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 | |
Chris Lattner | 128eb86 | 2007-11-05 19:06:59 +0000 | [diff] [blame] | 16 | <ul> |
Chris Lattner | 0e555b1 | 2007-11-05 20:04:56 +0000 | [diff] [blame] | 17 | <li><a href="index.html">Up to Tutorial Index</a></li> |
Chris Lattner | 128eb86 | 2007-11-05 19:06:59 +0000 | [diff] [blame] | 18 | <li>Chapter 6 |
| 19 | <ol> |
| 20 | <li><a href="#intro">Chapter 6 Introduction</a></li> |
| 21 | <li><a href="#idea">User-defined Operators: the Idea</a></li> |
| 22 | <li><a href="#binary">User-defined Binary Operators</a></li> |
| 23 | <li><a href="#unary">User-defined Unary Operators</a></li> |
| 24 | <li><a href="#example">Kicking the Tires</a></li> |
| 25 | <li><a href="#code">Full Code Listing</a></li> |
| 26 | </ol> |
| 27 | </li> |
Chris Lattner | 0e555b1 | 2007-11-05 20:04:56 +0000 | [diff] [blame] | 28 | <li><a href="LangImpl7.html">Chapter 7</a>: Extending the Language: Mutable |
| 29 | Variables / SSA Construction</li> |
Chris Lattner | 128eb86 | 2007-11-05 19:06:59 +0000 | [diff] [blame] | 30 | </ul> |
| 31 | |
Chris Lattner | c9d5d2c | 2007-11-01 06:49:54 +0000 | [diff] [blame] | 32 | <div class="doc_author"> |
| 33 | <p>Written by <a href="mailto:sabre@nondot.org">Chris Lattner</a></p> |
| 34 | </div> |
| 35 | |
| 36 | <!-- *********************************************************************** --> |
Chris Lattner | 128eb86 | 2007-11-05 19:06:59 +0000 | [diff] [blame] | 37 | <div class="doc_section"><a name="intro">Chapter 6 Introduction</a></div> |
Chris Lattner | c9d5d2c | 2007-11-01 06:49:54 +0000 | [diff] [blame] | 38 | <!-- *********************************************************************** --> |
| 39 | |
| 40 | <div class="doc_text"> |
| 41 | |
Chris Lattner | 128eb86 | 2007-11-05 19:06:59 +0000 | [diff] [blame] | 42 | <p>Welcome to Chapter 6 of the "<a href="index.html">Implementing a language |
| 43 | with LLVM</a>" tutorial. At this point in our tutorial, we now have a fully |
Chris Lattner | c9d5d2c | 2007-11-01 06:49:54 +0000 | [diff] [blame] | 44 | functional language that is fairly minimal, but also useful. One big problem |
| 45 | with it though is that it doesn't have many useful operators (like division, |
| 46 | logical negation, or even any comparisons other than less-than.</p> |
| 47 | |
Chris Lattner | 58f2c87 | 2007-11-02 05:42:52 +0000 | [diff] [blame] | 48 | <p>This chapter of the tutorial takes a wild digression into adding user-defined |
| 49 | operators to the simple and beautiful Kaleidoscope language, giving us a |
Chris Lattner | c9d5d2c | 2007-11-01 06:49:54 +0000 | [diff] [blame] | 50 | simple and ugly language in some ways, but also a powerful one at the same time. |
| 51 | One of the great things about creating your own language is that you get to |
| 52 | decide what is good or bad. In this tutorial we'll assume that it is okay and |
| 53 | use this as a way to show some interesting parsing techniques.</p> |
| 54 | |
Chris Lattner | 58f2c87 | 2007-11-02 05:42:52 +0000 | [diff] [blame] | 55 | <p>At the end of this tutorial, we'll <a href="#example">run through a nice |
| 56 | little example</a> that shows an example application that you can build with |
| 57 | Kaleidoscope and the feature set it now has.</p> |
| 58 | |
Chris Lattner | c9d5d2c | 2007-11-01 06:49:54 +0000 | [diff] [blame] | 59 | </div> |
| 60 | |
| 61 | <!-- *********************************************************************** --> |
Chris Lattner | 58f2c87 | 2007-11-02 05:42:52 +0000 | [diff] [blame] | 62 | <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] | 63 | <!-- *********************************************************************** --> |
| 64 | |
| 65 | <div class="doc_text"> |
| 66 | |
| 67 | <p> |
Chris Lattner | 58f2c87 | 2007-11-02 05:42:52 +0000 | [diff] [blame] | 68 | 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] | 69 | languages like C++. In C++, you are only allowed to redefine existing |
| 70 | operators: you can't programatically change the grammar, introduce new |
| 71 | operators, change precedence levels, etc. In this chapter, we will add this |
| 72 | capability to Kaleidoscope, which will allow us to round out the set of |
| 73 | operators that are supported, culminating in a more interesting example app.</p> |
| 74 | |
Chris Lattner | 58f2c87 | 2007-11-02 05:42:52 +0000 | [diff] [blame] | 75 | <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] | 76 | show the power and flexibility of using a hand-written parser. The parser we |
| 77 | are using so far is using recursive descent for most parts of the grammar, and |
| 78 | operator precedence parsing for the expressions. See <a |
| 79 | href="LangImpl2.html">Chapter 2</a> for details. Without using operator |
| 80 | precedence parsing, it would be very difficult to allow the programmer to |
| 81 | introduce new operators into the grammar: the grammar is dynamically extensible |
| 82 | as the JIT runs.</p> |
| 83 | |
| 84 | <p>The two specific features we'll add are programmable unary operators (right |
| 85 | now, Kaleidoscope has no unary operators at all) as well as binary operators. |
| 86 | An example of this is:</p> |
| 87 | |
| 88 | <div class="doc_code"> |
| 89 | <pre> |
| 90 | # Logical unary not. |
| 91 | def unary!(v) |
| 92 | if v then |
| 93 | 0 |
| 94 | else |
| 95 | 1; |
| 96 | |
| 97 | # Define > with the same precedence as <. |
| 98 | def binary> 10 (LHS RHS) |
| 99 | !(LHS < RHS); # alternatively, could just use "RHS < LHS" |
| 100 | |
| 101 | # Binary "logical or", (note that it does not "short circuit") |
| 102 | def binary| 5 (LHS RHS) |
| 103 | if LHS then |
| 104 | 1 |
| 105 | else if RHS then |
| 106 | 1 |
| 107 | else |
| 108 | 0; |
| 109 | |
| 110 | # Define = with slightly lower precedence than relationals. |
| 111 | def binary= 9 (LHS RHS) |
| 112 | !(LHS < RHS | LHS > RHS); |
| 113 | </pre> |
| 114 | </div> |
| 115 | |
| 116 | <p>Many languages aspire to being able to implement their standard runtime |
| 117 | library in the language itself. In Kaleidoscope, we can implement significant |
| 118 | parts of the language in the library!</p> |
| 119 | |
| 120 | <p>We will break down implementation of these features into two parts: |
Chris Lattner | 58f2c87 | 2007-11-02 05:42:52 +0000 | [diff] [blame] | 121 | implementing support for user-defined binary operators and adding unary |
Chris Lattner | c9d5d2c | 2007-11-01 06:49:54 +0000 | [diff] [blame] | 122 | operators.</p> |
| 123 | |
| 124 | </div> |
| 125 | |
| 126 | <!-- *********************************************************************** --> |
Chris Lattner | 58f2c87 | 2007-11-02 05:42:52 +0000 | [diff] [blame] | 127 | <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] | 128 | <!-- *********************************************************************** --> |
| 129 | |
| 130 | <div class="doc_text"> |
| 131 | |
Chris Lattner | 58f2c87 | 2007-11-02 05:42:52 +0000 | [diff] [blame] | 132 | <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] | 133 | current framework. We'll first add support for the unary/binary keywords:</p> |
| 134 | |
| 135 | <div class="doc_code"> |
| 136 | <pre> |
| 137 | enum Token { |
| 138 | ... |
| 139 | <b>// operators |
| 140 | tok_binary = -11, tok_unary = -12</b> |
| 141 | }; |
| 142 | ... |
| 143 | static int gettok() { |
| 144 | ... |
| 145 | if (IdentifierStr == "for") return tok_for; |
| 146 | if (IdentifierStr == "in") return tok_in; |
| 147 | <b>if (IdentifierStr == "binary") return tok_binary; |
| 148 | if (IdentifierStr == "unary") return tok_unary;</b> |
| 149 | return tok_identifier; |
| 150 | </pre> |
| 151 | </div> |
| 152 | |
| 153 | <p>This just adds lexer support for the unary and binary keywords, like we |
| 154 | did in <a href="LangImpl5.html#iflexer">previous chapters</a>. One nice thing |
| 155 | about our current AST is that we represent binary operators fully generally |
| 156 | with their ASCII code as the opcode. For our extended operators, we'll use the |
| 157 | same representation, so we don't need any new AST or parser support.</p> |
| 158 | |
| 159 | <p>On the other hand, we have to be able to represent the definitions of these |
| 160 | new operators, in the "def binary| 5" part of the function definition. In the |
| 161 | grammar so far, the "name" for the function definition is parsed as the |
| 162 | "prototype" production and into the <tt>PrototypeAST</tt> AST node. To |
| 163 | represent our new user-defined operators as prototypes, we have to extend |
| 164 | the <tt>PrototypeAST</tt> AST node like this:</p> |
| 165 | |
| 166 | <div class="doc_code"> |
| 167 | <pre> |
| 168 | /// PrototypeAST - This class represents the "prototype" for a function, |
| 169 | /// which captures its argument names as well as if it is an operator. |
| 170 | class PrototypeAST { |
| 171 | std::string Name; |
| 172 | std::vector<std::string> Args; |
| 173 | <b>bool isOperator; |
| 174 | unsigned Precedence; // Precedence if a binary op.</b> |
| 175 | public: |
| 176 | PrototypeAST(const std::string &name, const std::vector<std::string> &args, |
| 177 | <b>bool isoperator = false, unsigned prec = 0</b>) |
| 178 | : Name(name), Args(args), <b>isOperator(isoperator), Precedence(prec)</b> {} |
| 179 | |
| 180 | <b>bool isUnaryOp() const { return isOperator && Args.size() == 1; } |
| 181 | bool isBinaryOp() const { return isOperator && Args.size() == 2; } |
| 182 | |
| 183 | char getOperatorName() const { |
| 184 | assert(isUnaryOp() || isBinaryOp()); |
| 185 | return Name[Name.size()-1]; |
| 186 | } |
| 187 | |
| 188 | unsigned getBinaryPrecedence() const { return Precedence; }</b> |
| 189 | |
| 190 | Function *Codegen(); |
| 191 | }; |
| 192 | </pre> |
| 193 | </div> |
| 194 | |
| 195 | <p>Basically, in addition to knowing a name for the prototype, we now keep track |
| 196 | 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] | 197 | is at. The precedence is only used for binary operators (as you'll see below, |
| 198 | it just doesn't apply for unary operators). Now that we have a way to represent |
| 199 | the prototype for a user-defined operator, we need to parse it:</p> |
| 200 | |
| 201 | <div class="doc_code"> |
| 202 | <pre> |
| 203 | /// prototype |
| 204 | /// ::= id '(' id* ')' |
| 205 | <b>/// ::= binary LETTER number? (id, id)</b> |
| 206 | static PrototypeAST *ParsePrototype() { |
| 207 | std::string FnName; |
| 208 | |
| 209 | <b>int Kind = 0; // 0 = identifier, 1 = unary, 2 = binary. |
| 210 | unsigned BinaryPrecedence = 30;</b> |
| 211 | |
| 212 | switch (CurTok) { |
| 213 | default: |
| 214 | return ErrorP("Expected function name in prototype"); |
| 215 | case tok_identifier: |
| 216 | FnName = IdentifierStr; |
| 217 | Kind = 0; |
| 218 | getNextToken(); |
| 219 | break; |
| 220 | <b>case tok_binary: |
| 221 | getNextToken(); |
| 222 | if (!isascii(CurTok)) |
| 223 | return ErrorP("Expected binary operator"); |
| 224 | FnName = "binary"; |
| 225 | FnName += (char)CurTok; |
| 226 | Kind = 2; |
| 227 | getNextToken(); |
| 228 | |
| 229 | // Read the precedence if present. |
| 230 | if (CurTok == tok_number) { |
| 231 | if (NumVal < 1 || NumVal > 100) |
| 232 | return ErrorP("Invalid precedecnce: must be 1..100"); |
| 233 | BinaryPrecedence = (unsigned)NumVal; |
| 234 | getNextToken(); |
| 235 | } |
| 236 | break;</b> |
| 237 | } |
| 238 | |
| 239 | if (CurTok != '(') |
| 240 | return ErrorP("Expected '(' in prototype"); |
| 241 | |
| 242 | std::vector<std::string> ArgNames; |
| 243 | while (getNextToken() == tok_identifier) |
| 244 | ArgNames.push_back(IdentifierStr); |
| 245 | if (CurTok != ')') |
| 246 | return ErrorP("Expected ')' in prototype"); |
| 247 | |
| 248 | // success. |
| 249 | getNextToken(); // eat ')'. |
| 250 | |
| 251 | <b>// Verify right number of names for operator. |
| 252 | if (Kind && ArgNames.size() != Kind) |
| 253 | return ErrorP("Invalid number of operands for operator"); |
| 254 | |
| 255 | return new PrototypeAST(FnName, ArgNames, Kind != 0, BinaryPrecedence);</b> |
| 256 | } |
| 257 | </pre> |
| 258 | </div> |
| 259 | |
| 260 | <p>This is all fairly straight-forward parsing code, and we have already seen |
| 261 | a lot of similar code in the past. One interesting piece of this is the part |
| 262 | that sets up <tt>FnName</tt> for binary operators. This builds names like |
| 263 | "binary@" for a newly defined "@" operator. This takes advantage of the fact |
| 264 | that symbol names in the LLVM symbol table are allowed to have any character in |
| 265 | them, inluding embedded nul characters.</p> |
| 266 | |
| 267 | <p>The next interesting piece is codegen support for these binary operators. |
| 268 | Given our current structure, this is a simple addition of a default case for our |
| 269 | existing binary operator node:</p> |
| 270 | |
| 271 | <div class="doc_code"> |
| 272 | <pre> |
| 273 | Value *BinaryExprAST::Codegen() { |
| 274 | Value *L = LHS->Codegen(); |
| 275 | Value *R = RHS->Codegen(); |
| 276 | if (L == 0 || R == 0) return 0; |
| 277 | |
| 278 | switch (Op) { |
| 279 | case '+': return Builder.CreateAdd(L, R, "addtmp"); |
| 280 | case '-': return Builder.CreateSub(L, R, "subtmp"); |
| 281 | case '*': return Builder.CreateMul(L, R, "multmp"); |
| 282 | case '<': |
Chris Lattner | 7115521 | 2007-11-06 01:39:12 +0000 | [diff] [blame^] | 283 | L = Builder.CreateFCmpULT(L, R, "cmptmp"); |
Chris Lattner | 58f2c87 | 2007-11-02 05:42:52 +0000 | [diff] [blame] | 284 | // Convert bool 0/1 to double 0.0 or 1.0 |
| 285 | return Builder.CreateUIToFP(L, Type::DoubleTy, "booltmp"); |
| 286 | <b>default: break;</b> |
| 287 | } |
| 288 | |
| 289 | <b>// If it wasn't a builtin binary operator, it must be a user defined one. Emit |
| 290 | // a call to it. |
| 291 | Function *F = TheModule->getFunction(std::string("binary")+Op); |
| 292 | assert(F && "binary operator not found!"); |
| 293 | |
| 294 | Value *Ops[] = { L, R }; |
| 295 | return Builder.CreateCall(F, Ops, Ops+2, "binop");</b> |
| 296 | } |
| 297 | |
| 298 | </pre> |
| 299 | </div> |
| 300 | |
| 301 | <p>As you can see above, the new code is actually really simple. It just does |
| 302 | a lookup for the appropriate operator in the symbol table and generates a |
| 303 | function call to it. Since user-defined operators are just built as normal |
| 304 | functions (because the "prototype" boils down into a function with the right |
| 305 | name) everything falls into place.</p> |
| 306 | |
| 307 | <p>The final missing piece is a bit of top level magic, here:</p> |
| 308 | |
| 309 | <div class="doc_code"> |
| 310 | <pre> |
| 311 | Function *FunctionAST::Codegen() { |
| 312 | NamedValues.clear(); |
| 313 | |
| 314 | Function *TheFunction = Proto->Codegen(); |
| 315 | if (TheFunction == 0) |
| 316 | return 0; |
| 317 | |
| 318 | <b>// If this is an operator, install it. |
| 319 | if (Proto->isBinaryOp()) |
| 320 | BinopPrecedence[Proto->getOperatorName()] = Proto->getBinaryPrecedence();</b> |
| 321 | |
| 322 | // Create a new basic block to start insertion into. |
| 323 | BasicBlock *BB = new BasicBlock("entry", TheFunction); |
| 324 | Builder.SetInsertPoint(BB); |
| 325 | |
| 326 | if (Value *RetVal = Body->Codegen()) { |
| 327 | ... |
| 328 | </pre> |
| 329 | </div> |
| 330 | |
| 331 | <p>Basically, before codegening a function, if it is a user-defined operator, we |
| 332 | register it in the precedence table. This allows the binary operator parsing |
| 333 | logic we already have to handle it. Since it is a fully-general operator |
| 334 | precedence parser, this is all we need to do to "extend the grammar".</p> |
| 335 | |
| 336 | <p>With that, we have useful user-defined binary operators. This builds a lot |
| 337 | on the previous framework we built for other operators. Adding unary operators |
| 338 | is a bit more challenging, because we don't have any framework for it yet, lets |
| 339 | see what it takes.</p> |
| 340 | |
| 341 | </div> |
| 342 | |
| 343 | <!-- *********************************************************************** --> |
| 344 | <div class="doc_section"><a name="unary">User-defined Unary Operators</a></div> |
| 345 | <!-- *********************************************************************** --> |
| 346 | |
| 347 | <div class="doc_text"> |
| 348 | |
| 349 | <p>Since we don't currently support unary operators in the Kaleidoscope |
| 350 | langugage, we'll need to add everything for them. Above, we added simple |
| 351 | support for the 'unary' keyword to the lexer. In addition to that, we need an |
| 352 | AST node:</p> |
| 353 | |
| 354 | <div class="doc_code"> |
| 355 | <pre> |
| 356 | /// UnaryExprAST - Expression class for a unary operator. |
| 357 | class UnaryExprAST : public ExprAST { |
| 358 | char Opcode; |
| 359 | ExprAST *Operand; |
| 360 | public: |
| 361 | UnaryExprAST(char opcode, ExprAST *operand) |
| 362 | : Opcode(opcode), Operand(operand) {} |
| 363 | virtual Value *Codegen(); |
| 364 | }; |
| 365 | </pre> |
| 366 | </div> |
| 367 | |
| 368 | <p>This AST node is very simple and obvious by now. It directly mirrors the |
| 369 | binary operator AST node, except that it only has one child. With this, we |
| 370 | need to add the parsing logic. Parsing a unary operator is pretty simple: we'll |
| 371 | add a new function to do it:</p> |
| 372 | |
| 373 | <div class="doc_code"> |
| 374 | <pre> |
| 375 | /// unary |
| 376 | /// ::= primary |
| 377 | /// ::= '!' unary |
| 378 | static ExprAST *ParseUnary() { |
| 379 | // If the current token is not an operator, it must be a primary expr. |
| 380 | if (!isascii(CurTok) || CurTok == '(' || CurTok == ',') |
| 381 | return ParsePrimary(); |
| 382 | |
| 383 | // If this is a unary operator, read it. |
| 384 | int Opc = CurTok; |
| 385 | getNextToken(); |
| 386 | if (ExprAST *Operand = ParseUnary()) |
| 387 | return new UnaryExprAST(Opc, Operand); |
| 388 | return 0; |
| 389 | } |
| 390 | </pre> |
| 391 | </div> |
| 392 | |
| 393 | <p>The grammar we add is pretty straight-forward here. If we see a unary |
| 394 | operator when parsing a primary operator, we eat the operator as a prefix and |
| 395 | parse the remaining piece as another unary operator. This allows us to handle |
| 396 | multiple unary operators (e.g. "!!x"). Note that unary operators can't have |
| 397 | ambiguous parses like binary operators can, so there is no need for precedence |
| 398 | information.</p> |
| 399 | |
| 400 | <p>The problem with the above is that we need to call ParseUnary from somewhere. |
| 401 | To do this, we change previous callers of ParsePrimary to call ParseUnary |
| 402 | instead:</p> |
| 403 | |
| 404 | <div class="doc_code"> |
| 405 | <pre> |
| 406 | /// binoprhs |
| 407 | /// ::= ('+' unary)* |
| 408 | static ExprAST *ParseBinOpRHS(int ExprPrec, ExprAST *LHS) { |
| 409 | ... |
| 410 | <b>// Parse the unary expression after the binary operator. |
| 411 | ExprAST *RHS = ParseUnary(); |
| 412 | if (!RHS) return 0;</b> |
| 413 | ... |
| 414 | } |
| 415 | /// expression |
| 416 | /// ::= unary binoprhs |
| 417 | /// |
| 418 | static ExprAST *ParseExpression() { |
| 419 | <b>ExprAST *LHS = ParseUnary();</b> |
| 420 | if (!LHS) return 0; |
| 421 | |
| 422 | return ParseBinOpRHS(0, LHS); |
| 423 | } |
| 424 | </pre> |
| 425 | </div> |
| 426 | |
| 427 | <p>With these two simple changes, we now parse unary operators and build the |
| 428 | AST for them. Next up, we need to add parser support for prototypes, to parse |
| 429 | the unary operator prototype. We extend the binary operator code above |
| 430 | with:</p> |
| 431 | |
| 432 | <div class="doc_code"> |
| 433 | <pre> |
| 434 | /// prototype |
| 435 | /// ::= id '(' id* ')' |
| 436 | /// ::= binary LETTER number? (id, id) |
| 437 | <b>/// ::= unary LETTER (id)</b> |
| 438 | static PrototypeAST *ParsePrototype() { |
| 439 | std::string FnName; |
| 440 | |
| 441 | int Kind = 0; // 0 = identifier, 1 = unary, 2 = binary. |
| 442 | unsigned BinaryPrecedence = 30; |
| 443 | |
| 444 | switch (CurTok) { |
| 445 | default: |
| 446 | return ErrorP("Expected function name in prototype"); |
| 447 | case tok_identifier: |
| 448 | FnName = IdentifierStr; |
| 449 | Kind = 0; |
| 450 | getNextToken(); |
| 451 | break; |
| 452 | <b>case tok_unary: |
| 453 | getNextToken(); |
| 454 | if (!isascii(CurTok)) |
| 455 | return ErrorP("Expected unary operator"); |
| 456 | FnName = "unary"; |
| 457 | FnName += (char)CurTok; |
| 458 | Kind = 1; |
| 459 | getNextToken(); |
| 460 | break;</b> |
| 461 | case tok_binary: |
| 462 | ... |
| 463 | </pre> |
| 464 | </div> |
| 465 | |
| 466 | <p>As with binary operators, we name unary operators with a name that includes |
| 467 | the operator character. This assists us at code generation time. Speaking of, |
| 468 | the final piece we need to add is codegen support for unary operators. It looks |
| 469 | like this:</p> |
| 470 | |
| 471 | <div class="doc_code"> |
| 472 | <pre> |
| 473 | Value *UnaryExprAST::Codegen() { |
| 474 | Value *OperandV = Operand->Codegen(); |
| 475 | if (OperandV == 0) return 0; |
| 476 | |
| 477 | Function *F = TheModule->getFunction(std::string("unary")+Opcode); |
| 478 | if (F == 0) |
| 479 | return ErrorV("Unknown unary operator"); |
| 480 | |
| 481 | return Builder.CreateCall(F, OperandV, "unop"); |
| 482 | } |
| 483 | </pre> |
| 484 | </div> |
| 485 | |
| 486 | <p>This code is similar to, but simpler than, the code for binary operators. It |
| 487 | is simpler primarily because it doesn't need to handle any predefined operators. |
| 488 | </p> |
| 489 | |
| 490 | </div> |
| 491 | |
| 492 | <!-- *********************************************************************** --> |
| 493 | <div class="doc_section"><a name="example">Kicking the Tires</a></div> |
| 494 | <!-- *********************************************************************** --> |
| 495 | |
| 496 | <div class="doc_text"> |
| 497 | |
| 498 | <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] | 499 | 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] | 500 | can do a lot of interesting things, including I/O, math, and a bunch of other |
| 501 | things. For example, we can now add a nice sequencing operator (printd is |
| 502 | defined to print out the specified value and a newline):</p> |
| 503 | |
| 504 | <div class="doc_code"> |
| 505 | <pre> |
| 506 | ready> <b>extern printd(x);</b> |
| 507 | Read extern: declare double @printd(double) |
| 508 | ready> <b>def binary : 1 (x y) 0; # Low-precedence operator that ignores operands.</b> |
| 509 | .. |
| 510 | ready> <b>printd(123) : printd(456) : printd(789);</b> |
| 511 | 123.000000 |
| 512 | 456.000000 |
| 513 | 789.000000 |
| 514 | Evaluated to 0.000000 |
| 515 | </pre> |
| 516 | </div> |
| 517 | |
Chris Lattner | b5d81b3 | 2007-11-02 05:54:25 +0000 | [diff] [blame] | 518 | <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] | 519 | |
| 520 | <div class="doc_code"> |
| 521 | <pre> |
| 522 | # Logical unary not. |
| 523 | def unary!(v) |
| 524 | if v then |
| 525 | 0 |
| 526 | else |
| 527 | 1; |
| 528 | |
| 529 | # Unary negate. |
| 530 | def unary-(v) |
| 531 | 0-v; |
| 532 | |
| 533 | # Define > with the same precedence as >. We could also easily define |
| 534 | # <= etc. |
| 535 | def binary> 10 (LHS RHS) |
| 536 | !(LHS < RHS); |
| 537 | |
| 538 | # Binary logical or, which does not short circuit. |
| 539 | def binary| 5 (LHS RHS) |
| 540 | if LHS then |
| 541 | 1 |
| 542 | else if RHS then |
| 543 | 1 |
| 544 | else |
| 545 | 0; |
| 546 | |
| 547 | # Binary logical and, which does not short circuit. |
| 548 | def binary& 6 (LHS RHS) |
| 549 | if !LHS then |
| 550 | 0 |
| 551 | else |
| 552 | !!RHS; |
| 553 | |
| 554 | # Define = with slightly lower precedence than relationals. |
| 555 | def binary = 9 (LHS RHS) |
| 556 | !(LHS < RHS | LHS > RHS); |
| 557 | |
| 558 | </pre> |
| 559 | </div> |
Chris Lattner | c9d5d2c | 2007-11-01 06:49:54 +0000 | [diff] [blame] | 560 | |
| 561 | |
Chris Lattner | 58f2c87 | 2007-11-02 05:42:52 +0000 | [diff] [blame] | 562 | <p>Given the previous if/then/else support, we can also define interesting |
| 563 | functions for I/O. For example, the following prints out a character whose |
| 564 | "density" reflects the value passed in: the lower the value, the denser the |
| 565 | character:</p> |
| 566 | |
| 567 | <div class="doc_code"> |
| 568 | <pre> |
| 569 | ready> |
| 570 | <b> |
| 571 | extern putchard(char) |
| 572 | def printdensity(d) |
| 573 | if d > 8 then |
| 574 | putchard(32) # ' ' |
| 575 | else if d > 4 then |
| 576 | putchard(46) # '.' |
| 577 | else if d > 2 then |
| 578 | putchard(43) # '+' |
| 579 | else |
| 580 | putchard(42); # '*'</b> |
| 581 | ... |
| 582 | ready> <b>printdensity(1): printdensity(2): printdensity(3) : |
| 583 | printdensity(4): printdensity(5): printdensity(9): putchard(10);</b> |
| 584 | *++.. |
| 585 | Evaluated to 0.000000 |
| 586 | </pre> |
| 587 | </div> |
| 588 | |
| 589 | <p>Based on these simple primitive operations, we can start to define more |
| 590 | interesting things. For example, here's a little function that solves for the |
| 591 | number of iterations it takes for a function in the complex plane to |
| 592 | converge:</p> |
| 593 | |
| 594 | <div class="doc_code"> |
| 595 | <pre> |
| 596 | # determine whether the specific location diverges. |
| 597 | # Solve for z = z^2 + c in the complex plane. |
| 598 | def mandleconverger(real imag iters creal cimag) |
| 599 | if iters > 255 | (real*real + imag*imag > 4) then |
| 600 | iters |
| 601 | else |
| 602 | mandleconverger(real*real - imag*imag + creal, |
| 603 | 2*real*imag + cimag, |
| 604 | iters+1, creal, cimag); |
| 605 | |
| 606 | # return the number of iterations required for the iteration to escape |
| 607 | def mandleconverge(real imag) |
| 608 | mandleconverger(real, imag, 0, real, imag); |
| 609 | </pre> |
| 610 | </div> |
| 611 | |
Chris Lattner | b5d81b3 | 2007-11-02 05:54:25 +0000 | [diff] [blame] | 612 | <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] | 613 | for computation of the <a |
| 614 | href="http://en.wikipedia.org/wiki/Mandelbrot_set">Mandelbrot Set</a>. Our |
| 615 | <tt>mandelconverge</tt> function returns the number of iterations that it takes |
| 616 | for a complex orbit to escape, saturating to 255. This is not a very useful |
| 617 | function by itself, but if you plot its value over a two-dimensional plane, |
| 618 | you can see the mandelbrot set. Given that we are limited to using putchard |
| 619 | here, our amazing graphical output is limited, but we can whip together |
| 620 | something using the density plotter above:</p> |
| 621 | |
| 622 | <div class="doc_code"> |
| 623 | <pre> |
| 624 | # compute and plot the mandlebrot set with the specified 2 dimentional range |
| 625 | # info. |
| 626 | def mandelhelp(xmin xmax xstep ymin ymax ystep) |
| 627 | for y = ymin, y < ymax, ystep in ( |
| 628 | (for x = xmin, x < xmax, xstep in |
| 629 | printdensity(mandleconverge(x,y))) |
| 630 | : putchard(10) |
| 631 | ) |
| 632 | |
| 633 | # mandel - This is a convenient helper function for ploting the mandelbrot set |
| 634 | # from the specified position with the specified magnification. |
| 635 | def mandel(realstart imagstart realmag imagmag) |
| 636 | mandelhelp(realstart, realstart+realmag*78, realmag, |
| 637 | imagstart, imagstart+imagmag*40, imagmag); |
| 638 | </pre> |
| 639 | </div> |
| 640 | |
| 641 | <p>Given this, we can try plotting out the mandlebrot set! Lets try it out:</p> |
| 642 | |
| 643 | <div class="doc_code"> |
| 644 | <pre> |
| 645 | ready> <b>mandel(-2.3, -1.3, 0.05, 0.07);</b> |
| 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 | ********++++++++++++............ ...++++++++************** |
| 672 | *********++++++++++++++.......... ...+++++++*************** |
| 673 | **********++++++++++++++++........ .+++++++**************** |
| 674 | **********++++++++++++++++++++.... ... ..+++++++**************** |
| 675 | ***********++++++++++++++++++++++....... .......++++++++***************** |
| 676 | ************+++++++++++++++++++++++...... ......++++++++****************** |
| 677 | **************+++++++++++++++++++++++.... ....++++++++******************** |
| 678 | ***************+++++++++++++++++++++++..... ...+++++++++********************* |
| 679 | *****************++++++++++++++++++++++.... ...++++++++*********************** |
| 680 | *******************+++++++++++++++++++++......++++++++************************* |
| 681 | *********************++++++++++++++++++++++.++++++++*************************** |
| 682 | *************************+++++++++++++++++++++++******************************* |
| 683 | ******************************+++++++++++++************************************ |
| 684 | ******************************************************************************* |
| 685 | ******************************************************************************* |
| 686 | ******************************************************************************* |
| 687 | Evaluated to 0.000000 |
| 688 | ready> <b>mandel(-2, -1, 0.02, 0.04);</b> |
| 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 | * |
| 715 | *.......... ........... |
| 716 | *+++....++++................ |
| 717 | *++++++++++++++.............. |
| 718 | *++++++++++++++++............ ... |
| 719 | *++++++++++++++++++................. |
| 720 | **+++++++++++++++++++................ |
| 721 | **+++++++++++++++++++++.............. |
| 722 | ***++++++++++++++++++++++............ |
| 723 | ***++++++++++++++++++++++++......... ...... ........... |
| 724 | ****++++++++++++++++++++++++++...... ......................... |
| 725 | *****++++++++++++++++++++++++++++............................... |
| 726 | *****++++++++++++++++++++++++++++++++............................ |
| 727 | ******+++++++++++++++++++++++++++++++++++........................... |
| 728 | *******+++++++++++++++++++++++++++++++++++++++....................... |
| 729 | ********+++++++++++++++++++++++++++++++++++++++++++.................. |
| 730 | Evaluated to 0.000000 |
| 731 | ready> <b>mandel(-0.9, -1.4, 0.02, 0.03);</b> |
| 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 | ++............ ...... ....++++++++++ |
| 758 | .............. ...++++++++++ |
| 759 | .............. ....+++++++++ |
| 760 | .............. .....++++++++ |
| 761 | ............. ......++++++++ |
| 762 | ........... .......++++++++ |
| 763 | ......... ........+++++++ |
| 764 | ......... ........+++++++ |
| 765 | ......... ....+++++++ |
| 766 | ........ ...+++++++ |
| 767 | ....... ...+++++++ |
| 768 | ....+++++++ |
| 769 | .....+++++++ |
| 770 | ....+++++++ |
| 771 | ....+++++++ |
| 772 | ....+++++++ |
| 773 | Evaluated to 0.000000 |
| 774 | ready> <b>^D</b> |
| 775 | </pre> |
| 776 | </div> |
| 777 | |
| 778 | <p>At this point, you may be starting to realize that Kaleidoscope is a real |
| 779 | and powerful language. It may not be self-similar :), but it can be used to |
| 780 | plot things that are!</p> |
| 781 | |
| 782 | <p>With this, we conclude the "adding user-defined operators" chapter of the |
| 783 | tutorial. We successfully extended our language with the ability to extend the |
| 784 | language in the library, and showed how this can be used to build a simple but |
| 785 | interesting end user application in Kaleidoscope. At this point, Kaleidoscope |
| 786 | 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] | 787 | 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] | 788 | </p> |
| 789 | |
| 790 | <p>Strikingly, lack of this feature is an important limitation for some |
| 791 | languages, and it is not at all obvious how to <a href="LangImpl7.html">add |
| 792 | support for mutable variables</a> without having to add an "SSA construction" |
| 793 | phase to your front-end. In the next chapter, we will describe how you can |
| 794 | add this without building SSA in your front-end.</p> |
Chris Lattner | c9d5d2c | 2007-11-01 06:49:54 +0000 | [diff] [blame] | 795 | |
| 796 | </div> |
| 797 | |
| 798 | |
| 799 | <!-- *********************************************************************** --> |
| 800 | <div class="doc_section"><a name="code">Full Code Listing</a></div> |
| 801 | <!-- *********************************************************************** --> |
| 802 | |
| 803 | <div class="doc_text"> |
| 804 | |
| 805 | <p> |
| 806 | Here is the complete code listing for our running example, enhanced with the |
| 807 | if/then/else and for expressions.. To build this example, use: |
| 808 | </p> |
| 809 | |
| 810 | <div class="doc_code"> |
| 811 | <pre> |
| 812 | # Compile |
| 813 | g++ -g toy.cpp `llvm-config --cppflags --ldflags --libs core jit native` -O3 -o toy |
| 814 | # Run |
| 815 | ./toy |
| 816 | </pre> |
| 817 | </div> |
| 818 | |
| 819 | <p>Here is the code:</p> |
| 820 | |
| 821 | <div class="doc_code"> |
| 822 | <pre> |
Chris Lattner | 58f2c87 | 2007-11-02 05:42:52 +0000 | [diff] [blame] | 823 | #include "llvm/DerivedTypes.h" |
| 824 | #include "llvm/ExecutionEngine/ExecutionEngine.h" |
| 825 | #include "llvm/Module.h" |
| 826 | #include "llvm/ModuleProvider.h" |
| 827 | #include "llvm/PassManager.h" |
| 828 | #include "llvm/Analysis/Verifier.h" |
| 829 | #include "llvm/Target/TargetData.h" |
| 830 | #include "llvm/Transforms/Scalar.h" |
| 831 | #include "llvm/Support/LLVMBuilder.h" |
| 832 | #include <cstdio> |
| 833 | #include <string> |
| 834 | #include <map> |
| 835 | #include <vector> |
| 836 | using namespace llvm; |
| 837 | |
| 838 | //===----------------------------------------------------------------------===// |
| 839 | // Lexer |
| 840 | //===----------------------------------------------------------------------===// |
| 841 | |
| 842 | // The lexer returns tokens [0-255] if it is an unknown character, otherwise one |
| 843 | // of these for known things. |
| 844 | enum Token { |
| 845 | tok_eof = -1, |
| 846 | |
| 847 | // commands |
| 848 | tok_def = -2, tok_extern = -3, |
| 849 | |
| 850 | // primary |
| 851 | tok_identifier = -4, tok_number = -5, |
| 852 | |
| 853 | // control |
| 854 | tok_if = -6, tok_then = -7, tok_else = -8, |
| 855 | tok_for = -9, tok_in = -10, |
| 856 | |
| 857 | // operators |
| 858 | tok_binary = -11, tok_unary = -12 |
| 859 | }; |
| 860 | |
| 861 | static std::string IdentifierStr; // Filled in if tok_identifier |
| 862 | static double NumVal; // Filled in if tok_number |
| 863 | |
| 864 | /// gettok - Return the next token from standard input. |
| 865 | static int gettok() { |
| 866 | static int LastChar = ' '; |
| 867 | |
| 868 | // Skip any whitespace. |
| 869 | while (isspace(LastChar)) |
| 870 | LastChar = getchar(); |
| 871 | |
| 872 | if (isalpha(LastChar)) { // identifier: [a-zA-Z][a-zA-Z0-9]* |
| 873 | IdentifierStr = LastChar; |
| 874 | while (isalnum((LastChar = getchar()))) |
| 875 | IdentifierStr += LastChar; |
| 876 | |
| 877 | if (IdentifierStr == "def") return tok_def; |
| 878 | if (IdentifierStr == "extern") return tok_extern; |
| 879 | if (IdentifierStr == "if") return tok_if; |
| 880 | if (IdentifierStr == "then") return tok_then; |
| 881 | if (IdentifierStr == "else") return tok_else; |
| 882 | if (IdentifierStr == "for") return tok_for; |
| 883 | if (IdentifierStr == "in") return tok_in; |
| 884 | if (IdentifierStr == "binary") return tok_binary; |
| 885 | if (IdentifierStr == "unary") return tok_unary; |
| 886 | return tok_identifier; |
| 887 | } |
| 888 | |
| 889 | if (isdigit(LastChar) || LastChar == '.') { // Number: [0-9.]+ |
| 890 | std::string NumStr; |
| 891 | do { |
| 892 | NumStr += LastChar; |
| 893 | LastChar = getchar(); |
| 894 | } while (isdigit(LastChar) || LastChar == '.'); |
| 895 | |
| 896 | NumVal = strtod(NumStr.c_str(), 0); |
| 897 | return tok_number; |
| 898 | } |
| 899 | |
| 900 | if (LastChar == '#') { |
| 901 | // Comment until end of line. |
| 902 | do LastChar = getchar(); |
| 903 | while (LastChar != EOF && LastChar != '\n' & LastChar != '\r'); |
| 904 | |
| 905 | if (LastChar != EOF) |
| 906 | return gettok(); |
| 907 | } |
| 908 | |
| 909 | // Check for end of file. Don't eat the EOF. |
| 910 | if (LastChar == EOF) |
| 911 | return tok_eof; |
| 912 | |
| 913 | // Otherwise, just return the character as its ascii value. |
| 914 | int ThisChar = LastChar; |
| 915 | LastChar = getchar(); |
| 916 | return ThisChar; |
| 917 | } |
| 918 | |
| 919 | //===----------------------------------------------------------------------===// |
| 920 | // Abstract Syntax Tree (aka Parse Tree) |
| 921 | //===----------------------------------------------------------------------===// |
| 922 | |
| 923 | /// ExprAST - Base class for all expression nodes. |
| 924 | class ExprAST { |
| 925 | public: |
| 926 | virtual ~ExprAST() {} |
| 927 | virtual Value *Codegen() = 0; |
| 928 | }; |
| 929 | |
| 930 | /// NumberExprAST - Expression class for numeric literals like "1.0". |
| 931 | class NumberExprAST : public ExprAST { |
| 932 | double Val; |
| 933 | public: |
| 934 | NumberExprAST(double val) : Val(val) {} |
| 935 | virtual Value *Codegen(); |
| 936 | }; |
| 937 | |
| 938 | /// VariableExprAST - Expression class for referencing a variable, like "a". |
| 939 | class VariableExprAST : public ExprAST { |
| 940 | std::string Name; |
| 941 | public: |
| 942 | VariableExprAST(const std::string &name) : Name(name) {} |
| 943 | virtual Value *Codegen(); |
| 944 | }; |
| 945 | |
| 946 | /// UnaryExprAST - Expression class for a unary operator. |
| 947 | class UnaryExprAST : public ExprAST { |
| 948 | char Opcode; |
| 949 | ExprAST *Operand; |
| 950 | public: |
| 951 | UnaryExprAST(char opcode, ExprAST *operand) |
| 952 | : Opcode(opcode), Operand(operand) {} |
| 953 | virtual Value *Codegen(); |
| 954 | }; |
| 955 | |
| 956 | /// BinaryExprAST - Expression class for a binary operator. |
| 957 | class BinaryExprAST : public ExprAST { |
| 958 | char Op; |
| 959 | ExprAST *LHS, *RHS; |
| 960 | public: |
| 961 | BinaryExprAST(char op, ExprAST *lhs, ExprAST *rhs) |
| 962 | : Op(op), LHS(lhs), RHS(rhs) {} |
| 963 | virtual Value *Codegen(); |
| 964 | }; |
| 965 | |
| 966 | /// CallExprAST - Expression class for function calls. |
| 967 | class CallExprAST : public ExprAST { |
| 968 | std::string Callee; |
| 969 | std::vector<ExprAST*> Args; |
| 970 | public: |
| 971 | CallExprAST(const std::string &callee, std::vector<ExprAST*> &args) |
| 972 | : Callee(callee), Args(args) {} |
| 973 | virtual Value *Codegen(); |
| 974 | }; |
| 975 | |
| 976 | /// IfExprAST - Expression class for if/then/else. |
| 977 | class IfExprAST : public ExprAST { |
| 978 | ExprAST *Cond, *Then, *Else; |
| 979 | public: |
| 980 | IfExprAST(ExprAST *cond, ExprAST *then, ExprAST *_else) |
| 981 | : Cond(cond), Then(then), Else(_else) {} |
| 982 | virtual Value *Codegen(); |
| 983 | }; |
| 984 | |
| 985 | /// ForExprAST - Expression class for for/in. |
| 986 | class ForExprAST : public ExprAST { |
| 987 | std::string VarName; |
| 988 | ExprAST *Start, *End, *Step, *Body; |
| 989 | public: |
| 990 | ForExprAST(const std::string &varname, ExprAST *start, ExprAST *end, |
| 991 | ExprAST *step, ExprAST *body) |
| 992 | : VarName(varname), Start(start), End(end), Step(step), Body(body) {} |
| 993 | virtual Value *Codegen(); |
| 994 | }; |
| 995 | |
| 996 | /// PrototypeAST - This class represents the "prototype" for a function, |
| 997 | /// which captures its argument names as well as if it is an operator. |
| 998 | class PrototypeAST { |
| 999 | std::string Name; |
| 1000 | std::vector<std::string> Args; |
| 1001 | bool isOperator; |
| 1002 | unsigned Precedence; // Precedence if a binary op. |
| 1003 | public: |
| 1004 | PrototypeAST(const std::string &name, const std::vector<std::string> &args, |
| 1005 | bool isoperator = false, unsigned prec = 0) |
| 1006 | : Name(name), Args(args), isOperator(isoperator), Precedence(prec) {} |
| 1007 | |
| 1008 | bool isUnaryOp() const { return isOperator && Args.size() == 1; } |
| 1009 | bool isBinaryOp() const { return isOperator && Args.size() == 2; } |
| 1010 | |
| 1011 | char getOperatorName() const { |
| 1012 | assert(isUnaryOp() || isBinaryOp()); |
| 1013 | return Name[Name.size()-1]; |
| 1014 | } |
| 1015 | |
| 1016 | unsigned getBinaryPrecedence() const { return Precedence; } |
| 1017 | |
| 1018 | Function *Codegen(); |
| 1019 | }; |
| 1020 | |
| 1021 | /// FunctionAST - This class represents a function definition itself. |
| 1022 | class FunctionAST { |
| 1023 | PrototypeAST *Proto; |
| 1024 | ExprAST *Body; |
| 1025 | public: |
| 1026 | FunctionAST(PrototypeAST *proto, ExprAST *body) |
| 1027 | : Proto(proto), Body(body) {} |
| 1028 | |
| 1029 | Function *Codegen(); |
| 1030 | }; |
| 1031 | |
| 1032 | //===----------------------------------------------------------------------===// |
| 1033 | // Parser |
| 1034 | //===----------------------------------------------------------------------===// |
| 1035 | |
| 1036 | /// CurTok/getNextToken - Provide a simple token buffer. CurTok is the current |
| 1037 | /// token the parser it looking at. getNextToken reads another token from the |
| 1038 | /// lexer and updates CurTok with its results. |
| 1039 | static int CurTok; |
| 1040 | static int getNextToken() { |
| 1041 | return CurTok = gettok(); |
| 1042 | } |
| 1043 | |
| 1044 | /// BinopPrecedence - This holds the precedence for each binary operator that is |
| 1045 | /// defined. |
| 1046 | static std::map<char, int> BinopPrecedence; |
| 1047 | |
| 1048 | /// GetTokPrecedence - Get the precedence of the pending binary operator token. |
| 1049 | static int GetTokPrecedence() { |
| 1050 | if (!isascii(CurTok)) |
| 1051 | return -1; |
| 1052 | |
| 1053 | // Make sure it's a declared binop. |
| 1054 | int TokPrec = BinopPrecedence[CurTok]; |
| 1055 | if (TokPrec <= 0) return -1; |
| 1056 | return TokPrec; |
| 1057 | } |
| 1058 | |
| 1059 | /// Error* - These are little helper functions for error handling. |
| 1060 | ExprAST *Error(const char *Str) { fprintf(stderr, "Error: %s\n", Str);return 0;} |
| 1061 | PrototypeAST *ErrorP(const char *Str) { Error(Str); return 0; } |
| 1062 | FunctionAST *ErrorF(const char *Str) { Error(Str); return 0; } |
| 1063 | |
| 1064 | static ExprAST *ParseExpression(); |
| 1065 | |
| 1066 | /// identifierexpr |
Chris Lattner | 20a0c80 | 2007-11-05 17:54:34 +0000 | [diff] [blame] | 1067 | /// ::= identifier |
| 1068 | /// ::= identifier '(' expression* ')' |
Chris Lattner | 58f2c87 | 2007-11-02 05:42:52 +0000 | [diff] [blame] | 1069 | static ExprAST *ParseIdentifierExpr() { |
| 1070 | std::string IdName = IdentifierStr; |
| 1071 | |
Chris Lattner | 20a0c80 | 2007-11-05 17:54:34 +0000 | [diff] [blame] | 1072 | getNextToken(); // eat identifier. |
Chris Lattner | 58f2c87 | 2007-11-02 05:42:52 +0000 | [diff] [blame] | 1073 | |
| 1074 | if (CurTok != '(') // Simple variable ref. |
| 1075 | return new VariableExprAST(IdName); |
| 1076 | |
| 1077 | // Call. |
| 1078 | getNextToken(); // eat ( |
| 1079 | std::vector<ExprAST*> Args; |
| 1080 | if (CurTok != ')') { |
| 1081 | while (1) { |
| 1082 | ExprAST *Arg = ParseExpression(); |
| 1083 | if (!Arg) return 0; |
| 1084 | Args.push_back(Arg); |
| 1085 | |
| 1086 | if (CurTok == ')') break; |
| 1087 | |
| 1088 | if (CurTok != ',') |
| 1089 | return Error("Expected ')'"); |
| 1090 | getNextToken(); |
| 1091 | } |
| 1092 | } |
| 1093 | |
| 1094 | // Eat the ')'. |
| 1095 | getNextToken(); |
| 1096 | |
| 1097 | return new CallExprAST(IdName, Args); |
| 1098 | } |
| 1099 | |
| 1100 | /// numberexpr ::= number |
| 1101 | static ExprAST *ParseNumberExpr() { |
| 1102 | ExprAST *Result = new NumberExprAST(NumVal); |
| 1103 | getNextToken(); // consume the number |
| 1104 | return Result; |
| 1105 | } |
| 1106 | |
| 1107 | /// parenexpr ::= '(' expression ')' |
| 1108 | static ExprAST *ParseParenExpr() { |
| 1109 | getNextToken(); // eat (. |
| 1110 | ExprAST *V = ParseExpression(); |
| 1111 | if (!V) return 0; |
| 1112 | |
| 1113 | if (CurTok != ')') |
| 1114 | return Error("expected ')'"); |
| 1115 | getNextToken(); // eat ). |
| 1116 | return V; |
| 1117 | } |
| 1118 | |
| 1119 | /// ifexpr ::= 'if' expression 'then' expression 'else' expression |
| 1120 | static ExprAST *ParseIfExpr() { |
| 1121 | getNextToken(); // eat the if. |
| 1122 | |
| 1123 | // condition. |
| 1124 | ExprAST *Cond = ParseExpression(); |
| 1125 | if (!Cond) return 0; |
| 1126 | |
| 1127 | if (CurTok != tok_then) |
| 1128 | return Error("expected then"); |
| 1129 | getNextToken(); // eat the then |
| 1130 | |
| 1131 | ExprAST *Then = ParseExpression(); |
| 1132 | if (Then == 0) return 0; |
| 1133 | |
| 1134 | if (CurTok != tok_else) |
| 1135 | return Error("expected else"); |
| 1136 | |
| 1137 | getNextToken(); |
| 1138 | |
| 1139 | ExprAST *Else = ParseExpression(); |
| 1140 | if (!Else) return 0; |
| 1141 | |
| 1142 | return new IfExprAST(Cond, Then, Else); |
| 1143 | } |
| 1144 | |
Chris Lattner | 20a0c80 | 2007-11-05 17:54:34 +0000 | [diff] [blame] | 1145 | /// forexpr ::= 'for' identifier '=' expr ',' expr (',' expr)? 'in' expression |
Chris Lattner | 58f2c87 | 2007-11-02 05:42:52 +0000 | [diff] [blame] | 1146 | static ExprAST *ParseForExpr() { |
| 1147 | getNextToken(); // eat the for. |
| 1148 | |
| 1149 | if (CurTok != tok_identifier) |
| 1150 | return Error("expected identifier after for"); |
| 1151 | |
| 1152 | std::string IdName = IdentifierStr; |
Chris Lattner | 20a0c80 | 2007-11-05 17:54:34 +0000 | [diff] [blame] | 1153 | getNextToken(); // eat identifier. |
Chris Lattner | 58f2c87 | 2007-11-02 05:42:52 +0000 | [diff] [blame] | 1154 | |
| 1155 | if (CurTok != '=') |
| 1156 | return Error("expected '=' after for"); |
| 1157 | getNextToken(); // eat '='. |
| 1158 | |
| 1159 | |
| 1160 | ExprAST *Start = ParseExpression(); |
| 1161 | if (Start == 0) return 0; |
| 1162 | if (CurTok != ',') |
| 1163 | return Error("expected ',' after for start value"); |
| 1164 | getNextToken(); |
| 1165 | |
| 1166 | ExprAST *End = ParseExpression(); |
| 1167 | if (End == 0) return 0; |
| 1168 | |
| 1169 | // The step value is optional. |
| 1170 | ExprAST *Step = 0; |
| 1171 | if (CurTok == ',') { |
| 1172 | getNextToken(); |
| 1173 | Step = ParseExpression(); |
| 1174 | if (Step == 0) return 0; |
| 1175 | } |
| 1176 | |
| 1177 | if (CurTok != tok_in) |
| 1178 | return Error("expected 'in' after for"); |
| 1179 | getNextToken(); // eat 'in'. |
| 1180 | |
| 1181 | ExprAST *Body = ParseExpression(); |
| 1182 | if (Body == 0) return 0; |
| 1183 | |
| 1184 | return new ForExprAST(IdName, Start, End, Step, Body); |
| 1185 | } |
| 1186 | |
| 1187 | |
| 1188 | /// primary |
| 1189 | /// ::= identifierexpr |
| 1190 | /// ::= numberexpr |
| 1191 | /// ::= parenexpr |
| 1192 | /// ::= ifexpr |
| 1193 | /// ::= forexpr |
| 1194 | static ExprAST *ParsePrimary() { |
| 1195 | switch (CurTok) { |
| 1196 | default: return Error("unknown token when expecting an expression"); |
| 1197 | case tok_identifier: return ParseIdentifierExpr(); |
| 1198 | case tok_number: return ParseNumberExpr(); |
| 1199 | case '(': return ParseParenExpr(); |
| 1200 | case tok_if: return ParseIfExpr(); |
| 1201 | case tok_for: return ParseForExpr(); |
| 1202 | } |
| 1203 | } |
| 1204 | |
| 1205 | /// unary |
| 1206 | /// ::= primary |
| 1207 | /// ::= '!' unary |
| 1208 | static ExprAST *ParseUnary() { |
| 1209 | // If the current token is not an operator, it must be a primary expr. |
| 1210 | if (!isascii(CurTok) || CurTok == '(' || CurTok == ',') |
| 1211 | return ParsePrimary(); |
| 1212 | |
| 1213 | // If this is a unary operator, read it. |
| 1214 | int Opc = CurTok; |
| 1215 | getNextToken(); |
| 1216 | if (ExprAST *Operand = ParseUnary()) |
| 1217 | return new UnaryExprAST(Opc, Operand); |
| 1218 | return 0; |
| 1219 | } |
| 1220 | |
| 1221 | /// binoprhs |
| 1222 | /// ::= ('+' unary)* |
| 1223 | static ExprAST *ParseBinOpRHS(int ExprPrec, ExprAST *LHS) { |
| 1224 | // If this is a binop, find its precedence. |
| 1225 | while (1) { |
| 1226 | int TokPrec = GetTokPrecedence(); |
| 1227 | |
| 1228 | // If this is a binop that binds at least as tightly as the current binop, |
| 1229 | // consume it, otherwise we are done. |
| 1230 | if (TokPrec < ExprPrec) |
| 1231 | return LHS; |
| 1232 | |
| 1233 | // Okay, we know this is a binop. |
| 1234 | int BinOp = CurTok; |
| 1235 | getNextToken(); // eat binop |
| 1236 | |
| 1237 | // Parse the unary expression after the binary operator. |
| 1238 | ExprAST *RHS = ParseUnary(); |
| 1239 | if (!RHS) return 0; |
| 1240 | |
| 1241 | // If BinOp binds less tightly with RHS than the operator after RHS, let |
| 1242 | // the pending operator take RHS as its LHS. |
| 1243 | int NextPrec = GetTokPrecedence(); |
| 1244 | if (TokPrec < NextPrec) { |
| 1245 | RHS = ParseBinOpRHS(TokPrec+1, RHS); |
| 1246 | if (RHS == 0) return 0; |
| 1247 | } |
| 1248 | |
| 1249 | // Merge LHS/RHS. |
| 1250 | LHS = new BinaryExprAST(BinOp, LHS, RHS); |
| 1251 | } |
| 1252 | } |
| 1253 | |
| 1254 | /// expression |
| 1255 | /// ::= unary binoprhs |
| 1256 | /// |
| 1257 | static ExprAST *ParseExpression() { |
| 1258 | ExprAST *LHS = ParseUnary(); |
| 1259 | if (!LHS) return 0; |
| 1260 | |
| 1261 | return ParseBinOpRHS(0, LHS); |
| 1262 | } |
| 1263 | |
| 1264 | /// prototype |
| 1265 | /// ::= id '(' id* ')' |
| 1266 | /// ::= binary LETTER number? (id, id) |
| 1267 | /// ::= unary LETTER (id) |
| 1268 | static PrototypeAST *ParsePrototype() { |
| 1269 | std::string FnName; |
| 1270 | |
| 1271 | int Kind = 0; // 0 = identifier, 1 = unary, 2 = binary. |
| 1272 | unsigned BinaryPrecedence = 30; |
| 1273 | |
| 1274 | switch (CurTok) { |
| 1275 | default: |
| 1276 | return ErrorP("Expected function name in prototype"); |
| 1277 | case tok_identifier: |
| 1278 | FnName = IdentifierStr; |
| 1279 | Kind = 0; |
| 1280 | getNextToken(); |
| 1281 | break; |
| 1282 | case tok_unary: |
| 1283 | getNextToken(); |
| 1284 | if (!isascii(CurTok)) |
| 1285 | return ErrorP("Expected unary operator"); |
| 1286 | FnName = "unary"; |
| 1287 | FnName += (char)CurTok; |
| 1288 | Kind = 1; |
| 1289 | getNextToken(); |
| 1290 | break; |
| 1291 | case tok_binary: |
| 1292 | getNextToken(); |
| 1293 | if (!isascii(CurTok)) |
| 1294 | return ErrorP("Expected binary operator"); |
| 1295 | FnName = "binary"; |
| 1296 | FnName += (char)CurTok; |
| 1297 | Kind = 2; |
| 1298 | getNextToken(); |
| 1299 | |
| 1300 | // Read the precedence if present. |
| 1301 | if (CurTok == tok_number) { |
| 1302 | if (NumVal < 1 || NumVal > 100) |
| 1303 | return ErrorP("Invalid precedecnce: must be 1..100"); |
| 1304 | BinaryPrecedence = (unsigned)NumVal; |
| 1305 | getNextToken(); |
| 1306 | } |
| 1307 | break; |
| 1308 | } |
| 1309 | |
| 1310 | if (CurTok != '(') |
| 1311 | return ErrorP("Expected '(' in prototype"); |
| 1312 | |
| 1313 | std::vector<std::string> ArgNames; |
| 1314 | while (getNextToken() == tok_identifier) |
| 1315 | ArgNames.push_back(IdentifierStr); |
| 1316 | if (CurTok != ')') |
| 1317 | return ErrorP("Expected ')' in prototype"); |
| 1318 | |
| 1319 | // success. |
| 1320 | getNextToken(); // eat ')'. |
| 1321 | |
| 1322 | // Verify right number of names for operator. |
| 1323 | if (Kind && ArgNames.size() != Kind) |
| 1324 | return ErrorP("Invalid number of operands for operator"); |
| 1325 | |
| 1326 | return new PrototypeAST(FnName, ArgNames, Kind != 0, BinaryPrecedence); |
| 1327 | } |
| 1328 | |
| 1329 | /// definition ::= 'def' prototype expression |
| 1330 | static FunctionAST *ParseDefinition() { |
| 1331 | getNextToken(); // eat def. |
| 1332 | PrototypeAST *Proto = ParsePrototype(); |
| 1333 | if (Proto == 0) return 0; |
| 1334 | |
| 1335 | if (ExprAST *E = ParseExpression()) |
| 1336 | return new FunctionAST(Proto, E); |
| 1337 | return 0; |
| 1338 | } |
| 1339 | |
| 1340 | /// toplevelexpr ::= expression |
| 1341 | static FunctionAST *ParseTopLevelExpr() { |
| 1342 | if (ExprAST *E = ParseExpression()) { |
| 1343 | // Make an anonymous proto. |
| 1344 | PrototypeAST *Proto = new PrototypeAST("", std::vector<std::string>()); |
| 1345 | return new FunctionAST(Proto, E); |
| 1346 | } |
| 1347 | return 0; |
| 1348 | } |
| 1349 | |
| 1350 | /// external ::= 'extern' prototype |
| 1351 | static PrototypeAST *ParseExtern() { |
| 1352 | getNextToken(); // eat extern. |
| 1353 | return ParsePrototype(); |
| 1354 | } |
| 1355 | |
| 1356 | //===----------------------------------------------------------------------===// |
| 1357 | // Code Generation |
| 1358 | //===----------------------------------------------------------------------===// |
| 1359 | |
| 1360 | static Module *TheModule; |
| 1361 | static LLVMFoldingBuilder Builder; |
| 1362 | static std::map<std::string, Value*> NamedValues; |
| 1363 | static FunctionPassManager *TheFPM; |
| 1364 | |
| 1365 | Value *ErrorV(const char *Str) { Error(Str); return 0; } |
| 1366 | |
| 1367 | Value *NumberExprAST::Codegen() { |
| 1368 | return ConstantFP::get(Type::DoubleTy, APFloat(Val)); |
| 1369 | } |
| 1370 | |
| 1371 | Value *VariableExprAST::Codegen() { |
| 1372 | // Look this variable up in the function. |
| 1373 | Value *V = NamedValues[Name]; |
| 1374 | return V ? V : ErrorV("Unknown variable name"); |
| 1375 | } |
| 1376 | |
| 1377 | Value *UnaryExprAST::Codegen() { |
| 1378 | Value *OperandV = Operand->Codegen(); |
| 1379 | if (OperandV == 0) return 0; |
| 1380 | |
| 1381 | Function *F = TheModule->getFunction(std::string("unary")+Opcode); |
| 1382 | if (F == 0) |
| 1383 | return ErrorV("Unknown unary operator"); |
| 1384 | |
| 1385 | return Builder.CreateCall(F, OperandV, "unop"); |
| 1386 | } |
| 1387 | |
| 1388 | |
| 1389 | Value *BinaryExprAST::Codegen() { |
| 1390 | Value *L = LHS->Codegen(); |
| 1391 | Value *R = RHS->Codegen(); |
| 1392 | if (L == 0 || R == 0) return 0; |
| 1393 | |
| 1394 | switch (Op) { |
| 1395 | case '+': return Builder.CreateAdd(L, R, "addtmp"); |
| 1396 | case '-': return Builder.CreateSub(L, R, "subtmp"); |
| 1397 | case '*': return Builder.CreateMul(L, R, "multmp"); |
| 1398 | case '<': |
Chris Lattner | 7115521 | 2007-11-06 01:39:12 +0000 | [diff] [blame^] | 1399 | L = Builder.CreateFCmpULT(L, R, "cmptmp"); |
Chris Lattner | 58f2c87 | 2007-11-02 05:42:52 +0000 | [diff] [blame] | 1400 | // Convert bool 0/1 to double 0.0 or 1.0 |
| 1401 | return Builder.CreateUIToFP(L, Type::DoubleTy, "booltmp"); |
| 1402 | default: break; |
| 1403 | } |
| 1404 | |
| 1405 | // If it wasn't a builtin binary operator, it must be a user defined one. Emit |
| 1406 | // a call to it. |
| 1407 | Function *F = TheModule->getFunction(std::string("binary")+Op); |
| 1408 | assert(F && "binary operator not found!"); |
| 1409 | |
| 1410 | Value *Ops[] = { L, R }; |
| 1411 | return Builder.CreateCall(F, Ops, Ops+2, "binop"); |
| 1412 | } |
| 1413 | |
| 1414 | Value *CallExprAST::Codegen() { |
| 1415 | // Look up the name in the global module table. |
| 1416 | Function *CalleeF = TheModule->getFunction(Callee); |
| 1417 | if (CalleeF == 0) |
| 1418 | return ErrorV("Unknown function referenced"); |
| 1419 | |
| 1420 | // If argument mismatch error. |
| 1421 | if (CalleeF->arg_size() != Args.size()) |
| 1422 | return ErrorV("Incorrect # arguments passed"); |
| 1423 | |
| 1424 | std::vector<Value*> ArgsV; |
| 1425 | for (unsigned i = 0, e = Args.size(); i != e; ++i) { |
| 1426 | ArgsV.push_back(Args[i]->Codegen()); |
| 1427 | if (ArgsV.back() == 0) return 0; |
| 1428 | } |
| 1429 | |
| 1430 | return Builder.CreateCall(CalleeF, ArgsV.begin(), ArgsV.end(), "calltmp"); |
| 1431 | } |
| 1432 | |
| 1433 | Value *IfExprAST::Codegen() { |
| 1434 | Value *CondV = Cond->Codegen(); |
| 1435 | if (CondV == 0) return 0; |
| 1436 | |
| 1437 | // Convert condition to a bool by comparing equal to 0.0. |
| 1438 | CondV = Builder.CreateFCmpONE(CondV, |
| 1439 | ConstantFP::get(Type::DoubleTy, APFloat(0.0)), |
| 1440 | "ifcond"); |
| 1441 | |
| 1442 | Function *TheFunction = Builder.GetInsertBlock()->getParent(); |
| 1443 | |
| 1444 | // Create blocks for the then and else cases. Insert the 'then' block at the |
| 1445 | // end of the function. |
| 1446 | BasicBlock *ThenBB = new BasicBlock("then", TheFunction); |
| 1447 | BasicBlock *ElseBB = new BasicBlock("else"); |
| 1448 | BasicBlock *MergeBB = new BasicBlock("ifcont"); |
| 1449 | |
| 1450 | Builder.CreateCondBr(CondV, ThenBB, ElseBB); |
| 1451 | |
| 1452 | // Emit then value. |
| 1453 | Builder.SetInsertPoint(ThenBB); |
| 1454 | |
| 1455 | Value *ThenV = Then->Codegen(); |
| 1456 | if (ThenV == 0) return 0; |
| 1457 | |
| 1458 | Builder.CreateBr(MergeBB); |
| 1459 | // Codegen of 'Then' can change the current block, update ThenBB for the PHI. |
| 1460 | ThenBB = Builder.GetInsertBlock(); |
| 1461 | |
| 1462 | // Emit else block. |
| 1463 | TheFunction->getBasicBlockList().push_back(ElseBB); |
| 1464 | Builder.SetInsertPoint(ElseBB); |
| 1465 | |
| 1466 | Value *ElseV = Else->Codegen(); |
| 1467 | if (ElseV == 0) return 0; |
| 1468 | |
| 1469 | Builder.CreateBr(MergeBB); |
| 1470 | // Codegen of 'Else' can change the current block, update ElseBB for the PHI. |
| 1471 | ElseBB = Builder.GetInsertBlock(); |
| 1472 | |
| 1473 | // Emit merge block. |
| 1474 | TheFunction->getBasicBlockList().push_back(MergeBB); |
| 1475 | Builder.SetInsertPoint(MergeBB); |
| 1476 | PHINode *PN = Builder.CreatePHI(Type::DoubleTy, "iftmp"); |
| 1477 | |
| 1478 | PN->addIncoming(ThenV, ThenBB); |
| 1479 | PN->addIncoming(ElseV, ElseBB); |
| 1480 | return PN; |
| 1481 | } |
| 1482 | |
| 1483 | Value *ForExprAST::Codegen() { |
| 1484 | // Output this as: |
| 1485 | // ... |
| 1486 | // start = startexpr |
| 1487 | // goto loop |
| 1488 | // loop: |
| 1489 | // variable = phi [start, loopheader], [nextvariable, loopend] |
| 1490 | // ... |
| 1491 | // bodyexpr |
| 1492 | // ... |
| 1493 | // loopend: |
| 1494 | // step = stepexpr |
| 1495 | // nextvariable = variable + step |
| 1496 | // endcond = endexpr |
| 1497 | // br endcond, loop, endloop |
| 1498 | // outloop: |
| 1499 | |
| 1500 | // Emit the start code first, without 'variable' in scope. |
| 1501 | Value *StartVal = Start->Codegen(); |
| 1502 | if (StartVal == 0) return 0; |
| 1503 | |
| 1504 | // Make the new basic block for the loop header, inserting after current |
| 1505 | // block. |
| 1506 | Function *TheFunction = Builder.GetInsertBlock()->getParent(); |
| 1507 | BasicBlock *PreheaderBB = Builder.GetInsertBlock(); |
| 1508 | BasicBlock *LoopBB = new BasicBlock("loop", TheFunction); |
| 1509 | |
| 1510 | // Insert an explicit fall through from the current block to the LoopBB. |
| 1511 | Builder.CreateBr(LoopBB); |
| 1512 | |
| 1513 | // Start insertion in LoopBB. |
| 1514 | Builder.SetInsertPoint(LoopBB); |
| 1515 | |
| 1516 | // Start the PHI node with an entry for Start. |
| 1517 | PHINode *Variable = Builder.CreatePHI(Type::DoubleTy, VarName.c_str()); |
| 1518 | Variable->addIncoming(StartVal, PreheaderBB); |
| 1519 | |
| 1520 | // Within the loop, the variable is defined equal to the PHI node. If it |
| 1521 | // shadows an existing variable, we have to restore it, so save it now. |
| 1522 | Value *OldVal = NamedValues[VarName]; |
| 1523 | NamedValues[VarName] = Variable; |
| 1524 | |
| 1525 | // Emit the body of the loop. This, like any other expr, can change the |
| 1526 | // current BB. Note that we ignore the value computed by the body, but don't |
| 1527 | // allow an error. |
| 1528 | if (Body->Codegen() == 0) |
| 1529 | return 0; |
| 1530 | |
| 1531 | // Emit the step value. |
| 1532 | Value *StepVal; |
| 1533 | if (Step) { |
| 1534 | StepVal = Step->Codegen(); |
| 1535 | if (StepVal == 0) return 0; |
| 1536 | } else { |
| 1537 | // If not specified, use 1.0. |
| 1538 | StepVal = ConstantFP::get(Type::DoubleTy, APFloat(1.0)); |
| 1539 | } |
| 1540 | |
| 1541 | Value *NextVar = Builder.CreateAdd(Variable, StepVal, "nextvar"); |
| 1542 | |
| 1543 | // Compute the end condition. |
| 1544 | Value *EndCond = End->Codegen(); |
| 1545 | if (EndCond == 0) return EndCond; |
| 1546 | |
| 1547 | // Convert condition to a bool by comparing equal to 0.0. |
| 1548 | EndCond = Builder.CreateFCmpONE(EndCond, |
| 1549 | ConstantFP::get(Type::DoubleTy, APFloat(0.0)), |
| 1550 | "loopcond"); |
| 1551 | |
| 1552 | // Create the "after loop" block and insert it. |
| 1553 | BasicBlock *LoopEndBB = Builder.GetInsertBlock(); |
| 1554 | BasicBlock *AfterBB = new BasicBlock("afterloop", TheFunction); |
| 1555 | |
| 1556 | // Insert the conditional branch into the end of LoopEndBB. |
| 1557 | Builder.CreateCondBr(EndCond, LoopBB, AfterBB); |
| 1558 | |
| 1559 | // Any new code will be inserted in AfterBB. |
| 1560 | Builder.SetInsertPoint(AfterBB); |
| 1561 | |
| 1562 | // Add a new entry to the PHI node for the backedge. |
| 1563 | Variable->addIncoming(NextVar, LoopEndBB); |
| 1564 | |
| 1565 | // Restore the unshadowed variable. |
| 1566 | if (OldVal) |
| 1567 | NamedValues[VarName] = OldVal; |
| 1568 | else |
| 1569 | NamedValues.erase(VarName); |
| 1570 | |
| 1571 | |
| 1572 | // for expr always returns 0.0. |
| 1573 | return Constant::getNullValue(Type::DoubleTy); |
| 1574 | } |
| 1575 | |
| 1576 | Function *PrototypeAST::Codegen() { |
| 1577 | // Make the function type: double(double,double) etc. |
| 1578 | std::vector<const Type*> Doubles(Args.size(), Type::DoubleTy); |
| 1579 | FunctionType *FT = FunctionType::get(Type::DoubleTy, Doubles, false); |
| 1580 | |
| 1581 | Function *F = new Function(FT, Function::ExternalLinkage, Name, TheModule); |
| 1582 | |
| 1583 | // If F conflicted, there was already something named 'Name'. If it has a |
| 1584 | // body, don't allow redefinition or reextern. |
| 1585 | if (F->getName() != Name) { |
| 1586 | // Delete the one we just made and get the existing one. |
| 1587 | F->eraseFromParent(); |
| 1588 | F = TheModule->getFunction(Name); |
| 1589 | |
| 1590 | // If F already has a body, reject this. |
| 1591 | if (!F->empty()) { |
| 1592 | ErrorF("redefinition of function"); |
| 1593 | return 0; |
| 1594 | } |
| 1595 | |
| 1596 | // If F took a different number of args, reject. |
| 1597 | if (F->arg_size() != Args.size()) { |
| 1598 | ErrorF("redefinition of function with different # args"); |
| 1599 | return 0; |
| 1600 | } |
| 1601 | } |
| 1602 | |
| 1603 | // Set names for all arguments. |
| 1604 | unsigned Idx = 0; |
| 1605 | for (Function::arg_iterator AI = F->arg_begin(); Idx != Args.size(); |
| 1606 | ++AI, ++Idx) { |
| 1607 | AI->setName(Args[Idx]); |
| 1608 | |
| 1609 | // Add arguments to variable symbol table. |
| 1610 | NamedValues[Args[Idx]] = AI; |
| 1611 | } |
| 1612 | |
| 1613 | return F; |
| 1614 | } |
| 1615 | |
| 1616 | Function *FunctionAST::Codegen() { |
| 1617 | NamedValues.clear(); |
| 1618 | |
| 1619 | Function *TheFunction = Proto->Codegen(); |
| 1620 | if (TheFunction == 0) |
| 1621 | return 0; |
| 1622 | |
| 1623 | // If this is an operator, install it. |
| 1624 | if (Proto->isBinaryOp()) |
| 1625 | BinopPrecedence[Proto->getOperatorName()] = Proto->getBinaryPrecedence(); |
| 1626 | |
| 1627 | // Create a new basic block to start insertion into. |
| 1628 | BasicBlock *BB = new BasicBlock("entry", TheFunction); |
| 1629 | Builder.SetInsertPoint(BB); |
| 1630 | |
| 1631 | if (Value *RetVal = Body->Codegen()) { |
| 1632 | // Finish off the function. |
| 1633 | Builder.CreateRet(RetVal); |
| 1634 | |
| 1635 | // Validate the generated code, checking for consistency. |
| 1636 | verifyFunction(*TheFunction); |
| 1637 | |
| 1638 | // Optimize the function. |
| 1639 | TheFPM->run(*TheFunction); |
| 1640 | |
| 1641 | return TheFunction; |
| 1642 | } |
| 1643 | |
| 1644 | // Error reading body, remove function. |
| 1645 | TheFunction->eraseFromParent(); |
| 1646 | |
| 1647 | if (Proto->isBinaryOp()) |
| 1648 | BinopPrecedence.erase(Proto->getOperatorName()); |
| 1649 | return 0; |
| 1650 | } |
| 1651 | |
| 1652 | //===----------------------------------------------------------------------===// |
| 1653 | // Top-Level parsing and JIT Driver |
| 1654 | //===----------------------------------------------------------------------===// |
| 1655 | |
| 1656 | static ExecutionEngine *TheExecutionEngine; |
| 1657 | |
| 1658 | static void HandleDefinition() { |
| 1659 | if (FunctionAST *F = ParseDefinition()) { |
| 1660 | if (Function *LF = F->Codegen()) { |
| 1661 | fprintf(stderr, "Read function definition:"); |
| 1662 | LF->dump(); |
| 1663 | } |
| 1664 | } else { |
| 1665 | // Skip token for error recovery. |
| 1666 | getNextToken(); |
| 1667 | } |
| 1668 | } |
| 1669 | |
| 1670 | static void HandleExtern() { |
| 1671 | if (PrototypeAST *P = ParseExtern()) { |
| 1672 | if (Function *F = P->Codegen()) { |
| 1673 | fprintf(stderr, "Read extern: "); |
| 1674 | F->dump(); |
| 1675 | } |
| 1676 | } else { |
| 1677 | // Skip token for error recovery. |
| 1678 | getNextToken(); |
| 1679 | } |
| 1680 | } |
| 1681 | |
| 1682 | static void HandleTopLevelExpression() { |
| 1683 | // Evaluate a top level expression into an anonymous function. |
| 1684 | if (FunctionAST *F = ParseTopLevelExpr()) { |
| 1685 | if (Function *LF = F->Codegen()) { |
| 1686 | // JIT the function, returning a function pointer. |
| 1687 | void *FPtr = TheExecutionEngine->getPointerToFunction(LF); |
| 1688 | |
| 1689 | // Cast it to the right type (takes no arguments, returns a double) so we |
| 1690 | // can call it as a native function. |
| 1691 | double (*FP)() = (double (*)())FPtr; |
| 1692 | fprintf(stderr, "Evaluated to %f\n", FP()); |
| 1693 | } |
| 1694 | } else { |
| 1695 | // Skip token for error recovery. |
| 1696 | getNextToken(); |
| 1697 | } |
| 1698 | } |
| 1699 | |
| 1700 | /// top ::= definition | external | expression | ';' |
| 1701 | static void MainLoop() { |
| 1702 | while (1) { |
| 1703 | fprintf(stderr, "ready> "); |
| 1704 | switch (CurTok) { |
| 1705 | case tok_eof: return; |
| 1706 | case ';': getNextToken(); break; // ignore top level semicolons. |
| 1707 | case tok_def: HandleDefinition(); break; |
| 1708 | case tok_extern: HandleExtern(); break; |
| 1709 | default: HandleTopLevelExpression(); break; |
| 1710 | } |
| 1711 | } |
| 1712 | } |
| 1713 | |
| 1714 | |
| 1715 | |
| 1716 | //===----------------------------------------------------------------------===// |
| 1717 | // "Library" functions that can be "extern'd" from user code. |
| 1718 | //===----------------------------------------------------------------------===// |
| 1719 | |
| 1720 | /// putchard - putchar that takes a double and returns 0. |
| 1721 | extern "C" |
| 1722 | double putchard(double X) { |
| 1723 | putchar((char)X); |
| 1724 | return 0; |
| 1725 | } |
| 1726 | |
| 1727 | /// printd - printf that takes a double prints it as "%f\n", returning 0. |
| 1728 | extern "C" |
| 1729 | double printd(double X) { |
| 1730 | printf("%f\n", X); |
| 1731 | return 0; |
| 1732 | } |
| 1733 | |
| 1734 | //===----------------------------------------------------------------------===// |
| 1735 | // Main driver code. |
| 1736 | //===----------------------------------------------------------------------===// |
| 1737 | |
| 1738 | int main() { |
| 1739 | // Install standard binary operators. |
| 1740 | // 1 is lowest precedence. |
| 1741 | BinopPrecedence['<'] = 10; |
| 1742 | BinopPrecedence['+'] = 20; |
| 1743 | BinopPrecedence['-'] = 20; |
| 1744 | BinopPrecedence['*'] = 40; // highest. |
| 1745 | |
| 1746 | // Prime the first token. |
| 1747 | fprintf(stderr, "ready> "); |
| 1748 | getNextToken(); |
| 1749 | |
| 1750 | // Make the module, which holds all the code. |
| 1751 | TheModule = new Module("my cool jit"); |
| 1752 | |
| 1753 | // Create the JIT. |
| 1754 | TheExecutionEngine = ExecutionEngine::create(TheModule); |
| 1755 | |
| 1756 | { |
| 1757 | ExistingModuleProvider OurModuleProvider(TheModule); |
| 1758 | FunctionPassManager OurFPM(&OurModuleProvider); |
| 1759 | |
| 1760 | // Set up the optimizer pipeline. Start with registering info about how the |
| 1761 | // target lays out data structures. |
| 1762 | OurFPM.add(new TargetData(*TheExecutionEngine->getTargetData())); |
| 1763 | // Do simple "peephole" optimizations and bit-twiddling optzns. |
| 1764 | OurFPM.add(createInstructionCombiningPass()); |
| 1765 | // Reassociate expressions. |
| 1766 | OurFPM.add(createReassociatePass()); |
| 1767 | // Eliminate Common SubExpressions. |
| 1768 | OurFPM.add(createGVNPass()); |
| 1769 | // Simplify the control flow graph (deleting unreachable blocks, etc). |
| 1770 | OurFPM.add(createCFGSimplificationPass()); |
| 1771 | // Set the global so the code gen can use this. |
| 1772 | TheFPM = &OurFPM; |
| 1773 | |
| 1774 | // Run the main "interpreter loop" now. |
| 1775 | MainLoop(); |
| 1776 | |
| 1777 | TheFPM = 0; |
| 1778 | } // Free module provider and pass manager. |
| 1779 | |
| 1780 | |
| 1781 | // Print out all of the generated code. |
| 1782 | TheModule->dump(); |
| 1783 | return 0; |
| 1784 | } |
Chris Lattner | c9d5d2c | 2007-11-01 06:49:54 +0000 | [diff] [blame] | 1785 | </pre> |
| 1786 | </div> |
| 1787 | |
| 1788 | </div> |
| 1789 | |
| 1790 | <!-- *********************************************************************** --> |
| 1791 | <hr> |
| 1792 | <address> |
| 1793 | <a href="http://jigsaw.w3.org/css-validator/check/referer"><img |
| 1794 | src="http://jigsaw.w3.org/css-validator/images/vcss" alt="Valid CSS!"></a> |
| 1795 | <a href="http://validator.w3.org/check/referer"><img |
| 1796 | src="http://www.w3.org/Icons/valid-html401" alt="Valid HTML 4.01!"></a> |
| 1797 | |
| 1798 | <a href="mailto:sabre@nondot.org">Chris Lattner</a><br> |
| 1799 | <a href="http://llvm.org">The LLVM Compiler Infrastructure</a><br> |
| 1800 | Last modified: $Date: 2007-10-17 11:05:13 -0700 (Wed, 17 Oct 2007) $ |
| 1801 | </address> |
| 1802 | </body> |
| 1803 | </html> |