% % (c) The GRASP/AQUA Project, Glasgow University, 1992-1998 % \section[CoreUtils]{Utility functions on @Core@ syntax} \begin{code} module CoreUtils ( -- Construction mkInlineMe, mkSCC, mkCoerce, mkCoerce2, bindNonRec, needsCaseBinding, mkIfThenElse, mkAltExpr, mkPiType, mkPiTypes, -- Taking expressions apart findDefault, findAlt, isDefaultAlt, -- Properties of expressions exprType, coreAltType, exprIsDupable, exprIsTrivial, exprIsCheap, exprIsHNF,exprOkForSpeculation, exprIsBig, exprIsConApp_maybe, exprIsBottom, rhsIsStatic, -- Arity and eta expansion manifestArity, exprArity, exprEtaExpandArity, etaExpand, -- Size coreBindsSize, -- Hashing hashExpr, -- Equality cheapEqExpr, tcEqExpr, tcEqExprX, applyTypeToArgs, applyTypeToArg ) where #include "HsVersions.h" import GLAEXTS -- For `xori` import CoreSyn import CoreFVs ( exprFreeVars ) import PprCore ( pprCoreExpr ) import Var ( Var ) import VarSet ( unionVarSet ) import VarEnv import Name ( hashName ) import Packages ( HomeModules ) #if mingw32_TARGET_OS import Packages ( isDllName ) #endif import Literal ( hashLiteral, literalType, litIsDupable, litIsTrivial, isZeroLit, Literal( MachLabel ) ) import DataCon ( DataCon, dataConRepArity, dataConInstArgTys, isVanillaDataCon, dataConTyCon ) import PrimOp ( PrimOp(..), primOpOkForSpeculation, primOpIsCheap ) import Id ( Id, idType, globalIdDetails, idNewStrictness, mkWildId, idArity, idName, idUnfolding, idInfo, isOneShotBndr, isStateHackType, isDataConWorkId_maybe, mkSysLocal, isDataConWorkId, isBottomingId ) import IdInfo ( GlobalIdDetails(..), megaSeqIdInfo ) import NewDemand ( appIsBottom ) import Type ( Type, mkFunTy, mkForAllTy, splitFunTy_maybe, splitFunTy, tcEqTypeX, applyTys, isUnLiftedType, seqType, mkTyVarTy, splitForAllTy_maybe, isForAllTy, splitRecNewType_maybe, splitTyConApp_maybe, coreEqType, funResultTy, applyTy ) import TyCon ( tyConArity ) import TysWiredIn ( boolTy, trueDataCon, falseDataCon ) import CostCentre ( CostCentre ) import BasicTypes ( Arity ) import Unique ( Unique ) import Outputable import TysPrim ( alphaTy ) -- Debugging only import Util ( equalLength, lengthAtLeast, foldl2 ) \end{code} %************************************************************************ %* * \subsection{Find the type of a Core atom/expression} %* * %************************************************************************ \begin{code} exprType :: CoreExpr -> Type exprType (Var var) = idType var exprType (Lit lit) = literalType lit exprType (Let _ body) = exprType body exprType (Case _ _ ty alts) = ty exprType (Note (Coerce ty _) e) = ty -- **! should take usage from e exprType (Note other_note e) = exprType e exprType (Lam binder expr) = mkPiType binder (exprType expr) exprType e@(App _ _) = case collectArgs e of (fun, args) -> applyTypeToArgs e (exprType fun) args exprType other = pprTrace "exprType" (pprCoreExpr other) alphaTy coreAltType :: CoreAlt -> Type coreAltType (_,_,rhs) = exprType rhs \end{code} @mkPiType@ makes a (->) type or a forall type, depending on whether it is given a type variable or a term variable. We cleverly use the lbvarinfo field to figure out the right annotation for the arrove in case of a term variable. \begin{code} mkPiType :: Var -> Type -> Type -- The more polymorphic version mkPiTypes :: [Var] -> Type -> Type -- doesn't work... mkPiTypes vs ty = foldr mkPiType ty vs mkPiType v ty | isId v = mkFunTy (idType v) ty | otherwise = mkForAllTy v ty \end{code} \begin{code} applyTypeToArg :: Type -> CoreExpr -> Type applyTypeToArg fun_ty (Type arg_ty) = applyTy fun_ty arg_ty applyTypeToArg fun_ty other_arg = funResultTy fun_ty applyTypeToArgs :: CoreExpr -> Type -> [CoreExpr] -> Type -- A more efficient version of applyTypeToArg -- when we have several args -- The first argument is just for debugging applyTypeToArgs e op_ty [] = op_ty applyTypeToArgs e op_ty (Type ty : args) = -- Accumulate type arguments so we can instantiate all at once go [ty] args where go rev_tys (Type ty : args) = go (ty:rev_tys) args go rev_tys rest_args = applyTypeToArgs e op_ty' rest_args where op_ty' = applyTys op_ty (reverse rev_tys) applyTypeToArgs e op_ty (other_arg : args) = case (splitFunTy_maybe op_ty) of Just (_, res_ty) -> applyTypeToArgs e res_ty args Nothing -> pprPanic "applyTypeToArgs" (pprCoreExpr e) \end{code} %************************************************************************ %* * \subsection{Attaching notes} %* * %************************************************************************ mkNote removes redundant coercions, and SCCs where possible \begin{code} #ifdef UNUSED mkNote :: Note -> CoreExpr -> CoreExpr mkNote (Coerce to_ty from_ty) expr = mkCoerce2 to_ty from_ty expr mkNote (SCC cc) expr = mkSCC cc expr mkNote InlineMe expr = mkInlineMe expr mkNote note expr = Note note expr #endif -- Slide InlineCall in around the function -- No longer necessary I think (SLPJ Apr 99) -- mkNote InlineCall (App f a) = App (mkNote InlineCall f) a -- mkNote InlineCall (Var v) = Note InlineCall (Var v) -- mkNote InlineCall expr = expr \end{code} Drop trivial InlineMe's. This is somewhat important, because if we have an unfolding that looks like (Note InlineMe (Var v)), the InlineMe doesn't go away because it may not be *applied* to anything. We don't use exprIsTrivial here, though, because we sometimes generate worker/wrapper bindings like fw = ... f = inline_me (coerce t fw) As usual, the inline_me prevents the worker from getting inlined back into the wrapper. We want the split, so that the coerces can cancel at the call site. However, we can get left with tiresome type applications. Notably, consider f = /\ a -> let t = e in (t, w) Then lifting the let out of the big lambda gives t' = /\a -> e f = /\ a -> let t = inline_me (t' a) in (t, w) The inline_me is to stop the simplifier inlining t' right back into t's RHS. In the next phase we'll substitute for t (since its rhs is trivial) and *then* we could get rid of the inline_me. But it hardly seems worth it, so I don't bother. \begin{code} mkInlineMe (Var v) = Var v mkInlineMe e = Note InlineMe e \end{code} \begin{code} mkCoerce :: Type -> CoreExpr -> CoreExpr mkCoerce to_ty expr = mkCoerce2 to_ty (exprType expr) expr mkCoerce2 :: Type -> Type -> CoreExpr -> CoreExpr mkCoerce2 to_ty from_ty (Note (Coerce to_ty2 from_ty2) expr) = ASSERT( from_ty `coreEqType` to_ty2 ) mkCoerce2 to_ty from_ty2 expr mkCoerce2 to_ty from_ty expr | to_ty `coreEqType` from_ty = expr | otherwise = ASSERT( from_ty `coreEqType` exprType expr ) Note (Coerce to_ty from_ty) expr \end{code} \begin{code} mkSCC :: CostCentre -> Expr b -> Expr b -- Note: Nested SCC's *are* preserved for the benefit of -- cost centre stack profiling mkSCC cc (Lit lit) = Lit lit mkSCC cc (Lam x e) = Lam x (mkSCC cc e) -- Move _scc_ inside lambda mkSCC cc (Note (SCC cc') e) = Note (SCC cc) (Note (SCC cc') e) mkSCC cc (Note n e) = Note n (mkSCC cc e) -- Move _scc_ inside notes mkSCC cc expr = Note (SCC cc) expr \end{code} %************************************************************************ %* * \subsection{Other expression construction} %* * %************************************************************************ \begin{code} bindNonRec :: Id -> CoreExpr -> CoreExpr -> CoreExpr -- (bindNonRec x r b) produces either -- let x = r in b -- or -- case r of x { _DEFAULT_ -> b } -- -- depending on whether x is unlifted or not -- It's used by the desugarer to avoid building bindings -- that give Core Lint a heart attack. Actually the simplifier -- deals with them perfectly well. bindNonRec bndr rhs body | needsCaseBinding (idType bndr) rhs = Case rhs bndr (exprType body) [(DEFAULT,[],body)] | otherwise = Let (NonRec bndr rhs) body needsCaseBinding ty rhs = isUnLiftedType ty && not (exprOkForSpeculation rhs) -- Make a case expression instead of a let -- These can arise either from the desugarer, -- or from beta reductions: (\x.e) (x +# y) \end{code} \begin{code} mkAltExpr :: AltCon -> [CoreBndr] -> [Type] -> CoreExpr -- This guy constructs the value that the scrutinee must have -- when you are in one particular branch of a case mkAltExpr (DataAlt con) args inst_tys = mkConApp con (map Type inst_tys ++ map varToCoreExpr args) mkAltExpr (LitAlt lit) [] [] = Lit lit mkIfThenElse :: CoreExpr -> CoreExpr -> CoreExpr -> CoreExpr mkIfThenElse guard then_expr else_expr -- Not going to be refining, so okay to take the type of the "then" clause = Case guard (mkWildId boolTy) (exprType then_expr) [ (DataAlt falseDataCon, [], else_expr), -- Increasing order of tag! (DataAlt trueDataCon, [], then_expr) ] \end{code} %************************************************************************ %* * \subsection{Taking expressions apart} %* * %************************************************************************ The default alternative must be first, if it exists at all. This makes it easy to find, though it makes matching marginally harder. \begin{code} findDefault :: [CoreAlt] -> ([CoreAlt], Maybe CoreExpr) findDefault ((DEFAULT,args,rhs) : alts) = ASSERT( null args ) (alts, Just rhs) findDefault alts = (alts, Nothing) findAlt :: AltCon -> [CoreAlt] -> CoreAlt findAlt con alts = case alts of (deflt@(DEFAULT,_,_):alts) -> go alts deflt other -> go alts panic_deflt where panic_deflt = pprPanic "Missing alternative" (ppr con $$ vcat (map ppr alts)) go [] deflt = deflt go (alt@(con1,_,_) : alts) deflt = case con `cmpAltCon` con1 of LT -> deflt -- Missed it already; the alts are in increasing order EQ -> alt GT -> ASSERT( not (con1 == DEFAULT) ) go alts deflt isDefaultAlt :: CoreAlt -> Bool isDefaultAlt (DEFAULT, _, _) = True isDefaultAlt other = False \end{code} %************************************************************************ %* * \subsection{Figuring out things about expressions} %* * %************************************************************************ @exprIsTrivial@ is true of expressions we are unconditionally happy to duplicate; simple variables and constants, and type applications. Note that primop Ids aren't considered trivial unless @exprIsBottom@ is true of expressions that are guaranteed to diverge There used to be a gruesome test for (hasNoBinding v) in the Var case: exprIsTrivial (Var v) | hasNoBinding v = idArity v == 0 The idea here is that a constructor worker, like $wJust, is really short for (\x -> $wJust x), becuase $wJust has no binding. So it should be treated like a lambda. Ditto unsaturated primops. But now constructor workers are not "have-no-binding" Ids. And completely un-applied primops and foreign-call Ids are sufficiently rare that I plan to allow them to be duplicated and put up with saturating them. SCC notes. We do not treat (_scc_ "foo" x) as trivial, because a) it really generates code, (and a heap object when it's a function arg) to capture the cost centre b) see the note [SCC-and-exprIsTrivial] in Simplify.simplLazyBind \begin{code} exprIsTrivial (Var v) = True -- See notes above exprIsTrivial (Type _) = True exprIsTrivial (Lit lit) = litIsTrivial lit exprIsTrivial (App e arg) = not (isRuntimeArg arg) && exprIsTrivial e exprIsTrivial (Note (SCC _) e) = False -- See notes above exprIsTrivial (Note _ e) = exprIsTrivial e exprIsTrivial (Lam b body) = not (isRuntimeVar b) && exprIsTrivial body exprIsTrivial other = False \end{code} @exprIsDupable@ is true of expressions that can be duplicated at a modest cost in code size. This will only happen in different case branches, so there's no issue about duplicating work. That is, exprIsDupable returns True of (f x) even if f is very very expensive to call. Its only purpose is to avoid fruitless let-binding and then inlining of case join points \begin{code} exprIsDupable (Type _) = True exprIsDupable (Var v) = True exprIsDupable (Lit lit) = litIsDupable lit exprIsDupable (Note InlineMe e) = True exprIsDupable (Note _ e) = exprIsDupable e exprIsDupable expr = go expr 0 where go (Var v) n_args = True go (App f a) n_args = n_args < dupAppSize && exprIsDupable a && go f (n_args+1) go other n_args = False dupAppSize :: Int dupAppSize = 4 -- Size of application we are prepared to duplicate \end{code} @exprIsCheap@ looks at a Core expression and returns \tr{True} if it is obviously in weak head normal form, or is cheap to get to WHNF. [Note that that's not the same as exprIsDupable; an expression might be big, and hence not dupable, but still cheap.] By ``cheap'' we mean a computation we're willing to: push inside a lambda, or inline at more than one place That might mean it gets evaluated more than once, instead of being shared. The main examples of things which aren't WHNF but are ``cheap'' are: * case e of pi -> ei (where e, and all the ei are cheap) * let x = e in b (where e and b are cheap) * op x1 ... xn (where op is a cheap primitive operator) * error "foo" (because we are happy to substitute it inside a lambda) Notice that a variable is considered 'cheap': we can push it inside a lambda, because sharing will make sure it is only evaluated once. \begin{code} exprIsCheap :: CoreExpr -> Bool exprIsCheap (Lit lit) = True exprIsCheap (Type _) = True exprIsCheap (Var _) = True exprIsCheap (Note InlineMe e) = True exprIsCheap (Note _ e) = exprIsCheap e exprIsCheap (Lam x e) = isRuntimeVar x || exprIsCheap e exprIsCheap (Case e _ _ alts) = exprIsCheap e && and [exprIsCheap rhs | (_,_,rhs) <- alts] -- Experimentally, treat (case x of ...) as cheap -- (and case __coerce x etc.) -- This improves arities of overloaded functions where -- there is only dictionary selection (no construction) involved exprIsCheap (Let (NonRec x _) e) | isUnLiftedType (idType x) = exprIsCheap e | otherwise = False -- strict lets always have cheap right hand sides, and -- do no allocation. exprIsCheap other_expr = go other_expr 0 True where go (Var f) n_args args_cheap = (idAppIsCheap f n_args && args_cheap) -- A constructor, cheap primop, or partial application || idAppIsBottom f n_args -- Application of a function which -- always gives bottom; we treat this as cheap -- because it certainly doesn't need to be shared! go (App f a) n_args args_cheap | not (isRuntimeArg a) = go f n_args args_cheap | otherwise = go f (n_args + 1) (exprIsCheap a && args_cheap) go other n_args args_cheap = False idAppIsCheap :: Id -> Int -> Bool idAppIsCheap id n_val_args | n_val_args == 0 = True -- Just a type application of -- a variable (f t1 t2 t3) -- counts as WHNF | otherwise = case globalIdDetails id of DataConWorkId _ -> True RecordSelId {} -> n_val_args == 1 -- I'm experimenting with making record selection ClassOpId _ -> n_val_args == 1 -- look cheap, so we will substitute it inside a -- lambda. Particularly for dictionary field selection. -- BUT: Take care with (sel d x)! The (sel d) might be cheap, but -- there's no guarantee that (sel d x) will be too. Hence (n_val_args == 1) PrimOpId op -> primOpIsCheap op -- In principle we should worry about primops -- that return a type variable, since the result -- might be applied to something, but I'm not going -- to bother to check the number of args other -> n_val_args < idArity id \end{code} exprOkForSpeculation returns True of an expression that it is * safe to evaluate even if normal order eval might not evaluate the expression at all, or * safe *not* to evaluate even if normal order would do so It returns True iff the expression guarantees to terminate, soon, without raising an exception, without causing a side effect (e.g. writing a mutable variable) E.G. let x = case y# +# 1# of { r# -> I# r# } in E ==> case y# +# 1# of { r# -> let x = I# r# in E } We can only do this if the (y+1) is ok for speculation: it has no side effects, and can't diverge or raise an exception. \begin{code} exprOkForSpeculation :: CoreExpr -> Bool exprOkForSpeculation (Lit _) = True exprOkForSpeculation (Type _) = True exprOkForSpeculation (Var v) = isUnLiftedType (idType v) exprOkForSpeculation (Note _ e) = exprOkForSpeculation e exprOkForSpeculation other_expr = case collectArgs other_expr of (Var f, args) -> spec_ok (globalIdDetails f) args other -> False where spec_ok (DataConWorkId _) args = True -- The strictness of the constructor has already -- been expressed by its "wrapper", so we don't need -- to take the arguments into account spec_ok (PrimOpId op) args | isDivOp op, -- Special case for dividing operations that fail [arg1, Lit lit] <- args -- only if the divisor is zero = not (isZeroLit lit) && exprOkForSpeculation arg1 -- Often there is a literal divisor, and this -- can get rid of a thunk in an inner looop | otherwise = primOpOkForSpeculation op && all exprOkForSpeculation args -- A bit conservative: we don't really need -- to care about lazy arguments, but this is easy spec_ok other args = False isDivOp :: PrimOp -> Bool -- True of dyadic operators that can fail -- only if the second arg is zero -- This function probably belongs in PrimOp, or even in -- an automagically generated file.. but it's such a -- special case I thought I'd leave it here for now. isDivOp IntQuotOp = True isDivOp IntRemOp = True isDivOp WordQuotOp = True isDivOp WordRemOp = True isDivOp IntegerQuotRemOp = True isDivOp IntegerDivModOp = True isDivOp FloatDivOp = True isDivOp DoubleDivOp = True isDivOp other = False \end{code} \begin{code} exprIsBottom :: CoreExpr -> Bool -- True => definitely bottom exprIsBottom e = go 0 e where -- n is the number of args go n (Note _ e) = go n e go n (Let _ e) = go n e go n (Case e _ _ _) = go 0 e -- Just check the scrut go n (App e _) = go (n+1) e go n (Var v) = idAppIsBottom v n go n (Lit _) = False go n (Lam _ _) = False go n (Type _) = False idAppIsBottom :: Id -> Int -> Bool idAppIsBottom id n_val_args = appIsBottom (idNewStrictness id) n_val_args \end{code} @exprIsHNF@ returns true for expressions that are certainly *already* evaluated to *head* normal form. This is used to decide whether it's ok to change case x of _ -> e ===> e and to decide whether it's safe to discard a `seq` So, it does *not* treat variables as evaluated, unless they say they are. But it *does* treat partial applications and constructor applications as values, even if their arguments are non-trivial, provided the argument type is lifted; e.g. (:) (f x) (map f xs) is a value map (...redex...) is a value Because `seq` on such things completes immediately For unlifted argument types, we have to be careful: C (f x :: Int#) Suppose (f x) diverges; then C (f x) is not a value. True, but this form is illegal (see the invariants in CoreSyn). Args of unboxed type must be ok-for-speculation (or trivial). \begin{code} exprIsHNF :: CoreExpr -> Bool -- True => Value-lambda, constructor, PAP exprIsHNF (Var v) -- NB: There are no value args at this point = isDataConWorkId v -- Catches nullary constructors, -- so that [] and () are values, for example || idArity v > 0 -- Catches (e.g.) primops that don't have unfoldings || isEvaldUnfolding (idUnfolding v) -- Check the thing's unfolding; it might be bound to a value -- A worry: what if an Id's unfolding is just itself: -- then we could get an infinite loop... exprIsHNF (Lit l) = True exprIsHNF (Type ty) = True -- Types are honorary Values; -- we don't mind copying them exprIsHNF (Lam b e) = isRuntimeVar b || exprIsHNF e exprIsHNF (Note _ e) = exprIsHNF e exprIsHNF (App e (Type _)) = exprIsHNF e exprIsHNF (App e a) = app_is_value e [a] exprIsHNF other = False -- There is at least one value argument app_is_value (Var fun) args | isDataConWorkId fun -- Constructor apps are values || idArity fun > valArgCount args -- Under-applied function = check_args (idType fun) args app_is_value (App f a) as = app_is_value f (a:as) app_is_value other as = False -- 'check_args' checks that unlifted-type args -- are in fact guaranteed non-divergent check_args fun_ty [] = True check_args fun_ty (Type _ : args) = case splitForAllTy_maybe fun_ty of Just (_, ty) -> check_args ty args check_args fun_ty (arg : args) | isUnLiftedType arg_ty = exprOkForSpeculation arg | otherwise = check_args res_ty args where (arg_ty, res_ty) = splitFunTy fun_ty \end{code} \begin{code} exprIsConApp_maybe :: CoreExpr -> Maybe (DataCon, [CoreExpr]) exprIsConApp_maybe (Note (Coerce to_ty from_ty) expr) = -- Maybe this is over the top, but here we try to turn -- coerce (S,T) ( x, y ) -- effectively into -- ( coerce S x, coerce T y ) -- This happens in anger in PrelArrExts which has a coerce -- case coerce memcpy a b of -- (# r, s #) -> ... -- where the memcpy is in the IO monad, but the call is in -- the (ST s) monad case exprIsConApp_maybe expr of { Nothing -> Nothing ; Just (dc, args) -> case splitTyConApp_maybe to_ty of { Nothing -> Nothing ; Just (tc, tc_arg_tys) | tc /= dataConTyCon dc -> Nothing | not (isVanillaDataCon dc) -> Nothing | otherwise -> -- Type constructor must match -- We knock out existentials to keep matters simple(r) let arity = tyConArity tc val_args = drop arity args to_arg_tys = dataConInstArgTys dc tc_arg_tys mk_coerce ty arg = mkCoerce ty arg new_val_args = zipWith mk_coerce to_arg_tys val_args in ASSERT( all isTypeArg (take arity args) ) ASSERT( equalLength val_args to_arg_tys ) Just (dc, map Type tc_arg_tys ++ new_val_args) }} exprIsConApp_maybe (Note _ expr) = exprIsConApp_maybe expr -- We ignore InlineMe notes in case we have -- x = __inline_me__ (a,b) -- All part of making sure that INLINE pragmas never hurt -- Marcin tripped on this one when making dictionaries more inlinable -- -- In fact, we ignore all notes. For example, -- case _scc_ "foo" (C a b) of -- C a b -> e -- should be optimised away, but it will be only if we look -- through the SCC note. exprIsConApp_maybe expr = analyse (collectArgs expr) where analyse (Var fun, args) | Just con <- isDataConWorkId_maybe fun, args `lengthAtLeast` dataConRepArity con -- Might be > because the arity excludes type args = Just (con,args) -- Look through unfoldings, but only cheap ones, because -- we are effectively duplicating the unfolding analyse (Var fun, []) | let unf = idUnfolding fun, isCheapUnfolding unf = exprIsConApp_maybe (unfoldingTemplate unf) analyse other = Nothing \end{code} %************************************************************************ %* * \subsection{Eta reduction and expansion} %* * %************************************************************************ \begin{code} exprEtaExpandArity :: CoreExpr -> Arity {- The Arity returned is the number of value args the thing can be applied to without doing much work exprEtaExpandArity is used when eta expanding e ==> \xy -> e x y It returns 1 (or more) to: case x of p -> \s -> ... because for I/O ish things we really want to get that \s to the top. We are prepared to evaluate x each time round the loop in order to get that It's all a bit more subtle than it looks: 1. One-shot lambdas Consider one-shot lambdas let x = expensive in \y z -> E We want this to have arity 2 if the \y-abstraction is a 1-shot lambda Hence the ArityType returned by arityType 2. The state-transformer hack The one-shot lambda special cause is particularly important/useful for IO state transformers, where we often get let x = E in \ s -> ... and the \s is a real-world state token abstraction. Such abstractions are almost invariably 1-shot, so we want to pull the \s out, past the let x=E, even if E is expensive. So we treat state-token lambdas as one-shot even if they aren't really. The hack is in Id.isOneShotBndr. 3. Dealing with bottom Consider also f = \x -> error "foo" Here, arity 1 is fine. But if it is f = \x -> case x of True -> error "foo" False -> \y -> x+y then we want to get arity 2. Tecnically, this isn't quite right, because (f True) `seq` 1 should diverge, but it'll converge if we eta-expand f. Nevertheless, we do so; it improves some programs significantly, and increasing convergence isn't a bad thing. Hence the ABot/ATop in ArityType. Actually, the situation is worse. Consider f = \x -> case x of True -> \y -> x+y False -> \y -> x-y Can we eta-expand here? At first the answer looks like "yes of course", but consider (f bot) `seq` 1 This should diverge! But if we eta-expand, it won't. Again, we ignore this "problem", because being scrupulous would lose an important transformation for many programs. 4. Newtypes Non-recursive newtypes are transparent, and should not get in the way. We do (currently) eta-expand recursive newtypes too. So if we have, say newtype T = MkT ([T] -> Int) Suppose we have e = coerce T f where f has arity 1. Then: etaExpandArity e = 1; that is, etaExpandArity looks through the coerce. When we eta-expand e to arity 1: eta_expand 1 e T we want to get: coerce T (\x::[T] -> (coerce ([T]->Int) e) x) HOWEVER, note that if you use coerce bogusly you can ge coerce Int negate And since negate has arity 2, you might try to eta expand. But you can't decopose Int to a function type. Hence the final case in eta_expand. -} exprEtaExpandArity e = arityDepth (arityType e) -- A limited sort of function type data ArityType = AFun Bool ArityType -- True <=> one-shot | ATop -- Know nothing | ABot -- Diverges arityDepth :: ArityType -> Arity arityDepth (AFun _ ty) = 1 + arityDepth ty arityDepth ty = 0 andArityType ABot at2 = at2 andArityType ATop at2 = ATop andArityType (AFun t1 at1) (AFun t2 at2) = AFun (t1 && t2) (andArityType at1 at2) andArityType at1 at2 = andArityType at2 at1 arityType :: CoreExpr -> ArityType -- (go1 e) = [b1,..,bn] -- means expression can be rewritten \x_b1 -> ... \x_bn -> body -- where bi is True <=> the lambda is one-shot arityType (Note n e) = arityType e -- Not needed any more: etaExpand is cleverer -- | ok_note n = arityType e -- | otherwise = ATop arityType (Var v) = mk (idArity v) (arg_tys (idType v)) where mk :: Arity -> [Type] -> ArityType -- The argument types are only to steer the "state hack" -- Consider case x of -- True -> foo -- False -> \(s:RealWorld) -> e -- where foo has arity 1. Then we want the state hack to -- apply to foo too, so we can eta expand the case. mk 0 tys | isBottomingId v = ABot | otherwise = ATop mk n (ty:tys) = AFun (isStateHackType ty) (mk (n-1) tys) mk n [] = AFun False (mk (n-1) []) arg_tys :: Type -> [Type] -- Ignore for-alls arg_tys ty | Just (_, ty') <- splitForAllTy_maybe ty = arg_tys ty' | Just (arg,res) <- splitFunTy_maybe ty = arg : arg_tys res | otherwise = [] -- Lambdas; increase arity arityType (Lam x e) | isId x = AFun (isOneShotBndr x) (arityType e) | otherwise = arityType e -- Applications; decrease arity arityType (App f (Type _)) = arityType f arityType (App f a) = case arityType f of AFun one_shot xs | exprIsCheap a -> xs other -> ATop -- Case/Let; keep arity if either the expression is cheap -- or it's a 1-shot lambda -- The former is not really right for Haskell -- f x = case x of { (a,b) -> \y. e } -- ===> -- f x y = case x of { (a,b) -> e } -- The difference is observable using 'seq' arityType (Case scrut _ _ alts) = case foldr1 andArityType [arityType rhs | (_,_,rhs) <- alts] of xs@(AFun one_shot _) | one_shot -> xs xs | exprIsCheap scrut -> xs | otherwise -> ATop arityType (Let b e) = case arityType e of xs@(AFun one_shot _) | one_shot -> xs xs | all exprIsCheap (rhssOfBind b) -> xs | otherwise -> ATop arityType other = ATop {- NOT NEEDED ANY MORE: etaExpand is cleverer ok_note InlineMe = False ok_note other = True -- Notice that we do not look through __inline_me__ -- This may seem surprising, but consider -- f = _inline_me (\x -> e) -- We DO NOT want to eta expand this to -- f = \x -> (_inline_me (\x -> e)) x -- because the _inline_me gets dropped now it is applied, -- giving just -- f = \x -> e -- A Bad Idea -} \end{code} \begin{code} etaExpand :: Arity -- Result should have this number of value args -> [Unique] -> CoreExpr -> Type -- Expression and its type -> CoreExpr -- (etaExpand n us e ty) returns an expression with -- the same meaning as 'e', but with arity 'n'. -- -- Given e' = etaExpand n us e ty -- We should have -- ty = exprType e = exprType e' -- -- Note that SCCs are not treated specially. If we have -- etaExpand 2 (\x -> scc "foo" e) -- = (\xy -> (scc "foo" e) y) -- So the costs of evaluating 'e' (not 'e y') are attributed to "foo" etaExpand n us expr ty | manifestArity expr >= n = expr -- The no-op case | otherwise = eta_expand n us expr ty where -- manifestArity sees how many leading value lambdas there are manifestArity :: CoreExpr -> Arity manifestArity (Lam v e) | isId v = 1 + manifestArity e | otherwise = manifestArity e manifestArity (Note _ e) = manifestArity e manifestArity e = 0 -- etaExpand deals with for-alls. For example: -- etaExpand 1 E -- where E :: forall a. a -> a -- would return -- (/\b. \y::a -> E b y) -- -- It deals with coerces too, though they are now rare -- so perhaps the extra code isn't worth it eta_expand n us expr ty | n == 0 && -- The ILX code generator requires eta expansion for type arguments -- too, but alas the 'n' doesn't tell us how many of them there -- may be. So we eagerly eta expand any big lambdas, and just -- cross our fingers about possible loss of sharing in the ILX case. -- The Right Thing is probably to make 'arity' include -- type variables throughout the compiler. (ToDo.) not (isForAllTy ty) -- Saturated, so nothing to do = expr -- Short cut for the case where there already -- is a lambda; no point in gratuitously adding more eta_expand n us (Lam v body) ty | isTyVar v = Lam v (eta_expand n us body (applyTy ty (mkTyVarTy v))) | otherwise = Lam v (eta_expand (n-1) us body (funResultTy ty)) -- We used to have a special case that stepped inside Coerces here, -- thus: eta_expand n us (Note note@(Coerce _ ty) e) _ -- = Note note (eta_expand n us e ty) -- BUT this led to an infinite loop -- Example: newtype T = MkT (Int -> Int) -- eta_expand 1 (coerce (Int->Int) e) -- --> coerce (Int->Int) (eta_expand 1 T e) -- by the bogus eqn -- --> coerce (Int->Int) (coerce T -- (\x::Int -> eta_expand 1 (coerce (Int->Int) e))) -- by the splitNewType_maybe case below -- and round we go eta_expand n us expr ty = case splitForAllTy_maybe ty of { Just (tv,ty') -> Lam tv (eta_expand n us (App expr (Type (mkTyVarTy tv))) ty') ; Nothing -> case splitFunTy_maybe ty of { Just (arg_ty, res_ty) -> Lam arg1 (eta_expand (n-1) us2 (App expr (Var arg1)) res_ty) where arg1 = mkSysLocal FSLIT("eta") uniq arg_ty (uniq:us2) = us ; Nothing -> -- Given this: -- newtype T = MkT ([T] -> Int) -- Consider eta-expanding this -- eta_expand 1 e T -- We want to get -- coerce T (\x::[T] -> (coerce ([T]->Int) e) x) -- Only try this for recursive newtypes; the non-recursive kind -- are transparent anyway case splitRecNewType_maybe ty of { Just ty' -> mkCoerce2 ty ty' (eta_expand n us (mkCoerce2 ty' ty expr) ty') ; Nothing -> -- We have an expression of arity > 0, but its type isn't a function -- This *can* legitmately happen: e.g. coerce Int (\x. x) -- Essentially the programmer is playing fast and loose with types -- (Happy does this a lot). So we simply decline to eta-expand. expr }}} \end{code} exprArity is a cheap-and-cheerful version of exprEtaExpandArity. It tells how many things the expression can be applied to before doing any work. It doesn't look inside cases, lets, etc. The idea is that exprEtaExpandArity will do the hard work, leaving something that's easy for exprArity to grapple with. In particular, Simplify uses exprArity to compute the ArityInfo for the Id. Originally I thought that it was enough just to look for top-level lambdas, but it isn't. I've seen this foo = PrelBase.timesInt We want foo to get arity 2 even though the eta-expander will leave it unchanged, in the expectation that it'll be inlined. But occasionally it isn't, because foo is blacklisted (used in a rule). Similarly, see the ok_note check in exprEtaExpandArity. So f = __inline_me (\x -> e) won't be eta-expanded. And in any case it seems more robust to have exprArity be a bit more intelligent. But note that (\x y z -> f x y z) should have arity 3, regardless of f's arity. \begin{code} exprArity :: CoreExpr -> Arity exprArity e = go e where go (Var v) = idArity v go (Lam x e) | isId x = go e + 1 | otherwise = go e go (Note n e) = go e go (App e (Type t)) = go e go (App f a) | exprIsCheap a = (go f - 1) `max` 0 -- NB: exprIsCheap a! -- f (fac x) does not have arity 2, -- even if f has arity 3! -- NB: `max 0`! (\x y -> f x) has arity 2, even if f is -- unknown, hence arity 0 go _ = 0 \end{code} %************************************************************************ %* * \subsection{Equality} %* * %************************************************************************ @cheapEqExpr@ is a cheap equality test which bales out fast! True => definitely equal False => may or may not be equal \begin{code} cheapEqExpr :: Expr b -> Expr b -> Bool cheapEqExpr (Var v1) (Var v2) = v1==v2 cheapEqExpr (Lit lit1) (Lit lit2) = lit1 == lit2 cheapEqExpr (Type t1) (Type t2) = t1 `coreEqType` t2 cheapEqExpr (App f1 a1) (App f2 a2) = f1 `cheapEqExpr` f2 && a1 `cheapEqExpr` a2 cheapEqExpr _ _ = False exprIsBig :: Expr b -> Bool -- Returns True of expressions that are too big to be compared by cheapEqExpr exprIsBig (Lit _) = False exprIsBig (Var v) = False exprIsBig (Type t) = False exprIsBig (App f a) = exprIsBig f || exprIsBig a exprIsBig other = True \end{code} \begin{code} tcEqExpr :: CoreExpr -> CoreExpr -> Bool -- Used in rule matching, so does *not* look through -- newtypes, predicate types; hence tcEqExpr tcEqExpr e1 e2 = tcEqExprX rn_env e1 e2 where rn_env = mkRnEnv2 (mkInScopeSet (exprFreeVars e1 `unionVarSet` exprFreeVars e2)) tcEqExprX :: RnEnv2 -> CoreExpr -> CoreExpr -> Bool tcEqExprX env (Var v1) (Var v2) = rnOccL env v1 == rnOccR env v2 tcEqExprX env (Lit lit1) (Lit lit2) = lit1 == lit2 tcEqExprX env (App f1 a1) (App f2 a2) = tcEqExprX env f1 f2 && tcEqExprX env a1 a2 tcEqExprX env (Lam v1 e1) (Lam v2 e2) = tcEqExprX (rnBndr2 env v1 v2) e1 e2 tcEqExprX env (Let (NonRec v1 r1) e1) (Let (NonRec v2 r2) e2) = tcEqExprX env r1 r2 && tcEqExprX (rnBndr2 env v1 v2) e1 e2 tcEqExprX env (Let (Rec ps1) e1) (Let (Rec ps2) e2) = equalLength ps1 ps2 && and (zipWith eq_rhs ps1 ps2) && tcEqExprX env' e1 e2 where env' = foldl2 rn_bndr2 env ps2 ps2 rn_bndr2 env (b1,_) (b2,_) = rnBndr2 env b1 b2 eq_rhs (_,r1) (_,r2) = tcEqExprX env' r1 r2 tcEqExprX env (Case e1 v1 t1 a1) (Case e2 v2 t2 a2) = tcEqExprX env e1 e2 && tcEqTypeX env t1 t2 && equalLength a1 a2 && and (zipWith (eq_alt env') a1 a2) where env' = rnBndr2 env v1 v2 tcEqExprX env (Note n1 e1) (Note n2 e2) = eq_note env n1 n2 && tcEqExprX env e1 e2 tcEqExprX env (Type t1) (Type t2) = tcEqTypeX env t1 t2 tcEqExprX env e1 e2 = False eq_alt env (c1,vs1,r1) (c2,vs2,r2) = c1==c2 && tcEqExprX (rnBndrs2 env vs1 vs2) r1 r2 eq_note env (SCC cc1) (SCC cc2) = cc1 == cc2 eq_note env (Coerce t1 f1) (Coerce t2 f2) = tcEqTypeX env t1 t2 && tcEqTypeX env f1 f2 eq_note env InlineCall InlineCall = True eq_note env (CoreNote s1) (CoreNote s2) = s1 == s2 eq_note env other1 other2 = False \end{code} %************************************************************************ %* * \subsection{The size of an expression} %* * %************************************************************************ \begin{code} coreBindsSize :: [CoreBind] -> Int coreBindsSize bs = foldr ((+) . bindSize) 0 bs exprSize :: CoreExpr -> Int -- A measure of the size of the expressions -- It also forces the expression pretty drastically as a side effect exprSize (Var v) = v `seq` 1 exprSize (Lit lit) = lit `seq` 1 exprSize (App f a) = exprSize f + exprSize a exprSize (Lam b e) = varSize b + exprSize e exprSize (Let b e) = bindSize b + exprSize e exprSize (Case e b t as) = seqType t `seq` exprSize e + varSize b + 1 + foldr ((+) . altSize) 0 as exprSize (Note n e) = noteSize n + exprSize e exprSize (Type t) = seqType t `seq` 1 noteSize (SCC cc) = cc `seq` 1 noteSize (Coerce t1 t2) = seqType t1 `seq` seqType t2 `seq` 1 noteSize InlineCall = 1 noteSize InlineMe = 1 noteSize (CoreNote s) = s `seq` 1 -- hdaume: core annotations varSize :: Var -> Int varSize b | isTyVar b = 1 | otherwise = seqType (idType b) `seq` megaSeqIdInfo (idInfo b) `seq` 1 varsSize = foldr ((+) . varSize) 0 bindSize (NonRec b e) = varSize b + exprSize e bindSize (Rec prs) = foldr ((+) . pairSize) 0 prs pairSize (b,e) = varSize b + exprSize e altSize (c,bs,e) = c `seq` varsSize bs + exprSize e \end{code} %************************************************************************ %* * \subsection{Hashing} %* * %************************************************************************ \begin{code} hashExpr :: CoreExpr -> Int hashExpr e | hash < 0 = 77 -- Just in case we hit -maxInt | otherwise = hash where hash = abs (hash_expr e) -- Negative numbers kill UniqFM hash_expr (Note _ e) = hash_expr e hash_expr (Let (NonRec b r) e) = hashId b hash_expr (Let (Rec ((b,r):_)) e) = hashId b hash_expr (Case _ b _ _) = hashId b hash_expr (App f e) = hash_expr f * fast_hash_expr e hash_expr (Var v) = hashId v hash_expr (Lit lit) = hashLiteral lit hash_expr (Lam b _) = hashId b hash_expr (Type t) = trace "hash_expr: type" 1 -- Shouldn't happen fast_hash_expr (Var v) = hashId v fast_hash_expr (Lit lit) = hashLiteral lit fast_hash_expr (App f (Type _)) = fast_hash_expr f fast_hash_expr (App f a) = fast_hash_expr a fast_hash_expr (Lam b _) = hashId b fast_hash_expr other = 1 hashId :: Id -> Int hashId id = hashName (idName id) \end{code} %************************************************************************ %* * \subsection{Determining non-updatable right-hand-sides} %* * %************************************************************************ Top-level constructor applications can usually be allocated statically, but they can't if the constructor, or any of the arguments, come from another DLL (because we can't refer to static labels in other DLLs). If this happens we simply make the RHS into an updatable thunk, and 'exectute' it rather than allocating it statically. \begin{code} rhsIsStatic :: HomeModules -> CoreExpr -> Bool -- This function is called only on *top-level* right-hand sides -- Returns True if the RHS can be allocated statically, with -- no thunks involved at all. -- -- It's called (i) in TidyPgm.hasCafRefs to decide if the rhs is, or -- refers to, CAFs; and (ii) in CoreToStg to decide whether to put an -- update flag on it. -- -- The basic idea is that rhsIsStatic returns True only if the RHS is -- (a) a value lambda -- (b) a saturated constructor application with static args -- -- BUT watch out for -- (i) Any cross-DLL references kill static-ness completely -- because they must be 'executed' not statically allocated -- ("DLL" here really only refers to Windows DLLs, on other platforms, -- this is not necessary) -- -- (ii) We treat partial applications as redexes, because in fact we -- make a thunk for them that runs and builds a PAP -- at run-time. The only appliations that are treated as -- static are *saturated* applications of constructors. -- We used to try to be clever with nested structures like this: -- ys = (:) w ((:) w []) -- on the grounds that CorePrep will flatten ANF-ise it later. -- But supporting this special case made the function much more -- complicated, because the special case only applies if there are no -- enclosing type lambdas: -- ys = /\ a -> Foo (Baz ([] a)) -- Here the nested (Baz []) won't float out to top level in CorePrep. -- -- But in fact, even without -O, nested structures at top level are -- flattened by the simplifier, so we don't need to be super-clever here. -- -- Examples -- -- f = \x::Int. x+7 TRUE -- p = (True,False) TRUE -- -- d = (fst p, False) FALSE because there's a redex inside -- (this particular one doesn't happen but...) -- -- h = D# (1.0## /## 2.0##) FALSE (redex again) -- n = /\a. Nil a TRUE -- -- t = /\a. (:) (case w a of ...) (Nil a) FALSE (redex) -- -- -- This is a bit like CoreUtils.exprIsHNF, with the following differences: -- a) scc "foo" (\x -> ...) is updatable (so we catch the right SCC) -- -- b) (C x xs), where C is a contructors is updatable if the application is -- dynamic -- -- c) don't look through unfolding of f in (f x). -- -- When opt_RuntimeTypes is on, we keep type lambdas and treat -- them as making the RHS re-entrant (non-updatable). rhsIsStatic hmods rhs = is_static False rhs where is_static :: Bool -- True <=> in a constructor argument; must be atomic -> CoreExpr -> Bool is_static False (Lam b e) = isRuntimeVar b || is_static False e is_static in_arg (Note (SCC _) e) = False is_static in_arg (Note _ e) = is_static in_arg e is_static in_arg (Lit lit) = case lit of MachLabel _ _ -> False other -> True -- A MachLabel (foreign import "&foo") in an argument -- prevents a constructor application from being static. The -- reason is that it might give rise to unresolvable symbols -- in the object file: under Linux, references to "weak" -- symbols from the data segment give rise to "unresolvable -- relocation" errors at link time This might be due to a bug -- in the linker, but we'll work around it here anyway. -- SDM 24/2/2004 is_static in_arg other_expr = go other_expr 0 where go (Var f) n_val_args #if mingw32_TARGET_OS | not (isDllName hmods (idName f)) #endif = saturated_data_con f n_val_args || (in_arg && n_val_args == 0) -- A naked un-applied variable is *not* deemed a static RHS -- E.g. f = g -- Reason: better to update so that the indirection gets shorted -- out, and the true value will be seen -- NB: if you change this, you'll break the invariant that THUNK_STATICs -- are always updatable. If you do so, make sure that non-updatable -- ones have enough space for their static link field! go (App f a) n_val_args | isTypeArg a = go f n_val_args | not in_arg && is_static True a = go f (n_val_args + 1) -- The (not in_arg) checks that we aren't in a constructor argument; -- if we are, we don't allow (value) applications of any sort -- -- NB. In case you wonder, args are sometimes not atomic. eg. -- x = D# (1.0## /## 2.0##) -- can't float because /## can fail. go (Note (SCC _) f) n_val_args = False go (Note _ f) n_val_args = go f n_val_args go other n_val_args = False saturated_data_con f n_val_args = case isDataConWorkId_maybe f of Just dc -> n_val_args == dataConRepArity dc Nothing -> False \end{code}