X-Git-Url: http://git.megacz.com/?a=blobdiff_plain;f=ghc%2Fcompiler%2FsimplCore%2FSimplUtils.lhs;h=0d9be520fe2849566c99d5e851a4204e800299e7;hb=b16992d66aa5f610de586eb8a720214b8065bd65;hp=906568bcc54744ab3494de68581da601f78be378;hpb=43ff61950ef701fc7d6e8308fa7c5256f82f8dc5;p=ghc-hetmet.git diff --git a/ghc/compiler/simplCore/SimplUtils.lhs b/ghc/compiler/simplCore/SimplUtils.lhs index 906568b..0d9be52 100644 --- a/ghc/compiler/simplCore/SimplUtils.lhs +++ b/ghc/compiler/simplCore/SimplUtils.lhs @@ -5,46 +5,55 @@ \begin{code} module SimplUtils ( - simplBinder, simplBinders, simplRecBndrs, simplLetBndr, - simplLamBndrs, simplTopBndrs, - newId, mkLam, mkCase, + mkLam, prepareAlts, mkCase, + + -- Inlining, + preInlineUnconditionally, postInlineUnconditionally, activeInline, activeRule, + inlineMode, -- The continuation type SimplCont(..), DupFlag(..), LetRhsFlag(..), contIsDupable, contResultType, countValArgs, countArgs, pushContArgs, - mkBoringStop, mkStop, contIsRhs, contIsRhsOrArg, + mkBoringStop, mkRhsStop, contIsRhs, contIsRhsOrArg, getContArgs, interestingCallContext, interestingArg, isStrictType ) where #include "HsVersions.h" -import CmdLineOpts ( SimplifierSwitch(..), - opt_SimplDoLambdaEtaExpansion, opt_SimplDoEtaReduction, - opt_SimplCaseMerge, opt_UF_UpdateInPlace - ) +import SimplEnv +import DynFlags ( SimplifierSwitch(..), SimplifierMode(..), + DynFlag(..), dopt ) +import StaticFlags ( opt_UF_UpdateInPlace, opt_SimplNoPreInlining, + opt_RulesOff ) + import CoreSyn -import CoreUtils ( cheapEqExpr, exprType, - etaExpand, exprEtaExpandArity, bindNonRec, mkCoerce, - findDefault, exprOkForSpeculation, exprIsValue +import CoreFVs ( exprFreeVars ) +import CoreUtils ( cheapEqExpr, exprType, exprIsTrivial, exprIsCheap, + etaExpand, exprEtaExpandArity, bindNonRec, mkCoerce2, + findDefault, exprOkForSpeculation, exprIsHNF ) -import qualified Subst ( simplBndrs, simplBndr, simplLetId, simplLamBndr ) -import Id ( Id, idType, idInfo, isLocalId, - mkSysLocal, hasNoBinding, isDeadBinder, idNewDemandInfo, - idUnfolding, idNewStrictness +import Literal ( mkStringLit ) +import CoreUnfold ( smallEnoughToInline ) +import MkId ( eRROR_ID ) +import Id ( idType, isDataConWorkId, idOccInfo, isDictId, + mkSysLocal, isDeadBinder, idNewDemandInfo, isExportedId, + idUnfolding, idNewStrictness, idInlinePragma, ) import NewDemand ( isStrictDmd, isBotRes, splitStrictSig ) import SimplMonad -import Type ( Type, seqType, - splitTyConApp_maybe, tyConAppArgs, mkTyVarTys, - splitRepFunTys, isStrictType +import Type ( Type, splitFunTys, dropForAlls, isStrictType, + splitTyConApp_maybe, tyConAppArgs, mkTyVarTys ) -import OccName ( UserFS ) -import TyCon ( tyConDataConsIfAvailable, isAlgTyCon, isNewTyCon ) -import DataCon ( dataConRepArity, dataConSig, dataConArgTys ) -import Var ( mkSysTyVar, tyVarKind ) -import Util ( lengthExceeds, mapAccumL ) +import Name ( mkSysTvName ) +import TyCon ( tyConDataCons_maybe, isAlgTyCon, isNewTyCon ) +import DataCon ( dataConRepArity, dataConTyVars, dataConInstArgTys, isVanillaDataCon ) +import Var ( tyVarKind, mkTyVar ) +import VarSet +import BasicTypes ( TopLevelFlag(..), isNotTopLevel, OccInfo(..), isLoopBreaker, isOneOcc, + Activation, isAlwaysActive, isActive ) +import Util ( lengthExceeds ) import Outputable \end{code} @@ -77,14 +86,16 @@ data SimplCont -- Strict contexts InId [InAlt] SimplEnv -- The case binder, alts, and subst-env SimplCont - | ArgOf DupFlag -- An arbitrary strict context: the argument + | ArgOf LetRhsFlag -- An arbitrary strict context: the argument -- of a strict function, or a primitive-arg fn -- or a PrimOp - LetRhsFlag + -- No DupFlag because we never duplicate it + OutType -- arg_ty: type of the argument itself OutType -- cont_ty: the type of the expression being sought by the context -- f (error "foo") ==> coerce t (error "foo") -- when f is strict -- We need to know the type t, to which to coerce. + (SimplEnv -> OutExpr -> SimplM FloatsWithExpr) -- What to do with the result -- The result expression in the OutExprStuff has type cont_ty @@ -98,7 +109,7 @@ instance Outputable LetRhsFlag where instance Outputable SimplCont where ppr (Stop _ is_rhs _) = ptext SLIT("Stop") <> brackets (ppr is_rhs) ppr (ApplyTo dup arg se cont) = (ptext SLIT("ApplyTo") <+> ppr dup <+> ppr arg) $$ ppr cont - ppr (ArgOf dup _ _ _) = ptext SLIT("ArgOf...") <+> ppr dup + ppr (ArgOf _ _ _ _) = ptext SLIT("ArgOf...") ppr (Select dup bndr alts se cont) = (ptext SLIT("Select") <+> ppr dup <+> ppr bndr) $$ (nest 4 (ppr alts)) $$ ppr cont ppr (CoerceIt ty cont) = (ptext SLIT("CoerceIt") <+> ppr ty) $$ ppr cont @@ -112,15 +123,13 @@ instance Outputable DupFlag where ------------------- -mkBoringStop :: OutType -> SimplCont +mkBoringStop, mkRhsStop :: OutType -> SimplCont mkBoringStop ty = Stop ty AnArg (canUpdateInPlace ty) - -mkStop :: OutType -> LetRhsFlag -> SimplCont -mkStop ty is_rhs = Stop ty is_rhs (canUpdateInPlace ty) +mkRhsStop ty = Stop ty AnRhs (canUpdateInPlace ty) contIsRhs :: SimplCont -> Bool contIsRhs (Stop _ AnRhs _) = True -contIsRhs (ArgOf _ AnRhs _ _) = True +contIsRhs (ArgOf AnRhs _ _ _) = True contIsRhs other = False contIsRhsOrArg (Stop _ _ _) = True @@ -131,11 +140,10 @@ contIsRhsOrArg other = False contIsDupable :: SimplCont -> Bool contIsDupable (Stop _ _ _) = True contIsDupable (ApplyTo OkToDup _ _ _) = True -contIsDupable (ArgOf OkToDup _ _ _) = True contIsDupable (Select OkToDup _ _ _ _) = True contIsDupable (CoerceIt _ cont) = contIsDupable cont -contIsDupable (InlinePlease cont) = contIsDupable cont -contIsDupable other = False +contIsDupable (InlinePlease cont) = contIsDupable cont +contIsDupable other = False ------------------- discardableCont :: SimplCont -> Bool @@ -221,6 +229,9 @@ getContArgs chkr fun orig_cont -- * (error "Hello") arg -- * f (error "Hello") where f is strict -- etc + -- Then, especially in the first of these cases, we'd like to discard + -- the continuation, leaving just the bottoming expression. But the + -- type might not be right, so we may have to add a coerce. go acc ss inl cont | null ss && discardableCont cont = (reverse acc, discardCont cont, inl) | otherwise = (reverse acc, cont, inl) @@ -231,14 +242,14 @@ getContArgs chkr fun orig_cont computed_stricts = zipWith (||) fun_stricts arg_stricts ---------------------------- - (val_arg_tys, _) = splitRepFunTys (idType fun) + (val_arg_tys, _) = splitFunTys (dropForAlls (idType fun)) arg_stricts = map isStrictType val_arg_tys ++ repeat False -- These argument types are used as a cheap and cheerful way to find -- unboxed arguments, which must be strict. But it's an InType -- and so there might be a type variable where we expect a function -- type (the substitution hasn't happened yet). And we don't bother -- doing the type applications for a polymorphic function. - -- Hence the split*Rep*FunTys + -- Hence the splitFunTys*IgnoringForAlls* ---------------------------- -- If fun_stricts is finite, it means the function returns bottom @@ -272,6 +283,9 @@ interestingArg :: OutExpr -> Bool interestingArg (Var v) = hasSomeUnfolding (idUnfolding v) -- Was: isValueUnfolding (idUnfolding v') -- But that seems over-pessimistic + || isDataConWorkId v + -- This accounts for an argument like + -- () or [], which is definitely interesting interestingArg (Type _) = False interestingArg (App fn (Type _)) = interestingArg fn interestingArg (Note _ a) = interestingArg a @@ -364,9 +378,9 @@ interestingCallContext :: Bool -- False <=> no args at all interestingCallContext some_args some_val_args cont = interesting cont where - interesting (InlinePlease _) = True - interesting (Select _ _ _ _ _) = some_args - interesting (ApplyTo _ _ _ _) = True -- Can happen if we have (coerce t (f x)) y + interesting (InlinePlease _) = True + interesting (Select _ _ _ _ _) = some_args + interesting (ApplyTo _ _ _ _) = True -- Can happen if we have (coerce t (f x)) y -- Perhaps True is a bit over-keen, but I've -- seen (coerce f) x, where f has an INLINE prag, -- So we have to give some motivaiton for inlining it @@ -404,10 +418,10 @@ canUpdateInPlace ty | otherwise = case splitTyConApp_maybe ty of Nothing -> False - Just (tycon, _) -> case tyConDataConsIfAvailable tycon of - [dc] -> arity == 1 || arity == 2 - where - arity = dataConRepArity dc + Just (tycon, _) -> case tyConDataCons_maybe tycon of + Just [dc] -> arity == 1 || arity == 2 + where + arity = dataConRepArity dc other -> False \end{code} @@ -415,81 +429,343 @@ canUpdateInPlace ty %************************************************************************ %* * -\section{Dealing with a single binder} +\subsection{Decisions about inlining} %* * %************************************************************************ -These functions are in the monad only so that they can be made strict via seq. +Inlining is controlled partly by the SimplifierMode switch. This has two +settings: -\begin{code} -simplBinders :: SimplEnv -> [InBinder] -> SimplM (SimplEnv, [OutBinder]) -simplBinders env bndrs - = let - (subst', bndrs') = Subst.simplBndrs (getSubst env) bndrs - in - seqBndrs bndrs' `seq` - returnSmpl (setSubst env subst', bndrs') + SimplGently (a) Simplifying before specialiser/full laziness + (b) Simplifiying inside INLINE pragma + (c) Simplifying the LHS of a rule + (d) Simplifying a GHCi expression or Template + Haskell splice -simplBinder :: SimplEnv -> InBinder -> SimplM (SimplEnv, OutBinder) -simplBinder env bndr - = let - (subst', bndr') = Subst.simplBndr (getSubst env) bndr - in - seqBndr bndr' `seq` - returnSmpl (setSubst env subst', bndr') + SimplPhase n Used at all other times +The key thing about SimplGently is that it does no call-site inlining. +Before full laziness we must be careful not to inline wrappers, +because doing so inhibits floating + e.g. ...(case f x of ...)... + ==> ...(case (case x of I# x# -> fw x#) of ...)... + ==> ...(case x of I# x# -> case fw x# of ...)... +and now the redex (f x) isn't floatable any more. -simplLetBndr :: SimplEnv -> InBinder -> SimplM (SimplEnv, OutBinder) -simplLetBndr env id - = let - (subst', id') = Subst.simplLetId (getSubst env) id - in - seqBndr id' `seq` - returnSmpl (setSubst env subst', id') - -simplTopBndrs, simplLamBndrs, simplRecBndrs - :: SimplEnv -> [InBinder] -> SimplM (SimplEnv, [OutBinder]) -simplTopBndrs = simplBndrs simplTopBinder -simplRecBndrs = simplBndrs Subst.simplLetId -simplLamBndrs = simplBndrs Subst.simplLamBndr - --- For top-level binders, don't use simplLetId for GlobalIds. --- There are some of these, notably consructor wrappers, and we don't --- want to clone them or fiddle with them at all. --- Rather tiresomely, the specialiser may float a use of a constructor --- wrapper to before its definition (which shouldn't really matter) --- because it doesn't see the constructor wrapper as free in the binding --- it is floating (because it's a GlobalId). --- Then the simplifier brings all top level Ids into scope at the --- beginning, and we don't want to lose the IdInfo on the constructor --- wrappers. It would also be Bad to clone it! -simplTopBinder subst bndr - | isLocalId bndr = Subst.simplLetId subst bndr - | otherwise = (subst, bndr) - -simplBndrs simpl_bndr env bndrs - = let - (subst', bndrs') = mapAccumL simpl_bndr (getSubst env) bndrs - in - seqBndrs bndrs' `seq` - returnSmpl (setSubst env subst', bndrs') +The no-inling thing is also important for Template Haskell. You might be +compiling in one-shot mode with -O2; but when TH compiles a splice before +running it, we don't want to use -O2. Indeed, we don't want to inline +anything, because the byte-code interpreter might get confused about +unboxed tuples and suchlike. -seqBndrs [] = () -seqBndrs (b:bs) = seqBndr b `seq` seqBndrs bs +INLINE pragmas +~~~~~~~~~~~~~~ +SimplGently is also used as the mode to simplify inside an InlineMe note. -seqBndr b | isTyVar b = b `seq` () - | otherwise = seqType (idType b) `seq` - idInfo b `seq` - () +\begin{code} +inlineMode :: SimplifierMode +inlineMode = SimplGently \end{code} +It really is important to switch off inlinings inside such +expressions. Consider the following example + + let f = \pq -> BIG + in + let g = \y -> f y y + {-# INLINE g #-} + in ...g...g...g...g...g... + +Now, if that's the ONLY occurrence of f, it will be inlined inside g, +and thence copied multiple times when g is inlined. + + +This function may be inlinined in other modules, so we +don't want to remove (by inlining) calls to functions that have +specialisations, or that may have transformation rules in an importing +scope. + +E.g. {-# INLINE f #-} + f x = ...g... + +and suppose that g is strict *and* has specialisations. If we inline +g's wrapper, we deny f the chance of getting the specialised version +of g when f is inlined at some call site (perhaps in some other +module). + +It's also important not to inline a worker back into a wrapper. +A wrapper looks like + wraper = inline_me (\x -> ...worker... ) +Normally, the inline_me prevents the worker getting inlined into +the wrapper (initially, the worker's only call site!). But, +if the wrapper is sure to be called, the strictness analyser will +mark it 'demanded', so when the RHS is simplified, it'll get an ArgOf +continuation. That's why the keep_inline predicate returns True for +ArgOf continuations. It shouldn't do any harm not to dissolve the +inline-me note under these circumstances. + +Note that the result is that we do very little simplification +inside an InlineMe. + + all xs = foldr (&&) True xs + any p = all . map p {-# INLINE any #-} + +Problem: any won't get deforested, and so if it's exported and the +importer doesn't use the inlining, (eg passes it as an arg) then we +won't get deforestation at all. We havn't solved this problem yet! + + +preInlineUnconditionally +~~~~~~~~~~~~~~~~~~~~~~~~ +@preInlineUnconditionally@ examines a bndr to see if it is used just +once in a completely safe way, so that it is safe to discard the +binding inline its RHS at the (unique) usage site, REGARDLESS of how +big the RHS might be. If this is the case we don't simplify the RHS +first, but just inline it un-simplified. + +This is much better than first simplifying a perhaps-huge RHS and then +inlining and re-simplifying it. Indeed, it can be at least quadratically +better. Consider + + x1 = e1 + x2 = e2[x1] + x3 = e3[x2] + ...etc... + xN = eN[xN-1] + +We may end up simplifying e1 N times, e2 N-1 times, e3 N-3 times etc. +This can happen with cascades of functions too: + + f1 = \x1.e1 + f2 = \xs.e2[f1] + f3 = \xs.e3[f3] + ...etc... + +THE MAIN INVARIANT is this: + + ---- preInlineUnconditionally invariant ----- + IF preInlineUnconditionally chooses to inline x = + THEN doing the inlining should not change the occurrence + info for the free vars of + ---------------------------------------------- + +For example, it's tempting to look at trivial binding like + x = y +and inline it unconditionally. But suppose x is used many times, +but this is the unique occurrence of y. Then inlining x would change +y's occurrence info, which breaks the invariant. It matters: y +might have a BIG rhs, which will now be dup'd at every occurrenc of x. + + +Evne RHSs labelled InlineMe aren't caught here, because there might be +no benefit from inlining at the call site. + +[Sept 01] Don't unconditionally inline a top-level thing, because that +can simply make a static thing into something built dynamically. E.g. + x = (a,b) + main = \s -> h x + +[Remember that we treat \s as a one-shot lambda.] No point in +inlining x unless there is something interesting about the call site. + +But watch out: if you aren't careful, some useful foldr/build fusion +can be lost (most notably in spectral/hartel/parstof) because the +foldr didn't see the build. Doing the dynamic allocation isn't a big +deal, in fact, but losing the fusion can be. But the right thing here +seems to be to do a callSiteInline based on the fact that there is +something interesting about the call site (it's strict). Hmm. That +seems a bit fragile. + +Conclusion: inline top level things gaily until Phase 0 (the last +phase), at which point don't. \begin{code} -newId :: UserFS -> Type -> SimplM Id -newId fs ty = getUniqueSmpl `thenSmpl` \ uniq -> - returnSmpl (mkSysLocal fs uniq ty) +preInlineUnconditionally :: SimplEnv -> TopLevelFlag -> InId -> InExpr -> Bool +preInlineUnconditionally env top_lvl bndr rhs + | not active = False + | opt_SimplNoPreInlining = False + | otherwise = case idOccInfo bndr of + IAmDead -> True -- Happens in ((\x.1) v) + OneOcc in_lam True int_cxt -> try_once in_lam int_cxt + other -> False + where + phase = getMode env + active = case phase of + SimplGently -> isAlwaysActive prag + SimplPhase n -> isActive n prag + prag = idInlinePragma bndr + + try_once in_lam int_cxt -- There's one textual occurrence + | not in_lam = isNotTopLevel top_lvl || early_phase + | otherwise = int_cxt && canInlineInLam rhs + +-- Be very careful before inlining inside a lambda, becuase (a) we must not +-- invalidate occurrence information, and (b) we want to avoid pushing a +-- single allocation (here) into multiple allocations (inside lambda). +-- Inlining a *function* with a single *saturated* call would be ok, mind you. +-- || (if is_cheap && not (canInlineInLam rhs) then pprTrace "preinline" (ppr bndr <+> ppr rhs) ok else ok) +-- where +-- is_cheap = exprIsCheap rhs +-- ok = is_cheap && int_cxt + + -- int_cxt The context isn't totally boring + -- E.g. let f = \ab.BIG in \y. map f xs + -- Don't want to substitute for f, because then we allocate + -- its closure every time the \y is called + -- But: let f = \ab.BIG in \y. map (f y) xs + -- Now we do want to substitute for f, even though it's not + -- saturated, because we're going to allocate a closure for + -- (f y) every time round the loop anyhow. + + -- canInlineInLam => free vars of rhs are (Once in_lam) or Many, + -- so substituting rhs inside a lambda doesn't change the occ info. + -- Sadly, not quite the same as exprIsHNF. + canInlineInLam (Lit l) = True + canInlineInLam (Lam b e) = isRuntimeVar b || canInlineInLam e + canInlineInLam (Note _ e) = canInlineInLam e + canInlineInLam _ = False + + early_phase = case phase of + SimplPhase 0 -> False + other -> True +-- If we don't have this early_phase test, consider +-- x = length [1,2,3] +-- The full laziness pass carefully floats all the cons cells to +-- top level, and preInlineUnconditionally floats them all back in. +-- Result is (a) static allocation replaced by dynamic allocation +-- (b) many simplifier iterations because this tickles +-- a related problem; only one inlining per pass +-- +-- On the other hand, I have seen cases where top-level fusion is +-- lost if we don't inline top level thing (e.g. string constants) +-- Hence the test for phase zero (which is the phase for all the final +-- simplifications). Until phase zero we take no special notice of +-- top level things, but then we become more leery about inlining +-- them. + \end{code} +postInlineUnconditionally +~~~~~~~~~~~~~~~~~~~~~~~~~ +@postInlineUnconditionally@ decides whether to unconditionally inline +a thing based on the form of its RHS; in particular if it has a +trivial RHS. If so, we can inline and discard the binding altogether. + +NB: a loop breaker has must_keep_binding = True and non-loop-breakers +only have *forward* references Hence, it's safe to discard the binding + +NOTE: This isn't our last opportunity to inline. We're at the binding +site right now, and we'll get another opportunity when we get to the +ocurrence(s) + +Note that we do this unconditional inlining only for trival RHSs. +Don't inline even WHNFs inside lambdas; doing so may simply increase +allocation when the function is called. This isn't the last chance; see +NOTE above. + +NB: Even inline pragmas (e.g. IMustBeINLINEd) are ignored here Why? +Because we don't even want to inline them into the RHS of constructor +arguments. See NOTE above + +NB: At one time even NOINLINE was ignored here: if the rhs is trivial +it's best to inline it anyway. We often get a=E; b=a from desugaring, +with both a and b marked NOINLINE. But that seems incompatible with +our new view that inlining is like a RULE, so I'm sticking to the 'active' +story for now. + +\begin{code} +postInlineUnconditionally :: SimplEnv -> TopLevelFlag -> OutId -> OccInfo -> OutExpr -> Unfolding -> Bool +postInlineUnconditionally env top_lvl bndr occ_info rhs unfolding + | not active = False + | isLoopBreaker occ_info = False + | isExportedId bndr = False + | exprIsTrivial rhs = True + | otherwise + = case occ_info of + OneOcc in_lam one_br int_cxt + -> (one_br || smallEnoughToInline unfolding) -- Small enough to dup + -- ToDo: consider discount on smallEnoughToInline if int_cxt is true + -- + -- NB: Do we want to inline arbitrarily big things becuase + -- one_br is True? that can lead to inline cascades. But + -- preInlineUnconditionlly has dealt with all the common cases + -- so perhaps it's worth the risk. Here's an example + -- let f = if b then Left (\x.BIG) else Right (\y.BIG) + -- in \y. ....f.... + -- We can't preInlineUnconditionally because that woud invalidate + -- the occ info for b. Yet f is used just once, and duplicating + -- the case work is fine (exprIsCheap). + + && ((isNotTopLevel top_lvl && not in_lam) || + -- But outside a lambda, we want to be reasonably aggressive + -- about inlining into multiple branches of case + -- e.g. let x = + -- in case y of { C1 -> ..x..; C2 -> ..x..; C3 -> ... } + -- Inlining can be a big win if C3 is the hot-spot, even if + -- the uses in C1, C2 are not 'interesting' + -- An example that gets worse if you add int_cxt here is 'clausify' + + (isCheapUnfolding unfolding && int_cxt)) + -- isCheap => acceptable work duplication; in_lam may be true + -- int_cxt to prevent us inlining inside a lambda without some + -- good reason. See the notes on int_cxt in preInlineUnconditionally + + other -> False + -- The point here is that for *non-values* that occur + -- outside a lambda, the call-site inliner won't have + -- a chance (becuase it doesn't know that the thing + -- only occurs once). The pre-inliner won't have gotten + -- it either, if the thing occurs in more than one branch + -- So the main target is things like + -- let x = f y in + -- case v of + -- True -> case x of ... + -- False -> case x of ... + -- I'm not sure how important this is in practice + where + active = case getMode env of + SimplGently -> isAlwaysActive prag + SimplPhase n -> isActive n prag + prag = idInlinePragma bndr + +activeInline :: SimplEnv -> OutId -> OccInfo -> Bool +activeInline env id occ + = case getMode env of + SimplGently -> isOneOcc occ && isAlwaysActive prag + -- No inlining at all when doing gentle stuff, + -- except for local things that occur once + -- The reason is that too little clean-up happens if you + -- don't inline use-once things. Also a bit of inlining is *good* for + -- full laziness; it can expose constant sub-expressions. + -- Example in spectral/mandel/Mandel.hs, where the mandelset + -- function gets a useful let-float if you inline windowToViewport + + -- NB: we used to have a second exception, for data con wrappers. + -- On the grounds that we use gentle mode for rule LHSs, and + -- they match better when data con wrappers are inlined. + -- But that only really applies to the trivial wrappers (like (:)), + -- and they are now constructed as Compulsory unfoldings (in MkId) + -- so they'll happen anyway. + + SimplPhase n -> isActive n prag + where + prag = idInlinePragma id + +activeRule :: SimplEnv -> Maybe (Activation -> Bool) +-- Nothing => No rules at all +activeRule env + | opt_RulesOff = Nothing + | otherwise + = case getMode env of + SimplGently -> Just isAlwaysActive + -- Used to be Nothing (no rules in gentle mode) + -- Main motivation for changing is that I wanted + -- lift String ===> ... + -- to work in Template Haskell when simplifying + -- splices, so we get simpler code for literal strings + SimplPhase n -> Just (isActive n) +\end{code} + %************************************************************************ %* * @@ -510,15 +786,19 @@ Try three things \begin{code} mkLam env bndrs body cont - | opt_SimplDoEtaReduction, - Just etad_lam <- tryEtaReduce bndrs body - = tick (EtaReduction (head bndrs)) `thenSmpl_` - returnSmpl (emptyFloats env, etad_lam) - - | opt_SimplDoLambdaEtaExpansion, - any isRuntimeVar bndrs - = tryEtaExpansion body `thenSmpl` \ body' -> - returnSmpl (emptyFloats env, mkLams bndrs body') + = getDOptsSmpl `thenSmpl` \dflags -> + mkLam' dflags env bndrs body cont + where + mkLam' dflags env bndrs body cont + | dopt Opt_DoEtaReduction dflags, + Just etad_lam <- tryEtaReduce bndrs body + = tick (EtaReduction (head bndrs)) `thenSmpl_` + returnSmpl (emptyFloats env, etad_lam) + + | dopt Opt_DoLambdaEtaExpansion dflags, + any isRuntimeVar bndrs + = tryEtaExpansion body `thenSmpl` \ body' -> + returnSmpl (emptyFloats env, mkLams bndrs body') {- Sept 01: I'm experimenting with getting the full laziness pass to float out past big lambdsa @@ -531,8 +811,8 @@ mkLam env bndrs body cont returnSmpl (floats, mkLams bndrs body') -} - | otherwise - = returnSmpl (emptyFloats env, mkLams bndrs body) + | otherwise + = returnSmpl (emptyFloats env, mkLams bndrs body) \end{code} @@ -554,16 +834,27 @@ tryEtaReduce bndrs body -- efficient here: -- (a) we already have the binders -- (b) we can do the triviality test before computing the free vars - -- [in fact I take the simple path and look for just a variable] = go (reverse bndrs) body where go (b : bs) (App fun arg) | ok_arg b arg = go bs fun -- Loop round - go [] (Var fun) | ok_fun fun = Just (Var fun) -- Success! + go [] fun | ok_fun fun = Just fun -- Success! go _ _ = Nothing -- Failure! - ok_fun fun = not (fun `elem` bndrs) && not (hasNoBinding fun) - -- Note the awkward "hasNoBinding" test - -- Details with exprIsTrivial + ok_fun fun = exprIsTrivial fun + && not (any (`elemVarSet` (exprFreeVars fun)) bndrs) + && (exprIsHNF fun || all ok_lam bndrs) + ok_lam v = isTyVar v || isDictId v + -- The exprIsHNF is because eta reduction is not + -- valid in general: \x. bot /= bot + -- So we need to be sure that the "fun" is a value. + -- + -- However, we always want to reduce (/\a -> f a) to f + -- This came up in a RULE: foldr (build (/\a -> g a)) + -- did not match foldr (build (/\b -> ...something complex...)) + -- The type checker can insert these eta-expanded versions, + -- with both type and dictionary lambdas; hence the slightly + -- ad-hoc isDictTy + ok_arg b arg = varToCoreExpr b `cheapEqExpr` arg \end{code} @@ -781,6 +1072,143 @@ tryRhsTyLam env tyvars body -- Only does something if there's a let -} \end{code} +%************************************************************************ +%* * +\subsection{Case alternative filtering +%* * +%************************************************************************ + +prepareAlts does two things: + +1. Eliminate alternatives that cannot match, including the + DEFAULT alternative. + +2. If the DEFAULT alternative can match only one possible constructor, + then make that constructor explicit. + e.g. + case e of x { DEFAULT -> rhs } + ===> + case e of x { (a,b) -> rhs } + where the type is a single constructor type. This gives better code + when rhs also scrutinises x or e. + +It's a good idea do do this stuff before simplifying the alternatives, to +avoid simplifying alternatives we know can't happen, and to come up with +the list of constructors that are handled, to put into the IdInfo of the +case binder, for use when simplifying the alternatives. + +Eliminating the default alternative in (1) isn't so obvious, but it can +happen: + +data Colour = Red | Green | Blue + +f x = case x of + Red -> .. + Green -> .. + DEFAULT -> h x + +h y = case y of + Blue -> .. + DEFAULT -> [ case y of ... ] + +If we inline h into f, the default case of the inlined h can't happen. +If we don't notice this, we may end up filtering out *all* the cases +of the inner case y, which give us nowhere to go! + + +\begin{code} +prepareAlts :: OutExpr -- Scrutinee + -> InId -- Case binder (passed only to use in statistics) + -> [InAlt] -- Increasing order + -> SimplM ([InAlt], -- Better alternatives, still incresaing order + [AltCon]) -- These cases are handled + +prepareAlts scrut case_bndr alts + = let + (alts_wo_default, maybe_deflt) = findDefault alts + + impossible_cons = case scrut of + Var v -> otherCons (idUnfolding v) + other -> [] + + -- Filter out alternatives that can't possibly match + better_alts | null impossible_cons = alts_wo_default + | otherwise = [alt | alt@(con,_,_) <- alts_wo_default, + not (con `elem` impossible_cons)] + + -- "handled_cons" are handled either by the context, + -- or by a branch in this case expression + -- (Don't add DEFAULT to the handled_cons!!) + handled_cons = impossible_cons ++ [con | (con,_,_) <- better_alts] + in + -- Filter out the default, if it can't happen, + -- or replace it with "proper" alternative if there + -- is only one constructor left + prepareDefault scrut case_bndr handled_cons maybe_deflt `thenSmpl` \ deflt_alt -> + + returnSmpl (mergeAlts better_alts deflt_alt, handled_cons) + -- We need the mergeAlts in case the new default_alt + -- has turned into a constructor alternative. + +prepareDefault scrut case_bndr handled_cons (Just rhs) + | Just (tycon, inst_tys) <- splitTyConApp_maybe (exprType scrut), + -- Use exprType scrut here, rather than idType case_bndr, because + -- case_bndr is an InId, so exprType scrut may have more information + -- Test simpl013 is an example + isAlgTyCon tycon, -- It's a data type, tuple, or unboxed tuples. + not (isNewTyCon tycon), -- We can have a newtype, if we are just doing an eval: + -- case x of { DEFAULT -> e } + -- and we don't want to fill in a default for them! + Just all_cons <- tyConDataCons_maybe tycon, + not (null all_cons), -- This is a tricky corner case. If the data type has no constructors, + -- which GHC allows, then the case expression will have at most a default + -- alternative. We don't want to eliminate that alternative, because the + -- invariant is that there's always one alternative. It's more convenient + -- to leave + -- case x of { DEFAULT -> e } + -- as it is, rather than transform it to + -- error "case cant match" + -- which would be quite legitmate. But it's a really obscure corner, and + -- not worth wasting code on. + let handled_data_cons = [data_con | DataAlt data_con <- handled_cons], + let missing_cons = [con | con <- all_cons, + not (con `elem` handled_data_cons)] + = case missing_cons of + [] -> returnSmpl [] -- Eliminate the default alternative + -- if it can't match + + [con] -> -- It matches exactly one constructor, so fill it in + tick (FillInCaseDefault case_bndr) `thenSmpl_` + mk_args con inst_tys `thenSmpl` \ args -> + returnSmpl [(DataAlt con, args, rhs)] + + two_or_more -> returnSmpl [(DEFAULT, [], rhs)] + + | otherwise + = returnSmpl [(DEFAULT, [], rhs)] + +prepareDefault scrut case_bndr handled_cons Nothing + = returnSmpl [] + +mk_args missing_con inst_tys + = mk_tv_bndrs missing_con inst_tys `thenSmpl` \ (tv_bndrs, inst_tys') -> + getUniquesSmpl `thenSmpl` \ id_uniqs -> + let arg_tys = dataConInstArgTys missing_con inst_tys' + arg_ids = zipWith (mkSysLocal FSLIT("a")) id_uniqs arg_tys + in + returnSmpl (tv_bndrs ++ arg_ids) + +mk_tv_bndrs missing_con inst_tys + | isVanillaDataCon missing_con + = returnSmpl ([], inst_tys) + | otherwise + = getUniquesSmpl `thenSmpl` \ tv_uniqs -> + let new_tvs = zipWith mk tv_uniqs (dataConTyVars missing_con) + mk uniq tv = mkTyVar (mkSysTvName uniq FSLIT("t")) (tyVarKind tv) + in + returnSmpl (new_tvs, mkTyVarTys new_tvs) +\end{code} + %************************************************************************ %* * @@ -791,11 +1219,14 @@ tryRhsTyLam env tyvars body -- Only does something if there's a let mkCase puts a case expression back together, trying various transformations first. \begin{code} -mkCase :: OutExpr -> OutId -> [OutAlt] -> SimplM OutExpr - -mkCase scrut case_bndr alts - = mkAlts scrut case_bndr alts `thenSmpl` \ better_alts -> - mkCase1 scrut case_bndr better_alts +mkCase :: OutExpr -> OutId -> OutType + -> [OutAlt] -- Increasing order + -> SimplM OutExpr + +mkCase scrut case_bndr ty alts + = getDOptsSmpl `thenSmpl` \dflags -> + mkAlts dflags scrut case_bndr alts `thenSmpl` \ better_alts -> + mkCase1 scrut case_bndr ty better_alts \end{code} @@ -817,16 +1248,7 @@ mkAlts tries these things: a) all branches equal b) some branches equal to the DEFAULT (which occurs first) -2. If the DEFAULT alternative can match only one possible constructor, - then make that constructor explicit. - e.g. - case e of x { DEFAULT -> rhs } - ===> - case e of x { (a,b) -> rhs } - where the type is a single constructor type. This gives better code - when rhs also scrutinises x or e. - -3. Case merging: +2. Case merging: case e of b { ==> case e of b { p1 -> rhs1 p1 -> rhs1 ... ... @@ -869,7 +1291,7 @@ and similarly in cascade for all the join points! -------------------------------------------------- -- 1. Merge identical branches -------------------------------------------------- -mkAlts scrut case_bndr alts@((con1,bndrs1,rhs1) : con_alts) +mkAlts dflags scrut case_bndr alts@((con1,bndrs1,rhs1) : con_alts) | all isDeadBinder bndrs1, -- Remember the default length filtered_alts < length con_alts -- alternative comes first = tick (AltMerge case_bndr) `thenSmpl_` @@ -881,91 +1303,56 @@ mkAlts scrut case_bndr alts@((con1,bndrs1,rhs1) : con_alts) -------------------------------------------------- --- 2. Fill in missing constructor +-- 2. Merge nested cases -------------------------------------------------- -mkAlts scrut case_bndr alts - | (alts_no_deflt, Just rhs) <- findDefault alts, - -- There is a DEFAULT case - - Just (tycon, inst_tys) <- splitTyConApp_maybe (idType case_bndr), - isAlgTyCon tycon, -- It's a data type, tuple, or unboxed tuples. - not (isNewTyCon tycon), -- We can have a newtype, if we are just doing an eval: - -- case x of { DEFAULT -> e } - -- and we don't want to fill in a default for them! - - [missing_con] <- filter is_missing (tyConDataConsIfAvailable tycon) - -- There is just one missing constructor! - - = tick (FillInCaseDefault case_bndr) `thenSmpl_` - getUniquesSmpl `thenSmpl` \ tv_uniqs -> - getUniquesSmpl `thenSmpl` \ id_uniqs -> - let - (_,_,ex_tyvars,_,_,_) = dataConSig missing_con - ex_tyvars' = zipWith mk tv_uniqs ex_tyvars - mk uniq tv = mkSysTyVar uniq (tyVarKind tv) - arg_ids = zipWith (mkSysLocal SLIT("a")) id_uniqs arg_tys - arg_tys = dataConArgTys missing_con (inst_tys ++ mkTyVarTys ex_tyvars') - better_alts = (DataAlt missing_con, ex_tyvars' ++ arg_ids, rhs) : alts_no_deflt - in - returnSmpl better_alts - where - impossible_cons = case scrut of - Var v -> otherCons (idUnfolding v) - other -> [] - handled_data_cons = [data_con | DataAlt data_con <- impossible_cons] ++ - [data_con | (DataAlt data_con, _, _) <- alts] - is_missing con = not (con `elem` handled_data_cons) - --------------------------------------------------- --- 3. Merge nested cases --------------------------------------------------- - -mkAlts scrut outer_bndr outer_alts - | opt_SimplCaseMerge, +mkAlts dflags scrut outer_bndr outer_alts + | dopt Opt_CaseMerge dflags, (outer_alts_without_deflt, maybe_outer_deflt) <- findDefault outer_alts, - Just (Case (Var scrut_var) inner_bndr inner_alts) <- maybe_outer_deflt, + Just (Case (Var scrut_var) inner_bndr _ inner_alts) <- maybe_outer_deflt, scruting_same_var scrut_var - - = let -- Eliminate any inner alts which are shadowed by the outer ones - outer_cons = [con | (con,_,_) <- outer_alts_without_deflt] - - munged_inner_alts = [ (con, args, munge_rhs rhs) - | (con, args, rhs) <- inner_alts, - not (con `elem` outer_cons) -- Eliminate shadowed inner alts - ] - munge_rhs rhs = bindCaseBndr inner_bndr (Var outer_bndr) rhs - - (inner_con_alts, maybe_inner_default) = findDefault munged_inner_alts - - new_alts = add_default maybe_inner_default - (outer_alts_without_deflt ++ inner_con_alts) + = let + munged_inner_alts = [(con, args, munge_rhs rhs) | (con, args, rhs) <- inner_alts] + munge_rhs rhs = bindCaseBndr inner_bndr (Var outer_bndr) rhs + + new_alts = mergeAlts outer_alts_without_deflt munged_inner_alts + -- The merge keeps the inner DEFAULT at the front, if there is one + -- and eliminates any inner_alts that are shadowed by the outer_alts in tick (CaseMerge outer_bndr) `thenSmpl_` returnSmpl new_alts - -- Warning: don't call mkAlts recursively! - -- Firstly, there's no point, because inner alts have already had - -- mkCase applied to them, so they won't have a case in their default - -- Secondly, if you do, you get an infinite loop, because the bindCaseBndr - -- in munge_rhs may put a case into the DEFAULT branch! + -- Warning: don't call mkAlts recursively! + -- Firstly, there's no point, because inner alts have already had + -- mkCase applied to them, so they won't have a case in their default + -- Secondly, if you do, you get an infinite loop, because the bindCaseBndr + -- in munge_rhs may put a case into the DEFAULT branch! where - -- We are scrutinising the same variable if it's - -- the outer case-binder, or if the outer case scrutinises a variable - -- (and it's the same). Testing both allows us not to replace the - -- outer scrut-var with the outer case-binder (Simplify.simplCaseBinder). + -- We are scrutinising the same variable if it's + -- the outer case-binder, or if the outer case scrutinises a variable + -- (and it's the same). Testing both allows us not to replace the + -- outer scrut-var with the outer case-binder (Simplify.simplCaseBinder). scruting_same_var = case scrut of Var outer_scrut -> \ v -> v == outer_bndr || v == outer_scrut other -> \ v -> v == outer_bndr - add_default (Just rhs) alts = (DEFAULT,[],rhs) : alts - add_default Nothing alts = alts - - --------------------------------------------------- +------------------------------------------------ -- Catch-all --------------------------------------------------- - -mkAlts scrut case_bndr other_alts = returnSmpl other_alts +------------------------------------------------ + +mkAlts dflags scrut case_bndr other_alts = returnSmpl other_alts + + +--------------------------------- +mergeAlts :: [OutAlt] -> [OutAlt] -> [OutAlt] +-- Merge preserving order; alternatives in the first arg +-- shadow ones in the second +mergeAlts [] as2 = as2 +mergeAlts as1 [] = as1 +mergeAlts (a1:as1) (a2:as2) + = case a1 `cmpAlt` a2 of + LT -> a1 : mergeAlts as1 (a2:as2) + EQ -> a1 : mergeAlts as1 as2 -- Discard a2 + GT -> a2 : mergeAlts (a1:as1) as2 \end{code} @@ -1061,13 +1448,65 @@ So the case-elimination algorithm is: If so, then we can replace the case with one of the rhss. +Further notes about case elimination +~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ +Consider: test :: Integer -> IO () + test = print + +Turns out that this compiles to: + Print.test + = \ eta :: Integer + eta1 :: State# RealWorld -> + case PrelNum.< eta PrelNum.zeroInteger of wild { __DEFAULT -> + case hPutStr stdout + (PrelNum.jtos eta ($w[] @ Char)) + eta1 + of wild1 { (# new_s, a4 #) -> PrelIO.lvl23 new_s }} + +Notice the strange '<' which has no effect at all. This is a funny one. +It started like this: + +f x y = if x < 0 then jtos x + else if y==0 then "" else jtos x + +At a particular call site we have (f v 1). So we inline to get + + if v < 0 then jtos x + else if 1==0 then "" else jtos x + +Now simplify the 1==0 conditional: + + if v<0 then jtos v else jtos v + +Now common-up the two branches of the case: + + case (v<0) of DEFAULT -> jtos v + +Why don't we drop the case? Because it's strict in v. It's technically +wrong to drop even unnecessary evaluations, and in practice they +may be a result of 'seq' so we *definitely* don't want to drop those. +I don't really know how to improve this situation. + \begin{code} -------------------------------------------------- +-- 0. Check for empty alternatives +-------------------------------------------------- + +-- This isn't strictly an error. It's possible that the simplifer might "see" +-- that an inner case has no accessible alternatives before it "sees" that the +-- entire branch of an outer case is inaccessible. So we simply +-- put an error case here insteadd +mkCase1 scrut case_bndr ty [] + = pprTrace "mkCase1: null alts" (ppr case_bndr <+> ppr scrut) $ + return (mkApps (Var eRROR_ID) + [Type ty, Lit (mkStringLit "Impossible alternative")]) + +-------------------------------------------------- -- 1. Eliminate the case altogether if poss -------------------------------------------------- -mkCase1 scrut case_bndr [(con,bndrs,rhs)] +mkCase1 scrut case_bndr ty [(con,bndrs,rhs)] -- See if we can get rid of the case altogether -- See the extensive notes on case-elimination above -- mkCase made sure that if all the alternatives are equal, @@ -1084,7 +1523,7 @@ mkCase1 scrut case_bndr [(con,bndrs,rhs)] -- x -- This particular example shows up in default methods for -- comparision operations (e.g. in (>=) for Int.Int32) - || exprIsValue scrut -- It's already evaluated + || exprIsHNF scrut -- It's already evaluated || var_demanded_later scrut -- It'll be demanded later -- || not opt_SimplPedanticBottoms) -- Or we don't care! @@ -1094,6 +1533,7 @@ mkCase1 scrut case_bndr [(con,bndrs,rhs)] -- Here we must *not* discard the case, because dataToTag# just fetches the tag from -- the info pointer. So we'll be pedantic all the time, and see if that gives any -- other problems +-- Also we don't want to discard 'seq's = tick (CaseElim case_bndr) `thenSmpl_` returnSmpl (bindCaseBndr case_bndr scrut rhs) @@ -1108,7 +1548,7 @@ mkCase1 scrut case_bndr [(con,bndrs,rhs)] -- 2. Identity case -------------------------------------------------- -mkCase1 scrut case_bndr alts -- Identity case +mkCase1 scrut case_bndr ty alts -- Identity case | all identity_alt alts = tick (CaseIdentity case_bndr) `thenSmpl_` returnSmpl (re_note scrut) @@ -1131,14 +1571,14 @@ mkCase1 scrut case_bndr alts -- Identity case -- re_note wraps a coerce if it might be necessary re_note scrut = case head alts of - (_,_,rhs1@(Note _ _)) -> mkCoerce (exprType rhs1) (idType case_bndr) scrut + (_,_,rhs1@(Note _ _)) -> mkCoerce2 (exprType rhs1) (idType case_bndr) scrut other -> scrut -------------------------------------------------- -- Catch-all -------------------------------------------------- -mkCase1 scrut bndr alts = returnSmpl (Case scrut bndr alts) +mkCase1 scrut bndr ty alts = returnSmpl (Case scrut bndr ty alts) \end{code}