2 % (c) The AQUA Project, Glasgow University, 1993-1998
4 \section[Simplify]{The main module of the simplifier}
7 module Simplify ( simplTopBinds, simplExpr ) where
9 #include "HsVersions.h"
11 import DynFlags ( dopt, DynFlag(Opt_D_dump_inlinings),
15 import Type hiding ( substTy, extendTvSubst )
21 import TcGadt ( dataConCanMatch )
22 import DataCon ( dataConTyCon, dataConRepStrictness )
23 import TyCon ( tyConArity, isAlgTyCon, isNewTyCon, tyConDataCons_maybe )
25 import PprCore ( pprParendExpr, pprCoreExpr )
26 import CoreUnfold ( mkUnfolding, callSiteInline )
28 import Rules ( lookupRule )
29 import BasicTypes ( isMarkedStrict )
30 import CostCentre ( currentCCS )
31 import TysPrim ( realWorldStatePrimTy )
32 import PrelInfo ( realWorldPrimId )
33 import BasicTypes ( TopLevelFlag(..), isTopLevel,
34 RecFlag(..), isNonRuleLoopBreaker )
36 import Maybes ( orElse )
42 The guts of the simplifier is in this module, but the driver loop for
43 the simplifier is in SimplCore.lhs.
46 -----------------------------------------
47 *** IMPORTANT NOTE ***
48 -----------------------------------------
49 The simplifier used to guarantee that the output had no shadowing, but
50 it does not do so any more. (Actually, it never did!) The reason is
51 documented with simplifyArgs.
54 -----------------------------------------
55 *** IMPORTANT NOTE ***
56 -----------------------------------------
57 Many parts of the simplifier return a bunch of "floats" as well as an
58 expression. This is wrapped as a datatype SimplUtils.FloatsWith.
60 All "floats" are let-binds, not case-binds, but some non-rec lets may
61 be unlifted (with RHS ok-for-speculation).
65 -----------------------------------------
66 ORGANISATION OF FUNCTIONS
67 -----------------------------------------
69 - simplify all top-level binders
70 - for NonRec, call simplRecOrTopPair
71 - for Rec, call simplRecBind
74 ------------------------------
75 simplExpr (applied lambda) ==> simplNonRecBind
76 simplExpr (Let (NonRec ...) ..) ==> simplNonRecBind
77 simplExpr (Let (Rec ...) ..) ==> simplify binders; simplRecBind
79 ------------------------------
80 simplRecBind [binders already simplfied]
81 - use simplRecOrTopPair on each pair in turn
83 simplRecOrTopPair [binder already simplified]
84 Used for: recursive bindings (top level and nested)
85 top-level non-recursive bindings
87 - check for PreInlineUnconditionally
91 Used for: non-top-level non-recursive bindings
92 beta reductions (which amount to the same thing)
93 Because it can deal with strict arts, it takes a
94 "thing-inside" and returns an expression
96 - check for PreInlineUnconditionally
97 - simplify binder, including its IdInfo
106 simplNonRecX: [given a *simplified* RHS, but an *unsimplified* binder]
107 Used for: binding case-binder and constr args in a known-constructor case
108 - check for PreInLineUnconditionally
112 ------------------------------
113 simplLazyBind: [binder already simplified, RHS not]
114 Used for: recursive bindings (top level and nested)
115 top-level non-recursive bindings
116 non-top-level, but *lazy* non-recursive bindings
117 [must not be strict or unboxed]
118 Returns floats + an augmented environment, not an expression
119 - substituteIdInfo and add result to in-scope
120 [so that rules are available in rec rhs]
123 - float if exposes constructor or PAP
127 completeNonRecX: [binder and rhs both simplified]
128 - if the the thing needs case binding (unlifted and not ok-for-spec)
134 completeBind: [given a simplified RHS]
135 [used for both rec and non-rec bindings, top level and not]
136 - try PostInlineUnconditionally
137 - add unfolding [this is the only place we add an unfolding]
142 Right hand sides and arguments
143 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
144 In many ways we want to treat
145 (a) the right hand side of a let(rec), and
146 (b) a function argument
147 in the same way. But not always! In particular, we would
148 like to leave these arguments exactly as they are, so they
149 will match a RULE more easily.
154 It's harder to make the rule match if we ANF-ise the constructor,
155 or eta-expand the PAP:
157 f (let { a = g x; b = h x } in (a,b))
160 On the other hand if we see the let-defns
165 then we *do* want to ANF-ise and eta-expand, so that p and q
166 can be safely inlined.
168 Even floating lets out is a bit dubious. For let RHS's we float lets
169 out if that exposes a value, so that the value can be inlined more vigorously.
172 r = let x = e in (x,x)
174 Here, if we float the let out we'll expose a nice constructor. We did experiments
175 that showed this to be a generally good thing. But it was a bad thing to float
176 lets out unconditionally, because that meant they got allocated more often.
178 For function arguments, there's less reason to expose a constructor (it won't
179 get inlined). Just possibly it might make a rule match, but I'm pretty skeptical.
180 So for the moment we don't float lets out of function arguments either.
185 For eta expansion, we want to catch things like
187 case e of (a,b) -> \x -> case a of (p,q) -> \y -> r
189 If the \x was on the RHS of a let, we'd eta expand to bring the two
190 lambdas together. And in general that's a good thing to do. Perhaps
191 we should eta expand wherever we find a (value) lambda? Then the eta
192 expansion at a let RHS can concentrate solely on the PAP case.
195 %************************************************************************
197 \subsection{Bindings}
199 %************************************************************************
202 simplTopBinds :: SimplEnv -> [InBind] -> SimplM [OutBind]
204 simplTopBinds env binds
205 = do { -- Put all the top-level binders into scope at the start
206 -- so that if a transformation rule has unexpectedly brought
207 -- anything into scope, then we don't get a complaint about that.
208 -- It's rather as if the top-level binders were imported.
209 ; env <- simplRecBndrs env (bindersOfBinds binds)
210 ; dflags <- getDOptsSmpl
211 ; let dump_flag = dopt Opt_D_dump_inlinings dflags
212 ; env' <- simpl_binds dump_flag env binds
213 ; freeTick SimplifierDone
214 ; return (getFloats env') }
216 -- We need to track the zapped top-level binders, because
217 -- they should have their fragile IdInfo zapped (notably occurrence info)
218 -- That's why we run down binds and bndrs' simultaneously.
219 simpl_binds :: Bool -> SimplEnv -> [InBind] -> SimplM SimplEnv
220 simpl_binds dump env [] = return env
221 simpl_binds dump env (bind:binds) = do { env' <- trace dump bind $
223 ; simpl_binds dump env' binds }
225 trace True bind = pprTrace "SimplBind" (ppr (bindersOf bind))
226 trace False bind = \x -> x
228 simpl_bind env (NonRec b r) = simplRecOrTopPair env TopLevel b r
229 simpl_bind env (Rec pairs) = simplRecBind env TopLevel pairs
233 %************************************************************************
235 \subsection{Lazy bindings}
237 %************************************************************************
239 simplRecBind is used for
240 * recursive bindings only
243 simplRecBind :: SimplEnv -> TopLevelFlag
246 simplRecBind env top_lvl pairs
247 = do { env' <- go (zapFloats env) pairs
248 ; return (env `addRecFloats` env') }
249 -- addFloats adds the floats from env',
250 -- *and* updates env with the in-scope set from env'
252 go env [] = return env
254 go env ((bndr, rhs) : pairs)
255 = do { env <- simplRecOrTopPair env top_lvl bndr rhs
259 simplOrTopPair is used for
260 * recursive bindings (whether top level or not)
261 * top-level non-recursive bindings
263 It assumes the binder has already been simplified, but not its IdInfo.
266 simplRecOrTopPair :: SimplEnv
268 -> InId -> InExpr -- Binder and rhs
269 -> SimplM SimplEnv -- Returns an env that includes the binding
271 simplRecOrTopPair env top_lvl bndr rhs
272 | preInlineUnconditionally env top_lvl bndr rhs -- Check for unconditional inline
273 = do { tick (PreInlineUnconditionally bndr)
274 ; return (extendIdSubst env bndr (mkContEx env rhs)) }
277 = do { let bndr' = lookupRecBndr env bndr
278 (env', bndr'') = addLetIdInfo env bndr bndr'
279 ; simplLazyBind env' top_lvl Recursive bndr bndr'' rhs env' }
280 -- May not actually be recursive, but it doesn't matter
284 simplLazyBind is used for
285 * [simplRecOrTopPair] recursive bindings (whether top level or not)
286 * [simplRecOrTopPair] top-level non-recursive bindings
287 * [simplNonRecE] non-top-level *lazy* non-recursive bindings
290 1. It assumes that the binder is *already* simplified,
291 and is in scope, and its IdInfo too, except unfolding
293 2. It assumes that the binder type is lifted.
295 3. It does not check for pre-inline-unconditionallly;
296 that should have been done already.
299 simplLazyBind :: SimplEnv
300 -> TopLevelFlag -> RecFlag
301 -> InId -> OutId -- Binder, both pre-and post simpl
302 -- The OutId has IdInfo, except arity, unfolding
303 -> InExpr -> SimplEnv -- The RHS and its environment
306 simplLazyBind env top_lvl is_rec bndr bndr1 rhs rhs_se
307 = do { let rhs_env = rhs_se `setInScope` env
308 rhs_cont = mkRhsStop (idType bndr1)
310 -- Simplify the RHS; note the mkRhsStop, which tells
311 -- the simplifier that this is the RHS of a let.
312 ; (rhs_env1, rhs1) <- simplExprF rhs_env rhs rhs_cont
314 -- If any of the floats can't be floated, give up now
315 -- (The canFloat predicate says True for empty floats.)
316 ; if (not (canFloat top_lvl is_rec False rhs_env1))
317 then completeBind env top_lvl bndr bndr1
318 (wrapFloats rhs_env1 rhs1)
320 -- ANF-ise a constructor or PAP rhs
321 { (rhs_env2, rhs2) <- prepareRhs rhs_env1 rhs1
322 ; (env', rhs3) <- chooseRhsFloats top_lvl is_rec False env rhs_env2 rhs2
323 ; completeBind env' top_lvl bndr bndr1 rhs3 } }
325 chooseRhsFloats :: TopLevelFlag -> RecFlag -> Bool
326 -> SimplEnv -- Env for the let
327 -> SimplEnv -- Env for the RHS, with RHS floats in it
328 -> OutExpr -- ..and the RHS itself
329 -> SimplM (SimplEnv, OutExpr) -- New env for let, and RHS
331 chooseRhsFloats top_lvl is_rec is_strict env rhs_env rhs
332 | not (isEmptyFloats rhs_env) -- Something to float
333 , canFloat top_lvl is_rec is_strict rhs_env -- ...that can float
334 , (isTopLevel top_lvl || exprIsCheap rhs) -- ...and we want to float
335 = do { tick LetFloatFromLet -- Float
336 ; return (addFloats env rhs_env, rhs) } -- Add the floats to the main env
337 | otherwise -- Don't float
338 = return (env, wrapFloats rhs_env rhs) -- Wrap the floats around the RHS
342 %************************************************************************
344 \subsection{simplNonRec}
346 %************************************************************************
348 A specialised variant of simplNonRec used when the RHS is already simplified,
349 notably in knownCon. It uses case-binding where necessary.
352 simplNonRecX :: SimplEnv
353 -> InId -- Old binder
354 -> OutExpr -- Simplified RHS
357 simplNonRecX env bndr new_rhs
358 = do { (env, bndr') <- simplBinder env bndr
359 ; completeNonRecX env NotTopLevel NonRecursive
360 (isStrictBndr bndr) bndr bndr' new_rhs }
362 completeNonRecX :: SimplEnv
363 -> TopLevelFlag -> RecFlag -> Bool
364 -> InId -- Old binder
365 -> OutId -- New binder
366 -> OutExpr -- Simplified RHS
369 completeNonRecX env top_lvl is_rec is_strict old_bndr new_bndr new_rhs
370 = do { (env1, rhs1) <- prepareRhs (zapFloats env) new_rhs
371 ; (env2, rhs2) <- chooseRhsFloats top_lvl is_rec is_strict env env1 rhs1
372 ; completeBind env2 NotTopLevel old_bndr new_bndr rhs2 }
375 {- No, no, no! Do not try preInlineUnconditionally in completeNonRecX
376 Doing so risks exponential behaviour, because new_rhs has been simplified once already
377 In the cases described by the folowing commment, postInlineUnconditionally will
378 catch many of the relevant cases.
379 -- This happens; for example, the case_bndr during case of
380 -- known constructor: case (a,b) of x { (p,q) -> ... }
381 -- Here x isn't mentioned in the RHS, so we don't want to
382 -- create the (dead) let-binding let x = (a,b) in ...
384 -- Similarly, single occurrences can be inlined vigourously
385 -- e.g. case (f x, g y) of (a,b) -> ....
386 -- If a,b occur once we can avoid constructing the let binding for them.
388 Furthermore in the case-binding case preInlineUnconditionally risks extra thunks
389 -- Consider case I# (quotInt# x y) of
390 -- I# v -> let w = J# v in ...
391 -- If we gaily inline (quotInt# x y) for v, we end up building an
393 -- let w = J# (quotInt# x y) in ...
394 -- because quotInt# can fail.
396 | preInlineUnconditionally env NotTopLevel bndr new_rhs
397 = thing_inside (extendIdSubst env bndr (DoneEx new_rhs))
400 prepareRhs takes a putative RHS, checks whether it's a PAP or
401 constructor application and, if so, converts it to ANF, so that the
402 resulting thing can be inlined more easily. Thus
410 prepareRhs :: SimplEnv -> OutExpr -> SimplM (SimplEnv, OutExpr)
411 -- Adds new floats to the env iff that allows us to return a good RHS
413 prepareRhs env (Cast rhs co) -- Note [Float coercions]
414 = do { (env', rhs') <- makeTrivial env rhs
415 ; return (env', Cast rhs' co) }
418 | (Var fun, args) <- collectArgs rhs -- It's an application
419 , let n_args = valArgCount args
420 , n_args > 0 -- ...but not a trivial one
421 , isDataConWorkId fun || n_args < idArity fun -- ...and it's a constructor or PAP
422 = go env (Var fun) args
424 go env fun [] = return (env, fun)
425 go env fun (arg : args) = do { (env', arg') <- makeTrivial env arg
426 ; go env' (App fun arg') args }
428 prepareRhs env rhs -- The default case
432 Note [Float coercions]
433 ~~~~~~~~~~~~~~~~~~~~~~
434 When we find the binding
436 we'd like to transform it to
438 x = x `cast` co -- A trivial binding
439 There's a chance that e will be a constructor application or function, or something
440 like that, so moving the coerion to the usage site may well cancel the coersions
441 and lead to further optimisation. Example:
444 data instance T Int = T Int
446 foo :: Int -> Int -> Int
451 go n = case x of { T m -> go (n-m) }
452 -- This case should optimise
456 makeTrivial :: SimplEnv -> OutExpr -> SimplM (SimplEnv, OutExpr)
457 -- Binds the expression to a variable, if it's not trivial, returning the variable
461 | otherwise -- See Note [Take care] below
462 = do { var <- newId FSLIT("a") (exprType expr)
463 ; env <- completeNonRecX env NotTopLevel NonRecursive
465 ; return (env, substExpr env (Var var)) }
469 %************************************************************************
471 \subsection{Completing a lazy binding}
473 %************************************************************************
476 * deals only with Ids, not TyVars
477 * takes an already-simplified binder and RHS
478 * is used for both recursive and non-recursive bindings
479 * is used for both top-level and non-top-level bindings
481 It does the following:
482 - tries discarding a dead binding
483 - tries PostInlineUnconditionally
484 - add unfolding [this is the only place we add an unfolding]
487 It does *not* attempt to do let-to-case. Why? Because it is used for
488 - top-level bindings (when let-to-case is impossible)
489 - many situations where the "rhs" is known to be a WHNF
490 (so let-to-case is inappropriate).
492 Nor does it do the atomic-argument thing
495 completeBind :: SimplEnv
496 -> TopLevelFlag -- Flag stuck into unfolding
497 -> InId -- Old binder
498 -> OutId -> OutExpr -- New binder and RHS
500 -- completeBind may choose to do its work
501 -- * by extending the substitution (e.g. let x = y in ...)
502 -- * or by adding to the floats in the envt
504 completeBind env top_lvl old_bndr new_bndr new_rhs
505 | postInlineUnconditionally env top_lvl new_bndr occ_info new_rhs unfolding
506 -- Inline and discard the binding
507 = do { tick (PostInlineUnconditionally old_bndr)
508 ; -- pprTrace "postInlineUnconditionally" (ppr old_bndr <+> ppr new_bndr <+> ppr new_rhs) $
509 return (extendIdSubst env old_bndr (DoneEx new_rhs)) }
510 -- Use the substitution to make quite, quite sure that the
511 -- substitution will happen, since we are going to discard the binding
516 new_bndr_info = idInfo new_bndr `setArityInfo` exprArity new_rhs
519 -- Add the unfolding *only* for non-loop-breakers
520 -- Making loop breakers not have an unfolding at all
521 -- means that we can avoid tests in exprIsConApp, for example.
522 -- This is important: if exprIsConApp says 'yes' for a recursive
523 -- thing, then we can get into an infinite loop
526 -- If the unfolding is a value, the demand info may
527 -- go pear-shaped, so we nuke it. Example:
529 -- case x of (p,q) -> h p q x
530 -- Here x is certainly demanded. But after we've nuked
531 -- the case, we'll get just
532 -- let x = (a,b) in h a b x
533 -- and now x is not demanded (I'm assuming h is lazy)
534 -- This really happens. Similarly
535 -- let f = \x -> e in ...f..f...
536 -- After inlining f at some of its call sites the original binding may
537 -- (for example) be no longer strictly demanded.
538 -- The solution here is a bit ad hoc...
539 info_w_unf = new_bndr_info `setUnfoldingInfo` unfolding
540 final_info | loop_breaker = new_bndr_info
541 | isEvaldUnfolding unfolding = zapDemandInfo info_w_unf `orElse` info_w_unf
542 | otherwise = info_w_unf
544 final_id = new_bndr `setIdInfo` final_info
546 -- These seqs forces the Id, and hence its IdInfo,
547 -- and hence any inner substitutions
549 -- pprTrace "Binding" (ppr final_id <+> ppr unfolding) $
550 return (addNonRec env final_id new_rhs)
552 unfolding = mkUnfolding (isTopLevel top_lvl) new_rhs
553 loop_breaker = isNonRuleLoopBreaker occ_info
554 old_info = idInfo old_bndr
555 occ_info = occInfo old_info
560 %************************************************************************
562 \subsection[Simplify-simplExpr]{The main function: simplExpr}
564 %************************************************************************
566 The reason for this OutExprStuff stuff is that we want to float *after*
567 simplifying a RHS, not before. If we do so naively we get quadratic
568 behaviour as things float out.
570 To see why it's important to do it after, consider this (real) example:
584 a -- Can't inline a this round, cos it appears twice
588 Each of the ==> steps is a round of simplification. We'd save a
589 whole round if we float first. This can cascade. Consider
594 let f = let d1 = ..d.. in \y -> e
598 in \x -> ...(\y ->e)...
600 Only in this second round can the \y be applied, and it
601 might do the same again.
605 simplExpr :: SimplEnv -> CoreExpr -> SimplM CoreExpr
606 simplExpr env expr = simplExprC env expr (mkBoringStop expr_ty')
608 expr_ty' = substTy env (exprType expr)
609 -- The type in the Stop continuation, expr_ty', is usually not used
610 -- It's only needed when discarding continuations after finding
611 -- a function that returns bottom.
612 -- Hence the lazy substitution
615 simplExprC :: SimplEnv -> CoreExpr -> SimplCont -> SimplM CoreExpr
616 -- Simplify an expression, given a continuation
617 simplExprC env expr cont
618 = -- pprTrace "simplExprC" (ppr expr $$ ppr cont {- $$ ppr (seIdSubst env) -} $$ ppr (seFloats env) ) $
619 do { (env', expr') <- simplExprF (zapFloats env) expr cont
620 ; -- pprTrace "simplExprC ret" (ppr expr $$ ppr expr') $
621 -- pprTrace "simplExprC ret3" (ppr (seInScope env')) $
622 -- pprTrace "simplExprC ret4" (ppr (seFloats env')) $
623 return (wrapFloats env' expr') }
625 --------------------------------------------------
626 simplExprF :: SimplEnv -> InExpr -> SimplCont
627 -> SimplM (SimplEnv, OutExpr)
629 simplExprF env e cont
630 = -- pprTrace "simplExprF" (ppr e $$ ppr cont $$ ppr (seTvSubst env) $$ ppr (seIdSubst env) {- $$ ppr (seFloats env) -} ) $
631 simplExprF' env e cont
633 simplExprF' env (Var v) cont = simplVar env v cont
634 simplExprF' env (Lit lit) cont = rebuild env (Lit lit) cont
635 simplExprF' env (Note n expr) cont = simplNote env n expr cont
636 simplExprF' env (Cast body co) cont = simplCast env body co cont
637 simplExprF' env (App fun arg) cont = simplExprF env fun $
638 ApplyTo NoDup arg env cont
640 simplExprF' env expr@(Lam _ _) cont
641 = simplLam env (map zap bndrs) body cont
642 -- The main issue here is under-saturated lambdas
643 -- (\x1. \x2. e) arg1
644 -- Here x1 might have "occurs-once" occ-info, because occ-info
645 -- is computed assuming that a group of lambdas is applied
646 -- all at once. If there are too few args, we must zap the
649 n_args = countArgs cont
650 n_params = length bndrs
651 (bndrs, body) = collectBinders expr
652 zap | n_args >= n_params = \b -> b
653 | otherwise = \b -> if isTyVar b then b
655 -- NB: we count all the args incl type args
656 -- so we must count all the binders (incl type lambdas)
658 simplExprF' env (Type ty) cont
659 = ASSERT( contIsRhsOrArg cont )
660 do { ty' <- simplType env ty
661 ; rebuild env (Type ty') cont }
663 simplExprF' env (Case scrut bndr case_ty alts) cont
664 | not (switchIsOn (getSwitchChecker env) NoCaseOfCase)
665 = -- Simplify the scrutinee with a Select continuation
666 simplExprF env scrut (Select NoDup bndr alts env cont)
669 = -- If case-of-case is off, simply simplify the case expression
670 -- in a vanilla Stop context, and rebuild the result around it
671 do { case_expr' <- simplExprC env scrut case_cont
672 ; rebuild env case_expr' cont }
674 case_cont = Select NoDup bndr alts env (mkBoringStop case_ty')
675 case_ty' = substTy env case_ty -- c.f. defn of simplExpr
677 simplExprF' env (Let (Rec pairs) body) cont
678 = do { env <- simplRecBndrs env (map fst pairs)
679 -- NB: bndrs' don't have unfoldings or rules
680 -- We add them as we go down
682 ; env <- simplRecBind env NotTopLevel pairs
683 ; simplExprF env body cont }
685 simplExprF' env (Let (NonRec bndr rhs) body) cont
686 = simplNonRecE env bndr (rhs, env) ([], body) cont
688 ---------------------------------
689 simplType :: SimplEnv -> InType -> SimplM OutType
690 -- Kept monadic just so we can do the seqType
692 = -- pprTrace "simplType" (ppr ty $$ ppr (seTvSubst env)) $
693 seqType new_ty `seq` returnSmpl new_ty
695 new_ty = substTy env ty
699 %************************************************************************
701 \subsection{The main rebuilder}
703 %************************************************************************
706 rebuild :: SimplEnv -> OutExpr -> SimplCont -> SimplM (SimplEnv, OutExpr)
707 -- At this point the substitution in the SimplEnv should be irrelevant
708 -- only the in-scope set and floats should matter
709 rebuild env expr cont
710 = -- pprTrace "rebuild" (ppr expr $$ ppr cont $$ ppr (seFloats env)) $
712 Stop {} -> return (env, expr)
713 CoerceIt co cont -> rebuild env (mkCoerce co expr) cont
714 Select _ bndr alts se cont -> rebuildCase (se `setFloats` env) expr bndr alts cont
715 StrictArg fun ty info cont -> rebuildCall env (fun `App` expr) (funResultTy ty) info cont
716 StrictBind b bs body se cont -> do { env' <- simplNonRecX (se `setFloats` env) b expr
717 ; simplLam env' bs body cont }
718 ApplyTo _ arg se cont -> do { arg' <- simplExpr (se `setInScope` env) arg
719 ; rebuild env (App expr arg') cont }
723 %************************************************************************
727 %************************************************************************
730 simplCast :: SimplEnv -> InExpr -> Coercion -> SimplCont
731 -> SimplM (SimplEnv, OutExpr)
732 simplCast env body co cont
733 = do { co' <- simplType env co
734 ; simplExprF env body (addCoerce co' cont) }
736 addCoerce co cont = add_coerce co (coercionKind co) cont
738 add_coerce co (s1, k1) cont -- co :: ty~ty
739 | s1 `coreEqType` k1 = cont -- is a no-op
741 add_coerce co1 (s1, k2) (CoerceIt co2 cont)
742 | (l1, t1) <- coercionKind co2
743 -- coerce T1 S1 (coerce S1 K1 e)
746 -- coerce T1 K1 e, otherwise
748 -- For example, in the initial form of a worker
749 -- we may find (coerce T (coerce S (\x.e))) y
750 -- and we'd like it to simplify to e[y/x] in one round
752 , s1 `coreEqType` t1 = cont -- The coerces cancel out
753 | otherwise = CoerceIt (mkTransCoercion co1 co2) cont
755 add_coerce co (s1s2, t1t2) (ApplyTo dup arg arg_se cont)
756 | not (isTypeArg arg) -- This whole case only works for value args
757 -- Could upgrade to have equiv thing for type apps too
758 , isFunTy s1s2 -- t1t2 must be a function type, becuase it's applied
759 -- co : s1s2 :=: t1t2
760 -- (coerce (T1->T2) (S1->S2) F) E
762 -- coerce T2 S2 (F (coerce S1 T1 E))
764 -- t1t2 must be a function type, T1->T2, because it's applied
765 -- to something but s1s2 might conceivably not be
767 -- When we build the ApplyTo we can't mix the out-types
768 -- with the InExpr in the argument, so we simply substitute
769 -- to make it all consistent. It's a bit messy.
770 -- But it isn't a common case.
772 -- Example of use: Trac #995
773 = ApplyTo dup new_arg (zapSubstEnv env) (addCoerce co2 cont)
775 -- we split coercion t1->t2 :=: s1->s2 into t1 :=: s1 and
776 -- t2 :=: s2 with left and right on the curried form:
777 -- (->) t1 t2 :=: (->) s1 s2
778 [co1, co2] = decomposeCo 2 co
779 new_arg = mkCoerce (mkSymCoercion co1) arg'
780 arg' = substExpr arg_se arg
782 add_coerce co _ cont = CoerceIt co cont
786 %************************************************************************
790 %************************************************************************
793 simplLam :: SimplEnv -> [InId] -> InExpr -> SimplCont
794 -> SimplM (SimplEnv, OutExpr)
796 simplLam env [] body cont = simplExprF env body cont
798 -- Type-beta reduction
799 simplLam env (bndr:bndrs) body (ApplyTo _ (Type ty_arg) arg_se cont)
800 = ASSERT( isTyVar bndr )
801 do { tick (BetaReduction bndr)
802 ; ty_arg' <- simplType (arg_se `setInScope` env) ty_arg
803 ; simplLam (extendTvSubst env bndr ty_arg') bndrs body cont }
805 -- Ordinary beta reduction
806 simplLam env (bndr:bndrs) body (ApplyTo _ arg arg_se cont)
807 = do { tick (BetaReduction bndr)
808 ; simplNonRecE env bndr (arg, arg_se) (bndrs, body) cont }
810 -- Not enough args, so there are real lambdas left to put in the result
811 simplLam env bndrs body cont
812 = do { (env, bndrs') <- simplLamBndrs env bndrs
813 ; body' <- simplExpr env body
814 ; new_lam <- mkLam bndrs' body'
815 ; rebuild env new_lam cont }
818 simplNonRecE :: SimplEnv
819 -> InId -- The binder
820 -> (InExpr, SimplEnv) -- Rhs of binding (or arg of lambda)
821 -> ([InId], InExpr) -- Body of the let/lambda
824 -> SimplM (SimplEnv, OutExpr)
826 -- simplNonRecE is used for
827 -- * non-top-level non-recursive lets in expressions
830 -- It deals with strict bindings, via the StrictBind continuation,
831 -- which may abort the whole process
833 -- The "body" of the binding comes as a pair of ([InId],InExpr)
834 -- representing a lambda; so we recurse back to simplLam
835 -- Why? Because of the binder-occ-info-zapping done before
836 -- the call to simplLam in simplExprF (Lam ...)
838 simplNonRecE env bndr (rhs, rhs_se) (bndrs, body) cont
839 | preInlineUnconditionally env NotTopLevel bndr rhs
840 = do { tick (PreInlineUnconditionally bndr)
841 ; simplLam (extendIdSubst env bndr (mkContEx rhs_se rhs)) bndrs body cont }
844 = do { simplExprF (rhs_se `setFloats` env) rhs
845 (StrictBind bndr bndrs body env cont) }
848 = do { (env, bndr') <- simplBinder env bndr
849 ; env <- simplLazyBind env NotTopLevel NonRecursive bndr bndr' rhs rhs_se
850 ; simplLam env bndrs body cont }
854 %************************************************************************
858 %************************************************************************
861 -- Hack alert: we only distinguish subsumed cost centre stacks for the
862 -- purposes of inlining. All other CCCSs are mapped to currentCCS.
863 simplNote env (SCC cc) e cont
864 = do { e' <- simplExpr (setEnclosingCC env currentCCS) e
865 ; rebuild env (mkSCC cc e') cont }
867 -- See notes with SimplMonad.inlineMode
868 simplNote env InlineMe e cont
869 | contIsRhsOrArg cont -- Totally boring continuation; see notes above
870 = do { -- Don't inline inside an INLINE expression
871 e' <- simplExpr (setMode inlineMode env) e
872 ; rebuild env (mkInlineMe e') cont }
874 | otherwise -- Dissolve the InlineMe note if there's
875 -- an interesting context of any kind to combine with
876 -- (even a type application -- anything except Stop)
877 = simplExprF env e cont
879 simplNote env (CoreNote s) e cont
880 = simplExpr env e `thenSmpl` \ e' ->
881 rebuild env (Note (CoreNote s) e') cont
885 %************************************************************************
887 \subsection{Dealing with calls}
889 %************************************************************************
892 simplVar env var cont
893 = case substId env var of
894 DoneEx e -> simplExprF (zapSubstEnv env) e cont
895 ContEx tvs ids e -> simplExprF (setSubstEnv env tvs ids) e cont
896 DoneId var1 -> completeCall (zapSubstEnv env) var1 cont
897 -- Note [zapSubstEnv]
898 -- The template is already simplified, so don't re-substitute.
899 -- This is VITAL. Consider
901 -- let y = \z -> ...x... in
903 -- We'll clone the inner \x, adding x->x' in the id_subst
904 -- Then when we inline y, we must *not* replace x by x' in
905 -- the inlined copy!!
907 ---------------------------------------------------------
908 -- Dealing with a call site
910 completeCall env var cont
911 = do { dflags <- getDOptsSmpl
912 ; let (args,call_cont) = contArgs cont
913 -- The args are OutExprs, obtained by *lazily* substituting
914 -- in the args found in cont. These args are only examined
915 -- to limited depth (unless a rule fires). But we must do
916 -- the substitution; rule matching on un-simplified args would
919 ------------- First try rules ----------------
920 -- Do this before trying inlining. Some functions have
921 -- rules *and* are strict; in this case, we don't want to
922 -- inline the wrapper of the non-specialised thing; better
923 -- to call the specialised thing instead.
925 -- We used to use the black-listing mechanism to ensure that inlining of
926 -- the wrapper didn't occur for things that have specialisations till a
927 -- later phase, so but now we just try RULES first
929 -- You might think that we shouldn't apply rules for a loop breaker:
930 -- doing so might give rise to an infinite loop, because a RULE is
931 -- rather like an extra equation for the function:
932 -- RULE: f (g x) y = x+y
935 -- But it's too drastic to disable rules for loop breakers.
936 -- Even the foldr/build rule would be disabled, because foldr
937 -- is recursive, and hence a loop breaker:
938 -- foldr k z (build g) = g k z
939 -- So it's up to the programmer: rules can cause divergence
940 ; let in_scope = getInScope env
942 maybe_rule = case activeRule env of
943 Nothing -> Nothing -- No rules apply
944 Just act_fn -> lookupRule act_fn in_scope
946 ; case maybe_rule of {
947 Just (rule, rule_rhs) ->
948 tick (RuleFired (ru_name rule)) `thenSmpl_`
949 (if dopt Opt_D_dump_inlinings dflags then
950 pprTrace "Rule fired" (vcat [
951 text "Rule:" <+> ftext (ru_name rule),
952 text "Before:" <+> ppr var <+> sep (map pprParendExpr args),
953 text "After: " <+> pprCoreExpr rule_rhs,
954 text "Cont: " <+> ppr call_cont])
957 simplExprF env rule_rhs (dropArgs (ruleArity rule) cont)
958 -- The ruleArity says how many args the rule consumed
960 ; Nothing -> do -- No rules
962 ------------- Next try inlining ----------------
963 { let arg_infos = [interestingArg arg | arg <- args, isValArg arg]
964 n_val_args = length arg_infos
965 interesting_cont = interestingCallContext (notNull args)
968 active_inline = activeInline env var
969 maybe_inline = callSiteInline dflags active_inline
970 var arg_infos interesting_cont
971 ; case maybe_inline of {
972 Just unfolding -- There is an inlining!
973 -> do { tick (UnfoldingDone var)
974 ; (if dopt Opt_D_dump_inlinings dflags then
975 pprTrace "Inlining done" (vcat [
976 text "Before:" <+> ppr var <+> sep (map pprParendExpr args),
977 text "Inlined fn: " <+> nest 2 (ppr unfolding),
978 text "Cont: " <+> ppr call_cont])
981 simplExprF env unfolding cont }
983 ; Nothing -> -- No inlining!
985 ------------- No inlining! ----------------
986 -- Next, look for rules or specialisations that match
988 rebuildCall env (Var var) (idType var)
989 (mkArgInfo var n_val_args call_cont) cont
992 rebuildCall :: SimplEnv
993 -> OutExpr -> OutType -- Function and its type
994 -> (Bool, [Bool]) -- See SimplUtils.mkArgInfo
996 -> SimplM (SimplEnv, OutExpr)
997 rebuildCall env fun fun_ty (has_rules, []) cont
998 -- When we run out of strictness args, it means
999 -- that the call is definitely bottom; see SimplUtils.mkArgInfo
1000 -- Then we want to discard the entire strict continuation. E.g.
1001 -- * case (error "hello") of { ... }
1002 -- * (error "Hello") arg
1003 -- * f (error "Hello") where f is strict
1005 -- Then, especially in the first of these cases, we'd like to discard
1006 -- the continuation, leaving just the bottoming expression. But the
1007 -- type might not be right, so we may have to add a coerce.
1008 | not (contIsTrivial cont) -- Only do thia if there is a non-trivial
1009 = return (env, mk_coerce fun) -- contination to discard, else we do it
1010 where -- again and again!
1011 cont_ty = contResultType cont
1012 co = mkUnsafeCoercion fun_ty cont_ty
1013 mk_coerce expr | cont_ty `coreEqType` fun_ty = fun
1014 | otherwise = mkCoerce co fun
1016 rebuildCall env fun fun_ty info (ApplyTo _ (Type arg_ty) se cont)
1017 = do { ty' <- simplType (se `setInScope` env) arg_ty
1018 ; rebuildCall env (fun `App` Type ty') (applyTy fun_ty ty') info cont }
1020 rebuildCall env fun fun_ty (has_rules, str:strs) (ApplyTo _ arg arg_se cont)
1021 | str || isStrictType arg_ty -- Strict argument
1022 = -- pprTrace "Strict Arg" (ppr arg $$ ppr (seIdSubst env) $$ ppr (seInScope env)) $
1023 simplExprF (arg_se `setFloats` env) arg
1024 (StrictArg fun fun_ty (has_rules, strs) cont)
1027 | otherwise -- Lazy argument
1028 -- DO NOT float anything outside, hence simplExprC
1029 -- There is no benefit (unlike in a let-binding), and we'd
1030 -- have to be very careful about bogus strictness through
1031 -- floating a demanded let.
1032 = do { arg' <- simplExprC (arg_se `setInScope` env) arg
1033 (mkLazyArgStop arg_ty has_rules)
1034 ; rebuildCall env (fun `App` arg') res_ty (has_rules, strs) cont }
1036 (arg_ty, res_ty) = splitFunTy fun_ty
1038 rebuildCall env fun fun_ty info cont
1039 = rebuild env fun cont
1044 This part of the simplifier may break the no-shadowing invariant
1046 f (...(\a -> e)...) (case y of (a,b) -> e')
1047 where f is strict in its second arg
1048 If we simplify the innermost one first we get (...(\a -> e)...)
1049 Simplifying the second arg makes us float the case out, so we end up with
1050 case y of (a,b) -> f (...(\a -> e)...) e'
1051 So the output does not have the no-shadowing invariant. However, there is
1052 no danger of getting name-capture, because when the first arg was simplified
1053 we used an in-scope set that at least mentioned all the variables free in its
1054 static environment, and that is enough.
1056 We can't just do innermost first, or we'd end up with a dual problem:
1057 case x of (a,b) -> f e (...(\a -> e')...)
1059 I spent hours trying to recover the no-shadowing invariant, but I just could
1060 not think of an elegant way to do it. The simplifier is already knee-deep in
1061 continuations. We have to keep the right in-scope set around; AND we have
1062 to get the effect that finding (error "foo") in a strict arg position will
1063 discard the entire application and replace it with (error "foo"). Getting
1064 all this at once is TOO HARD!
1066 %************************************************************************
1068 Rebuilding a cse expression
1070 %************************************************************************
1072 Blob of helper functions for the "case-of-something-else" situation.
1075 ---------------------------------------------------------
1076 -- Eliminate the case if possible
1078 rebuildCase :: SimplEnv
1079 -> OutExpr -- Scrutinee
1080 -> InId -- Case binder
1081 -> [InAlt] -- Alternatives (inceasing order)
1083 -> SimplM (SimplEnv, OutExpr)
1085 rebuildCase env scrut case_bndr alts cont
1086 | Just (con,args) <- exprIsConApp_maybe scrut
1087 -- Works when the scrutinee is a variable with a known unfolding
1088 -- as well as when it's an explicit constructor application
1089 = knownCon env scrut (DataAlt con) args case_bndr alts cont
1091 | Lit lit <- scrut -- No need for same treatment as constructors
1092 -- because literals are inlined more vigorously
1093 = knownCon env scrut (LitAlt lit) [] case_bndr alts cont
1096 = do { -- Prepare the continuation;
1097 -- The new subst_env is in place
1098 (env, dup_cont, nodup_cont) <- prepareCaseCont env alts cont
1100 -- Simplify the alternatives
1101 ; (case_bndr', alts') <- simplAlts env scrut case_bndr alts dup_cont
1102 ; let res_ty' = contResultType dup_cont
1103 ; case_expr <- mkCase scrut case_bndr' res_ty' alts'
1105 -- Notice that rebuildDone returns the in-scope set from env, not alt_env
1106 -- The case binder *not* scope over the whole returned case-expression
1107 ; rebuild env case_expr nodup_cont }
1110 simplCaseBinder checks whether the scrutinee is a variable, v. If so,
1111 try to eliminate uses of v in the RHSs in favour of case_bndr; that
1112 way, there's a chance that v will now only be used once, and hence
1115 Note [no-case-of-case]
1116 ~~~~~~~~~~~~~~~~~~~~~~
1117 There is a time we *don't* want to do that, namely when
1118 -fno-case-of-case is on. This happens in the first simplifier pass,
1119 and enhances full laziness. Here's the bad case:
1120 f = \ y -> ...(case x of I# v -> ...(case x of ...) ... )
1121 If we eliminate the inner case, we trap it inside the I# v -> arm,
1122 which might prevent some full laziness happening. I've seen this
1123 in action in spectral/cichelli/Prog.hs:
1124 [(m,n) | m <- [1..max], n <- [1..max]]
1125 Hence the check for NoCaseOfCase.
1127 Note [Suppressing the case binder-swap]
1128 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1129 There is another situation when it might make sense to suppress the
1130 case-expression binde-swap. If we have
1132 case x of w1 { DEFAULT -> case x of w2 { A -> e1; B -> e2 }
1133 ...other cases .... }
1135 We'll perform the binder-swap for the outer case, giving
1137 case x of w1 { DEFAULT -> case w1 of w2 { A -> e1; B -> e2 }
1138 ...other cases .... }
1140 But there is no point in doing it for the inner case, because w1 can't
1141 be inlined anyway. Furthermore, doing the case-swapping involves
1142 zapping w2's occurrence info (see paragraphs that follow), and that
1143 forces us to bind w2 when doing case merging. So we get
1145 case x of w1 { A -> let w2 = w1 in e1
1146 B -> let w2 = w1 in e2
1147 ...other cases .... }
1149 This is plain silly in the common case where w2 is dead.
1151 Even so, I can't see a good way to implement this idea. I tried
1152 not doing the binder-swap if the scrutinee was already evaluated
1153 but that failed big-time:
1157 case v of w { MkT x ->
1158 case x of x1 { I# y1 ->
1159 case x of x2 { I# y2 -> ...
1161 Notice that because MkT is strict, x is marked "evaluated". But to
1162 eliminate the last case, we must either make sure that x (as well as
1163 x1) has unfolding MkT y1. THe straightforward thing to do is to do
1164 the binder-swap. So this whole note is a no-op.
1168 If we replace the scrutinee, v, by tbe case binder, then we have to nuke
1169 any occurrence info (eg IAmDead) in the case binder, because the
1170 case-binder now effectively occurs whenever v does. AND we have to do
1171 the same for the pattern-bound variables! Example:
1173 (case x of { (a,b) -> a }) (case x of { (p,q) -> q })
1175 Here, b and p are dead. But when we move the argment inside the first
1176 case RHS, and eliminate the second case, we get
1178 case x of { (a,b) -> a b }
1180 Urk! b is alive! Reason: the scrutinee was a variable, and case elimination
1183 Indeed, this can happen anytime the case binder isn't dead:
1184 case <any> of x { (a,b) ->
1185 case x of { (p,q) -> p } }
1186 Here (a,b) both look dead, but come alive after the inner case is eliminated.
1187 The point is that we bring into the envt a binding
1189 after the outer case, and that makes (a,b) alive. At least we do unless
1190 the case binder is guaranteed dead.
1194 Consider case (v `cast` co) of x { I# ->
1195 ... (case (v `cast` co) of {...}) ...
1196 We'd like to eliminate the inner case. We can get this neatly by
1197 arranging that inside the outer case we add the unfolding
1198 v |-> x `cast` (sym co)
1199 to v. Then we should inline v at the inner case, cancel the casts, and away we go
1202 simplCaseBinder :: SimplEnv -> OutExpr -> InId -> SimplM (SimplEnv, OutId)
1203 simplCaseBinder env scrut case_bndr
1204 | switchIsOn (getSwitchChecker env) NoCaseOfCase
1205 -- See Note [no-case-of-case]
1206 = do { (env, case_bndr') <- simplBinder env case_bndr
1207 ; return (env, case_bndr') }
1209 simplCaseBinder env (Var v) case_bndr
1210 -- Failed try [see Note 2 above]
1211 -- not (isEvaldUnfolding (idUnfolding v))
1212 = do { (env, case_bndr') <- simplBinder env (zapOccInfo case_bndr)
1213 ; return (modifyInScope env v case_bndr', case_bndr') }
1214 -- We could extend the substitution instead, but it would be
1215 -- a hack because then the substitution wouldn't be idempotent
1216 -- any more (v is an OutId). And this does just as well.
1218 simplCaseBinder env (Cast (Var v) co) case_bndr -- Note [Case of cast]
1219 = do { (env, case_bndr') <- simplBinder env (zapOccInfo case_bndr)
1220 ; let rhs = Cast (Var case_bndr') (mkSymCoercion co)
1221 ; return (addBinderUnfolding env v rhs, case_bndr') }
1223 simplCaseBinder env other_scrut case_bndr
1224 = do { (env, case_bndr') <- simplBinder env case_bndr
1225 ; return (env, case_bndr') }
1227 zapOccInfo :: InId -> InId -- See Note [zapOccInfo]
1228 zapOccInfo b = b `setIdOccInfo` NoOccInfo
1232 simplAlts does two things:
1234 1. Eliminate alternatives that cannot match, including the
1235 DEFAULT alternative.
1237 2. If the DEFAULT alternative can match only one possible constructor,
1238 then make that constructor explicit.
1240 case e of x { DEFAULT -> rhs }
1242 case e of x { (a,b) -> rhs }
1243 where the type is a single constructor type. This gives better code
1244 when rhs also scrutinises x or e.
1246 Here "cannot match" includes knowledge from GADTs
1248 It's a good idea do do this stuff before simplifying the alternatives, to
1249 avoid simplifying alternatives we know can't happen, and to come up with
1250 the list of constructors that are handled, to put into the IdInfo of the
1251 case binder, for use when simplifying the alternatives.
1253 Eliminating the default alternative in (1) isn't so obvious, but it can
1256 data Colour = Red | Green | Blue
1265 DEFAULT -> [ case y of ... ]
1267 If we inline h into f, the default case of the inlined h can't happen.
1268 If we don't notice this, we may end up filtering out *all* the cases
1269 of the inner case y, which give us nowhere to go!
1273 simplAlts :: SimplEnv
1275 -> InId -- Case binder
1276 -> [InAlt] -> SimplCont
1277 -> SimplM (OutId, [OutAlt]) -- Includes the continuation
1278 -- Like simplExpr, this just returns the simplified alternatives;
1279 -- it not return an environment
1281 simplAlts env scrut case_bndr alts cont'
1282 = -- pprTrace "simplAlts" (ppr alts $$ ppr (seIdSubst env)) $
1283 do { let alt_env = zapFloats env
1284 ; (alt_env, case_bndr') <- simplCaseBinder alt_env scrut case_bndr
1286 ; default_alts <- prepareDefault alt_env case_bndr' imposs_deflt_cons cont' maybe_deflt
1288 ; let inst_tys = tyConAppArgs (idType case_bndr')
1289 trimmed_alts = filter (is_possible inst_tys) alts_wo_default
1290 in_alts = mergeAlts default_alts trimmed_alts
1291 -- We need the mergeAlts in case the new default_alt
1292 -- has turned into a constructor alternative.
1294 ; alts' <- mapM (simplAlt alt_env imposs_cons case_bndr' cont') in_alts
1295 ; return (case_bndr', alts') }
1297 (alts_wo_default, maybe_deflt) = findDefault alts
1298 imposs_cons = case scrut of
1299 Var v -> otherCons (idUnfolding v)
1302 -- "imposs_deflt_cons" are handled either by the context,
1303 -- OR by a branch in this case expression. (Don't include DEFAULT!!)
1304 imposs_deflt_cons = nub (imposs_cons ++ [con | (con,_,_) <- alts_wo_default])
1306 is_possible :: [Type] -> CoreAlt -> Bool
1307 is_possible tys (con, _, _) | con `elem` imposs_cons = False
1308 is_possible tys (DataAlt con, _, _) = dataConCanMatch tys con
1309 is_possible tys alt = True
1311 ------------------------------------
1312 prepareDefault :: SimplEnv
1313 -> OutId -- Case binder; need just for its type. Note that as an
1314 -- OutId, it has maximum information; this is important.
1315 -- Test simpl013 is an example
1316 -> [AltCon] -- These cons can't happen when matching the default
1319 -> SimplM [InAlt] -- One branch or none; still unsimplified
1320 -- We use a list because it's what mergeAlts expects
1322 prepareDefault env case_bndr' imposs_cons cont Nothing
1323 = return [] -- No default branch
1325 prepareDefault env case_bndr' imposs_cons cont (Just rhs)
1326 | -- This branch handles the case where we are
1327 -- scrutinisng an algebraic data type
1328 Just (tycon, inst_tys) <- splitTyConApp_maybe (idType case_bndr'),
1329 isAlgTyCon tycon, -- It's a data type, tuple, or unboxed tuples.
1330 not (isNewTyCon tycon), -- We can have a newtype, if we are just doing an eval:
1331 -- case x of { DEFAULT -> e }
1332 -- and we don't want to fill in a default for them!
1333 Just all_cons <- tyConDataCons_maybe tycon,
1334 not (null all_cons), -- This is a tricky corner case. If the data type has no constructors,
1335 -- which GHC allows, then the case expression will have at most a default
1336 -- alternative. We don't want to eliminate that alternative, because the
1337 -- invariant is that there's always one alternative. It's more convenient
1339 -- case x of { DEFAULT -> e }
1340 -- as it is, rather than transform it to
1341 -- error "case cant match"
1342 -- which would be quite legitmate. But it's a really obscure corner, and
1343 -- not worth wasting code on.
1345 let imposs_data_cons = [con | DataAlt con <- imposs_cons] -- We now know it's a data type
1346 is_possible con = not (con `elem` imposs_data_cons)
1347 && dataConCanMatch inst_tys con
1348 = case filter is_possible all_cons of
1349 [] -> return [] -- Eliminate the default alternative
1350 -- altogether if it can't match
1352 [con] -> -- It matches exactly one constructor, so fill it in
1353 do { tick (FillInCaseDefault case_bndr')
1354 ; us <- getUniquesSmpl
1355 ; let (ex_tvs, co_tvs, arg_ids) =
1356 dataConRepInstPat us con inst_tys
1357 ; return [(DataAlt con, ex_tvs ++ co_tvs ++ arg_ids, rhs)] }
1359 two_or_more -> return [(DEFAULT, [], rhs)]
1362 = return [(DEFAULT, [], rhs)]
1364 ------------------------------------
1365 simplAlt :: SimplEnv
1366 -> [AltCon] -- These constructors can't be present when
1367 -- matching this alternative
1368 -> OutId -- The case binder
1373 -- Simplify an alternative, returning the type refinement for the
1374 -- alternative, if the alternative does any refinement at all
1376 simplAlt env handled_cons case_bndr' cont' (DEFAULT, bndrs, rhs)
1377 = ASSERT( null bndrs )
1378 do { let env' = addBinderOtherCon env case_bndr' handled_cons
1379 -- Record the constructors that the case-binder *can't* be.
1380 ; rhs' <- simplExprC env' rhs cont'
1381 ; return (DEFAULT, [], rhs') }
1383 simplAlt env handled_cons case_bndr' cont' (LitAlt lit, bndrs, rhs)
1384 = ASSERT( null bndrs )
1385 do { let env' = addBinderUnfolding env case_bndr' (Lit lit)
1386 ; rhs' <- simplExprC env' rhs cont'
1387 ; return (LitAlt lit, [], rhs') }
1389 simplAlt env handled_cons case_bndr' cont' (DataAlt con, vs, rhs)
1390 = do { -- Deal with the pattern-bound variables
1391 -- Mark the ones that are in ! positions in the data constructor
1392 -- as certainly-evaluated.
1393 -- NB: it happens that simplBinders does *not* erase the OtherCon
1394 -- form of unfolding, so it's ok to add this info before
1395 -- doing simplBinders
1396 (env, vs') <- simplBinders env (add_evals con vs)
1398 -- Bind the case-binder to (con args)
1399 ; let inst_tys' = tyConAppArgs (idType case_bndr')
1400 con_args = map Type inst_tys' ++ varsToCoreExprs vs'
1401 env' = addBinderUnfolding env case_bndr' (mkConApp con con_args)
1403 ; rhs' <- simplExprC env' rhs cont'
1404 ; return (DataAlt con, vs', rhs') }
1406 -- add_evals records the evaluated-ness of the bound variables of
1407 -- a case pattern. This is *important*. Consider
1408 -- data T = T !Int !Int
1410 -- case x of { T a b -> T (a+1) b }
1412 -- We really must record that b is already evaluated so that we don't
1413 -- go and re-evaluate it when constructing the result.
1414 -- See Note [Data-con worker strictness] in MkId.lhs
1415 add_evals dc vs = cat_evals dc vs (dataConRepStrictness dc)
1417 cat_evals dc vs strs
1421 go (v:vs) strs | isTyVar v = v : go vs strs
1422 go (v:vs) (str:strs)
1423 | isMarkedStrict str = evald_v : go vs strs
1424 | otherwise = zapped_v : go vs strs
1426 zapped_v = zap_occ_info v
1427 evald_v = zapped_v `setIdUnfolding` evaldUnfolding
1428 go _ _ = pprPanic "cat_evals" (ppr dc $$ ppr vs $$ ppr strs)
1430 -- If the case binder is alive, then we add the unfolding
1432 -- to the envt; so vs are now very much alive
1433 -- Note [Aug06] I can't see why this actually matters
1434 zap_occ_info | isDeadBinder case_bndr' = \id -> id
1435 | otherwise = zapOccInfo
1437 addBinderUnfolding :: SimplEnv -> Id -> CoreExpr -> SimplEnv
1438 addBinderUnfolding env bndr rhs
1439 = modifyInScope env bndr (bndr `setIdUnfolding` mkUnfolding False rhs)
1441 addBinderOtherCon :: SimplEnv -> Id -> [AltCon] -> SimplEnv
1442 addBinderOtherCon env bndr cons
1443 = modifyInScope env bndr (bndr `setIdUnfolding` mkOtherCon cons)
1447 %************************************************************************
1449 \subsection{Known constructor}
1451 %************************************************************************
1453 We are a bit careful with occurrence info. Here's an example
1455 (\x* -> case x of (a*, b) -> f a) (h v, e)
1457 where the * means "occurs once". This effectively becomes
1458 case (h v, e) of (a*, b) -> f a)
1460 let a* = h v; b = e in f a
1464 All this should happen in one sweep.
1467 knownCon :: SimplEnv -> OutExpr -> AltCon -> [OutExpr]
1468 -> InId -> [InAlt] -> SimplCont
1469 -> SimplM (SimplEnv, OutExpr)
1471 knownCon env scrut con args bndr alts cont
1472 = do { tick (KnownBranch bndr)
1473 ; knownAlt env scrut args bndr (findAlt con alts) cont }
1475 knownAlt env scrut args bndr (DEFAULT, bs, rhs) cont
1477 do { env <- simplNonRecX env bndr scrut
1478 -- This might give rise to a binding with non-atomic args
1479 -- like x = Node (f x) (g x)
1480 -- but simplNonRecX will atomic-ify it
1481 ; simplExprF env rhs cont }
1483 knownAlt env scrut args bndr (LitAlt lit, bs, rhs) cont
1485 do { env <- simplNonRecX env bndr scrut
1486 ; simplExprF env rhs cont }
1488 knownAlt env scrut args bndr (DataAlt dc, bs, rhs) cont
1489 = do { let dead_bndr = isDeadBinder bndr
1490 n_drop_tys = tyConArity (dataConTyCon dc)
1491 ; env <- bind_args env dead_bndr bs (drop n_drop_tys args)
1493 -- It's useful to bind bndr to scrut, rather than to a fresh
1494 -- binding x = Con arg1 .. argn
1495 -- because very often the scrut is a variable, so we avoid
1496 -- creating, and then subsequently eliminating, a let-binding
1497 -- BUT, if scrut is a not a variable, we must be careful
1498 -- about duplicating the arg redexes; in that case, make
1499 -- a new con-app from the args
1500 bndr_rhs = case scrut of
1503 con_app = mkConApp dc (take n_drop_tys args ++ con_args)
1504 con_args = [substExpr env (varToCoreExpr b) | b <- bs]
1505 -- args are aready OutExprs, but bs are InIds
1507 ; env <- simplNonRecX env bndr bndr_rhs
1508 ; -- pprTrace "knownCon2" (ppr bs $$ ppr rhs $$ ppr (seIdSubst env)) $
1509 simplExprF env rhs cont }
1512 bind_args env dead_bndr [] _ = return env
1514 bind_args env dead_bndr (b:bs) (Type ty : args)
1515 = ASSERT( isTyVar b )
1516 bind_args (extendTvSubst env b ty) dead_bndr bs args
1518 bind_args env dead_bndr (b:bs) (arg : args)
1520 do { let b' = if dead_bndr then b else zapOccInfo b
1521 -- Note that the binder might be "dead", because it doesn't occur
1522 -- in the RHS; and simplNonRecX may therefore discard it via postInlineUnconditionally
1523 -- Nevertheless we must keep it if the case-binder is alive, because it may
1524 -- be used in the con_app. See Note [zapOccInfo]
1525 ; env <- simplNonRecX env b' arg
1526 ; bind_args env dead_bndr bs args }
1528 bind_args _ _ _ _ = panic "bind_args"
1532 %************************************************************************
1534 \subsection{Duplicating continuations}
1536 %************************************************************************
1539 prepareCaseCont :: SimplEnv
1540 -> [InAlt] -> SimplCont
1541 -> SimplM (SimplEnv, SimplCont,SimplCont)
1542 -- Return a duplicatable continuation, a non-duplicable part
1543 -- plus some extra bindings (that scope over the entire
1546 -- No need to make it duplicatable if there's only one alternative
1547 prepareCaseCont env [alt] cont = return (env, cont, mkBoringStop (contResultType cont))
1548 prepareCaseCont env alts cont = mkDupableCont env cont
1552 mkDupableCont :: SimplEnv -> SimplCont
1553 -> SimplM (SimplEnv, SimplCont, SimplCont)
1555 mkDupableCont env cont
1556 | contIsDupable cont
1557 = returnSmpl (env, cont, mkBoringStop (contResultType cont))
1559 mkDupableCont env (Stop {}) = panic "mkDupableCont" -- Handled by previous eqn
1561 mkDupableCont env (CoerceIt ty cont)
1562 = do { (env, dup, nodup) <- mkDupableCont env cont
1563 ; return (env, CoerceIt ty dup, nodup) }
1565 mkDupableCont env cont@(StrictBind bndr _ _ se _)
1566 = return (env, mkBoringStop (substTy se (idType bndr)), cont)
1567 -- See Note [Duplicating strict continuations]
1569 mkDupableCont env cont@(StrictArg _ fun_ty _ _)
1570 = return (env, mkBoringStop (funArgTy fun_ty), cont)
1571 -- See Note [Duplicating strict continuations]
1573 mkDupableCont env (ApplyTo _ arg se cont)
1574 = -- e.g. [...hole...] (...arg...)
1576 -- let a = ...arg...
1577 -- in [...hole...] a
1578 do { (env, dup_cont, nodup_cont) <- mkDupableCont env cont
1579 ; arg <- simplExpr (se `setInScope` env) arg
1580 ; (env, arg) <- makeTrivial env arg
1581 ; let app_cont = ApplyTo OkToDup arg (zapSubstEnv env) dup_cont
1582 ; return (env, app_cont, nodup_cont) }
1584 mkDupableCont env cont@(Select _ case_bndr [(_,bs,rhs)] se case_cont)
1585 -- See Note [Single-alternative case]
1586 -- | not (exprIsDupable rhs && contIsDupable case_cont)
1587 -- | not (isDeadBinder case_bndr)
1588 | all isDeadBinder bs
1589 = return (env, mkBoringStop scrut_ty, cont)
1591 scrut_ty = substTy se (idType case_bndr)
1593 mkDupableCont env (Select _ case_bndr alts se cont)
1594 = -- e.g. (case [...hole...] of { pi -> ei })
1596 -- let ji = \xij -> ei
1597 -- in case [...hole...] of { pi -> ji xij }
1598 do { tick (CaseOfCase case_bndr)
1599 ; (env, dup_cont, nodup_cont) <- mkDupableCont env cont
1600 -- NB: call mkDupableCont here, *not* prepareCaseCont
1601 -- We must make a duplicable continuation, whereas prepareCaseCont
1602 -- doesn't when there is a single case branch
1604 ; let alt_env = se `setInScope` env
1605 ; (alt_env, case_bndr') <- simplBinder alt_env case_bndr
1606 ; alts' <- mapM (simplAlt alt_env [] case_bndr' dup_cont) alts
1607 -- Safe to say that there are no handled-cons for the DEFAULT case
1608 -- NB: simplBinder does not zap deadness occ-info, so
1609 -- a dead case_bndr' will still advertise its deadness
1610 -- This is really important because in
1611 -- case e of b { (# a,b #) -> ... }
1612 -- b is always dead, and indeed we are not allowed to bind b to (# a,b #),
1613 -- which might happen if e was an explicit unboxed pair and b wasn't marked dead.
1614 -- In the new alts we build, we have the new case binder, so it must retain
1616 -- NB: we don't use alt_env further; it has the substEnv for
1617 -- the alternatives, and we don't want that
1619 ; (env, alts') <- mkDupableAlts env case_bndr' alts'
1620 ; return (env, -- Note [Duplicated env]
1621 Select OkToDup case_bndr' alts' (zapSubstEnv env)
1622 (mkBoringStop (contResultType dup_cont)),
1626 mkDupableAlts :: SimplEnv -> OutId -> [InAlt]
1627 -> SimplM (SimplEnv, [InAlt])
1628 -- Absorbs the continuation into the new alternatives
1630 mkDupableAlts env case_bndr' alts
1633 go env [] = return (env, [])
1635 = do { (env, alt') <- mkDupableAlt env case_bndr' alt
1636 ; (env, alts') <- go env alts
1637 ; return (env, alt' : alts' ) }
1639 mkDupableAlt env case_bndr' (con, bndrs', rhs')
1640 | exprIsDupable rhs' -- Note [Small alternative rhs]
1641 = return (env, (con, bndrs', rhs'))
1643 = do { let rhs_ty' = exprType rhs'
1644 used_bndrs' = filter abstract_over (case_bndr' : bndrs')
1646 | isTyVar bndr = True -- Abstract over all type variables just in case
1647 | otherwise = not (isDeadBinder bndr)
1648 -- The deadness info on the new Ids is preserved by simplBinders
1650 ; (final_bndrs', final_args) -- Note [Join point abstraction]
1651 <- if (any isId used_bndrs')
1652 then return (used_bndrs', varsToCoreExprs used_bndrs')
1653 else do { rw_id <- newId FSLIT("w") realWorldStatePrimTy
1654 ; return ([rw_id], [Var realWorldPrimId]) }
1656 ; join_bndr <- newId FSLIT("$j") (mkPiTypes final_bndrs' rhs_ty')
1657 -- Note [Funky mkPiTypes]
1659 ; let -- We make the lambdas into one-shot-lambdas. The
1660 -- join point is sure to be applied at most once, and doing so
1661 -- prevents the body of the join point being floated out by
1662 -- the full laziness pass
1663 really_final_bndrs = map one_shot final_bndrs'
1664 one_shot v | isId v = setOneShotLambda v
1666 join_rhs = mkLams really_final_bndrs rhs'
1667 join_call = mkApps (Var join_bndr) final_args
1669 ; return (addNonRec env join_bndr join_rhs, (con, bndrs', join_call)) }
1670 -- See Note [Duplicated env]
1673 Note [Duplicated env]
1674 ~~~~~~~~~~~~~~~~~~~~~
1675 Some of the alternatives are simplified, but have not been turned into a join point
1676 So they *must* have an zapped subst-env. So we can't use completeNonRecX to
1677 bind the join point, because it might to do PostInlineUnconditionally, and
1678 we'd lose that when zapping the subst-env. We could have a per-alt subst-env,
1679 but zapping it (as we do in mkDupableCont, the Select case) is safe, and
1680 at worst delays the join-point inlining.
1682 Note [Small alterantive rhs]
1683 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1684 It is worth checking for a small RHS because otherwise we
1685 get extra let bindings that may cause an extra iteration of the simplifier to
1686 inline back in place. Quite often the rhs is just a variable or constructor.
1687 The Ord instance of Maybe in PrelMaybe.lhs, for example, took several extra
1688 iterations because the version with the let bindings looked big, and so wasn't
1689 inlined, but after the join points had been inlined it looked smaller, and so
1692 NB: we have to check the size of rhs', not rhs.
1693 Duplicating a small InAlt might invalidate occurrence information
1694 However, if it *is* dupable, we return the *un* simplified alternative,
1695 because otherwise we'd need to pair it up with an empty subst-env....
1696 but we only have one env shared between all the alts.
1697 (Remember we must zap the subst-env before re-simplifying something).
1698 Rather than do this we simply agree to re-simplify the original (small) thing later.
1700 Note [Funky mkPiTypes]
1701 ~~~~~~~~~~~~~~~~~~~~~~
1702 Notice the funky mkPiTypes. If the contructor has existentials
1703 it's possible that the join point will be abstracted over
1704 type varaibles as well as term variables.
1705 Example: Suppose we have
1706 data T = forall t. C [t]
1708 case (case e of ...) of
1710 We get the join point
1711 let j :: forall t. [t] -> ...
1712 j = /\t \xs::[t] -> rhs
1714 case (case e of ...) of
1715 C t xs::[t] -> j t xs
1717 Note [Join point abstaction]
1718 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1719 If we try to lift a primitive-typed something out
1720 for let-binding-purposes, we will *caseify* it (!),
1721 with potentially-disastrous strictness results. So
1722 instead we turn it into a function: \v -> e
1723 where v::State# RealWorld#. The value passed to this function
1724 is realworld#, which generates (almost) no code.
1726 There's a slight infelicity here: we pass the overall
1727 case_bndr to all the join points if it's used in *any* RHS,
1728 because we don't know its usage in each RHS separately
1730 We used to say "&& isUnLiftedType rhs_ty'" here, but now
1731 we make the join point into a function whenever used_bndrs'
1732 is empty. This makes the join-point more CPR friendly.
1733 Consider: let j = if .. then I# 3 else I# 4
1734 in case .. of { A -> j; B -> j; C -> ... }
1736 Now CPR doesn't w/w j because it's a thunk, so
1737 that means that the enclosing function can't w/w either,
1738 which is a lose. Here's the example that happened in practice:
1739 kgmod :: Int -> Int -> Int
1740 kgmod x y = if x > 0 && y < 0 || x < 0 && y > 0
1744 I have seen a case alternative like this:
1746 It's a bit silly to add the realWorld dummy arg in this case, making
1749 (the \v alone is enough to make CPR happy) but I think it's rare
1751 Note [Duplicating strict continuations]
1752 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1753 Do *not* duplicate StrictBind and StritArg continuations. We gain
1754 nothing by propagating them into the expressions, and we do lose a
1755 lot. Here's an example:
1756 && (case x of { T -> F; F -> T }) E
1757 Now, && is strict so we end up simplifying the case with
1758 an ArgOf continuation. If we let-bind it, we get
1760 let $j = \v -> && v E
1761 in simplExpr (case x of { T -> F; F -> T })
1763 And after simplifying more we get
1765 let $j = \v -> && v E
1766 in case x of { T -> $j F; F -> $j T }
1767 Which is a Very Bad Thing
1769 The desire not to duplicate is the entire reason that
1770 mkDupableCont returns a pair of continuations.
1772 The original plan had:
1773 e.g. (...strict-fn...) [...hole...]
1775 let $j = \a -> ...strict-fn...
1778 Note [Single-alternative cases]
1779 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1780 This case is just like the ArgOf case. Here's an example:
1784 case (case x of I# x' ->
1786 True -> I# (negate# x')
1787 False -> I# x') of y {
1789 Because the (case x) has only one alternative, we'll transform to
1791 case (case x' <# 0# of
1792 True -> I# (negate# x')
1793 False -> I# x') of y {
1795 But now we do *NOT* want to make a join point etc, giving
1797 let $j = \y -> MkT y
1799 True -> $j (I# (negate# x'))
1801 In this case the $j will inline again, but suppose there was a big
1802 strict computation enclosing the orginal call to MkT. Then, it won't
1803 "see" the MkT any more, because it's big and won't get duplicated.
1804 And, what is worse, nothing was gained by the case-of-case transform.
1806 When should use this case of mkDupableCont?
1807 However, matching on *any* single-alternative case is a *disaster*;
1808 e.g. case (case ....) of (a,b) -> (# a,b #)
1809 We must push the outer case into the inner one!
1812 * Match [(DEFAULT,_,_)], but in the common case of Int,
1813 the alternative-filling-in code turned the outer case into
1814 case (...) of y { I# _ -> MkT y }
1816 * Match on single alternative plus (not (isDeadBinder case_bndr))
1817 Rationale: pushing the case inwards won't eliminate the construction.
1818 But there's a risk of
1819 case (...) of y { (a,b) -> let z=(a,b) in ... }
1820 Now y looks dead, but it'll come alive again. Still, this
1821 seems like the best option at the moment.
1823 * Match on single alternative plus (all (isDeadBinder bndrs))
1824 Rationale: this is essentially seq.
1826 * Match when the rhs is *not* duplicable, and hence would lead to a
1827 join point. This catches the disaster-case above. We can test
1828 the *un-simplified* rhs, which is fine. It might get bigger or
1829 smaller after simplification; if it gets smaller, this case might
1830 fire next time round. NB also that we must test contIsDupable
1831 case_cont *btoo, because case_cont might be big!
1833 HOWEVER: I found that this version doesn't work well, because
1834 we can get let x = case (...) of { small } in ...case x...
1835 When x is inlined into its full context, we find that it was a bad
1836 idea to have pushed the outer case inside the (...) case.