Geek Answers Handbook

Haskell compiler code generation

I am writing a compiler for a small imperative language. The target language is Java bytecode, and the compiler is implemented in Haskell.

I wrote an interface for the language - I have lexer, parser and typechecker. I am having trouble figuring out how to generate the code.

I keep a data structure representing a stack of local variables. I can query this structure with the name of a local variable and get its position on the stack. This data structure is passed when I walk through the syntax tree, and variables are pushed and pushed as they enter and exit new areas.

I'm having trouble figuring out how to emit a bytecode. Emitting strings on terminals and combining them at higher levels seems like a bad decision, both with clarity and in terms of performance.

tl; dr How to emit bytecode when waling a syntax tree?

+5
source share
2 answers

Haskell , c, , . , , (, , ) -, .

LIR ( ), (x86_64 ):

data LIRInst = LIRRegAssignInst LIRReg LIRExpr
             | LIRRegOffAssignInst LIRReg LIRReg LIRSize LIROperand
             | LIRStoreInst LIRMemAddr LIROperand
             | LIRLoadInst LIRReg LIRMemAddr
             | LIREnterInst LIRInt
             | LIRJumpLabelInst LIRLabel
             | LIRIfInst LIRRelExpr LIRLabel LIRLabel -- false, then true
             | LIRCallInst LIRLabel LIRLabel -- method label, return label
             | LIRCalloutInst String
             | LIRRetInst [LIRLabel] String -- list of successors, and the name of the method returning from
             | LIRLabelInst LIRLabel
             deriving (Show, Eq, Typeable)

, ( - State Monad -):

newtype LIRTranslator a = LIRTranslator
    { runLIR :: Namespace -> (a, Namespace) }

instance Monad LIRTranslator where
    return a = LIRTranslator (\s -> (a, s))
    m >>= f = LIRTranslator (\s ->
        let (a, s') = runLIR m s
        in runLIR (f a) s')

, "" :

data Namespace = Namespace
    { temp         :: Int                       -- id for new temporaries
    , labels       :: Int                       -- id for new labels
    , scope        :: [(LIRLabel, LIRLabel)]    -- current program scope
    , encMethod    :: String                    -- current enclosing method
    , blockindex   :: [Int]                     -- index into the SymbolTree
    , successorMap :: Map.Map String [LIRLabel]
    , ivarStack    :: [(LIRReg, [CFGInst])]     -- stack of ivars (see motioned code)
    }

, :

-- |Increment our translator label counter
incLabel :: LIRTranslator Int
incLabel = LIRTranslator (\ns@(Namespace{ labels = l }) -> (l, ns{ labels = (l+1) }))

, AST, , :

translateBlock :: SymbolTree -> ASTBlock -> LIRTranslator [LIRInst]
translateBlock st (DecafBlock _ [] _) = withBlock (return [])
translateBlock st block =
    withBlock (do b <- getBlock
                  let st' = select b st
                  declarations <- mapM (translateVarDeclaration st') (blockVars block)
                  statements <- mapM (translateStm st') (blockStms block)
                  return (concat declarations ++ concat statements))

( )

-- | Given a SymbolTree, Translate a single DecafMethodStm into [LIRInst]
translateStm st (DecafMethodStm mc _) =
    do (instructions, operand) <- translateMethodCall st mc
       final <- motionCode instructions
       return final

( )

translateMethodPrologue :: SymbolTree -> DecafMethod -> LIRTranslator [LIRInst]
translateMethodPrologue st (DecafMethod _ ident args _ _) =
    do let numRegVars = min (length args) 6
           regvars = map genRegVar (zip [LRDI, LRSI, LRDX, LRCX, LR8, LR9] args)
       stackvars <- mapM genStackVar (zip [1..] (drop numRegVars args))
       return (regvars ++ stackvars)
  where
    genRegVar (reg, arg) =
        LIRRegAssignInst (symVar arg st) (LIROperExpr $ LIRRegOperand reg)
    genStackVar (index, arg) =
        do let mem = LIRMemAddr LRBP Nothing ((index + 1) * 8) qword -- ^ [rbp] = old rbp; [rbp + 8] = ret address; [rbp + 16] = first stack param
                                  return $ LIRLoadInst (symVar arg st) mem

LIR. , ; , , ( ) .

+4

, :   1) - ( )   2) , , gotos .., (      , )   3) /   4) JAR

, . , . , JVM ( , ).

, , . , . , . . , .

+2

More articles:


All Articles