //   goto loop | 
 | 1472 |   // loop:  | 
 | 1473 |   //   variable = phi [start, loopheader], [nextvariable, loopend] | 
 | 1474 |   //   ... | 
 | 1475 |   //   bodyexpr | 
 | 1476 |   //   ... | 
 | 1477 |   // loopend: | 
 | 1478 |   //   step = stepexpr | 
 | 1479 |   //   nextvariable = variable + step | 
 | 1480 |   //   endcond = endexpr | 
 | 1481 |   //   br endcond, loop, endloop | 
 | 1482 |   // outloop: | 
 | 1483 |    | 
 | 1484 |   // Emit the start code first, without 'variable' in scope. | 
 | 1485 |   Value *StartVal = Start->Codegen(); | 
 | 1486 |   if (StartVal == 0) return 0; | 
 | 1487 |    | 
 | 1488 |   // Make the new basic block for the loop header, inserting after current | 
 | 1489 |   // block. | 
 | 1490 |   Function *TheFunction = Builder.GetInsertBlock()->getParent(); | 
 | 1491 |   BasicBlock *PreheaderBB = Builder.GetInsertBlock(); | 
 | 1492 |   BasicBlock *LoopBB = new BasicBlock("loop", TheFunction); | 
 | 1493 |    | 
 | 1494 |   // Insert an explicit fall through from the current block to the LoopBB. | 
 | 1495 |   Builder.CreateBr(LoopBB); | 
 | 1496 |  | 
 | 1497 |   // Start insertion in LoopBB. | 
 | 1498 |   Builder.SetInsertPoint(LoopBB); | 
 | 1499 |    | 
 | 1500 |   // Start the PHI node with an entry for Start. | 
 | 1501 |   PHINode *Variable = Builder.CreatePHI(Type::DoubleTy, VarName.c_str()); | 
 | 1502 |   Variable->addIncoming(StartVal, PreheaderBB); | 
 | 1503 |    | 
 | 1504 |   // Within the loop, the variable is defined equal to the PHI node.  If it | 
 | 1505 |   // shadows an existing variable, we have to restore it, so save it now. | 
 | 1506 |   Value *OldVal = NamedValues[VarName]; | 
 | 1507 |   NamedValues[VarName] = Variable; | 
 | 1508 |    | 
 | 1509 |   // Emit the body of the loop.  This, like any other expr, can change the | 
 | 1510 |   // current BB.  Note that we ignore the value computed by the body, but don't | 
 | 1511 |   // allow an error. | 
 | 1512 |   if (Body->Codegen() == 0) | 
 | 1513 |     return 0; | 
 | 1514 |    | 
 | 1515 |   // Emit the step value. | 
 | 1516 |   Value *StepVal; | 
 | 1517 |   if (Step) { | 
 | 1518 |     StepVal = Step->Codegen(); | 
 | 1519 |     if (StepVal == 0) return 0; | 
 | 1520 |   } else { | 
 | 1521 |     // If not specified, use 1.0. | 
 | 1522 |     StepVal = ConstantFP::get(Type::DoubleTy, APFloat(1.0)); | 
 | 1523 |   } | 
 | 1524 |    | 
 | 1525 |   Value *NextVar = Builder.CreateAdd(Variable, StepVal, "nextvar"); | 
 | 1526 |  | 
 | 1527 |   // Compute the end condition. | 
 | 1528 |   Value *EndCond = End->Codegen(); | 
 | 1529 |   if (EndCond == 0) return EndCond; | 
 | 1530 |    | 
 | 1531 |   // Convert condition to a bool by comparing equal to 0.0. | 
 | 1532 |   EndCond = Builder.CreateFCmpONE(EndCond,  | 
 | 1533 |                                   ConstantFP::get(Type::DoubleTy, APFloat(0.0)), | 
 | 1534 |                                   "loopcond"); | 
 | 1535 |    | 
 | 1536 |   // Create the "after loop" block and insert it. | 
 | 1537 |   BasicBlock *LoopEndBB = Builder.GetInsertBlock(); | 
 | 1538 |   BasicBlock *AfterBB = new BasicBlock("afterloop", TheFunction); | 
 | 1539 |    | 
 | 1540 |   // Insert the conditional branch into the end of LoopEndBB. | 
 | 1541 |   Builder.CreateCondBr(EndCond, LoopBB, AfterBB); | 
 | 1542 |    | 
 | 1543 |   // Any new code will be inserted in AfterBB. | 
 | 1544 |   Builder.SetInsertPoint(AfterBB); | 
 | 1545 |    | 
 | 1546 |   // Add a new entry to the PHI node for the backedge. | 
 | 1547 |   Variable->addIncoming(NextVar, LoopEndBB); | 
 | 1548 |    | 
 | 1549 |   // Restore the unshadowed variable. | 
 | 1550 |   if (OldVal) | 
 | 1551 |     NamedValues[VarName] = OldVal; | 
 | 1552 |   else | 
 | 1553 |     NamedValues.erase(VarName); | 
 | 1554 |  | 
 | 1555 |    | 
 | 1556 |   // for expr always returns 0.0. | 
 | 1557 |   return Constant::getNullValue(Type::DoubleTy); | 
 | 1558 | } | 
 | 1559 |  | 
 | 1560 | Function *PrototypeAST::Codegen() { | 
 | 1561 |   // Make the function type:  double(double,double) etc. | 
 | 1562 |   std::vector<const Type*> Doubles(Args.size(), Type::DoubleTy); | 
 | 1563 |   FunctionType *FT = FunctionType::get(Type::DoubleTy, Doubles, false); | 
 | 1564 |    | 
 | 1565 |   Function *F = new Function(FT, Function::ExternalLinkage, Name, TheModule); | 
 | 1566 |    | 
 | 1567 |   // If F conflicted, there was already something named 'Name'.  If it has a | 
 | 1568 |   // body, don't allow redefinition or reextern. | 
 | 1569 |   if (F->getName() != Name) { | 
 | 1570 |     // Delete the one we just made and get the existing one. | 
 | 1571 |     F->eraseFromParent(); | 
 | 1572 |     F = TheModule->getFunction(Name); | 
 | 1573 |      | 
 | 1574 |     // If F already has a body, reject this. | 
 | 1575 |     if (!F->empty()) { | 
 | 1576 |       ErrorF("redefinition of function"); | 
 | 1577 |       return 0; | 
 | 1578 |     } | 
 | 1579 |      | 
 | 1580 |     // If F took a different number of args, reject. | 
 | 1581 |     if (F->arg_size() != Args.size()) { | 
 | 1582 |       ErrorF("redefinition of function with different # args"); | 
 | 1583 |       return 0; | 
 | 1584 |     } | 
 | 1585 |   } | 
 | 1586 |    | 
 | 1587 |   // Set names for all arguments. | 
 | 1588 |   unsigned Idx = 0; | 
 | 1589 |   for (Function::arg_iterator AI = F->arg_begin(); Idx != Args.size(); | 
 | 1590 |        ++AI, ++Idx) { | 
 | 1591 |     AI->setName(Args[Idx]); | 
 | 1592 |      | 
 | 1593 |     // Add arguments to variable symbol table. | 
 | 1594 |     NamedValues[Args[Idx]] = AI; | 
 | 1595 |   } | 
 | 1596 |    | 
 | 1597 |   return F; | 
 | 1598 | } | 
 | 1599 |  | 
 | 1600 | Function *FunctionAST::Codegen() { | 
 | 1601 |   NamedValues.clear(); | 
 | 1602 |    | 
 | 1603 |   Function *TheFunction = Proto->Codegen(); | 
 | 1604 |   if (TheFunction == 0) | 
 | 1605 |     return 0; | 
 | 1606 |    | 
 | 1607 |   // If this is an operator, install it. | 
 | 1608 |   if (Proto->isBinaryOp()) | 
 | 1609 |     BinopPrecedence[Proto->getOperatorName()] = Proto->getBinaryPrecedence(); | 
 | 1610 |    | 
 | 1611 |   // Create a new basic block to start insertion into. | 
 | 1612 |   BasicBlock *BB = new BasicBlock("entry", TheFunction); | 
 | 1613 |   Builder.SetInsertPoint(BB); | 
 | 1614 |    | 
 | 1615 |   if (Value *RetVal = Body->Codegen()) { | 
 | 1616 |     // Finish off the function. | 
 | 1617 |     Builder.CreateRet(RetVal); | 
 | 1618 |  | 
 | 1619 |     // Validate the generated code, checking for consistency. | 
 | 1620 |     verifyFunction(*TheFunction); | 
 | 1621 |  | 
 | 1622 |     // Optimize the function. | 
 | 1623 |     TheFPM->run(*TheFunction); | 
 | 1624 |      | 
 | 1625 |     return TheFunction; | 
 | 1626 |   } | 
 | 1627 |    | 
 | 1628 |   // Error reading body, remove function. | 
 | 1629 |   TheFunction->eraseFromParent(); | 
 | 1630 |  | 
 | 1631 |   if (Proto->isBinaryOp()) | 
 | 1632 |     BinopPrecedence.erase(Proto->getOperatorName()); | 
 | 1633 |   return 0; | 
 | 1634 | } | 
 | 1635 |  | 
 | 1636 | //===----------------------------------------------------------------------===// | 
 | 1637 | // Top-Level parsing and JIT Driver | 
 | 1638 | //===----------------------------------------------------------------------===// | 
 | 1639 |  | 
 | 1640 | static ExecutionEngine *TheExecutionEngine; | 
 | 1641 |  | 
 | 1642 | static void HandleDefinition() { | 
 | 1643 |   if (FunctionAST *F = ParseDefinition()) { | 
 | 1644 |     if (Function *LF = F->Codegen()) { | 
 | 1645 |       fprintf(stderr, "Read function definition:"); | 
 | 1646 |       LF->dump(); | 
 | 1647 |     } | 
 | 1648 |   } else { | 
 | 1649 |     // Skip token for error recovery. | 
 | 1650 |     getNextToken(); | 
 | 1651 |   } | 
 | 1652 | } | 
 | 1653 |  | 
 | 1654 | static void HandleExtern() { | 
 | 1655 |   if (PrototypeAST *P = ParseExtern()) { | 
 | 1656 |     if (Function *F = P->Codegen()) { | 
 | 1657 |       fprintf(stderr, "Read extern: "); | 
 | 1658 |       F->dump(); | 
 | 1659 |     } | 
 | 1660 |   } else { | 
 | 1661 |     // Skip token for error recovery. | 
 | 1662 |     getNextToken(); | 
 | 1663 |   } | 
 | 1664 | } | 
 | 1665 |  | 
 | 1666 | static void HandleTopLevelExpression() { | 
 | 1667 |   // Evaluate a top level expression into an anonymous function. | 
 | 1668 |   if (FunctionAST *F = ParseTopLevelExpr()) { | 
 | 1669 |     if (Function *LF = F->Codegen()) { | 
 | 1670 |       // JIT the function, returning a function pointer. | 
 | 1671 |       void *FPtr = TheExecutionEngine->getPointerToFunction(LF); | 
 | 1672 |        | 
 | 1673 |       // Cast it to the right type (takes no arguments, returns a double) so we | 
 | 1674 |       // can call it as a native function. | 
 | 1675 |       double (*FP)() = (double (*)())FPtr; | 
 | 1676 |       fprintf(stderr, "Evaluated to %f\n", FP()); | 
 | 1677 |     } | 
 | 1678 |   } else { | 
 | 1679 |     // Skip token for error recovery. | 
 | 1680 |     getNextToken(); | 
 | 1681 |   } | 
 | 1682 | } | 
 | 1683 |  | 
 | 1684 | /// top ::= definition | external | expression | ';' | 
 | 1685 | static void MainLoop() { | 
 | 1686 |   while (1) { | 
 | 1687 |     fprintf(stderr, "ready> "); | 
 | 1688 |     switch (CurTok) { | 
 | 1689 |     case tok_eof:    return; | 
 | 1690 |     case ';':        getNextToken(); break;  // ignore top level semicolons. | 
 | 1691 |     case tok_def:    HandleDefinition(); break; | 
 | 1692 |     case tok_extern: HandleExtern(); break; | 
 | 1693 |     default:         HandleTopLevelExpression(); break; | 
 | 1694 |     } | 
 | 1695 |   } | 
 | 1696 | } | 
 | 1697 |  | 
 | 1698 |  | 
 | 1699 |  | 
 | 1700 | //===----------------------------------------------------------------------===// | 
 | 1701 | // "Library" functions that can be "extern'd" from user code. | 
 | 1702 | //===----------------------------------------------------------------------===// | 
 | 1703 |  | 
 | 1704 | /// putchard - putchar that takes a double and returns 0. | 
 | 1705 | extern "C"  | 
 | 1706 | double putchard(double X) { | 
 | 1707 |   putchar((char)X); | 
 | 1708 |   return 0; | 
 | 1709 | } | 
 | 1710 |  | 
 | 1711 | /// printd - printf that takes a double prints it as "%f\n", returning 0. | 
 | 1712 | extern "C"  | 
 | 1713 | double printd(double X) { | 
 | 1714 |   printf("%f\n", X); | 
 | 1715 |   return 0; | 
 | 1716 | } | 
 | 1717 |  | 
 | 1718 | //===----------------------------------------------------------------------===// | 
 | 1719 | // Main driver code. | 
 | 1720 | //===----------------------------------------------------------------------===// | 
 | 1721 |  | 
 | 1722 | int main() { | 
 | 1723 |   // Install standard binary operators. | 
 | 1724 |   // 1 is lowest precedence. | 
 | 1725 |   BinopPrecedence['<'] = 10; | 
 | 1726 |   BinopPrecedence['+'] = 20; | 
 | 1727 |   BinopPrecedence['-'] = 20; | 
 | 1728 |   BinopPrecedence['*'] = 40;  // highest. | 
 | 1729 |  | 
 | 1730 |   // Prime the first token. | 
 | 1731 |   fprintf(stderr, "ready> "); | 
 | 1732 |   getNextToken(); | 
 | 1733 |  | 
 | 1734 |   // Make the module, which holds all the code. | 
 | 1735 |   TheModule = new Module("my cool jit"); | 
 | 1736 |    | 
 | 1737 |   // Create the JIT. | 
 | 1738 |   TheExecutionEngine = ExecutionEngine::create(TheModule); | 
 | 1739 |  | 
 | 1740 |   { | 
 | 1741 |     ExistingModuleProvider OurModuleProvider(TheModule); | 
 | 1742 |     FunctionPassManager OurFPM(&OurModuleProvider); | 
 | 1743 |        | 
 | 1744 |     // Set up the optimizer pipeline.  Start with registering info about how the | 
 | 1745 |     // target lays out data structures. | 
 | 1746 |     OurFPM.add(new TargetData(*TheExecutionEngine->getTargetData())); | 
 | 1747 |     // Do simple "peephole" optimizations and bit-twiddling optzns. | 
 | 1748 |     OurFPM.add(createInstructionCombiningPass()); | 
 | 1749 |     // Reassociate expressions. | 
 | 1750 |     OurFPM.add(createReassociatePass()); | 
 | 1751 |     // Eliminate Common SubExpressions. | 
 | 1752 |     OurFPM.add(createGVNPass()); | 
 | 1753 |     // Simplify the control flow graph (deleting unreachable blocks, etc). | 
 | 1754 |     OurFPM.add(createCFGSimplificationPass()); | 
 | 1755 |     // Set the global so the code gen can use this. | 
 | 1756 |     TheFPM = &OurFPM; | 
 | 1757 |  | 
 | 1758 |     // Run the main "interpreter loop" now. | 
 | 1759 |     MainLoop(); | 
 | 1760 |      | 
 | 1761 |     TheFPM = 0; | 
 | 1762 |   }  // Free module provider and pass manager. | 
 | 1763 |                                     | 
 | 1764 |                                     | 
 | 1765 |   // Print out all of the generated code. | 
 | 1766 |   TheModule->dump(); | 
 | 1767 |   return 0; | 
 | 1768 | } | 
| Chris Lattner | c9d5d2c | 2007-11-01 06:49:54 +0000 | [diff] [blame] | 1769 | </pre> | 
 | 1770 | </div> | 
 | 1771 |  | 
 | 1772 | </div> | 
 | 1773 |  | 
 | 1774 | <!-- *********************************************************************** --> | 
 | 1775 | <hr> | 
 | 1776 | <address> | 
 | 1777 |   <a href="http://jigsaw.w3.org/css-validator/check/referer"><img | 
 | 1778 |   src="http://jigsaw.w3.org/css-validator/images/vcss" alt="Valid CSS!"></a> | 
 | 1779 |   <a href="http://validator.w3.org/check/referer"><img | 
 | 1780 |   src="http://www.w3.org/Icons/valid-html401" alt="Valid HTML 4.01!"></a> | 
 | 1781 |  | 
 | 1782 |   <a href="mailto:sabre@nondot.org">Chris Lattner</a><br> | 
 | 1783 |   <a href="http://llvm.org">The LLVM Compiler Infrastructure</a><br> | 
 | 1784 |   Last modified: $Date: 2007-10-17 11:05:13 -0700 (Wed, 17 Oct 2007) $ | 
 | 1785 | </address> | 
 | 1786 | </body> | 
 | 1787 | </html> |