1 <!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01//EN"
2 "http://www.w3.org/TR/html4/strict.dtd">
6 <title>Kaleidoscope: Extending the Language: Mutable Variables / SSA
8 <meta http-equiv="Content-Type" content="text/html; charset=utf-8">
9 <meta name="author" content="Chris Lattner">
10 <link rel="stylesheet" href="../llvm.css" type="text/css">
15 <div class="doc_title">Kaleidoscope: Extending the Language: Mutable Variables</div>
18 <li><a href="index.html">Up to Tutorial Index</a></li>
21 <li><a href="#intro">Chapter 7 Introduction</a></li>
22 <li><a href="#why">Why is this a hard problem?</a></li>
23 <li><a href="#memory">Memory in LLVM</a></li>
24 <li><a href="#kalvars">Mutable Variables in Kaleidoscope</a></li>
25 <li><a href="#adjustments">Adjusting Existing Variables for
27 <li><a href="#assignment">New Assignment Operator</a></li>
28 <li><a href="#localvars">User-defined Local Variables</a></li>
29 <li><a href="#code">Full Code Listing</a></li>
32 <li><a href="LangImpl8.html">Chapter 8</a>: Conclusion and other useful LLVM
36 <div class="doc_author">
37 <p>Written by <a href="mailto:sabre@nondot.org">Chris Lattner</a></p>
40 <!-- *********************************************************************** -->
41 <div class="doc_section"><a name="intro">Chapter 7 Introduction</a></div>
42 <!-- *********************************************************************** -->
44 <div class="doc_text">
46 <p>Welcome to Chapter 7 of the "<a href="index.html">Implementing a language
47 with LLVM</a>" tutorial. In chapters 1 through 6, we've built a very
48 respectable, albeit simple, <a
49 href="http://en.wikipedia.org/wiki/Functional_programming">functional
50 programming language</a>. In our journey, we learned some parsing techniques,
51 how to build and represent an AST, how to build LLVM IR, and how to optimize
52 the resultant code and JIT compile it.</p>
54 <p>While Kaleidoscope is interesting as a functional language, this makes it
55 "too easy" to generate LLVM IR for it. In particular, a functional language
56 makes it very easy to build LLVM IR directly in <a
57 href="http://en.wikipedia.org/wiki/Static_single_assignment_form">SSA form</a>.
58 Since LLVM requires that the input code be in SSA form, this is a very nice
59 property and it is often unclear to newcomers how to generate code for an
60 imperative language with mutable variables.</p>
62 <p>The short (and happy) summary of this chapter is that there is no need for
63 your front-end to build SSA form: LLVM provides highly tuned and well tested
64 support for this, though the way it works is a bit unexpected for some.</p>
68 <!-- *********************************************************************** -->
69 <div class="doc_section"><a name="why">Why is this a hard problem?</a></div>
70 <!-- *********************************************************************** -->
72 <div class="doc_text">
75 To understand why mutable variables cause complexities in SSA construction,
76 consider this extremely simple C example:
79 <div class="doc_code">
82 int test(_Bool Condition) {
93 <p>In this case, we have the variable "X", whose value depends on the path
94 executed in the program. Because there are two different possible values for X
95 before the return instruction, a PHI node is inserted to merge the two values.
96 The LLVM IR that we want for this example looks like this:</p>
98 <div class="doc_code">
100 @G = weak global i32 0 ; type of @G is i32*
101 @H = weak global i32 0 ; type of @H is i32*
103 define i32 @test(i1 %Condition) {
105 br i1 %Condition, label %cond_true, label %cond_false
116 %X.2 = phi i32 [ %X.1, %cond_false ], [ %X.0, %cond_true ]
122 <p>In this example, the loads from the G and H global variables are explicit in
123 the LLVM IR, and they live in the then/else branches of the if statement
124 (cond_true/cond_false). In order to merge the incoming values, the X.2 phi node
125 in the cond_next block selects the right value to use based on where control
126 flow is coming from: if control flow comes from the cond_false block, X.2 gets
127 the value of X.1. Alternatively, if control flow comes from cond_tree, it gets
128 the value of X.0. The intent of this chapter is not to explain the details of
129 SSA form. For more information, see one of the many <a
130 href="http://en.wikipedia.org/wiki/Static_single_assignment_form">online
133 <p>The question for this article is "who places phi nodes when lowering
134 assignments to mutable variables?". The issue here is that LLVM
135 <em>requires</em> that its IR be in SSA form: there is no "non-ssa" mode for it.
136 However, SSA construction requires non-trivial algorithms and data structures,
137 so it is inconvenient and wasteful for every front-end to have to reproduce this
142 <!-- *********************************************************************** -->
143 <div class="doc_section"><a name="memory">Memory in LLVM</a></div>
144 <!-- *********************************************************************** -->
146 <div class="doc_text">
148 <p>The 'trick' here is that while LLVM does require all register values to be
149 in SSA form, it does not require (or permit) memory objects to be in SSA form.
150 In the example above, note that the loads from G and H are direct accesses to
151 G and H: they are not renamed or versioned. This differs from some other
152 compiler systems, which do try to version memory objects. In LLVM, instead of
153 encoding dataflow analysis of memory into the LLVM IR, it is handled with <a
154 href="../WritingAnLLVMPass.html">Analysis Passes</a> which are computed on
158 With this in mind, the high-level idea is that we want to make a stack variable
159 (which lives in memory, because it is on the stack) for each mutable object in
160 a function. To take advantage of this trick, we need to talk about how LLVM
161 represents stack variables.
164 <p>In LLVM, all memory accesses are explicit with load/store instructions, and
165 it is carefully designed to not have (or need) an "address-of" operator. Notice
166 how the type of the @G/@H global variables is actually "i32*" even though the
167 variable is defined as "i32". What this means is that @G defines <em>space</em>
168 for an i32 in the global data area, but its <em>name</em> actually refers to the
169 address for that space. Stack variables work the same way, but instead of being
170 declared with global variable definitions, they are declared with the
171 <a href="../LangRef.html#i_alloca">LLVM alloca instruction</a>:</p>
173 <div class="doc_code">
175 define i32 @test(i1 %Condition) {
177 %X = alloca i32 ; type of %X is i32*.
179 %tmp = load i32* %X ; load the stack value %X from the stack.
180 %tmp2 = add i32 %tmp, 1 ; increment it
181 store i32 %tmp2, i32* %X ; store it back
186 <p>This code shows an example of how you can declare and manipulate a stack
187 variable in the LLVM IR. Stack memory allocated with the alloca instruction is
188 fully general: you can pass the address of the stack slot to functions, you can
189 store it in other variables, etc. In our example above, we could rewrite the
190 example to use the alloca technique to avoid using a PHI node:</p>
192 <div class="doc_code">
194 @G = weak global i32 0 ; type of @G is i32*
195 @H = weak global i32 0 ; type of @H is i32*
197 define i32 @test(i1 %Condition) {
199 %X = alloca i32 ; type of %X is i32*.
200 br i1 %Condition, label %cond_true, label %cond_false
204 store i32 %X.0, i32* %X ; Update X
209 store i32 %X.1, i32* %X ; Update X
213 %X.2 = load i32* %X ; Read X
219 <p>With this, we have discovered a way to handle arbitrary mutable variables
220 without the need to create Phi nodes at all:</p>
223 <li>Each mutable variable becomes a stack allocation.</li>
224 <li>Each read of the variable becomes a load from the stack.</li>
225 <li>Each update of the variable becomes a store to the stack.</li>
226 <li>Taking the address of a variable just uses the stack address directly.</li>
229 <p>While this solution has solved our immediate problem, it introduced another
230 one: we have now apparently introduced a lot of stack traffic for very simple
231 and common operations, a major performance problem. Fortunately for us, the
232 LLVM optimizer has a highly-tuned optimization pass named "mem2reg" that handles
233 this case, promoting allocas like this into SSA registers, inserting Phi nodes
234 as appropriate. If you run this example through the pass, for example, you'll
237 <div class="doc_code">
239 $ <b>llvm-as < example.ll | opt -mem2reg | llvm-dis</b>
240 @G = weak global i32 0
241 @H = weak global i32 0
243 define i32 @test(i1 %Condition) {
245 br i1 %Condition, label %cond_true, label %cond_false
256 %X.01 = phi i32 [ %X.1, %cond_false ], [ %X.0, %cond_true ]
262 <p>The mem2reg pass implements the standard "iterated dominator frontier"
263 algorithm for constructing SSA form and has a number of optimizations that speed
264 up very common degenerate cases. mem2reg really is the answer for dealing with
265 mutable variables, and we highly recommend that you depend on it. Note that
266 mem2reg only works on variables in certain circumstances:</p>
269 <li>mem2reg is alloca-driven: it looks for allocas and if it can handle them, it
270 promotes them. It does not apply to global variables or heap allocations.</li>
272 <li>mem2reg only looks for alloca instructions in the entry block of the
273 function. Being in the entry block guarantees that the alloca is only executed
274 once, which makes analysis simpler.</li>
276 <li>mem2reg only promotes allocas whose uses are direct loads and stores. If
277 the address of the stack object is passed to a function, or if any funny pointer
278 arithmetic is involved, the alloca will not be promoted.</li>
280 <li>mem2reg only works on allocas of <a
281 href="../LangRef.html#t_classifications">first class</a>
282 values (such as pointers, scalars and vectors), and only if the array size
283 of the allocation is 1 (or missing in the .ll file). mem2reg is not capable of
284 promoting structs or arrays to registers. Note that the "scalarrepl" pass is
285 more powerful and can promote structs, "unions", and arrays in many cases.</li>
290 All of these properties are easy to satisfy for most imperative languages, and
291 we'll illustrate this below with Kaleidoscope. The final question you may be
292 asking is: should I bother with this nonsense for my front-end? Wouldn't it be
293 better if I just did SSA construction directly, avoiding use of the mem2reg
294 optimization pass? In short, we strongly recommend that use you this technique
295 for building SSA form, unless there is an extremely good reason not to. Using
296 this technique is:</p>
299 <li>Proven and well tested: llvm-gcc and clang both use this technique for local
300 mutable variables. As such, the most common clients of LLVM are using this to
301 handle a bulk of their variables. You can be sure that bugs are found fast and
304 <li>Extremely Fast: mem2reg has a number of special cases that make it fast in
305 common cases as well as fully general. For example, it has fast-paths for
306 variables that are only used in a single block, variables that only have one
307 assignment point, good heuristics to avoid insertion of unneeded phi nodes, etc.
310 <li>Needed for debug info generation: <a href="../SourceLevelDebugging.html">
311 Debug information in LLVM</a> relies on having the address of the variable
312 exposed to attach debug info to it. This technique dovetails very naturally
313 with this style of debug info.</li>
316 <p>If nothing else, this makes it much easier to get your front-end up and
317 running, and is very simple to implement. Lets extend Kaleidoscope with mutable
323 <!-- *********************************************************************** -->
324 <div class="doc_section"><a name="kalvars">Mutable Variables in
325 Kaleidoscope</a></div>
326 <!-- *********************************************************************** -->
328 <div class="doc_text">
330 <p>Now that we know the sort of problem we want to tackle, lets see what this
331 looks like in the context of our little Kaleidoscope language. We're going to
332 add two features:</p>
335 <li>The ability to mutate variables with the '=' operator.</li>
336 <li>The ability to define new variables.</li>
339 <p>While the first item is really what this is about, we only have variables
340 for incoming arguments and for induction variables, and redefining them only
341 goes so far :). Also, the ability to define new variables is a
342 useful thing regardless of whether you will be mutating them. Here's a
343 motivating example that shows how we could use these:</p>
345 <div class="doc_code">
347 # Define ':' for sequencing: as a low-precedence operator that ignores operands
348 # and just returns the RHS.
349 def binary : 1 (x y) y;
351 # Recursive fib, we could do this before.
360 <b>var a = 1, b = 1, c in</b>
361 (for i = 3, i &;t; x in
373 In order to mutate variables, we have to change our existing variables to use
374 the "alloca trick". Once we have that, we'll add our new operator, then extend
375 Kaleidoscope to support new variable definitions.
380 <!-- *********************************************************************** -->
381 <div class="doc_section"><a name="adjustments">Adjusting Existing Variables for
383 <!-- *********************************************************************** -->
385 <div class="doc_text">
388 The symbol table in Kaleidoscope is managed at code generation time by the
389 '<tt>NamedValues</tt>' map. This map currently keeps track of the LLVM "Value*"
390 that holds the double value for the named variable. In order to support
391 mutation, we need to change this slightly, so that it <tt>NamedValues</tt> holds
392 the <em>memory location</em> of the variable in question. Note that this
393 change is a refactoring: it changes the structure of the code, but does not
394 (by itself) change the behavior of the compiler. All of these changes are
395 isolated in the Kaleidoscope code generator.</p>
398 At this point in Kaleidoscope's development, it only supports variables for two
399 things: incoming arguments to functions and the induction variable of 'for'
400 loops. For consistency, we'll allow mutation of these variables in addition to
401 other user-defined variables. This means that these will both need memory
405 <p>To start our transformation of Kaleidoscope, we'll change the NamedValues
406 map to map to AllocaInst* instead of Value*. Once we do this, the C++ compiler
407 will tell use what parts of the code we need to update:</p>
409 <div class="doc_code">
411 static std::map<std::string, AllocaInst*> NamedValues;
415 <p>Also, since we will need to create these alloca's, we'll use a helper
416 function that ensures that the allocas are created in the entry block of the
419 <div class="doc_code">
421 /// CreateEntryBlockAlloca - Create an alloca instruction in the entry block of
422 /// the function. This is used for mutable variables etc.
423 static AllocaInst *CreateEntryBlockAlloca(Function *TheFunction,
424 const std::string &VarName) {
425 LLVMBuilder TmpB(&TheFunction->getEntryBlock(),
426 TheFunction->getEntryBlock().begin());
427 return TmpB.CreateAlloca(Type::DoubleTy, 0, VarName.c_str());
432 <p>This funny looking code creates an LLVMBuilder object that is pointing at
433 the first instruction (.begin()) of the entry block. It then creates an alloca
434 with the expected name and returns it. Because all values in Kaleidoscope are
435 doubles, there is no need to pass in a type to use.</p>
437 <p>With this in place, the first functionality change we want to make is to
438 variable references. In our new scheme, variables live on the stack, so code
439 generating a reference to them actually needs to produce a load from the stack
442 <div class="doc_code">
444 Value *VariableExprAST::Codegen() {
445 // Look this variable up in the function.
446 Value *V = NamedValues[Name];
447 if (V == 0) return ErrorV("Unknown variable name");
450 return Builder.CreateLoad(V, Name.c_str());
455 <p>As you can see, this is pretty straight-forward. Next we need to update the
456 things that define the variables to set up the alloca. We'll start with
457 <tt>ForExprAST::Codegen</tt> (see the <a href="#code">full code listing</a> for
458 the unabridged code):</p>
460 <div class="doc_code">
462 Function *TheFunction = Builder.GetInsertBlock()->getParent();
464 <b>// Create an alloca for the variable in the entry block.
465 AllocaInst *Alloca = CreateEntryBlockAlloca(TheFunction, VarName);</b>
467 // Emit the start code first, without 'variable' in scope.
468 Value *StartVal = Start->Codegen();
469 if (StartVal == 0) return 0;
471 <b>// Store the value into the alloca.
472 Builder.CreateStore(StartVal, Alloca);</b>
475 // Compute the end condition.
476 Value *EndCond = End->Codegen();
477 if (EndCond == 0) return EndCond;
479 <b>// Reload, increment, and restore the alloca. This handles the case where
480 // the body of the loop mutates the variable.
481 Value *CurVar = Builder.CreateLoad(Alloca);
482 Value *NextVar = Builder.CreateAdd(CurVar, StepVal, "nextvar");
483 Builder.CreateStore(NextVar, Alloca);</b>
488 <p>This code is virtually identical to the code <a
489 href="LangImpl5.html#forcodegen">before we allowed mutable variables</a>. The
490 big difference is that we no longer have to construct a PHI node, and we use
491 load/store to access the variable as needed.</p>
493 <p>To support mutable argument variables, we need to also make allocas for them.
494 The code for this is also pretty simple:</p>
496 <div class="doc_code">
498 /// CreateArgumentAllocas - Create an alloca for each argument and register the
499 /// argument in the symbol table so that references to it will succeed.
500 void PrototypeAST::CreateArgumentAllocas(Function *F) {
501 Function::arg_iterator AI = F->arg_begin();
502 for (unsigned Idx = 0, e = Args.size(); Idx != e; ++Idx, ++AI) {
503 // Create an alloca for this variable.
504 AllocaInst *Alloca = CreateEntryBlockAlloca(F, Args[Idx]);
506 // Store the initial value into the alloca.
507 Builder.CreateStore(AI, Alloca);
509 // Add arguments to variable symbol table.
510 NamedValues[Args[Idx]] = Alloca;
516 <p>For each argument, we make an alloca, store the input value to the function
517 into the alloca, and register the alloca as the memory location for the
518 argument. This method gets invoked by <tt>FunctionAST::Codegen</tt> right after
519 it sets up the entry block for the function.</p>
521 <p>The final missing piece is adding the 'mem2reg' pass, which allows us to get
522 good codegen once again:</p>
524 <div class="doc_code">
526 // Set up the optimizer pipeline. Start with registering info about how the
527 // target lays out data structures.
528 OurFPM.add(new TargetData(*TheExecutionEngine->getTargetData()));
529 <b>// Promote allocas to registers.
530 OurFPM.add(createPromoteMemoryToRegisterPass());</b>
531 // Do simple "peephole" optimizations and bit-twiddling optzns.
532 OurFPM.add(createInstructionCombiningPass());
533 // Reassociate expressions.
534 OurFPM.add(createReassociatePass());
538 <p>It is interesting to see what the code looks like before and after the
539 mem2reg optimization runs. For example, this is the before/after code for our
540 recursive fib. Before the optimization:</p>
542 <div class="doc_code">
544 define double @fib(double %x) {
546 <b>%x1 = alloca double
547 store double %x, double* %x1
548 %x2 = load double* %x1</b>
549 %multmp = fcmp ult double %x2, 3.000000e+00
550 %booltmp = uitofp i1 %multmp to double
551 %ifcond = fcmp one double %booltmp, 0.000000e+00
552 br i1 %ifcond, label %then, label %else
554 then: ; preds = %entry
557 else: ; preds = %entry
558 <b>%x3 = load double* %x1</b>
559 %subtmp = sub double %x3, 1.000000e+00
560 %calltmp = call double @fib( double %subtmp )
561 <b>%x4 = load double* %x1</b>
562 %subtmp5 = sub double %x4, 2.000000e+00
563 %calltmp6 = call double @fib( double %subtmp5 )
564 %addtmp = add double %calltmp, %calltmp6
567 ifcont: ; preds = %else, %then
568 %iftmp = phi double [ 1.000000e+00, %then ], [ %addtmp, %else ]
574 <p>Here there is only one variable (x, the input argument) but you can still
575 see the extremely simple-minded code generation strategy we are using. In the
576 entry block, an alloca is created, and the initial input value is stored into
577 it. Each reference to the variable does a reload from the stack. Also, note
578 that we didn't modify the if/then/else expression, so it still inserts a PHI
579 node. While we could make an alloca for it, it is actually easier to create a
580 PHI node for it, so we still just make the PHI.</p>
582 <p>Here is the code after the mem2reg pass runs:</p>
584 <div class="doc_code">
586 define double @fib(double %x) {
588 %multmp = fcmp ult double <b>%x</b>, 3.000000e+00
589 %booltmp = uitofp i1 %multmp to double
590 %ifcond = fcmp one double %booltmp, 0.000000e+00
591 br i1 %ifcond, label %then, label %else
597 %subtmp = sub double <b>%x</b>, 1.000000e+00
598 %calltmp = call double @fib( double %subtmp )
599 %subtmp5 = sub double <b>%x</b>, 2.000000e+00
600 %calltmp6 = call double @fib( double %subtmp5 )
601 %addtmp = add double %calltmp, %calltmp6
604 ifcont: ; preds = %else, %then
605 %iftmp = phi double [ 1.000000e+00, %then ], [ %addtmp, %else ]
611 <p>This is a trivial case for mem2reg, since there are no redefinitions of the
612 variable. The point of showing this is to calm your tension about inserting
613 such blatent inefficiencies :).</p>
615 <p>After the rest of the optimizers run, we get:</p>
617 <div class="doc_code">
619 define double @fib(double %x) {
621 %multmp = fcmp ult double %x, 3.000000e+00
622 %booltmp = uitofp i1 %multmp to double
623 %ifcond = fcmp ueq double %booltmp, 0.000000e+00
624 br i1 %ifcond, label %else, label %ifcont
627 %subtmp = sub double %x, 1.000000e+00
628 %calltmp = call double @fib( double %subtmp )
629 %subtmp5 = sub double %x, 2.000000e+00
630 %calltmp6 = call double @fib( double %subtmp5 )
631 %addtmp = add double %calltmp, %calltmp6
635 ret double 1.000000e+00
640 <p>Here we see that the simplifycfg pass decided to clone the return instruction
641 into the end of the 'else' block. This allowed it to eliminate some branches
642 and the PHI node.</p>
644 <p>Now that all symbol table references are updated to use stack variables,
645 we'll add the assignment operator.</p>
649 <!-- *********************************************************************** -->
650 <div class="doc_section"><a name="assignment">New Assignment Operator</a></div>
651 <!-- *********************************************************************** -->
653 <div class="doc_text">
655 <p>With our current framework, adding a new assignment operator is really
656 simple. We will parse it just like any other binary operator, but handle it
657 internally (instead of allowing the user to define it). The first step is to
658 set a precedence:</p>
660 <div class="doc_code">
663 // Install standard binary operators.
664 // 1 is lowest precedence.
665 <b>BinopPrecedence['='] = 2;</b>
666 BinopPrecedence['<'] = 10;
667 BinopPrecedence['+'] = 20;
668 BinopPrecedence['-'] = 20;
672 <p>Now that the parser knows the precedence of the binary operator, it takes
673 care of all the parsing and AST generation. We just need to implement codegen
674 for the assignment operator. This looks like:</p>
676 <div class="doc_code">
678 Value *BinaryExprAST::Codegen() {
679 // Special case '=' because we don't want to emit the LHS as an expression.
681 // Assignment requires the LHS to be an identifier.
682 VariableExprAST *LHSE = dynamic_cast<VariableExprAST*>(LHS);
684 return ErrorV("destination of '=' must be a variable");
688 <p>Unlike the rest of the binary operators, our assignment operator doesn't
689 follow the "emit LHS, emit RHS, do computation" model. As such, it is handled
690 as a special case before the other binary operators are handled. The other
691 strange thing about it is that it requires the LHS to be a variable directly.
694 <div class="doc_code">
697 Value *Val = RHS->Codegen();
698 if (Val == 0) return 0;
701 Value *Variable = NamedValues[LHSE->getName()];
702 if (Variable == 0) return ErrorV("Unknown variable name");
704 Builder.CreateStore(Val, Variable);
711 <p>Once it has the variable, codegen'ing the assignment is straight-forward:
712 we emit the RHS of the assignment, create a store, and return the computed
713 value. Returning a value allows for chained assignments like "X = (Y = Z)".</p>
715 <p>Now that we have an assignment operator, we can mutate loop variables and
716 arguments. For example, we can now run code like this:</p>
718 <div class="doc_code">
720 # Function to print a double.
723 # Define ':' for sequencing: as a low-precedence operator that ignores operands
724 # and just returns the RHS.
725 def binary : 1 (x y) y;
736 <p>When run, this example prints "123" and then "4", showing that we did
737 actually mutate the value! Okay, we have now officially implemented our goal:
738 getting this to work requires SSA construction in the general case. However,
739 to be really useful, we want the ability to define our own local variables, lets
745 <!-- *********************************************************************** -->
746 <div class="doc_section"><a name="localvars">User-defined Local
748 <!-- *********************************************************************** -->
750 <div class="doc_text">
752 <p>Adding var/in is just like any other other extensions we made to
753 Kaleidoscope: we extend the lexer, the parser, the AST and the code generator.
754 The first step for adding our new 'var/in' construct is to extend the lexer.
755 As before, this is pretty trivial, the code looks like this:</p>
757 <div class="doc_code">
766 static int gettok() {
768 if (IdentifierStr == "in") return tok_in;
769 if (IdentifierStr == "binary") return tok_binary;
770 if (IdentifierStr == "unary") return tok_unary;
771 <b>if (IdentifierStr == "var") return tok_var;</b>
772 return tok_identifier;
777 <p>The next step is to define the AST node that we will construct. For var/in,
778 it will look like this:</p>
780 <div class="doc_code">
782 /// VarExprAST - Expression class for var/in
783 class VarExprAST : public ExprAST {
784 std::vector<std::pair<std::string, ExprAST*> > VarNames;
787 VarExprAST(const std::vector<std::pair<std::string, ExprAST*> > &varnames,
789 : VarNames(varnames), Body(body) {}
791 virtual Value *Codegen();
796 <p>var/in allows a list of names to be defined all at once, and each name can
797 optionally have an initializer value. As such, we capture this information in
798 the VarNames vector. Also, var/in has a body, this body is allowed to access
799 the variables defined by the let/in.</p>
801 <p>With this ready, we can define the parser pieces. First thing we do is add
802 it as a primary expression:</p>
804 <div class="doc_code">
807 /// ::= identifierexpr
812 <b>/// ::= varexpr</b>
813 static ExprAST *ParsePrimary() {
815 default: return Error("unknown token when expecting an expression");
816 case tok_identifier: return ParseIdentifierExpr();
817 case tok_number: return ParseNumberExpr();
818 case '(': return ParseParenExpr();
819 case tok_if: return ParseIfExpr();
820 case tok_for: return ParseForExpr();
821 <b>case tok_var: return ParseVarExpr();</b>
827 <p>Next we define ParseVarExpr:</p>
829 <div class="doc_code">
831 /// varexpr ::= 'var' identifier ('=' expression)?
832 // (',' identifier ('=' expression)?)* 'in' expression
833 static ExprAST *ParseVarExpr() {
834 getNextToken(); // eat the var.
836 std::vector<std::pair<std::string, ExprAST*> > VarNames;
838 // At least one variable name is required.
839 if (CurTok != tok_identifier)
840 return Error("expected identifier after var");
844 <p>The first part of this code parses the list of identifier/expr pairs into the
845 local <tt>VarNames</tt> vector.
847 <div class="doc_code">
850 std::string Name = IdentifierStr;
851 getNextToken(); // eat identifier.
853 // Read the optional initializer.
856 getNextToken(); // eat the '='.
858 Init = ParseExpression();
859 if (Init == 0) return 0;
862 VarNames.push_back(std::make_pair(Name, Init));
864 // End of var list, exit loop.
865 if (CurTok != ',') break;
866 getNextToken(); // eat the ','.
868 if (CurTok != tok_identifier)
869 return Error("expected identifier list after var");
874 <p>Once all the variables are parsed, we then parse the body and create the
877 <div class="doc_code">
879 // At this point, we have to have 'in'.
880 if (CurTok != tok_in)
881 return Error("expected 'in' keyword after 'var'");
882 getNextToken(); // eat 'in'.
884 ExprAST *Body = ParseExpression();
885 if (Body == 0) return 0;
887 return new VarExprAST(VarNames, Body);
892 <p>Now that we can parse and represent the code, we need to support emission of
893 LLVM IR for it. This code starts out with:</p>
895 <div class="doc_code">
897 Value *VarExprAST::Codegen() {
898 std::vector<AllocaInst *> OldBindings;
900 Function *TheFunction = Builder.GetInsertBlock()->getParent();
902 // Register all variables and emit their initializer.
903 for (unsigned i = 0, e = VarNames.size(); i != e; ++i) {
904 const std::string &VarName = VarNames[i].first;
905 ExprAST *Init = VarNames[i].second;
909 <p>Basically it loops over all the variables, installing them one at a time.
910 For each variable we put into the symbol table, we remember the previous value
911 that we replace in OldBindings.</p>
913 <div class="doc_code">
915 // Emit the initializer before adding the variable to scope, this prevents
916 // the initializer from referencing the variable itself, and permits stuff
919 // var a = a in ... # refers to outer 'a'.
922 InitVal = Init->Codegen();
923 if (InitVal == 0) return 0;
924 } else { // If not specified, use 0.0.
925 InitVal = ConstantFP::get(Type::DoubleTy, APFloat(0.0));
928 AllocaInst *Alloca = CreateEntryBlockAlloca(TheFunction, VarName);
929 Builder.CreateStore(InitVal, Alloca);
931 // Remember the old variable binding so that we can restore the binding when
933 OldBindings.push_back(NamedValues[VarName]);
935 // Remember this binding.
936 NamedValues[VarName] = Alloca;
941 <p>There are more comments here than code. The basic idea is that we emit the
942 initializer, create the alloca, then update the symbol table to point to it.
943 Once all the variables are installed in the symbol table, we evaluate the body
944 of the var/in expression:</p>
946 <div class="doc_code">
948 // Codegen the body, now that all vars are in scope.
949 Value *BodyVal = Body->Codegen();
950 if (BodyVal == 0) return 0;
954 <p>Finally, before returning, we restore the previous variable bindings:</p>
956 <div class="doc_code">
958 // Pop all our variables from scope.
959 for (unsigned i = 0, e = VarNames.size(); i != e; ++i)
960 NamedValues[VarNames[i].first] = OldBindings[i];
962 // Return the body computation.
968 <p>The end result of all of this is that we get properly scoped variable
969 definitions, and we even (trivially) allow mutation of them :).</p>
971 <p>With this, we completed what we set out to do. Our nice iterative fib
972 example from the intro compiles and runs just fine. The mem2reg pass optimizes
973 all of our stack variables into SSA registers, inserting PHI nodes where needed,
974 and our front-end remains simple: no iterated dominator frontier computation
975 anywhere in sight.</p>
979 <!-- *********************************************************************** -->
980 <div class="doc_section"><a name="code">Full Code Listing</a></div>
981 <!-- *********************************************************************** -->
983 <div class="doc_text">
986 Here is the complete code listing for our running example, enhanced with mutable
987 variables and var/in support. To build this example, use:
990 <div class="doc_code">
993 g++ -g toy.cpp `llvm-config --cppflags --ldflags --libs core jit native` -O3 -o toy
999 <p>Here is the code:</p>
1001 <div class="doc_code">
1003 #include "llvm/DerivedTypes.h"
1004 #include "llvm/ExecutionEngine/ExecutionEngine.h"
1005 #include "llvm/Module.h"
1006 #include "llvm/ModuleProvider.h"
1007 #include "llvm/PassManager.h"
1008 #include "llvm/Analysis/Verifier.h"
1009 #include "llvm/Target/TargetData.h"
1010 #include "llvm/Transforms/Scalar.h"
1011 #include "llvm/Support/LLVMBuilder.h"
1012 #include <cstdio>
1013 #include <string>
1014 #include <map>
1015 #include <vector>
1016 using namespace llvm;
1018 //===----------------------------------------------------------------------===//
1020 //===----------------------------------------------------------------------===//
1022 // The lexer returns tokens [0-255] if it is an unknown character, otherwise one
1023 // of these for known things.
1028 tok_def = -2, tok_extern = -3,
1031 tok_identifier = -4, tok_number = -5,
1034 tok_if = -6, tok_then = -7, tok_else = -8,
1035 tok_for = -9, tok_in = -10,
1038 tok_binary = -11, tok_unary = -12,
1044 static std::string IdentifierStr; // Filled in if tok_identifier
1045 static double NumVal; // Filled in if tok_number
1047 /// gettok - Return the next token from standard input.
1048 static int gettok() {
1049 static int LastChar = ' ';
1051 // Skip any whitespace.
1052 while (isspace(LastChar))
1053 LastChar = getchar();
1055 if (isalpha(LastChar)) { // identifier: [a-zA-Z][a-zA-Z0-9]*
1056 IdentifierStr = LastChar;
1057 while (isalnum((LastChar = getchar())))
1058 IdentifierStr += LastChar;
1060 if (IdentifierStr == "def") return tok_def;
1061 if (IdentifierStr == "extern") return tok_extern;
1062 if (IdentifierStr == "if") return tok_if;
1063 if (IdentifierStr == "then") return tok_then;
1064 if (IdentifierStr == "else") return tok_else;
1065 if (IdentifierStr == "for") return tok_for;
1066 if (IdentifierStr == "in") return tok_in;
1067 if (IdentifierStr == "binary") return tok_binary;
1068 if (IdentifierStr == "unary") return tok_unary;
1069 if (IdentifierStr == "var") return tok_var;
1070 return tok_identifier;
1073 if (isdigit(LastChar) || LastChar == '.') { // Number: [0-9.]+
1077 LastChar = getchar();
1078 } while (isdigit(LastChar) || LastChar == '.');
1080 NumVal = strtod(NumStr.c_str(), 0);
1084 if (LastChar == '#') {
1085 // Comment until end of line.
1086 do LastChar = getchar();
1087 while (LastChar != EOF && LastChar != '\n' & LastChar != '\r');
1089 if (LastChar != EOF)
1093 // Check for end of file. Don't eat the EOF.
1094 if (LastChar == EOF)
1097 // Otherwise, just return the character as its ascii value.
1098 int ThisChar = LastChar;
1099 LastChar = getchar();
1103 //===----------------------------------------------------------------------===//
1104 // Abstract Syntax Tree (aka Parse Tree)
1105 //===----------------------------------------------------------------------===//
1107 /// ExprAST - Base class for all expression nodes.
1110 virtual ~ExprAST() {}
1111 virtual Value *Codegen() = 0;
1114 /// NumberExprAST - Expression class for numeric literals like "1.0".
1115 class NumberExprAST : public ExprAST {
1118 NumberExprAST(double val) : Val(val) {}
1119 virtual Value *Codegen();
1122 /// VariableExprAST - Expression class for referencing a variable, like "a".
1123 class VariableExprAST : public ExprAST {
1126 VariableExprAST(const std::string &name) : Name(name) {}
1127 const std::string &getName() const { return Name; }
1128 virtual Value *Codegen();
1131 /// UnaryExprAST - Expression class for a unary operator.
1132 class UnaryExprAST : public ExprAST {
1136 UnaryExprAST(char opcode, ExprAST *operand)
1137 : Opcode(opcode), Operand(operand) {}
1138 virtual Value *Codegen();
1141 /// BinaryExprAST - Expression class for a binary operator.
1142 class BinaryExprAST : public ExprAST {
1146 BinaryExprAST(char op, ExprAST *lhs, ExprAST *rhs)
1147 : Op(op), LHS(lhs), RHS(rhs) {}
1148 virtual Value *Codegen();
1151 /// CallExprAST - Expression class for function calls.
1152 class CallExprAST : public ExprAST {
1154 std::vector<ExprAST*> Args;
1156 CallExprAST(const std::string &callee, std::vector<ExprAST*> &args)
1157 : Callee(callee), Args(args) {}
1158 virtual Value *Codegen();
1161 /// IfExprAST - Expression class for if/then/else.
1162 class IfExprAST : public ExprAST {
1163 ExprAST *Cond, *Then, *Else;
1165 IfExprAST(ExprAST *cond, ExprAST *then, ExprAST *_else)
1166 : Cond(cond), Then(then), Else(_else) {}
1167 virtual Value *Codegen();
1170 /// ForExprAST - Expression class for for/in.
1171 class ForExprAST : public ExprAST {
1172 std::string VarName;
1173 ExprAST *Start, *End, *Step, *Body;
1175 ForExprAST(const std::string &varname, ExprAST *start, ExprAST *end,
1176 ExprAST *step, ExprAST *body)
1177 : VarName(varname), Start(start), End(end), Step(step), Body(body) {}
1178 virtual Value *Codegen();
1181 /// VarExprAST - Expression class for var/in
1182 class VarExprAST : public ExprAST {
1183 std::vector<std::pair<std::string, ExprAST*> > VarNames;
1186 VarExprAST(const std::vector<std::pair<std::string, ExprAST*> > &varnames,
1188 : VarNames(varnames), Body(body) {}
1190 virtual Value *Codegen();
1193 /// PrototypeAST - This class represents the "prototype" for a function,
1194 /// which captures its argument names as well as if it is an operator.
1195 class PrototypeAST {
1197 std::vector<std::string> Args;
1199 unsigned Precedence; // Precedence if a binary op.
1201 PrototypeAST(const std::string &name, const std::vector<std::string> &args,
1202 bool isoperator = false, unsigned prec = 0)
1203 : Name(name), Args(args), isOperator(isoperator), Precedence(prec) {}
1205 bool isUnaryOp() const { return isOperator && Args.size() == 1; }
1206 bool isBinaryOp() const { return isOperator && Args.size() == 2; }
1208 char getOperatorName() const {
1209 assert(isUnaryOp() || isBinaryOp());
1210 return Name[Name.size()-1];
1213 unsigned getBinaryPrecedence() const { return Precedence; }
1215 Function *Codegen();
1217 void CreateArgumentAllocas(Function *F);
1220 /// FunctionAST - This class represents a function definition itself.
1222 PrototypeAST *Proto;
1225 FunctionAST(PrototypeAST *proto, ExprAST *body)
1226 : Proto(proto), Body(body) {}
1228 Function *Codegen();
1231 //===----------------------------------------------------------------------===//
1233 //===----------------------------------------------------------------------===//
1235 /// CurTok/getNextToken - Provide a simple token buffer. CurTok is the current
1236 /// token the parser it looking at. getNextToken reads another token from the
1237 /// lexer and updates CurTok with its results.
1239 static int getNextToken() {
1240 return CurTok = gettok();
1243 /// BinopPrecedence - This holds the precedence for each binary operator that is
1245 static std::map<char, int> BinopPrecedence;
1247 /// GetTokPrecedence - Get the precedence of the pending binary operator token.
1248 static int GetTokPrecedence() {
1249 if (!isascii(CurTok))
1252 // Make sure it's a declared binop.
1253 int TokPrec = BinopPrecedence[CurTok];
1254 if (TokPrec <= 0) return -1;
1258 /// Error* - These are little helper functions for error handling.
1259 ExprAST *Error(const char *Str) { fprintf(stderr, "Error: %s\n", Str);return 0;}
1260 PrototypeAST *ErrorP(const char *Str) { Error(Str); return 0; }
1261 FunctionAST *ErrorF(const char *Str) { Error(Str); return 0; }
1263 static ExprAST *ParseExpression();
1267 /// ::= identifier '(' expression* ')'
1268 static ExprAST *ParseIdentifierExpr() {
1269 std::string IdName = IdentifierStr;
1271 getNextToken(); // eat identifier.
1273 if (CurTok != '(') // Simple variable ref.
1274 return new VariableExprAST(IdName);
1277 getNextToken(); // eat (
1278 std::vector<ExprAST*> Args;
1279 if (CurTok != ')') {
1281 ExprAST *Arg = ParseExpression();
1283 Args.push_back(Arg);
1285 if (CurTok == ')') break;
1288 return Error("Expected ')'");
1296 return new CallExprAST(IdName, Args);
1299 /// numberexpr ::= number
1300 static ExprAST *ParseNumberExpr() {
1301 ExprAST *Result = new NumberExprAST(NumVal);
1302 getNextToken(); // consume the number
1306 /// parenexpr ::= '(' expression ')'
1307 static ExprAST *ParseParenExpr() {
1308 getNextToken(); // eat (.
1309 ExprAST *V = ParseExpression();
1313 return Error("expected ')'");
1314 getNextToken(); // eat ).
1318 /// ifexpr ::= 'if' expression 'then' expression 'else' expression
1319 static ExprAST *ParseIfExpr() {
1320 getNextToken(); // eat the if.
1323 ExprAST *Cond = ParseExpression();
1324 if (!Cond) return 0;
1326 if (CurTok != tok_then)
1327 return Error("expected then");
1328 getNextToken(); // eat the then
1330 ExprAST *Then = ParseExpression();
1331 if (Then == 0) return 0;
1333 if (CurTok != tok_else)
1334 return Error("expected else");
1338 ExprAST *Else = ParseExpression();
1339 if (!Else) return 0;
1341 return new IfExprAST(Cond, Then, Else);
1344 /// forexpr ::= 'for' identifier '=' expr ',' expr (',' expr)? 'in' expression
1345 static ExprAST *ParseForExpr() {
1346 getNextToken(); // eat the for.
1348 if (CurTok != tok_identifier)
1349 return Error("expected identifier after for");
1351 std::string IdName = IdentifierStr;
1352 getNextToken(); // eat identifier.
1355 return Error("expected '=' after for");
1356 getNextToken(); // eat '='.
1359 ExprAST *Start = ParseExpression();
1360 if (Start == 0) return 0;
1362 return Error("expected ',' after for start value");
1365 ExprAST *End = ParseExpression();
1366 if (End == 0) return 0;
1368 // The step value is optional.
1370 if (CurTok == ',') {
1372 Step = ParseExpression();
1373 if (Step == 0) return 0;
1376 if (CurTok != tok_in)
1377 return Error("expected 'in' after for");
1378 getNextToken(); // eat 'in'.
1380 ExprAST *Body = ParseExpression();
1381 if (Body == 0) return 0;
1383 return new ForExprAST(IdName, Start, End, Step, Body);
1386 /// varexpr ::= 'var' identifier ('=' expression)?
1387 // (',' identifier ('=' expression)?)* 'in' expression
1388 static ExprAST *ParseVarExpr() {
1389 getNextToken(); // eat the var.
1391 std::vector<std::pair<std::string, ExprAST*> > VarNames;
1393 // At least one variable name is required.
1394 if (CurTok != tok_identifier)
1395 return Error("expected identifier after var");
1398 std::string Name = IdentifierStr;
1399 getNextToken(); // eat identifier.
1401 // Read the optional initializer.
1403 if (CurTok == '=') {
1404 getNextToken(); // eat the '='.
1406 Init = ParseExpression();
1407 if (Init == 0) return 0;
1410 VarNames.push_back(std::make_pair(Name, Init));
1412 // End of var list, exit loop.
1413 if (CurTok != ',') break;
1414 getNextToken(); // eat the ','.
1416 if (CurTok != tok_identifier)
1417 return Error("expected identifier list after var");
1420 // At this point, we have to have 'in'.
1421 if (CurTok != tok_in)
1422 return Error("expected 'in' keyword after 'var'");
1423 getNextToken(); // eat 'in'.
1425 ExprAST *Body = ParseExpression();
1426 if (Body == 0) return 0;
1428 return new VarExprAST(VarNames, Body);
1433 /// ::= identifierexpr
1439 static ExprAST *ParsePrimary() {
1441 default: return Error("unknown token when expecting an expression");
1442 case tok_identifier: return ParseIdentifierExpr();
1443 case tok_number: return ParseNumberExpr();
1444 case '(': return ParseParenExpr();
1445 case tok_if: return ParseIfExpr();
1446 case tok_for: return ParseForExpr();
1447 case tok_var: return ParseVarExpr();
1454 static ExprAST *ParseUnary() {
1455 // If the current token is not an operator, it must be a primary expr.
1456 if (!isascii(CurTok) || CurTok == '(' || CurTok == ',')
1457 return ParsePrimary();
1459 // If this is a unary operator, read it.
1462 if (ExprAST *Operand = ParseUnary())
1463 return new UnaryExprAST(Opc, Operand);
1468 /// ::= ('+' unary)*
1469 static ExprAST *ParseBinOpRHS(int ExprPrec, ExprAST *LHS) {
1470 // If this is a binop, find its precedence.
1472 int TokPrec = GetTokPrecedence();
1474 // If this is a binop that binds at least as tightly as the current binop,
1475 // consume it, otherwise we are done.
1476 if (TokPrec < ExprPrec)
1479 // Okay, we know this is a binop.
1481 getNextToken(); // eat binop
1483 // Parse the unary expression after the binary operator.
1484 ExprAST *RHS = ParseUnary();
1487 // If BinOp binds less tightly with RHS than the operator after RHS, let
1488 // the pending operator take RHS as its LHS.
1489 int NextPrec = GetTokPrecedence();
1490 if (TokPrec < NextPrec) {
1491 RHS = ParseBinOpRHS(TokPrec+1, RHS);
1492 if (RHS == 0) return 0;
1496 LHS = new BinaryExprAST(BinOp, LHS, RHS);
1501 /// ::= unary binoprhs
1503 static ExprAST *ParseExpression() {
1504 ExprAST *LHS = ParseUnary();
1507 return ParseBinOpRHS(0, LHS);
1511 /// ::= id '(' id* ')'
1512 /// ::= binary LETTER number? (id, id)
1513 /// ::= unary LETTER (id)
1514 static PrototypeAST *ParsePrototype() {
1517 int Kind = 0; // 0 = identifier, 1 = unary, 2 = binary.
1518 unsigned BinaryPrecedence = 30;
1522 return ErrorP("Expected function name in prototype");
1523 case tok_identifier:
1524 FnName = IdentifierStr;
1530 if (!isascii(CurTok))
1531 return ErrorP("Expected unary operator");
1533 FnName += (char)CurTok;
1539 if (!isascii(CurTok))
1540 return ErrorP("Expected binary operator");
1542 FnName += (char)CurTok;
1546 // Read the precedence if present.
1547 if (CurTok == tok_number) {
1548 if (NumVal < 1 || NumVal > 100)
1549 return ErrorP("Invalid precedecnce: must be 1..100");
1550 BinaryPrecedence = (unsigned)NumVal;
1557 return ErrorP("Expected '(' in prototype");
1559 std::vector<std::string> ArgNames;
1560 while (getNextToken() == tok_identifier)
1561 ArgNames.push_back(IdentifierStr);
1563 return ErrorP("Expected ')' in prototype");
1566 getNextToken(); // eat ')'.
1568 // Verify right number of names for operator.
1569 if (Kind && ArgNames.size() != Kind)
1570 return ErrorP("Invalid number of operands for operator");
1572 return new PrototypeAST(FnName, ArgNames, Kind != 0, BinaryPrecedence);
1575 /// definition ::= 'def' prototype expression
1576 static FunctionAST *ParseDefinition() {
1577 getNextToken(); // eat def.
1578 PrototypeAST *Proto = ParsePrototype();
1579 if (Proto == 0) return 0;
1581 if (ExprAST *E = ParseExpression())
1582 return new FunctionAST(Proto, E);
1586 /// toplevelexpr ::= expression
1587 static FunctionAST *ParseTopLevelExpr() {
1588 if (ExprAST *E = ParseExpression()) {
1589 // Make an anonymous proto.
1590 PrototypeAST *Proto = new PrototypeAST("", std::vector<std::string>());
1591 return new FunctionAST(Proto, E);
1596 /// external ::= 'extern' prototype
1597 static PrototypeAST *ParseExtern() {
1598 getNextToken(); // eat extern.
1599 return ParsePrototype();
1602 //===----------------------------------------------------------------------===//
1604 //===----------------------------------------------------------------------===//
1606 static Module *TheModule;
1607 static LLVMFoldingBuilder Builder;
1608 static std::map<std::string, AllocaInst*> NamedValues;
1609 static FunctionPassManager *TheFPM;
1611 Value *ErrorV(const char *Str) { Error(Str); return 0; }
1613 /// CreateEntryBlockAlloca - Create an alloca instruction in the entry block of
1614 /// the function. This is used for mutable variables etc.
1615 static AllocaInst *CreateEntryBlockAlloca(Function *TheFunction,
1616 const std::string &VarName) {
1617 LLVMBuilder TmpB(&TheFunction->getEntryBlock(),
1618 TheFunction->getEntryBlock().begin());
1619 return TmpB.CreateAlloca(Type::DoubleTy, 0, VarName.c_str());
1623 Value *NumberExprAST::Codegen() {
1624 return ConstantFP::get(Type::DoubleTy, APFloat(Val));
1627 Value *VariableExprAST::Codegen() {
1628 // Look this variable up in the function.
1629 Value *V = NamedValues[Name];
1630 if (V == 0) return ErrorV("Unknown variable name");
1633 return Builder.CreateLoad(V, Name.c_str());
1636 Value *UnaryExprAST::Codegen() {
1637 Value *OperandV = Operand->Codegen();
1638 if (OperandV == 0) return 0;
1640 Function *F = TheModule->getFunction(std::string("unary")+Opcode);
1642 return ErrorV("Unknown unary operator");
1644 return Builder.CreateCall(F, OperandV, "unop");
1648 Value *BinaryExprAST::Codegen() {
1649 // Special case '=' because we don't want to emit the LHS as an expression.
1651 // Assignment requires the LHS to be an identifier.
1652 VariableExprAST *LHSE = dynamic_cast<VariableExprAST*>(LHS);
1654 return ErrorV("destination of '=' must be a variable");
1656 Value *Val = RHS->Codegen();
1657 if (Val == 0) return 0;
1659 // Look up the name.
1660 Value *Variable = NamedValues[LHSE->getName()];
1661 if (Variable == 0) return ErrorV("Unknown variable name");
1663 Builder.CreateStore(Val, Variable);
1668 Value *L = LHS->Codegen();
1669 Value *R = RHS->Codegen();
1670 if (L == 0 || R == 0) return 0;
1673 case '+': return Builder.CreateAdd(L, R, "addtmp");
1674 case '-': return Builder.CreateSub(L, R, "subtmp");
1675 case '*': return Builder.CreateMul(L, R, "multmp");
1677 L = Builder.CreateFCmpULT(L, R, "multmp");
1678 // Convert bool 0/1 to double 0.0 or 1.0
1679 return Builder.CreateUIToFP(L, Type::DoubleTy, "booltmp");
1683 // If it wasn't a builtin binary operator, it must be a user defined one. Emit
1685 Function *F = TheModule->getFunction(std::string("binary")+Op);
1686 assert(F && "binary operator not found!");
1688 Value *Ops[] = { L, R };
1689 return Builder.CreateCall(F, Ops, Ops+2, "binop");
1692 Value *CallExprAST::Codegen() {
1693 // Look up the name in the global module table.
1694 Function *CalleeF = TheModule->getFunction(Callee);
1696 return ErrorV("Unknown function referenced");
1698 // If argument mismatch error.
1699 if (CalleeF->arg_size() != Args.size())
1700 return ErrorV("Incorrect # arguments passed");
1702 std::vector<Value*> ArgsV;
1703 for (unsigned i = 0, e = Args.size(); i != e; ++i) {
1704 ArgsV.push_back(Args[i]->Codegen());
1705 if (ArgsV.back() == 0) return 0;
1708 return Builder.CreateCall(CalleeF, ArgsV.begin(), ArgsV.end(), "calltmp");
1711 Value *IfExprAST::Codegen() {
1712 Value *CondV = Cond->Codegen();
1713 if (CondV == 0) return 0;
1715 // Convert condition to a bool by comparing equal to 0.0.
1716 CondV = Builder.CreateFCmpONE(CondV,
1717 ConstantFP::get(Type::DoubleTy, APFloat(0.0)),
1720 Function *TheFunction = Builder.GetInsertBlock()->getParent();
1722 // Create blocks for the then and else cases. Insert the 'then' block at the
1723 // end of the function.
1724 BasicBlock *ThenBB = new BasicBlock("then", TheFunction);
1725 BasicBlock *ElseBB = new BasicBlock("else");
1726 BasicBlock *MergeBB = new BasicBlock("ifcont");
1728 Builder.CreateCondBr(CondV, ThenBB, ElseBB);
1731 Builder.SetInsertPoint(ThenBB);
1733 Value *ThenV = Then->Codegen();
1734 if (ThenV == 0) return 0;
1736 Builder.CreateBr(MergeBB);
1737 // Codegen of 'Then' can change the current block, update ThenBB for the PHI.
1738 ThenBB = Builder.GetInsertBlock();
1741 TheFunction->getBasicBlockList().push_back(ElseBB);
1742 Builder.SetInsertPoint(ElseBB);
1744 Value *ElseV = Else->Codegen();
1745 if (ElseV == 0) return 0;
1747 Builder.CreateBr(MergeBB);
1748 // Codegen of 'Else' can change the current block, update ElseBB for the PHI.
1749 ElseBB = Builder.GetInsertBlock();
1751 // Emit merge block.
1752 TheFunction->getBasicBlockList().push_back(MergeBB);
1753 Builder.SetInsertPoint(MergeBB);
1754 PHINode *PN = Builder.CreatePHI(Type::DoubleTy, "iftmp");
1756 PN->addIncoming(ThenV, ThenBB);
1757 PN->addIncoming(ElseV, ElseBB);
1761 Value *ForExprAST::Codegen() {
1763 // var = alloca double
1765 // start = startexpr
1766 // store start -> var
1774 // endcond = endexpr
1776 // curvar = load var
1777 // nextvar = curvar + step
1778 // store nextvar -> var
1779 // br endcond, loop, endloop
1782 Function *TheFunction = Builder.GetInsertBlock()->getParent();
1784 // Create an alloca for the variable in the entry block.
1785 AllocaInst *Alloca = CreateEntryBlockAlloca(TheFunction, VarName);
1787 // Emit the start code first, without 'variable' in scope.
1788 Value *StartVal = Start->Codegen();
1789 if (StartVal == 0) return 0;
1791 // Store the value into the alloca.
1792 Builder.CreateStore(StartVal, Alloca);
1794 // Make the new basic block for the loop header, inserting after current
1796 BasicBlock *PreheaderBB = Builder.GetInsertBlock();
1797 BasicBlock *LoopBB = new BasicBlock("loop", TheFunction);
1799 // Insert an explicit fall through from the current block to the LoopBB.
1800 Builder.CreateBr(LoopBB);
1802 // Start insertion in LoopBB.
1803 Builder.SetInsertPoint(LoopBB);
1805 // Within the loop, the variable is defined equal to the PHI node. If it
1806 // shadows an existing variable, we have to restore it, so save it now.
1807 AllocaInst *OldVal = NamedValues[VarName];
1808 NamedValues[VarName] = Alloca;
1810 // Emit the body of the loop. This, like any other expr, can change the
1811 // current BB. Note that we ignore the value computed by the body, but don't
1813 if (Body->Codegen() == 0)
1816 // Emit the step value.
1819 StepVal = Step->Codegen();
1820 if (StepVal == 0) return 0;
1822 // If not specified, use 1.0.
1823 StepVal = ConstantFP::get(Type::DoubleTy, APFloat(1.0));
1826 // Compute the end condition.
1827 Value *EndCond = End->Codegen();
1828 if (EndCond == 0) return EndCond;
1830 // Reload, increment, and restore the alloca. This handles the case where
1831 // the body of the loop mutates the variable.
1832 Value *CurVar = Builder.CreateLoad(Alloca, VarName.c_str());
1833 Value *NextVar = Builder.CreateAdd(CurVar, StepVal, "nextvar");
1834 Builder.CreateStore(NextVar, Alloca);
1836 // Convert condition to a bool by comparing equal to 0.0.
1837 EndCond = Builder.CreateFCmpONE(EndCond,
1838 ConstantFP::get(Type::DoubleTy, APFloat(0.0)),
1841 // Create the "after loop" block and insert it.
1842 BasicBlock *LoopEndBB = Builder.GetInsertBlock();
1843 BasicBlock *AfterBB = new BasicBlock("afterloop", TheFunction);
1845 // Insert the conditional branch into the end of LoopEndBB.
1846 Builder.CreateCondBr(EndCond, LoopBB, AfterBB);
1848 // Any new code will be inserted in AfterBB.
1849 Builder.SetInsertPoint(AfterBB);
1851 // Restore the unshadowed variable.
1853 NamedValues[VarName] = OldVal;
1855 NamedValues.erase(VarName);
1858 // for expr always returns 0.0.
1859 return Constant::getNullValue(Type::DoubleTy);
1862 Value *VarExprAST::Codegen() {
1863 std::vector<AllocaInst *> OldBindings;
1865 Function *TheFunction = Builder.GetInsertBlock()->getParent();
1867 // Register all variables and emit their initializer.
1868 for (unsigned i = 0, e = VarNames.size(); i != e; ++i) {
1869 const std::string &VarName = VarNames[i].first;
1870 ExprAST *Init = VarNames[i].second;
1872 // Emit the initializer before adding the variable to scope, this prevents
1873 // the initializer from referencing the variable itself, and permits stuff
1876 // var a = a in ... # refers to outer 'a'.
1879 InitVal = Init->Codegen();
1880 if (InitVal == 0) return 0;
1881 } else { // If not specified, use 0.0.
1882 InitVal = ConstantFP::get(Type::DoubleTy, APFloat(0.0));
1885 AllocaInst *Alloca = CreateEntryBlockAlloca(TheFunction, VarName);
1886 Builder.CreateStore(InitVal, Alloca);
1888 // Remember the old variable binding so that we can restore the binding when
1890 OldBindings.push_back(NamedValues[VarName]);
1892 // Remember this binding.
1893 NamedValues[VarName] = Alloca;
1896 // Codegen the body, now that all vars are in scope.
1897 Value *BodyVal = Body->Codegen();
1898 if (BodyVal == 0) return 0;
1900 // Pop all our variables from scope.
1901 for (unsigned i = 0, e = VarNames.size(); i != e; ++i)
1902 NamedValues[VarNames[i].first] = OldBindings[i];
1904 // Return the body computation.
1909 Function *PrototypeAST::Codegen() {
1910 // Make the function type: double(double,double) etc.
1911 std::vector<const Type*> Doubles(Args.size(), Type::DoubleTy);
1912 FunctionType *FT = FunctionType::get(Type::DoubleTy, Doubles, false);
1914 Function *F = new Function(FT, Function::ExternalLinkage, Name, TheModule);
1916 // If F conflicted, there was already something named 'Name'. If it has a
1917 // body, don't allow redefinition or reextern.
1918 if (F->getName() != Name) {
1919 // Delete the one we just made and get the existing one.
1920 F->eraseFromParent();
1921 F = TheModule->getFunction(Name);
1923 // If F already has a body, reject this.
1924 if (!F->empty()) {
1925 ErrorF("redefinition of function");
1929 // If F took a different number of args, reject.
1930 if (F->arg_size() != Args.size()) {
1931 ErrorF("redefinition of function with different # args");
1936 // Set names for all arguments.
1938 for (Function::arg_iterator AI = F->arg_begin(); Idx != Args.size();
1940 AI->setName(Args[Idx]);
1945 /// CreateArgumentAllocas - Create an alloca for each argument and register the
1946 /// argument in the symbol table so that references to it will succeed.
1947 void PrototypeAST::CreateArgumentAllocas(Function *F) {
1948 Function::arg_iterator AI = F->arg_begin();
1949 for (unsigned Idx = 0, e = Args.size(); Idx != e; ++Idx, ++AI) {
1950 // Create an alloca for this variable.
1951 AllocaInst *Alloca = CreateEntryBlockAlloca(F, Args[Idx]);
1953 // Store the initial value into the alloca.
1954 Builder.CreateStore(AI, Alloca);
1956 // Add arguments to variable symbol table.
1957 NamedValues[Args[Idx]] = Alloca;
1962 Function *FunctionAST::Codegen() {
1963 NamedValues.clear();
1965 Function *TheFunction = Proto->Codegen();
1966 if (TheFunction == 0)
1969 // If this is an operator, install it.
1970 if (Proto->isBinaryOp())
1971 BinopPrecedence[Proto->getOperatorName()] = Proto->getBinaryPrecedence();
1973 // Create a new basic block to start insertion into.
1974 BasicBlock *BB = new BasicBlock("entry", TheFunction);
1975 Builder.SetInsertPoint(BB);
1977 // Add all arguments to the symbol table and create their allocas.
1978 Proto->CreateArgumentAllocas(TheFunction);
1980 if (Value *RetVal = Body->Codegen()) {
1981 // Finish off the function.
1982 Builder.CreateRet(RetVal);
1984 // Validate the generated code, checking for consistency.
1985 verifyFunction(*TheFunction);
1987 // Optimize the function.
1988 TheFPM->run(*TheFunction);
1993 // Error reading body, remove function.
1994 TheFunction->eraseFromParent();
1996 if (Proto->isBinaryOp())
1997 BinopPrecedence.erase(Proto->getOperatorName());
2001 //===----------------------------------------------------------------------===//
2002 // Top-Level parsing and JIT Driver
2003 //===----------------------------------------------------------------------===//
2005 static ExecutionEngine *TheExecutionEngine;
2007 static void HandleDefinition() {
2008 if (FunctionAST *F = ParseDefinition()) {
2009 if (Function *LF = F->Codegen()) {
2010 fprintf(stderr, "Read function definition:");
2014 // Skip token for error recovery.
2019 static void HandleExtern() {
2020 if (PrototypeAST *P = ParseExtern()) {
2021 if (Function *F = P->Codegen()) {
2022 fprintf(stderr, "Read extern: ");
2026 // Skip token for error recovery.
2031 static void HandleTopLevelExpression() {
2032 // Evaluate a top level expression into an anonymous function.
2033 if (FunctionAST *F = ParseTopLevelExpr()) {
2034 if (Function *LF = F->Codegen()) {
2035 // JIT the function, returning a function pointer.
2036 void *FPtr = TheExecutionEngine->getPointerToFunction(LF);
2038 // Cast it to the right type (takes no arguments, returns a double) so we
2039 // can call it as a native function.
2040 double (*FP)() = (double (*)())FPtr;
2041 fprintf(stderr, "Evaluated to %f\n", FP());
2044 // Skip token for error recovery.
2049 /// top ::= definition | external | expression | ';'
2050 static void MainLoop() {
2052 fprintf(stderr, "ready> ");
2054 case tok_eof: return;
2055 case ';': getNextToken(); break; // ignore top level semicolons.
2056 case tok_def: HandleDefinition(); break;
2057 case tok_extern: HandleExtern(); break;
2058 default: HandleTopLevelExpression(); break;
2065 //===----------------------------------------------------------------------===//
2066 // "Library" functions that can be "extern'd" from user code.
2067 //===----------------------------------------------------------------------===//
2069 /// putchard - putchar that takes a double and returns 0.
2071 double putchard(double X) {
2076 /// printd - printf that takes a double prints it as "%f\n", returning 0.
2078 double printd(double X) {
2083 //===----------------------------------------------------------------------===//
2084 // Main driver code.
2085 //===----------------------------------------------------------------------===//
2088 // Install standard binary operators.
2089 // 1 is lowest precedence.
2090 BinopPrecedence['='] = 2;
2091 BinopPrecedence['<'] = 10;
2092 BinopPrecedence['+'] = 20;
2093 BinopPrecedence['-'] = 20;
2094 BinopPrecedence['*'] = 40; // highest.
2096 // Prime the first token.
2097 fprintf(stderr, "ready> ");
2100 // Make the module, which holds all the code.
2101 TheModule = new Module("my cool jit");
2104 TheExecutionEngine = ExecutionEngine::create(TheModule);
2107 ExistingModuleProvider OurModuleProvider(TheModule);
2108 FunctionPassManager OurFPM(&OurModuleProvider);
2110 // Set up the optimizer pipeline. Start with registering info about how the
2111 // target lays out data structures.
2112 OurFPM.add(new TargetData(*TheExecutionEngine->getTargetData()));
2113 // Promote allocas to registers.
2114 OurFPM.add(createPromoteMemoryToRegisterPass());
2115 // Do simple "peephole" optimizations and bit-twiddling optzns.
2116 OurFPM.add(createInstructionCombiningPass());
2117 // Reassociate expressions.
2118 OurFPM.add(createReassociatePass());
2119 // Eliminate Common SubExpressions.
2120 OurFPM.add(createGVNPass());
2121 // Simplify the control flow graph (deleting unreachable blocks, etc).
2122 OurFPM.add(createCFGSimplificationPass());
2124 // Set the global so the code gen can use this.
2125 TheFPM = &OurFPM;
2127 // Run the main "interpreter loop" now.
2131 } // Free module provider and pass manager.
2134 // Print out all of the generated code.
2135 TheModule->dump();
2143 <!-- *********************************************************************** -->
2146 <a href="http://jigsaw.w3.org/css-validator/check/referer"><img
2147 src="http://jigsaw.w3.org/css-validator/images/vcss" alt="Valid CSS!"></a>
2148 <a href="http://validator.w3.org/check/referer"><img
2149 src="http://www.w3.org/Icons/valid-html401" alt="Valid HTML 4.01!"></a>
2151 <a href="mailto:sabre@nondot.org">Chris Lattner</a><br>
2152 <a href="http://llvm.org">The LLVM Compiler Infrastructure</a><br>
2153 Last modified: $Date: 2007-10-17 11:05:13 -0700 (Wed, 17 Oct 2007) $